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("$FreeBSD$"); 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 122 123 extern int in6_mcast_loop; 124 125 struct 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 }; 132 133 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, 134 struct ucred *, int); 135 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *, 136 struct socket *, struct sockopt *); 137 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *); 138 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, 139 struct ucred *, int, int, int); 140 141 static int ip6_copyexthdr(struct mbuf **, caddr_t, int); 142 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 143 struct ip6_frag **); 144 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); 145 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); 146 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *, 147 struct ifnet *, struct in6_addr *, u_long *, int *, u_int); 148 static 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) 186 187 static void 188 in6_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=0x%04x\n", __func__, m->m_len, plen, offset, 200 m->m_pkthdr.csum_flags); 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 */ 227 int 228 ip6_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 } 486 skip_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 struct flentry *fle; 526 527 /* 528 * The flow table returns route entries valid for up to 30 529 * seconds; we rely on the remainder of ip_output() taking no 530 * longer than that long for the stability of ro_rt. The 531 * flow ID assignment must have happened before this point. 532 */ 533 fle = flowtable_lookup_mbuf(V_ip6_ft, m, AF_INET6); 534 if (fle != NULL) 535 flow_to_route_in6(fle, ro); 536 } 537 #endif 538 again: 539 /* 540 * if specified, try to fill in the traffic class field. 541 * do not override if a non-zero value is already set. 542 * we check the diffserv field and the ecn field separately. 543 */ 544 if (opt && opt->ip6po_tclass >= 0) { 545 int mask = 0; 546 547 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 548 mask |= 0xfc; 549 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 550 mask |= 0x03; 551 if (mask != 0) 552 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 553 } 554 555 /* fill in or override the hop limit field, if necessary. */ 556 if (opt && opt->ip6po_hlim != -1) 557 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 558 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 559 if (im6o != NULL) 560 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 561 else 562 ip6->ip6_hlim = V_ip6_defmcasthlim; 563 } 564 565 #ifdef IPSEC 566 /* 567 * We may re-inject packets into the stack here. 568 */ 569 if (needipsec && needipsectun) { 570 struct ipsec_output_state state; 571 572 /* 573 * All the extension headers will become inaccessible 574 * (since they can be encrypted). 575 * Don't panic, we need no more updates to extension headers 576 * on inner IPv6 packet (since they are now encapsulated). 577 * 578 * IPv6 [ESP|AH] IPv6 [extension headers] payload 579 */ 580 bzero(&exthdrs, sizeof(exthdrs)); 581 exthdrs.ip6e_ip6 = m; 582 583 bzero(&state, sizeof(state)); 584 state.m = m; 585 state.ro = (struct route *)ro; 586 state.dst = (struct sockaddr *)dst; 587 588 error = ipsec6_output_tunnel(&state, sp, flags); 589 590 m = state.m; 591 ro = (struct route_in6 *)state.ro; 592 dst = (struct sockaddr_in6 *)state.dst; 593 if (error == EJUSTRETURN) { 594 /* 595 * We had a SP with a level of 'use' and no SA. We 596 * will just continue to process the packet without 597 * IPsec processing. 598 */ 599 ; 600 } else if (error) { 601 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 602 m0 = m = NULL; 603 m = NULL; 604 switch (error) { 605 case EHOSTUNREACH: 606 case ENETUNREACH: 607 case EMSGSIZE: 608 case ENOBUFS: 609 case ENOMEM: 610 break; 611 default: 612 printf("[%s:%d] (ipsec): error code %d\n", 613 __func__, __LINE__, error); 614 /* FALLTHROUGH */ 615 case ENOENT: 616 /* don't show these error codes to the user */ 617 error = 0; 618 break; 619 } 620 goto bad; 621 } else { 622 /* 623 * In the FAST IPSec case we have already 624 * re-injected the packet and it has been freed 625 * by the ipsec_done() function. So, just clean 626 * up after ourselves. 627 */ 628 m = NULL; 629 goto done; 630 } 631 632 exthdrs.ip6e_ip6 = m; 633 } 634 #endif /* IPSEC */ 635 636 /* adjust pointer */ 637 ip6 = mtod(m, struct ip6_hdr *); 638 639 if (ro->ro_rt && fwd_tag == NULL) { 640 rt = ro->ro_rt; 641 ifp = ro->ro_rt->rt_ifp; 642 } else { 643 if (fwd_tag == NULL) { 644 bzero(&dst_sa, sizeof(dst_sa)); 645 dst_sa.sin6_family = AF_INET6; 646 dst_sa.sin6_len = sizeof(dst_sa); 647 dst_sa.sin6_addr = ip6->ip6_dst; 648 } 649 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp, 650 &rt, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 651 if (error != 0) { 652 if (ifp != NULL) 653 in6_ifstat_inc(ifp, ifs6_out_discard); 654 goto bad; 655 } 656 } 657 if (rt == NULL) { 658 /* 659 * If in6_selectroute() does not return a route entry, 660 * dst may not have been updated. 661 */ 662 *dst = dst_sa; /* XXX */ 663 } 664 665 /* 666 * then rt (for unicast) and ifp must be non-NULL valid values. 667 */ 668 if ((flags & IPV6_FORWARDING) == 0) { 669 /* XXX: the FORWARDING flag can be set for mrouting. */ 670 in6_ifstat_inc(ifp, ifs6_out_request); 671 } 672 if (rt != NULL) { 673 ia = (struct in6_ifaddr *)(rt->rt_ifa); 674 rt->rt_use++; 675 } 676 677 678 /* 679 * The outgoing interface must be in the zone of source and 680 * destination addresses. 681 */ 682 origifp = ifp; 683 684 src0 = ip6->ip6_src; 685 if (in6_setscope(&src0, origifp, &zone)) 686 goto badscope; 687 bzero(&src_sa, sizeof(src_sa)); 688 src_sa.sin6_family = AF_INET6; 689 src_sa.sin6_len = sizeof(src_sa); 690 src_sa.sin6_addr = ip6->ip6_src; 691 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 692 goto badscope; 693 694 dst0 = ip6->ip6_dst; 695 if (in6_setscope(&dst0, origifp, &zone)) 696 goto badscope; 697 /* re-initialize to be sure */ 698 bzero(&dst_sa, sizeof(dst_sa)); 699 dst_sa.sin6_family = AF_INET6; 700 dst_sa.sin6_len = sizeof(dst_sa); 701 dst_sa.sin6_addr = ip6->ip6_dst; 702 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) { 703 goto badscope; 704 } 705 706 /* We should use ia_ifp to support the case of 707 * sending packets to an address of our own. 708 */ 709 if (ia != NULL && ia->ia_ifp) 710 ifp = ia->ia_ifp; 711 712 /* scope check is done. */ 713 goto routefound; 714 715 badscope: 716 IP6STAT_INC(ip6s_badscope); 717 in6_ifstat_inc(origifp, ifs6_out_discard); 718 if (error == 0) 719 error = EHOSTUNREACH; /* XXX */ 720 goto bad; 721 722 routefound: 723 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 724 if (opt && opt->ip6po_nextroute.ro_rt) { 725 /* 726 * The nexthop is explicitly specified by the 727 * application. We assume the next hop is an IPv6 728 * address. 729 */ 730 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 731 } 732 else if ((rt->rt_flags & RTF_GATEWAY)) 733 dst = (struct sockaddr_in6 *)rt->rt_gateway; 734 } 735 736 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 737 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 738 } else { 739 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 740 in6_ifstat_inc(ifp, ifs6_out_mcast); 741 /* 742 * Confirm that the outgoing interface supports multicast. 743 */ 744 if (!(ifp->if_flags & IFF_MULTICAST)) { 745 IP6STAT_INC(ip6s_noroute); 746 in6_ifstat_inc(ifp, ifs6_out_discard); 747 error = ENETUNREACH; 748 goto bad; 749 } 750 if ((im6o == NULL && in6_mcast_loop) || 751 (im6o && im6o->im6o_multicast_loop)) { 752 /* 753 * Loop back multicast datagram if not expressly 754 * forbidden to do so, even if we have not joined 755 * the address; protocols will filter it later, 756 * thus deferring a hash lookup and lock acquisition 757 * at the expense of an m_copym(). 758 */ 759 ip6_mloopback(ifp, m, dst); 760 } else { 761 /* 762 * If we are acting as a multicast router, perform 763 * multicast forwarding as if the packet had just 764 * arrived on the interface to which we are about 765 * to send. The multicast forwarding function 766 * recursively calls this function, using the 767 * IPV6_FORWARDING flag to prevent infinite recursion. 768 * 769 * Multicasts that are looped back by ip6_mloopback(), 770 * above, will be forwarded by the ip6_input() routine, 771 * if necessary. 772 */ 773 if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 774 /* 775 * XXX: ip6_mforward expects that rcvif is NULL 776 * when it is called from the originating path. 777 * However, it may not always be the case. 778 */ 779 m->m_pkthdr.rcvif = NULL; 780 if (ip6_mforward(ip6, ifp, m) != 0) { 781 m_freem(m); 782 goto done; 783 } 784 } 785 } 786 /* 787 * Multicasts with a hoplimit of zero may be looped back, 788 * above, but must not be transmitted on a network. 789 * Also, multicasts addressed to the loopback interface 790 * are not sent -- the above call to ip6_mloopback() will 791 * loop back a copy if this host actually belongs to the 792 * destination group on the loopback interface. 793 */ 794 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 795 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 796 m_freem(m); 797 goto done; 798 } 799 } 800 801 /* 802 * Fill the outgoing inteface to tell the upper layer 803 * to increment per-interface statistics. 804 */ 805 if (ifpp) 806 *ifpp = ifp; 807 808 /* Determine path MTU. */ 809 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 810 &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0) 811 goto bad; 812 813 /* 814 * The caller of this function may specify to use the minimum MTU 815 * in some cases. 816 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 817 * setting. The logic is a bit complicated; by default, unicast 818 * packets will follow path MTU while multicast packets will be sent at 819 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 820 * including unicast ones will be sent at the minimum MTU. Multicast 821 * packets will always be sent at the minimum MTU unless 822 * IP6PO_MINMTU_DISABLE is explicitly specified. 823 * See RFC 3542 for more details. 824 */ 825 if (mtu > IPV6_MMTU) { 826 if ((flags & IPV6_MINMTU)) 827 mtu = IPV6_MMTU; 828 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 829 mtu = IPV6_MMTU; 830 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 831 (opt == NULL || 832 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 833 mtu = IPV6_MMTU; 834 } 835 } 836 837 /* 838 * clear embedded scope identifiers if necessary. 839 * in6_clearscope will touch the addresses only when necessary. 840 */ 841 in6_clearscope(&ip6->ip6_src); 842 in6_clearscope(&ip6->ip6_dst); 843 844 /* 845 * If the outgoing packet contains a hop-by-hop options header, 846 * it must be examined and processed even by the source node. 847 * (RFC 2460, section 4.) 848 */ 849 if (exthdrs.ip6e_hbh) { 850 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 851 u_int32_t dummy; /* XXX unused */ 852 u_int32_t plen = 0; /* XXX: ip6_process will check the value */ 853 854 #ifdef DIAGNOSTIC 855 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len) 856 panic("ip6e_hbh is not contiguous"); 857 #endif 858 /* 859 * XXX: if we have to send an ICMPv6 error to the sender, 860 * we need the M_LOOP flag since icmp6_error() expects 861 * the IPv6 and the hop-by-hop options header are 862 * contiguous unless the flag is set. 863 */ 864 m->m_flags |= M_LOOP; 865 m->m_pkthdr.rcvif = ifp; 866 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 867 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 868 &dummy, &plen) < 0) { 869 /* m was already freed at this point */ 870 error = EINVAL;/* better error? */ 871 goto done; 872 } 873 m->m_flags &= ~M_LOOP; /* XXX */ 874 m->m_pkthdr.rcvif = NULL; 875 } 876 877 /* Jump over all PFIL processing if hooks are not active. */ 878 if (!PFIL_HOOKED(&V_inet6_pfil_hook)) 879 goto passout; 880 881 odst = ip6->ip6_dst; 882 /* Run through list of hooks for output packets. */ 883 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp); 884 if (error != 0 || m == NULL) 885 goto done; 886 ip6 = mtod(m, struct ip6_hdr *); 887 888 /* See if destination IP address was changed by packet filter. */ 889 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) { 890 m->m_flags |= M_SKIP_FIREWALL; 891 /* If destination is now ourself drop to ip6_input(). */ 892 if (in6_localip(&ip6->ip6_dst)) { 893 m->m_flags |= M_FASTFWD_OURS; 894 if (m->m_pkthdr.rcvif == NULL) 895 m->m_pkthdr.rcvif = V_loif; 896 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 897 m->m_pkthdr.csum_flags |= 898 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; 899 m->m_pkthdr.csum_data = 0xffff; 900 } 901 #ifdef SCTP 902 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) 903 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 904 #endif 905 error = netisr_queue(NETISR_IPV6, m); 906 goto done; 907 } else 908 goto again; /* Redo the routing table lookup. */ 909 } 910 911 /* See if local, if yes, send it to netisr. */ 912 if (m->m_flags & M_FASTFWD_OURS) { 913 if (m->m_pkthdr.rcvif == NULL) 914 m->m_pkthdr.rcvif = V_loif; 915 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 916 m->m_pkthdr.csum_flags |= 917 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; 918 m->m_pkthdr.csum_data = 0xffff; 919 } 920 #ifdef SCTP 921 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) 922 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 923 #endif 924 error = netisr_queue(NETISR_IPV6, m); 925 goto done; 926 } 927 /* Or forward to some other address? */ 928 if ((m->m_flags & M_IP6_NEXTHOP) && 929 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { 930 dst = (struct sockaddr_in6 *)&ro->ro_dst; 931 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6)); 932 m->m_flags |= M_SKIP_FIREWALL; 933 m->m_flags &= ~M_IP6_NEXTHOP; 934 m_tag_delete(m, fwd_tag); 935 goto again; 936 } 937 938 passout: 939 /* 940 * Send the packet to the outgoing interface. 941 * If necessary, do IPv6 fragmentation before sending. 942 * 943 * the logic here is rather complex: 944 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 945 * 1-a: send as is if tlen <= path mtu 946 * 1-b: fragment if tlen > path mtu 947 * 948 * 2: if user asks us not to fragment (dontfrag == 1) 949 * 2-a: send as is if tlen <= interface mtu 950 * 2-b: error if tlen > interface mtu 951 * 952 * 3: if we always need to attach fragment header (alwaysfrag == 1) 953 * always fragment 954 * 955 * 4: if dontfrag == 1 && alwaysfrag == 1 956 * error, as we cannot handle this conflicting request 957 */ 958 sw_csum = m->m_pkthdr.csum_flags; 959 if (!hdrsplit) { 960 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0; 961 sw_csum &= ~ifp->if_hwassist; 962 } else 963 tso = 0; 964 /* 965 * If we added extension headers, we will not do TSO and calculate the 966 * checksums ourselves for now. 967 * XXX-BZ Need a framework to know when the NIC can handle it, even 968 * with ext. hdrs. 969 */ 970 if (sw_csum & CSUM_DELAY_DATA_IPV6) { 971 sw_csum &= ~CSUM_DELAY_DATA_IPV6; 972 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr)); 973 } 974 #ifdef SCTP 975 if (sw_csum & CSUM_SCTP_IPV6) { 976 sw_csum &= ~CSUM_SCTP_IPV6; 977 sctp_delayed_cksum(m, sizeof(struct ip6_hdr)); 978 } 979 #endif 980 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 981 tlen = m->m_pkthdr.len; 982 983 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso) 984 dontfrag = 1; 985 else 986 dontfrag = 0; 987 if (dontfrag && alwaysfrag) { /* case 4 */ 988 /* conflicting request - can't transmit */ 989 error = EMSGSIZE; 990 goto bad; 991 } 992 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* case 2-b */ 993 /* 994 * Even if the DONTFRAG option is specified, we cannot send the 995 * packet when the data length is larger than the MTU of the 996 * outgoing interface. 997 * Notify the error by sending IPV6_PATHMTU ancillary data as 998 * well as returning an error code (the latter is not described 999 * in the API spec.) 1000 */ 1001 u_int32_t mtu32; 1002 struct ip6ctlparam ip6cp; 1003 1004 mtu32 = (u_int32_t)mtu; 1005 bzero(&ip6cp, sizeof(ip6cp)); 1006 ip6cp.ip6c_cmdarg = (void *)&mtu32; 1007 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, 1008 (void *)&ip6cp); 1009 1010 error = EMSGSIZE; 1011 goto bad; 1012 } 1013 1014 /* 1015 * transmit packet without fragmentation 1016 */ 1017 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */ 1018 struct in6_ifaddr *ia6; 1019 1020 ip6 = mtod(m, struct ip6_hdr *); 1021 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1022 if (ia6) { 1023 /* Record statistics for this interface address. */ 1024 ia6->ia_ifa.if_opackets++; 1025 ia6->ia_ifa.if_obytes += m->m_pkthdr.len; 1026 ifa_free(&ia6->ia_ifa); 1027 } 1028 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 1029 goto done; 1030 } 1031 1032 /* 1033 * try to fragment the packet. case 1-b and 3 1034 */ 1035 if (mtu < IPV6_MMTU) { 1036 /* path MTU cannot be less than IPV6_MMTU */ 1037 error = EMSGSIZE; 1038 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1039 goto bad; 1040 } else if (ip6->ip6_plen == 0) { 1041 /* jumbo payload cannot be fragmented */ 1042 error = EMSGSIZE; 1043 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1044 goto bad; 1045 } else { 1046 struct mbuf **mnext, *m_frgpart; 1047 struct ip6_frag *ip6f; 1048 u_int32_t id = htonl(ip6_randomid()); 1049 u_char nextproto; 1050 1051 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len; 1052 1053 /* 1054 * Too large for the destination or interface; 1055 * fragment if possible. 1056 * Must be able to put at least 8 bytes per fragment. 1057 */ 1058 hlen = unfragpartlen; 1059 if (mtu > IPV6_MAXPACKET) 1060 mtu = IPV6_MAXPACKET; 1061 1062 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 1063 if (len < 8) { 1064 error = EMSGSIZE; 1065 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1066 goto bad; 1067 } 1068 1069 /* 1070 * Verify that we have any chance at all of being able to queue 1071 * the packet or packet fragments 1072 */ 1073 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen) 1074 < tlen /* - hlen */)) { 1075 error = ENOBUFS; 1076 IP6STAT_INC(ip6s_odropped); 1077 goto bad; 1078 } 1079 1080 1081 /* 1082 * If the interface will not calculate checksums on 1083 * fragmented packets, then do it here. 1084 * XXX-BZ handle the hw offloading case. Need flags. 1085 */ 1086 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 1087 in6_delayed_cksum(m, plen, hlen); 1088 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6; 1089 } 1090 #ifdef SCTP 1091 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) { 1092 sctp_delayed_cksum(m, hlen); 1093 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6; 1094 } 1095 #endif 1096 mnext = &m->m_nextpkt; 1097 1098 /* 1099 * Change the next header field of the last header in the 1100 * unfragmentable part. 1101 */ 1102 if (exthdrs.ip6e_rthdr) { 1103 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 1104 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 1105 } else if (exthdrs.ip6e_dest1) { 1106 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 1107 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 1108 } else if (exthdrs.ip6e_hbh) { 1109 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 1110 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 1111 } else { 1112 nextproto = ip6->ip6_nxt; 1113 ip6->ip6_nxt = IPPROTO_FRAGMENT; 1114 } 1115 1116 /* 1117 * Loop through length of segment after first fragment, 1118 * make new header and copy data of each part and link onto 1119 * chain. 1120 */ 1121 m0 = m; 1122 for (off = hlen; off < tlen; off += len) { 1123 m = m_gethdr(M_NOWAIT, MT_DATA); 1124 if (!m) { 1125 error = ENOBUFS; 1126 IP6STAT_INC(ip6s_odropped); 1127 goto sendorfree; 1128 } 1129 m->m_flags = m0->m_flags & M_COPYFLAGS; /* incl. FIB */ 1130 *mnext = m; 1131 mnext = &m->m_nextpkt; 1132 m->m_data += max_linkhdr; 1133 mhip6 = mtod(m, struct ip6_hdr *); 1134 *mhip6 = *ip6; 1135 m->m_len = sizeof(*mhip6); 1136 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 1137 if (error) { 1138 IP6STAT_INC(ip6s_odropped); 1139 goto sendorfree; 1140 } 1141 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7)); 1142 if (off + len >= tlen) 1143 len = tlen - off; 1144 else 1145 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 1146 mhip6->ip6_plen = htons((u_short)(len + hlen + 1147 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 1148 if ((m_frgpart = m_copy(m0, off, len)) == 0) { 1149 error = ENOBUFS; 1150 IP6STAT_INC(ip6s_odropped); 1151 goto sendorfree; 1152 } 1153 m_cat(m, m_frgpart); 1154 m->m_pkthdr.len = len + hlen + sizeof(*ip6f); 1155 m->m_pkthdr.rcvif = NULL; 1156 ip6f->ip6f_reserved = 0; 1157 ip6f->ip6f_ident = id; 1158 ip6f->ip6f_nxt = nextproto; 1159 IP6STAT_INC(ip6s_ofragments); 1160 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 1161 } 1162 1163 in6_ifstat_inc(ifp, ifs6_out_fragok); 1164 } 1165 1166 /* 1167 * Remove leading garbages. 1168 */ 1169 sendorfree: 1170 m = m0->m_nextpkt; 1171 m0->m_nextpkt = 0; 1172 m_freem(m0); 1173 for (m0 = m; m; m = m0) { 1174 m0 = m->m_nextpkt; 1175 m->m_nextpkt = 0; 1176 if (error == 0) { 1177 /* Record statistics for this interface address. */ 1178 if (ia) { 1179 ia->ia_ifa.if_opackets++; 1180 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1181 } 1182 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt); 1183 } else 1184 m_freem(m); 1185 } 1186 1187 if (error == 0) 1188 IP6STAT_INC(ip6s_fragmented); 1189 1190 done: 1191 if (ro == &ip6route) 1192 RO_RTFREE(ro); 1193 if (ro_pmtu == &ip6route) 1194 RO_RTFREE(ro_pmtu); 1195 #ifdef IPSEC 1196 if (sp != NULL) 1197 KEY_FREESP(&sp); 1198 #endif 1199 1200 return (error); 1201 1202 freehdrs: 1203 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1204 m_freem(exthdrs.ip6e_dest1); 1205 m_freem(exthdrs.ip6e_rthdr); 1206 m_freem(exthdrs.ip6e_dest2); 1207 /* FALLTHROUGH */ 1208 bad: 1209 if (m) 1210 m_freem(m); 1211 goto done; 1212 } 1213 1214 static int 1215 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen) 1216 { 1217 struct mbuf *m; 1218 1219 if (hlen > MCLBYTES) 1220 return (ENOBUFS); /* XXX */ 1221 1222 if (hlen > MLEN) 1223 m = m_getcl(M_NOWAIT, MT_DATA, 0); 1224 else 1225 m = m_get(M_NOWAIT, MT_DATA); 1226 if (m == NULL) 1227 return (ENOBUFS); 1228 m->m_len = hlen; 1229 if (hdr) 1230 bcopy(hdr, mtod(m, caddr_t), hlen); 1231 1232 *mp = m; 1233 return (0); 1234 } 1235 1236 /* 1237 * Insert jumbo payload option. 1238 */ 1239 static int 1240 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1241 { 1242 struct mbuf *mopt; 1243 u_char *optbuf; 1244 u_int32_t v; 1245 1246 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1247 1248 /* 1249 * If there is no hop-by-hop options header, allocate new one. 1250 * If there is one but it doesn't have enough space to store the 1251 * jumbo payload option, allocate a cluster to store the whole options. 1252 * Otherwise, use it to store the options. 1253 */ 1254 if (exthdrs->ip6e_hbh == 0) { 1255 mopt = m_get(M_NOWAIT, MT_DATA); 1256 if (mopt == NULL) 1257 return (ENOBUFS); 1258 mopt->m_len = JUMBOOPTLEN; 1259 optbuf = mtod(mopt, u_char *); 1260 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1261 exthdrs->ip6e_hbh = mopt; 1262 } else { 1263 struct ip6_hbh *hbh; 1264 1265 mopt = exthdrs->ip6e_hbh; 1266 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1267 /* 1268 * XXX assumption: 1269 * - exthdrs->ip6e_hbh is not referenced from places 1270 * other than exthdrs. 1271 * - exthdrs->ip6e_hbh is not an mbuf chain. 1272 */ 1273 int oldoptlen = mopt->m_len; 1274 struct mbuf *n; 1275 1276 /* 1277 * XXX: give up if the whole (new) hbh header does 1278 * not fit even in an mbuf cluster. 1279 */ 1280 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1281 return (ENOBUFS); 1282 1283 /* 1284 * As a consequence, we must always prepare a cluster 1285 * at this point. 1286 */ 1287 n = m_getcl(M_NOWAIT, MT_DATA, 0); 1288 if (n == NULL) 1289 return (ENOBUFS); 1290 n->m_len = oldoptlen + JUMBOOPTLEN; 1291 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1292 oldoptlen); 1293 optbuf = mtod(n, caddr_t) + oldoptlen; 1294 m_freem(mopt); 1295 mopt = exthdrs->ip6e_hbh = n; 1296 } else { 1297 optbuf = mtod(mopt, u_char *) + mopt->m_len; 1298 mopt->m_len += JUMBOOPTLEN; 1299 } 1300 optbuf[0] = IP6OPT_PADN; 1301 optbuf[1] = 1; 1302 1303 /* 1304 * Adjust the header length according to the pad and 1305 * the jumbo payload option. 1306 */ 1307 hbh = mtod(mopt, struct ip6_hbh *); 1308 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1309 } 1310 1311 /* fill in the option. */ 1312 optbuf[2] = IP6OPT_JUMBO; 1313 optbuf[3] = 4; 1314 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1315 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1316 1317 /* finally, adjust the packet header length */ 1318 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1319 1320 return (0); 1321 #undef JUMBOOPTLEN 1322 } 1323 1324 /* 1325 * Insert fragment header and copy unfragmentable header portions. 1326 */ 1327 static int 1328 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1329 struct ip6_frag **frghdrp) 1330 { 1331 struct mbuf *n, *mlast; 1332 1333 if (hlen > sizeof(struct ip6_hdr)) { 1334 n = m_copym(m0, sizeof(struct ip6_hdr), 1335 hlen - sizeof(struct ip6_hdr), M_NOWAIT); 1336 if (n == 0) 1337 return (ENOBUFS); 1338 m->m_next = n; 1339 } else 1340 n = m; 1341 1342 /* Search for the last mbuf of unfragmentable part. */ 1343 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1344 ; 1345 1346 if ((mlast->m_flags & M_EXT) == 0 && 1347 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1348 /* use the trailing space of the last mbuf for the fragment hdr */ 1349 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1350 mlast->m_len); 1351 mlast->m_len += sizeof(struct ip6_frag); 1352 m->m_pkthdr.len += sizeof(struct ip6_frag); 1353 } else { 1354 /* allocate a new mbuf for the fragment header */ 1355 struct mbuf *mfrg; 1356 1357 mfrg = m_get(M_NOWAIT, MT_DATA); 1358 if (mfrg == NULL) 1359 return (ENOBUFS); 1360 mfrg->m_len = sizeof(struct ip6_frag); 1361 *frghdrp = mtod(mfrg, struct ip6_frag *); 1362 mlast->m_next = mfrg; 1363 } 1364 1365 return (0); 1366 } 1367 1368 static int 1369 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro, 1370 struct ifnet *ifp, struct in6_addr *dst, u_long *mtup, 1371 int *alwaysfragp, u_int fibnum) 1372 { 1373 u_int32_t mtu = 0; 1374 int alwaysfrag = 0; 1375 int error = 0; 1376 1377 if (ro_pmtu != ro) { 1378 /* The first hop and the final destination may differ. */ 1379 struct sockaddr_in6 *sa6_dst = 1380 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1381 if (ro_pmtu->ro_rt && 1382 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || 1383 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { 1384 RTFREE(ro_pmtu->ro_rt); 1385 ro_pmtu->ro_rt = (struct rtentry *)NULL; 1386 } 1387 if (ro_pmtu->ro_rt == NULL) { 1388 bzero(sa6_dst, sizeof(*sa6_dst)); 1389 sa6_dst->sin6_family = AF_INET6; 1390 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1391 sa6_dst->sin6_addr = *dst; 1392 1393 in6_rtalloc(ro_pmtu, fibnum); 1394 } 1395 } 1396 if (ro_pmtu->ro_rt) { 1397 u_int32_t ifmtu; 1398 struct in_conninfo inc; 1399 1400 bzero(&inc, sizeof(inc)); 1401 inc.inc_flags |= INC_ISIPV6; 1402 inc.inc6_faddr = *dst; 1403 1404 if (ifp == NULL) 1405 ifp = ro_pmtu->ro_rt->rt_ifp; 1406 ifmtu = IN6_LINKMTU(ifp); 1407 mtu = tcp_hc_getmtu(&inc); 1408 if (mtu) 1409 mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu); 1410 else 1411 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; 1412 if (mtu == 0) 1413 mtu = ifmtu; 1414 else if (mtu < IPV6_MMTU) { 1415 /* 1416 * RFC2460 section 5, last paragraph: 1417 * if we record ICMPv6 too big message with 1418 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1419 * or smaller, with framgent header attached. 1420 * (fragment header is needed regardless from the 1421 * packet size, for translators to identify packets) 1422 */ 1423 alwaysfrag = 1; 1424 mtu = IPV6_MMTU; 1425 } else if (mtu > ifmtu) { 1426 /* 1427 * The MTU on the route is larger than the MTU on 1428 * the interface! This shouldn't happen, unless the 1429 * MTU of the interface has been changed after the 1430 * interface was brought up. Change the MTU in the 1431 * route to match the interface MTU (as long as the 1432 * field isn't locked). 1433 */ 1434 mtu = ifmtu; 1435 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; 1436 } 1437 } else if (ifp) { 1438 mtu = IN6_LINKMTU(ifp); 1439 } else 1440 error = EHOSTUNREACH; /* XXX */ 1441 1442 *mtup = mtu; 1443 if (alwaysfragp) 1444 *alwaysfragp = alwaysfrag; 1445 return (error); 1446 } 1447 1448 /* 1449 * IP6 socket option processing. 1450 */ 1451 int 1452 ip6_ctloutput(struct socket *so, struct sockopt *sopt) 1453 { 1454 int optdatalen, uproto; 1455 void *optdata; 1456 struct inpcb *in6p = sotoinpcb(so); 1457 int error, optval; 1458 int level, op, optname; 1459 int optlen; 1460 struct thread *td; 1461 1462 level = sopt->sopt_level; 1463 op = sopt->sopt_dir; 1464 optname = sopt->sopt_name; 1465 optlen = sopt->sopt_valsize; 1466 td = sopt->sopt_td; 1467 error = 0; 1468 optval = 0; 1469 uproto = (int)so->so_proto->pr_protocol; 1470 1471 if (level != IPPROTO_IPV6) { 1472 error = EINVAL; 1473 1474 if (sopt->sopt_level == SOL_SOCKET && 1475 sopt->sopt_dir == SOPT_SET) { 1476 switch (sopt->sopt_name) { 1477 case SO_REUSEADDR: 1478 INP_WLOCK(in6p); 1479 if (IN_MULTICAST(ntohl(in6p->inp_laddr.s_addr))) { 1480 if ((so->so_options & 1481 (SO_REUSEADDR | SO_REUSEPORT)) != 0) 1482 in6p->inp_flags2 |= INP_REUSEPORT; 1483 else 1484 in6p->inp_flags2 &= ~INP_REUSEPORT; 1485 } 1486 INP_WUNLOCK(in6p); 1487 error = 0; 1488 break; 1489 case SO_REUSEPORT: 1490 INP_WLOCK(in6p); 1491 if ((so->so_options & SO_REUSEPORT) != 0) 1492 in6p->inp_flags2 |= INP_REUSEPORT; 1493 else 1494 in6p->inp_flags2 &= ~INP_REUSEPORT; 1495 INP_WUNLOCK(in6p); 1496 error = 0; 1497 break; 1498 case SO_SETFIB: 1499 INP_WLOCK(in6p); 1500 in6p->inp_inc.inc_fibnum = so->so_fibnum; 1501 INP_WUNLOCK(in6p); 1502 error = 0; 1503 break; 1504 default: 1505 break; 1506 } 1507 } 1508 } else { /* level == IPPROTO_IPV6 */ 1509 switch (op) { 1510 1511 case SOPT_SET: 1512 switch (optname) { 1513 case IPV6_2292PKTOPTIONS: 1514 #ifdef IPV6_PKTOPTIONS 1515 case IPV6_PKTOPTIONS: 1516 #endif 1517 { 1518 struct mbuf *m; 1519 1520 error = soopt_getm(sopt, &m); /* XXX */ 1521 if (error != 0) 1522 break; 1523 error = soopt_mcopyin(sopt, m); /* XXX */ 1524 if (error != 0) 1525 break; 1526 error = ip6_pcbopts(&in6p->in6p_outputopts, 1527 m, so, sopt); 1528 m_freem(m); /* XXX */ 1529 break; 1530 } 1531 1532 /* 1533 * Use of some Hop-by-Hop options or some 1534 * Destination options, might require special 1535 * privilege. That is, normal applications 1536 * (without special privilege) might be forbidden 1537 * from setting certain options in outgoing packets, 1538 * and might never see certain options in received 1539 * packets. [RFC 2292 Section 6] 1540 * KAME specific note: 1541 * KAME prevents non-privileged users from sending or 1542 * receiving ANY hbh/dst options in order to avoid 1543 * overhead of parsing options in the kernel. 1544 */ 1545 case IPV6_RECVHOPOPTS: 1546 case IPV6_RECVDSTOPTS: 1547 case IPV6_RECVRTHDRDSTOPTS: 1548 if (td != NULL) { 1549 error = priv_check(td, 1550 PRIV_NETINET_SETHDROPTS); 1551 if (error) 1552 break; 1553 } 1554 /* FALLTHROUGH */ 1555 case IPV6_UNICAST_HOPS: 1556 case IPV6_HOPLIMIT: 1557 case IPV6_FAITH: 1558 1559 case IPV6_RECVPKTINFO: 1560 case IPV6_RECVHOPLIMIT: 1561 case IPV6_RECVRTHDR: 1562 case IPV6_RECVPATHMTU: 1563 case IPV6_RECVTCLASS: 1564 case IPV6_V6ONLY: 1565 case IPV6_AUTOFLOWLABEL: 1566 case IPV6_BINDANY: 1567 if (optname == IPV6_BINDANY && td != NULL) { 1568 error = priv_check(td, 1569 PRIV_NETINET_BINDANY); 1570 if (error) 1571 break; 1572 } 1573 1574 if (optlen != sizeof(int)) { 1575 error = EINVAL; 1576 break; 1577 } 1578 error = sooptcopyin(sopt, &optval, 1579 sizeof optval, sizeof optval); 1580 if (error) 1581 break; 1582 switch (optname) { 1583 1584 case IPV6_UNICAST_HOPS: 1585 if (optval < -1 || optval >= 256) 1586 error = EINVAL; 1587 else { 1588 /* -1 = kernel default */ 1589 in6p->in6p_hops = optval; 1590 if ((in6p->inp_vflag & 1591 INP_IPV4) != 0) 1592 in6p->inp_ip_ttl = optval; 1593 } 1594 break; 1595 #define OPTSET(bit) \ 1596 do { \ 1597 INP_WLOCK(in6p); \ 1598 if (optval) \ 1599 in6p->inp_flags |= (bit); \ 1600 else \ 1601 in6p->inp_flags &= ~(bit); \ 1602 INP_WUNLOCK(in6p); \ 1603 } while (/*CONSTCOND*/ 0) 1604 #define OPTSET2292(bit) \ 1605 do { \ 1606 INP_WLOCK(in6p); \ 1607 in6p->inp_flags |= IN6P_RFC2292; \ 1608 if (optval) \ 1609 in6p->inp_flags |= (bit); \ 1610 else \ 1611 in6p->inp_flags &= ~(bit); \ 1612 INP_WUNLOCK(in6p); \ 1613 } while (/*CONSTCOND*/ 0) 1614 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0) 1615 1616 case IPV6_RECVPKTINFO: 1617 /* cannot mix with RFC2292 */ 1618 if (OPTBIT(IN6P_RFC2292)) { 1619 error = EINVAL; 1620 break; 1621 } 1622 OPTSET(IN6P_PKTINFO); 1623 break; 1624 1625 case IPV6_HOPLIMIT: 1626 { 1627 struct ip6_pktopts **optp; 1628 1629 /* cannot mix with RFC2292 */ 1630 if (OPTBIT(IN6P_RFC2292)) { 1631 error = EINVAL; 1632 break; 1633 } 1634 optp = &in6p->in6p_outputopts; 1635 error = ip6_pcbopt(IPV6_HOPLIMIT, 1636 (u_char *)&optval, sizeof(optval), 1637 optp, (td != NULL) ? td->td_ucred : 1638 NULL, uproto); 1639 break; 1640 } 1641 1642 case IPV6_RECVHOPLIMIT: 1643 /* cannot mix with RFC2292 */ 1644 if (OPTBIT(IN6P_RFC2292)) { 1645 error = EINVAL; 1646 break; 1647 } 1648 OPTSET(IN6P_HOPLIMIT); 1649 break; 1650 1651 case IPV6_RECVHOPOPTS: 1652 /* cannot mix with RFC2292 */ 1653 if (OPTBIT(IN6P_RFC2292)) { 1654 error = EINVAL; 1655 break; 1656 } 1657 OPTSET(IN6P_HOPOPTS); 1658 break; 1659 1660 case IPV6_RECVDSTOPTS: 1661 /* cannot mix with RFC2292 */ 1662 if (OPTBIT(IN6P_RFC2292)) { 1663 error = EINVAL; 1664 break; 1665 } 1666 OPTSET(IN6P_DSTOPTS); 1667 break; 1668 1669 case IPV6_RECVRTHDRDSTOPTS: 1670 /* cannot mix with RFC2292 */ 1671 if (OPTBIT(IN6P_RFC2292)) { 1672 error = EINVAL; 1673 break; 1674 } 1675 OPTSET(IN6P_RTHDRDSTOPTS); 1676 break; 1677 1678 case IPV6_RECVRTHDR: 1679 /* cannot mix with RFC2292 */ 1680 if (OPTBIT(IN6P_RFC2292)) { 1681 error = EINVAL; 1682 break; 1683 } 1684 OPTSET(IN6P_RTHDR); 1685 break; 1686 1687 case IPV6_FAITH: 1688 OPTSET(INP_FAITH); 1689 break; 1690 1691 case IPV6_RECVPATHMTU: 1692 /* 1693 * We ignore this option for TCP 1694 * sockets. 1695 * (RFC3542 leaves this case 1696 * unspecified.) 1697 */ 1698 if (uproto != IPPROTO_TCP) 1699 OPTSET(IN6P_MTU); 1700 break; 1701 1702 case IPV6_V6ONLY: 1703 /* 1704 * make setsockopt(IPV6_V6ONLY) 1705 * available only prior to bind(2). 1706 * see ipng mailing list, Jun 22 2001. 1707 */ 1708 if (in6p->inp_lport || 1709 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1710 error = EINVAL; 1711 break; 1712 } 1713 OPTSET(IN6P_IPV6_V6ONLY); 1714 if (optval) 1715 in6p->inp_vflag &= ~INP_IPV4; 1716 else 1717 in6p->inp_vflag |= INP_IPV4; 1718 break; 1719 case IPV6_RECVTCLASS: 1720 /* cannot mix with RFC2292 XXX */ 1721 if (OPTBIT(IN6P_RFC2292)) { 1722 error = EINVAL; 1723 break; 1724 } 1725 OPTSET(IN6P_TCLASS); 1726 break; 1727 case IPV6_AUTOFLOWLABEL: 1728 OPTSET(IN6P_AUTOFLOWLABEL); 1729 break; 1730 1731 case IPV6_BINDANY: 1732 OPTSET(INP_BINDANY); 1733 break; 1734 } 1735 break; 1736 1737 case IPV6_TCLASS: 1738 case IPV6_DONTFRAG: 1739 case IPV6_USE_MIN_MTU: 1740 case IPV6_PREFER_TEMPADDR: 1741 if (optlen != sizeof(optval)) { 1742 error = EINVAL; 1743 break; 1744 } 1745 error = sooptcopyin(sopt, &optval, 1746 sizeof optval, sizeof optval); 1747 if (error) 1748 break; 1749 { 1750 struct ip6_pktopts **optp; 1751 optp = &in6p->in6p_outputopts; 1752 error = ip6_pcbopt(optname, 1753 (u_char *)&optval, sizeof(optval), 1754 optp, (td != NULL) ? td->td_ucred : 1755 NULL, uproto); 1756 break; 1757 } 1758 1759 case IPV6_2292PKTINFO: 1760 case IPV6_2292HOPLIMIT: 1761 case IPV6_2292HOPOPTS: 1762 case IPV6_2292DSTOPTS: 1763 case IPV6_2292RTHDR: 1764 /* RFC 2292 */ 1765 if (optlen != sizeof(int)) { 1766 error = EINVAL; 1767 break; 1768 } 1769 error = sooptcopyin(sopt, &optval, 1770 sizeof optval, sizeof optval); 1771 if (error) 1772 break; 1773 switch (optname) { 1774 case IPV6_2292PKTINFO: 1775 OPTSET2292(IN6P_PKTINFO); 1776 break; 1777 case IPV6_2292HOPLIMIT: 1778 OPTSET2292(IN6P_HOPLIMIT); 1779 break; 1780 case IPV6_2292HOPOPTS: 1781 /* 1782 * Check super-user privilege. 1783 * See comments for IPV6_RECVHOPOPTS. 1784 */ 1785 if (td != NULL) { 1786 error = priv_check(td, 1787 PRIV_NETINET_SETHDROPTS); 1788 if (error) 1789 return (error); 1790 } 1791 OPTSET2292(IN6P_HOPOPTS); 1792 break; 1793 case IPV6_2292DSTOPTS: 1794 if (td != NULL) { 1795 error = priv_check(td, 1796 PRIV_NETINET_SETHDROPTS); 1797 if (error) 1798 return (error); 1799 } 1800 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1801 break; 1802 case IPV6_2292RTHDR: 1803 OPTSET2292(IN6P_RTHDR); 1804 break; 1805 } 1806 break; 1807 case IPV6_PKTINFO: 1808 case IPV6_HOPOPTS: 1809 case IPV6_RTHDR: 1810 case IPV6_DSTOPTS: 1811 case IPV6_RTHDRDSTOPTS: 1812 case IPV6_NEXTHOP: 1813 { 1814 /* new advanced API (RFC3542) */ 1815 u_char *optbuf; 1816 u_char optbuf_storage[MCLBYTES]; 1817 int optlen; 1818 struct ip6_pktopts **optp; 1819 1820 /* cannot mix with RFC2292 */ 1821 if (OPTBIT(IN6P_RFC2292)) { 1822 error = EINVAL; 1823 break; 1824 } 1825 1826 /* 1827 * We only ensure valsize is not too large 1828 * here. Further validation will be done 1829 * later. 1830 */ 1831 error = sooptcopyin(sopt, optbuf_storage, 1832 sizeof(optbuf_storage), 0); 1833 if (error) 1834 break; 1835 optlen = sopt->sopt_valsize; 1836 optbuf = optbuf_storage; 1837 optp = &in6p->in6p_outputopts; 1838 error = ip6_pcbopt(optname, optbuf, optlen, 1839 optp, (td != NULL) ? td->td_ucred : NULL, 1840 uproto); 1841 break; 1842 } 1843 #undef OPTSET 1844 1845 case IPV6_MULTICAST_IF: 1846 case IPV6_MULTICAST_HOPS: 1847 case IPV6_MULTICAST_LOOP: 1848 case IPV6_JOIN_GROUP: 1849 case IPV6_LEAVE_GROUP: 1850 case IPV6_MSFILTER: 1851 case MCAST_BLOCK_SOURCE: 1852 case MCAST_UNBLOCK_SOURCE: 1853 case MCAST_JOIN_GROUP: 1854 case MCAST_LEAVE_GROUP: 1855 case MCAST_JOIN_SOURCE_GROUP: 1856 case MCAST_LEAVE_SOURCE_GROUP: 1857 error = ip6_setmoptions(in6p, sopt); 1858 break; 1859 1860 case IPV6_PORTRANGE: 1861 error = sooptcopyin(sopt, &optval, 1862 sizeof optval, sizeof optval); 1863 if (error) 1864 break; 1865 1866 INP_WLOCK(in6p); 1867 switch (optval) { 1868 case IPV6_PORTRANGE_DEFAULT: 1869 in6p->inp_flags &= ~(INP_LOWPORT); 1870 in6p->inp_flags &= ~(INP_HIGHPORT); 1871 break; 1872 1873 case IPV6_PORTRANGE_HIGH: 1874 in6p->inp_flags &= ~(INP_LOWPORT); 1875 in6p->inp_flags |= INP_HIGHPORT; 1876 break; 1877 1878 case IPV6_PORTRANGE_LOW: 1879 in6p->inp_flags &= ~(INP_HIGHPORT); 1880 in6p->inp_flags |= INP_LOWPORT; 1881 break; 1882 1883 default: 1884 error = EINVAL; 1885 break; 1886 } 1887 INP_WUNLOCK(in6p); 1888 break; 1889 1890 #ifdef IPSEC 1891 case IPV6_IPSEC_POLICY: 1892 { 1893 caddr_t req; 1894 struct mbuf *m; 1895 1896 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1897 break; 1898 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1899 break; 1900 req = mtod(m, caddr_t); 1901 error = ipsec_set_policy(in6p, optname, req, 1902 m->m_len, (sopt->sopt_td != NULL) ? 1903 sopt->sopt_td->td_ucred : NULL); 1904 m_freem(m); 1905 break; 1906 } 1907 #endif /* IPSEC */ 1908 1909 default: 1910 error = ENOPROTOOPT; 1911 break; 1912 } 1913 break; 1914 1915 case SOPT_GET: 1916 switch (optname) { 1917 1918 case IPV6_2292PKTOPTIONS: 1919 #ifdef IPV6_PKTOPTIONS 1920 case IPV6_PKTOPTIONS: 1921 #endif 1922 /* 1923 * RFC3542 (effectively) deprecated the 1924 * semantics of the 2292-style pktoptions. 1925 * Since it was not reliable in nature (i.e., 1926 * applications had to expect the lack of some 1927 * information after all), it would make sense 1928 * to simplify this part by always returning 1929 * empty data. 1930 */ 1931 sopt->sopt_valsize = 0; 1932 break; 1933 1934 case IPV6_RECVHOPOPTS: 1935 case IPV6_RECVDSTOPTS: 1936 case IPV6_RECVRTHDRDSTOPTS: 1937 case IPV6_UNICAST_HOPS: 1938 case IPV6_RECVPKTINFO: 1939 case IPV6_RECVHOPLIMIT: 1940 case IPV6_RECVRTHDR: 1941 case IPV6_RECVPATHMTU: 1942 1943 case IPV6_FAITH: 1944 case IPV6_V6ONLY: 1945 case IPV6_PORTRANGE: 1946 case IPV6_RECVTCLASS: 1947 case IPV6_AUTOFLOWLABEL: 1948 case IPV6_BINDANY: 1949 switch (optname) { 1950 1951 case IPV6_RECVHOPOPTS: 1952 optval = OPTBIT(IN6P_HOPOPTS); 1953 break; 1954 1955 case IPV6_RECVDSTOPTS: 1956 optval = OPTBIT(IN6P_DSTOPTS); 1957 break; 1958 1959 case IPV6_RECVRTHDRDSTOPTS: 1960 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1961 break; 1962 1963 case IPV6_UNICAST_HOPS: 1964 optval = in6p->in6p_hops; 1965 break; 1966 1967 case IPV6_RECVPKTINFO: 1968 optval = OPTBIT(IN6P_PKTINFO); 1969 break; 1970 1971 case IPV6_RECVHOPLIMIT: 1972 optval = OPTBIT(IN6P_HOPLIMIT); 1973 break; 1974 1975 case IPV6_RECVRTHDR: 1976 optval = OPTBIT(IN6P_RTHDR); 1977 break; 1978 1979 case IPV6_RECVPATHMTU: 1980 optval = OPTBIT(IN6P_MTU); 1981 break; 1982 1983 case IPV6_FAITH: 1984 optval = OPTBIT(INP_FAITH); 1985 break; 1986 1987 case IPV6_V6ONLY: 1988 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1989 break; 1990 1991 case IPV6_PORTRANGE: 1992 { 1993 int flags; 1994 flags = in6p->inp_flags; 1995 if (flags & INP_HIGHPORT) 1996 optval = IPV6_PORTRANGE_HIGH; 1997 else if (flags & INP_LOWPORT) 1998 optval = IPV6_PORTRANGE_LOW; 1999 else 2000 optval = 0; 2001 break; 2002 } 2003 case IPV6_RECVTCLASS: 2004 optval = OPTBIT(IN6P_TCLASS); 2005 break; 2006 2007 case IPV6_AUTOFLOWLABEL: 2008 optval = OPTBIT(IN6P_AUTOFLOWLABEL); 2009 break; 2010 2011 case IPV6_BINDANY: 2012 optval = OPTBIT(INP_BINDANY); 2013 break; 2014 } 2015 if (error) 2016 break; 2017 error = sooptcopyout(sopt, &optval, 2018 sizeof optval); 2019 break; 2020 2021 case IPV6_PATHMTU: 2022 { 2023 u_long pmtu = 0; 2024 struct ip6_mtuinfo mtuinfo; 2025 struct route_in6 sro; 2026 2027 bzero(&sro, sizeof(sro)); 2028 2029 if (!(so->so_state & SS_ISCONNECTED)) 2030 return (ENOTCONN); 2031 /* 2032 * XXX: we dot not consider the case of source 2033 * routing, or optional information to specify 2034 * the outgoing interface. 2035 */ 2036 error = ip6_getpmtu(&sro, NULL, NULL, 2037 &in6p->in6p_faddr, &pmtu, NULL, 2038 so->so_fibnum); 2039 if (sro.ro_rt) 2040 RTFREE(sro.ro_rt); 2041 if (error) 2042 break; 2043 if (pmtu > IPV6_MAXPACKET) 2044 pmtu = IPV6_MAXPACKET; 2045 2046 bzero(&mtuinfo, sizeof(mtuinfo)); 2047 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 2048 optdata = (void *)&mtuinfo; 2049 optdatalen = sizeof(mtuinfo); 2050 error = sooptcopyout(sopt, optdata, 2051 optdatalen); 2052 break; 2053 } 2054 2055 case IPV6_2292PKTINFO: 2056 case IPV6_2292HOPLIMIT: 2057 case IPV6_2292HOPOPTS: 2058 case IPV6_2292RTHDR: 2059 case IPV6_2292DSTOPTS: 2060 switch (optname) { 2061 case IPV6_2292PKTINFO: 2062 optval = OPTBIT(IN6P_PKTINFO); 2063 break; 2064 case IPV6_2292HOPLIMIT: 2065 optval = OPTBIT(IN6P_HOPLIMIT); 2066 break; 2067 case IPV6_2292HOPOPTS: 2068 optval = OPTBIT(IN6P_HOPOPTS); 2069 break; 2070 case IPV6_2292RTHDR: 2071 optval = OPTBIT(IN6P_RTHDR); 2072 break; 2073 case IPV6_2292DSTOPTS: 2074 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 2075 break; 2076 } 2077 error = sooptcopyout(sopt, &optval, 2078 sizeof optval); 2079 break; 2080 case IPV6_PKTINFO: 2081 case IPV6_HOPOPTS: 2082 case IPV6_RTHDR: 2083 case IPV6_DSTOPTS: 2084 case IPV6_RTHDRDSTOPTS: 2085 case IPV6_NEXTHOP: 2086 case IPV6_TCLASS: 2087 case IPV6_DONTFRAG: 2088 case IPV6_USE_MIN_MTU: 2089 case IPV6_PREFER_TEMPADDR: 2090 error = ip6_getpcbopt(in6p->in6p_outputopts, 2091 optname, sopt); 2092 break; 2093 2094 case IPV6_MULTICAST_IF: 2095 case IPV6_MULTICAST_HOPS: 2096 case IPV6_MULTICAST_LOOP: 2097 case IPV6_MSFILTER: 2098 error = ip6_getmoptions(in6p, sopt); 2099 break; 2100 2101 #ifdef IPSEC 2102 case IPV6_IPSEC_POLICY: 2103 { 2104 caddr_t req = NULL; 2105 size_t len = 0; 2106 struct mbuf *m = NULL; 2107 struct mbuf **mp = &m; 2108 size_t ovalsize = sopt->sopt_valsize; 2109 caddr_t oval = (caddr_t)sopt->sopt_val; 2110 2111 error = soopt_getm(sopt, &m); /* XXX */ 2112 if (error != 0) 2113 break; 2114 error = soopt_mcopyin(sopt, m); /* XXX */ 2115 if (error != 0) 2116 break; 2117 sopt->sopt_valsize = ovalsize; 2118 sopt->sopt_val = oval; 2119 if (m) { 2120 req = mtod(m, caddr_t); 2121 len = m->m_len; 2122 } 2123 error = ipsec_get_policy(in6p, req, len, mp); 2124 if (error == 0) 2125 error = soopt_mcopyout(sopt, m); /* XXX */ 2126 if (error == 0 && m) 2127 m_freem(m); 2128 break; 2129 } 2130 #endif /* IPSEC */ 2131 2132 default: 2133 error = ENOPROTOOPT; 2134 break; 2135 } 2136 break; 2137 } 2138 } 2139 return (error); 2140 } 2141 2142 int 2143 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt) 2144 { 2145 int error = 0, optval, optlen; 2146 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2147 struct inpcb *in6p = sotoinpcb(so); 2148 int level, op, optname; 2149 2150 level = sopt->sopt_level; 2151 op = sopt->sopt_dir; 2152 optname = sopt->sopt_name; 2153 optlen = sopt->sopt_valsize; 2154 2155 if (level != IPPROTO_IPV6) { 2156 return (EINVAL); 2157 } 2158 2159 switch (optname) { 2160 case IPV6_CHECKSUM: 2161 /* 2162 * For ICMPv6 sockets, no modification allowed for checksum 2163 * offset, permit "no change" values to help existing apps. 2164 * 2165 * RFC3542 says: "An attempt to set IPV6_CHECKSUM 2166 * for an ICMPv6 socket will fail." 2167 * The current behavior does not meet RFC3542. 2168 */ 2169 switch (op) { 2170 case SOPT_SET: 2171 if (optlen != sizeof(int)) { 2172 error = EINVAL; 2173 break; 2174 } 2175 error = sooptcopyin(sopt, &optval, sizeof(optval), 2176 sizeof(optval)); 2177 if (error) 2178 break; 2179 if ((optval % 2) != 0) { 2180 /* the API assumes even offset values */ 2181 error = EINVAL; 2182 } else if (so->so_proto->pr_protocol == 2183 IPPROTO_ICMPV6) { 2184 if (optval != icmp6off) 2185 error = EINVAL; 2186 } else 2187 in6p->in6p_cksum = optval; 2188 break; 2189 2190 case SOPT_GET: 2191 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2192 optval = icmp6off; 2193 else 2194 optval = in6p->in6p_cksum; 2195 2196 error = sooptcopyout(sopt, &optval, sizeof(optval)); 2197 break; 2198 2199 default: 2200 error = EINVAL; 2201 break; 2202 } 2203 break; 2204 2205 default: 2206 error = ENOPROTOOPT; 2207 break; 2208 } 2209 2210 return (error); 2211 } 2212 2213 /* 2214 * Set up IP6 options in pcb for insertion in output packets or 2215 * specifying behavior of outgoing packets. 2216 */ 2217 static int 2218 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, 2219 struct socket *so, struct sockopt *sopt) 2220 { 2221 struct ip6_pktopts *opt = *pktopt; 2222 int error = 0; 2223 struct thread *td = sopt->sopt_td; 2224 2225 /* turn off any old options. */ 2226 if (opt) { 2227 #ifdef DIAGNOSTIC 2228 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2229 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2230 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2231 printf("ip6_pcbopts: all specified options are cleared.\n"); 2232 #endif 2233 ip6_clearpktopts(opt, -1); 2234 } else 2235 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2236 *pktopt = NULL; 2237 2238 if (!m || m->m_len == 0) { 2239 /* 2240 * Only turning off any previous options, regardless of 2241 * whether the opt is just created or given. 2242 */ 2243 free(opt, M_IP6OPT); 2244 return (0); 2245 } 2246 2247 /* set options specified by user. */ 2248 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ? 2249 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) { 2250 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2251 free(opt, M_IP6OPT); 2252 return (error); 2253 } 2254 *pktopt = opt; 2255 return (0); 2256 } 2257 2258 /* 2259 * initialize ip6_pktopts. beware that there are non-zero default values in 2260 * the struct. 2261 */ 2262 void 2263 ip6_initpktopts(struct ip6_pktopts *opt) 2264 { 2265 2266 bzero(opt, sizeof(*opt)); 2267 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2268 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2269 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2270 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; 2271 } 2272 2273 static int 2274 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2275 struct ucred *cred, int uproto) 2276 { 2277 struct ip6_pktopts *opt; 2278 2279 if (*pktopt == NULL) { 2280 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2281 M_WAITOK); 2282 ip6_initpktopts(*pktopt); 2283 } 2284 opt = *pktopt; 2285 2286 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); 2287 } 2288 2289 static int 2290 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 2291 { 2292 void *optdata = NULL; 2293 int optdatalen = 0; 2294 struct ip6_ext *ip6e; 2295 int error = 0; 2296 struct in6_pktinfo null_pktinfo; 2297 int deftclass = 0, on; 2298 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2299 int defpreftemp = IP6PO_TEMPADDR_SYSTEM; 2300 2301 switch (optname) { 2302 case IPV6_PKTINFO: 2303 if (pktopt && pktopt->ip6po_pktinfo) 2304 optdata = (void *)pktopt->ip6po_pktinfo; 2305 else { 2306 /* XXX: we don't have to do this every time... */ 2307 bzero(&null_pktinfo, sizeof(null_pktinfo)); 2308 optdata = (void *)&null_pktinfo; 2309 } 2310 optdatalen = sizeof(struct in6_pktinfo); 2311 break; 2312 case IPV6_TCLASS: 2313 if (pktopt && pktopt->ip6po_tclass >= 0) 2314 optdata = (void *)&pktopt->ip6po_tclass; 2315 else 2316 optdata = (void *)&deftclass; 2317 optdatalen = sizeof(int); 2318 break; 2319 case IPV6_HOPOPTS: 2320 if (pktopt && pktopt->ip6po_hbh) { 2321 optdata = (void *)pktopt->ip6po_hbh; 2322 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2323 optdatalen = (ip6e->ip6e_len + 1) << 3; 2324 } 2325 break; 2326 case IPV6_RTHDR: 2327 if (pktopt && pktopt->ip6po_rthdr) { 2328 optdata = (void *)pktopt->ip6po_rthdr; 2329 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2330 optdatalen = (ip6e->ip6e_len + 1) << 3; 2331 } 2332 break; 2333 case IPV6_RTHDRDSTOPTS: 2334 if (pktopt && pktopt->ip6po_dest1) { 2335 optdata = (void *)pktopt->ip6po_dest1; 2336 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2337 optdatalen = (ip6e->ip6e_len + 1) << 3; 2338 } 2339 break; 2340 case IPV6_DSTOPTS: 2341 if (pktopt && pktopt->ip6po_dest2) { 2342 optdata = (void *)pktopt->ip6po_dest2; 2343 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2344 optdatalen = (ip6e->ip6e_len + 1) << 3; 2345 } 2346 break; 2347 case IPV6_NEXTHOP: 2348 if (pktopt && pktopt->ip6po_nexthop) { 2349 optdata = (void *)pktopt->ip6po_nexthop; 2350 optdatalen = pktopt->ip6po_nexthop->sa_len; 2351 } 2352 break; 2353 case IPV6_USE_MIN_MTU: 2354 if (pktopt) 2355 optdata = (void *)&pktopt->ip6po_minmtu; 2356 else 2357 optdata = (void *)&defminmtu; 2358 optdatalen = sizeof(int); 2359 break; 2360 case IPV6_DONTFRAG: 2361 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2362 on = 1; 2363 else 2364 on = 0; 2365 optdata = (void *)&on; 2366 optdatalen = sizeof(on); 2367 break; 2368 case IPV6_PREFER_TEMPADDR: 2369 if (pktopt) 2370 optdata = (void *)&pktopt->ip6po_prefer_tempaddr; 2371 else 2372 optdata = (void *)&defpreftemp; 2373 optdatalen = sizeof(int); 2374 break; 2375 default: /* should not happen */ 2376 #ifdef DIAGNOSTIC 2377 panic("ip6_getpcbopt: unexpected option\n"); 2378 #endif 2379 return (ENOPROTOOPT); 2380 } 2381 2382 error = sooptcopyout(sopt, optdata, optdatalen); 2383 2384 return (error); 2385 } 2386 2387 void 2388 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2389 { 2390 if (pktopt == NULL) 2391 return; 2392 2393 if (optname == -1 || optname == IPV6_PKTINFO) { 2394 if (pktopt->ip6po_pktinfo) 2395 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2396 pktopt->ip6po_pktinfo = NULL; 2397 } 2398 if (optname == -1 || optname == IPV6_HOPLIMIT) 2399 pktopt->ip6po_hlim = -1; 2400 if (optname == -1 || optname == IPV6_TCLASS) 2401 pktopt->ip6po_tclass = -1; 2402 if (optname == -1 || optname == IPV6_NEXTHOP) { 2403 if (pktopt->ip6po_nextroute.ro_rt) { 2404 RTFREE(pktopt->ip6po_nextroute.ro_rt); 2405 pktopt->ip6po_nextroute.ro_rt = NULL; 2406 } 2407 if (pktopt->ip6po_nexthop) 2408 free(pktopt->ip6po_nexthop, M_IP6OPT); 2409 pktopt->ip6po_nexthop = NULL; 2410 } 2411 if (optname == -1 || optname == IPV6_HOPOPTS) { 2412 if (pktopt->ip6po_hbh) 2413 free(pktopt->ip6po_hbh, M_IP6OPT); 2414 pktopt->ip6po_hbh = NULL; 2415 } 2416 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2417 if (pktopt->ip6po_dest1) 2418 free(pktopt->ip6po_dest1, M_IP6OPT); 2419 pktopt->ip6po_dest1 = NULL; 2420 } 2421 if (optname == -1 || optname == IPV6_RTHDR) { 2422 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2423 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2424 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2425 if (pktopt->ip6po_route.ro_rt) { 2426 RTFREE(pktopt->ip6po_route.ro_rt); 2427 pktopt->ip6po_route.ro_rt = NULL; 2428 } 2429 } 2430 if (optname == -1 || optname == IPV6_DSTOPTS) { 2431 if (pktopt->ip6po_dest2) 2432 free(pktopt->ip6po_dest2, M_IP6OPT); 2433 pktopt->ip6po_dest2 = NULL; 2434 } 2435 } 2436 2437 #define PKTOPT_EXTHDRCPY(type) \ 2438 do {\ 2439 if (src->type) {\ 2440 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2441 dst->type = malloc(hlen, M_IP6OPT, canwait);\ 2442 if (dst->type == NULL && canwait == M_NOWAIT)\ 2443 goto bad;\ 2444 bcopy(src->type, dst->type, hlen);\ 2445 }\ 2446 } while (/*CONSTCOND*/ 0) 2447 2448 static int 2449 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2450 { 2451 if (dst == NULL || src == NULL) { 2452 printf("ip6_clearpktopts: invalid argument\n"); 2453 return (EINVAL); 2454 } 2455 2456 dst->ip6po_hlim = src->ip6po_hlim; 2457 dst->ip6po_tclass = src->ip6po_tclass; 2458 dst->ip6po_flags = src->ip6po_flags; 2459 dst->ip6po_minmtu = src->ip6po_minmtu; 2460 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr; 2461 if (src->ip6po_pktinfo) { 2462 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2463 M_IP6OPT, canwait); 2464 if (dst->ip6po_pktinfo == NULL) 2465 goto bad; 2466 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2467 } 2468 if (src->ip6po_nexthop) { 2469 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2470 M_IP6OPT, canwait); 2471 if (dst->ip6po_nexthop == NULL) 2472 goto bad; 2473 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop, 2474 src->ip6po_nexthop->sa_len); 2475 } 2476 PKTOPT_EXTHDRCPY(ip6po_hbh); 2477 PKTOPT_EXTHDRCPY(ip6po_dest1); 2478 PKTOPT_EXTHDRCPY(ip6po_dest2); 2479 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2480 return (0); 2481 2482 bad: 2483 ip6_clearpktopts(dst, -1); 2484 return (ENOBUFS); 2485 } 2486 #undef PKTOPT_EXTHDRCPY 2487 2488 struct ip6_pktopts * 2489 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2490 { 2491 int error; 2492 struct ip6_pktopts *dst; 2493 2494 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2495 if (dst == NULL) 2496 return (NULL); 2497 ip6_initpktopts(dst); 2498 2499 if ((error = copypktopts(dst, src, canwait)) != 0) { 2500 free(dst, M_IP6OPT); 2501 return (NULL); 2502 } 2503 2504 return (dst); 2505 } 2506 2507 void 2508 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2509 { 2510 if (pktopt == NULL) 2511 return; 2512 2513 ip6_clearpktopts(pktopt, -1); 2514 2515 free(pktopt, M_IP6OPT); 2516 } 2517 2518 /* 2519 * Set IPv6 outgoing packet options based on advanced API. 2520 */ 2521 int 2522 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2523 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto) 2524 { 2525 struct cmsghdr *cm = 0; 2526 2527 if (control == NULL || opt == NULL) 2528 return (EINVAL); 2529 2530 ip6_initpktopts(opt); 2531 if (stickyopt) { 2532 int error; 2533 2534 /* 2535 * If stickyopt is provided, make a local copy of the options 2536 * for this particular packet, then override them by ancillary 2537 * objects. 2538 * XXX: copypktopts() does not copy the cached route to a next 2539 * hop (if any). This is not very good in terms of efficiency, 2540 * but we can allow this since this option should be rarely 2541 * used. 2542 */ 2543 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2544 return (error); 2545 } 2546 2547 /* 2548 * XXX: Currently, we assume all the optional information is stored 2549 * in a single mbuf. 2550 */ 2551 if (control->m_next) 2552 return (EINVAL); 2553 2554 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2555 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2556 int error; 2557 2558 if (control->m_len < CMSG_LEN(0)) 2559 return (EINVAL); 2560 2561 cm = mtod(control, struct cmsghdr *); 2562 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2563 return (EINVAL); 2564 if (cm->cmsg_level != IPPROTO_IPV6) 2565 continue; 2566 2567 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2568 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); 2569 if (error) 2570 return (error); 2571 } 2572 2573 return (0); 2574 } 2575 2576 /* 2577 * Set a particular packet option, as a sticky option or an ancillary data 2578 * item. "len" can be 0 only when it's a sticky option. 2579 * We have 4 cases of combination of "sticky" and "cmsg": 2580 * "sticky=0, cmsg=0": impossible 2581 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2582 * "sticky=1, cmsg=0": RFC3542 socket option 2583 * "sticky=1, cmsg=1": RFC2292 socket option 2584 */ 2585 static int 2586 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2587 struct ucred *cred, int sticky, int cmsg, int uproto) 2588 { 2589 int minmtupolicy, preftemp; 2590 int error; 2591 2592 if (!sticky && !cmsg) { 2593 #ifdef DIAGNOSTIC 2594 printf("ip6_setpktopt: impossible case\n"); 2595 #endif 2596 return (EINVAL); 2597 } 2598 2599 /* 2600 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2601 * not be specified in the context of RFC3542. Conversely, 2602 * RFC3542 types should not be specified in the context of RFC2292. 2603 */ 2604 if (!cmsg) { 2605 switch (optname) { 2606 case IPV6_2292PKTINFO: 2607 case IPV6_2292HOPLIMIT: 2608 case IPV6_2292NEXTHOP: 2609 case IPV6_2292HOPOPTS: 2610 case IPV6_2292DSTOPTS: 2611 case IPV6_2292RTHDR: 2612 case IPV6_2292PKTOPTIONS: 2613 return (ENOPROTOOPT); 2614 } 2615 } 2616 if (sticky && cmsg) { 2617 switch (optname) { 2618 case IPV6_PKTINFO: 2619 case IPV6_HOPLIMIT: 2620 case IPV6_NEXTHOP: 2621 case IPV6_HOPOPTS: 2622 case IPV6_DSTOPTS: 2623 case IPV6_RTHDRDSTOPTS: 2624 case IPV6_RTHDR: 2625 case IPV6_USE_MIN_MTU: 2626 case IPV6_DONTFRAG: 2627 case IPV6_TCLASS: 2628 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */ 2629 return (ENOPROTOOPT); 2630 } 2631 } 2632 2633 switch (optname) { 2634 case IPV6_2292PKTINFO: 2635 case IPV6_PKTINFO: 2636 { 2637 struct ifnet *ifp = NULL; 2638 struct in6_pktinfo *pktinfo; 2639 2640 if (len != sizeof(struct in6_pktinfo)) 2641 return (EINVAL); 2642 2643 pktinfo = (struct in6_pktinfo *)buf; 2644 2645 /* 2646 * An application can clear any sticky IPV6_PKTINFO option by 2647 * doing a "regular" setsockopt with ipi6_addr being 2648 * in6addr_any and ipi6_ifindex being zero. 2649 * [RFC 3542, Section 6] 2650 */ 2651 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2652 pktinfo->ipi6_ifindex == 0 && 2653 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2654 ip6_clearpktopts(opt, optname); 2655 break; 2656 } 2657 2658 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2659 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2660 return (EINVAL); 2661 } 2662 2663 /* validate the interface index if specified. */ 2664 if (pktinfo->ipi6_ifindex > V_if_index || 2665 pktinfo->ipi6_ifindex < 0) { 2666 return (ENXIO); 2667 } 2668 if (pktinfo->ipi6_ifindex) { 2669 ifp = ifnet_byindex(pktinfo->ipi6_ifindex); 2670 if (ifp == NULL) 2671 return (ENXIO); 2672 } 2673 2674 /* 2675 * We store the address anyway, and let in6_selectsrc() 2676 * validate the specified address. This is because ipi6_addr 2677 * may not have enough information about its scope zone, and 2678 * we may need additional information (such as outgoing 2679 * interface or the scope zone of a destination address) to 2680 * disambiguate the scope. 2681 * XXX: the delay of the validation may confuse the 2682 * application when it is used as a sticky option. 2683 */ 2684 if (opt->ip6po_pktinfo == NULL) { 2685 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2686 M_IP6OPT, M_NOWAIT); 2687 if (opt->ip6po_pktinfo == NULL) 2688 return (ENOBUFS); 2689 } 2690 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo)); 2691 break; 2692 } 2693 2694 case IPV6_2292HOPLIMIT: 2695 case IPV6_HOPLIMIT: 2696 { 2697 int *hlimp; 2698 2699 /* 2700 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2701 * to simplify the ordering among hoplimit options. 2702 */ 2703 if (optname == IPV6_HOPLIMIT && sticky) 2704 return (ENOPROTOOPT); 2705 2706 if (len != sizeof(int)) 2707 return (EINVAL); 2708 hlimp = (int *)buf; 2709 if (*hlimp < -1 || *hlimp > 255) 2710 return (EINVAL); 2711 2712 opt->ip6po_hlim = *hlimp; 2713 break; 2714 } 2715 2716 case IPV6_TCLASS: 2717 { 2718 int tclass; 2719 2720 if (len != sizeof(int)) 2721 return (EINVAL); 2722 tclass = *(int *)buf; 2723 if (tclass < -1 || tclass > 255) 2724 return (EINVAL); 2725 2726 opt->ip6po_tclass = tclass; 2727 break; 2728 } 2729 2730 case IPV6_2292NEXTHOP: 2731 case IPV6_NEXTHOP: 2732 if (cred != NULL) { 2733 error = priv_check_cred(cred, 2734 PRIV_NETINET_SETHDROPTS, 0); 2735 if (error) 2736 return (error); 2737 } 2738 2739 if (len == 0) { /* just remove the option */ 2740 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2741 break; 2742 } 2743 2744 /* check if cmsg_len is large enough for sa_len */ 2745 if (len < sizeof(struct sockaddr) || len < *buf) 2746 return (EINVAL); 2747 2748 switch (((struct sockaddr *)buf)->sa_family) { 2749 case AF_INET6: 2750 { 2751 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 2752 int error; 2753 2754 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 2755 return (EINVAL); 2756 2757 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 2758 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 2759 return (EINVAL); 2760 } 2761 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone)) 2762 != 0) { 2763 return (error); 2764 } 2765 break; 2766 } 2767 case AF_LINK: /* should eventually be supported */ 2768 default: 2769 return (EAFNOSUPPORT); 2770 } 2771 2772 /* turn off the previous option, then set the new option. */ 2773 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2774 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 2775 if (opt->ip6po_nexthop == NULL) 2776 return (ENOBUFS); 2777 bcopy(buf, opt->ip6po_nexthop, *buf); 2778 break; 2779 2780 case IPV6_2292HOPOPTS: 2781 case IPV6_HOPOPTS: 2782 { 2783 struct ip6_hbh *hbh; 2784 int hbhlen; 2785 2786 /* 2787 * XXX: We don't allow a non-privileged user to set ANY HbH 2788 * options, since per-option restriction has too much 2789 * overhead. 2790 */ 2791 if (cred != NULL) { 2792 error = priv_check_cred(cred, 2793 PRIV_NETINET_SETHDROPTS, 0); 2794 if (error) 2795 return (error); 2796 } 2797 2798 if (len == 0) { 2799 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2800 break; /* just remove the option */ 2801 } 2802 2803 /* message length validation */ 2804 if (len < sizeof(struct ip6_hbh)) 2805 return (EINVAL); 2806 hbh = (struct ip6_hbh *)buf; 2807 hbhlen = (hbh->ip6h_len + 1) << 3; 2808 if (len != hbhlen) 2809 return (EINVAL); 2810 2811 /* turn off the previous option, then set the new option. */ 2812 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2813 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 2814 if (opt->ip6po_hbh == NULL) 2815 return (ENOBUFS); 2816 bcopy(hbh, opt->ip6po_hbh, hbhlen); 2817 2818 break; 2819 } 2820 2821 case IPV6_2292DSTOPTS: 2822 case IPV6_DSTOPTS: 2823 case IPV6_RTHDRDSTOPTS: 2824 { 2825 struct ip6_dest *dest, **newdest = NULL; 2826 int destlen; 2827 2828 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */ 2829 error = priv_check_cred(cred, 2830 PRIV_NETINET_SETHDROPTS, 0); 2831 if (error) 2832 return (error); 2833 } 2834 2835 if (len == 0) { 2836 ip6_clearpktopts(opt, optname); 2837 break; /* just remove the option */ 2838 } 2839 2840 /* message length validation */ 2841 if (len < sizeof(struct ip6_dest)) 2842 return (EINVAL); 2843 dest = (struct ip6_dest *)buf; 2844 destlen = (dest->ip6d_len + 1) << 3; 2845 if (len != destlen) 2846 return (EINVAL); 2847 2848 /* 2849 * Determine the position that the destination options header 2850 * should be inserted; before or after the routing header. 2851 */ 2852 switch (optname) { 2853 case IPV6_2292DSTOPTS: 2854 /* 2855 * The old advacned API is ambiguous on this point. 2856 * Our approach is to determine the position based 2857 * according to the existence of a routing header. 2858 * Note, however, that this depends on the order of the 2859 * extension headers in the ancillary data; the 1st 2860 * part of the destination options header must appear 2861 * before the routing header in the ancillary data, 2862 * too. 2863 * RFC3542 solved the ambiguity by introducing 2864 * separate ancillary data or option types. 2865 */ 2866 if (opt->ip6po_rthdr == NULL) 2867 newdest = &opt->ip6po_dest1; 2868 else 2869 newdest = &opt->ip6po_dest2; 2870 break; 2871 case IPV6_RTHDRDSTOPTS: 2872 newdest = &opt->ip6po_dest1; 2873 break; 2874 case IPV6_DSTOPTS: 2875 newdest = &opt->ip6po_dest2; 2876 break; 2877 } 2878 2879 /* turn off the previous option, then set the new option. */ 2880 ip6_clearpktopts(opt, optname); 2881 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 2882 if (*newdest == NULL) 2883 return (ENOBUFS); 2884 bcopy(dest, *newdest, destlen); 2885 2886 break; 2887 } 2888 2889 case IPV6_2292RTHDR: 2890 case IPV6_RTHDR: 2891 { 2892 struct ip6_rthdr *rth; 2893 int rthlen; 2894 2895 if (len == 0) { 2896 ip6_clearpktopts(opt, IPV6_RTHDR); 2897 break; /* just remove the option */ 2898 } 2899 2900 /* message length validation */ 2901 if (len < sizeof(struct ip6_rthdr)) 2902 return (EINVAL); 2903 rth = (struct ip6_rthdr *)buf; 2904 rthlen = (rth->ip6r_len + 1) << 3; 2905 if (len != rthlen) 2906 return (EINVAL); 2907 2908 switch (rth->ip6r_type) { 2909 case IPV6_RTHDR_TYPE_0: 2910 if (rth->ip6r_len == 0) /* must contain one addr */ 2911 return (EINVAL); 2912 if (rth->ip6r_len % 2) /* length must be even */ 2913 return (EINVAL); 2914 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 2915 return (EINVAL); 2916 break; 2917 default: 2918 return (EINVAL); /* not supported */ 2919 } 2920 2921 /* turn off the previous option */ 2922 ip6_clearpktopts(opt, IPV6_RTHDR); 2923 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 2924 if (opt->ip6po_rthdr == NULL) 2925 return (ENOBUFS); 2926 bcopy(rth, opt->ip6po_rthdr, rthlen); 2927 2928 break; 2929 } 2930 2931 case IPV6_USE_MIN_MTU: 2932 if (len != sizeof(int)) 2933 return (EINVAL); 2934 minmtupolicy = *(int *)buf; 2935 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 2936 minmtupolicy != IP6PO_MINMTU_DISABLE && 2937 minmtupolicy != IP6PO_MINMTU_ALL) { 2938 return (EINVAL); 2939 } 2940 opt->ip6po_minmtu = minmtupolicy; 2941 break; 2942 2943 case IPV6_DONTFRAG: 2944 if (len != sizeof(int)) 2945 return (EINVAL); 2946 2947 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 2948 /* 2949 * we ignore this option for TCP sockets. 2950 * (RFC3542 leaves this case unspecified.) 2951 */ 2952 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 2953 } else 2954 opt->ip6po_flags |= IP6PO_DONTFRAG; 2955 break; 2956 2957 case IPV6_PREFER_TEMPADDR: 2958 if (len != sizeof(int)) 2959 return (EINVAL); 2960 preftemp = *(int *)buf; 2961 if (preftemp != IP6PO_TEMPADDR_SYSTEM && 2962 preftemp != IP6PO_TEMPADDR_NOTPREFER && 2963 preftemp != IP6PO_TEMPADDR_PREFER) { 2964 return (EINVAL); 2965 } 2966 opt->ip6po_prefer_tempaddr = preftemp; 2967 break; 2968 2969 default: 2970 return (ENOPROTOOPT); 2971 } /* end of switch */ 2972 2973 return (0); 2974 } 2975 2976 /* 2977 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 2978 * packet to the input queue of a specified interface. Note that this 2979 * calls the output routine of the loopback "driver", but with an interface 2980 * pointer that might NOT be &loif -- easier than replicating that code here. 2981 */ 2982 void 2983 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst) 2984 { 2985 struct mbuf *copym; 2986 struct ip6_hdr *ip6; 2987 2988 copym = m_copy(m, 0, M_COPYALL); 2989 if (copym == NULL) 2990 return; 2991 2992 /* 2993 * Make sure to deep-copy IPv6 header portion in case the data 2994 * is in an mbuf cluster, so that we can safely override the IPv6 2995 * header portion later. 2996 */ 2997 if ((copym->m_flags & M_EXT) != 0 || 2998 copym->m_len < sizeof(struct ip6_hdr)) { 2999 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3000 if (copym == NULL) 3001 return; 3002 } 3003 3004 #ifdef DIAGNOSTIC 3005 if (copym->m_len < sizeof(*ip6)) { 3006 m_freem(copym); 3007 return; 3008 } 3009 #endif 3010 3011 ip6 = mtod(copym, struct ip6_hdr *); 3012 /* 3013 * clear embedded scope identifiers if necessary. 3014 * in6_clearscope will touch the addresses only when necessary. 3015 */ 3016 in6_clearscope(&ip6->ip6_src); 3017 in6_clearscope(&ip6->ip6_dst); 3018 3019 (void)if_simloop(ifp, copym, dst->sin6_family, 0); 3020 } 3021 3022 /* 3023 * Chop IPv6 header off from the payload. 3024 */ 3025 static int 3026 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3027 { 3028 struct mbuf *mh; 3029 struct ip6_hdr *ip6; 3030 3031 ip6 = mtod(m, struct ip6_hdr *); 3032 if (m->m_len > sizeof(*ip6)) { 3033 mh = m_gethdr(M_NOWAIT, MT_DATA); 3034 if (mh == NULL) { 3035 m_freem(m); 3036 return ENOBUFS; 3037 } 3038 m_move_pkthdr(mh, m); 3039 MH_ALIGN(mh, sizeof(*ip6)); 3040 m->m_len -= sizeof(*ip6); 3041 m->m_data += sizeof(*ip6); 3042 mh->m_next = m; 3043 m = mh; 3044 m->m_len = sizeof(*ip6); 3045 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 3046 } 3047 exthdrs->ip6e_ip6 = m; 3048 return 0; 3049 } 3050 3051 /* 3052 * Compute IPv6 extension header length. 3053 */ 3054 int 3055 ip6_optlen(struct inpcb *in6p) 3056 { 3057 int len; 3058 3059 if (!in6p->in6p_outputopts) 3060 return 0; 3061 3062 len = 0; 3063 #define elen(x) \ 3064 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3065 3066 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3067 if (in6p->in6p_outputopts->ip6po_rthdr) 3068 /* dest1 is valid with rthdr only */ 3069 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3070 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3071 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3072 return len; 3073 #undef elen 3074 } 3075