ip6_output.c revision 262743
1/*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $ 30 */ 31 32/*- 33 * Copyright (c) 1982, 1986, 1988, 1990, 1993 34 * The Regents of the University of California. All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 4. Neither the name of the University nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 * 60 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 61 */ 62 63#include <sys/cdefs.h> 64__FBSDID("$FreeBSD: stable/10/sys/netinet6/ip6_output.c 262743 2014-03-04 15:14:47Z glebius $"); 65 66#include "opt_inet.h" 67#include "opt_inet6.h" 68#include "opt_ipfw.h" 69#include "opt_ipsec.h" 70#include "opt_sctp.h" 71#include "opt_route.h" 72 73#include <sys/param.h> 74#include <sys/kernel.h> 75#include <sys/malloc.h> 76#include <sys/mbuf.h> 77#include <sys/errno.h> 78#include <sys/priv.h> 79#include <sys/proc.h> 80#include <sys/protosw.h> 81#include <sys/socket.h> 82#include <sys/socketvar.h> 83#include <sys/syslog.h> 84#include <sys/ucred.h> 85 86#include <machine/in_cksum.h> 87 88#include <net/if.h> 89#include <net/netisr.h> 90#include <net/route.h> 91#include <net/pfil.h> 92#include <net/vnet.h> 93 94#include <netinet/in.h> 95#include <netinet/in_var.h> 96#include <netinet/ip_var.h> 97#include <netinet6/in6_var.h> 98#include <netinet/ip6.h> 99#include <netinet/icmp6.h> 100#include <netinet6/ip6_var.h> 101#include <netinet/in_pcb.h> 102#include <netinet/tcp_var.h> 103#include <netinet6/nd6.h> 104 105#ifdef IPSEC 106#include <netipsec/ipsec.h> 107#include <netipsec/ipsec6.h> 108#include <netipsec/key.h> 109#include <netinet6/ip6_ipsec.h> 110#endif /* IPSEC */ 111#ifdef SCTP 112#include <netinet/sctp.h> 113#include <netinet/sctp_crc32.h> 114#endif 115 116#include <netinet6/ip6protosw.h> 117#include <netinet6/scope6_var.h> 118 119#ifdef FLOWTABLE 120#include <net/flowtable.h> 121#endif 122 123extern int in6_mcast_loop; 124 125struct ip6_exthdrs { 126 struct mbuf *ip6e_ip6; 127 struct mbuf *ip6e_hbh; 128 struct mbuf *ip6e_dest1; 129 struct mbuf *ip6e_rthdr; 130 struct mbuf *ip6e_dest2; 131}; 132 133static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, 134 struct ucred *, int); 135static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *, 136 struct socket *, struct sockopt *); 137static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *); 138static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, 139 struct ucred *, int, int, int); 140 141static int ip6_copyexthdr(struct mbuf **, caddr_t, int); 142static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 143 struct ip6_frag **); 144static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); 145static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); 146static int ip6_getpmtu(struct route_in6 *, struct route_in6 *, 147 struct ifnet *, struct in6_addr *, u_long *, int *, u_int); 148static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int); 149 150 151/* 152 * Make an extension header from option data. hp is the source, and 153 * mp is the destination. 154 */ 155#define MAKE_EXTHDR(hp, mp) \ 156 do { \ 157 if (hp) { \ 158 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 159 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 160 ((eh)->ip6e_len + 1) << 3); \ 161 if (error) \ 162 goto freehdrs; \ 163 } \ 164 } while (/*CONSTCOND*/ 0) 165 166/* 167 * Form a chain of extension headers. 168 * m is the extension header mbuf 169 * mp is the previous mbuf in the chain 170 * p is the next header 171 * i is the type of option. 172 */ 173#define MAKE_CHAIN(m, mp, p, i)\ 174 do {\ 175 if (m) {\ 176 if (!hdrsplit) \ 177 panic("assumption failed: hdr not split"); \ 178 *mtod((m), u_char *) = *(p);\ 179 *(p) = (i);\ 180 p = mtod((m), u_char *);\ 181 (m)->m_next = (mp)->m_next;\ 182 (mp)->m_next = (m);\ 183 (mp) = (m);\ 184 }\ 185 } while (/*CONSTCOND*/ 0) 186 187static void 188in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset) 189{ 190 u_short csum; 191 192 csum = in_cksum_skip(m, offset + plen, offset); 193 if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0) 194 csum = 0xffff; 195 offset += m->m_pkthdr.csum_data; /* checksum offset */ 196 197 if (offset + sizeof(u_short) > m->m_len) { 198 printf("%s: delayed m_pullup, m->len: %d plen %u off %u " 199 "csum_flags=%b\n", __func__, m->m_len, plen, offset, 200 (int)m->m_pkthdr.csum_flags, CSUM_BITS); 201 /* 202 * XXX this should not happen, but if it does, the correct 203 * behavior may be to insert the checksum in the appropriate 204 * next mbuf in the chain. 205 */ 206 return; 207 } 208 *(u_short *)(m->m_data + offset) = csum; 209} 210 211/* 212 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 213 * header (with pri, len, nxt, hlim, src, dst). 214 * This function may modify ver and hlim only. 215 * The mbuf chain containing the packet will be freed. 216 * The mbuf opt, if present, will not be freed. 217 * If route_in6 ro is present and has ro_rt initialized, route lookup would be 218 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL, 219 * then result of route lookup is stored in ro->ro_rt. 220 * 221 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and 222 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 223 * which is rt_rmx.rmx_mtu. 224 * 225 * ifpp - XXX: just for statistics 226 */ 227int 228ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, 229 struct route_in6 *ro, int flags, struct ip6_moptions *im6o, 230 struct ifnet **ifpp, struct inpcb *inp) 231{ 232 struct ip6_hdr *ip6, *mhip6; 233 struct ifnet *ifp, *origifp; 234 struct mbuf *m = m0; 235 struct mbuf *mprev = NULL; 236 int hlen, tlen, len, off; 237 struct route_in6 ip6route; 238 struct rtentry *rt = NULL; 239 struct sockaddr_in6 *dst, src_sa, dst_sa; 240 struct in6_addr odst; 241 int error = 0; 242 struct in6_ifaddr *ia = NULL; 243 u_long mtu; 244 int alwaysfrag, dontfrag; 245 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 246 struct ip6_exthdrs exthdrs; 247 struct in6_addr finaldst, src0, dst0; 248 u_int32_t zone; 249 struct route_in6 *ro_pmtu = NULL; 250 int hdrsplit = 0; 251 int needipsec = 0; 252 int sw_csum, tso; 253#ifdef IPSEC 254 struct ipsec_output_state state; 255 struct ip6_rthdr *rh = NULL; 256 int needipsectun = 0; 257 int segleft_org = 0; 258 struct secpolicy *sp = NULL; 259#endif /* IPSEC */ 260 struct m_tag *fwd_tag = NULL; 261 262 ip6 = mtod(m, struct ip6_hdr *); 263 if (ip6 == NULL) { 264 printf ("ip6 is NULL"); 265 goto bad; 266 } 267 268 if (inp != NULL) 269 M_SETFIB(m, inp->inp_inc.inc_fibnum); 270 271 finaldst = ip6->ip6_dst; 272 bzero(&exthdrs, sizeof(exthdrs)); 273 if (opt) { 274 /* Hop-by-Hop options header */ 275 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 276 /* Destination options header(1st part) */ 277 if (opt->ip6po_rthdr) { 278 /* 279 * Destination options header(1st part) 280 * This only makes sense with a routing header. 281 * See Section 9.2 of RFC 3542. 282 * Disabling this part just for MIP6 convenience is 283 * a bad idea. We need to think carefully about a 284 * way to make the advanced API coexist with MIP6 285 * options, which might automatically be inserted in 286 * the kernel. 287 */ 288 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 289 } 290 /* Routing header */ 291 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 292 /* Destination options header(2nd part) */ 293 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 294 } 295 296#ifdef IPSEC 297 /* 298 * IPSec checking which handles several cases. 299 * FAST IPSEC: We re-injected the packet. 300 */ 301 switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp, &sp)) 302 { 303 case 1: /* Bad packet */ 304 goto freehdrs; 305 case -1: /* Do IPSec */ 306 needipsec = 1; 307 /* 308 * Do delayed checksums now, as we may send before returning. 309 */ 310 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 311 plen = m->m_pkthdr.len - sizeof(*ip6); 312 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr)); 313 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6; 314 } 315#ifdef SCTP 316 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) { 317 sctp_delayed_cksum(m, sizeof(struct ip6_hdr)); 318 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6; 319 } 320#endif 321 case 0: /* No IPSec */ 322 default: 323 break; 324 } 325#endif /* IPSEC */ 326 327 /* 328 * Calculate the total length of the extension header chain. 329 * Keep the length of the unfragmentable part for fragmentation. 330 */ 331 optlen = 0; 332 if (exthdrs.ip6e_hbh) 333 optlen += exthdrs.ip6e_hbh->m_len; 334 if (exthdrs.ip6e_dest1) 335 optlen += exthdrs.ip6e_dest1->m_len; 336 if (exthdrs.ip6e_rthdr) 337 optlen += exthdrs.ip6e_rthdr->m_len; 338 unfragpartlen = optlen + sizeof(struct ip6_hdr); 339 340 /* NOTE: we don't add AH/ESP length here. do that later. */ 341 if (exthdrs.ip6e_dest2) 342 optlen += exthdrs.ip6e_dest2->m_len; 343 344 /* 345 * If we need IPsec, or there is at least one extension header, 346 * separate IP6 header from the payload. 347 */ 348 if ((needipsec || optlen) && !hdrsplit) { 349 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 350 m = NULL; 351 goto freehdrs; 352 } 353 m = exthdrs.ip6e_ip6; 354 hdrsplit++; 355 } 356 357 /* adjust pointer */ 358 ip6 = mtod(m, struct ip6_hdr *); 359 360 /* adjust mbuf packet header length */ 361 m->m_pkthdr.len += optlen; 362 plen = m->m_pkthdr.len - sizeof(*ip6); 363 364 /* If this is a jumbo payload, insert a jumbo payload option. */ 365 if (plen > IPV6_MAXPACKET) { 366 if (!hdrsplit) { 367 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 368 m = NULL; 369 goto freehdrs; 370 } 371 m = exthdrs.ip6e_ip6; 372 hdrsplit++; 373 } 374 /* adjust pointer */ 375 ip6 = mtod(m, struct ip6_hdr *); 376 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 377 goto freehdrs; 378 ip6->ip6_plen = 0; 379 } else 380 ip6->ip6_plen = htons(plen); 381 382 /* 383 * Concatenate headers and fill in next header fields. 384 * Here we have, on "m" 385 * IPv6 payload 386 * and we insert headers accordingly. Finally, we should be getting: 387 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 388 * 389 * during the header composing process, "m" points to IPv6 header. 390 * "mprev" points to an extension header prior to esp. 391 */ 392 u_char *nexthdrp = &ip6->ip6_nxt; 393 mprev = m; 394 395 /* 396 * we treat dest2 specially. this makes IPsec processing 397 * much easier. the goal here is to make mprev point the 398 * mbuf prior to dest2. 399 * 400 * result: IPv6 dest2 payload 401 * m and mprev will point to IPv6 header. 402 */ 403 if (exthdrs.ip6e_dest2) { 404 if (!hdrsplit) 405 panic("assumption failed: hdr not split"); 406 exthdrs.ip6e_dest2->m_next = m->m_next; 407 m->m_next = exthdrs.ip6e_dest2; 408 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 409 ip6->ip6_nxt = IPPROTO_DSTOPTS; 410 } 411 412 /* 413 * result: IPv6 hbh dest1 rthdr dest2 payload 414 * m will point to IPv6 header. mprev will point to the 415 * extension header prior to dest2 (rthdr in the above case). 416 */ 417 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 418 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 419 IPPROTO_DSTOPTS); 420 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 421 IPPROTO_ROUTING); 422 423#ifdef IPSEC 424 if (!needipsec) 425 goto skip_ipsec2; 426 427 /* 428 * pointers after IPsec headers are not valid any more. 429 * other pointers need a great care too. 430 * (IPsec routines should not mangle mbufs prior to AH/ESP) 431 */ 432 exthdrs.ip6e_dest2 = NULL; 433 434 if (exthdrs.ip6e_rthdr) { 435 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); 436 segleft_org = rh->ip6r_segleft; 437 rh->ip6r_segleft = 0; 438 } 439 440 bzero(&state, sizeof(state)); 441 state.m = m; 442 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, 443 &needipsectun); 444 m = state.m; 445 if (error == EJUSTRETURN) { 446 /* 447 * We had a SP with a level of 'use' and no SA. We 448 * will just continue to process the packet without 449 * IPsec processing. 450 */ 451 ; 452 } else if (error) { 453 /* mbuf is already reclaimed in ipsec6_output_trans. */ 454 m = NULL; 455 switch (error) { 456 case EHOSTUNREACH: 457 case ENETUNREACH: 458 case EMSGSIZE: 459 case ENOBUFS: 460 case ENOMEM: 461 break; 462 default: 463 printf("[%s:%d] (ipsec): error code %d\n", 464 __func__, __LINE__, error); 465 /* FALLTHROUGH */ 466 case ENOENT: 467 /* don't show these error codes to the user */ 468 error = 0; 469 break; 470 } 471 goto bad; 472 } else if (!needipsectun) { 473 /* 474 * In the FAST IPSec case we have already 475 * re-injected the packet and it has been freed 476 * by the ipsec_done() function. So, just clean 477 * up after ourselves. 478 */ 479 m = NULL; 480 goto done; 481 } 482 if (exthdrs.ip6e_rthdr) { 483 /* ah6_output doesn't modify mbuf chain */ 484 rh->ip6r_segleft = segleft_org; 485 } 486skip_ipsec2:; 487#endif /* IPSEC */ 488 489 /* 490 * If there is a routing header, discard the packet. 491 */ 492 if (exthdrs.ip6e_rthdr) { 493 error = EINVAL; 494 goto bad; 495 } 496 497 /* Source address validation */ 498 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 499 (flags & IPV6_UNSPECSRC) == 0) { 500 error = EOPNOTSUPP; 501 IP6STAT_INC(ip6s_badscope); 502 goto bad; 503 } 504 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 505 error = EOPNOTSUPP; 506 IP6STAT_INC(ip6s_badscope); 507 goto bad; 508 } 509 510 IP6STAT_INC(ip6s_localout); 511 512 /* 513 * Route packet. 514 */ 515 if (ro == 0) { 516 ro = &ip6route; 517 bzero((caddr_t)ro, sizeof(*ro)); 518 } 519 ro_pmtu = ro; 520 if (opt && opt->ip6po_rthdr) 521 ro = &opt->ip6po_route; 522 dst = (struct sockaddr_in6 *)&ro->ro_dst; 523#ifdef FLOWTABLE 524 if (ro->ro_rt == NULL) 525 (void )flowtable_lookup(AF_INET6, m, (struct route *)ro); 526#endif 527again: 528 /* 529 * if specified, try to fill in the traffic class field. 530 * do not override if a non-zero value is already set. 531 * we check the diffserv field and the ecn field separately. 532 */ 533 if (opt && opt->ip6po_tclass >= 0) { 534 int mask = 0; 535 536 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 537 mask |= 0xfc; 538 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 539 mask |= 0x03; 540 if (mask != 0) 541 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 542 } 543 544 /* fill in or override the hop limit field, if necessary. */ 545 if (opt && opt->ip6po_hlim != -1) 546 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 547 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 548 if (im6o != NULL) 549 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 550 else 551 ip6->ip6_hlim = V_ip6_defmcasthlim; 552 } 553 554#ifdef IPSEC 555 /* 556 * We may re-inject packets into the stack here. 557 */ 558 if (needipsec && needipsectun) { 559 struct ipsec_output_state state; 560 561 /* 562 * All the extension headers will become inaccessible 563 * (since they can be encrypted). 564 * Don't panic, we need no more updates to extension headers 565 * on inner IPv6 packet (since they are now encapsulated). 566 * 567 * IPv6 [ESP|AH] IPv6 [extension headers] payload 568 */ 569 bzero(&exthdrs, sizeof(exthdrs)); 570 exthdrs.ip6e_ip6 = m; 571 572 bzero(&state, sizeof(state)); 573 state.m = m; 574 state.ro = (struct route *)ro; 575 state.dst = (struct sockaddr *)dst; 576 577 error = ipsec6_output_tunnel(&state, sp, flags); 578 579 m = state.m; 580 ro = (struct route_in6 *)state.ro; 581 dst = (struct sockaddr_in6 *)state.dst; 582 if (error == EJUSTRETURN) { 583 /* 584 * We had a SP with a level of 'use' and no SA. We 585 * will just continue to process the packet without 586 * IPsec processing. 587 */ 588 ; 589 } else if (error) { 590 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 591 m0 = m = NULL; 592 m = NULL; 593 switch (error) { 594 case EHOSTUNREACH: 595 case ENETUNREACH: 596 case EMSGSIZE: 597 case ENOBUFS: 598 case ENOMEM: 599 break; 600 default: 601 printf("[%s:%d] (ipsec): error code %d\n", 602 __func__, __LINE__, error); 603 /* FALLTHROUGH */ 604 case ENOENT: 605 /* don't show these error codes to the user */ 606 error = 0; 607 break; 608 } 609 goto bad; 610 } else { 611 /* 612 * In the FAST IPSec case we have already 613 * re-injected the packet and it has been freed 614 * by the ipsec_done() function. So, just clean 615 * up after ourselves. 616 */ 617 m = NULL; 618 goto done; 619 } 620 621 exthdrs.ip6e_ip6 = m; 622 } 623#endif /* IPSEC */ 624 625 /* adjust pointer */ 626 ip6 = mtod(m, struct ip6_hdr *); 627 628 if (ro->ro_rt && fwd_tag == NULL) { 629 rt = ro->ro_rt; 630 ifp = ro->ro_rt->rt_ifp; 631 } else { 632 if (fwd_tag == NULL) { 633 bzero(&dst_sa, sizeof(dst_sa)); 634 dst_sa.sin6_family = AF_INET6; 635 dst_sa.sin6_len = sizeof(dst_sa); 636 dst_sa.sin6_addr = ip6->ip6_dst; 637 } 638 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp, 639 &rt, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 640 if (error != 0) { 641 if (ifp != NULL) 642 in6_ifstat_inc(ifp, ifs6_out_discard); 643 goto bad; 644 } 645 } 646 if (rt == NULL) { 647 /* 648 * If in6_selectroute() does not return a route entry, 649 * dst may not have been updated. 650 */ 651 *dst = dst_sa; /* XXX */ 652 } 653 654 /* 655 * then rt (for unicast) and ifp must be non-NULL valid values. 656 */ 657 if ((flags & IPV6_FORWARDING) == 0) { 658 /* XXX: the FORWARDING flag can be set for mrouting. */ 659 in6_ifstat_inc(ifp, ifs6_out_request); 660 } 661 if (rt != NULL) { 662 ia = (struct in6_ifaddr *)(rt->rt_ifa); 663 rt->rt_use++; 664 } 665 666 667 /* 668 * The outgoing interface must be in the zone of source and 669 * destination addresses. 670 */ 671 origifp = ifp; 672 673 src0 = ip6->ip6_src; 674 if (in6_setscope(&src0, origifp, &zone)) 675 goto badscope; 676 bzero(&src_sa, sizeof(src_sa)); 677 src_sa.sin6_family = AF_INET6; 678 src_sa.sin6_len = sizeof(src_sa); 679 src_sa.sin6_addr = ip6->ip6_src; 680 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 681 goto badscope; 682 683 dst0 = ip6->ip6_dst; 684 if (in6_setscope(&dst0, origifp, &zone)) 685 goto badscope; 686 /* re-initialize to be sure */ 687 bzero(&dst_sa, sizeof(dst_sa)); 688 dst_sa.sin6_family = AF_INET6; 689 dst_sa.sin6_len = sizeof(dst_sa); 690 dst_sa.sin6_addr = ip6->ip6_dst; 691 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) { 692 goto badscope; 693 } 694 695 /* We should use ia_ifp to support the case of 696 * sending packets to an address of our own. 697 */ 698 if (ia != NULL && ia->ia_ifp) 699 ifp = ia->ia_ifp; 700 701 /* scope check is done. */ 702 goto routefound; 703 704 badscope: 705 IP6STAT_INC(ip6s_badscope); 706 in6_ifstat_inc(origifp, ifs6_out_discard); 707 if (error == 0) 708 error = EHOSTUNREACH; /* XXX */ 709 goto bad; 710 711 routefound: 712 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 713 if (opt && opt->ip6po_nextroute.ro_rt) { 714 /* 715 * The nexthop is explicitly specified by the 716 * application. We assume the next hop is an IPv6 717 * address. 718 */ 719 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 720 } 721 else if ((rt->rt_flags & RTF_GATEWAY)) 722 dst = (struct sockaddr_in6 *)rt->rt_gateway; 723 } 724 725 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 726 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 727 } else { 728 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 729 in6_ifstat_inc(ifp, ifs6_out_mcast); 730 /* 731 * Confirm that the outgoing interface supports multicast. 732 */ 733 if (!(ifp->if_flags & IFF_MULTICAST)) { 734 IP6STAT_INC(ip6s_noroute); 735 in6_ifstat_inc(ifp, ifs6_out_discard); 736 error = ENETUNREACH; 737 goto bad; 738 } 739 if ((im6o == NULL && in6_mcast_loop) || 740 (im6o && im6o->im6o_multicast_loop)) { 741 /* 742 * Loop back multicast datagram if not expressly 743 * forbidden to do so, even if we have not joined 744 * the address; protocols will filter it later, 745 * thus deferring a hash lookup and lock acquisition 746 * at the expense of an m_copym(). 747 */ 748 ip6_mloopback(ifp, m, dst); 749 } else { 750 /* 751 * If we are acting as a multicast router, perform 752 * multicast forwarding as if the packet had just 753 * arrived on the interface to which we are about 754 * to send. The multicast forwarding function 755 * recursively calls this function, using the 756 * IPV6_FORWARDING flag to prevent infinite recursion. 757 * 758 * Multicasts that are looped back by ip6_mloopback(), 759 * above, will be forwarded by the ip6_input() routine, 760 * if necessary. 761 */ 762 if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 763 /* 764 * XXX: ip6_mforward expects that rcvif is NULL 765 * when it is called from the originating path. 766 * However, it may not always be the case. 767 */ 768 m->m_pkthdr.rcvif = NULL; 769 if (ip6_mforward(ip6, ifp, m) != 0) { 770 m_freem(m); 771 goto done; 772 } 773 } 774 } 775 /* 776 * Multicasts with a hoplimit of zero may be looped back, 777 * above, but must not be transmitted on a network. 778 * Also, multicasts addressed to the loopback interface 779 * are not sent -- the above call to ip6_mloopback() will 780 * loop back a copy if this host actually belongs to the 781 * destination group on the loopback interface. 782 */ 783 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 784 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 785 m_freem(m); 786 goto done; 787 } 788 } 789 790 /* 791 * Fill the outgoing inteface to tell the upper layer 792 * to increment per-interface statistics. 793 */ 794 if (ifpp) 795 *ifpp = ifp; 796 797 /* Determine path MTU. */ 798 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 799 &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0) 800 goto bad; 801 802 /* 803 * The caller of this function may specify to use the minimum MTU 804 * in some cases. 805 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 806 * setting. The logic is a bit complicated; by default, unicast 807 * packets will follow path MTU while multicast packets will be sent at 808 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 809 * including unicast ones will be sent at the minimum MTU. Multicast 810 * packets will always be sent at the minimum MTU unless 811 * IP6PO_MINMTU_DISABLE is explicitly specified. 812 * See RFC 3542 for more details. 813 */ 814 if (mtu > IPV6_MMTU) { 815 if ((flags & IPV6_MINMTU)) 816 mtu = IPV6_MMTU; 817 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 818 mtu = IPV6_MMTU; 819 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 820 (opt == NULL || 821 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 822 mtu = IPV6_MMTU; 823 } 824 } 825 826 /* 827 * clear embedded scope identifiers if necessary. 828 * in6_clearscope will touch the addresses only when necessary. 829 */ 830 in6_clearscope(&ip6->ip6_src); 831 in6_clearscope(&ip6->ip6_dst); 832 833 /* 834 * If the outgoing packet contains a hop-by-hop options header, 835 * it must be examined and processed even by the source node. 836 * (RFC 2460, section 4.) 837 */ 838 if (exthdrs.ip6e_hbh) { 839 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 840 u_int32_t dummy; /* XXX unused */ 841 u_int32_t plen = 0; /* XXX: ip6_process will check the value */ 842 843#ifdef DIAGNOSTIC 844 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len) 845 panic("ip6e_hbh is not contiguous"); 846#endif 847 /* 848 * XXX: if we have to send an ICMPv6 error to the sender, 849 * we need the M_LOOP flag since icmp6_error() expects 850 * the IPv6 and the hop-by-hop options header are 851 * contiguous unless the flag is set. 852 */ 853 m->m_flags |= M_LOOP; 854 m->m_pkthdr.rcvif = ifp; 855 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 856 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 857 &dummy, &plen) < 0) { 858 /* m was already freed at this point */ 859 error = EINVAL;/* better error? */ 860 goto done; 861 } 862 m->m_flags &= ~M_LOOP; /* XXX */ 863 m->m_pkthdr.rcvif = NULL; 864 } 865 866 /* Jump over all PFIL processing if hooks are not active. */ 867 if (!PFIL_HOOKED(&V_inet6_pfil_hook)) 868 goto passout; 869 870 odst = ip6->ip6_dst; 871 /* Run through list of hooks for output packets. */ 872 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp); 873 if (error != 0 || m == NULL) 874 goto done; 875 ip6 = mtod(m, struct ip6_hdr *); 876 877 /* See if destination IP address was changed by packet filter. */ 878 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) { 879 m->m_flags |= M_SKIP_FIREWALL; 880 /* If destination is now ourself drop to ip6_input(). */ 881 if (in6_localip(&ip6->ip6_dst)) { 882 m->m_flags |= M_FASTFWD_OURS; 883 if (m->m_pkthdr.rcvif == NULL) 884 m->m_pkthdr.rcvif = V_loif; 885 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 886 m->m_pkthdr.csum_flags |= 887 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; 888 m->m_pkthdr.csum_data = 0xffff; 889 } 890#ifdef SCTP 891 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) 892 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 893#endif 894 error = netisr_queue(NETISR_IPV6, m); 895 goto done; 896 } else 897 goto again; /* Redo the routing table lookup. */ 898 } 899 900 /* See if local, if yes, send it to netisr. */ 901 if (m->m_flags & M_FASTFWD_OURS) { 902 if (m->m_pkthdr.rcvif == NULL) 903 m->m_pkthdr.rcvif = V_loif; 904 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 905 m->m_pkthdr.csum_flags |= 906 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; 907 m->m_pkthdr.csum_data = 0xffff; 908 } 909#ifdef SCTP 910 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) 911 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 912#endif 913 error = netisr_queue(NETISR_IPV6, m); 914 goto done; 915 } 916 /* Or forward to some other address? */ 917 if ((m->m_flags & M_IP6_NEXTHOP) && 918 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { 919 dst = (struct sockaddr_in6 *)&ro->ro_dst; 920 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6)); 921 m->m_flags |= M_SKIP_FIREWALL; 922 m->m_flags &= ~M_IP6_NEXTHOP; 923 m_tag_delete(m, fwd_tag); 924 goto again; 925 } 926 927passout: 928 /* 929 * Send the packet to the outgoing interface. 930 * If necessary, do IPv6 fragmentation before sending. 931 * 932 * the logic here is rather complex: 933 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 934 * 1-a: send as is if tlen <= path mtu 935 * 1-b: fragment if tlen > path mtu 936 * 937 * 2: if user asks us not to fragment (dontfrag == 1) 938 * 2-a: send as is if tlen <= interface mtu 939 * 2-b: error if tlen > interface mtu 940 * 941 * 3: if we always need to attach fragment header (alwaysfrag == 1) 942 * always fragment 943 * 944 * 4: if dontfrag == 1 && alwaysfrag == 1 945 * error, as we cannot handle this conflicting request 946 */ 947 sw_csum = m->m_pkthdr.csum_flags; 948 if (!hdrsplit) { 949 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0; 950 sw_csum &= ~ifp->if_hwassist; 951 } else 952 tso = 0; 953 /* 954 * If we added extension headers, we will not do TSO and calculate the 955 * checksums ourselves for now. 956 * XXX-BZ Need a framework to know when the NIC can handle it, even 957 * with ext. hdrs. 958 */ 959 if (sw_csum & CSUM_DELAY_DATA_IPV6) { 960 sw_csum &= ~CSUM_DELAY_DATA_IPV6; 961 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr)); 962 } 963#ifdef SCTP 964 if (sw_csum & CSUM_SCTP_IPV6) { 965 sw_csum &= ~CSUM_SCTP_IPV6; 966 sctp_delayed_cksum(m, sizeof(struct ip6_hdr)); 967 } 968#endif 969 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 970 tlen = m->m_pkthdr.len; 971 972 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso) 973 dontfrag = 1; 974 else 975 dontfrag = 0; 976 if (dontfrag && alwaysfrag) { /* case 4 */ 977 /* conflicting request - can't transmit */ 978 error = EMSGSIZE; 979 goto bad; 980 } 981 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* case 2-b */ 982 /* 983 * Even if the DONTFRAG option is specified, we cannot send the 984 * packet when the data length is larger than the MTU of the 985 * outgoing interface. 986 * Notify the error by sending IPV6_PATHMTU ancillary data as 987 * well as returning an error code (the latter is not described 988 * in the API spec.) 989 */ 990 u_int32_t mtu32; 991 struct ip6ctlparam ip6cp; 992 993 mtu32 = (u_int32_t)mtu; 994 bzero(&ip6cp, sizeof(ip6cp)); 995 ip6cp.ip6c_cmdarg = (void *)&mtu32; 996 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 997 (void *)&ip6cp); 998 999 error = EMSGSIZE; 1000 goto bad; 1001 } 1002 1003 /* 1004 * transmit packet without fragmentation 1005 */ 1006 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */ 1007 struct in6_ifaddr *ia6; 1008 1009 ip6 = mtod(m, struct ip6_hdr *); 1010 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1011 if (ia6) { 1012 /* Record statistics for this interface address. */ 1013 ia6->ia_ifa.if_opackets++; 1014 ia6->ia_ifa.if_obytes += m->m_pkthdr.len; 1015 ifa_free(&ia6->ia_ifa); 1016 } 1017 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 1018 goto done; 1019 } 1020 1021 /* 1022 * try to fragment the packet. case 1-b and 3 1023 */ 1024 if (mtu < IPV6_MMTU) { 1025 /* path MTU cannot be less than IPV6_MMTU */ 1026 error = EMSGSIZE; 1027 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1028 goto bad; 1029 } else if (ip6->ip6_plen == 0) { 1030 /* jumbo payload cannot be fragmented */ 1031 error = EMSGSIZE; 1032 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1033 goto bad; 1034 } else { 1035 struct mbuf **mnext, *m_frgpart; 1036 struct ip6_frag *ip6f; 1037 u_int32_t id = htonl(ip6_randomid()); 1038 u_char nextproto; 1039 1040 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len; 1041 1042 /* 1043 * Too large for the destination or interface; 1044 * fragment if possible. 1045 * Must be able to put at least 8 bytes per fragment. 1046 */ 1047 hlen = unfragpartlen; 1048 if (mtu > IPV6_MAXPACKET) 1049 mtu = IPV6_MAXPACKET; 1050 1051 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 1052 if (len < 8) { 1053 error = EMSGSIZE; 1054 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1055 goto bad; 1056 } 1057 1058 /* 1059 * Verify that we have any chance at all of being able to queue 1060 * the packet or packet fragments 1061 */ 1062 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen) 1063 < tlen /* - hlen */)) { 1064 error = ENOBUFS; 1065 IP6STAT_INC(ip6s_odropped); 1066 goto bad; 1067 } 1068 1069 1070 /* 1071 * If the interface will not calculate checksums on 1072 * fragmented packets, then do it here. 1073 * XXX-BZ handle the hw offloading case. Need flags. 1074 */ 1075 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 1076 in6_delayed_cksum(m, plen, hlen); 1077 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6; 1078 } 1079#ifdef SCTP 1080 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) { 1081 sctp_delayed_cksum(m, hlen); 1082 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6; 1083 } 1084#endif 1085 mnext = &m->m_nextpkt; 1086 1087 /* 1088 * Change the next header field of the last header in the 1089 * unfragmentable part. 1090 */ 1091 if (exthdrs.ip6e_rthdr) { 1092 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 1093 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 1094 } else if (exthdrs.ip6e_dest1) { 1095 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 1096 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 1097 } else if (exthdrs.ip6e_hbh) { 1098 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 1099 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 1100 } else { 1101 nextproto = ip6->ip6_nxt; 1102 ip6->ip6_nxt = IPPROTO_FRAGMENT; 1103 } 1104 1105 /* 1106 * Loop through length of segment after first fragment, 1107 * make new header and copy data of each part and link onto 1108 * chain. 1109 */ 1110 m0 = m; 1111 for (off = hlen; off < tlen; off += len) { 1112 m = m_gethdr(M_NOWAIT, MT_DATA); 1113 if (!m) { 1114 error = ENOBUFS; 1115 IP6STAT_INC(ip6s_odropped); 1116 goto sendorfree; 1117 } 1118 m->m_flags = m0->m_flags & M_COPYFLAGS; 1119 *mnext = m; 1120 mnext = &m->m_nextpkt; 1121 m->m_data += max_linkhdr; 1122 mhip6 = mtod(m, struct ip6_hdr *); 1123 *mhip6 = *ip6; 1124 m->m_len = sizeof(*mhip6); 1125 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 1126 if (error) { 1127 IP6STAT_INC(ip6s_odropped); 1128 goto sendorfree; 1129 } 1130 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7)); 1131 if (off + len >= tlen) 1132 len = tlen - off; 1133 else 1134 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1135 mhip6->ip6_plen = htons((u_short)(len + hlen + 1136 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1137 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 1138 error = ENOBUFS; 1139 IP6STAT_INC(ip6s_odropped); 1140 goto sendorfree; 1141 } 1142 m_cat(m, m_frgpart); 1143 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1144 m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum; 1145 m->m_pkthdr.rcvif = NULL; 1146 ip6f->ip6f_reserved = 0; 1147 ip6f->ip6f_ident = id; 1148 ip6f->ip6f_nxt = nextproto; 1149 IP6STAT_INC(ip6s_ofragments); 1150 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1151 } 1152 1153 in6_ifstat_inc(ifp, ifs6_out_fragok); 1154 } 1155 1156 /* 1157 * Remove leading garbages. 1158 */ 1159sendorfree: 1160 m = m0->m_nextpkt; 1161 m0->m_nextpkt = 0; 1162 m_freem(m0); 1163 for (m0 = m; m; m = m0) { 1164 m0 = m->m_nextpkt; 1165 m->m_nextpkt = 0; 1166 if (error == 0) { 1167 /* Record statistics for this interface address. */ 1168 if (ia) { 1169 ia->ia_ifa.if_opackets++; 1170 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1171 } 1172 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 1173 } else 1174 m_freem(m); 1175 } 1176 1177 if (error == 0) 1178 IP6STAT_INC(ip6s_fragmented); 1179 1180done: 1181 if (ro == &ip6route) 1182 RO_RTFREE(ro); 1183 if (ro_pmtu == &ip6route) 1184 RO_RTFREE(ro_pmtu); 1185#ifdef IPSEC 1186 if (sp != NULL) 1187 KEY_FREESP(&sp); 1188#endif 1189 1190 return (error); 1191 1192freehdrs: 1193 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1194 m_freem(exthdrs.ip6e_dest1); 1195 m_freem(exthdrs.ip6e_rthdr); 1196 m_freem(exthdrs.ip6e_dest2); 1197 /* FALLTHROUGH */ 1198bad: 1199 if (m) 1200 m_freem(m); 1201 goto done; 1202} 1203 1204static int 1205ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen) 1206{ 1207 struct mbuf *m; 1208 1209 if (hlen > MCLBYTES) 1210 return (ENOBUFS); /* XXX */ 1211 1212 if (hlen > MLEN) 1213 m = m_getcl(M_NOWAIT, MT_DATA, 0); 1214 else 1215 m = m_get(M_NOWAIT, MT_DATA); 1216 if (m == NULL) 1217 return (ENOBUFS); 1218 m->m_len = hlen; 1219 if (hdr) 1220 bcopy(hdr, mtod(m, caddr_t), hlen); 1221 1222 *mp = m; 1223 return (0); 1224} 1225 1226/* 1227 * Insert jumbo payload option. 1228 */ 1229static int 1230ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1231{ 1232 struct mbuf *mopt; 1233 u_char *optbuf; 1234 u_int32_t v; 1235 1236#define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1237 1238 /* 1239 * If there is no hop-by-hop options header, allocate new one. 1240 * If there is one but it doesn't have enough space to store the 1241 * jumbo payload option, allocate a cluster to store the whole options. 1242 * Otherwise, use it to store the options. 1243 */ 1244 if (exthdrs->ip6e_hbh == 0) { 1245 mopt = m_get(M_NOWAIT, MT_DATA); 1246 if (mopt == NULL) 1247 return (ENOBUFS); 1248 mopt->m_len = JUMBOOPTLEN; 1249 optbuf = mtod(mopt, u_char *); 1250 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1251 exthdrs->ip6e_hbh = mopt; 1252 } else { 1253 struct ip6_hbh *hbh; 1254 1255 mopt = exthdrs->ip6e_hbh; 1256 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1257 /* 1258 * XXX assumption: 1259 * - exthdrs->ip6e_hbh is not referenced from places 1260 * other than exthdrs. 1261 * - exthdrs->ip6e_hbh is not an mbuf chain. 1262 */ 1263 int oldoptlen = mopt->m_len; 1264 struct mbuf *n; 1265 1266 /* 1267 * XXX: give up if the whole (new) hbh header does 1268 * not fit even in an mbuf cluster. 1269 */ 1270 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1271 return (ENOBUFS); 1272 1273 /* 1274 * As a consequence, we must always prepare a cluster 1275 * at this point. 1276 */ 1277 n = m_getcl(M_NOWAIT, MT_DATA, 0); 1278 if (n == NULL) 1279 return (ENOBUFS); 1280 n->m_len = oldoptlen + JUMBOOPTLEN; 1281 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1282 oldoptlen); 1283 optbuf = mtod(n, caddr_t) + oldoptlen; 1284 m_freem(mopt); 1285 mopt = exthdrs->ip6e_hbh = n; 1286 } else { 1287 optbuf = mtod(mopt, u_char *) + mopt->m_len; 1288 mopt->m_len += JUMBOOPTLEN; 1289 } 1290 optbuf[0] = IP6OPT_PADN; 1291 optbuf[1] = 1; 1292 1293 /* 1294 * Adjust the header length according to the pad and 1295 * the jumbo payload option. 1296 */ 1297 hbh = mtod(mopt, struct ip6_hbh *); 1298 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1299 } 1300 1301 /* fill in the option. */ 1302 optbuf[2] = IP6OPT_JUMBO; 1303 optbuf[3] = 4; 1304 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1305 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1306 1307 /* finally, adjust the packet header length */ 1308 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1309 1310 return (0); 1311#undef JUMBOOPTLEN 1312} 1313 1314/* 1315 * Insert fragment header and copy unfragmentable header portions. 1316 */ 1317static int 1318ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1319 struct ip6_frag **frghdrp) 1320{ 1321 struct mbuf *n, *mlast; 1322 1323 if (hlen > sizeof(struct ip6_hdr)) { 1324 n = m_copym(m0, sizeof(struct ip6_hdr), 1325 hlen - sizeof(struct ip6_hdr), M_NOWAIT); 1326 if (n == 0) 1327 return (ENOBUFS); 1328 m->m_next = n; 1329 } else 1330 n = m; 1331 1332 /* Search for the last mbuf of unfragmentable part. */ 1333 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1334 ; 1335 1336 if ((mlast->m_flags & M_EXT) == 0 && 1337 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1338 /* use the trailing space of the last mbuf for the fragment hdr */ 1339 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1340 mlast->m_len); 1341 mlast->m_len += sizeof(struct ip6_frag); 1342 m->m_pkthdr.len += sizeof(struct ip6_frag); 1343 } else { 1344 /* allocate a new mbuf for the fragment header */ 1345 struct mbuf *mfrg; 1346 1347 mfrg = m_get(M_NOWAIT, MT_DATA); 1348 if (mfrg == NULL) 1349 return (ENOBUFS); 1350 mfrg->m_len = sizeof(struct ip6_frag); 1351 *frghdrp = mtod(mfrg, struct ip6_frag *); 1352 mlast->m_next = mfrg; 1353 } 1354 1355 return (0); 1356} 1357 1358static int 1359ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro, 1360 struct ifnet *ifp, struct in6_addr *dst, u_long *mtup, 1361 int *alwaysfragp, u_int fibnum) 1362{ 1363 u_int32_t mtu = 0; 1364 int alwaysfrag = 0; 1365 int error = 0; 1366 1367 if (ro_pmtu != ro) { 1368 /* The first hop and the final destination may differ. */ 1369 struct sockaddr_in6 *sa6_dst = 1370 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1371 if (ro_pmtu->ro_rt && 1372 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || 1373 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { 1374 RTFREE(ro_pmtu->ro_rt); 1375 ro_pmtu->ro_rt = (struct rtentry *)NULL; 1376 } 1377 if (ro_pmtu->ro_rt == NULL) { 1378 bzero(sa6_dst, sizeof(*sa6_dst)); 1379 sa6_dst->sin6_family = AF_INET6; 1380 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1381 sa6_dst->sin6_addr = *dst; 1382 1383 in6_rtalloc(ro_pmtu, fibnum); 1384 } 1385 } 1386 if (ro_pmtu->ro_rt) { 1387 u_int32_t ifmtu; 1388 struct in_conninfo inc; 1389 1390 bzero(&inc, sizeof(inc)); 1391 inc.inc_flags |= INC_ISIPV6; 1392 inc.inc6_faddr = *dst; 1393 1394 if (ifp == NULL) 1395 ifp = ro_pmtu->ro_rt->rt_ifp; 1396 ifmtu = IN6_LINKMTU(ifp); 1397 mtu = tcp_hc_getmtu(&inc); 1398 if (mtu) 1399 mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu); 1400 else 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 framgent 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 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1426 } 1427 } else if (ifp) { 1428 mtu = IN6_LINKMTU(ifp); 1429 } else 1430 error = EHOSTUNREACH; /* XXX */ 1431 1432 *mtup = mtu; 1433 if (alwaysfragp) 1434 *alwaysfragp = alwaysfrag; 1435 return (error); 1436} 1437 1438/* 1439 * IP6 socket option processing. 1440 */ 1441int 1442ip6_ctloutput(struct socket *so, struct sockopt *sopt) 1443{ 1444 int optdatalen, uproto; 1445 void *optdata; 1446 struct inpcb *in6p = sotoinpcb(so); 1447 int error, optval; 1448 int level, op, optname; 1449 int optlen; 1450 struct thread *td; 1451 1452 level = sopt->sopt_level; 1453 op = sopt->sopt_dir; 1454 optname = sopt->sopt_name; 1455 optlen = sopt->sopt_valsize; 1456 td = sopt->sopt_td; 1457 error = 0; 1458 optval = 0; 1459 uproto = (int)so->so_proto->pr_protocol; 1460 1461 if (level != IPPROTO_IPV6) { 1462 error = EINVAL; 1463 1464 if (sopt->sopt_level == SOL_SOCKET && 1465 sopt->sopt_dir == SOPT_SET) { 1466 switch (sopt->sopt_name) { 1467 case SO_REUSEADDR: 1468 INP_WLOCK(in6p); 1469 if ((so->so_options & SO_REUSEADDR) != 0) 1470 in6p->inp_flags2 |= INP_REUSEADDR; 1471 else 1472 in6p->inp_flags2 &= ~INP_REUSEADDR; 1473 INP_WUNLOCK(in6p); 1474 error = 0; 1475 break; 1476 case SO_REUSEPORT: 1477 INP_WLOCK(in6p); 1478 if ((so->so_options & SO_REUSEPORT) != 0) 1479 in6p->inp_flags2 |= INP_REUSEPORT; 1480 else 1481 in6p->inp_flags2 &= ~INP_REUSEPORT; 1482 INP_WUNLOCK(in6p); 1483 error = 0; 1484 break; 1485 case SO_SETFIB: 1486 INP_WLOCK(in6p); 1487 in6p->inp_inc.inc_fibnum = so->so_fibnum; 1488 INP_WUNLOCK(in6p); 1489 error = 0; 1490 break; 1491 default: 1492 break; 1493 } 1494 } 1495 } else { /* level == IPPROTO_IPV6 */ 1496 switch (op) { 1497 1498 case SOPT_SET: 1499 switch (optname) { 1500 case IPV6_2292PKTOPTIONS: 1501#ifdef IPV6_PKTOPTIONS 1502 case IPV6_PKTOPTIONS: 1503#endif 1504 { 1505 struct mbuf *m; 1506 1507 error = soopt_getm(sopt, &m); /* XXX */ 1508 if (error != 0) 1509 break; 1510 error = soopt_mcopyin(sopt, m); /* XXX */ 1511 if (error != 0) 1512 break; 1513 error = ip6_pcbopts(&in6p->in6p_outputopts, 1514 m, so, sopt); 1515 m_freem(m); /* XXX */ 1516 break; 1517 } 1518 1519 /* 1520 * Use of some Hop-by-Hop options or some 1521 * Destination options, might require special 1522 * privilege. That is, normal applications 1523 * (without special privilege) might be forbidden 1524 * from setting certain options in outgoing packets, 1525 * and might never see certain options in received 1526 * packets. [RFC 2292 Section 6] 1527 * KAME specific note: 1528 * KAME prevents non-privileged users from sending or 1529 * receiving ANY hbh/dst options in order to avoid 1530 * overhead of parsing options in the kernel. 1531 */ 1532 case IPV6_RECVHOPOPTS: 1533 case IPV6_RECVDSTOPTS: 1534 case IPV6_RECVRTHDRDSTOPTS: 1535 if (td != NULL) { 1536 error = priv_check(td, 1537 PRIV_NETINET_SETHDROPTS); 1538 if (error) 1539 break; 1540 } 1541 /* FALLTHROUGH */ 1542 case IPV6_UNICAST_HOPS: 1543 case IPV6_HOPLIMIT: 1544 case IPV6_FAITH: 1545 1546 case IPV6_RECVPKTINFO: 1547 case IPV6_RECVHOPLIMIT: 1548 case IPV6_RECVRTHDR: 1549 case IPV6_RECVPATHMTU: 1550 case IPV6_RECVTCLASS: 1551 case IPV6_V6ONLY: 1552 case IPV6_AUTOFLOWLABEL: 1553 case IPV6_BINDANY: 1554 if (optname == IPV6_BINDANY && td != NULL) { 1555 error = priv_check(td, 1556 PRIV_NETINET_BINDANY); 1557 if (error) 1558 break; 1559 } 1560 1561 if (optlen != sizeof(int)) { 1562 error = EINVAL; 1563 break; 1564 } 1565 error = sooptcopyin(sopt, &optval, 1566 sizeof optval, sizeof optval); 1567 if (error) 1568 break; 1569 switch (optname) { 1570 1571 case IPV6_UNICAST_HOPS: 1572 if (optval < -1 || optval >= 256) 1573 error = EINVAL; 1574 else { 1575 /* -1 = kernel default */ 1576 in6p->in6p_hops = optval; 1577 if ((in6p->inp_vflag & 1578 INP_IPV4) != 0) 1579 in6p->inp_ip_ttl = optval; 1580 } 1581 break; 1582#define OPTSET(bit) \ 1583do { \ 1584 INP_WLOCK(in6p); \ 1585 if (optval) \ 1586 in6p->inp_flags |= (bit); \ 1587 else \ 1588 in6p->inp_flags &= ~(bit); \ 1589 INP_WUNLOCK(in6p); \ 1590} while (/*CONSTCOND*/ 0) 1591#define OPTSET2292(bit) \ 1592do { \ 1593 INP_WLOCK(in6p); \ 1594 in6p->inp_flags |= IN6P_RFC2292; \ 1595 if (optval) \ 1596 in6p->inp_flags |= (bit); \ 1597 else \ 1598 in6p->inp_flags &= ~(bit); \ 1599 INP_WUNLOCK(in6p); \ 1600} while (/*CONSTCOND*/ 0) 1601#define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0) 1602 1603 case IPV6_RECVPKTINFO: 1604 /* cannot mix with RFC2292 */ 1605 if (OPTBIT(IN6P_RFC2292)) { 1606 error = EINVAL; 1607 break; 1608 } 1609 OPTSET(IN6P_PKTINFO); 1610 break; 1611 1612 case IPV6_HOPLIMIT: 1613 { 1614 struct ip6_pktopts **optp; 1615 1616 /* cannot mix with RFC2292 */ 1617 if (OPTBIT(IN6P_RFC2292)) { 1618 error = EINVAL; 1619 break; 1620 } 1621 optp = &in6p->in6p_outputopts; 1622 error = ip6_pcbopt(IPV6_HOPLIMIT, 1623 (u_char *)&optval, sizeof(optval), 1624 optp, (td != NULL) ? td->td_ucred : 1625 NULL, uproto); 1626 break; 1627 } 1628 1629 case IPV6_RECVHOPLIMIT: 1630 /* cannot mix with RFC2292 */ 1631 if (OPTBIT(IN6P_RFC2292)) { 1632 error = EINVAL; 1633 break; 1634 } 1635 OPTSET(IN6P_HOPLIMIT); 1636 break; 1637 1638 case IPV6_RECVHOPOPTS: 1639 /* cannot mix with RFC2292 */ 1640 if (OPTBIT(IN6P_RFC2292)) { 1641 error = EINVAL; 1642 break; 1643 } 1644 OPTSET(IN6P_HOPOPTS); 1645 break; 1646 1647 case IPV6_RECVDSTOPTS: 1648 /* cannot mix with RFC2292 */ 1649 if (OPTBIT(IN6P_RFC2292)) { 1650 error = EINVAL; 1651 break; 1652 } 1653 OPTSET(IN6P_DSTOPTS); 1654 break; 1655 1656 case IPV6_RECVRTHDRDSTOPTS: 1657 /* cannot mix with RFC2292 */ 1658 if (OPTBIT(IN6P_RFC2292)) { 1659 error = EINVAL; 1660 break; 1661 } 1662 OPTSET(IN6P_RTHDRDSTOPTS); 1663 break; 1664 1665 case IPV6_RECVRTHDR: 1666 /* cannot mix with RFC2292 */ 1667 if (OPTBIT(IN6P_RFC2292)) { 1668 error = EINVAL; 1669 break; 1670 } 1671 OPTSET(IN6P_RTHDR); 1672 break; 1673 1674 case IPV6_FAITH: 1675 OPTSET(INP_FAITH); 1676 break; 1677 1678 case IPV6_RECVPATHMTU: 1679 /* 1680 * We ignore this option for TCP 1681 * sockets. 1682 * (RFC3542 leaves this case 1683 * unspecified.) 1684 */ 1685 if (uproto != IPPROTO_TCP) 1686 OPTSET(IN6P_MTU); 1687 break; 1688 1689 case IPV6_V6ONLY: 1690 /* 1691 * make setsockopt(IPV6_V6ONLY) 1692 * available only prior to bind(2). 1693 * see ipng mailing list, Jun 22 2001. 1694 */ 1695 if (in6p->inp_lport || 1696 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1697 error = EINVAL; 1698 break; 1699 } 1700 OPTSET(IN6P_IPV6_V6ONLY); 1701 if (optval) 1702 in6p->inp_vflag &= ~INP_IPV4; 1703 else 1704 in6p->inp_vflag |= INP_IPV4; 1705 break; 1706 case IPV6_RECVTCLASS: 1707 /* cannot mix with RFC2292 XXX */ 1708 if (OPTBIT(IN6P_RFC2292)) { 1709 error = EINVAL; 1710 break; 1711 } 1712 OPTSET(IN6P_TCLASS); 1713 break; 1714 case IPV6_AUTOFLOWLABEL: 1715 OPTSET(IN6P_AUTOFLOWLABEL); 1716 break; 1717 1718 case IPV6_BINDANY: 1719 OPTSET(INP_BINDANY); 1720 break; 1721 } 1722 break; 1723 1724 case IPV6_TCLASS: 1725 case IPV6_DONTFRAG: 1726 case IPV6_USE_MIN_MTU: 1727 case IPV6_PREFER_TEMPADDR: 1728 if (optlen != sizeof(optval)) { 1729 error = EINVAL; 1730 break; 1731 } 1732 error = sooptcopyin(sopt, &optval, 1733 sizeof optval, sizeof optval); 1734 if (error) 1735 break; 1736 { 1737 struct ip6_pktopts **optp; 1738 optp = &in6p->in6p_outputopts; 1739 error = ip6_pcbopt(optname, 1740 (u_char *)&optval, sizeof(optval), 1741 optp, (td != NULL) ? td->td_ucred : 1742 NULL, uproto); 1743 break; 1744 } 1745 1746 case IPV6_2292PKTINFO: 1747 case IPV6_2292HOPLIMIT: 1748 case IPV6_2292HOPOPTS: 1749 case IPV6_2292DSTOPTS: 1750 case IPV6_2292RTHDR: 1751 /* RFC 2292 */ 1752 if (optlen != sizeof(int)) { 1753 error = EINVAL; 1754 break; 1755 } 1756 error = sooptcopyin(sopt, &optval, 1757 sizeof optval, sizeof optval); 1758 if (error) 1759 break; 1760 switch (optname) { 1761 case IPV6_2292PKTINFO: 1762 OPTSET2292(IN6P_PKTINFO); 1763 break; 1764 case IPV6_2292HOPLIMIT: 1765 OPTSET2292(IN6P_HOPLIMIT); 1766 break; 1767 case IPV6_2292HOPOPTS: 1768 /* 1769 * Check super-user privilege. 1770 * See comments for IPV6_RECVHOPOPTS. 1771 */ 1772 if (td != NULL) { 1773 error = priv_check(td, 1774 PRIV_NETINET_SETHDROPTS); 1775 if (error) 1776 return (error); 1777 } 1778 OPTSET2292(IN6P_HOPOPTS); 1779 break; 1780 case IPV6_2292DSTOPTS: 1781 if (td != NULL) { 1782 error = priv_check(td, 1783 PRIV_NETINET_SETHDROPTS); 1784 if (error) 1785 return (error); 1786 } 1787 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1788 break; 1789 case IPV6_2292RTHDR: 1790 OPTSET2292(IN6P_RTHDR); 1791 break; 1792 } 1793 break; 1794 case IPV6_PKTINFO: 1795 case IPV6_HOPOPTS: 1796 case IPV6_RTHDR: 1797 case IPV6_DSTOPTS: 1798 case IPV6_RTHDRDSTOPTS: 1799 case IPV6_NEXTHOP: 1800 { 1801 /* new advanced API (RFC3542) */ 1802 u_char *optbuf; 1803 u_char optbuf_storage[MCLBYTES]; 1804 int optlen; 1805 struct ip6_pktopts **optp; 1806 1807 /* cannot mix with RFC2292 */ 1808 if (OPTBIT(IN6P_RFC2292)) { 1809 error = EINVAL; 1810 break; 1811 } 1812 1813 /* 1814 * We only ensure valsize is not too large 1815 * here. Further validation will be done 1816 * later. 1817 */ 1818 error = sooptcopyin(sopt, optbuf_storage, 1819 sizeof(optbuf_storage), 0); 1820 if (error) 1821 break; 1822 optlen = sopt->sopt_valsize; 1823 optbuf = optbuf_storage; 1824 optp = &in6p->in6p_outputopts; 1825 error = ip6_pcbopt(optname, optbuf, optlen, 1826 optp, (td != NULL) ? td->td_ucred : NULL, 1827 uproto); 1828 break; 1829 } 1830#undef OPTSET 1831 1832 case IPV6_MULTICAST_IF: 1833 case IPV6_MULTICAST_HOPS: 1834 case IPV6_MULTICAST_LOOP: 1835 case IPV6_JOIN_GROUP: 1836 case IPV6_LEAVE_GROUP: 1837 case IPV6_MSFILTER: 1838 case MCAST_BLOCK_SOURCE: 1839 case MCAST_UNBLOCK_SOURCE: 1840 case MCAST_JOIN_GROUP: 1841 case MCAST_LEAVE_GROUP: 1842 case MCAST_JOIN_SOURCE_GROUP: 1843 case MCAST_LEAVE_SOURCE_GROUP: 1844 error = ip6_setmoptions(in6p, sopt); 1845 break; 1846 1847 case IPV6_PORTRANGE: 1848 error = sooptcopyin(sopt, &optval, 1849 sizeof optval, sizeof optval); 1850 if (error) 1851 break; 1852 1853 INP_WLOCK(in6p); 1854 switch (optval) { 1855 case IPV6_PORTRANGE_DEFAULT: 1856 in6p->inp_flags &= ~(INP_LOWPORT); 1857 in6p->inp_flags &= ~(INP_HIGHPORT); 1858 break; 1859 1860 case IPV6_PORTRANGE_HIGH: 1861 in6p->inp_flags &= ~(INP_LOWPORT); 1862 in6p->inp_flags |= INP_HIGHPORT; 1863 break; 1864 1865 case IPV6_PORTRANGE_LOW: 1866 in6p->inp_flags &= ~(INP_HIGHPORT); 1867 in6p->inp_flags |= INP_LOWPORT; 1868 break; 1869 1870 default: 1871 error = EINVAL; 1872 break; 1873 } 1874 INP_WUNLOCK(in6p); 1875 break; 1876 1877#ifdef IPSEC 1878 case IPV6_IPSEC_POLICY: 1879 { 1880 caddr_t req; 1881 struct mbuf *m; 1882 1883 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1884 break; 1885 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1886 break; 1887 req = mtod(m, caddr_t); 1888 error = ipsec_set_policy(in6p, optname, req, 1889 m->m_len, (sopt->sopt_td != NULL) ? 1890 sopt->sopt_td->td_ucred : NULL); 1891 m_freem(m); 1892 break; 1893 } 1894#endif /* IPSEC */ 1895 1896 default: 1897 error = ENOPROTOOPT; 1898 break; 1899 } 1900 break; 1901 1902 case SOPT_GET: 1903 switch (optname) { 1904 1905 case IPV6_2292PKTOPTIONS: 1906#ifdef IPV6_PKTOPTIONS 1907 case IPV6_PKTOPTIONS: 1908#endif 1909 /* 1910 * RFC3542 (effectively) deprecated the 1911 * semantics of the 2292-style pktoptions. 1912 * Since it was not reliable in nature (i.e., 1913 * applications had to expect the lack of some 1914 * information after all), it would make sense 1915 * to simplify this part by always returning 1916 * empty data. 1917 */ 1918 sopt->sopt_valsize = 0; 1919 break; 1920 1921 case IPV6_RECVHOPOPTS: 1922 case IPV6_RECVDSTOPTS: 1923 case IPV6_RECVRTHDRDSTOPTS: 1924 case IPV6_UNICAST_HOPS: 1925 case IPV6_RECVPKTINFO: 1926 case IPV6_RECVHOPLIMIT: 1927 case IPV6_RECVRTHDR: 1928 case IPV6_RECVPATHMTU: 1929 1930 case IPV6_FAITH: 1931 case IPV6_V6ONLY: 1932 case IPV6_PORTRANGE: 1933 case IPV6_RECVTCLASS: 1934 case IPV6_AUTOFLOWLABEL: 1935 case IPV6_BINDANY: 1936 switch (optname) { 1937 1938 case IPV6_RECVHOPOPTS: 1939 optval = OPTBIT(IN6P_HOPOPTS); 1940 break; 1941 1942 case IPV6_RECVDSTOPTS: 1943 optval = OPTBIT(IN6P_DSTOPTS); 1944 break; 1945 1946 case IPV6_RECVRTHDRDSTOPTS: 1947 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1948 break; 1949 1950 case IPV6_UNICAST_HOPS: 1951 optval = in6p->in6p_hops; 1952 break; 1953 1954 case IPV6_RECVPKTINFO: 1955 optval = OPTBIT(IN6P_PKTINFO); 1956 break; 1957 1958 case IPV6_RECVHOPLIMIT: 1959 optval = OPTBIT(IN6P_HOPLIMIT); 1960 break; 1961 1962 case IPV6_RECVRTHDR: 1963 optval = OPTBIT(IN6P_RTHDR); 1964 break; 1965 1966 case IPV6_RECVPATHMTU: 1967 optval = OPTBIT(IN6P_MTU); 1968 break; 1969 1970 case IPV6_FAITH: 1971 optval = OPTBIT(INP_FAITH); 1972 break; 1973 1974 case IPV6_V6ONLY: 1975 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1976 break; 1977 1978 case IPV6_PORTRANGE: 1979 { 1980 int flags; 1981 flags = in6p->inp_flags; 1982 if (flags & INP_HIGHPORT) 1983 optval = IPV6_PORTRANGE_HIGH; 1984 else if (flags & INP_LOWPORT) 1985 optval = IPV6_PORTRANGE_LOW; 1986 else 1987 optval = 0; 1988 break; 1989 } 1990 case IPV6_RECVTCLASS: 1991 optval = OPTBIT(IN6P_TCLASS); 1992 break; 1993 1994 case IPV6_AUTOFLOWLABEL: 1995 optval = OPTBIT(IN6P_AUTOFLOWLABEL); 1996 break; 1997 1998 case IPV6_BINDANY: 1999 optval = OPTBIT(INP_BINDANY); 2000 break; 2001 } 2002 if (error) 2003 break; 2004 error = sooptcopyout(sopt, &optval, 2005 sizeof optval); 2006 break; 2007 2008 case IPV6_PATHMTU: 2009 { 2010 u_long pmtu = 0; 2011 struct ip6_mtuinfo mtuinfo; 2012 struct route_in6 sro; 2013 2014 bzero(&sro, sizeof(sro)); 2015 2016 if (!(so->so_state & SS_ISCONNECTED)) 2017 return (ENOTCONN); 2018 /* 2019 * XXX: we dot not consider the case of source 2020 * routing, or optional information to specify 2021 * the outgoing interface. 2022 */ 2023 error = ip6_getpmtu(&sro, NULL, NULL, 2024 &in6p->in6p_faddr, &pmtu, NULL, 2025 so->so_fibnum); 2026 if (sro.ro_rt) 2027 RTFREE(sro.ro_rt); 2028 if (error) 2029 break; 2030 if (pmtu > IPV6_MAXPACKET) 2031 pmtu = IPV6_MAXPACKET; 2032 2033 bzero(&mtuinfo, sizeof(mtuinfo)); 2034 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 2035 optdata = (void *)&mtuinfo; 2036 optdatalen = sizeof(mtuinfo); 2037 error = sooptcopyout(sopt, optdata, 2038 optdatalen); 2039 break; 2040 } 2041 2042 case IPV6_2292PKTINFO: 2043 case IPV6_2292HOPLIMIT: 2044 case IPV6_2292HOPOPTS: 2045 case IPV6_2292RTHDR: 2046 case IPV6_2292DSTOPTS: 2047 switch (optname) { 2048 case IPV6_2292PKTINFO: 2049 optval = OPTBIT(IN6P_PKTINFO); 2050 break; 2051 case IPV6_2292HOPLIMIT: 2052 optval = OPTBIT(IN6P_HOPLIMIT); 2053 break; 2054 case IPV6_2292HOPOPTS: 2055 optval = OPTBIT(IN6P_HOPOPTS); 2056 break; 2057 case IPV6_2292RTHDR: 2058 optval = OPTBIT(IN6P_RTHDR); 2059 break; 2060 case IPV6_2292DSTOPTS: 2061 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 2062 break; 2063 } 2064 error = sooptcopyout(sopt, &optval, 2065 sizeof optval); 2066 break; 2067 case IPV6_PKTINFO: 2068 case IPV6_HOPOPTS: 2069 case IPV6_RTHDR: 2070 case IPV6_DSTOPTS: 2071 case IPV6_RTHDRDSTOPTS: 2072 case IPV6_NEXTHOP: 2073 case IPV6_TCLASS: 2074 case IPV6_DONTFRAG: 2075 case IPV6_USE_MIN_MTU: 2076 case IPV6_PREFER_TEMPADDR: 2077 error = ip6_getpcbopt(in6p->in6p_outputopts, 2078 optname, sopt); 2079 break; 2080 2081 case IPV6_MULTICAST_IF: 2082 case IPV6_MULTICAST_HOPS: 2083 case IPV6_MULTICAST_LOOP: 2084 case IPV6_MSFILTER: 2085 error = ip6_getmoptions(in6p, sopt); 2086 break; 2087 2088#ifdef IPSEC 2089 case IPV6_IPSEC_POLICY: 2090 { 2091 caddr_t req = NULL; 2092 size_t len = 0; 2093 struct mbuf *m = NULL; 2094 struct mbuf **mp = &m; 2095 size_t ovalsize = sopt->sopt_valsize; 2096 caddr_t oval = (caddr_t)sopt->sopt_val; 2097 2098 error = soopt_getm(sopt, &m); /* XXX */ 2099 if (error != 0) 2100 break; 2101 error = soopt_mcopyin(sopt, m); /* XXX */ 2102 if (error != 0) 2103 break; 2104 sopt->sopt_valsize = ovalsize; 2105 sopt->sopt_val = oval; 2106 if (m) { 2107 req = mtod(m, caddr_t); 2108 len = m->m_len; 2109 } 2110 error = ipsec_get_policy(in6p, req, len, mp); 2111 if (error == 0) 2112 error = soopt_mcopyout(sopt, m); /* XXX */ 2113 if (error == 0 && m) 2114 m_freem(m); 2115 break; 2116 } 2117#endif /* IPSEC */ 2118 2119 default: 2120 error = ENOPROTOOPT; 2121 break; 2122 } 2123 break; 2124 } 2125 } 2126 return (error); 2127} 2128 2129int 2130ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt) 2131{ 2132 int error = 0, optval, optlen; 2133 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2134 struct inpcb *in6p = sotoinpcb(so); 2135 int level, op, optname; 2136 2137 level = sopt->sopt_level; 2138 op = sopt->sopt_dir; 2139 optname = sopt->sopt_name; 2140 optlen = sopt->sopt_valsize; 2141 2142 if (level != IPPROTO_IPV6) { 2143 return (EINVAL); 2144 } 2145 2146 switch (optname) { 2147 case IPV6_CHECKSUM: 2148 /* 2149 * For ICMPv6 sockets, no modification allowed for checksum 2150 * offset, permit "no change" values to help existing apps. 2151 * 2152 * RFC3542 says: "An attempt to set IPV6_CHECKSUM 2153 * for an ICMPv6 socket will fail." 2154 * The current behavior does not meet RFC3542. 2155 */ 2156 switch (op) { 2157 case SOPT_SET: 2158 if (optlen != sizeof(int)) { 2159 error = EINVAL; 2160 break; 2161 } 2162 error = sooptcopyin(sopt, &optval, sizeof(optval), 2163 sizeof(optval)); 2164 if (error) 2165 break; 2166 if ((optval % 2) != 0) { 2167 /* the API assumes even offset values */ 2168 error = EINVAL; 2169 } else if (so->so_proto->pr_protocol == 2170 IPPROTO_ICMPV6) { 2171 if (optval != icmp6off) 2172 error = EINVAL; 2173 } else 2174 in6p->in6p_cksum = optval; 2175 break; 2176 2177 case SOPT_GET: 2178 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2179 optval = icmp6off; 2180 else 2181 optval = in6p->in6p_cksum; 2182 2183 error = sooptcopyout(sopt, &optval, sizeof(optval)); 2184 break; 2185 2186 default: 2187 error = EINVAL; 2188 break; 2189 } 2190 break; 2191 2192 default: 2193 error = ENOPROTOOPT; 2194 break; 2195 } 2196 2197 return (error); 2198} 2199 2200/* 2201 * Set up IP6 options in pcb for insertion in output packets or 2202 * specifying behavior of outgoing packets. 2203 */ 2204static int 2205ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, 2206 struct socket *so, struct sockopt *sopt) 2207{ 2208 struct ip6_pktopts *opt = *pktopt; 2209 int error = 0; 2210 struct thread *td = sopt->sopt_td; 2211 2212 /* turn off any old options. */ 2213 if (opt) { 2214#ifdef DIAGNOSTIC 2215 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2216 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2217 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2218 printf("ip6_pcbopts: all specified options are cleared.\n"); 2219#endif 2220 ip6_clearpktopts(opt, -1); 2221 } else 2222 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2223 *pktopt = NULL; 2224 2225 if (!m || m->m_len == 0) { 2226 /* 2227 * Only turning off any previous options, regardless of 2228 * whether the opt is just created or given. 2229 */ 2230 free(opt, M_IP6OPT); 2231 return (0); 2232 } 2233 2234 /* set options specified by user. */ 2235 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ? 2236 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) { 2237 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2238 free(opt, M_IP6OPT); 2239 return (error); 2240 } 2241 *pktopt = opt; 2242 return (0); 2243} 2244 2245/* 2246 * initialize ip6_pktopts. beware that there are non-zero default values in 2247 * the struct. 2248 */ 2249void 2250ip6_initpktopts(struct ip6_pktopts *opt) 2251{ 2252 2253 bzero(opt, sizeof(*opt)); 2254 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2255 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2256 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2257 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; 2258} 2259 2260static int 2261ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2262 struct ucred *cred, int uproto) 2263{ 2264 struct ip6_pktopts *opt; 2265 2266 if (*pktopt == NULL) { 2267 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2268 M_WAITOK); 2269 ip6_initpktopts(*pktopt); 2270 } 2271 opt = *pktopt; 2272 2273 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); 2274} 2275 2276static int 2277ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 2278{ 2279 void *optdata = NULL; 2280 int optdatalen = 0; 2281 struct ip6_ext *ip6e; 2282 int error = 0; 2283 struct in6_pktinfo null_pktinfo; 2284 int deftclass = 0, on; 2285 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2286 int defpreftemp = IP6PO_TEMPADDR_SYSTEM; 2287 2288 switch (optname) { 2289 case IPV6_PKTINFO: 2290 if (pktopt && pktopt->ip6po_pktinfo) 2291 optdata = (void *)pktopt->ip6po_pktinfo; 2292 else { 2293 /* XXX: we don't have to do this every time... */ 2294 bzero(&null_pktinfo, sizeof(null_pktinfo)); 2295 optdata = (void *)&null_pktinfo; 2296 } 2297 optdatalen = sizeof(struct in6_pktinfo); 2298 break; 2299 case IPV6_TCLASS: 2300 if (pktopt && pktopt->ip6po_tclass >= 0) 2301 optdata = (void *)&pktopt->ip6po_tclass; 2302 else 2303 optdata = (void *)&deftclass; 2304 optdatalen = sizeof(int); 2305 break; 2306 case IPV6_HOPOPTS: 2307 if (pktopt && pktopt->ip6po_hbh) { 2308 optdata = (void *)pktopt->ip6po_hbh; 2309 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2310 optdatalen = (ip6e->ip6e_len + 1) << 3; 2311 } 2312 break; 2313 case IPV6_RTHDR: 2314 if (pktopt && pktopt->ip6po_rthdr) { 2315 optdata = (void *)pktopt->ip6po_rthdr; 2316 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2317 optdatalen = (ip6e->ip6e_len + 1) << 3; 2318 } 2319 break; 2320 case IPV6_RTHDRDSTOPTS: 2321 if (pktopt && pktopt->ip6po_dest1) { 2322 optdata = (void *)pktopt->ip6po_dest1; 2323 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2324 optdatalen = (ip6e->ip6e_len + 1) << 3; 2325 } 2326 break; 2327 case IPV6_DSTOPTS: 2328 if (pktopt && pktopt->ip6po_dest2) { 2329 optdata = (void *)pktopt->ip6po_dest2; 2330 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2331 optdatalen = (ip6e->ip6e_len + 1) << 3; 2332 } 2333 break; 2334 case IPV6_NEXTHOP: 2335 if (pktopt && pktopt->ip6po_nexthop) { 2336 optdata = (void *)pktopt->ip6po_nexthop; 2337 optdatalen = pktopt->ip6po_nexthop->sa_len; 2338 } 2339 break; 2340 case IPV6_USE_MIN_MTU: 2341 if (pktopt) 2342 optdata = (void *)&pktopt->ip6po_minmtu; 2343 else 2344 optdata = (void *)&defminmtu; 2345 optdatalen = sizeof(int); 2346 break; 2347 case IPV6_DONTFRAG: 2348 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2349 on = 1; 2350 else 2351 on = 0; 2352 optdata = (void *)&on; 2353 optdatalen = sizeof(on); 2354 break; 2355 case IPV6_PREFER_TEMPADDR: 2356 if (pktopt) 2357 optdata = (void *)&pktopt->ip6po_prefer_tempaddr; 2358 else 2359 optdata = (void *)&defpreftemp; 2360 optdatalen = sizeof(int); 2361 break; 2362 default: /* should not happen */ 2363#ifdef DIAGNOSTIC 2364 panic("ip6_getpcbopt: unexpected option\n"); 2365#endif 2366 return (ENOPROTOOPT); 2367 } 2368 2369 error = sooptcopyout(sopt, optdata, optdatalen); 2370 2371 return (error); 2372} 2373 2374void 2375ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2376{ 2377 if (pktopt == NULL) 2378 return; 2379 2380 if (optname == -1 || optname == IPV6_PKTINFO) { 2381 if (pktopt->ip6po_pktinfo) 2382 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2383 pktopt->ip6po_pktinfo = NULL; 2384 } 2385 if (optname == -1 || optname == IPV6_HOPLIMIT) 2386 pktopt->ip6po_hlim = -1; 2387 if (optname == -1 || optname == IPV6_TCLASS) 2388 pktopt->ip6po_tclass = -1; 2389 if (optname == -1 || optname == IPV6_NEXTHOP) { 2390 if (pktopt->ip6po_nextroute.ro_rt) { 2391 RTFREE(pktopt->ip6po_nextroute.ro_rt); 2392 pktopt->ip6po_nextroute.ro_rt = NULL; 2393 } 2394 if (pktopt->ip6po_nexthop) 2395 free(pktopt->ip6po_nexthop, M_IP6OPT); 2396 pktopt->ip6po_nexthop = NULL; 2397 } 2398 if (optname == -1 || optname == IPV6_HOPOPTS) { 2399 if (pktopt->ip6po_hbh) 2400 free(pktopt->ip6po_hbh, M_IP6OPT); 2401 pktopt->ip6po_hbh = NULL; 2402 } 2403 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2404 if (pktopt->ip6po_dest1) 2405 free(pktopt->ip6po_dest1, M_IP6OPT); 2406 pktopt->ip6po_dest1 = NULL; 2407 } 2408 if (optname == -1 || optname == IPV6_RTHDR) { 2409 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2410 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2411 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2412 if (pktopt->ip6po_route.ro_rt) { 2413 RTFREE(pktopt->ip6po_route.ro_rt); 2414 pktopt->ip6po_route.ro_rt = NULL; 2415 } 2416 } 2417 if (optname == -1 || optname == IPV6_DSTOPTS) { 2418 if (pktopt->ip6po_dest2) 2419 free(pktopt->ip6po_dest2, M_IP6OPT); 2420 pktopt->ip6po_dest2 = NULL; 2421 } 2422} 2423 2424#define PKTOPT_EXTHDRCPY(type) \ 2425do {\ 2426 if (src->type) {\ 2427 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2428 dst->type = malloc(hlen, M_IP6OPT, canwait);\ 2429 if (dst->type == NULL && canwait == M_NOWAIT)\ 2430 goto bad;\ 2431 bcopy(src->type, dst->type, hlen);\ 2432 }\ 2433} while (/*CONSTCOND*/ 0) 2434 2435static int 2436copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2437{ 2438 if (dst == NULL || src == NULL) { 2439 printf("ip6_clearpktopts: invalid argument\n"); 2440 return (EINVAL); 2441 } 2442 2443 dst->ip6po_hlim = src->ip6po_hlim; 2444 dst->ip6po_tclass = src->ip6po_tclass; 2445 dst->ip6po_flags = src->ip6po_flags; 2446 dst->ip6po_minmtu = src->ip6po_minmtu; 2447 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr; 2448 if (src->ip6po_pktinfo) { 2449 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2450 M_IP6OPT, canwait); 2451 if (dst->ip6po_pktinfo == NULL) 2452 goto bad; 2453 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2454 } 2455 if (src->ip6po_nexthop) { 2456 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2457 M_IP6OPT, canwait); 2458 if (dst->ip6po_nexthop == NULL) 2459 goto bad; 2460 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop, 2461 src->ip6po_nexthop->sa_len); 2462 } 2463 PKTOPT_EXTHDRCPY(ip6po_hbh); 2464 PKTOPT_EXTHDRCPY(ip6po_dest1); 2465 PKTOPT_EXTHDRCPY(ip6po_dest2); 2466 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2467 return (0); 2468 2469 bad: 2470 ip6_clearpktopts(dst, -1); 2471 return (ENOBUFS); 2472} 2473#undef PKTOPT_EXTHDRCPY 2474 2475struct ip6_pktopts * 2476ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2477{ 2478 int error; 2479 struct ip6_pktopts *dst; 2480 2481 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2482 if (dst == NULL) 2483 return (NULL); 2484 ip6_initpktopts(dst); 2485 2486 if ((error = copypktopts(dst, src, canwait)) != 0) { 2487 free(dst, M_IP6OPT); 2488 return (NULL); 2489 } 2490 2491 return (dst); 2492} 2493 2494void 2495ip6_freepcbopts(struct ip6_pktopts *pktopt) 2496{ 2497 if (pktopt == NULL) 2498 return; 2499 2500 ip6_clearpktopts(pktopt, -1); 2501 2502 free(pktopt, M_IP6OPT); 2503} 2504 2505/* 2506 * Set IPv6 outgoing packet options based on advanced API. 2507 */ 2508int 2509ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2510 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto) 2511{ 2512 struct cmsghdr *cm = 0; 2513 2514 if (control == NULL || opt == NULL) 2515 return (EINVAL); 2516 2517 ip6_initpktopts(opt); 2518 if (stickyopt) { 2519 int error; 2520 2521 /* 2522 * If stickyopt is provided, make a local copy of the options 2523 * for this particular packet, then override them by ancillary 2524 * objects. 2525 * XXX: copypktopts() does not copy the cached route to a next 2526 * hop (if any). This is not very good in terms of efficiency, 2527 * but we can allow this since this option should be rarely 2528 * used. 2529 */ 2530 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2531 return (error); 2532 } 2533 2534 /* 2535 * XXX: Currently, we assume all the optional information is stored 2536 * in a single mbuf. 2537 */ 2538 if (control->m_next) 2539 return (EINVAL); 2540 2541 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2542 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2543 int error; 2544 2545 if (control->m_len < CMSG_LEN(0)) 2546 return (EINVAL); 2547 2548 cm = mtod(control, struct cmsghdr *); 2549 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2550 return (EINVAL); 2551 if (cm->cmsg_level != IPPROTO_IPV6) 2552 continue; 2553 2554 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2555 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); 2556 if (error) 2557 return (error); 2558 } 2559 2560 return (0); 2561} 2562 2563/* 2564 * Set a particular packet option, as a sticky option or an ancillary data 2565 * item. "len" can be 0 only when it's a sticky option. 2566 * We have 4 cases of combination of "sticky" and "cmsg": 2567 * "sticky=0, cmsg=0": impossible 2568 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2569 * "sticky=1, cmsg=0": RFC3542 socket option 2570 * "sticky=1, cmsg=1": RFC2292 socket option 2571 */ 2572static int 2573ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2574 struct ucred *cred, int sticky, int cmsg, int uproto) 2575{ 2576 int minmtupolicy, preftemp; 2577 int error; 2578 2579 if (!sticky && !cmsg) { 2580#ifdef DIAGNOSTIC 2581 printf("ip6_setpktopt: impossible case\n"); 2582#endif 2583 return (EINVAL); 2584 } 2585 2586 /* 2587 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2588 * not be specified in the context of RFC3542. Conversely, 2589 * RFC3542 types should not be specified in the context of RFC2292. 2590 */ 2591 if (!cmsg) { 2592 switch (optname) { 2593 case IPV6_2292PKTINFO: 2594 case IPV6_2292HOPLIMIT: 2595 case IPV6_2292NEXTHOP: 2596 case IPV6_2292HOPOPTS: 2597 case IPV6_2292DSTOPTS: 2598 case IPV6_2292RTHDR: 2599 case IPV6_2292PKTOPTIONS: 2600 return (ENOPROTOOPT); 2601 } 2602 } 2603 if (sticky && cmsg) { 2604 switch (optname) { 2605 case IPV6_PKTINFO: 2606 case IPV6_HOPLIMIT: 2607 case IPV6_NEXTHOP: 2608 case IPV6_HOPOPTS: 2609 case IPV6_DSTOPTS: 2610 case IPV6_RTHDRDSTOPTS: 2611 case IPV6_RTHDR: 2612 case IPV6_USE_MIN_MTU: 2613 case IPV6_DONTFRAG: 2614 case IPV6_TCLASS: 2615 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */ 2616 return (ENOPROTOOPT); 2617 } 2618 } 2619 2620 switch (optname) { 2621 case IPV6_2292PKTINFO: 2622 case IPV6_PKTINFO: 2623 { 2624 struct ifnet *ifp = NULL; 2625 struct in6_pktinfo *pktinfo; 2626 2627 if (len != sizeof(struct in6_pktinfo)) 2628 return (EINVAL); 2629 2630 pktinfo = (struct in6_pktinfo *)buf; 2631 2632 /* 2633 * An application can clear any sticky IPV6_PKTINFO option by 2634 * doing a "regular" setsockopt with ipi6_addr being 2635 * in6addr_any and ipi6_ifindex being zero. 2636 * [RFC 3542, Section 6] 2637 */ 2638 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2639 pktinfo->ipi6_ifindex == 0 && 2640 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2641 ip6_clearpktopts(opt, optname); 2642 break; 2643 } 2644 2645 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2646 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2647 return (EINVAL); 2648 } 2649 2650 /* validate the interface index if specified. */ 2651 if (pktinfo->ipi6_ifindex > V_if_index || 2652 pktinfo->ipi6_ifindex < 0) { 2653 return (ENXIO); 2654 } 2655 if (pktinfo->ipi6_ifindex) { 2656 ifp = ifnet_byindex(pktinfo->ipi6_ifindex); 2657 if (ifp == NULL) 2658 return (ENXIO); 2659 } 2660 2661 /* 2662 * We store the address anyway, and let in6_selectsrc() 2663 * validate the specified address. This is because ipi6_addr 2664 * may not have enough information about its scope zone, and 2665 * we may need additional information (such as outgoing 2666 * interface or the scope zone of a destination address) to 2667 * disambiguate the scope. 2668 * XXX: the delay of the validation may confuse the 2669 * application when it is used as a sticky option. 2670 */ 2671 if (opt->ip6po_pktinfo == NULL) { 2672 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2673 M_IP6OPT, M_NOWAIT); 2674 if (opt->ip6po_pktinfo == NULL) 2675 return (ENOBUFS); 2676 } 2677 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo)); 2678 break; 2679 } 2680 2681 case IPV6_2292HOPLIMIT: 2682 case IPV6_HOPLIMIT: 2683 { 2684 int *hlimp; 2685 2686 /* 2687 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2688 * to simplify the ordering among hoplimit options. 2689 */ 2690 if (optname == IPV6_HOPLIMIT && sticky) 2691 return (ENOPROTOOPT); 2692 2693 if (len != sizeof(int)) 2694 return (EINVAL); 2695 hlimp = (int *)buf; 2696 if (*hlimp < -1 || *hlimp > 255) 2697 return (EINVAL); 2698 2699 opt->ip6po_hlim = *hlimp; 2700 break; 2701 } 2702 2703 case IPV6_TCLASS: 2704 { 2705 int tclass; 2706 2707 if (len != sizeof(int)) 2708 return (EINVAL); 2709 tclass = *(int *)buf; 2710 if (tclass < -1 || tclass > 255) 2711 return (EINVAL); 2712 2713 opt->ip6po_tclass = tclass; 2714 break; 2715 } 2716 2717 case IPV6_2292NEXTHOP: 2718 case IPV6_NEXTHOP: 2719 if (cred != NULL) { 2720 error = priv_check_cred(cred, 2721 PRIV_NETINET_SETHDROPTS, 0); 2722 if (error) 2723 return (error); 2724 } 2725 2726 if (len == 0) { /* just remove the option */ 2727 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2728 break; 2729 } 2730 2731 /* check if cmsg_len is large enough for sa_len */ 2732 if (len < sizeof(struct sockaddr) || len < *buf) 2733 return (EINVAL); 2734 2735 switch (((struct sockaddr *)buf)->sa_family) { 2736 case AF_INET6: 2737 { 2738 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 2739 int error; 2740 2741 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 2742 return (EINVAL); 2743 2744 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 2745 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 2746 return (EINVAL); 2747 } 2748 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone)) 2749 != 0) { 2750 return (error); 2751 } 2752 break; 2753 } 2754 case AF_LINK: /* should eventually be supported */ 2755 default: 2756 return (EAFNOSUPPORT); 2757 } 2758 2759 /* turn off the previous option, then set the new option. */ 2760 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2761 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 2762 if (opt->ip6po_nexthop == NULL) 2763 return (ENOBUFS); 2764 bcopy(buf, opt->ip6po_nexthop, *buf); 2765 break; 2766 2767 case IPV6_2292HOPOPTS: 2768 case IPV6_HOPOPTS: 2769 { 2770 struct ip6_hbh *hbh; 2771 int hbhlen; 2772 2773 /* 2774 * XXX: We don't allow a non-privileged user to set ANY HbH 2775 * options, since per-option restriction has too much 2776 * overhead. 2777 */ 2778 if (cred != NULL) { 2779 error = priv_check_cred(cred, 2780 PRIV_NETINET_SETHDROPTS, 0); 2781 if (error) 2782 return (error); 2783 } 2784 2785 if (len == 0) { 2786 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2787 break; /* just remove the option */ 2788 } 2789 2790 /* message length validation */ 2791 if (len < sizeof(struct ip6_hbh)) 2792 return (EINVAL); 2793 hbh = (struct ip6_hbh *)buf; 2794 hbhlen = (hbh->ip6h_len + 1) << 3; 2795 if (len != hbhlen) 2796 return (EINVAL); 2797 2798 /* turn off the previous option, then set the new option. */ 2799 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2800 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 2801 if (opt->ip6po_hbh == NULL) 2802 return (ENOBUFS); 2803 bcopy(hbh, opt->ip6po_hbh, hbhlen); 2804 2805 break; 2806 } 2807 2808 case IPV6_2292DSTOPTS: 2809 case IPV6_DSTOPTS: 2810 case IPV6_RTHDRDSTOPTS: 2811 { 2812 struct ip6_dest *dest, **newdest = NULL; 2813 int destlen; 2814 2815 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */ 2816 error = priv_check_cred(cred, 2817 PRIV_NETINET_SETHDROPTS, 0); 2818 if (error) 2819 return (error); 2820 } 2821 2822 if (len == 0) { 2823 ip6_clearpktopts(opt, optname); 2824 break; /* just remove the option */ 2825 } 2826 2827 /* message length validation */ 2828 if (len < sizeof(struct ip6_dest)) 2829 return (EINVAL); 2830 dest = (struct ip6_dest *)buf; 2831 destlen = (dest->ip6d_len + 1) << 3; 2832 if (len != destlen) 2833 return (EINVAL); 2834 2835 /* 2836 * Determine the position that the destination options header 2837 * should be inserted; before or after the routing header. 2838 */ 2839 switch (optname) { 2840 case IPV6_2292DSTOPTS: 2841 /* 2842 * The old advacned API is ambiguous on this point. 2843 * Our approach is to determine the position based 2844 * according to the existence of a routing header. 2845 * Note, however, that this depends on the order of the 2846 * extension headers in the ancillary data; the 1st 2847 * part of the destination options header must appear 2848 * before the routing header in the ancillary data, 2849 * too. 2850 * RFC3542 solved the ambiguity by introducing 2851 * separate ancillary data or option types. 2852 */ 2853 if (opt->ip6po_rthdr == NULL) 2854 newdest = &opt->ip6po_dest1; 2855 else 2856 newdest = &opt->ip6po_dest2; 2857 break; 2858 case IPV6_RTHDRDSTOPTS: 2859 newdest = &opt->ip6po_dest1; 2860 break; 2861 case IPV6_DSTOPTS: 2862 newdest = &opt->ip6po_dest2; 2863 break; 2864 } 2865 2866 /* turn off the previous option, then set the new option. */ 2867 ip6_clearpktopts(opt, optname); 2868 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 2869 if (*newdest == NULL) 2870 return (ENOBUFS); 2871 bcopy(dest, *newdest, destlen); 2872 2873 break; 2874 } 2875 2876 case IPV6_2292RTHDR: 2877 case IPV6_RTHDR: 2878 { 2879 struct ip6_rthdr *rth; 2880 int rthlen; 2881 2882 if (len == 0) { 2883 ip6_clearpktopts(opt, IPV6_RTHDR); 2884 break; /* just remove the option */ 2885 } 2886 2887 /* message length validation */ 2888 if (len < sizeof(struct ip6_rthdr)) 2889 return (EINVAL); 2890 rth = (struct ip6_rthdr *)buf; 2891 rthlen = (rth->ip6r_len + 1) << 3; 2892 if (len != rthlen) 2893 return (EINVAL); 2894 2895 switch (rth->ip6r_type) { 2896 case IPV6_RTHDR_TYPE_0: 2897 if (rth->ip6r_len == 0) /* must contain one addr */ 2898 return (EINVAL); 2899 if (rth->ip6r_len % 2) /* length must be even */ 2900 return (EINVAL); 2901 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 2902 return (EINVAL); 2903 break; 2904 default: 2905 return (EINVAL); /* not supported */ 2906 } 2907 2908 /* turn off the previous option */ 2909 ip6_clearpktopts(opt, IPV6_RTHDR); 2910 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 2911 if (opt->ip6po_rthdr == NULL) 2912 return (ENOBUFS); 2913 bcopy(rth, opt->ip6po_rthdr, rthlen); 2914 2915 break; 2916 } 2917 2918 case IPV6_USE_MIN_MTU: 2919 if (len != sizeof(int)) 2920 return (EINVAL); 2921 minmtupolicy = *(int *)buf; 2922 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 2923 minmtupolicy != IP6PO_MINMTU_DISABLE && 2924 minmtupolicy != IP6PO_MINMTU_ALL) { 2925 return (EINVAL); 2926 } 2927 opt->ip6po_minmtu = minmtupolicy; 2928 break; 2929 2930 case IPV6_DONTFRAG: 2931 if (len != sizeof(int)) 2932 return (EINVAL); 2933 2934 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 2935 /* 2936 * we ignore this option for TCP sockets. 2937 * (RFC3542 leaves this case unspecified.) 2938 */ 2939 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 2940 } else 2941 opt->ip6po_flags |= IP6PO_DONTFRAG; 2942 break; 2943 2944 case IPV6_PREFER_TEMPADDR: 2945 if (len != sizeof(int)) 2946 return (EINVAL); 2947 preftemp = *(int *)buf; 2948 if (preftemp != IP6PO_TEMPADDR_SYSTEM && 2949 preftemp != IP6PO_TEMPADDR_NOTPREFER && 2950 preftemp != IP6PO_TEMPADDR_PREFER) { 2951 return (EINVAL); 2952 } 2953 opt->ip6po_prefer_tempaddr = preftemp; 2954 break; 2955 2956 default: 2957 return (ENOPROTOOPT); 2958 } /* end of switch */ 2959 2960 return (0); 2961} 2962 2963/* 2964 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 2965 * packet to the input queue of a specified interface. Note that this 2966 * calls the output routine of the loopback "driver", but with an interface 2967 * pointer that might NOT be &loif -- easier than replicating that code here. 2968 */ 2969void 2970ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst) 2971{ 2972 struct mbuf *copym; 2973 struct ip6_hdr *ip6; 2974 2975 copym = m_copy(m, 0, M_COPYALL); 2976 if (copym == NULL) 2977 return; 2978 2979 /* 2980 * Make sure to deep-copy IPv6 header portion in case the data 2981 * is in an mbuf cluster, so that we can safely override the IPv6 2982 * header portion later. 2983 */ 2984 if ((copym->m_flags & M_EXT) != 0 || 2985 copym->m_len < sizeof(struct ip6_hdr)) { 2986 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 2987 if (copym == NULL) 2988 return; 2989 } 2990 2991#ifdef DIAGNOSTIC 2992 if (copym->m_len < sizeof(*ip6)) { 2993 m_freem(copym); 2994 return; 2995 } 2996#endif 2997 2998 ip6 = mtod(copym, struct ip6_hdr *); 2999 /* 3000 * clear embedded scope identifiers if necessary. 3001 * in6_clearscope will touch the addresses only when necessary. 3002 */ 3003 in6_clearscope(&ip6->ip6_src); 3004 in6_clearscope(&ip6->ip6_dst); 3005 3006 (void)if_simloop(ifp, copym, dst->sin6_family, 0); 3007} 3008 3009/* 3010 * Chop IPv6 header off from the payload. 3011 */ 3012static int 3013ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3014{ 3015 struct mbuf *mh; 3016 struct ip6_hdr *ip6; 3017 3018 ip6 = mtod(m, struct ip6_hdr *); 3019 if (m->m_len > sizeof(*ip6)) { 3020 mh = m_gethdr(M_NOWAIT, MT_DATA); 3021 if (mh == NULL) { 3022 m_freem(m); 3023 return ENOBUFS; 3024 } 3025 m_move_pkthdr(mh, m); 3026 MH_ALIGN(mh, sizeof(*ip6)); 3027 m->m_len -= sizeof(*ip6); 3028 m->m_data += sizeof(*ip6); 3029 mh->m_next = m; 3030 m = mh; 3031 m->m_len = sizeof(*ip6); 3032 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 3033 } 3034 exthdrs->ip6e_ip6 = m; 3035 return 0; 3036} 3037 3038/* 3039 * Compute IPv6 extension header length. 3040 */ 3041int 3042ip6_optlen(struct inpcb *in6p) 3043{ 3044 int len; 3045 3046 if (!in6p->in6p_outputopts) 3047 return 0; 3048 3049 len = 0; 3050#define elen(x) \ 3051 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3052 3053 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3054 if (in6p->in6p_outputopts->ip6po_rthdr) 3055 /* dest1 is valid with rthdr only */ 3056 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3057 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3058 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3059 return len; 3060#undef elen 3061} 3062