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 #include "opt_rss.h" 73 74 #include <sys/param.h> 75 #include <sys/kernel.h> 76 #include <sys/malloc.h> 77 #include <sys/mbuf.h> 78 #include <sys/errno.h> 79 #include <sys/priv.h> 80 #include <sys/proc.h> 81 #include <sys/protosw.h> 82 #include <sys/socket.h> 83 #include <sys/socketvar.h> 84 #include <sys/syslog.h> 85 #include <sys/ucred.h> 86 87 #include <machine/in_cksum.h> 88 89 #include <net/if.h> 90 #include <net/if_var.h> 91 #include <net/netisr.h> 92 #include <net/route.h> 93 #include <net/pfil.h> 94 #include <net/rss_config.h> 95 #include <net/vnet.h> 96 97 #include <netinet/in.h> 98 #include <netinet/in_var.h> 99 #include <netinet/ip_var.h> 100 #include <netinet6/in6_var.h> 101 #include <netinet/ip6.h> 102 #include <netinet/icmp6.h> 103 #include <netinet6/ip6_var.h> 104 #include <netinet/in_pcb.h> 105 #include <netinet/tcp_var.h> 106 #include <netinet6/nd6.h> 107 #include <netinet6/in6_rss.h> 108 109 #ifdef IPSEC 110 #include <netipsec/ipsec.h> 111 #include <netipsec/ipsec6.h> 112 #include <netipsec/key.h> 113 #include <netinet6/ip6_ipsec.h> 114 #endif /* IPSEC */ 115 #ifdef SCTP 116 #include <netinet/sctp.h> 117 #include <netinet/sctp_crc32.h> 118 #endif 119 120 #include <netinet6/ip6protosw.h> 121 #include <netinet6/scope6_var.h> 122 123 #ifdef FLOWTABLE 124 #include <net/flowtable.h> 125 #endif 126 127 extern int in6_mcast_loop; 128 129 struct ip6_exthdrs { 130 struct mbuf *ip6e_ip6; 131 struct mbuf *ip6e_hbh; 132 struct mbuf *ip6e_dest1; 133 struct mbuf *ip6e_rthdr; 134 struct mbuf *ip6e_dest2; 135 }; 136 137 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, 138 struct ucred *, int); 139 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *, 140 struct socket *, struct sockopt *); 141 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *); 142 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, 143 struct ucred *, int, int, int); 144 145 static int ip6_copyexthdr(struct mbuf **, caddr_t, int); 146 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 147 struct ip6_frag **); 148 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); 149 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); 150 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *, 151 struct ifnet *, struct in6_addr *, u_long *, int *, u_int); 152 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int); 153 154 155 /* 156 * Make an extension header from option data. hp is the source, and 157 * mp is the destination. 158 */ 159 #define MAKE_EXTHDR(hp, mp) \ 160 do { \ 161 if (hp) { \ 162 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 163 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 164 ((eh)->ip6e_len + 1) << 3); \ 165 if (error) \ 166 goto freehdrs; \ 167 } \ 168 } while (/*CONSTCOND*/ 0) 169 170 /* 171 * Form a chain of extension headers. 172 * m is the extension header mbuf 173 * mp is the previous mbuf in the chain 174 * p is the next header 175 * i is the type of option. 176 */ 177 #define MAKE_CHAIN(m, mp, p, i)\ 178 do {\ 179 if (m) {\ 180 if (!hdrsplit) \ 181 panic("assumption failed: hdr not split"); \ 182 *mtod((m), u_char *) = *(p);\ 183 *(p) = (i);\ 184 p = mtod((m), u_char *);\ 185 (m)->m_next = (mp)->m_next;\ 186 (mp)->m_next = (m);\ 187 (mp) = (m);\ 188 }\ 189 } while (/*CONSTCOND*/ 0) 190 191 void 192 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset) 193 { 194 u_short csum; 195 196 csum = in_cksum_skip(m, offset + plen, offset); 197 if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0) 198 csum = 0xffff; 199 offset += m->m_pkthdr.csum_data; /* checksum offset */ 200 201 if (offset + sizeof(u_short) > m->m_len) { 202 printf("%s: delayed m_pullup, m->len: %d plen %u off %u " 203 "csum_flags=%b\n", __func__, m->m_len, plen, offset, 204 (int)m->m_pkthdr.csum_flags, CSUM_BITS); 205 /* 206 * XXX this should not happen, but if it does, the correct 207 * behavior may be to insert the checksum in the appropriate 208 * next mbuf in the chain. 209 */ 210 return; 211 } 212 *(u_short *)(m->m_data + offset) = csum; 213 } 214 215 int 216 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto, 217 int mtu, uint32_t id) 218 { 219 struct mbuf *m, **mnext, *m_frgpart; 220 struct ip6_hdr *ip6, *mhip6; 221 struct ip6_frag *ip6f; 222 int off; 223 int error; 224 int tlen = m0->m_pkthdr.len; 225 226 m = m0; 227 ip6 = mtod(m, struct ip6_hdr *); 228 mnext = &m->m_nextpkt; 229 230 for (off = hlen; off < tlen; off += mtu) { 231 m = m_gethdr(M_NOWAIT, MT_DATA); 232 if (!m) { 233 IP6STAT_INC(ip6s_odropped); 234 return (ENOBUFS); 235 } 236 m->m_flags = m0->m_flags & M_COPYFLAGS; 237 *mnext = m; 238 mnext = &m->m_nextpkt; 239 m->m_data += max_linkhdr; 240 mhip6 = mtod(m, struct ip6_hdr *); 241 *mhip6 = *ip6; 242 m->m_len = sizeof(*mhip6); 243 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 244 if (error) { 245 IP6STAT_INC(ip6s_odropped); 246 return (error); 247 } 248 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7)); 249 if (off + mtu >= tlen) 250 mtu = tlen - off; 251 else 252 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 253 mhip6->ip6_plen = htons((u_short)(mtu + hlen + 254 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 255 if ((m_frgpart = m_copy(m0, off, mtu)) == 0) { 256 IP6STAT_INC(ip6s_odropped); 257 return (ENOBUFS); 258 } 259 m_cat(m, m_frgpart); 260 m->m_pkthdr.len = mtu + hlen + sizeof(*ip6f); 261 m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum; 262 m->m_pkthdr.rcvif = NULL; 263 ip6f->ip6f_reserved = 0; 264 ip6f->ip6f_ident = id; 265 ip6f->ip6f_nxt = nextproto; 266 IP6STAT_INC(ip6s_ofragments); 267 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 268 } 269 270 return (0); 271 } 272 273 /* 274 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 275 * header (with pri, len, nxt, hlim, src, dst). 276 * This function may modify ver and hlim only. 277 * The mbuf chain containing the packet will be freed. 278 * The mbuf opt, if present, will not be freed. 279 * If route_in6 ro is present and has ro_rt initialized, route lookup would be 280 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL, 281 * then result of route lookup is stored in ro->ro_rt. 282 * 283 * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and 284 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 285 * which is rt_mtu. 286 * 287 * ifpp - XXX: just for statistics 288 */ 289 /* 290 * XXX TODO: no flowid is assigned for outbound flows? 291 */ 292 int 293 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, 294 struct route_in6 *ro, int flags, struct ip6_moptions *im6o, 295 struct ifnet **ifpp, struct inpcb *inp) 296 { 297 struct ip6_hdr *ip6; 298 struct ifnet *ifp, *origifp; 299 struct mbuf *m = m0; 300 struct mbuf *mprev = NULL; 301 int hlen, tlen, len; 302 struct route_in6 ip6route; 303 struct rtentry *rt = NULL; 304 struct sockaddr_in6 *dst, src_sa, dst_sa; 305 struct in6_addr odst; 306 int error = 0; 307 struct in6_ifaddr *ia = NULL; 308 u_long mtu; 309 int alwaysfrag, dontfrag; 310 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 311 struct ip6_exthdrs exthdrs; 312 struct in6_addr finaldst, src0, dst0; 313 u_int32_t zone; 314 struct route_in6 *ro_pmtu = NULL; 315 int hdrsplit = 0; 316 int sw_csum, tso; 317 int needfiblookup; 318 uint32_t fibnum; 319 struct m_tag *fwd_tag = NULL; 320 uint32_t id; 321 322 ip6 = mtod(m, struct ip6_hdr *); 323 if (ip6 == NULL) { 324 printf ("ip6 is NULL"); 325 goto bad; 326 } 327 328 if (inp != NULL) { 329 M_SETFIB(m, inp->inp_inc.inc_fibnum); 330 if ((flags & IP_NODEFAULTFLOWID) == 0) { 331 /* unconditionally set flowid */ 332 m->m_pkthdr.flowid = inp->inp_flowid; 333 M_HASHTYPE_SET(m, inp->inp_flowtype); 334 } 335 } 336 337 finaldst = ip6->ip6_dst; 338 bzero(&exthdrs, sizeof(exthdrs)); 339 if (opt) { 340 /* Hop-by-Hop options header */ 341 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 342 /* Destination options header(1st part) */ 343 if (opt->ip6po_rthdr) { 344 /* 345 * Destination options header(1st part) 346 * This only makes sense with a routing header. 347 * See Section 9.2 of RFC 3542. 348 * Disabling this part just for MIP6 convenience is 349 * a bad idea. We need to think carefully about a 350 * way to make the advanced API coexist with MIP6 351 * options, which might automatically be inserted in 352 * the kernel. 353 */ 354 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 355 } 356 /* Routing header */ 357 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 358 /* Destination options header(2nd part) */ 359 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 360 } 361 362 #ifdef IPSEC 363 /* 364 * IPSec checking which handles several cases. 365 * FAST IPSEC: We re-injected the packet. 366 * XXX: need scope argument. 367 */ 368 switch(ip6_ipsec_output(&m, inp, &error)) 369 { 370 case 1: /* Bad packet */ 371 goto freehdrs; 372 case -1: /* IPSec done */ 373 goto done; 374 case 0: /* No IPSec */ 375 default: 376 break; 377 } 378 #endif /* IPSEC */ 379 380 /* 381 * Calculate the total length of the extension header chain. 382 * Keep the length of the unfragmentable part for fragmentation. 383 */ 384 optlen = 0; 385 if (exthdrs.ip6e_hbh) 386 optlen += exthdrs.ip6e_hbh->m_len; 387 if (exthdrs.ip6e_dest1) 388 optlen += exthdrs.ip6e_dest1->m_len; 389 if (exthdrs.ip6e_rthdr) 390 optlen += exthdrs.ip6e_rthdr->m_len; 391 unfragpartlen = optlen + sizeof(struct ip6_hdr); 392 393 /* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */ 394 if (exthdrs.ip6e_dest2) 395 optlen += exthdrs.ip6e_dest2->m_len; 396 397 /* 398 * If there is at least one extension header, 399 * separate IP6 header from the payload. 400 */ 401 if (optlen && !hdrsplit) { 402 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 403 m = NULL; 404 goto freehdrs; 405 } 406 m = exthdrs.ip6e_ip6; 407 hdrsplit++; 408 } 409 410 /* adjust pointer */ 411 ip6 = mtod(m, struct ip6_hdr *); 412 413 /* adjust mbuf packet header length */ 414 m->m_pkthdr.len += optlen; 415 plen = m->m_pkthdr.len - sizeof(*ip6); 416 417 /* If this is a jumbo payload, insert a jumbo payload option. */ 418 if (plen > IPV6_MAXPACKET) { 419 if (!hdrsplit) { 420 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 421 m = NULL; 422 goto freehdrs; 423 } 424 m = exthdrs.ip6e_ip6; 425 hdrsplit++; 426 } 427 /* adjust pointer */ 428 ip6 = mtod(m, struct ip6_hdr *); 429 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 430 goto freehdrs; 431 ip6->ip6_plen = 0; 432 } else 433 ip6->ip6_plen = htons(plen); 434 435 /* 436 * Concatenate headers and fill in next header fields. 437 * Here we have, on "m" 438 * IPv6 payload 439 * and we insert headers accordingly. Finally, we should be getting: 440 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 441 * 442 * during the header composing process, "m" points to IPv6 header. 443 * "mprev" points to an extension header prior to esp. 444 */ 445 u_char *nexthdrp = &ip6->ip6_nxt; 446 mprev = m; 447 448 /* 449 * we treat dest2 specially. this makes IPsec processing 450 * much easier. the goal here is to make mprev point the 451 * mbuf prior to dest2. 452 * 453 * result: IPv6 dest2 payload 454 * m and mprev will point to IPv6 header. 455 */ 456 if (exthdrs.ip6e_dest2) { 457 if (!hdrsplit) 458 panic("assumption failed: hdr not split"); 459 exthdrs.ip6e_dest2->m_next = m->m_next; 460 m->m_next = exthdrs.ip6e_dest2; 461 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 462 ip6->ip6_nxt = IPPROTO_DSTOPTS; 463 } 464 465 /* 466 * result: IPv6 hbh dest1 rthdr dest2 payload 467 * m will point to IPv6 header. mprev will point to the 468 * extension header prior to dest2 (rthdr in the above case). 469 */ 470 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 471 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 472 IPPROTO_DSTOPTS); 473 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 474 IPPROTO_ROUTING); 475 476 /* 477 * If there is a routing header, discard the packet. 478 */ 479 if (exthdrs.ip6e_rthdr) { 480 error = EINVAL; 481 goto bad; 482 } 483 484 /* Source address validation */ 485 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 486 (flags & IPV6_UNSPECSRC) == 0) { 487 error = EOPNOTSUPP; 488 IP6STAT_INC(ip6s_badscope); 489 goto bad; 490 } 491 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 492 error = EOPNOTSUPP; 493 IP6STAT_INC(ip6s_badscope); 494 goto bad; 495 } 496 497 IP6STAT_INC(ip6s_localout); 498 499 /* 500 * Route packet. 501 */ 502 if (ro == 0) { 503 ro = &ip6route; 504 bzero((caddr_t)ro, sizeof(*ro)); 505 } 506 ro_pmtu = ro; 507 if (opt && opt->ip6po_rthdr) 508 ro = &opt->ip6po_route; 509 dst = (struct sockaddr_in6 *)&ro->ro_dst; 510 #ifdef FLOWTABLE 511 if (ro->ro_rt == NULL) 512 (void )flowtable_lookup(AF_INET6, m, (struct route *)ro); 513 #endif 514 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m); 515 again: 516 /* 517 * if specified, try to fill in the traffic class field. 518 * do not override if a non-zero value is already set. 519 * we check the diffserv field and the ecn field separately. 520 */ 521 if (opt && opt->ip6po_tclass >= 0) { 522 int mask = 0; 523 524 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 525 mask |= 0xfc; 526 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 527 mask |= 0x03; 528 if (mask != 0) 529 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 530 } 531 532 /* fill in or override the hop limit field, if necessary. */ 533 if (opt && opt->ip6po_hlim != -1) 534 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 535 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 536 if (im6o != NULL) 537 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 538 else 539 ip6->ip6_hlim = V_ip6_defmcasthlim; 540 } 541 542 /* adjust pointer */ 543 ip6 = mtod(m, struct ip6_hdr *); 544 545 if (ro->ro_rt && fwd_tag == NULL) { 546 rt = ro->ro_rt; 547 ifp = ro->ro_rt->rt_ifp; 548 } else { 549 if (fwd_tag == NULL) { 550 bzero(&dst_sa, sizeof(dst_sa)); 551 dst_sa.sin6_family = AF_INET6; 552 dst_sa.sin6_len = sizeof(dst_sa); 553 dst_sa.sin6_addr = ip6->ip6_dst; 554 } 555 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp, 556 &rt, fibnum); 557 if (error != 0) { 558 if (ifp != NULL) 559 in6_ifstat_inc(ifp, ifs6_out_discard); 560 goto bad; 561 } 562 } 563 if (rt == NULL) { 564 /* 565 * If in6_selectroute() does not return a route entry, 566 * dst may not have been updated. 567 */ 568 *dst = dst_sa; /* XXX */ 569 } 570 571 /* 572 * then rt (for unicast) and ifp must be non-NULL valid values. 573 */ 574 if ((flags & IPV6_FORWARDING) == 0) { 575 /* XXX: the FORWARDING flag can be set for mrouting. */ 576 in6_ifstat_inc(ifp, ifs6_out_request); 577 } 578 if (rt != NULL) { 579 ia = (struct in6_ifaddr *)(rt->rt_ifa); 580 counter_u64_add(rt->rt_pksent, 1); 581 } 582 583 584 /* 585 * The outgoing interface must be in the zone of source and 586 * destination addresses. 587 */ 588 origifp = ifp; 589 590 src0 = ip6->ip6_src; 591 if (in6_setscope(&src0, origifp, &zone)) 592 goto badscope; 593 bzero(&src_sa, sizeof(src_sa)); 594 src_sa.sin6_family = AF_INET6; 595 src_sa.sin6_len = sizeof(src_sa); 596 src_sa.sin6_addr = ip6->ip6_src; 597 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 598 goto badscope; 599 600 dst0 = ip6->ip6_dst; 601 if (in6_setscope(&dst0, origifp, &zone)) 602 goto badscope; 603 /* re-initialize to be sure */ 604 bzero(&dst_sa, sizeof(dst_sa)); 605 dst_sa.sin6_family = AF_INET6; 606 dst_sa.sin6_len = sizeof(dst_sa); 607 dst_sa.sin6_addr = ip6->ip6_dst; 608 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) { 609 goto badscope; 610 } 611 612 /* We should use ia_ifp to support the case of 613 * sending packets to an address of our own. 614 */ 615 if (ia != NULL && ia->ia_ifp) 616 ifp = ia->ia_ifp; 617 618 /* scope check is done. */ 619 goto routefound; 620 621 badscope: 622 IP6STAT_INC(ip6s_badscope); 623 in6_ifstat_inc(origifp, ifs6_out_discard); 624 if (error == 0) 625 error = EHOSTUNREACH; /* XXX */ 626 goto bad; 627 628 routefound: 629 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 630 if (opt && opt->ip6po_nextroute.ro_rt) { 631 /* 632 * The nexthop is explicitly specified by the 633 * application. We assume the next hop is an IPv6 634 * address. 635 */ 636 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 637 } 638 else if ((rt->rt_flags & RTF_GATEWAY)) 639 dst = (struct sockaddr_in6 *)rt->rt_gateway; 640 } 641 642 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 643 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 644 } else { 645 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 646 in6_ifstat_inc(ifp, ifs6_out_mcast); 647 /* 648 * Confirm that the outgoing interface supports multicast. 649 */ 650 if (!(ifp->if_flags & IFF_MULTICAST)) { 651 IP6STAT_INC(ip6s_noroute); 652 in6_ifstat_inc(ifp, ifs6_out_discard); 653 error = ENETUNREACH; 654 goto bad; 655 } 656 if ((im6o == NULL && in6_mcast_loop) || 657 (im6o && im6o->im6o_multicast_loop)) { 658 /* 659 * Loop back multicast datagram if not expressly 660 * forbidden to do so, even if we have not joined 661 * the address; protocols will filter it later, 662 * thus deferring a hash lookup and lock acquisition 663 * at the expense of an m_copym(). 664 */ 665 ip6_mloopback(ifp, m); 666 } else { 667 /* 668 * If we are acting as a multicast router, perform 669 * multicast forwarding as if the packet had just 670 * arrived on the interface to which we are about 671 * to send. The multicast forwarding function 672 * recursively calls this function, using the 673 * IPV6_FORWARDING flag to prevent infinite recursion. 674 * 675 * Multicasts that are looped back by ip6_mloopback(), 676 * above, will be forwarded by the ip6_input() routine, 677 * if necessary. 678 */ 679 if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 680 /* 681 * XXX: ip6_mforward expects that rcvif is NULL 682 * when it is called from the originating path. 683 * However, it may not always be the case. 684 */ 685 m->m_pkthdr.rcvif = NULL; 686 if (ip6_mforward(ip6, ifp, m) != 0) { 687 m_freem(m); 688 goto done; 689 } 690 } 691 } 692 /* 693 * Multicasts with a hoplimit of zero may be looped back, 694 * above, but must not be transmitted on a network. 695 * Also, multicasts addressed to the loopback interface 696 * are not sent -- the above call to ip6_mloopback() will 697 * loop back a copy if this host actually belongs to the 698 * destination group on the loopback interface. 699 */ 700 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 701 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 702 m_freem(m); 703 goto done; 704 } 705 } 706 707 /* 708 * Fill the outgoing inteface to tell the upper layer 709 * to increment per-interface statistics. 710 */ 711 if (ifpp) 712 *ifpp = ifp; 713 714 /* Determine path MTU. */ 715 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, 716 &alwaysfrag, fibnum)) != 0) 717 goto bad; 718 719 /* 720 * The caller of this function may specify to use the minimum MTU 721 * in some cases. 722 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 723 * setting. The logic is a bit complicated; by default, unicast 724 * packets will follow path MTU while multicast packets will be sent at 725 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 726 * including unicast ones will be sent at the minimum MTU. Multicast 727 * packets will always be sent at the minimum MTU unless 728 * IP6PO_MINMTU_DISABLE is explicitly specified. 729 * See RFC 3542 for more details. 730 */ 731 if (mtu > IPV6_MMTU) { 732 if ((flags & IPV6_MINMTU)) 733 mtu = IPV6_MMTU; 734 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 735 mtu = IPV6_MMTU; 736 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 737 (opt == NULL || 738 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 739 mtu = IPV6_MMTU; 740 } 741 } 742 743 /* 744 * clear embedded scope identifiers if necessary. 745 * in6_clearscope will touch the addresses only when necessary. 746 */ 747 in6_clearscope(&ip6->ip6_src); 748 in6_clearscope(&ip6->ip6_dst); 749 750 /* 751 * If the outgoing packet contains a hop-by-hop options header, 752 * it must be examined and processed even by the source node. 753 * (RFC 2460, section 4.) 754 */ 755 if (exthdrs.ip6e_hbh) { 756 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 757 u_int32_t dummy; /* XXX unused */ 758 u_int32_t plen = 0; /* XXX: ip6_process will check the value */ 759 760 #ifdef DIAGNOSTIC 761 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len) 762 panic("ip6e_hbh is not contiguous"); 763 #endif 764 /* 765 * XXX: if we have to send an ICMPv6 error to the sender, 766 * we need the M_LOOP flag since icmp6_error() expects 767 * the IPv6 and the hop-by-hop options header are 768 * contiguous unless the flag is set. 769 */ 770 m->m_flags |= M_LOOP; 771 m->m_pkthdr.rcvif = ifp; 772 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 773 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 774 &dummy, &plen) < 0) { 775 /* m was already freed at this point */ 776 error = EINVAL;/* better error? */ 777 goto done; 778 } 779 m->m_flags &= ~M_LOOP; /* XXX */ 780 m->m_pkthdr.rcvif = NULL; 781 } 782 783 /* Jump over all PFIL processing if hooks are not active. */ 784 if (!PFIL_HOOKED(&V_inet6_pfil_hook)) 785 goto passout; 786 787 odst = ip6->ip6_dst; 788 /* Run through list of hooks for output packets. */ 789 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp); 790 if (error != 0 || m == NULL) 791 goto done; 792 ip6 = mtod(m, struct ip6_hdr *); 793 794 needfiblookup = 0; 795 /* See if destination IP address was changed by packet filter. */ 796 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) { 797 m->m_flags |= M_SKIP_FIREWALL; 798 /* If destination is now ourself drop to ip6_input(). */ 799 if (in6_localip(&ip6->ip6_dst)) { 800 m->m_flags |= M_FASTFWD_OURS; 801 if (m->m_pkthdr.rcvif == NULL) 802 m->m_pkthdr.rcvif = V_loif; 803 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 804 m->m_pkthdr.csum_flags |= 805 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; 806 m->m_pkthdr.csum_data = 0xffff; 807 } 808 #ifdef SCTP 809 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) 810 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 811 #endif 812 error = netisr_queue(NETISR_IPV6, m); 813 goto done; 814 } else 815 needfiblookup = 1; /* Redo the routing table lookup. */ 816 } 817 /* See if fib was changed by packet filter. */ 818 if (fibnum != M_GETFIB(m)) { 819 m->m_flags |= M_SKIP_FIREWALL; 820 fibnum = M_GETFIB(m); 821 RO_RTFREE(ro); 822 needfiblookup = 1; 823 } 824 if (needfiblookup) 825 goto again; 826 827 /* See if local, if yes, send it to netisr. */ 828 if (m->m_flags & M_FASTFWD_OURS) { 829 if (m->m_pkthdr.rcvif == NULL) 830 m->m_pkthdr.rcvif = V_loif; 831 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 832 m->m_pkthdr.csum_flags |= 833 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; 834 m->m_pkthdr.csum_data = 0xffff; 835 } 836 #ifdef SCTP 837 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) 838 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 839 #endif 840 error = netisr_queue(NETISR_IPV6, m); 841 goto done; 842 } 843 /* Or forward to some other address? */ 844 if ((m->m_flags & M_IP6_NEXTHOP) && 845 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { 846 dst = (struct sockaddr_in6 *)&ro->ro_dst; 847 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6)); 848 m->m_flags |= M_SKIP_FIREWALL; 849 m->m_flags &= ~M_IP6_NEXTHOP; 850 m_tag_delete(m, fwd_tag); 851 goto again; 852 } 853 854 passout: 855 /* 856 * Send the packet to the outgoing interface. 857 * If necessary, do IPv6 fragmentation before sending. 858 * 859 * the logic here is rather complex: 860 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 861 * 1-a: send as is if tlen <= path mtu 862 * 1-b: fragment if tlen > path mtu 863 * 864 * 2: if user asks us not to fragment (dontfrag == 1) 865 * 2-a: send as is if tlen <= interface mtu 866 * 2-b: error if tlen > interface mtu 867 * 868 * 3: if we always need to attach fragment header (alwaysfrag == 1) 869 * always fragment 870 * 871 * 4: if dontfrag == 1 && alwaysfrag == 1 872 * error, as we cannot handle this conflicting request 873 */ 874 sw_csum = m->m_pkthdr.csum_flags; 875 if (!hdrsplit) { 876 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0; 877 sw_csum &= ~ifp->if_hwassist; 878 } else 879 tso = 0; 880 /* 881 * If we added extension headers, we will not do TSO and calculate the 882 * checksums ourselves for now. 883 * XXX-BZ Need a framework to know when the NIC can handle it, even 884 * with ext. hdrs. 885 */ 886 if (sw_csum & CSUM_DELAY_DATA_IPV6) { 887 sw_csum &= ~CSUM_DELAY_DATA_IPV6; 888 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr)); 889 } 890 #ifdef SCTP 891 if (sw_csum & CSUM_SCTP_IPV6) { 892 sw_csum &= ~CSUM_SCTP_IPV6; 893 sctp_delayed_cksum(m, sizeof(struct ip6_hdr)); 894 } 895 #endif 896 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 897 tlen = m->m_pkthdr.len; 898 899 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso) 900 dontfrag = 1; 901 else 902 dontfrag = 0; 903 if (dontfrag && alwaysfrag) { /* case 4 */ 904 /* conflicting request - can't transmit */ 905 error = EMSGSIZE; 906 goto bad; 907 } 908 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* case 2-b */ 909 /* 910 * Even if the DONTFRAG option is specified, we cannot send the 911 * packet when the data length is larger than the MTU of the 912 * outgoing interface. 913 * Notify the error by sending IPV6_PATHMTU ancillary data if 914 * application wanted to know the MTU value. Also return an 915 * error code (this is not described in the API spec). 916 */ 917 if (inp != NULL) 918 ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu); 919 error = EMSGSIZE; 920 goto bad; 921 } 922 923 /* 924 * transmit packet without fragmentation 925 */ 926 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */ 927 struct in6_ifaddr *ia6; 928 929 ip6 = mtod(m, struct ip6_hdr *); 930 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 931 if (ia6) { 932 /* Record statistics for this interface address. */ 933 counter_u64_add(ia6->ia_ifa.ifa_opackets, 1); 934 counter_u64_add(ia6->ia_ifa.ifa_obytes, 935 m->m_pkthdr.len); 936 ifa_free(&ia6->ia_ifa); 937 } 938 error = nd6_output_ifp(ifp, origifp, m, dst); 939 goto done; 940 } 941 942 /* 943 * try to fragment the packet. case 1-b and 3 944 */ 945 if (mtu < IPV6_MMTU) { 946 /* path MTU cannot be less than IPV6_MMTU */ 947 error = EMSGSIZE; 948 in6_ifstat_inc(ifp, ifs6_out_fragfail); 949 goto bad; 950 } else if (ip6->ip6_plen == 0) { 951 /* jumbo payload cannot be fragmented */ 952 error = EMSGSIZE; 953 in6_ifstat_inc(ifp, ifs6_out_fragfail); 954 goto bad; 955 } else { 956 u_char nextproto; 957 958 /* 959 * Too large for the destination or interface; 960 * fragment if possible. 961 * Must be able to put at least 8 bytes per fragment. 962 */ 963 hlen = unfragpartlen; 964 if (mtu > IPV6_MAXPACKET) 965 mtu = IPV6_MAXPACKET; 966 967 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 968 if (len < 8) { 969 error = EMSGSIZE; 970 in6_ifstat_inc(ifp, ifs6_out_fragfail); 971 goto bad; 972 } 973 974 /* 975 * If the interface will not calculate checksums on 976 * fragmented packets, then do it here. 977 * XXX-BZ handle the hw offloading case. Need flags. 978 */ 979 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 980 in6_delayed_cksum(m, plen, hlen); 981 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6; 982 } 983 #ifdef SCTP 984 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) { 985 sctp_delayed_cksum(m, hlen); 986 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6; 987 } 988 #endif 989 /* 990 * Change the next header field of the last header in the 991 * unfragmentable part. 992 */ 993 if (exthdrs.ip6e_rthdr) { 994 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 995 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 996 } else if (exthdrs.ip6e_dest1) { 997 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 998 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 999 } else if (exthdrs.ip6e_hbh) { 1000 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 1001 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 1002 } else { 1003 nextproto = ip6->ip6_nxt; 1004 ip6->ip6_nxt = IPPROTO_FRAGMENT; 1005 } 1006 1007 /* 1008 * Loop through length of segment after first fragment, 1009 * make new header and copy data of each part and link onto 1010 * chain. 1011 */ 1012 m0 = m; 1013 id = htonl(ip6_randomid()); 1014 if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id))) 1015 goto sendorfree; 1016 1017 in6_ifstat_inc(ifp, ifs6_out_fragok); 1018 } 1019 1020 /* 1021 * Remove leading garbages. 1022 */ 1023 sendorfree: 1024 m = m0->m_nextpkt; 1025 m0->m_nextpkt = 0; 1026 m_freem(m0); 1027 for (m0 = m; m; m = m0) { 1028 m0 = m->m_nextpkt; 1029 m->m_nextpkt = 0; 1030 if (error == 0) { 1031 /* Record statistics for this interface address. */ 1032 if (ia) { 1033 counter_u64_add(ia->ia_ifa.ifa_opackets, 1); 1034 counter_u64_add(ia->ia_ifa.ifa_obytes, 1035 m->m_pkthdr.len); 1036 } 1037 error = nd6_output_ifp(ifp, origifp, m, dst); 1038 } else 1039 m_freem(m); 1040 } 1041 1042 if (error == 0) 1043 IP6STAT_INC(ip6s_fragmented); 1044 1045 done: 1046 if (ro == &ip6route) 1047 RO_RTFREE(ro); 1048 if (ro_pmtu == &ip6route) 1049 RO_RTFREE(ro_pmtu); 1050 return (error); 1051 1052 freehdrs: 1053 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1054 m_freem(exthdrs.ip6e_dest1); 1055 m_freem(exthdrs.ip6e_rthdr); 1056 m_freem(exthdrs.ip6e_dest2); 1057 /* FALLTHROUGH */ 1058 bad: 1059 if (m) 1060 m_freem(m); 1061 goto done; 1062 } 1063 1064 static int 1065 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen) 1066 { 1067 struct mbuf *m; 1068 1069 if (hlen > MCLBYTES) 1070 return (ENOBUFS); /* XXX */ 1071 1072 if (hlen > MLEN) 1073 m = m_getcl(M_NOWAIT, MT_DATA, 0); 1074 else 1075 m = m_get(M_NOWAIT, MT_DATA); 1076 if (m == NULL) 1077 return (ENOBUFS); 1078 m->m_len = hlen; 1079 if (hdr) 1080 bcopy(hdr, mtod(m, caddr_t), hlen); 1081 1082 *mp = m; 1083 return (0); 1084 } 1085 1086 /* 1087 * Insert jumbo payload option. 1088 */ 1089 static int 1090 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1091 { 1092 struct mbuf *mopt; 1093 u_char *optbuf; 1094 u_int32_t v; 1095 1096 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1097 1098 /* 1099 * If there is no hop-by-hop options header, allocate new one. 1100 * If there is one but it doesn't have enough space to store the 1101 * jumbo payload option, allocate a cluster to store the whole options. 1102 * Otherwise, use it to store the options. 1103 */ 1104 if (exthdrs->ip6e_hbh == 0) { 1105 mopt = m_get(M_NOWAIT, MT_DATA); 1106 if (mopt == NULL) 1107 return (ENOBUFS); 1108 mopt->m_len = JUMBOOPTLEN; 1109 optbuf = mtod(mopt, u_char *); 1110 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1111 exthdrs->ip6e_hbh = mopt; 1112 } else { 1113 struct ip6_hbh *hbh; 1114 1115 mopt = exthdrs->ip6e_hbh; 1116 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1117 /* 1118 * XXX assumption: 1119 * - exthdrs->ip6e_hbh is not referenced from places 1120 * other than exthdrs. 1121 * - exthdrs->ip6e_hbh is not an mbuf chain. 1122 */ 1123 int oldoptlen = mopt->m_len; 1124 struct mbuf *n; 1125 1126 /* 1127 * XXX: give up if the whole (new) hbh header does 1128 * not fit even in an mbuf cluster. 1129 */ 1130 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1131 return (ENOBUFS); 1132 1133 /* 1134 * As a consequence, we must always prepare a cluster 1135 * at this point. 1136 */ 1137 n = m_getcl(M_NOWAIT, MT_DATA, 0); 1138 if (n == NULL) 1139 return (ENOBUFS); 1140 n->m_len = oldoptlen + JUMBOOPTLEN; 1141 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1142 oldoptlen); 1143 optbuf = mtod(n, caddr_t) + oldoptlen; 1144 m_freem(mopt); 1145 mopt = exthdrs->ip6e_hbh = n; 1146 } else { 1147 optbuf = mtod(mopt, u_char *) + mopt->m_len; 1148 mopt->m_len += JUMBOOPTLEN; 1149 } 1150 optbuf[0] = IP6OPT_PADN; 1151 optbuf[1] = 1; 1152 1153 /* 1154 * Adjust the header length according to the pad and 1155 * the jumbo payload option. 1156 */ 1157 hbh = mtod(mopt, struct ip6_hbh *); 1158 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1159 } 1160 1161 /* fill in the option. */ 1162 optbuf[2] = IP6OPT_JUMBO; 1163 optbuf[3] = 4; 1164 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1165 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1166 1167 /* finally, adjust the packet header length */ 1168 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1169 1170 return (0); 1171 #undef JUMBOOPTLEN 1172 } 1173 1174 /* 1175 * Insert fragment header and copy unfragmentable header portions. 1176 */ 1177 static int 1178 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1179 struct ip6_frag **frghdrp) 1180 { 1181 struct mbuf *n, *mlast; 1182 1183 if (hlen > sizeof(struct ip6_hdr)) { 1184 n = m_copym(m0, sizeof(struct ip6_hdr), 1185 hlen - sizeof(struct ip6_hdr), M_NOWAIT); 1186 if (n == 0) 1187 return (ENOBUFS); 1188 m->m_next = n; 1189 } else 1190 n = m; 1191 1192 /* Search for the last mbuf of unfragmentable part. */ 1193 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1194 ; 1195 1196 if (M_WRITABLE(mlast) && 1197 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1198 /* use the trailing space of the last mbuf for the fragment hdr */ 1199 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1200 mlast->m_len); 1201 mlast->m_len += sizeof(struct ip6_frag); 1202 m->m_pkthdr.len += sizeof(struct ip6_frag); 1203 } else { 1204 /* allocate a new mbuf for the fragment header */ 1205 struct mbuf *mfrg; 1206 1207 mfrg = m_get(M_NOWAIT, MT_DATA); 1208 if (mfrg == NULL) 1209 return (ENOBUFS); 1210 mfrg->m_len = sizeof(struct ip6_frag); 1211 *frghdrp = mtod(mfrg, struct ip6_frag *); 1212 mlast->m_next = mfrg; 1213 } 1214 1215 return (0); 1216 } 1217 1218 static int 1219 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro, 1220 struct ifnet *ifp, struct in6_addr *dst, u_long *mtup, 1221 int *alwaysfragp, u_int fibnum) 1222 { 1223 u_int32_t mtu = 0; 1224 int alwaysfrag = 0; 1225 int error = 0; 1226 1227 if (ro_pmtu != ro) { 1228 /* The first hop and the final destination may differ. */ 1229 struct sockaddr_in6 *sa6_dst = 1230 (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1231 if (ro_pmtu->ro_rt && 1232 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || 1233 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { 1234 RTFREE(ro_pmtu->ro_rt); 1235 ro_pmtu->ro_rt = (struct rtentry *)NULL; 1236 } 1237 if (ro_pmtu->ro_rt == NULL) { 1238 bzero(sa6_dst, sizeof(*sa6_dst)); 1239 sa6_dst->sin6_family = AF_INET6; 1240 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1241 sa6_dst->sin6_addr = *dst; 1242 1243 in6_rtalloc(ro_pmtu, fibnum); 1244 } 1245 } 1246 if (ro_pmtu->ro_rt) { 1247 u_int32_t ifmtu; 1248 struct in_conninfo inc; 1249 1250 bzero(&inc, sizeof(inc)); 1251 inc.inc_flags |= INC_ISIPV6; 1252 inc.inc6_faddr = *dst; 1253 1254 if (ifp == NULL) 1255 ifp = ro_pmtu->ro_rt->rt_ifp; 1256 ifmtu = IN6_LINKMTU(ifp); 1257 mtu = tcp_hc_getmtu(&inc); 1258 if (mtu) 1259 mtu = min(mtu, ro_pmtu->ro_rt->rt_mtu); 1260 else 1261 mtu = ro_pmtu->ro_rt->rt_mtu; 1262 if (mtu == 0) 1263 mtu = ifmtu; 1264 else if (mtu < IPV6_MMTU) { 1265 /* 1266 * RFC2460 section 5, last paragraph: 1267 * if we record ICMPv6 too big message with 1268 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1269 * or smaller, with framgent header attached. 1270 * (fragment header is needed regardless from the 1271 * packet size, for translators to identify packets) 1272 */ 1273 alwaysfrag = 1; 1274 mtu = IPV6_MMTU; 1275 } 1276 } else if (ifp) { 1277 mtu = IN6_LINKMTU(ifp); 1278 } else 1279 error = EHOSTUNREACH; /* XXX */ 1280 1281 *mtup = mtu; 1282 if (alwaysfragp) 1283 *alwaysfragp = alwaysfrag; 1284 return (error); 1285 } 1286 1287 /* 1288 * IP6 socket option processing. 1289 */ 1290 int 1291 ip6_ctloutput(struct socket *so, struct sockopt *sopt) 1292 { 1293 int optdatalen, uproto; 1294 void *optdata; 1295 struct inpcb *in6p = sotoinpcb(so); 1296 int error, optval; 1297 int level, op, optname; 1298 int optlen; 1299 struct thread *td; 1300 #ifdef RSS 1301 uint32_t rss_bucket; 1302 int retval; 1303 #endif 1304 1305 level = sopt->sopt_level; 1306 op = sopt->sopt_dir; 1307 optname = sopt->sopt_name; 1308 optlen = sopt->sopt_valsize; 1309 td = sopt->sopt_td; 1310 error = 0; 1311 optval = 0; 1312 uproto = (int)so->so_proto->pr_protocol; 1313 1314 if (level != IPPROTO_IPV6) { 1315 error = EINVAL; 1316 1317 if (sopt->sopt_level == SOL_SOCKET && 1318 sopt->sopt_dir == SOPT_SET) { 1319 switch (sopt->sopt_name) { 1320 case SO_REUSEADDR: 1321 INP_WLOCK(in6p); 1322 if ((so->so_options & SO_REUSEADDR) != 0) 1323 in6p->inp_flags2 |= INP_REUSEADDR; 1324 else 1325 in6p->inp_flags2 &= ~INP_REUSEADDR; 1326 INP_WUNLOCK(in6p); 1327 error = 0; 1328 break; 1329 case SO_REUSEPORT: 1330 INP_WLOCK(in6p); 1331 if ((so->so_options & SO_REUSEPORT) != 0) 1332 in6p->inp_flags2 |= INP_REUSEPORT; 1333 else 1334 in6p->inp_flags2 &= ~INP_REUSEPORT; 1335 INP_WUNLOCK(in6p); 1336 error = 0; 1337 break; 1338 case SO_SETFIB: 1339 INP_WLOCK(in6p); 1340 in6p->inp_inc.inc_fibnum = so->so_fibnum; 1341 INP_WUNLOCK(in6p); 1342 error = 0; 1343 break; 1344 default: 1345 break; 1346 } 1347 } 1348 } else { /* level == IPPROTO_IPV6 */ 1349 switch (op) { 1350 1351 case SOPT_SET: 1352 switch (optname) { 1353 case IPV6_2292PKTOPTIONS: 1354 #ifdef IPV6_PKTOPTIONS 1355 case IPV6_PKTOPTIONS: 1356 #endif 1357 { 1358 struct mbuf *m; 1359 1360 error = soopt_getm(sopt, &m); /* XXX */ 1361 if (error != 0) 1362 break; 1363 error = soopt_mcopyin(sopt, m); /* XXX */ 1364 if (error != 0) 1365 break; 1366 error = ip6_pcbopts(&in6p->in6p_outputopts, 1367 m, so, sopt); 1368 m_freem(m); /* XXX */ 1369 break; 1370 } 1371 1372 /* 1373 * Use of some Hop-by-Hop options or some 1374 * Destination options, might require special 1375 * privilege. That is, normal applications 1376 * (without special privilege) might be forbidden 1377 * from setting certain options in outgoing packets, 1378 * and might never see certain options in received 1379 * packets. [RFC 2292 Section 6] 1380 * KAME specific note: 1381 * KAME prevents non-privileged users from sending or 1382 * receiving ANY hbh/dst options in order to avoid 1383 * overhead of parsing options in the kernel. 1384 */ 1385 case IPV6_RECVHOPOPTS: 1386 case IPV6_RECVDSTOPTS: 1387 case IPV6_RECVRTHDRDSTOPTS: 1388 if (td != NULL) { 1389 error = priv_check(td, 1390 PRIV_NETINET_SETHDROPTS); 1391 if (error) 1392 break; 1393 } 1394 /* FALLTHROUGH */ 1395 case IPV6_UNICAST_HOPS: 1396 case IPV6_HOPLIMIT: 1397 1398 case IPV6_RECVPKTINFO: 1399 case IPV6_RECVHOPLIMIT: 1400 case IPV6_RECVRTHDR: 1401 case IPV6_RECVPATHMTU: 1402 case IPV6_RECVTCLASS: 1403 case IPV6_RECVFLOWID: 1404 #ifdef RSS 1405 case IPV6_RECVRSSBUCKETID: 1406 #endif 1407 case IPV6_V6ONLY: 1408 case IPV6_AUTOFLOWLABEL: 1409 case IPV6_BINDANY: 1410 case IPV6_BINDMULTI: 1411 #ifdef RSS 1412 case IPV6_RSS_LISTEN_BUCKET: 1413 #endif 1414 if (optname == IPV6_BINDANY && td != NULL) { 1415 error = priv_check(td, 1416 PRIV_NETINET_BINDANY); 1417 if (error) 1418 break; 1419 } 1420 1421 if (optlen != sizeof(int)) { 1422 error = EINVAL; 1423 break; 1424 } 1425 error = sooptcopyin(sopt, &optval, 1426 sizeof optval, sizeof optval); 1427 if (error) 1428 break; 1429 switch (optname) { 1430 1431 case IPV6_UNICAST_HOPS: 1432 if (optval < -1 || optval >= 256) 1433 error = EINVAL; 1434 else { 1435 /* -1 = kernel default */ 1436 in6p->in6p_hops = optval; 1437 if ((in6p->inp_vflag & 1438 INP_IPV4) != 0) 1439 in6p->inp_ip_ttl = optval; 1440 } 1441 break; 1442 #define OPTSET(bit) \ 1443 do { \ 1444 INP_WLOCK(in6p); \ 1445 if (optval) \ 1446 in6p->inp_flags |= (bit); \ 1447 else \ 1448 in6p->inp_flags &= ~(bit); \ 1449 INP_WUNLOCK(in6p); \ 1450 } while (/*CONSTCOND*/ 0) 1451 #define OPTSET2292(bit) \ 1452 do { \ 1453 INP_WLOCK(in6p); \ 1454 in6p->inp_flags |= IN6P_RFC2292; \ 1455 if (optval) \ 1456 in6p->inp_flags |= (bit); \ 1457 else \ 1458 in6p->inp_flags &= ~(bit); \ 1459 INP_WUNLOCK(in6p); \ 1460 } while (/*CONSTCOND*/ 0) 1461 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0) 1462 1463 #define OPTSET2(bit, val) do { \ 1464 INP_WLOCK(in6p); \ 1465 if (val) \ 1466 in6p->inp_flags2 |= bit; \ 1467 else \ 1468 in6p->inp_flags2 &= ~bit; \ 1469 INP_WUNLOCK(in6p); \ 1470 } while (0) 1471 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0) 1472 1473 case IPV6_RECVPKTINFO: 1474 /* cannot mix with RFC2292 */ 1475 if (OPTBIT(IN6P_RFC2292)) { 1476 error = EINVAL; 1477 break; 1478 } 1479 OPTSET(IN6P_PKTINFO); 1480 break; 1481 1482 case IPV6_HOPLIMIT: 1483 { 1484 struct ip6_pktopts **optp; 1485 1486 /* cannot mix with RFC2292 */ 1487 if (OPTBIT(IN6P_RFC2292)) { 1488 error = EINVAL; 1489 break; 1490 } 1491 optp = &in6p->in6p_outputopts; 1492 error = ip6_pcbopt(IPV6_HOPLIMIT, 1493 (u_char *)&optval, sizeof(optval), 1494 optp, (td != NULL) ? td->td_ucred : 1495 NULL, uproto); 1496 break; 1497 } 1498 1499 case IPV6_RECVHOPLIMIT: 1500 /* cannot mix with RFC2292 */ 1501 if (OPTBIT(IN6P_RFC2292)) { 1502 error = EINVAL; 1503 break; 1504 } 1505 OPTSET(IN6P_HOPLIMIT); 1506 break; 1507 1508 case IPV6_RECVHOPOPTS: 1509 /* cannot mix with RFC2292 */ 1510 if (OPTBIT(IN6P_RFC2292)) { 1511 error = EINVAL; 1512 break; 1513 } 1514 OPTSET(IN6P_HOPOPTS); 1515 break; 1516 1517 case IPV6_RECVDSTOPTS: 1518 /* cannot mix with RFC2292 */ 1519 if (OPTBIT(IN6P_RFC2292)) { 1520 error = EINVAL; 1521 break; 1522 } 1523 OPTSET(IN6P_DSTOPTS); 1524 break; 1525 1526 case IPV6_RECVRTHDRDSTOPTS: 1527 /* cannot mix with RFC2292 */ 1528 if (OPTBIT(IN6P_RFC2292)) { 1529 error = EINVAL; 1530 break; 1531 } 1532 OPTSET(IN6P_RTHDRDSTOPTS); 1533 break; 1534 1535 case IPV6_RECVRTHDR: 1536 /* cannot mix with RFC2292 */ 1537 if (OPTBIT(IN6P_RFC2292)) { 1538 error = EINVAL; 1539 break; 1540 } 1541 OPTSET(IN6P_RTHDR); 1542 break; 1543 1544 case IPV6_RECVPATHMTU: 1545 /* 1546 * We ignore this option for TCP 1547 * sockets. 1548 * (RFC3542 leaves this case 1549 * unspecified.) 1550 */ 1551 if (uproto != IPPROTO_TCP) 1552 OPTSET(IN6P_MTU); 1553 break; 1554 1555 case IPV6_RECVFLOWID: 1556 OPTSET2(INP_RECVFLOWID, optval); 1557 break; 1558 1559 #ifdef RSS 1560 case IPV6_RECVRSSBUCKETID: 1561 OPTSET2(INP_RECVRSSBUCKETID, optval); 1562 break; 1563 #endif 1564 1565 case IPV6_V6ONLY: 1566 /* 1567 * make setsockopt(IPV6_V6ONLY) 1568 * available only prior to bind(2). 1569 * see ipng mailing list, Jun 22 2001. 1570 */ 1571 if (in6p->inp_lport || 1572 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1573 error = EINVAL; 1574 break; 1575 } 1576 OPTSET(IN6P_IPV6_V6ONLY); 1577 if (optval) 1578 in6p->inp_vflag &= ~INP_IPV4; 1579 else 1580 in6p->inp_vflag |= INP_IPV4; 1581 break; 1582 case IPV6_RECVTCLASS: 1583 /* cannot mix with RFC2292 XXX */ 1584 if (OPTBIT(IN6P_RFC2292)) { 1585 error = EINVAL; 1586 break; 1587 } 1588 OPTSET(IN6P_TCLASS); 1589 break; 1590 case IPV6_AUTOFLOWLABEL: 1591 OPTSET(IN6P_AUTOFLOWLABEL); 1592 break; 1593 1594 case IPV6_BINDANY: 1595 OPTSET(INP_BINDANY); 1596 break; 1597 1598 case IPV6_BINDMULTI: 1599 OPTSET2(INP_BINDMULTI, optval); 1600 break; 1601 #ifdef RSS 1602 case IPV6_RSS_LISTEN_BUCKET: 1603 if ((optval >= 0) && 1604 (optval < rss_getnumbuckets())) { 1605 in6p->inp_rss_listen_bucket = optval; 1606 OPTSET2(INP_RSS_BUCKET_SET, 1); 1607 } else { 1608 error = EINVAL; 1609 } 1610 break; 1611 #endif 1612 } 1613 break; 1614 1615 case IPV6_TCLASS: 1616 case IPV6_DONTFRAG: 1617 case IPV6_USE_MIN_MTU: 1618 case IPV6_PREFER_TEMPADDR: 1619 if (optlen != sizeof(optval)) { 1620 error = EINVAL; 1621 break; 1622 } 1623 error = sooptcopyin(sopt, &optval, 1624 sizeof optval, sizeof optval); 1625 if (error) 1626 break; 1627 { 1628 struct ip6_pktopts **optp; 1629 optp = &in6p->in6p_outputopts; 1630 error = ip6_pcbopt(optname, 1631 (u_char *)&optval, sizeof(optval), 1632 optp, (td != NULL) ? td->td_ucred : 1633 NULL, uproto); 1634 break; 1635 } 1636 1637 case IPV6_2292PKTINFO: 1638 case IPV6_2292HOPLIMIT: 1639 case IPV6_2292HOPOPTS: 1640 case IPV6_2292DSTOPTS: 1641 case IPV6_2292RTHDR: 1642 /* RFC 2292 */ 1643 if (optlen != sizeof(int)) { 1644 error = EINVAL; 1645 break; 1646 } 1647 error = sooptcopyin(sopt, &optval, 1648 sizeof optval, sizeof optval); 1649 if (error) 1650 break; 1651 switch (optname) { 1652 case IPV6_2292PKTINFO: 1653 OPTSET2292(IN6P_PKTINFO); 1654 break; 1655 case IPV6_2292HOPLIMIT: 1656 OPTSET2292(IN6P_HOPLIMIT); 1657 break; 1658 case IPV6_2292HOPOPTS: 1659 /* 1660 * Check super-user privilege. 1661 * See comments for IPV6_RECVHOPOPTS. 1662 */ 1663 if (td != NULL) { 1664 error = priv_check(td, 1665 PRIV_NETINET_SETHDROPTS); 1666 if (error) 1667 return (error); 1668 } 1669 OPTSET2292(IN6P_HOPOPTS); 1670 break; 1671 case IPV6_2292DSTOPTS: 1672 if (td != NULL) { 1673 error = priv_check(td, 1674 PRIV_NETINET_SETHDROPTS); 1675 if (error) 1676 return (error); 1677 } 1678 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1679 break; 1680 case IPV6_2292RTHDR: 1681 OPTSET2292(IN6P_RTHDR); 1682 break; 1683 } 1684 break; 1685 case IPV6_PKTINFO: 1686 case IPV6_HOPOPTS: 1687 case IPV6_RTHDR: 1688 case IPV6_DSTOPTS: 1689 case IPV6_RTHDRDSTOPTS: 1690 case IPV6_NEXTHOP: 1691 { 1692 /* new advanced API (RFC3542) */ 1693 u_char *optbuf; 1694 u_char optbuf_storage[MCLBYTES]; 1695 int optlen; 1696 struct ip6_pktopts **optp; 1697 1698 /* cannot mix with RFC2292 */ 1699 if (OPTBIT(IN6P_RFC2292)) { 1700 error = EINVAL; 1701 break; 1702 } 1703 1704 /* 1705 * We only ensure valsize is not too large 1706 * here. Further validation will be done 1707 * later. 1708 */ 1709 error = sooptcopyin(sopt, optbuf_storage, 1710 sizeof(optbuf_storage), 0); 1711 if (error) 1712 break; 1713 optlen = sopt->sopt_valsize; 1714 optbuf = optbuf_storage; 1715 optp = &in6p->in6p_outputopts; 1716 error = ip6_pcbopt(optname, optbuf, optlen, 1717 optp, (td != NULL) ? td->td_ucred : NULL, 1718 uproto); 1719 break; 1720 } 1721 #undef OPTSET 1722 1723 case IPV6_MULTICAST_IF: 1724 case IPV6_MULTICAST_HOPS: 1725 case IPV6_MULTICAST_LOOP: 1726 case IPV6_JOIN_GROUP: 1727 case IPV6_LEAVE_GROUP: 1728 case IPV6_MSFILTER: 1729 case MCAST_BLOCK_SOURCE: 1730 case MCAST_UNBLOCK_SOURCE: 1731 case MCAST_JOIN_GROUP: 1732 case MCAST_LEAVE_GROUP: 1733 case MCAST_JOIN_SOURCE_GROUP: 1734 case MCAST_LEAVE_SOURCE_GROUP: 1735 error = ip6_setmoptions(in6p, sopt); 1736 break; 1737 1738 case IPV6_PORTRANGE: 1739 error = sooptcopyin(sopt, &optval, 1740 sizeof optval, sizeof optval); 1741 if (error) 1742 break; 1743 1744 INP_WLOCK(in6p); 1745 switch (optval) { 1746 case IPV6_PORTRANGE_DEFAULT: 1747 in6p->inp_flags &= ~(INP_LOWPORT); 1748 in6p->inp_flags &= ~(INP_HIGHPORT); 1749 break; 1750 1751 case IPV6_PORTRANGE_HIGH: 1752 in6p->inp_flags &= ~(INP_LOWPORT); 1753 in6p->inp_flags |= INP_HIGHPORT; 1754 break; 1755 1756 case IPV6_PORTRANGE_LOW: 1757 in6p->inp_flags &= ~(INP_HIGHPORT); 1758 in6p->inp_flags |= INP_LOWPORT; 1759 break; 1760 1761 default: 1762 error = EINVAL; 1763 break; 1764 } 1765 INP_WUNLOCK(in6p); 1766 break; 1767 1768 #ifdef IPSEC 1769 case IPV6_IPSEC_POLICY: 1770 { 1771 caddr_t req; 1772 struct mbuf *m; 1773 1774 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1775 break; 1776 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1777 break; 1778 req = mtod(m, caddr_t); 1779 error = ipsec_set_policy(in6p, optname, req, 1780 m->m_len, (sopt->sopt_td != NULL) ? 1781 sopt->sopt_td->td_ucred : NULL); 1782 m_freem(m); 1783 break; 1784 } 1785 #endif /* IPSEC */ 1786 1787 default: 1788 error = ENOPROTOOPT; 1789 break; 1790 } 1791 break; 1792 1793 case SOPT_GET: 1794 switch (optname) { 1795 1796 case IPV6_2292PKTOPTIONS: 1797 #ifdef IPV6_PKTOPTIONS 1798 case IPV6_PKTOPTIONS: 1799 #endif 1800 /* 1801 * RFC3542 (effectively) deprecated the 1802 * semantics of the 2292-style pktoptions. 1803 * Since it was not reliable in nature (i.e., 1804 * applications had to expect the lack of some 1805 * information after all), it would make sense 1806 * to simplify this part by always returning 1807 * empty data. 1808 */ 1809 sopt->sopt_valsize = 0; 1810 break; 1811 1812 case IPV6_RECVHOPOPTS: 1813 case IPV6_RECVDSTOPTS: 1814 case IPV6_RECVRTHDRDSTOPTS: 1815 case IPV6_UNICAST_HOPS: 1816 case IPV6_RECVPKTINFO: 1817 case IPV6_RECVHOPLIMIT: 1818 case IPV6_RECVRTHDR: 1819 case IPV6_RECVPATHMTU: 1820 1821 case IPV6_V6ONLY: 1822 case IPV6_PORTRANGE: 1823 case IPV6_RECVTCLASS: 1824 case IPV6_AUTOFLOWLABEL: 1825 case IPV6_BINDANY: 1826 case IPV6_FLOWID: 1827 case IPV6_FLOWTYPE: 1828 case IPV6_RECVFLOWID: 1829 #ifdef RSS 1830 case IPV6_RSSBUCKETID: 1831 case IPV6_RECVRSSBUCKETID: 1832 #endif 1833 switch (optname) { 1834 1835 case IPV6_RECVHOPOPTS: 1836 optval = OPTBIT(IN6P_HOPOPTS); 1837 break; 1838 1839 case IPV6_RECVDSTOPTS: 1840 optval = OPTBIT(IN6P_DSTOPTS); 1841 break; 1842 1843 case IPV6_RECVRTHDRDSTOPTS: 1844 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1845 break; 1846 1847 case IPV6_UNICAST_HOPS: 1848 optval = in6p->in6p_hops; 1849 break; 1850 1851 case IPV6_RECVPKTINFO: 1852 optval = OPTBIT(IN6P_PKTINFO); 1853 break; 1854 1855 case IPV6_RECVHOPLIMIT: 1856 optval = OPTBIT(IN6P_HOPLIMIT); 1857 break; 1858 1859 case IPV6_RECVRTHDR: 1860 optval = OPTBIT(IN6P_RTHDR); 1861 break; 1862 1863 case IPV6_RECVPATHMTU: 1864 optval = OPTBIT(IN6P_MTU); 1865 break; 1866 1867 case IPV6_V6ONLY: 1868 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1869 break; 1870 1871 case IPV6_PORTRANGE: 1872 { 1873 int flags; 1874 flags = in6p->inp_flags; 1875 if (flags & INP_HIGHPORT) 1876 optval = IPV6_PORTRANGE_HIGH; 1877 else if (flags & INP_LOWPORT) 1878 optval = IPV6_PORTRANGE_LOW; 1879 else 1880 optval = 0; 1881 break; 1882 } 1883 case IPV6_RECVTCLASS: 1884 optval = OPTBIT(IN6P_TCLASS); 1885 break; 1886 1887 case IPV6_AUTOFLOWLABEL: 1888 optval = OPTBIT(IN6P_AUTOFLOWLABEL); 1889 break; 1890 1891 case IPV6_BINDANY: 1892 optval = OPTBIT(INP_BINDANY); 1893 break; 1894 1895 case IPV6_FLOWID: 1896 optval = in6p->inp_flowid; 1897 break; 1898 1899 case IPV6_FLOWTYPE: 1900 optval = in6p->inp_flowtype; 1901 break; 1902 1903 case IPV6_RECVFLOWID: 1904 optval = OPTBIT2(INP_RECVFLOWID); 1905 break; 1906 #ifdef RSS 1907 case IPV6_RSSBUCKETID: 1908 retval = 1909 rss_hash2bucket(in6p->inp_flowid, 1910 in6p->inp_flowtype, 1911 &rss_bucket); 1912 if (retval == 0) 1913 optval = rss_bucket; 1914 else 1915 error = EINVAL; 1916 break; 1917 1918 case IPV6_RECVRSSBUCKETID: 1919 optval = OPTBIT2(INP_RECVRSSBUCKETID); 1920 break; 1921 #endif 1922 1923 case IPV6_BINDMULTI: 1924 optval = OPTBIT2(INP_BINDMULTI); 1925 break; 1926 1927 } 1928 if (error) 1929 break; 1930 error = sooptcopyout(sopt, &optval, 1931 sizeof optval); 1932 break; 1933 1934 case IPV6_PATHMTU: 1935 { 1936 u_long pmtu = 0; 1937 struct ip6_mtuinfo mtuinfo; 1938 struct route_in6 sro; 1939 1940 bzero(&sro, sizeof(sro)); 1941 1942 if (!(so->so_state & SS_ISCONNECTED)) 1943 return (ENOTCONN); 1944 /* 1945 * XXX: we dot not consider the case of source 1946 * routing, or optional information to specify 1947 * the outgoing interface. 1948 */ 1949 error = ip6_getpmtu(&sro, NULL, NULL, 1950 &in6p->in6p_faddr, &pmtu, NULL, 1951 so->so_fibnum); 1952 if (sro.ro_rt) 1953 RTFREE(sro.ro_rt); 1954 if (error) 1955 break; 1956 if (pmtu > IPV6_MAXPACKET) 1957 pmtu = IPV6_MAXPACKET; 1958 1959 bzero(&mtuinfo, sizeof(mtuinfo)); 1960 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 1961 optdata = (void *)&mtuinfo; 1962 optdatalen = sizeof(mtuinfo); 1963 error = sooptcopyout(sopt, optdata, 1964 optdatalen); 1965 break; 1966 } 1967 1968 case IPV6_2292PKTINFO: 1969 case IPV6_2292HOPLIMIT: 1970 case IPV6_2292HOPOPTS: 1971 case IPV6_2292RTHDR: 1972 case IPV6_2292DSTOPTS: 1973 switch (optname) { 1974 case IPV6_2292PKTINFO: 1975 optval = OPTBIT(IN6P_PKTINFO); 1976 break; 1977 case IPV6_2292HOPLIMIT: 1978 optval = OPTBIT(IN6P_HOPLIMIT); 1979 break; 1980 case IPV6_2292HOPOPTS: 1981 optval = OPTBIT(IN6P_HOPOPTS); 1982 break; 1983 case IPV6_2292RTHDR: 1984 optval = OPTBIT(IN6P_RTHDR); 1985 break; 1986 case IPV6_2292DSTOPTS: 1987 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 1988 break; 1989 } 1990 error = sooptcopyout(sopt, &optval, 1991 sizeof optval); 1992 break; 1993 case IPV6_PKTINFO: 1994 case IPV6_HOPOPTS: 1995 case IPV6_RTHDR: 1996 case IPV6_DSTOPTS: 1997 case IPV6_RTHDRDSTOPTS: 1998 case IPV6_NEXTHOP: 1999 case IPV6_TCLASS: 2000 case IPV6_DONTFRAG: 2001 case IPV6_USE_MIN_MTU: 2002 case IPV6_PREFER_TEMPADDR: 2003 error = ip6_getpcbopt(in6p->in6p_outputopts, 2004 optname, sopt); 2005 break; 2006 2007 case IPV6_MULTICAST_IF: 2008 case IPV6_MULTICAST_HOPS: 2009 case IPV6_MULTICAST_LOOP: 2010 case IPV6_MSFILTER: 2011 error = ip6_getmoptions(in6p, sopt); 2012 break; 2013 2014 #ifdef IPSEC 2015 case IPV6_IPSEC_POLICY: 2016 { 2017 caddr_t req = NULL; 2018 size_t len = 0; 2019 struct mbuf *m = NULL; 2020 struct mbuf **mp = &m; 2021 size_t ovalsize = sopt->sopt_valsize; 2022 caddr_t oval = (caddr_t)sopt->sopt_val; 2023 2024 error = soopt_getm(sopt, &m); /* XXX */ 2025 if (error != 0) 2026 break; 2027 error = soopt_mcopyin(sopt, m); /* XXX */ 2028 if (error != 0) 2029 break; 2030 sopt->sopt_valsize = ovalsize; 2031 sopt->sopt_val = oval; 2032 if (m) { 2033 req = mtod(m, caddr_t); 2034 len = m->m_len; 2035 } 2036 error = ipsec_get_policy(in6p, req, len, mp); 2037 if (error == 0) 2038 error = soopt_mcopyout(sopt, m); /* XXX */ 2039 if (error == 0 && m) 2040 m_freem(m); 2041 break; 2042 } 2043 #endif /* IPSEC */ 2044 2045 default: 2046 error = ENOPROTOOPT; 2047 break; 2048 } 2049 break; 2050 } 2051 } 2052 return (error); 2053 } 2054 2055 int 2056 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt) 2057 { 2058 int error = 0, optval, optlen; 2059 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2060 struct inpcb *in6p = sotoinpcb(so); 2061 int level, op, optname; 2062 2063 level = sopt->sopt_level; 2064 op = sopt->sopt_dir; 2065 optname = sopt->sopt_name; 2066 optlen = sopt->sopt_valsize; 2067 2068 if (level != IPPROTO_IPV6) { 2069 return (EINVAL); 2070 } 2071 2072 switch (optname) { 2073 case IPV6_CHECKSUM: 2074 /* 2075 * For ICMPv6 sockets, no modification allowed for checksum 2076 * offset, permit "no change" values to help existing apps. 2077 * 2078 * RFC3542 says: "An attempt to set IPV6_CHECKSUM 2079 * for an ICMPv6 socket will fail." 2080 * The current behavior does not meet RFC3542. 2081 */ 2082 switch (op) { 2083 case SOPT_SET: 2084 if (optlen != sizeof(int)) { 2085 error = EINVAL; 2086 break; 2087 } 2088 error = sooptcopyin(sopt, &optval, sizeof(optval), 2089 sizeof(optval)); 2090 if (error) 2091 break; 2092 if ((optval % 2) != 0) { 2093 /* the API assumes even offset values */ 2094 error = EINVAL; 2095 } else if (so->so_proto->pr_protocol == 2096 IPPROTO_ICMPV6) { 2097 if (optval != icmp6off) 2098 error = EINVAL; 2099 } else 2100 in6p->in6p_cksum = optval; 2101 break; 2102 2103 case SOPT_GET: 2104 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2105 optval = icmp6off; 2106 else 2107 optval = in6p->in6p_cksum; 2108 2109 error = sooptcopyout(sopt, &optval, sizeof(optval)); 2110 break; 2111 2112 default: 2113 error = EINVAL; 2114 break; 2115 } 2116 break; 2117 2118 default: 2119 error = ENOPROTOOPT; 2120 break; 2121 } 2122 2123 return (error); 2124 } 2125 2126 /* 2127 * Set up IP6 options in pcb for insertion in output packets or 2128 * specifying behavior of outgoing packets. 2129 */ 2130 static int 2131 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, 2132 struct socket *so, struct sockopt *sopt) 2133 { 2134 struct ip6_pktopts *opt = *pktopt; 2135 int error = 0; 2136 struct thread *td = sopt->sopt_td; 2137 2138 /* turn off any old options. */ 2139 if (opt) { 2140 #ifdef DIAGNOSTIC 2141 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2142 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2143 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2144 printf("ip6_pcbopts: all specified options are cleared.\n"); 2145 #endif 2146 ip6_clearpktopts(opt, -1); 2147 } else 2148 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2149 *pktopt = NULL; 2150 2151 if (!m || m->m_len == 0) { 2152 /* 2153 * Only turning off any previous options, regardless of 2154 * whether the opt is just created or given. 2155 */ 2156 free(opt, M_IP6OPT); 2157 return (0); 2158 } 2159 2160 /* set options specified by user. */ 2161 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ? 2162 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) { 2163 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2164 free(opt, M_IP6OPT); 2165 return (error); 2166 } 2167 *pktopt = opt; 2168 return (0); 2169 } 2170 2171 /* 2172 * initialize ip6_pktopts. beware that there are non-zero default values in 2173 * the struct. 2174 */ 2175 void 2176 ip6_initpktopts(struct ip6_pktopts *opt) 2177 { 2178 2179 bzero(opt, sizeof(*opt)); 2180 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2181 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2182 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2183 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; 2184 } 2185 2186 static int 2187 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2188 struct ucred *cred, int uproto) 2189 { 2190 struct ip6_pktopts *opt; 2191 2192 if (*pktopt == NULL) { 2193 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2194 M_WAITOK); 2195 ip6_initpktopts(*pktopt); 2196 } 2197 opt = *pktopt; 2198 2199 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); 2200 } 2201 2202 static int 2203 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 2204 { 2205 void *optdata = NULL; 2206 int optdatalen = 0; 2207 struct ip6_ext *ip6e; 2208 int error = 0; 2209 struct in6_pktinfo null_pktinfo; 2210 int deftclass = 0, on; 2211 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2212 int defpreftemp = IP6PO_TEMPADDR_SYSTEM; 2213 2214 switch (optname) { 2215 case IPV6_PKTINFO: 2216 optdata = (void *)&null_pktinfo; 2217 if (pktopt && pktopt->ip6po_pktinfo) { 2218 bcopy(pktopt->ip6po_pktinfo, &null_pktinfo, 2219 sizeof(null_pktinfo)); 2220 in6_clearscope(&null_pktinfo.ipi6_addr); 2221 } else { 2222 /* XXX: we don't have to do this every time... */ 2223 bzero(&null_pktinfo, sizeof(null_pktinfo)); 2224 } 2225 optdatalen = sizeof(struct in6_pktinfo); 2226 break; 2227 case IPV6_TCLASS: 2228 if (pktopt && pktopt->ip6po_tclass >= 0) 2229 optdata = (void *)&pktopt->ip6po_tclass; 2230 else 2231 optdata = (void *)&deftclass; 2232 optdatalen = sizeof(int); 2233 break; 2234 case IPV6_HOPOPTS: 2235 if (pktopt && pktopt->ip6po_hbh) { 2236 optdata = (void *)pktopt->ip6po_hbh; 2237 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2238 optdatalen = (ip6e->ip6e_len + 1) << 3; 2239 } 2240 break; 2241 case IPV6_RTHDR: 2242 if (pktopt && pktopt->ip6po_rthdr) { 2243 optdata = (void *)pktopt->ip6po_rthdr; 2244 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2245 optdatalen = (ip6e->ip6e_len + 1) << 3; 2246 } 2247 break; 2248 case IPV6_RTHDRDSTOPTS: 2249 if (pktopt && pktopt->ip6po_dest1) { 2250 optdata = (void *)pktopt->ip6po_dest1; 2251 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2252 optdatalen = (ip6e->ip6e_len + 1) << 3; 2253 } 2254 break; 2255 case IPV6_DSTOPTS: 2256 if (pktopt && pktopt->ip6po_dest2) { 2257 optdata = (void *)pktopt->ip6po_dest2; 2258 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2259 optdatalen = (ip6e->ip6e_len + 1) << 3; 2260 } 2261 break; 2262 case IPV6_NEXTHOP: 2263 if (pktopt && pktopt->ip6po_nexthop) { 2264 optdata = (void *)pktopt->ip6po_nexthop; 2265 optdatalen = pktopt->ip6po_nexthop->sa_len; 2266 } 2267 break; 2268 case IPV6_USE_MIN_MTU: 2269 if (pktopt) 2270 optdata = (void *)&pktopt->ip6po_minmtu; 2271 else 2272 optdata = (void *)&defminmtu; 2273 optdatalen = sizeof(int); 2274 break; 2275 case IPV6_DONTFRAG: 2276 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2277 on = 1; 2278 else 2279 on = 0; 2280 optdata = (void *)&on; 2281 optdatalen = sizeof(on); 2282 break; 2283 case IPV6_PREFER_TEMPADDR: 2284 if (pktopt) 2285 optdata = (void *)&pktopt->ip6po_prefer_tempaddr; 2286 else 2287 optdata = (void *)&defpreftemp; 2288 optdatalen = sizeof(int); 2289 break; 2290 default: /* should not happen */ 2291 #ifdef DIAGNOSTIC 2292 panic("ip6_getpcbopt: unexpected option\n"); 2293 #endif 2294 return (ENOPROTOOPT); 2295 } 2296 2297 error = sooptcopyout(sopt, optdata, optdatalen); 2298 2299 return (error); 2300 } 2301 2302 void 2303 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2304 { 2305 if (pktopt == NULL) 2306 return; 2307 2308 if (optname == -1 || optname == IPV6_PKTINFO) { 2309 if (pktopt->ip6po_pktinfo) 2310 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2311 pktopt->ip6po_pktinfo = NULL; 2312 } 2313 if (optname == -1 || optname == IPV6_HOPLIMIT) 2314 pktopt->ip6po_hlim = -1; 2315 if (optname == -1 || optname == IPV6_TCLASS) 2316 pktopt->ip6po_tclass = -1; 2317 if (optname == -1 || optname == IPV6_NEXTHOP) { 2318 if (pktopt->ip6po_nextroute.ro_rt) { 2319 RTFREE(pktopt->ip6po_nextroute.ro_rt); 2320 pktopt->ip6po_nextroute.ro_rt = NULL; 2321 } 2322 if (pktopt->ip6po_nexthop) 2323 free(pktopt->ip6po_nexthop, M_IP6OPT); 2324 pktopt->ip6po_nexthop = NULL; 2325 } 2326 if (optname == -1 || optname == IPV6_HOPOPTS) { 2327 if (pktopt->ip6po_hbh) 2328 free(pktopt->ip6po_hbh, M_IP6OPT); 2329 pktopt->ip6po_hbh = NULL; 2330 } 2331 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2332 if (pktopt->ip6po_dest1) 2333 free(pktopt->ip6po_dest1, M_IP6OPT); 2334 pktopt->ip6po_dest1 = NULL; 2335 } 2336 if (optname == -1 || optname == IPV6_RTHDR) { 2337 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2338 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2339 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2340 if (pktopt->ip6po_route.ro_rt) { 2341 RTFREE(pktopt->ip6po_route.ro_rt); 2342 pktopt->ip6po_route.ro_rt = NULL; 2343 } 2344 } 2345 if (optname == -1 || optname == IPV6_DSTOPTS) { 2346 if (pktopt->ip6po_dest2) 2347 free(pktopt->ip6po_dest2, M_IP6OPT); 2348 pktopt->ip6po_dest2 = NULL; 2349 } 2350 } 2351 2352 #define PKTOPT_EXTHDRCPY(type) \ 2353 do {\ 2354 if (src->type) {\ 2355 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2356 dst->type = malloc(hlen, M_IP6OPT, canwait);\ 2357 if (dst->type == NULL && canwait == M_NOWAIT)\ 2358 goto bad;\ 2359 bcopy(src->type, dst->type, hlen);\ 2360 }\ 2361 } while (/*CONSTCOND*/ 0) 2362 2363 static int 2364 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2365 { 2366 if (dst == NULL || src == NULL) { 2367 printf("ip6_clearpktopts: invalid argument\n"); 2368 return (EINVAL); 2369 } 2370 2371 dst->ip6po_hlim = src->ip6po_hlim; 2372 dst->ip6po_tclass = src->ip6po_tclass; 2373 dst->ip6po_flags = src->ip6po_flags; 2374 dst->ip6po_minmtu = src->ip6po_minmtu; 2375 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr; 2376 if (src->ip6po_pktinfo) { 2377 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2378 M_IP6OPT, canwait); 2379 if (dst->ip6po_pktinfo == NULL) 2380 goto bad; 2381 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2382 } 2383 if (src->ip6po_nexthop) { 2384 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2385 M_IP6OPT, canwait); 2386 if (dst->ip6po_nexthop == NULL) 2387 goto bad; 2388 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop, 2389 src->ip6po_nexthop->sa_len); 2390 } 2391 PKTOPT_EXTHDRCPY(ip6po_hbh); 2392 PKTOPT_EXTHDRCPY(ip6po_dest1); 2393 PKTOPT_EXTHDRCPY(ip6po_dest2); 2394 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2395 return (0); 2396 2397 bad: 2398 ip6_clearpktopts(dst, -1); 2399 return (ENOBUFS); 2400 } 2401 #undef PKTOPT_EXTHDRCPY 2402 2403 struct ip6_pktopts * 2404 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2405 { 2406 int error; 2407 struct ip6_pktopts *dst; 2408 2409 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2410 if (dst == NULL) 2411 return (NULL); 2412 ip6_initpktopts(dst); 2413 2414 if ((error = copypktopts(dst, src, canwait)) != 0) { 2415 free(dst, M_IP6OPT); 2416 return (NULL); 2417 } 2418 2419 return (dst); 2420 } 2421 2422 void 2423 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2424 { 2425 if (pktopt == NULL) 2426 return; 2427 2428 ip6_clearpktopts(pktopt, -1); 2429 2430 free(pktopt, M_IP6OPT); 2431 } 2432 2433 /* 2434 * Set IPv6 outgoing packet options based on advanced API. 2435 */ 2436 int 2437 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2438 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto) 2439 { 2440 struct cmsghdr *cm = 0; 2441 2442 if (control == NULL || opt == NULL) 2443 return (EINVAL); 2444 2445 ip6_initpktopts(opt); 2446 if (stickyopt) { 2447 int error; 2448 2449 /* 2450 * If stickyopt is provided, make a local copy of the options 2451 * for this particular packet, then override them by ancillary 2452 * objects. 2453 * XXX: copypktopts() does not copy the cached route to a next 2454 * hop (if any). This is not very good in terms of efficiency, 2455 * but we can allow this since this option should be rarely 2456 * used. 2457 */ 2458 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2459 return (error); 2460 } 2461 2462 /* 2463 * XXX: Currently, we assume all the optional information is stored 2464 * in a single mbuf. 2465 */ 2466 if (control->m_next) 2467 return (EINVAL); 2468 2469 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2470 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2471 int error; 2472 2473 if (control->m_len < CMSG_LEN(0)) 2474 return (EINVAL); 2475 2476 cm = mtod(control, struct cmsghdr *); 2477 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2478 return (EINVAL); 2479 if (cm->cmsg_level != IPPROTO_IPV6) 2480 continue; 2481 2482 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2483 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); 2484 if (error) 2485 return (error); 2486 } 2487 2488 return (0); 2489 } 2490 2491 /* 2492 * Set a particular packet option, as a sticky option or an ancillary data 2493 * item. "len" can be 0 only when it's a sticky option. 2494 * We have 4 cases of combination of "sticky" and "cmsg": 2495 * "sticky=0, cmsg=0": impossible 2496 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2497 * "sticky=1, cmsg=0": RFC3542 socket option 2498 * "sticky=1, cmsg=1": RFC2292 socket option 2499 */ 2500 static int 2501 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2502 struct ucred *cred, int sticky, int cmsg, int uproto) 2503 { 2504 int minmtupolicy, preftemp; 2505 int error; 2506 2507 if (!sticky && !cmsg) { 2508 #ifdef DIAGNOSTIC 2509 printf("ip6_setpktopt: impossible case\n"); 2510 #endif 2511 return (EINVAL); 2512 } 2513 2514 /* 2515 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2516 * not be specified in the context of RFC3542. Conversely, 2517 * RFC3542 types should not be specified in the context of RFC2292. 2518 */ 2519 if (!cmsg) { 2520 switch (optname) { 2521 case IPV6_2292PKTINFO: 2522 case IPV6_2292HOPLIMIT: 2523 case IPV6_2292NEXTHOP: 2524 case IPV6_2292HOPOPTS: 2525 case IPV6_2292DSTOPTS: 2526 case IPV6_2292RTHDR: 2527 case IPV6_2292PKTOPTIONS: 2528 return (ENOPROTOOPT); 2529 } 2530 } 2531 if (sticky && cmsg) { 2532 switch (optname) { 2533 case IPV6_PKTINFO: 2534 case IPV6_HOPLIMIT: 2535 case IPV6_NEXTHOP: 2536 case IPV6_HOPOPTS: 2537 case IPV6_DSTOPTS: 2538 case IPV6_RTHDRDSTOPTS: 2539 case IPV6_RTHDR: 2540 case IPV6_USE_MIN_MTU: 2541 case IPV6_DONTFRAG: 2542 case IPV6_TCLASS: 2543 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */ 2544 return (ENOPROTOOPT); 2545 } 2546 } 2547 2548 switch (optname) { 2549 case IPV6_2292PKTINFO: 2550 case IPV6_PKTINFO: 2551 { 2552 struct ifnet *ifp = NULL; 2553 struct in6_pktinfo *pktinfo; 2554 2555 if (len != sizeof(struct in6_pktinfo)) 2556 return (EINVAL); 2557 2558 pktinfo = (struct in6_pktinfo *)buf; 2559 2560 /* 2561 * An application can clear any sticky IPV6_PKTINFO option by 2562 * doing a "regular" setsockopt with ipi6_addr being 2563 * in6addr_any and ipi6_ifindex being zero. 2564 * [RFC 3542, Section 6] 2565 */ 2566 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2567 pktinfo->ipi6_ifindex == 0 && 2568 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2569 ip6_clearpktopts(opt, optname); 2570 break; 2571 } 2572 2573 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2574 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2575 return (EINVAL); 2576 } 2577 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr)) 2578 return (EINVAL); 2579 /* validate the interface index if specified. */ 2580 if (pktinfo->ipi6_ifindex > V_if_index) 2581 return (ENXIO); 2582 if (pktinfo->ipi6_ifindex) { 2583 ifp = ifnet_byindex(pktinfo->ipi6_ifindex); 2584 if (ifp == NULL) 2585 return (ENXIO); 2586 } 2587 if (ifp != NULL && ( 2588 ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) 2589 return (ENETDOWN); 2590 2591 if (ifp != NULL && 2592 !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2593 struct in6_ifaddr *ia; 2594 2595 in6_setscope(&pktinfo->ipi6_addr, ifp, NULL); 2596 ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr); 2597 if (ia == NULL) 2598 return (EADDRNOTAVAIL); 2599 ifa_free(&ia->ia_ifa); 2600 } 2601 /* 2602 * We store the address anyway, and let in6_selectsrc() 2603 * validate the specified address. This is because ipi6_addr 2604 * may not have enough information about its scope zone, and 2605 * we may need additional information (such as outgoing 2606 * interface or the scope zone of a destination address) to 2607 * disambiguate the scope. 2608 * XXX: the delay of the validation may confuse the 2609 * application when it is used as a sticky option. 2610 */ 2611 if (opt->ip6po_pktinfo == NULL) { 2612 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2613 M_IP6OPT, M_NOWAIT); 2614 if (opt->ip6po_pktinfo == NULL) 2615 return (ENOBUFS); 2616 } 2617 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo)); 2618 break; 2619 } 2620 2621 case IPV6_2292HOPLIMIT: 2622 case IPV6_HOPLIMIT: 2623 { 2624 int *hlimp; 2625 2626 /* 2627 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2628 * to simplify the ordering among hoplimit options. 2629 */ 2630 if (optname == IPV6_HOPLIMIT && sticky) 2631 return (ENOPROTOOPT); 2632 2633 if (len != sizeof(int)) 2634 return (EINVAL); 2635 hlimp = (int *)buf; 2636 if (*hlimp < -1 || *hlimp > 255) 2637 return (EINVAL); 2638 2639 opt->ip6po_hlim = *hlimp; 2640 break; 2641 } 2642 2643 case IPV6_TCLASS: 2644 { 2645 int tclass; 2646 2647 if (len != sizeof(int)) 2648 return (EINVAL); 2649 tclass = *(int *)buf; 2650 if (tclass < -1 || tclass > 255) 2651 return (EINVAL); 2652 2653 opt->ip6po_tclass = tclass; 2654 break; 2655 } 2656 2657 case IPV6_2292NEXTHOP: 2658 case IPV6_NEXTHOP: 2659 if (cred != NULL) { 2660 error = priv_check_cred(cred, 2661 PRIV_NETINET_SETHDROPTS, 0); 2662 if (error) 2663 return (error); 2664 } 2665 2666 if (len == 0) { /* just remove the option */ 2667 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2668 break; 2669 } 2670 2671 /* check if cmsg_len is large enough for sa_len */ 2672 if (len < sizeof(struct sockaddr) || len < *buf) 2673 return (EINVAL); 2674 2675 switch (((struct sockaddr *)buf)->sa_family) { 2676 case AF_INET6: 2677 { 2678 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 2679 int error; 2680 2681 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 2682 return (EINVAL); 2683 2684 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 2685 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 2686 return (EINVAL); 2687 } 2688 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone)) 2689 != 0) { 2690 return (error); 2691 } 2692 break; 2693 } 2694 case AF_LINK: /* should eventually be supported */ 2695 default: 2696 return (EAFNOSUPPORT); 2697 } 2698 2699 /* turn off the previous option, then set the new option. */ 2700 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2701 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 2702 if (opt->ip6po_nexthop == NULL) 2703 return (ENOBUFS); 2704 bcopy(buf, opt->ip6po_nexthop, *buf); 2705 break; 2706 2707 case IPV6_2292HOPOPTS: 2708 case IPV6_HOPOPTS: 2709 { 2710 struct ip6_hbh *hbh; 2711 int hbhlen; 2712 2713 /* 2714 * XXX: We don't allow a non-privileged user to set ANY HbH 2715 * options, since per-option restriction has too much 2716 * overhead. 2717 */ 2718 if (cred != NULL) { 2719 error = priv_check_cred(cred, 2720 PRIV_NETINET_SETHDROPTS, 0); 2721 if (error) 2722 return (error); 2723 } 2724 2725 if (len == 0) { 2726 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2727 break; /* just remove the option */ 2728 } 2729 2730 /* message length validation */ 2731 if (len < sizeof(struct ip6_hbh)) 2732 return (EINVAL); 2733 hbh = (struct ip6_hbh *)buf; 2734 hbhlen = (hbh->ip6h_len + 1) << 3; 2735 if (len != hbhlen) 2736 return (EINVAL); 2737 2738 /* turn off the previous option, then set the new option. */ 2739 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2740 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 2741 if (opt->ip6po_hbh == NULL) 2742 return (ENOBUFS); 2743 bcopy(hbh, opt->ip6po_hbh, hbhlen); 2744 2745 break; 2746 } 2747 2748 case IPV6_2292DSTOPTS: 2749 case IPV6_DSTOPTS: 2750 case IPV6_RTHDRDSTOPTS: 2751 { 2752 struct ip6_dest *dest, **newdest = NULL; 2753 int destlen; 2754 2755 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */ 2756 error = priv_check_cred(cred, 2757 PRIV_NETINET_SETHDROPTS, 0); 2758 if (error) 2759 return (error); 2760 } 2761 2762 if (len == 0) { 2763 ip6_clearpktopts(opt, optname); 2764 break; /* just remove the option */ 2765 } 2766 2767 /* message length validation */ 2768 if (len < sizeof(struct ip6_dest)) 2769 return (EINVAL); 2770 dest = (struct ip6_dest *)buf; 2771 destlen = (dest->ip6d_len + 1) << 3; 2772 if (len != destlen) 2773 return (EINVAL); 2774 2775 /* 2776 * Determine the position that the destination options header 2777 * should be inserted; before or after the routing header. 2778 */ 2779 switch (optname) { 2780 case IPV6_2292DSTOPTS: 2781 /* 2782 * The old advacned API is ambiguous on this point. 2783 * Our approach is to determine the position based 2784 * according to the existence of a routing header. 2785 * Note, however, that this depends on the order of the 2786 * extension headers in the ancillary data; the 1st 2787 * part of the destination options header must appear 2788 * before the routing header in the ancillary data, 2789 * too. 2790 * RFC3542 solved the ambiguity by introducing 2791 * separate ancillary data or option types. 2792 */ 2793 if (opt->ip6po_rthdr == NULL) 2794 newdest = &opt->ip6po_dest1; 2795 else 2796 newdest = &opt->ip6po_dest2; 2797 break; 2798 case IPV6_RTHDRDSTOPTS: 2799 newdest = &opt->ip6po_dest1; 2800 break; 2801 case IPV6_DSTOPTS: 2802 newdest = &opt->ip6po_dest2; 2803 break; 2804 } 2805 2806 /* turn off the previous option, then set the new option. */ 2807 ip6_clearpktopts(opt, optname); 2808 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 2809 if (*newdest == NULL) 2810 return (ENOBUFS); 2811 bcopy(dest, *newdest, destlen); 2812 2813 break; 2814 } 2815 2816 case IPV6_2292RTHDR: 2817 case IPV6_RTHDR: 2818 { 2819 struct ip6_rthdr *rth; 2820 int rthlen; 2821 2822 if (len == 0) { 2823 ip6_clearpktopts(opt, IPV6_RTHDR); 2824 break; /* just remove the option */ 2825 } 2826 2827 /* message length validation */ 2828 if (len < sizeof(struct ip6_rthdr)) 2829 return (EINVAL); 2830 rth = (struct ip6_rthdr *)buf; 2831 rthlen = (rth->ip6r_len + 1) << 3; 2832 if (len != rthlen) 2833 return (EINVAL); 2834 2835 switch (rth->ip6r_type) { 2836 case IPV6_RTHDR_TYPE_0: 2837 if (rth->ip6r_len == 0) /* must contain one addr */ 2838 return (EINVAL); 2839 if (rth->ip6r_len % 2) /* length must be even */ 2840 return (EINVAL); 2841 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 2842 return (EINVAL); 2843 break; 2844 default: 2845 return (EINVAL); /* not supported */ 2846 } 2847 2848 /* turn off the previous option */ 2849 ip6_clearpktopts(opt, IPV6_RTHDR); 2850 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 2851 if (opt->ip6po_rthdr == NULL) 2852 return (ENOBUFS); 2853 bcopy(rth, opt->ip6po_rthdr, rthlen); 2854 2855 break; 2856 } 2857 2858 case IPV6_USE_MIN_MTU: 2859 if (len != sizeof(int)) 2860 return (EINVAL); 2861 minmtupolicy = *(int *)buf; 2862 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 2863 minmtupolicy != IP6PO_MINMTU_DISABLE && 2864 minmtupolicy != IP6PO_MINMTU_ALL) { 2865 return (EINVAL); 2866 } 2867 opt->ip6po_minmtu = minmtupolicy; 2868 break; 2869 2870 case IPV6_DONTFRAG: 2871 if (len != sizeof(int)) 2872 return (EINVAL); 2873 2874 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 2875 /* 2876 * we ignore this option for TCP sockets. 2877 * (RFC3542 leaves this case unspecified.) 2878 */ 2879 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 2880 } else 2881 opt->ip6po_flags |= IP6PO_DONTFRAG; 2882 break; 2883 2884 case IPV6_PREFER_TEMPADDR: 2885 if (len != sizeof(int)) 2886 return (EINVAL); 2887 preftemp = *(int *)buf; 2888 if (preftemp != IP6PO_TEMPADDR_SYSTEM && 2889 preftemp != IP6PO_TEMPADDR_NOTPREFER && 2890 preftemp != IP6PO_TEMPADDR_PREFER) { 2891 return (EINVAL); 2892 } 2893 opt->ip6po_prefer_tempaddr = preftemp; 2894 break; 2895 2896 default: 2897 return (ENOPROTOOPT); 2898 } /* end of switch */ 2899 2900 return (0); 2901 } 2902 2903 /* 2904 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 2905 * packet to the input queue of a specified interface. Note that this 2906 * calls the output routine of the loopback "driver", but with an interface 2907 * pointer that might NOT be &loif -- easier than replicating that code here. 2908 */ 2909 void 2910 ip6_mloopback(struct ifnet *ifp, const struct mbuf *m) 2911 { 2912 struct mbuf *copym; 2913 struct ip6_hdr *ip6; 2914 2915 copym = m_copy(m, 0, M_COPYALL); 2916 if (copym == NULL) 2917 return; 2918 2919 /* 2920 * Make sure to deep-copy IPv6 header portion in case the data 2921 * is in an mbuf cluster, so that we can safely override the IPv6 2922 * header portion later. 2923 */ 2924 if (!M_WRITABLE(copym) || 2925 copym->m_len < sizeof(struct ip6_hdr)) { 2926 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 2927 if (copym == NULL) 2928 return; 2929 } 2930 ip6 = mtod(copym, struct ip6_hdr *); 2931 /* 2932 * clear embedded scope identifiers if necessary. 2933 * in6_clearscope will touch the addresses only when necessary. 2934 */ 2935 in6_clearscope(&ip6->ip6_src); 2936 in6_clearscope(&ip6->ip6_dst); 2937 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 2938 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 | 2939 CSUM_PSEUDO_HDR; 2940 copym->m_pkthdr.csum_data = 0xffff; 2941 } 2942 if_simloop(ifp, copym, AF_INET6, 0); 2943 } 2944 2945 /* 2946 * Chop IPv6 header off from the payload. 2947 */ 2948 static int 2949 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 2950 { 2951 struct mbuf *mh; 2952 struct ip6_hdr *ip6; 2953 2954 ip6 = mtod(m, struct ip6_hdr *); 2955 if (m->m_len > sizeof(*ip6)) { 2956 mh = m_gethdr(M_NOWAIT, MT_DATA); 2957 if (mh == NULL) { 2958 m_freem(m); 2959 return ENOBUFS; 2960 } 2961 m_move_pkthdr(mh, m); 2962 M_ALIGN(mh, sizeof(*ip6)); 2963 m->m_len -= sizeof(*ip6); 2964 m->m_data += sizeof(*ip6); 2965 mh->m_next = m; 2966 m = mh; 2967 m->m_len = sizeof(*ip6); 2968 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 2969 } 2970 exthdrs->ip6e_ip6 = m; 2971 return 0; 2972 } 2973 2974 /* 2975 * Compute IPv6 extension header length. 2976 */ 2977 int 2978 ip6_optlen(struct inpcb *in6p) 2979 { 2980 int len; 2981 2982 if (!in6p->in6p_outputopts) 2983 return 0; 2984 2985 len = 0; 2986 #define elen(x) \ 2987 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 2988 2989 len += elen(in6p->in6p_outputopts->ip6po_hbh); 2990 if (in6p->in6p_outputopts->ip6po_rthdr) 2991 /* dest1 is valid with rthdr only */ 2992 len += elen(in6p->in6p_outputopts->ip6po_dest1); 2993 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 2994 len += elen(in6p->in6p_outputopts->ip6po_dest2); 2995 return len; 2996 #undef elen 2997 } 2998