1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $ 32 */ 33 34 /*- 35 * Copyright (c) 1982, 1986, 1988, 1990, 1993 36 * The Regents of the University of California. All rights reserved. 37 * 38 * Redistribution and use in source and binary forms, with or without 39 * modification, are permitted provided that the following conditions 40 * are met: 41 * 1. Redistributions of source code must retain the above copyright 42 * notice, this list of conditions and the following disclaimer. 43 * 2. Redistributions in binary form must reproduce the above copyright 44 * notice, this list of conditions and the following disclaimer in the 45 * documentation and/or other materials provided with the distribution. 46 * 3. Neither the name of the University nor the names of its contributors 47 * may be used to endorse or promote products derived from this software 48 * without specific prior written permission. 49 * 50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 60 * SUCH DAMAGE. 61 * 62 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 63 */ 64 65 #include <sys/cdefs.h> 66 __FBSDID("$FreeBSD$"); 67 68 #include "opt_inet.h" 69 #include "opt_inet6.h" 70 #include "opt_ipsec.h" 71 #include "opt_kern_tls.h" 72 #include "opt_ratelimit.h" 73 #include "opt_route.h" 74 #include "opt_rss.h" 75 #include "opt_sctp.h" 76 77 #include <sys/param.h> 78 #include <sys/kernel.h> 79 #include <sys/ktls.h> 80 #include <sys/malloc.h> 81 #include <sys/mbuf.h> 82 #include <sys/errno.h> 83 #include <sys/priv.h> 84 #include <sys/proc.h> 85 #include <sys/protosw.h> 86 #include <sys/socket.h> 87 #include <sys/socketvar.h> 88 #include <sys/syslog.h> 89 #include <sys/ucred.h> 90 91 #include <machine/in_cksum.h> 92 93 #include <net/if.h> 94 #include <net/if_var.h> 95 #include <net/if_llatbl.h> 96 #include <net/netisr.h> 97 #include <net/route.h> 98 #include <net/route/nhop.h> 99 #include <net/pfil.h> 100 #include <net/rss_config.h> 101 #include <net/vnet.h> 102 103 #include <netinet/in.h> 104 #include <netinet/in_var.h> 105 #include <netinet/ip_var.h> 106 #include <netinet6/in6_fib.h> 107 #include <netinet6/in6_var.h> 108 #include <netinet/ip6.h> 109 #include <netinet/icmp6.h> 110 #include <netinet6/ip6_var.h> 111 #include <netinet/in_pcb.h> 112 #include <netinet/tcp_var.h> 113 #include <netinet6/nd6.h> 114 #include <netinet6/in6_rss.h> 115 116 #include <netipsec/ipsec_support.h> 117 #if defined(SCTP) || defined(SCTP_SUPPORT) 118 #include <netinet/sctp.h> 119 #include <netinet/sctp_crc32.h> 120 #endif 121 122 #include <netinet6/ip6protosw.h> 123 #include <netinet6/scope6_var.h> 124 125 extern int in6_mcast_loop; 126 127 struct ip6_exthdrs { 128 struct mbuf *ip6e_ip6; 129 struct mbuf *ip6e_hbh; 130 struct mbuf *ip6e_dest1; 131 struct mbuf *ip6e_rthdr; 132 struct mbuf *ip6e_dest2; 133 }; 134 135 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options"); 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 inpcb *, 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 *, int, 151 struct ifnet *, const struct in6_addr *, u_long *, int *, u_int, 152 u_int); 153 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long, 154 u_long *, int *, u_int); 155 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *); 156 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int); 157 158 /* 159 * Make an extension header from option data. hp is the source, 160 * mp is the destination, and _ol is the optlen. 161 */ 162 #define MAKE_EXTHDR(hp, mp, _ol) \ 163 do { \ 164 if (hp) { \ 165 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 166 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 167 ((eh)->ip6e_len + 1) << 3); \ 168 if (error) \ 169 goto freehdrs; \ 170 (_ol) += (*(mp))->m_len; \ 171 } \ 172 } while (/*CONSTCOND*/ 0) 173 174 /* 175 * Form a chain of extension headers. 176 * m is the extension header mbuf 177 * mp is the previous mbuf in the chain 178 * p is the next header 179 * i is the type of option. 180 */ 181 #define MAKE_CHAIN(m, mp, p, i)\ 182 do {\ 183 if (m) {\ 184 if (!hdrsplit) \ 185 panic("%s:%d: assumption failed: "\ 186 "hdr not split: hdrsplit %d exthdrs %p",\ 187 __func__, __LINE__, hdrsplit, &exthdrs);\ 188 *mtod((m), u_char *) = *(p);\ 189 *(p) = (i);\ 190 p = mtod((m), u_char *);\ 191 (m)->m_next = (mp)->m_next;\ 192 (mp)->m_next = (m);\ 193 (mp) = (m);\ 194 }\ 195 } while (/*CONSTCOND*/ 0) 196 197 void 198 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset) 199 { 200 u_short csum; 201 202 csum = in_cksum_skip(m, offset + plen, offset); 203 if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0) 204 csum = 0xffff; 205 offset += m->m_pkthdr.csum_data; /* checksum offset */ 206 207 if (offset + sizeof(csum) > m->m_len) 208 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum); 209 else 210 *(u_short *)mtodo(m, offset) = csum; 211 } 212 213 static int 214 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags, 215 int plen, int optlen, bool frag) 216 { 217 218 KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p " 219 "csum_flags %#x frag %d\n", 220 __func__, __LINE__, plen, optlen, m, ifp, csum_flags, frag)); 221 222 if ((csum_flags & CSUM_DELAY_DATA_IPV6) || 223 #if defined(SCTP) || defined(SCTP_SUPPORT) 224 (csum_flags & CSUM_SCTP_IPV6) || 225 #endif 226 (!frag && (ifp->if_capenable & IFCAP_NOMAP) == 0)) { 227 m = mb_unmapped_to_ext(m); 228 if (m == NULL) { 229 if (frag) 230 in6_ifstat_inc(ifp, ifs6_out_fragfail); 231 else 232 IP6STAT_INC(ip6s_odropped); 233 return (ENOBUFS); 234 } 235 if (csum_flags & CSUM_DELAY_DATA_IPV6) { 236 in6_delayed_cksum(m, plen - optlen, 237 sizeof(struct ip6_hdr) + optlen); 238 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6; 239 } 240 #if defined(SCTP) || defined(SCTP_SUPPORT) 241 if (csum_flags & CSUM_SCTP_IPV6) { 242 sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen); 243 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6; 244 } 245 #endif 246 } 247 248 return (0); 249 } 250 251 int 252 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto, 253 int fraglen , uint32_t id) 254 { 255 struct mbuf *m, **mnext, *m_frgpart; 256 struct ip6_hdr *ip6, *mhip6; 257 struct ip6_frag *ip6f; 258 int off; 259 int error; 260 int tlen = m0->m_pkthdr.len; 261 262 KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8")); 263 264 m = m0; 265 ip6 = mtod(m, struct ip6_hdr *); 266 mnext = &m->m_nextpkt; 267 268 for (off = hlen; off < tlen; off += fraglen) { 269 m = m_gethdr(M_NOWAIT, MT_DATA); 270 if (!m) { 271 IP6STAT_INC(ip6s_odropped); 272 return (ENOBUFS); 273 } 274 275 /* 276 * Make sure the complete packet header gets copied 277 * from the originating mbuf to the newly created 278 * mbuf. This also ensures that existing firewall 279 * classification(s), VLAN tags and so on get copied 280 * to the resulting fragmented packet(s): 281 */ 282 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) { 283 m_free(m); 284 IP6STAT_INC(ip6s_odropped); 285 return (ENOBUFS); 286 } 287 288 *mnext = m; 289 mnext = &m->m_nextpkt; 290 m->m_data += max_linkhdr; 291 mhip6 = mtod(m, struct ip6_hdr *); 292 *mhip6 = *ip6; 293 m->m_len = sizeof(*mhip6); 294 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 295 if (error) { 296 IP6STAT_INC(ip6s_odropped); 297 return (error); 298 } 299 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7)); 300 if (off + fraglen >= tlen) 301 fraglen = tlen - off; 302 else 303 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 304 mhip6->ip6_plen = htons((u_short)(fraglen + hlen + 305 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 306 if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) { 307 IP6STAT_INC(ip6s_odropped); 308 return (ENOBUFS); 309 } 310 m_cat(m, m_frgpart); 311 m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f); 312 ip6f->ip6f_reserved = 0; 313 ip6f->ip6f_ident = id; 314 ip6f->ip6f_nxt = nextproto; 315 IP6STAT_INC(ip6s_ofragments); 316 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 317 } 318 319 return (0); 320 } 321 322 static int 323 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp, 324 struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro, 325 bool stamp_tag) 326 { 327 #ifdef KERN_TLS 328 struct ktls_session *tls = NULL; 329 #endif 330 struct m_snd_tag *mst; 331 int error; 332 333 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0); 334 mst = NULL; 335 336 #ifdef KERN_TLS 337 /* 338 * If this is an unencrypted TLS record, save a reference to 339 * the record. This local reference is used to call 340 * ktls_output_eagain after the mbuf has been freed (thus 341 * dropping the mbuf's reference) in if_output. 342 */ 343 if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) { 344 tls = ktls_hold(m->m_next->m_epg_tls); 345 mst = tls->snd_tag; 346 347 /* 348 * If a TLS session doesn't have a valid tag, it must 349 * have had an earlier ifp mismatch, so drop this 350 * packet. 351 */ 352 if (mst == NULL) { 353 error = EAGAIN; 354 goto done; 355 } 356 /* 357 * Always stamp tags that include NIC ktls. 358 */ 359 stamp_tag = true; 360 } 361 #endif 362 #ifdef RATELIMIT 363 if (inp != NULL && mst == NULL) { 364 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 || 365 (inp->inp_snd_tag != NULL && 366 inp->inp_snd_tag->ifp != ifp)) 367 in_pcboutput_txrtlmt(inp, ifp, m); 368 369 if (inp->inp_snd_tag != NULL) 370 mst = inp->inp_snd_tag; 371 } 372 #endif 373 if (stamp_tag && mst != NULL) { 374 KASSERT(m->m_pkthdr.rcvif == NULL, 375 ("trying to add a send tag to a forwarded packet")); 376 if (mst->ifp != ifp) { 377 error = EAGAIN; 378 goto done; 379 } 380 381 /* stamp send tag on mbuf */ 382 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst); 383 m->m_pkthdr.csum_flags |= CSUM_SND_TAG; 384 } 385 386 error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro); 387 388 done: 389 /* Check for route change invalidating send tags. */ 390 #ifdef KERN_TLS 391 if (tls != NULL) { 392 if (error == EAGAIN) 393 error = ktls_output_eagain(inp, tls); 394 ktls_free(tls); 395 } 396 #endif 397 #ifdef RATELIMIT 398 if (error == EAGAIN) 399 in_pcboutput_eagain(inp); 400 #endif 401 return (error); 402 } 403 404 /* 405 * IP6 output. 406 * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len, 407 * nxt, hlim, src, dst). 408 * This function may modify ver and hlim only. 409 * The mbuf chain containing the packet will be freed. 410 * The mbuf opt, if present, will not be freed. 411 * If route_in6 ro is present and has ro_nh initialized, route lookup would be 412 * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL, 413 * then result of route lookup is stored in ro->ro_nh. 414 * 415 * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu 416 * is uint32_t. So we use u_long to hold largest one, which is rt_mtu. 417 * 418 * ifpp - XXX: just for statistics 419 */ 420 /* 421 * XXX TODO: no flowid is assigned for outbound flows? 422 */ 423 int 424 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, 425 struct route_in6 *ro, int flags, struct ip6_moptions *im6o, 426 struct ifnet **ifpp, struct inpcb *inp) 427 { 428 struct ip6_hdr *ip6; 429 struct ifnet *ifp, *origifp; 430 struct mbuf *m = m0; 431 struct mbuf *mprev; 432 struct route_in6 *ro_pmtu; 433 struct nhop_object *nh; 434 struct sockaddr_in6 *dst, sin6, src_sa, dst_sa; 435 struct in6_addr odst; 436 u_char *nexthdrp; 437 int tlen, len; 438 int error = 0; 439 struct in6_ifaddr *ia = NULL; 440 u_long mtu; 441 int alwaysfrag, dontfrag; 442 u_int32_t optlen, plen = 0, unfragpartlen; 443 struct ip6_exthdrs exthdrs; 444 struct in6_addr src0, dst0; 445 u_int32_t zone; 446 bool hdrsplit; 447 int sw_csum, tso; 448 int needfiblookup; 449 uint32_t fibnum; 450 struct m_tag *fwd_tag = NULL; 451 uint32_t id; 452 453 NET_EPOCH_ASSERT(); 454 455 if (inp != NULL) { 456 INP_LOCK_ASSERT(inp); 457 M_SETFIB(m, inp->inp_inc.inc_fibnum); 458 if ((flags & IP_NODEFAULTFLOWID) == 0) { 459 /* Unconditionally set flowid. */ 460 m->m_pkthdr.flowid = inp->inp_flowid; 461 M_HASHTYPE_SET(m, inp->inp_flowtype); 462 } 463 #ifdef NUMA 464 m->m_pkthdr.numa_domain = inp->inp_numa_domain; 465 #endif 466 } 467 468 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 469 /* 470 * IPSec checking which handles several cases. 471 * FAST IPSEC: We re-injected the packet. 472 * XXX: need scope argument. 473 */ 474 if (IPSEC_ENABLED(ipv6)) { 475 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) { 476 if (error == EINPROGRESS) 477 error = 0; 478 goto done; 479 } 480 } 481 #endif /* IPSEC */ 482 483 /* Source address validation. */ 484 ip6 = mtod(m, struct ip6_hdr *); 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 /* 498 * If we are given packet options to add extension headers prepare them. 499 * Calculate the total length of the extension header chain. 500 * Keep the length of the unfragmentable part for fragmentation. 501 */ 502 bzero(&exthdrs, sizeof(exthdrs)); 503 optlen = 0; 504 unfragpartlen = sizeof(struct ip6_hdr); 505 if (opt) { 506 /* Hop-by-Hop options header. */ 507 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen); 508 509 /* Destination options header (1st part). */ 510 if (opt->ip6po_rthdr) { 511 #ifndef RTHDR_SUPPORT_IMPLEMENTED 512 /* 513 * If there is a routing header, discard the packet 514 * right away here. RH0/1 are obsolete and we do not 515 * currently support RH2/3/4. 516 * People trying to use RH253/254 may want to disable 517 * this check. 518 * The moment we do support any routing header (again) 519 * this block should check the routing type more 520 * selectively. 521 */ 522 error = EINVAL; 523 goto bad; 524 #endif 525 526 /* 527 * Destination options header (1st part). 528 * This only makes sense with a routing header. 529 * See Section 9.2 of RFC 3542. 530 * Disabling this part just for MIP6 convenience is 531 * a bad idea. We need to think carefully about a 532 * way to make the advanced API coexist with MIP6 533 * options, which might automatically be inserted in 534 * the kernel. 535 */ 536 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1, 537 optlen); 538 } 539 /* Routing header. */ 540 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen); 541 542 unfragpartlen += optlen; 543 544 /* 545 * NOTE: we don't add AH/ESP length here (done in 546 * ip6_ipsec_output()). 547 */ 548 549 /* Destination options header (2nd part). */ 550 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen); 551 } 552 553 /* 554 * If there is at least one extension header, 555 * separate IP6 header from the payload. 556 */ 557 hdrsplit = false; 558 if (optlen) { 559 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 560 m = NULL; 561 goto freehdrs; 562 } 563 m = exthdrs.ip6e_ip6; 564 ip6 = mtod(m, struct ip6_hdr *); 565 hdrsplit = true; 566 } 567 568 /* Adjust mbuf packet header length. */ 569 m->m_pkthdr.len += optlen; 570 plen = m->m_pkthdr.len - sizeof(*ip6); 571 572 /* If this is a jumbo payload, insert a jumbo payload option. */ 573 if (plen > IPV6_MAXPACKET) { 574 if (!hdrsplit) { 575 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 576 m = NULL; 577 goto freehdrs; 578 } 579 m = exthdrs.ip6e_ip6; 580 ip6 = mtod(m, struct ip6_hdr *); 581 hdrsplit = true; 582 } 583 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 584 goto freehdrs; 585 ip6->ip6_plen = 0; 586 } else 587 ip6->ip6_plen = htons(plen); 588 nexthdrp = &ip6->ip6_nxt; 589 590 if (optlen) { 591 /* 592 * Concatenate headers and fill in next header fields. 593 * Here we have, on "m" 594 * IPv6 payload 595 * and we insert headers accordingly. 596 * Finally, we should be getting: 597 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]. 598 * 599 * During the header composing process "m" points to IPv6 600 * header. "mprev" points to an extension header prior to esp. 601 */ 602 mprev = m; 603 604 /* 605 * We treat dest2 specially. This makes IPsec processing 606 * much easier. The goal here is to make mprev point the 607 * mbuf prior to dest2. 608 * 609 * Result: IPv6 dest2 payload. 610 * m and mprev will point to IPv6 header. 611 */ 612 if (exthdrs.ip6e_dest2) { 613 if (!hdrsplit) 614 panic("%s:%d: assumption failed: " 615 "hdr not split: hdrsplit %d exthdrs %p", 616 __func__, __LINE__, hdrsplit, &exthdrs); 617 exthdrs.ip6e_dest2->m_next = m->m_next; 618 m->m_next = exthdrs.ip6e_dest2; 619 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 620 ip6->ip6_nxt = IPPROTO_DSTOPTS; 621 } 622 623 /* 624 * Result: IPv6 hbh dest1 rthdr dest2 payload. 625 * m will point to IPv6 header. mprev will point to the 626 * extension header prior to dest2 (rthdr in the above case). 627 */ 628 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 629 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 630 IPPROTO_DSTOPTS); 631 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 632 IPPROTO_ROUTING); 633 } 634 635 IP6STAT_INC(ip6s_localout); 636 637 /* Route packet. */ 638 ro_pmtu = ro; 639 if (opt && opt->ip6po_rthdr) 640 ro = &opt->ip6po_route; 641 if (ro != NULL) 642 dst = (struct sockaddr_in6 *)&ro->ro_dst; 643 else 644 dst = &sin6; 645 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m); 646 647 again: 648 /* 649 * If specified, try to fill in the traffic class field. 650 * Do not override if a non-zero value is already set. 651 * We check the diffserv field and the ECN field separately. 652 */ 653 if (opt && opt->ip6po_tclass >= 0) { 654 int mask = 0; 655 656 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 657 mask |= 0xfc; 658 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 659 mask |= 0x03; 660 if (mask != 0) 661 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 662 } 663 664 /* Fill in or override the hop limit field, if necessary. */ 665 if (opt && opt->ip6po_hlim != -1) 666 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 667 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 668 if (im6o != NULL) 669 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 670 else 671 ip6->ip6_hlim = V_ip6_defmcasthlim; 672 } 673 674 if (ro == NULL || ro->ro_nh == NULL) { 675 bzero(dst, sizeof(*dst)); 676 dst->sin6_family = AF_INET6; 677 dst->sin6_len = sizeof(*dst); 678 dst->sin6_addr = ip6->ip6_dst; 679 } 680 /* 681 * Validate route against routing table changes. 682 * Make sure that the address family is set in route. 683 */ 684 nh = NULL; 685 ifp = NULL; 686 mtu = 0; 687 if (ro != NULL) { 688 if (ro->ro_nh != NULL && inp != NULL) { 689 ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */ 690 NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, 691 fibnum); 692 } 693 if (ro->ro_nh != NULL && fwd_tag == NULL && 694 (!NH_IS_VALID(ro->ro_nh) || 695 ro->ro_dst.sin6_family != AF_INET6 || 696 !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst))) 697 RO_INVALIDATE_CACHE(ro); 698 699 if (ro->ro_nh != NULL && fwd_tag == NULL && 700 ro->ro_dst.sin6_family == AF_INET6 && 701 IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) { 702 nh = ro->ro_nh; 703 ifp = nh->nh_ifp; 704 } else { 705 if (ro->ro_lle) 706 LLE_FREE(ro->ro_lle); /* zeros ro_lle */ 707 ro->ro_lle = NULL; 708 if (fwd_tag == NULL) { 709 bzero(&dst_sa, sizeof(dst_sa)); 710 dst_sa.sin6_family = AF_INET6; 711 dst_sa.sin6_len = sizeof(dst_sa); 712 dst_sa.sin6_addr = ip6->ip6_dst; 713 } 714 error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp, 715 &nh, fibnum, m->m_pkthdr.flowid); 716 if (error != 0) { 717 IP6STAT_INC(ip6s_noroute); 718 if (ifp != NULL) 719 in6_ifstat_inc(ifp, ifs6_out_discard); 720 goto bad; 721 } 722 if (ifp != NULL) 723 mtu = ifp->if_mtu; 724 } 725 if (nh == NULL) { 726 /* 727 * If in6_selectroute() does not return a nexthop 728 * dst may not have been updated. 729 */ 730 *dst = dst_sa; /* XXX */ 731 } else { 732 if (nh->nh_flags & NHF_HOST) 733 mtu = nh->nh_mtu; 734 ia = (struct in6_ifaddr *)(nh->nh_ifa); 735 counter_u64_add(nh->nh_pksent, 1); 736 } 737 } else { 738 struct nhop_object *nh; 739 struct in6_addr kdst; 740 uint32_t scopeid; 741 742 if (fwd_tag == NULL) { 743 bzero(&dst_sa, sizeof(dst_sa)); 744 dst_sa.sin6_family = AF_INET6; 745 dst_sa.sin6_len = sizeof(dst_sa); 746 dst_sa.sin6_addr = ip6->ip6_dst; 747 } 748 749 if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) && 750 im6o != NULL && 751 (ifp = im6o->im6o_multicast_ifp) != NULL) { 752 /* We do not need a route lookup. */ 753 *dst = dst_sa; /* XXX */ 754 goto nonh6lookup; 755 } 756 757 in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid); 758 759 if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) || 760 IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) { 761 if (scopeid > 0) { 762 ifp = in6_getlinkifnet(scopeid); 763 if (ifp == NULL) { 764 error = EHOSTUNREACH; 765 goto bad; 766 } 767 *dst = dst_sa; /* XXX */ 768 goto nonh6lookup; 769 } 770 } 771 772 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0); 773 if (nh == NULL) { 774 IP6STAT_INC(ip6s_noroute); 775 /* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */ 776 error = EHOSTUNREACH;; 777 goto bad; 778 } 779 780 ifp = nh->nh_ifp; 781 mtu = nh->nh_mtu; 782 ia = ifatoia6(nh->nh_ifa); 783 if (nh->nh_flags & NHF_GATEWAY) 784 dst->sin6_addr = nh->gw6_sa.sin6_addr; 785 nonh6lookup: 786 ; 787 } 788 789 /* Then nh (for unicast) and ifp must be non-NULL valid values. */ 790 if ((flags & IPV6_FORWARDING) == 0) { 791 /* XXX: the FORWARDING flag can be set for mrouting. */ 792 in6_ifstat_inc(ifp, ifs6_out_request); 793 } 794 795 /* Setup data structures for scope ID checks. */ 796 src0 = ip6->ip6_src; 797 bzero(&src_sa, sizeof(src_sa)); 798 src_sa.sin6_family = AF_INET6; 799 src_sa.sin6_len = sizeof(src_sa); 800 src_sa.sin6_addr = ip6->ip6_src; 801 802 dst0 = ip6->ip6_dst; 803 /* Re-initialize to be sure. */ 804 bzero(&dst_sa, sizeof(dst_sa)); 805 dst_sa.sin6_family = AF_INET6; 806 dst_sa.sin6_len = sizeof(dst_sa); 807 dst_sa.sin6_addr = ip6->ip6_dst; 808 809 /* Check for valid scope ID. */ 810 if (in6_setscope(&src0, ifp, &zone) == 0 && 811 sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id && 812 in6_setscope(&dst0, ifp, &zone) == 0 && 813 sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) { 814 /* 815 * The outgoing interface is in the zone of the source 816 * and destination addresses. 817 * 818 * Because the loopback interface cannot receive 819 * packets with a different scope ID than its own, 820 * there is a trick to pretend the outgoing packet 821 * was received by the real network interface, by 822 * setting "origifp" different from "ifp". This is 823 * only allowed when "ifp" is a loopback network 824 * interface. Refer to code in nd6_output_ifp() for 825 * more details. 826 */ 827 origifp = ifp; 828 829 /* 830 * We should use ia_ifp to support the case of sending 831 * packets to an address of our own. 832 */ 833 if (ia != NULL && ia->ia_ifp) 834 ifp = ia->ia_ifp; 835 836 } else if ((ifp->if_flags & IFF_LOOPBACK) == 0 || 837 sa6_recoverscope(&src_sa) != 0 || 838 sa6_recoverscope(&dst_sa) != 0 || 839 dst_sa.sin6_scope_id == 0 || 840 (src_sa.sin6_scope_id != 0 && 841 src_sa.sin6_scope_id != dst_sa.sin6_scope_id) || 842 (origifp = ifnet_byindex(dst_sa.sin6_scope_id)) == NULL) { 843 /* 844 * If the destination network interface is not a 845 * loopback interface, or the destination network 846 * address has no scope ID, or the source address has 847 * a scope ID set which is different from the 848 * destination address one, or there is no network 849 * interface representing this scope ID, the address 850 * pair is considered invalid. 851 */ 852 IP6STAT_INC(ip6s_badscope); 853 in6_ifstat_inc(ifp, ifs6_out_discard); 854 if (error == 0) 855 error = EHOSTUNREACH; /* XXX */ 856 goto bad; 857 } 858 /* All scope ID checks are successful. */ 859 860 if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 861 if (opt && opt->ip6po_nextroute.ro_nh) { 862 /* 863 * The nexthop is explicitly specified by the 864 * application. We assume the next hop is an IPv6 865 * address. 866 */ 867 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 868 } 869 else if ((nh->nh_flags & NHF_GATEWAY)) 870 dst = &nh->gw6_sa; 871 } 872 873 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 874 m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */ 875 } else { 876 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 877 in6_ifstat_inc(ifp, ifs6_out_mcast); 878 879 /* Confirm that the outgoing interface supports multicast. */ 880 if (!(ifp->if_flags & IFF_MULTICAST)) { 881 IP6STAT_INC(ip6s_noroute); 882 in6_ifstat_inc(ifp, ifs6_out_discard); 883 error = ENETUNREACH; 884 goto bad; 885 } 886 if ((im6o == NULL && in6_mcast_loop) || 887 (im6o && im6o->im6o_multicast_loop)) { 888 /* 889 * Loop back multicast datagram if not expressly 890 * forbidden to do so, even if we have not joined 891 * the address; protocols will filter it later, 892 * thus deferring a hash lookup and lock acquisition 893 * at the expense of an m_copym(). 894 */ 895 ip6_mloopback(ifp, m); 896 } else { 897 /* 898 * If we are acting as a multicast router, perform 899 * multicast forwarding as if the packet had just 900 * arrived on the interface to which we are about 901 * to send. The multicast forwarding function 902 * recursively calls this function, using the 903 * IPV6_FORWARDING flag to prevent infinite recursion. 904 * 905 * Multicasts that are looped back by ip6_mloopback(), 906 * above, will be forwarded by the ip6_input() routine, 907 * if necessary. 908 */ 909 if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 910 /* 911 * XXX: ip6_mforward expects that rcvif is NULL 912 * when it is called from the originating path. 913 * However, it may not always be the case. 914 */ 915 m->m_pkthdr.rcvif = NULL; 916 if (ip6_mforward(ip6, ifp, m) != 0) { 917 m_freem(m); 918 goto done; 919 } 920 } 921 } 922 /* 923 * Multicasts with a hoplimit of zero may be looped back, 924 * above, but must not be transmitted on a network. 925 * Also, multicasts addressed to the loopback interface 926 * are not sent -- the above call to ip6_mloopback() will 927 * loop back a copy if this host actually belongs to the 928 * destination group on the loopback interface. 929 */ 930 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 931 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 932 m_freem(m); 933 goto done; 934 } 935 } 936 937 /* 938 * Fill the outgoing inteface to tell the upper layer 939 * to increment per-interface statistics. 940 */ 941 if (ifpp) 942 *ifpp = ifp; 943 944 /* Determine path MTU. */ 945 if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst, 946 &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0) 947 goto bad; 948 KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p " 949 "alwaysfrag %d fibnum %u\n", __func__, __LINE__, mtu, ro_pmtu, ro, 950 ifp, alwaysfrag, fibnum)); 951 952 /* 953 * The caller of this function may specify to use the minimum MTU 954 * in some cases. 955 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 956 * setting. The logic is a bit complicated; by default, unicast 957 * packets will follow path MTU while multicast packets will be sent at 958 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 959 * including unicast ones will be sent at the minimum MTU. Multicast 960 * packets will always be sent at the minimum MTU unless 961 * IP6PO_MINMTU_DISABLE is explicitly specified. 962 * See RFC 3542 for more details. 963 */ 964 if (mtu > IPV6_MMTU) { 965 if ((flags & IPV6_MINMTU)) 966 mtu = IPV6_MMTU; 967 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 968 mtu = IPV6_MMTU; 969 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 970 (opt == NULL || 971 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 972 mtu = IPV6_MMTU; 973 } 974 } 975 976 /* 977 * Clear embedded scope identifiers if necessary. 978 * in6_clearscope() will touch the addresses only when necessary. 979 */ 980 in6_clearscope(&ip6->ip6_src); 981 in6_clearscope(&ip6->ip6_dst); 982 983 /* 984 * If the outgoing packet contains a hop-by-hop options header, 985 * it must be examined and processed even by the source node. 986 * (RFC 2460, section 4.) 987 */ 988 if (exthdrs.ip6e_hbh) { 989 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 990 u_int32_t dummy; /* XXX unused */ 991 u_int32_t plen = 0; /* XXX: ip6_process will check the value */ 992 993 #ifdef DIAGNOSTIC 994 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len) 995 panic("ip6e_hbh is not contiguous"); 996 #endif 997 /* 998 * XXX: if we have to send an ICMPv6 error to the sender, 999 * we need the M_LOOP flag since icmp6_error() expects 1000 * the IPv6 and the hop-by-hop options header are 1001 * contiguous unless the flag is set. 1002 */ 1003 m->m_flags |= M_LOOP; 1004 m->m_pkthdr.rcvif = ifp; 1005 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 1006 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 1007 &dummy, &plen) < 0) { 1008 /* m was already freed at this point. */ 1009 error = EINVAL;/* better error? */ 1010 goto done; 1011 } 1012 m->m_flags &= ~M_LOOP; /* XXX */ 1013 m->m_pkthdr.rcvif = NULL; 1014 } 1015 1016 /* Jump over all PFIL processing if hooks are not active. */ 1017 if (!PFIL_HOOKED_OUT(V_inet6_pfil_head)) 1018 goto passout; 1019 1020 odst = ip6->ip6_dst; 1021 /* Run through list of hooks for output packets. */ 1022 switch (pfil_run_hooks(V_inet6_pfil_head, &m, ifp, PFIL_OUT, inp)) { 1023 case PFIL_PASS: 1024 ip6 = mtod(m, struct ip6_hdr *); 1025 break; 1026 case PFIL_DROPPED: 1027 error = EACCES; 1028 /* FALLTHROUGH */ 1029 case PFIL_CONSUMED: 1030 goto done; 1031 } 1032 1033 needfiblookup = 0; 1034 /* See if destination IP address was changed by packet filter. */ 1035 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) { 1036 m->m_flags |= M_SKIP_FIREWALL; 1037 /* If destination is now ourself drop to ip6_input(). */ 1038 if (in6_localip(&ip6->ip6_dst)) { 1039 m->m_flags |= M_FASTFWD_OURS; 1040 if (m->m_pkthdr.rcvif == NULL) 1041 m->m_pkthdr.rcvif = V_loif; 1042 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 1043 m->m_pkthdr.csum_flags |= 1044 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; 1045 m->m_pkthdr.csum_data = 0xffff; 1046 } 1047 #if defined(SCTP) || defined(SCTP_SUPPORT) 1048 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) 1049 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 1050 #endif 1051 error = netisr_queue(NETISR_IPV6, m); 1052 goto done; 1053 } else { 1054 if (ro != NULL) 1055 RO_INVALIDATE_CACHE(ro); 1056 needfiblookup = 1; /* Redo the routing table lookup. */ 1057 } 1058 } 1059 /* See if fib was changed by packet filter. */ 1060 if (fibnum != M_GETFIB(m)) { 1061 m->m_flags |= M_SKIP_FIREWALL; 1062 fibnum = M_GETFIB(m); 1063 if (ro != NULL) 1064 RO_INVALIDATE_CACHE(ro); 1065 needfiblookup = 1; 1066 } 1067 if (needfiblookup) 1068 goto again; 1069 1070 /* See if local, if yes, send it to netisr. */ 1071 if (m->m_flags & M_FASTFWD_OURS) { 1072 if (m->m_pkthdr.rcvif == NULL) 1073 m->m_pkthdr.rcvif = V_loif; 1074 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 1075 m->m_pkthdr.csum_flags |= 1076 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; 1077 m->m_pkthdr.csum_data = 0xffff; 1078 } 1079 #if defined(SCTP) || defined(SCTP_SUPPORT) 1080 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) 1081 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 1082 #endif 1083 error = netisr_queue(NETISR_IPV6, m); 1084 goto done; 1085 } 1086 /* Or forward to some other address? */ 1087 if ((m->m_flags & M_IP6_NEXTHOP) && 1088 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { 1089 if (ro != NULL) 1090 dst = (struct sockaddr_in6 *)&ro->ro_dst; 1091 else 1092 dst = &sin6; 1093 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6)); 1094 m->m_flags |= M_SKIP_FIREWALL; 1095 m->m_flags &= ~M_IP6_NEXTHOP; 1096 m_tag_delete(m, fwd_tag); 1097 goto again; 1098 } 1099 1100 passout: 1101 /* 1102 * Send the packet to the outgoing interface. 1103 * If necessary, do IPv6 fragmentation before sending. 1104 * 1105 * The logic here is rather complex: 1106 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 1107 * 1-a: send as is if tlen <= path mtu 1108 * 1-b: fragment if tlen > path mtu 1109 * 1110 * 2: if user asks us not to fragment (dontfrag == 1) 1111 * 2-a: send as is if tlen <= interface mtu 1112 * 2-b: error if tlen > interface mtu 1113 * 1114 * 3: if we always need to attach fragment header (alwaysfrag == 1) 1115 * always fragment 1116 * 1117 * 4: if dontfrag == 1 && alwaysfrag == 1 1118 * error, as we cannot handle this conflicting request. 1119 */ 1120 sw_csum = m->m_pkthdr.csum_flags; 1121 if (!hdrsplit) { 1122 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0; 1123 sw_csum &= ~ifp->if_hwassist; 1124 } else 1125 tso = 0; 1126 /* 1127 * If we added extension headers, we will not do TSO and calculate the 1128 * checksums ourselves for now. 1129 * XXX-BZ Need a framework to know when the NIC can handle it, even 1130 * with ext. hdrs. 1131 */ 1132 error = ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen, false); 1133 if (error != 0) 1134 goto bad; 1135 /* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */ 1136 tlen = m->m_pkthdr.len; 1137 1138 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso) 1139 dontfrag = 1; 1140 else 1141 dontfrag = 0; 1142 if (dontfrag && alwaysfrag) { /* Case 4. */ 1143 /* Conflicting request - can't transmit. */ 1144 error = EMSGSIZE; 1145 goto bad; 1146 } 1147 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* Case 2-b. */ 1148 /* 1149 * Even if the DONTFRAG option is specified, we cannot send the 1150 * packet when the data length is larger than the MTU of the 1151 * outgoing interface. 1152 * Notify the error by sending IPV6_PATHMTU ancillary data if 1153 * application wanted to know the MTU value. Also return an 1154 * error code (this is not described in the API spec). 1155 */ 1156 if (inp != NULL) 1157 ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu); 1158 error = EMSGSIZE; 1159 goto bad; 1160 } 1161 1162 /* Transmit packet without fragmentation. */ 1163 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* Cases 1-a and 2-a. */ 1164 struct in6_ifaddr *ia6; 1165 1166 ip6 = mtod(m, struct ip6_hdr *); 1167 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 1168 if (ia6) { 1169 /* Record statistics for this interface address. */ 1170 counter_u64_add(ia6->ia_ifa.ifa_opackets, 1); 1171 counter_u64_add(ia6->ia_ifa.ifa_obytes, 1172 m->m_pkthdr.len); 1173 ifa_free(&ia6->ia_ifa); 1174 } 1175 error = ip6_output_send(inp, ifp, origifp, m, dst, ro, 1176 (flags & IP_NO_SND_TAG_RL) ? false : true); 1177 goto done; 1178 } 1179 1180 /* Try to fragment the packet. Cases 1-b and 3. */ 1181 if (mtu < IPV6_MMTU) { 1182 /* Path MTU cannot be less than IPV6_MMTU. */ 1183 error = EMSGSIZE; 1184 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1185 goto bad; 1186 } else if (ip6->ip6_plen == 0) { 1187 /* Jumbo payload cannot be fragmented. */ 1188 error = EMSGSIZE; 1189 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1190 goto bad; 1191 } else { 1192 u_char nextproto; 1193 1194 /* 1195 * Too large for the destination or interface; 1196 * fragment if possible. 1197 * Must be able to put at least 8 bytes per fragment. 1198 */ 1199 if (mtu > IPV6_MAXPACKET) 1200 mtu = IPV6_MAXPACKET; 1201 1202 len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7; 1203 if (len < 8) { 1204 error = EMSGSIZE; 1205 in6_ifstat_inc(ifp, ifs6_out_fragfail); 1206 goto bad; 1207 } 1208 1209 /* 1210 * If the interface will not calculate checksums on 1211 * fragmented packets, then do it here. 1212 * XXX-BZ handle the hw offloading case. Need flags. 1213 */ 1214 error = ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags, 1215 plen, optlen, true); 1216 if (error != 0) 1217 goto bad; 1218 1219 /* 1220 * Change the next header field of the last header in the 1221 * unfragmentable part. 1222 */ 1223 if (exthdrs.ip6e_rthdr) { 1224 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 1225 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 1226 } else if (exthdrs.ip6e_dest1) { 1227 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 1228 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 1229 } else if (exthdrs.ip6e_hbh) { 1230 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 1231 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 1232 } else { 1233 ip6 = mtod(m, struct ip6_hdr *); 1234 nextproto = ip6->ip6_nxt; 1235 ip6->ip6_nxt = IPPROTO_FRAGMENT; 1236 } 1237 1238 /* 1239 * Loop through length of segment after first fragment, 1240 * make new header and copy data of each part and link onto 1241 * chain. 1242 */ 1243 m0 = m; 1244 id = htonl(ip6_randomid()); 1245 error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id); 1246 if (error != 0) 1247 goto sendorfree; 1248 1249 in6_ifstat_inc(ifp, ifs6_out_fragok); 1250 } 1251 1252 /* Remove leading garbage. */ 1253 sendorfree: 1254 m = m0->m_nextpkt; 1255 m0->m_nextpkt = 0; 1256 m_freem(m0); 1257 for (; m; m = m0) { 1258 m0 = m->m_nextpkt; 1259 m->m_nextpkt = 0; 1260 if (error == 0) { 1261 /* Record statistics for this interface address. */ 1262 if (ia) { 1263 counter_u64_add(ia->ia_ifa.ifa_opackets, 1); 1264 counter_u64_add(ia->ia_ifa.ifa_obytes, 1265 m->m_pkthdr.len); 1266 } 1267 error = ip6_output_send(inp, ifp, origifp, m, dst, ro, 1268 true); 1269 } else 1270 m_freem(m); 1271 } 1272 1273 if (error == 0) 1274 IP6STAT_INC(ip6s_fragmented); 1275 1276 done: 1277 return (error); 1278 1279 freehdrs: 1280 m_freem(exthdrs.ip6e_hbh); /* m_freem() checks if mbuf is NULL. */ 1281 m_freem(exthdrs.ip6e_dest1); 1282 m_freem(exthdrs.ip6e_rthdr); 1283 m_freem(exthdrs.ip6e_dest2); 1284 /* FALLTHROUGH */ 1285 bad: 1286 if (m) 1287 m_freem(m); 1288 goto done; 1289 } 1290 1291 static int 1292 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen) 1293 { 1294 struct mbuf *m; 1295 1296 if (hlen > MCLBYTES) 1297 return (ENOBUFS); /* XXX */ 1298 1299 if (hlen > MLEN) 1300 m = m_getcl(M_NOWAIT, MT_DATA, 0); 1301 else 1302 m = m_get(M_NOWAIT, MT_DATA); 1303 if (m == NULL) 1304 return (ENOBUFS); 1305 m->m_len = hlen; 1306 if (hdr) 1307 bcopy(hdr, mtod(m, caddr_t), hlen); 1308 1309 *mp = m; 1310 return (0); 1311 } 1312 1313 /* 1314 * Insert jumbo payload option. 1315 */ 1316 static int 1317 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1318 { 1319 struct mbuf *mopt; 1320 u_char *optbuf; 1321 u_int32_t v; 1322 1323 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1324 1325 /* 1326 * If there is no hop-by-hop options header, allocate new one. 1327 * If there is one but it doesn't have enough space to store the 1328 * jumbo payload option, allocate a cluster to store the whole options. 1329 * Otherwise, use it to store the options. 1330 */ 1331 if (exthdrs->ip6e_hbh == NULL) { 1332 mopt = m_get(M_NOWAIT, MT_DATA); 1333 if (mopt == NULL) 1334 return (ENOBUFS); 1335 mopt->m_len = JUMBOOPTLEN; 1336 optbuf = mtod(mopt, u_char *); 1337 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1338 exthdrs->ip6e_hbh = mopt; 1339 } else { 1340 struct ip6_hbh *hbh; 1341 1342 mopt = exthdrs->ip6e_hbh; 1343 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1344 /* 1345 * XXX assumption: 1346 * - exthdrs->ip6e_hbh is not referenced from places 1347 * other than exthdrs. 1348 * - exthdrs->ip6e_hbh is not an mbuf chain. 1349 */ 1350 int oldoptlen = mopt->m_len; 1351 struct mbuf *n; 1352 1353 /* 1354 * XXX: give up if the whole (new) hbh header does 1355 * not fit even in an mbuf cluster. 1356 */ 1357 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1358 return (ENOBUFS); 1359 1360 /* 1361 * As a consequence, we must always prepare a cluster 1362 * at this point. 1363 */ 1364 n = m_getcl(M_NOWAIT, MT_DATA, 0); 1365 if (n == NULL) 1366 return (ENOBUFS); 1367 n->m_len = oldoptlen + JUMBOOPTLEN; 1368 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1369 oldoptlen); 1370 optbuf = mtod(n, caddr_t) + oldoptlen; 1371 m_freem(mopt); 1372 mopt = exthdrs->ip6e_hbh = n; 1373 } else { 1374 optbuf = mtod(mopt, u_char *) + mopt->m_len; 1375 mopt->m_len += JUMBOOPTLEN; 1376 } 1377 optbuf[0] = IP6OPT_PADN; 1378 optbuf[1] = 1; 1379 1380 /* 1381 * Adjust the header length according to the pad and 1382 * the jumbo payload option. 1383 */ 1384 hbh = mtod(mopt, struct ip6_hbh *); 1385 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1386 } 1387 1388 /* fill in the option. */ 1389 optbuf[2] = IP6OPT_JUMBO; 1390 optbuf[3] = 4; 1391 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1392 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1393 1394 /* finally, adjust the packet header length */ 1395 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1396 1397 return (0); 1398 #undef JUMBOOPTLEN 1399 } 1400 1401 /* 1402 * Insert fragment header and copy unfragmentable header portions. 1403 */ 1404 static int 1405 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1406 struct ip6_frag **frghdrp) 1407 { 1408 struct mbuf *n, *mlast; 1409 1410 if (hlen > sizeof(struct ip6_hdr)) { 1411 n = m_copym(m0, sizeof(struct ip6_hdr), 1412 hlen - sizeof(struct ip6_hdr), M_NOWAIT); 1413 if (n == NULL) 1414 return (ENOBUFS); 1415 m->m_next = n; 1416 } else 1417 n = m; 1418 1419 /* Search for the last mbuf of unfragmentable part. */ 1420 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1421 ; 1422 1423 if (M_WRITABLE(mlast) && 1424 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1425 /* use the trailing space of the last mbuf for the fragment hdr */ 1426 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1427 mlast->m_len); 1428 mlast->m_len += sizeof(struct ip6_frag); 1429 m->m_pkthdr.len += sizeof(struct ip6_frag); 1430 } else { 1431 /* allocate a new mbuf for the fragment header */ 1432 struct mbuf *mfrg; 1433 1434 mfrg = m_get(M_NOWAIT, MT_DATA); 1435 if (mfrg == NULL) 1436 return (ENOBUFS); 1437 mfrg->m_len = sizeof(struct ip6_frag); 1438 *frghdrp = mtod(mfrg, struct ip6_frag *); 1439 mlast->m_next = mfrg; 1440 } 1441 1442 return (0); 1443 } 1444 1445 /* 1446 * Calculates IPv6 path mtu for destination @dst. 1447 * Resulting MTU is stored in @mtup. 1448 * 1449 * Returns 0 on success. 1450 */ 1451 static int 1452 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup) 1453 { 1454 struct epoch_tracker et; 1455 struct nhop_object *nh; 1456 struct in6_addr kdst; 1457 uint32_t scopeid; 1458 int error; 1459 1460 in6_splitscope(dst, &kdst, &scopeid); 1461 1462 NET_EPOCH_ENTER(et); 1463 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0); 1464 if (nh != NULL) 1465 error = ip6_calcmtu(nh->nh_ifp, dst, nh->nh_mtu, mtup, NULL, 0); 1466 else 1467 error = EHOSTUNREACH; 1468 NET_EPOCH_EXIT(et); 1469 1470 return (error); 1471 } 1472 1473 /* 1474 * Calculates IPv6 path MTU for @dst based on transmit @ifp, 1475 * and cached data in @ro_pmtu. 1476 * MTU from (successful) route lookup is saved (along with dst) 1477 * inside @ro_pmtu to avoid subsequent route lookups after packet 1478 * filter processing. 1479 * 1480 * Stores mtu and always-frag value into @mtup and @alwaysfragp. 1481 * Returns 0 on success. 1482 */ 1483 static int 1484 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup, 1485 struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup, 1486 int *alwaysfragp, u_int fibnum, u_int proto) 1487 { 1488 struct nhop_object *nh; 1489 struct in6_addr kdst; 1490 uint32_t scopeid; 1491 struct sockaddr_in6 *sa6_dst, sin6; 1492 u_long mtu; 1493 1494 NET_EPOCH_ASSERT(); 1495 1496 mtu = 0; 1497 if (ro_pmtu == NULL || do_lookup) { 1498 /* 1499 * Here ro_pmtu has final destination address, while 1500 * ro might represent immediate destination. 1501 * Use ro_pmtu destination since mtu might differ. 1502 */ 1503 if (ro_pmtu != NULL) { 1504 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1505 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst)) 1506 ro_pmtu->ro_mtu = 0; 1507 } else 1508 sa6_dst = &sin6; 1509 1510 if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) { 1511 bzero(sa6_dst, sizeof(*sa6_dst)); 1512 sa6_dst->sin6_family = AF_INET6; 1513 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1514 sa6_dst->sin6_addr = *dst; 1515 1516 in6_splitscope(dst, &kdst, &scopeid); 1517 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0); 1518 if (nh != NULL) { 1519 mtu = nh->nh_mtu; 1520 if (ro_pmtu != NULL) 1521 ro_pmtu->ro_mtu = mtu; 1522 } 1523 } else 1524 mtu = ro_pmtu->ro_mtu; 1525 } 1526 1527 if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL) 1528 mtu = ro_pmtu->ro_nh->nh_mtu; 1529 1530 return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto)); 1531 } 1532 1533 /* 1534 * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and 1535 * hostcache data for @dst. 1536 * Stores mtu and always-frag value into @mtup and @alwaysfragp. 1537 * 1538 * Returns 0 on success. 1539 */ 1540 static int 1541 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu, 1542 u_long *mtup, int *alwaysfragp, u_int proto) 1543 { 1544 u_long mtu = 0; 1545 int alwaysfrag = 0; 1546 int error = 0; 1547 1548 if (rt_mtu > 0) { 1549 u_int32_t ifmtu; 1550 struct in_conninfo inc; 1551 1552 bzero(&inc, sizeof(inc)); 1553 inc.inc_flags |= INC_ISIPV6; 1554 inc.inc6_faddr = *dst; 1555 1556 ifmtu = IN6_LINKMTU(ifp); 1557 1558 /* TCP is known to react to pmtu changes so skip hc */ 1559 if (proto != IPPROTO_TCP) 1560 mtu = tcp_hc_getmtu(&inc); 1561 1562 if (mtu) 1563 mtu = min(mtu, rt_mtu); 1564 else 1565 mtu = rt_mtu; 1566 if (mtu == 0) 1567 mtu = ifmtu; 1568 else if (mtu < IPV6_MMTU) { 1569 /* 1570 * RFC2460 section 5, last paragraph: 1571 * if we record ICMPv6 too big message with 1572 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1573 * or smaller, with framgent header attached. 1574 * (fragment header is needed regardless from the 1575 * packet size, for translators to identify packets) 1576 */ 1577 alwaysfrag = 1; 1578 mtu = IPV6_MMTU; 1579 } 1580 } else if (ifp) { 1581 mtu = IN6_LINKMTU(ifp); 1582 } else 1583 error = EHOSTUNREACH; /* XXX */ 1584 1585 *mtup = mtu; 1586 if (alwaysfragp) 1587 *alwaysfragp = alwaysfrag; 1588 return (error); 1589 } 1590 1591 /* 1592 * IP6 socket option processing. 1593 */ 1594 int 1595 ip6_ctloutput(struct socket *so, struct sockopt *sopt) 1596 { 1597 int optdatalen, uproto; 1598 void *optdata; 1599 struct inpcb *inp = sotoinpcb(so); 1600 int error, optval; 1601 int level, op, optname; 1602 int optlen; 1603 struct thread *td; 1604 #ifdef RSS 1605 uint32_t rss_bucket; 1606 int retval; 1607 #endif 1608 1609 /* 1610 * Don't use more than a quarter of mbuf clusters. N.B.: 1611 * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow 1612 * on LP64 architectures, so cast to u_long to avoid undefined 1613 * behavior. ILP32 architectures cannot have nmbclusters 1614 * large enough to overflow for other reasons. 1615 */ 1616 #define IPV6_PKTOPTIONS_MBUF_LIMIT ((u_long)nmbclusters * MCLBYTES / 4) 1617 1618 level = sopt->sopt_level; 1619 op = sopt->sopt_dir; 1620 optname = sopt->sopt_name; 1621 optlen = sopt->sopt_valsize; 1622 td = sopt->sopt_td; 1623 error = 0; 1624 optval = 0; 1625 uproto = (int)so->so_proto->pr_protocol; 1626 1627 if (level != IPPROTO_IPV6) { 1628 error = EINVAL; 1629 1630 if (sopt->sopt_level == SOL_SOCKET && 1631 sopt->sopt_dir == SOPT_SET) { 1632 switch (sopt->sopt_name) { 1633 case SO_REUSEADDR: 1634 INP_WLOCK(inp); 1635 if ((so->so_options & SO_REUSEADDR) != 0) 1636 inp->inp_flags2 |= INP_REUSEADDR; 1637 else 1638 inp->inp_flags2 &= ~INP_REUSEADDR; 1639 INP_WUNLOCK(inp); 1640 error = 0; 1641 break; 1642 case SO_REUSEPORT: 1643 INP_WLOCK(inp); 1644 if ((so->so_options & SO_REUSEPORT) != 0) 1645 inp->inp_flags2 |= INP_REUSEPORT; 1646 else 1647 inp->inp_flags2 &= ~INP_REUSEPORT; 1648 INP_WUNLOCK(inp); 1649 error = 0; 1650 break; 1651 case SO_REUSEPORT_LB: 1652 INP_WLOCK(inp); 1653 if ((so->so_options & SO_REUSEPORT_LB) != 0) 1654 inp->inp_flags2 |= INP_REUSEPORT_LB; 1655 else 1656 inp->inp_flags2 &= ~INP_REUSEPORT_LB; 1657 INP_WUNLOCK(inp); 1658 error = 0; 1659 break; 1660 case SO_SETFIB: 1661 INP_WLOCK(inp); 1662 inp->inp_inc.inc_fibnum = so->so_fibnum; 1663 INP_WUNLOCK(inp); 1664 error = 0; 1665 break; 1666 case SO_MAX_PACING_RATE: 1667 #ifdef RATELIMIT 1668 INP_WLOCK(inp); 1669 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED; 1670 INP_WUNLOCK(inp); 1671 error = 0; 1672 #else 1673 error = EOPNOTSUPP; 1674 #endif 1675 break; 1676 default: 1677 break; 1678 } 1679 } 1680 } else { /* level == IPPROTO_IPV6 */ 1681 switch (op) { 1682 case SOPT_SET: 1683 switch (optname) { 1684 case IPV6_2292PKTOPTIONS: 1685 #ifdef IPV6_PKTOPTIONS 1686 case IPV6_PKTOPTIONS: 1687 #endif 1688 { 1689 struct mbuf *m; 1690 1691 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) { 1692 printf("ip6_ctloutput: mbuf limit hit\n"); 1693 error = ENOBUFS; 1694 break; 1695 } 1696 1697 error = soopt_getm(sopt, &m); /* XXX */ 1698 if (error != 0) 1699 break; 1700 error = soopt_mcopyin(sopt, m); /* XXX */ 1701 if (error != 0) 1702 break; 1703 INP_WLOCK(inp); 1704 error = ip6_pcbopts(&inp->in6p_outputopts, m, 1705 so, sopt); 1706 INP_WUNLOCK(inp); 1707 m_freem(m); /* XXX */ 1708 break; 1709 } 1710 1711 /* 1712 * Use of some Hop-by-Hop options or some 1713 * Destination options, might require special 1714 * privilege. That is, normal applications 1715 * (without special privilege) might be forbidden 1716 * from setting certain options in outgoing packets, 1717 * and might never see certain options in received 1718 * packets. [RFC 2292 Section 6] 1719 * KAME specific note: 1720 * KAME prevents non-privileged users from sending or 1721 * receiving ANY hbh/dst options in order to avoid 1722 * overhead of parsing options in the kernel. 1723 */ 1724 case IPV6_RECVHOPOPTS: 1725 case IPV6_RECVDSTOPTS: 1726 case IPV6_RECVRTHDRDSTOPTS: 1727 if (td != NULL) { 1728 error = priv_check(td, 1729 PRIV_NETINET_SETHDROPTS); 1730 if (error) 1731 break; 1732 } 1733 /* FALLTHROUGH */ 1734 case IPV6_UNICAST_HOPS: 1735 case IPV6_HOPLIMIT: 1736 1737 case IPV6_RECVPKTINFO: 1738 case IPV6_RECVHOPLIMIT: 1739 case IPV6_RECVRTHDR: 1740 case IPV6_RECVPATHMTU: 1741 case IPV6_RECVTCLASS: 1742 case IPV6_RECVFLOWID: 1743 #ifdef RSS 1744 case IPV6_RECVRSSBUCKETID: 1745 #endif 1746 case IPV6_V6ONLY: 1747 case IPV6_AUTOFLOWLABEL: 1748 case IPV6_ORIGDSTADDR: 1749 case IPV6_BINDANY: 1750 case IPV6_BINDMULTI: 1751 #ifdef RSS 1752 case IPV6_RSS_LISTEN_BUCKET: 1753 #endif 1754 if (optname == IPV6_BINDANY && td != NULL) { 1755 error = priv_check(td, 1756 PRIV_NETINET_BINDANY); 1757 if (error) 1758 break; 1759 } 1760 1761 if (optlen != sizeof(int)) { 1762 error = EINVAL; 1763 break; 1764 } 1765 error = sooptcopyin(sopt, &optval, 1766 sizeof optval, sizeof optval); 1767 if (error) 1768 break; 1769 switch (optname) { 1770 case IPV6_UNICAST_HOPS: 1771 if (optval < -1 || optval >= 256) 1772 error = EINVAL; 1773 else { 1774 /* -1 = kernel default */ 1775 inp->in6p_hops = optval; 1776 if ((inp->inp_vflag & 1777 INP_IPV4) != 0) 1778 inp->inp_ip_ttl = optval; 1779 } 1780 break; 1781 #define OPTSET(bit) \ 1782 do { \ 1783 INP_WLOCK(inp); \ 1784 if (optval) \ 1785 inp->inp_flags |= (bit); \ 1786 else \ 1787 inp->inp_flags &= ~(bit); \ 1788 INP_WUNLOCK(inp); \ 1789 } while (/*CONSTCOND*/ 0) 1790 #define OPTSET2292(bit) \ 1791 do { \ 1792 INP_WLOCK(inp); \ 1793 inp->inp_flags |= IN6P_RFC2292; \ 1794 if (optval) \ 1795 inp->inp_flags |= (bit); \ 1796 else \ 1797 inp->inp_flags &= ~(bit); \ 1798 INP_WUNLOCK(inp); \ 1799 } while (/*CONSTCOND*/ 0) 1800 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0) 1801 1802 #define OPTSET2_N(bit, val) do { \ 1803 if (val) \ 1804 inp->inp_flags2 |= bit; \ 1805 else \ 1806 inp->inp_flags2 &= ~bit; \ 1807 } while (0) 1808 #define OPTSET2(bit, val) do { \ 1809 INP_WLOCK(inp); \ 1810 OPTSET2_N(bit, val); \ 1811 INP_WUNLOCK(inp); \ 1812 } while (0) 1813 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0) 1814 #define OPTSET2292_EXCLUSIVE(bit) \ 1815 do { \ 1816 INP_WLOCK(inp); \ 1817 if (OPTBIT(IN6P_RFC2292)) { \ 1818 error = EINVAL; \ 1819 } else { \ 1820 if (optval) \ 1821 inp->inp_flags |= (bit); \ 1822 else \ 1823 inp->inp_flags &= ~(bit); \ 1824 } \ 1825 INP_WUNLOCK(inp); \ 1826 } while (/*CONSTCOND*/ 0) 1827 1828 case IPV6_RECVPKTINFO: 1829 OPTSET2292_EXCLUSIVE(IN6P_PKTINFO); 1830 break; 1831 1832 case IPV6_HOPLIMIT: 1833 { 1834 struct ip6_pktopts **optp; 1835 1836 /* cannot mix with RFC2292 */ 1837 if (OPTBIT(IN6P_RFC2292)) { 1838 error = EINVAL; 1839 break; 1840 } 1841 INP_WLOCK(inp); 1842 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 1843 INP_WUNLOCK(inp); 1844 return (ECONNRESET); 1845 } 1846 optp = &inp->in6p_outputopts; 1847 error = ip6_pcbopt(IPV6_HOPLIMIT, 1848 (u_char *)&optval, sizeof(optval), 1849 optp, (td != NULL) ? td->td_ucred : 1850 NULL, uproto); 1851 INP_WUNLOCK(inp); 1852 break; 1853 } 1854 1855 case IPV6_RECVHOPLIMIT: 1856 OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT); 1857 break; 1858 1859 case IPV6_RECVHOPOPTS: 1860 OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS); 1861 break; 1862 1863 case IPV6_RECVDSTOPTS: 1864 OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS); 1865 break; 1866 1867 case IPV6_RECVRTHDRDSTOPTS: 1868 OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS); 1869 break; 1870 1871 case IPV6_RECVRTHDR: 1872 OPTSET2292_EXCLUSIVE(IN6P_RTHDR); 1873 break; 1874 1875 case IPV6_RECVPATHMTU: 1876 /* 1877 * We ignore this option for TCP 1878 * sockets. 1879 * (RFC3542 leaves this case 1880 * unspecified.) 1881 */ 1882 if (uproto != IPPROTO_TCP) 1883 OPTSET(IN6P_MTU); 1884 break; 1885 1886 case IPV6_RECVFLOWID: 1887 OPTSET2(INP_RECVFLOWID, optval); 1888 break; 1889 1890 #ifdef RSS 1891 case IPV6_RECVRSSBUCKETID: 1892 OPTSET2(INP_RECVRSSBUCKETID, optval); 1893 break; 1894 #endif 1895 1896 case IPV6_V6ONLY: 1897 INP_WLOCK(inp); 1898 if (inp->inp_lport || 1899 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 1900 /* 1901 * The socket is already bound. 1902 */ 1903 INP_WUNLOCK(inp); 1904 error = EINVAL; 1905 break; 1906 } 1907 if (optval) { 1908 inp->inp_flags |= IN6P_IPV6_V6ONLY; 1909 inp->inp_vflag &= ~INP_IPV4; 1910 } else { 1911 inp->inp_flags &= ~IN6P_IPV6_V6ONLY; 1912 inp->inp_vflag |= INP_IPV4; 1913 } 1914 INP_WUNLOCK(inp); 1915 break; 1916 case IPV6_RECVTCLASS: 1917 /* cannot mix with RFC2292 XXX */ 1918 OPTSET2292_EXCLUSIVE(IN6P_TCLASS); 1919 break; 1920 case IPV6_AUTOFLOWLABEL: 1921 OPTSET(IN6P_AUTOFLOWLABEL); 1922 break; 1923 1924 case IPV6_ORIGDSTADDR: 1925 OPTSET2(INP_ORIGDSTADDR, optval); 1926 break; 1927 case IPV6_BINDANY: 1928 OPTSET(INP_BINDANY); 1929 break; 1930 1931 case IPV6_BINDMULTI: 1932 OPTSET2(INP_BINDMULTI, optval); 1933 break; 1934 #ifdef RSS 1935 case IPV6_RSS_LISTEN_BUCKET: 1936 if ((optval >= 0) && 1937 (optval < rss_getnumbuckets())) { 1938 INP_WLOCK(inp); 1939 inp->inp_rss_listen_bucket = optval; 1940 OPTSET2_N(INP_RSS_BUCKET_SET, 1); 1941 INP_WUNLOCK(inp); 1942 } else { 1943 error = EINVAL; 1944 } 1945 break; 1946 #endif 1947 } 1948 break; 1949 1950 case IPV6_TCLASS: 1951 case IPV6_DONTFRAG: 1952 case IPV6_USE_MIN_MTU: 1953 case IPV6_PREFER_TEMPADDR: 1954 if (optlen != sizeof(optval)) { 1955 error = EINVAL; 1956 break; 1957 } 1958 error = sooptcopyin(sopt, &optval, 1959 sizeof optval, sizeof optval); 1960 if (error) 1961 break; 1962 { 1963 struct ip6_pktopts **optp; 1964 INP_WLOCK(inp); 1965 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 1966 INP_WUNLOCK(inp); 1967 return (ECONNRESET); 1968 } 1969 optp = &inp->in6p_outputopts; 1970 error = ip6_pcbopt(optname, 1971 (u_char *)&optval, sizeof(optval), 1972 optp, (td != NULL) ? td->td_ucred : 1973 NULL, uproto); 1974 INP_WUNLOCK(inp); 1975 break; 1976 } 1977 1978 case IPV6_2292PKTINFO: 1979 case IPV6_2292HOPLIMIT: 1980 case IPV6_2292HOPOPTS: 1981 case IPV6_2292DSTOPTS: 1982 case IPV6_2292RTHDR: 1983 /* RFC 2292 */ 1984 if (optlen != sizeof(int)) { 1985 error = EINVAL; 1986 break; 1987 } 1988 error = sooptcopyin(sopt, &optval, 1989 sizeof optval, sizeof optval); 1990 if (error) 1991 break; 1992 switch (optname) { 1993 case IPV6_2292PKTINFO: 1994 OPTSET2292(IN6P_PKTINFO); 1995 break; 1996 case IPV6_2292HOPLIMIT: 1997 OPTSET2292(IN6P_HOPLIMIT); 1998 break; 1999 case IPV6_2292HOPOPTS: 2000 /* 2001 * Check super-user privilege. 2002 * See comments for IPV6_RECVHOPOPTS. 2003 */ 2004 if (td != NULL) { 2005 error = priv_check(td, 2006 PRIV_NETINET_SETHDROPTS); 2007 if (error) 2008 return (error); 2009 } 2010 OPTSET2292(IN6P_HOPOPTS); 2011 break; 2012 case IPV6_2292DSTOPTS: 2013 if (td != NULL) { 2014 error = priv_check(td, 2015 PRIV_NETINET_SETHDROPTS); 2016 if (error) 2017 return (error); 2018 } 2019 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 2020 break; 2021 case IPV6_2292RTHDR: 2022 OPTSET2292(IN6P_RTHDR); 2023 break; 2024 } 2025 break; 2026 case IPV6_PKTINFO: 2027 case IPV6_HOPOPTS: 2028 case IPV6_RTHDR: 2029 case IPV6_DSTOPTS: 2030 case IPV6_RTHDRDSTOPTS: 2031 case IPV6_NEXTHOP: 2032 { 2033 /* new advanced API (RFC3542) */ 2034 u_char *optbuf; 2035 u_char optbuf_storage[MCLBYTES]; 2036 int optlen; 2037 struct ip6_pktopts **optp; 2038 2039 /* cannot mix with RFC2292 */ 2040 if (OPTBIT(IN6P_RFC2292)) { 2041 error = EINVAL; 2042 break; 2043 } 2044 2045 /* 2046 * We only ensure valsize is not too large 2047 * here. Further validation will be done 2048 * later. 2049 */ 2050 error = sooptcopyin(sopt, optbuf_storage, 2051 sizeof(optbuf_storage), 0); 2052 if (error) 2053 break; 2054 optlen = sopt->sopt_valsize; 2055 optbuf = optbuf_storage; 2056 INP_WLOCK(inp); 2057 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 2058 INP_WUNLOCK(inp); 2059 return (ECONNRESET); 2060 } 2061 optp = &inp->in6p_outputopts; 2062 error = ip6_pcbopt(optname, optbuf, optlen, 2063 optp, (td != NULL) ? td->td_ucred : NULL, 2064 uproto); 2065 INP_WUNLOCK(inp); 2066 break; 2067 } 2068 #undef OPTSET 2069 2070 case IPV6_MULTICAST_IF: 2071 case IPV6_MULTICAST_HOPS: 2072 case IPV6_MULTICAST_LOOP: 2073 case IPV6_JOIN_GROUP: 2074 case IPV6_LEAVE_GROUP: 2075 case IPV6_MSFILTER: 2076 case MCAST_BLOCK_SOURCE: 2077 case MCAST_UNBLOCK_SOURCE: 2078 case MCAST_JOIN_GROUP: 2079 case MCAST_LEAVE_GROUP: 2080 case MCAST_JOIN_SOURCE_GROUP: 2081 case MCAST_LEAVE_SOURCE_GROUP: 2082 error = ip6_setmoptions(inp, sopt); 2083 break; 2084 2085 case IPV6_PORTRANGE: 2086 error = sooptcopyin(sopt, &optval, 2087 sizeof optval, sizeof optval); 2088 if (error) 2089 break; 2090 2091 INP_WLOCK(inp); 2092 switch (optval) { 2093 case IPV6_PORTRANGE_DEFAULT: 2094 inp->inp_flags &= ~(INP_LOWPORT); 2095 inp->inp_flags &= ~(INP_HIGHPORT); 2096 break; 2097 2098 case IPV6_PORTRANGE_HIGH: 2099 inp->inp_flags &= ~(INP_LOWPORT); 2100 inp->inp_flags |= INP_HIGHPORT; 2101 break; 2102 2103 case IPV6_PORTRANGE_LOW: 2104 inp->inp_flags &= ~(INP_HIGHPORT); 2105 inp->inp_flags |= INP_LOWPORT; 2106 break; 2107 2108 default: 2109 error = EINVAL; 2110 break; 2111 } 2112 INP_WUNLOCK(inp); 2113 break; 2114 2115 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 2116 case IPV6_IPSEC_POLICY: 2117 if (IPSEC_ENABLED(ipv6)) { 2118 error = IPSEC_PCBCTL(ipv6, inp, sopt); 2119 break; 2120 } 2121 /* FALLTHROUGH */ 2122 #endif /* IPSEC */ 2123 2124 default: 2125 error = ENOPROTOOPT; 2126 break; 2127 } 2128 break; 2129 2130 case SOPT_GET: 2131 switch (optname) { 2132 case IPV6_2292PKTOPTIONS: 2133 #ifdef IPV6_PKTOPTIONS 2134 case IPV6_PKTOPTIONS: 2135 #endif 2136 /* 2137 * RFC3542 (effectively) deprecated the 2138 * semantics of the 2292-style pktoptions. 2139 * Since it was not reliable in nature (i.e., 2140 * applications had to expect the lack of some 2141 * information after all), it would make sense 2142 * to simplify this part by always returning 2143 * empty data. 2144 */ 2145 sopt->sopt_valsize = 0; 2146 break; 2147 2148 case IPV6_RECVHOPOPTS: 2149 case IPV6_RECVDSTOPTS: 2150 case IPV6_RECVRTHDRDSTOPTS: 2151 case IPV6_UNICAST_HOPS: 2152 case IPV6_RECVPKTINFO: 2153 case IPV6_RECVHOPLIMIT: 2154 case IPV6_RECVRTHDR: 2155 case IPV6_RECVPATHMTU: 2156 2157 case IPV6_V6ONLY: 2158 case IPV6_PORTRANGE: 2159 case IPV6_RECVTCLASS: 2160 case IPV6_AUTOFLOWLABEL: 2161 case IPV6_BINDANY: 2162 case IPV6_FLOWID: 2163 case IPV6_FLOWTYPE: 2164 case IPV6_RECVFLOWID: 2165 #ifdef RSS 2166 case IPV6_RSSBUCKETID: 2167 case IPV6_RECVRSSBUCKETID: 2168 #endif 2169 case IPV6_BINDMULTI: 2170 switch (optname) { 2171 case IPV6_RECVHOPOPTS: 2172 optval = OPTBIT(IN6P_HOPOPTS); 2173 break; 2174 2175 case IPV6_RECVDSTOPTS: 2176 optval = OPTBIT(IN6P_DSTOPTS); 2177 break; 2178 2179 case IPV6_RECVRTHDRDSTOPTS: 2180 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 2181 break; 2182 2183 case IPV6_UNICAST_HOPS: 2184 optval = inp->in6p_hops; 2185 break; 2186 2187 case IPV6_RECVPKTINFO: 2188 optval = OPTBIT(IN6P_PKTINFO); 2189 break; 2190 2191 case IPV6_RECVHOPLIMIT: 2192 optval = OPTBIT(IN6P_HOPLIMIT); 2193 break; 2194 2195 case IPV6_RECVRTHDR: 2196 optval = OPTBIT(IN6P_RTHDR); 2197 break; 2198 2199 case IPV6_RECVPATHMTU: 2200 optval = OPTBIT(IN6P_MTU); 2201 break; 2202 2203 case IPV6_V6ONLY: 2204 optval = OPTBIT(IN6P_IPV6_V6ONLY); 2205 break; 2206 2207 case IPV6_PORTRANGE: 2208 { 2209 int flags; 2210 flags = inp->inp_flags; 2211 if (flags & INP_HIGHPORT) 2212 optval = IPV6_PORTRANGE_HIGH; 2213 else if (flags & INP_LOWPORT) 2214 optval = IPV6_PORTRANGE_LOW; 2215 else 2216 optval = 0; 2217 break; 2218 } 2219 case IPV6_RECVTCLASS: 2220 optval = OPTBIT(IN6P_TCLASS); 2221 break; 2222 2223 case IPV6_AUTOFLOWLABEL: 2224 optval = OPTBIT(IN6P_AUTOFLOWLABEL); 2225 break; 2226 2227 case IPV6_ORIGDSTADDR: 2228 optval = OPTBIT2(INP_ORIGDSTADDR); 2229 break; 2230 2231 case IPV6_BINDANY: 2232 optval = OPTBIT(INP_BINDANY); 2233 break; 2234 2235 case IPV6_FLOWID: 2236 optval = inp->inp_flowid; 2237 break; 2238 2239 case IPV6_FLOWTYPE: 2240 optval = inp->inp_flowtype; 2241 break; 2242 2243 case IPV6_RECVFLOWID: 2244 optval = OPTBIT2(INP_RECVFLOWID); 2245 break; 2246 #ifdef RSS 2247 case IPV6_RSSBUCKETID: 2248 retval = 2249 rss_hash2bucket(inp->inp_flowid, 2250 inp->inp_flowtype, 2251 &rss_bucket); 2252 if (retval == 0) 2253 optval = rss_bucket; 2254 else 2255 error = EINVAL; 2256 break; 2257 2258 case IPV6_RECVRSSBUCKETID: 2259 optval = OPTBIT2(INP_RECVRSSBUCKETID); 2260 break; 2261 #endif 2262 2263 case IPV6_BINDMULTI: 2264 optval = OPTBIT2(INP_BINDMULTI); 2265 break; 2266 } 2267 if (error) 2268 break; 2269 error = sooptcopyout(sopt, &optval, 2270 sizeof optval); 2271 break; 2272 2273 case IPV6_PATHMTU: 2274 { 2275 u_long pmtu = 0; 2276 struct ip6_mtuinfo mtuinfo; 2277 struct in6_addr addr; 2278 2279 if (!(so->so_state & SS_ISCONNECTED)) 2280 return (ENOTCONN); 2281 /* 2282 * XXX: we dot not consider the case of source 2283 * routing, or optional information to specify 2284 * the outgoing interface. 2285 * Copy faddr out of inp to avoid holding lock 2286 * on inp during route lookup. 2287 */ 2288 INP_RLOCK(inp); 2289 bcopy(&inp->in6p_faddr, &addr, sizeof(addr)); 2290 INP_RUNLOCK(inp); 2291 error = ip6_getpmtu_ctl(so->so_fibnum, 2292 &addr, &pmtu); 2293 if (error) 2294 break; 2295 if (pmtu > IPV6_MAXPACKET) 2296 pmtu = IPV6_MAXPACKET; 2297 2298 bzero(&mtuinfo, sizeof(mtuinfo)); 2299 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 2300 optdata = (void *)&mtuinfo; 2301 optdatalen = sizeof(mtuinfo); 2302 error = sooptcopyout(sopt, optdata, 2303 optdatalen); 2304 break; 2305 } 2306 2307 case IPV6_2292PKTINFO: 2308 case IPV6_2292HOPLIMIT: 2309 case IPV6_2292HOPOPTS: 2310 case IPV6_2292RTHDR: 2311 case IPV6_2292DSTOPTS: 2312 switch (optname) { 2313 case IPV6_2292PKTINFO: 2314 optval = OPTBIT(IN6P_PKTINFO); 2315 break; 2316 case IPV6_2292HOPLIMIT: 2317 optval = OPTBIT(IN6P_HOPLIMIT); 2318 break; 2319 case IPV6_2292HOPOPTS: 2320 optval = OPTBIT(IN6P_HOPOPTS); 2321 break; 2322 case IPV6_2292RTHDR: 2323 optval = OPTBIT(IN6P_RTHDR); 2324 break; 2325 case IPV6_2292DSTOPTS: 2326 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 2327 break; 2328 } 2329 error = sooptcopyout(sopt, &optval, 2330 sizeof optval); 2331 break; 2332 case IPV6_PKTINFO: 2333 case IPV6_HOPOPTS: 2334 case IPV6_RTHDR: 2335 case IPV6_DSTOPTS: 2336 case IPV6_RTHDRDSTOPTS: 2337 case IPV6_NEXTHOP: 2338 case IPV6_TCLASS: 2339 case IPV6_DONTFRAG: 2340 case IPV6_USE_MIN_MTU: 2341 case IPV6_PREFER_TEMPADDR: 2342 error = ip6_getpcbopt(inp, optname, sopt); 2343 break; 2344 2345 case IPV6_MULTICAST_IF: 2346 case IPV6_MULTICAST_HOPS: 2347 case IPV6_MULTICAST_LOOP: 2348 case IPV6_MSFILTER: 2349 error = ip6_getmoptions(inp, sopt); 2350 break; 2351 2352 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 2353 case IPV6_IPSEC_POLICY: 2354 if (IPSEC_ENABLED(ipv6)) { 2355 error = IPSEC_PCBCTL(ipv6, inp, sopt); 2356 break; 2357 } 2358 /* FALLTHROUGH */ 2359 #endif /* IPSEC */ 2360 default: 2361 error = ENOPROTOOPT; 2362 break; 2363 } 2364 break; 2365 } 2366 } 2367 return (error); 2368 } 2369 2370 int 2371 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt) 2372 { 2373 int error = 0, optval, optlen; 2374 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2375 struct inpcb *inp = sotoinpcb(so); 2376 int level, op, optname; 2377 2378 level = sopt->sopt_level; 2379 op = sopt->sopt_dir; 2380 optname = sopt->sopt_name; 2381 optlen = sopt->sopt_valsize; 2382 2383 if (level != IPPROTO_IPV6) { 2384 return (EINVAL); 2385 } 2386 2387 switch (optname) { 2388 case IPV6_CHECKSUM: 2389 /* 2390 * For ICMPv6 sockets, no modification allowed for checksum 2391 * offset, permit "no change" values to help existing apps. 2392 * 2393 * RFC3542 says: "An attempt to set IPV6_CHECKSUM 2394 * for an ICMPv6 socket will fail." 2395 * The current behavior does not meet RFC3542. 2396 */ 2397 switch (op) { 2398 case SOPT_SET: 2399 if (optlen != sizeof(int)) { 2400 error = EINVAL; 2401 break; 2402 } 2403 error = sooptcopyin(sopt, &optval, sizeof(optval), 2404 sizeof(optval)); 2405 if (error) 2406 break; 2407 if (optval < -1 || (optval % 2) != 0) { 2408 /* 2409 * The API assumes non-negative even offset 2410 * values or -1 as a special value. 2411 */ 2412 error = EINVAL; 2413 } else if (so->so_proto->pr_protocol == 2414 IPPROTO_ICMPV6) { 2415 if (optval != icmp6off) 2416 error = EINVAL; 2417 } else 2418 inp->in6p_cksum = optval; 2419 break; 2420 2421 case SOPT_GET: 2422 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2423 optval = icmp6off; 2424 else 2425 optval = inp->in6p_cksum; 2426 2427 error = sooptcopyout(sopt, &optval, sizeof(optval)); 2428 break; 2429 2430 default: 2431 error = EINVAL; 2432 break; 2433 } 2434 break; 2435 2436 default: 2437 error = ENOPROTOOPT; 2438 break; 2439 } 2440 2441 return (error); 2442 } 2443 2444 /* 2445 * Set up IP6 options in pcb for insertion in output packets or 2446 * specifying behavior of outgoing packets. 2447 */ 2448 static int 2449 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, 2450 struct socket *so, struct sockopt *sopt) 2451 { 2452 struct ip6_pktopts *opt = *pktopt; 2453 int error = 0; 2454 struct thread *td = sopt->sopt_td; 2455 2456 /* turn off any old options. */ 2457 if (opt) { 2458 #ifdef DIAGNOSTIC 2459 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2460 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2461 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2462 printf("ip6_pcbopts: all specified options are cleared.\n"); 2463 #endif 2464 ip6_clearpktopts(opt, -1); 2465 } else { 2466 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT); 2467 if (opt == NULL) 2468 return (ENOMEM); 2469 } 2470 *pktopt = NULL; 2471 2472 if (!m || m->m_len == 0) { 2473 /* 2474 * Only turning off any previous options, regardless of 2475 * whether the opt is just created or given. 2476 */ 2477 free(opt, M_IP6OPT); 2478 return (0); 2479 } 2480 2481 /* set options specified by user. */ 2482 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ? 2483 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) { 2484 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2485 free(opt, M_IP6OPT); 2486 return (error); 2487 } 2488 *pktopt = opt; 2489 return (0); 2490 } 2491 2492 /* 2493 * initialize ip6_pktopts. beware that there are non-zero default values in 2494 * the struct. 2495 */ 2496 void 2497 ip6_initpktopts(struct ip6_pktopts *opt) 2498 { 2499 2500 bzero(opt, sizeof(*opt)); 2501 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2502 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2503 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2504 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; 2505 } 2506 2507 static int 2508 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2509 struct ucred *cred, int uproto) 2510 { 2511 struct ip6_pktopts *opt; 2512 2513 if (*pktopt == NULL) { 2514 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2515 M_NOWAIT); 2516 if (*pktopt == NULL) 2517 return (ENOBUFS); 2518 ip6_initpktopts(*pktopt); 2519 } 2520 opt = *pktopt; 2521 2522 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); 2523 } 2524 2525 #define GET_PKTOPT_VAR(field, lenexpr) do { \ 2526 if (pktopt && pktopt->field) { \ 2527 INP_RUNLOCK(inp); \ 2528 optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK); \ 2529 malloc_optdata = true; \ 2530 INP_RLOCK(inp); \ 2531 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \ 2532 INP_RUNLOCK(inp); \ 2533 free(optdata, M_TEMP); \ 2534 return (ECONNRESET); \ 2535 } \ 2536 pktopt = inp->in6p_outputopts; \ 2537 if (pktopt && pktopt->field) { \ 2538 optdatalen = min(lenexpr, sopt->sopt_valsize); \ 2539 bcopy(&pktopt->field, optdata, optdatalen); \ 2540 } else { \ 2541 free(optdata, M_TEMP); \ 2542 optdata = NULL; \ 2543 malloc_optdata = false; \ 2544 } \ 2545 } \ 2546 } while(0) 2547 2548 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field, \ 2549 (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3) 2550 2551 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field, \ 2552 pktopt->field->sa_len) 2553 2554 static int 2555 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt) 2556 { 2557 void *optdata = NULL; 2558 bool malloc_optdata = false; 2559 int optdatalen = 0; 2560 int error = 0; 2561 struct in6_pktinfo null_pktinfo; 2562 int deftclass = 0, on; 2563 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2564 int defpreftemp = IP6PO_TEMPADDR_SYSTEM; 2565 struct ip6_pktopts *pktopt; 2566 2567 INP_RLOCK(inp); 2568 pktopt = inp->in6p_outputopts; 2569 2570 switch (optname) { 2571 case IPV6_PKTINFO: 2572 optdata = (void *)&null_pktinfo; 2573 if (pktopt && pktopt->ip6po_pktinfo) { 2574 bcopy(pktopt->ip6po_pktinfo, &null_pktinfo, 2575 sizeof(null_pktinfo)); 2576 in6_clearscope(&null_pktinfo.ipi6_addr); 2577 } else { 2578 /* XXX: we don't have to do this every time... */ 2579 bzero(&null_pktinfo, sizeof(null_pktinfo)); 2580 } 2581 optdatalen = sizeof(struct in6_pktinfo); 2582 break; 2583 case IPV6_TCLASS: 2584 if (pktopt && pktopt->ip6po_tclass >= 0) 2585 deftclass = pktopt->ip6po_tclass; 2586 optdata = (void *)&deftclass; 2587 optdatalen = sizeof(int); 2588 break; 2589 case IPV6_HOPOPTS: 2590 GET_PKTOPT_EXT_HDR(ip6po_hbh); 2591 break; 2592 case IPV6_RTHDR: 2593 GET_PKTOPT_EXT_HDR(ip6po_rthdr); 2594 break; 2595 case IPV6_RTHDRDSTOPTS: 2596 GET_PKTOPT_EXT_HDR(ip6po_dest1); 2597 break; 2598 case IPV6_DSTOPTS: 2599 GET_PKTOPT_EXT_HDR(ip6po_dest2); 2600 break; 2601 case IPV6_NEXTHOP: 2602 GET_PKTOPT_SOCKADDR(ip6po_nexthop); 2603 break; 2604 case IPV6_USE_MIN_MTU: 2605 if (pktopt) 2606 defminmtu = pktopt->ip6po_minmtu; 2607 optdata = (void *)&defminmtu; 2608 optdatalen = sizeof(int); 2609 break; 2610 case IPV6_DONTFRAG: 2611 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2612 on = 1; 2613 else 2614 on = 0; 2615 optdata = (void *)&on; 2616 optdatalen = sizeof(on); 2617 break; 2618 case IPV6_PREFER_TEMPADDR: 2619 if (pktopt) 2620 defpreftemp = pktopt->ip6po_prefer_tempaddr; 2621 optdata = (void *)&defpreftemp; 2622 optdatalen = sizeof(int); 2623 break; 2624 default: /* should not happen */ 2625 #ifdef DIAGNOSTIC 2626 panic("ip6_getpcbopt: unexpected option\n"); 2627 #endif 2628 INP_RUNLOCK(inp); 2629 return (ENOPROTOOPT); 2630 } 2631 INP_RUNLOCK(inp); 2632 2633 error = sooptcopyout(sopt, optdata, optdatalen); 2634 if (malloc_optdata) 2635 free(optdata, M_TEMP); 2636 2637 return (error); 2638 } 2639 2640 void 2641 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2642 { 2643 if (pktopt == NULL) 2644 return; 2645 2646 if (optname == -1 || optname == IPV6_PKTINFO) { 2647 if (pktopt->ip6po_pktinfo) 2648 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2649 pktopt->ip6po_pktinfo = NULL; 2650 } 2651 if (optname == -1 || optname == IPV6_HOPLIMIT) 2652 pktopt->ip6po_hlim = -1; 2653 if (optname == -1 || optname == IPV6_TCLASS) 2654 pktopt->ip6po_tclass = -1; 2655 if (optname == -1 || optname == IPV6_NEXTHOP) { 2656 if (pktopt->ip6po_nextroute.ro_nh) { 2657 NH_FREE(pktopt->ip6po_nextroute.ro_nh); 2658 pktopt->ip6po_nextroute.ro_nh = NULL; 2659 } 2660 if (pktopt->ip6po_nexthop) 2661 free(pktopt->ip6po_nexthop, M_IP6OPT); 2662 pktopt->ip6po_nexthop = NULL; 2663 } 2664 if (optname == -1 || optname == IPV6_HOPOPTS) { 2665 if (pktopt->ip6po_hbh) 2666 free(pktopt->ip6po_hbh, M_IP6OPT); 2667 pktopt->ip6po_hbh = NULL; 2668 } 2669 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2670 if (pktopt->ip6po_dest1) 2671 free(pktopt->ip6po_dest1, M_IP6OPT); 2672 pktopt->ip6po_dest1 = NULL; 2673 } 2674 if (optname == -1 || optname == IPV6_RTHDR) { 2675 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2676 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2677 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2678 if (pktopt->ip6po_route.ro_nh) { 2679 NH_FREE(pktopt->ip6po_route.ro_nh); 2680 pktopt->ip6po_route.ro_nh = NULL; 2681 } 2682 } 2683 if (optname == -1 || optname == IPV6_DSTOPTS) { 2684 if (pktopt->ip6po_dest2) 2685 free(pktopt->ip6po_dest2, M_IP6OPT); 2686 pktopt->ip6po_dest2 = NULL; 2687 } 2688 } 2689 2690 #define PKTOPT_EXTHDRCPY(type) \ 2691 do {\ 2692 if (src->type) {\ 2693 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2694 dst->type = malloc(hlen, M_IP6OPT, canwait);\ 2695 if (dst->type == NULL)\ 2696 goto bad;\ 2697 bcopy(src->type, dst->type, hlen);\ 2698 }\ 2699 } while (/*CONSTCOND*/ 0) 2700 2701 static int 2702 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2703 { 2704 if (dst == NULL || src == NULL) { 2705 printf("ip6_clearpktopts: invalid argument\n"); 2706 return (EINVAL); 2707 } 2708 2709 dst->ip6po_hlim = src->ip6po_hlim; 2710 dst->ip6po_tclass = src->ip6po_tclass; 2711 dst->ip6po_flags = src->ip6po_flags; 2712 dst->ip6po_minmtu = src->ip6po_minmtu; 2713 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr; 2714 if (src->ip6po_pktinfo) { 2715 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2716 M_IP6OPT, canwait); 2717 if (dst->ip6po_pktinfo == NULL) 2718 goto bad; 2719 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2720 } 2721 if (src->ip6po_nexthop) { 2722 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2723 M_IP6OPT, canwait); 2724 if (dst->ip6po_nexthop == NULL) 2725 goto bad; 2726 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop, 2727 src->ip6po_nexthop->sa_len); 2728 } 2729 PKTOPT_EXTHDRCPY(ip6po_hbh); 2730 PKTOPT_EXTHDRCPY(ip6po_dest1); 2731 PKTOPT_EXTHDRCPY(ip6po_dest2); 2732 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2733 return (0); 2734 2735 bad: 2736 ip6_clearpktopts(dst, -1); 2737 return (ENOBUFS); 2738 } 2739 #undef PKTOPT_EXTHDRCPY 2740 2741 struct ip6_pktopts * 2742 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2743 { 2744 int error; 2745 struct ip6_pktopts *dst; 2746 2747 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2748 if (dst == NULL) 2749 return (NULL); 2750 ip6_initpktopts(dst); 2751 2752 if ((error = copypktopts(dst, src, canwait)) != 0) { 2753 free(dst, M_IP6OPT); 2754 return (NULL); 2755 } 2756 2757 return (dst); 2758 } 2759 2760 void 2761 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2762 { 2763 if (pktopt == NULL) 2764 return; 2765 2766 ip6_clearpktopts(pktopt, -1); 2767 2768 free(pktopt, M_IP6OPT); 2769 } 2770 2771 /* 2772 * Set IPv6 outgoing packet options based on advanced API. 2773 */ 2774 int 2775 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2776 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto) 2777 { 2778 struct cmsghdr *cm = NULL; 2779 2780 if (control == NULL || opt == NULL) 2781 return (EINVAL); 2782 2783 ip6_initpktopts(opt); 2784 if (stickyopt) { 2785 int error; 2786 2787 /* 2788 * If stickyopt is provided, make a local copy of the options 2789 * for this particular packet, then override them by ancillary 2790 * objects. 2791 * XXX: copypktopts() does not copy the cached route to a next 2792 * hop (if any). This is not very good in terms of efficiency, 2793 * but we can allow this since this option should be rarely 2794 * used. 2795 */ 2796 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2797 return (error); 2798 } 2799 2800 /* 2801 * XXX: Currently, we assume all the optional information is stored 2802 * in a single mbuf. 2803 */ 2804 if (control->m_next) 2805 return (EINVAL); 2806 2807 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2808 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2809 int error; 2810 2811 if (control->m_len < CMSG_LEN(0)) 2812 return (EINVAL); 2813 2814 cm = mtod(control, struct cmsghdr *); 2815 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2816 return (EINVAL); 2817 if (cm->cmsg_level != IPPROTO_IPV6) 2818 continue; 2819 2820 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2821 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); 2822 if (error) 2823 return (error); 2824 } 2825 2826 return (0); 2827 } 2828 2829 /* 2830 * Set a particular packet option, as a sticky option or an ancillary data 2831 * item. "len" can be 0 only when it's a sticky option. 2832 * We have 4 cases of combination of "sticky" and "cmsg": 2833 * "sticky=0, cmsg=0": impossible 2834 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2835 * "sticky=1, cmsg=0": RFC3542 socket option 2836 * "sticky=1, cmsg=1": RFC2292 socket option 2837 */ 2838 static int 2839 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2840 struct ucred *cred, int sticky, int cmsg, int uproto) 2841 { 2842 int minmtupolicy, preftemp; 2843 int error; 2844 2845 if (!sticky && !cmsg) { 2846 #ifdef DIAGNOSTIC 2847 printf("ip6_setpktopt: impossible case\n"); 2848 #endif 2849 return (EINVAL); 2850 } 2851 2852 /* 2853 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2854 * not be specified in the context of RFC3542. Conversely, 2855 * RFC3542 types should not be specified in the context of RFC2292. 2856 */ 2857 if (!cmsg) { 2858 switch (optname) { 2859 case IPV6_2292PKTINFO: 2860 case IPV6_2292HOPLIMIT: 2861 case IPV6_2292NEXTHOP: 2862 case IPV6_2292HOPOPTS: 2863 case IPV6_2292DSTOPTS: 2864 case IPV6_2292RTHDR: 2865 case IPV6_2292PKTOPTIONS: 2866 return (ENOPROTOOPT); 2867 } 2868 } 2869 if (sticky && cmsg) { 2870 switch (optname) { 2871 case IPV6_PKTINFO: 2872 case IPV6_HOPLIMIT: 2873 case IPV6_NEXTHOP: 2874 case IPV6_HOPOPTS: 2875 case IPV6_DSTOPTS: 2876 case IPV6_RTHDRDSTOPTS: 2877 case IPV6_RTHDR: 2878 case IPV6_USE_MIN_MTU: 2879 case IPV6_DONTFRAG: 2880 case IPV6_TCLASS: 2881 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */ 2882 return (ENOPROTOOPT); 2883 } 2884 } 2885 2886 switch (optname) { 2887 case IPV6_2292PKTINFO: 2888 case IPV6_PKTINFO: 2889 { 2890 struct ifnet *ifp = NULL; 2891 struct in6_pktinfo *pktinfo; 2892 2893 if (len != sizeof(struct in6_pktinfo)) 2894 return (EINVAL); 2895 2896 pktinfo = (struct in6_pktinfo *)buf; 2897 2898 /* 2899 * An application can clear any sticky IPV6_PKTINFO option by 2900 * doing a "regular" setsockopt with ipi6_addr being 2901 * in6addr_any and ipi6_ifindex being zero. 2902 * [RFC 3542, Section 6] 2903 */ 2904 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2905 pktinfo->ipi6_ifindex == 0 && 2906 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2907 ip6_clearpktopts(opt, optname); 2908 break; 2909 } 2910 2911 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2912 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2913 return (EINVAL); 2914 } 2915 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr)) 2916 return (EINVAL); 2917 /* validate the interface index if specified. */ 2918 if (pktinfo->ipi6_ifindex > V_if_index) 2919 return (ENXIO); 2920 if (pktinfo->ipi6_ifindex) { 2921 ifp = ifnet_byindex(pktinfo->ipi6_ifindex); 2922 if (ifp == NULL) 2923 return (ENXIO); 2924 } 2925 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL || 2926 (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0)) 2927 return (ENETDOWN); 2928 2929 if (ifp != NULL && 2930 !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2931 struct in6_ifaddr *ia; 2932 2933 in6_setscope(&pktinfo->ipi6_addr, ifp, NULL); 2934 ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr); 2935 if (ia == NULL) 2936 return (EADDRNOTAVAIL); 2937 ifa_free(&ia->ia_ifa); 2938 } 2939 /* 2940 * We store the address anyway, and let in6_selectsrc() 2941 * validate the specified address. This is because ipi6_addr 2942 * may not have enough information about its scope zone, and 2943 * we may need additional information (such as outgoing 2944 * interface or the scope zone of a destination address) to 2945 * disambiguate the scope. 2946 * XXX: the delay of the validation may confuse the 2947 * application when it is used as a sticky option. 2948 */ 2949 if (opt->ip6po_pktinfo == NULL) { 2950 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2951 M_IP6OPT, M_NOWAIT); 2952 if (opt->ip6po_pktinfo == NULL) 2953 return (ENOBUFS); 2954 } 2955 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo)); 2956 break; 2957 } 2958 2959 case IPV6_2292HOPLIMIT: 2960 case IPV6_HOPLIMIT: 2961 { 2962 int *hlimp; 2963 2964 /* 2965 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2966 * to simplify the ordering among hoplimit options. 2967 */ 2968 if (optname == IPV6_HOPLIMIT && sticky) 2969 return (ENOPROTOOPT); 2970 2971 if (len != sizeof(int)) 2972 return (EINVAL); 2973 hlimp = (int *)buf; 2974 if (*hlimp < -1 || *hlimp > 255) 2975 return (EINVAL); 2976 2977 opt->ip6po_hlim = *hlimp; 2978 break; 2979 } 2980 2981 case IPV6_TCLASS: 2982 { 2983 int tclass; 2984 2985 if (len != sizeof(int)) 2986 return (EINVAL); 2987 tclass = *(int *)buf; 2988 if (tclass < -1 || tclass > 255) 2989 return (EINVAL); 2990 2991 opt->ip6po_tclass = tclass; 2992 break; 2993 } 2994 2995 case IPV6_2292NEXTHOP: 2996 case IPV6_NEXTHOP: 2997 if (cred != NULL) { 2998 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS); 2999 if (error) 3000 return (error); 3001 } 3002 3003 if (len == 0) { /* just remove the option */ 3004 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3005 break; 3006 } 3007 3008 /* check if cmsg_len is large enough for sa_len */ 3009 if (len < sizeof(struct sockaddr) || len < *buf) 3010 return (EINVAL); 3011 3012 switch (((struct sockaddr *)buf)->sa_family) { 3013 case AF_INET6: 3014 { 3015 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 3016 int error; 3017 3018 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 3019 return (EINVAL); 3020 3021 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 3022 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 3023 return (EINVAL); 3024 } 3025 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone)) 3026 != 0) { 3027 return (error); 3028 } 3029 break; 3030 } 3031 case AF_LINK: /* should eventually be supported */ 3032 default: 3033 return (EAFNOSUPPORT); 3034 } 3035 3036 /* turn off the previous option, then set the new option. */ 3037 ip6_clearpktopts(opt, IPV6_NEXTHOP); 3038 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 3039 if (opt->ip6po_nexthop == NULL) 3040 return (ENOBUFS); 3041 bcopy(buf, opt->ip6po_nexthop, *buf); 3042 break; 3043 3044 case IPV6_2292HOPOPTS: 3045 case IPV6_HOPOPTS: 3046 { 3047 struct ip6_hbh *hbh; 3048 int hbhlen; 3049 3050 /* 3051 * XXX: We don't allow a non-privileged user to set ANY HbH 3052 * options, since per-option restriction has too much 3053 * overhead. 3054 */ 3055 if (cred != NULL) { 3056 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS); 3057 if (error) 3058 return (error); 3059 } 3060 3061 if (len == 0) { 3062 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3063 break; /* just remove the option */ 3064 } 3065 3066 /* message length validation */ 3067 if (len < sizeof(struct ip6_hbh)) 3068 return (EINVAL); 3069 hbh = (struct ip6_hbh *)buf; 3070 hbhlen = (hbh->ip6h_len + 1) << 3; 3071 if (len != hbhlen) 3072 return (EINVAL); 3073 3074 /* turn off the previous option, then set the new option. */ 3075 ip6_clearpktopts(opt, IPV6_HOPOPTS); 3076 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 3077 if (opt->ip6po_hbh == NULL) 3078 return (ENOBUFS); 3079 bcopy(hbh, opt->ip6po_hbh, hbhlen); 3080 3081 break; 3082 } 3083 3084 case IPV6_2292DSTOPTS: 3085 case IPV6_DSTOPTS: 3086 case IPV6_RTHDRDSTOPTS: 3087 { 3088 struct ip6_dest *dest, **newdest = NULL; 3089 int destlen; 3090 3091 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */ 3092 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS); 3093 if (error) 3094 return (error); 3095 } 3096 3097 if (len == 0) { 3098 ip6_clearpktopts(opt, optname); 3099 break; /* just remove the option */ 3100 } 3101 3102 /* message length validation */ 3103 if (len < sizeof(struct ip6_dest)) 3104 return (EINVAL); 3105 dest = (struct ip6_dest *)buf; 3106 destlen = (dest->ip6d_len + 1) << 3; 3107 if (len != destlen) 3108 return (EINVAL); 3109 3110 /* 3111 * Determine the position that the destination options header 3112 * should be inserted; before or after the routing header. 3113 */ 3114 switch (optname) { 3115 case IPV6_2292DSTOPTS: 3116 /* 3117 * The old advacned API is ambiguous on this point. 3118 * Our approach is to determine the position based 3119 * according to the existence of a routing header. 3120 * Note, however, that this depends on the order of the 3121 * extension headers in the ancillary data; the 1st 3122 * part of the destination options header must appear 3123 * before the routing header in the ancillary data, 3124 * too. 3125 * RFC3542 solved the ambiguity by introducing 3126 * separate ancillary data or option types. 3127 */ 3128 if (opt->ip6po_rthdr == NULL) 3129 newdest = &opt->ip6po_dest1; 3130 else 3131 newdest = &opt->ip6po_dest2; 3132 break; 3133 case IPV6_RTHDRDSTOPTS: 3134 newdest = &opt->ip6po_dest1; 3135 break; 3136 case IPV6_DSTOPTS: 3137 newdest = &opt->ip6po_dest2; 3138 break; 3139 } 3140 3141 /* turn off the previous option, then set the new option. */ 3142 ip6_clearpktopts(opt, optname); 3143 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 3144 if (*newdest == NULL) 3145 return (ENOBUFS); 3146 bcopy(dest, *newdest, destlen); 3147 3148 break; 3149 } 3150 3151 case IPV6_2292RTHDR: 3152 case IPV6_RTHDR: 3153 { 3154 struct ip6_rthdr *rth; 3155 int rthlen; 3156 3157 if (len == 0) { 3158 ip6_clearpktopts(opt, IPV6_RTHDR); 3159 break; /* just remove the option */ 3160 } 3161 3162 /* message length validation */ 3163 if (len < sizeof(struct ip6_rthdr)) 3164 return (EINVAL); 3165 rth = (struct ip6_rthdr *)buf; 3166 rthlen = (rth->ip6r_len + 1) << 3; 3167 if (len != rthlen) 3168 return (EINVAL); 3169 3170 switch (rth->ip6r_type) { 3171 case IPV6_RTHDR_TYPE_0: 3172 if (rth->ip6r_len == 0) /* must contain one addr */ 3173 return (EINVAL); 3174 if (rth->ip6r_len % 2) /* length must be even */ 3175 return (EINVAL); 3176 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 3177 return (EINVAL); 3178 break; 3179 default: 3180 return (EINVAL); /* not supported */ 3181 } 3182 3183 /* turn off the previous option */ 3184 ip6_clearpktopts(opt, IPV6_RTHDR); 3185 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 3186 if (opt->ip6po_rthdr == NULL) 3187 return (ENOBUFS); 3188 bcopy(rth, opt->ip6po_rthdr, rthlen); 3189 3190 break; 3191 } 3192 3193 case IPV6_USE_MIN_MTU: 3194 if (len != sizeof(int)) 3195 return (EINVAL); 3196 minmtupolicy = *(int *)buf; 3197 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 3198 minmtupolicy != IP6PO_MINMTU_DISABLE && 3199 minmtupolicy != IP6PO_MINMTU_ALL) { 3200 return (EINVAL); 3201 } 3202 opt->ip6po_minmtu = minmtupolicy; 3203 break; 3204 3205 case IPV6_DONTFRAG: 3206 if (len != sizeof(int)) 3207 return (EINVAL); 3208 3209 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 3210 /* 3211 * we ignore this option for TCP sockets. 3212 * (RFC3542 leaves this case unspecified.) 3213 */ 3214 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 3215 } else 3216 opt->ip6po_flags |= IP6PO_DONTFRAG; 3217 break; 3218 3219 case IPV6_PREFER_TEMPADDR: 3220 if (len != sizeof(int)) 3221 return (EINVAL); 3222 preftemp = *(int *)buf; 3223 if (preftemp != IP6PO_TEMPADDR_SYSTEM && 3224 preftemp != IP6PO_TEMPADDR_NOTPREFER && 3225 preftemp != IP6PO_TEMPADDR_PREFER) { 3226 return (EINVAL); 3227 } 3228 opt->ip6po_prefer_tempaddr = preftemp; 3229 break; 3230 3231 default: 3232 return (ENOPROTOOPT); 3233 } /* end of switch */ 3234 3235 return (0); 3236 } 3237 3238 /* 3239 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3240 * packet to the input queue of a specified interface. Note that this 3241 * calls the output routine of the loopback "driver", but with an interface 3242 * pointer that might NOT be &loif -- easier than replicating that code here. 3243 */ 3244 void 3245 ip6_mloopback(struct ifnet *ifp, struct mbuf *m) 3246 { 3247 struct mbuf *copym; 3248 struct ip6_hdr *ip6; 3249 3250 copym = m_copym(m, 0, M_COPYALL, M_NOWAIT); 3251 if (copym == NULL) 3252 return; 3253 3254 /* 3255 * Make sure to deep-copy IPv6 header portion in case the data 3256 * is in an mbuf cluster, so that we can safely override the IPv6 3257 * header portion later. 3258 */ 3259 if (!M_WRITABLE(copym) || 3260 copym->m_len < sizeof(struct ip6_hdr)) { 3261 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3262 if (copym == NULL) 3263 return; 3264 } 3265 ip6 = mtod(copym, struct ip6_hdr *); 3266 /* 3267 * clear embedded scope identifiers if necessary. 3268 * in6_clearscope will touch the addresses only when necessary. 3269 */ 3270 in6_clearscope(&ip6->ip6_src); 3271 in6_clearscope(&ip6->ip6_dst); 3272 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 3273 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 | 3274 CSUM_PSEUDO_HDR; 3275 copym->m_pkthdr.csum_data = 0xffff; 3276 } 3277 if_simloop(ifp, copym, AF_INET6, 0); 3278 } 3279 3280 /* 3281 * Chop IPv6 header off from the payload. 3282 */ 3283 static int 3284 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3285 { 3286 struct mbuf *mh; 3287 struct ip6_hdr *ip6; 3288 3289 ip6 = mtod(m, struct ip6_hdr *); 3290 if (m->m_len > sizeof(*ip6)) { 3291 mh = m_gethdr(M_NOWAIT, MT_DATA); 3292 if (mh == NULL) { 3293 m_freem(m); 3294 return ENOBUFS; 3295 } 3296 m_move_pkthdr(mh, m); 3297 M_ALIGN(mh, sizeof(*ip6)); 3298 m->m_len -= sizeof(*ip6); 3299 m->m_data += sizeof(*ip6); 3300 mh->m_next = m; 3301 m = mh; 3302 m->m_len = sizeof(*ip6); 3303 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 3304 } 3305 exthdrs->ip6e_ip6 = m; 3306 return 0; 3307 } 3308 3309 /* 3310 * Compute IPv6 extension header length. 3311 */ 3312 int 3313 ip6_optlen(struct inpcb *inp) 3314 { 3315 int len; 3316 3317 if (!inp->in6p_outputopts) 3318 return 0; 3319 3320 len = 0; 3321 #define elen(x) \ 3322 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3323 3324 len += elen(inp->in6p_outputopts->ip6po_hbh); 3325 if (inp->in6p_outputopts->ip6po_rthdr) 3326 /* dest1 is valid with rthdr only */ 3327 len += elen(inp->in6p_outputopts->ip6po_dest1); 3328 len += elen(inp->in6p_outputopts->ip6po_rthdr); 3329 len += elen(inp->in6p_outputopts->ip6po_dest2); 3330 return len; 3331 #undef elen 3332 } 3333