1 /*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $ 30 */ 31 32 /*- 33 * Copyright (c) 1982, 1986, 1988, 1990, 1993 34 * The Regents of the University of California. All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 4. Neither the name of the University nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 * 60 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 61 */ 62 63 #include <sys/cdefs.h> 64 __FBSDID("$FreeBSD$"); 65 66 #include "opt_inet.h" 67 #include "opt_inet6.h" 68 #include "opt_ratelimit.h" 69 #include "opt_ipsec.h" 70 #include "opt_sctp.h" 71 #include "opt_route.h" 72 #include "opt_rss.h" 73 74 #include <sys/param.h> 75 #include <sys/kernel.h> 76 #include <sys/malloc.h> 77 #include <sys/mbuf.h> 78 #include <sys/errno.h> 79 #include <sys/priv.h> 80 #include <sys/proc.h> 81 #include <sys/protosw.h> 82 #include <sys/socket.h> 83 #include <sys/socketvar.h> 84 #include <sys/syslog.h> 85 #include <sys/ucred.h> 86 87 #include <machine/in_cksum.h> 88 89 #include <net/if.h> 90 #include <net/if_var.h> 91 #include <net/if_llatbl.h> 92 #include <net/netisr.h> 93 #include <net/route.h> 94 #include <net/pfil.h> 95 #include <net/rss_config.h> 96 #include <net/vnet.h> 97 98 #include <netinet/in.h> 99 #include <netinet/in_var.h> 100 #include <netinet/ip_var.h> 101 #include <netinet6/in6_fib.h> 102 #include <netinet6/in6_var.h> 103 #include <netinet/ip6.h> 104 #include <netinet/icmp6.h> 105 #include <netinet6/ip6_var.h> 106 #include <netinet/in_pcb.h> 107 #include <netinet/tcp_var.h> 108 #include <netinet6/nd6.h> 109 #include <netinet6/in6_rss.h> 110 111 #include <netipsec/ipsec_support.h> 112 #ifdef SCTP 113 #include <netinet/sctp.h> 114 #include <netinet/sctp_crc32.h> 115 #endif 116 117 #include <netinet6/ip6protosw.h> 118 #include <netinet6/scope6_var.h> 119 120 #ifdef FLOWTABLE 121 #include <net/flowtable.h> 122 #endif 123 124 extern int in6_mcast_loop; 125 126 struct ip6_exthdrs { 127 struct mbuf *ip6e_ip6; 128 struct mbuf *ip6e_hbh; 129 struct mbuf *ip6e_dest1; 130 struct mbuf *ip6e_rthdr; 131 struct mbuf *ip6e_dest2; 132 }; 133 134 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options"); 135 136 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, 137 struct ucred *, int); 138 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *, 139 struct socket *, struct sockopt *); 140 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *); 141 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, 142 struct ucred *, int, int, int); 143 144 static int ip6_copyexthdr(struct mbuf **, caddr_t, int); 145 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 146 struct ip6_frag **); 147 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); 148 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); 149 static int ip6_getpmtu(struct route_in6 *, int, 150 struct ifnet *, const struct in6_addr *, u_long *, int *, u_int, 151 u_int); 152 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long, 153 u_long *, int *, u_int); 154 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *); 155 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int); 156 157 158 /* 159 * Make an extension header from option data. hp is the source, and 160 * mp is the destination. 161 */ 162 #define MAKE_EXTHDR(hp, mp) \ 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 } \ 171 } while (/*CONSTCOND*/ 0) 172 173 /* 174 * Form a chain of extension headers. 175 * m is the extension header mbuf 176 * mp is the previous mbuf in the chain 177 * p is the next header 178 * i is the type of option. 179 */ 180 #define MAKE_CHAIN(m, mp, p, i)\ 181 do {\ 182 if (m) {\ 183 if (!hdrsplit) \ 184 panic("assumption failed: hdr not split"); \ 185 *mtod((m), u_char *) = *(p);\ 186 *(p) = (i);\ 187 p = mtod((m), u_char *);\ 188 (m)->m_next = (mp)->m_next;\ 189 (mp)->m_next = (m);\ 190 (mp) = (m);\ 191 }\ 192 } while (/*CONSTCOND*/ 0) 193 194 void 195 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset) 196 { 197 u_short csum; 198 199 csum = in_cksum_skip(m, offset + plen, offset); 200 if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0) 201 csum = 0xffff; 202 offset += m->m_pkthdr.csum_data; /* checksum offset */ 203 204 if (offset + sizeof(u_short) > m->m_len) { 205 printf("%s: delayed m_pullup, m->len: %d plen %u off %u " 206 "csum_flags=%b\n", __func__, m->m_len, plen, offset, 207 (int)m->m_pkthdr.csum_flags, CSUM_BITS); 208 /* 209 * XXX this should not happen, but if it does, the correct 210 * behavior may be to insert the checksum in the appropriate 211 * next mbuf in the chain. 212 */ 213 return; 214 } 215 *(u_short *)(m->m_data + offset) = csum; 216 } 217 218 int 219 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto, 220 int mtu, uint32_t id) 221 { 222 struct mbuf *m, **mnext, *m_frgpart; 223 struct ip6_hdr *ip6, *mhip6; 224 struct ip6_frag *ip6f; 225 int off; 226 int error; 227 int tlen = m0->m_pkthdr.len; 228 229 m = m0; 230 ip6 = mtod(m, struct ip6_hdr *); 231 mnext = &m->m_nextpkt; 232 233 for (off = hlen; off < tlen; off += mtu) { 234 m = m_gethdr(M_NOWAIT, MT_DATA); 235 if (!m) { 236 IP6STAT_INC(ip6s_odropped); 237 return (ENOBUFS); 238 } 239 m->m_flags = m0->m_flags & M_COPYFLAGS; 240 *mnext = m; 241 mnext = &m->m_nextpkt; 242 m->m_data += max_linkhdr; 243 mhip6 = mtod(m, struct ip6_hdr *); 244 *mhip6 = *ip6; 245 m->m_len = sizeof(*mhip6); 246 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 247 if (error) { 248 IP6STAT_INC(ip6s_odropped); 249 return (error); 250 } 251 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7)); 252 if (off + mtu >= tlen) 253 mtu = tlen - off; 254 else 255 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 256 mhip6->ip6_plen = htons((u_short)(mtu + hlen + 257 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 258 if ((m_frgpart = m_copym(m0, off, mtu, M_NOWAIT)) == NULL) { 259 IP6STAT_INC(ip6s_odropped); 260 return (ENOBUFS); 261 } 262 m_cat(m, m_frgpart); 263 m->m_pkthdr.len = mtu + hlen + sizeof(*ip6f); 264 m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum; 265 m->m_pkthdr.rcvif = NULL; 266 ip6f->ip6f_reserved = 0; 267 ip6f->ip6f_ident = id; 268 ip6f->ip6f_nxt = nextproto; 269 IP6STAT_INC(ip6s_ofragments); 270 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 271 } 272 273 return (0); 274 } 275 276 /* 277 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 278 * header (with pri, len, nxt, hlim, src, dst). 279 * This function may modify ver and hlim only. 280 * The mbuf chain containing the packet will be freed. 281 * The mbuf opt, if present, will not be freed. 282 * If route_in6 ro is present and has ro_rt initialized, route lookup would be 283 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL, 284 * then result of route lookup is stored in ro->ro_rt. 285 * 286 * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and 287 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 288 * which is rt_mtu. 289 * 290 * ifpp - XXX: just for statistics 291 */ 292 /* 293 * XXX TODO: no flowid is assigned for outbound flows? 294 */ 295 int 296 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, 297 struct route_in6 *ro, int flags, struct ip6_moptions *im6o, 298 struct ifnet **ifpp, struct inpcb *inp) 299 { 300 struct ip6_hdr *ip6; 301 struct ifnet *ifp, *origifp; 302 struct mbuf *m = m0; 303 struct mbuf *mprev = NULL; 304 int hlen, tlen, len; 305 struct route_in6 ip6route; 306 struct rtentry *rt = NULL; 307 struct sockaddr_in6 *dst, src_sa, dst_sa; 308 struct in6_addr odst; 309 int error = 0; 310 struct in6_ifaddr *ia = NULL; 311 u_long mtu; 312 int alwaysfrag, dontfrag; 313 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 314 struct ip6_exthdrs exthdrs; 315 struct in6_addr src0, dst0; 316 u_int32_t zone; 317 struct route_in6 *ro_pmtu = NULL; 318 int hdrsplit = 0; 319 int sw_csum, tso; 320 int needfiblookup; 321 uint32_t fibnum; 322 struct m_tag *fwd_tag = NULL; 323 uint32_t id; 324 325 if (inp != NULL) { 326 M_SETFIB(m, inp->inp_inc.inc_fibnum); 327 if ((flags & IP_NODEFAULTFLOWID) == 0) { 328 /* unconditionally set flowid */ 329 m->m_pkthdr.flowid = inp->inp_flowid; 330 M_HASHTYPE_SET(m, inp->inp_flowtype); 331 } 332 } 333 334 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 335 /* 336 * IPSec checking which handles several cases. 337 * FAST IPSEC: We re-injected the packet. 338 * XXX: need scope argument. 339 */ 340 if (IPSEC_ENABLED(ipv6)) { 341 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) { 342 if (error == EINPROGRESS) 343 error = 0; 344 goto done; 345 } 346 } 347 #endif /* IPSEC */ 348 349 bzero(&exthdrs, sizeof(exthdrs)); 350 if (opt) { 351 /* Hop-by-Hop options header */ 352 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 353 /* Destination options header(1st part) */ 354 if (opt->ip6po_rthdr) { 355 /* 356 * Destination options header(1st part) 357 * This only makes sense with a routing header. 358 * See Section 9.2 of RFC 3542. 359 * Disabling this part just for MIP6 convenience is 360 * a bad idea. We need to think carefully about a 361 * way to make the advanced API coexist with MIP6 362 * options, which might automatically be inserted in 363 * the kernel. 364 */ 365 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 366 } 367 /* Routing header */ 368 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 369 /* Destination options header(2nd part) */ 370 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 371 } 372 373 /* 374 * Calculate the total length of the extension header chain. 375 * Keep the length of the unfragmentable part for fragmentation. 376 */ 377 optlen = 0; 378 if (exthdrs.ip6e_hbh) 379 optlen += exthdrs.ip6e_hbh->m_len; 380 if (exthdrs.ip6e_dest1) 381 optlen += exthdrs.ip6e_dest1->m_len; 382 if (exthdrs.ip6e_rthdr) 383 optlen += exthdrs.ip6e_rthdr->m_len; 384 unfragpartlen = optlen + sizeof(struct ip6_hdr); 385 386 /* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */ 387 if (exthdrs.ip6e_dest2) 388 optlen += exthdrs.ip6e_dest2->m_len; 389 390 /* 391 * If there is at least one extension header, 392 * separate IP6 header from the payload. 393 */ 394 if (optlen && !hdrsplit) { 395 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 396 m = NULL; 397 goto freehdrs; 398 } 399 m = exthdrs.ip6e_ip6; 400 hdrsplit++; 401 } 402 403 ip6 = mtod(m, struct ip6_hdr *); 404 405 /* adjust mbuf packet header length */ 406 m->m_pkthdr.len += optlen; 407 plen = m->m_pkthdr.len - sizeof(*ip6); 408 409 /* If this is a jumbo payload, insert a jumbo payload option. */ 410 if (plen > IPV6_MAXPACKET) { 411 if (!hdrsplit) { 412 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 413 m = NULL; 414 goto freehdrs; 415 } 416 m = exthdrs.ip6e_ip6; 417 hdrsplit++; 418 } 419 /* adjust pointer */ 420 ip6 = mtod(m, struct ip6_hdr *); 421 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 422 goto freehdrs; 423 ip6->ip6_plen = 0; 424 } else 425 ip6->ip6_plen = htons(plen); 426 427 /* 428 * Concatenate headers and fill in next header fields. 429 * Here we have, on "m" 430 * IPv6 payload 431 * and we insert headers accordingly. Finally, we should be getting: 432 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 433 * 434 * during the header composing process, "m" points to IPv6 header. 435 * "mprev" points to an extension header prior to esp. 436 */ 437 u_char *nexthdrp = &ip6->ip6_nxt; 438 mprev = m; 439 440 /* 441 * we treat dest2 specially. this makes IPsec processing 442 * much easier. the goal here is to make mprev point the 443 * mbuf prior to dest2. 444 * 445 * result: IPv6 dest2 payload 446 * m and mprev will point to IPv6 header. 447 */ 448 if (exthdrs.ip6e_dest2) { 449 if (!hdrsplit) 450 panic("assumption failed: hdr not split"); 451 exthdrs.ip6e_dest2->m_next = m->m_next; 452 m->m_next = exthdrs.ip6e_dest2; 453 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 454 ip6->ip6_nxt = IPPROTO_DSTOPTS; 455 } 456 457 /* 458 * result: IPv6 hbh dest1 rthdr dest2 payload 459 * m will point to IPv6 header. mprev will point to the 460 * extension header prior to dest2 (rthdr in the above case). 461 */ 462 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 463 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 464 IPPROTO_DSTOPTS); 465 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 466 IPPROTO_ROUTING); 467 468 /* 469 * If there is a routing header, discard the packet. 470 */ 471 if (exthdrs.ip6e_rthdr) { 472 error = EINVAL; 473 goto bad; 474 } 475 476 /* Source address validation */ 477 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 478 (flags & IPV6_UNSPECSRC) == 0) { 479 error = EOPNOTSUPP; 480 IP6STAT_INC(ip6s_badscope); 481 goto bad; 482 } 483 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 484 error = EOPNOTSUPP; 485 IP6STAT_INC(ip6s_badscope); 486 goto bad; 487 } 488 489 IP6STAT_INC(ip6s_localout); 490 491 /* 492 * Route packet. 493 */ 494 if (ro == NULL) { 495 ro = &ip6route; 496 bzero((caddr_t)ro, sizeof(*ro)); 497 } else 498 ro->ro_flags |= RT_LLE_CACHE; 499 ro_pmtu = ro; 500 if (opt && opt->ip6po_rthdr) 501 ro = &opt->ip6po_route; 502 dst = (struct sockaddr_in6 *)&ro->ro_dst; 503 #ifdef FLOWTABLE 504 if (ro->ro_rt == NULL) 505 (void )flowtable_lookup(AF_INET6, m, (struct route *)ro); 506 #endif 507 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m); 508 again: 509 /* 510 * if specified, try to fill in the traffic class field. 511 * do not override if a non-zero value is already set. 512 * we check the diffserv field and the ecn field separately. 513 */ 514 if (opt && opt->ip6po_tclass >= 0) { 515 int mask = 0; 516 517 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 518 mask |= 0xfc; 519 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 520 mask |= 0x03; 521 if (mask != 0) 522 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 523 } 524 525 /* fill in or override the hop limit field, if necessary. */ 526 if (opt && opt->ip6po_hlim != -1) 527 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 528 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 529 if (im6o != NULL) 530 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 531 else 532 ip6->ip6_hlim = V_ip6_defmcasthlim; 533 } 534 /* 535 * Validate route against routing table additions; 536 * a better/more specific route might have been added. 537 * Make sure address family is set in route. 538 */ 539 if (inp) { 540 ro->ro_dst.sin6_family = AF_INET6; 541 RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum); 542 } 543 if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) && 544 ro->ro_dst.sin6_family == AF_INET6 && 545 IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) { 546 rt = ro->ro_rt; 547 ifp = ro->ro_rt->rt_ifp; 548 } else { 549 if (ro->ro_lle) 550 LLE_FREE(ro->ro_lle); /* zeros ro_lle */ 551 ro->ro_lle = NULL; 552 if (fwd_tag == NULL) { 553 bzero(&dst_sa, sizeof(dst_sa)); 554 dst_sa.sin6_family = AF_INET6; 555 dst_sa.sin6_len = sizeof(dst_sa); 556 dst_sa.sin6_addr = ip6->ip6_dst; 557 } 558 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp, 559 &rt, fibnum); 560 if (error != 0) { 561 if (ifp != NULL) 562 in6_ifstat_inc(ifp, ifs6_out_discard); 563 goto bad; 564 } 565 } 566 if (rt == NULL) { 567 /* 568 * If in6_selectroute() does not return a route entry, 569 * dst may not have been updated. 570 */ 571 *dst = dst_sa; /* XXX */ 572 } 573 574 /* 575 * then rt (for unicast) and ifp must be non-NULL valid values. 576 */ 577 if ((flags & IPV6_FORWARDING) == 0) { 578 /* XXX: the FORWARDING flag can be set for mrouting. */ 579 in6_ifstat_inc(ifp, ifs6_out_request); 580 } 581 if (rt != NULL) { 582 ia = (struct in6_ifaddr *)(rt->rt_ifa); 583 counter_u64_add(rt->rt_pksent, 1); 584 } 585 586 587 /* 588 * The outgoing interface must be in the zone of source and 589 * destination addresses. 590 */ 591 origifp = ifp; 592 593 src0 = ip6->ip6_src; 594 if (in6_setscope(&src0, origifp, &zone)) 595 goto badscope; 596 bzero(&src_sa, sizeof(src_sa)); 597 src_sa.sin6_family = AF_INET6; 598 src_sa.sin6_len = sizeof(src_sa); 599 src_sa.sin6_addr = ip6->ip6_src; 600 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 601 goto badscope; 602 603 dst0 = ip6->ip6_dst; 604 if (in6_setscope(&dst0, origifp, &zone)) 605 goto badscope; 606 /* re-initialize to be sure */ 607 bzero(&dst_sa, sizeof(dst_sa)); 608 dst_sa.sin6_family = AF_INET6; 609 dst_sa.sin6_len = sizeof(dst_sa); 610 dst_sa.sin6_addr = ip6->ip6_dst; 611 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) { 612 goto badscope; 613 } 614 615 /* We should use ia_ifp to support the case of 616 * sending packets to an address of our own. 617 */ 618 if (ia != NULL && ia->ia_ifp) 619 ifp = ia->ia_ifp; 620 621 /* scope check is done. */ 622 goto routefound; 623 624 badscope: 625 IP6STAT_INC(ip6s_badscope); 626 in6_ifstat_inc(origifp, ifs6_out_discard); 627 if (error == 0) 628 error = EHOSTUNREACH; /* XXX */ 629 goto bad; 630 631 routefound: 632 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 633 if (opt && opt->ip6po_nextroute.ro_rt) { 634 /* 635 * The nexthop is explicitly specified by the 636 * application. We assume the next hop is an IPv6 637 * address. 638 */ 639 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 640 } 641 else if ((rt->rt_flags & RTF_GATEWAY)) 642 dst = (struct sockaddr_in6 *)rt->rt_gateway; 643 } 644 645 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 646 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 647 } else { 648 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 649 in6_ifstat_inc(ifp, ifs6_out_mcast); 650 /* 651 * Confirm that the outgoing interface supports multicast. 652 */ 653 if (!(ifp->if_flags & IFF_MULTICAST)) { 654 IP6STAT_INC(ip6s_noroute); 655 in6_ifstat_inc(ifp, ifs6_out_discard); 656 error = ENETUNREACH; 657 goto bad; 658 } 659 if ((im6o == NULL && in6_mcast_loop) || 660 (im6o && im6o->im6o_multicast_loop)) { 661 /* 662 * Loop back multicast datagram if not expressly 663 * forbidden to do so, even if we have not joined 664 * the address; protocols will filter it later, 665 * thus deferring a hash lookup and lock acquisition 666 * at the expense of an m_copym(). 667 */ 668 ip6_mloopback(ifp, m); 669 } else { 670 /* 671 * If we are acting as a multicast router, perform 672 * multicast forwarding as if the packet had just 673 * arrived on the interface to which we are about 674 * to send. The multicast forwarding function 675 * recursively calls this function, using the 676 * IPV6_FORWARDING flag to prevent infinite recursion. 677 * 678 * Multicasts that are looped back by ip6_mloopback(), 679 * above, will be forwarded by the ip6_input() routine, 680 * if necessary. 681 */ 682 if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 683 /* 684 * XXX: ip6_mforward expects that rcvif is NULL 685 * when it is called from the originating path. 686 * However, it may not always be the case. 687 */ 688 m->m_pkthdr.rcvif = NULL; 689 if (ip6_mforward(ip6, ifp, m) != 0) { 690 m_freem(m); 691 goto done; 692 } 693 } 694 } 695 /* 696 * Multicasts with a hoplimit of zero may be looped back, 697 * above, but must not be transmitted on a network. 698 * Also, multicasts addressed to the loopback interface 699 * are not sent -- the above call to ip6_mloopback() will 700 * loop back a copy if this host actually belongs to the 701 * destination group on the loopback interface. 702 */ 703 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 704 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 705 m_freem(m); 706 goto done; 707 } 708 } 709 710 /* 711 * Fill the outgoing inteface to tell the upper layer 712 * to increment per-interface statistics. 713 */ 714 if (ifpp) 715 *ifpp = ifp; 716 717 /* Determine path MTU. */ 718 if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst, 719 &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0) 720 goto bad; 721 722 /* 723 * The caller of this function may specify to use the minimum MTU 724 * in some cases. 725 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 726 * setting. The logic is a bit complicated; by default, unicast 727 * packets will follow path MTU while multicast packets will be sent at 728 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 729 * including unicast ones will be sent at the minimum MTU. Multicast 730 * packets will always be sent at the minimum MTU unless 731 * IP6PO_MINMTU_DISABLE is explicitly specified. 732 * See RFC 3542 for more details. 733 */ 734 if (mtu > IPV6_MMTU) { 735 if ((flags & IPV6_MINMTU)) 736 mtu = IPV6_MMTU; 737 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 738 mtu = IPV6_MMTU; 739 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 740 (opt == NULL || 741 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 742 mtu = IPV6_MMTU; 743 } 744 } 745 746 /* 747 * clear embedded scope identifiers if necessary. 748 * in6_clearscope will touch the addresses only when necessary. 749 */ 750 in6_clearscope(&ip6->ip6_src); 751 in6_clearscope(&ip6->ip6_dst); 752 753 /* 754 * If the outgoing packet contains a hop-by-hop options header, 755 * it must be examined and processed even by the source node. 756 * (RFC 2460, section 4.) 757 */ 758 if (exthdrs.ip6e_hbh) { 759 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 760 u_int32_t dummy; /* XXX unused */ 761 u_int32_t plen = 0; /* XXX: ip6_process will check the value */ 762 763 #ifdef DIAGNOSTIC 764 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len) 765 panic("ip6e_hbh is not contiguous"); 766 #endif 767 /* 768 * XXX: if we have to send an ICMPv6 error to the sender, 769 * we need the M_LOOP flag since icmp6_error() expects 770 * the IPv6 and the hop-by-hop options header are 771 * contiguous unless the flag is set. 772 */ 773 m->m_flags |= M_LOOP; 774 m->m_pkthdr.rcvif = ifp; 775 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 776 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 777 &dummy, &plen) < 0) { 778 /* m was already freed at this point */ 779 error = EINVAL;/* better error? */ 780 goto done; 781 } 782 m->m_flags &= ~M_LOOP; /* XXX */ 783 m->m_pkthdr.rcvif = NULL; 784 } 785 786 /* Jump over all PFIL processing if hooks are not active. */ 787 if (!PFIL_HOOKED(&V_inet6_pfil_hook)) 788 goto passout; 789 790 odst = ip6->ip6_dst; 791 /* Run through list of hooks for output packets. */ 792 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp); 793 if (error != 0 || m == NULL) 794 goto done; 795 /* adjust pointer */ 796 ip6 = mtod(m, struct ip6_hdr *); 797 798 needfiblookup = 0; 799 /* See if destination IP address was changed by packet filter. */ 800 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) { 801 m->m_flags |= M_SKIP_FIREWALL; 802 /* If destination is now ourself drop to ip6_input(). */ 803 if (in6_localip(&ip6->ip6_dst)) { 804 m->m_flags |= M_FASTFWD_OURS; 805 if (m->m_pkthdr.rcvif == NULL) 806 m->m_pkthdr.rcvif = V_loif; 807 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 808 m->m_pkthdr.csum_flags |= 809 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; 810 m->m_pkthdr.csum_data = 0xffff; 811 } 812 #ifdef SCTP 813 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) 814 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 815 #endif 816 error = netisr_queue(NETISR_IPV6, m); 817 goto done; 818 } else { 819 RO_RTFREE(ro); 820 needfiblookup = 1; /* Redo the routing table lookup. */ 821 if (ro->ro_lle) 822 LLE_FREE(ro->ro_lle); /* zeros ro_lle */ 823 ro->ro_lle = NULL; 824 } 825 } 826 /* See if fib was changed by packet filter. */ 827 if (fibnum != M_GETFIB(m)) { 828 m->m_flags |= M_SKIP_FIREWALL; 829 fibnum = M_GETFIB(m); 830 RO_RTFREE(ro); 831 needfiblookup = 1; 832 if (ro->ro_lle) 833 LLE_FREE(ro->ro_lle); /* zeros ro_lle */ 834 ro->ro_lle = NULL; 835 } 836 if (needfiblookup) 837 goto again; 838 839 /* See if local, if yes, send it to netisr. */ 840 if (m->m_flags & M_FASTFWD_OURS) { 841 if (m->m_pkthdr.rcvif == NULL) 842 m->m_pkthdr.rcvif = V_loif; 843 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 844 m->m_pkthdr.csum_flags |= 845 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; 846 m->m_pkthdr.csum_data = 0xffff; 847 } 848 #ifdef SCTP 849 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) 850 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 851 #endif 852 error = netisr_queue(NETISR_IPV6, m); 853 goto done; 854 } 855 /* Or forward to some other address? */ 856 if ((m->m_flags & M_IP6_NEXTHOP) && 857 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { 858 dst = (struct sockaddr_in6 *)&ro->ro_dst; 859 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6)); 860 m->m_flags |= M_SKIP_FIREWALL; 861 m->m_flags &= ~M_IP6_NEXTHOP; 862 m_tag_delete(m, fwd_tag); 863 goto again; 864 } 865 866 passout: 867 /* 868 * Send the packet to the outgoing interface. 869 * If necessary, do IPv6 fragmentation before sending. 870 * 871 * the logic here is rather complex: 872 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 873 * 1-a: send as is if tlen <= path mtu 874 * 1-b: fragment if tlen > path mtu 875 * 876 * 2: if user asks us not to fragment (dontfrag == 1) 877 * 2-a: send as is if tlen <= interface mtu 878 * 2-b: error if tlen > interface mtu 879 * 880 * 3: if we always need to attach fragment header (alwaysfrag == 1) 881 * always fragment 882 * 883 * 4: if dontfrag == 1 && alwaysfrag == 1 884 * error, as we cannot handle this conflicting request 885 */ 886 sw_csum = m->m_pkthdr.csum_flags; 887 if (!hdrsplit) { 888 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0; 889 sw_csum &= ~ifp->if_hwassist; 890 } else 891 tso = 0; 892 /* 893 * If we added extension headers, we will not do TSO and calculate the 894 * checksums ourselves for now. 895 * XXX-BZ Need a framework to know when the NIC can handle it, even 896 * with ext. hdrs. 897 */ 898 if (sw_csum & CSUM_DELAY_DATA_IPV6) { 899 sw_csum &= ~CSUM_DELAY_DATA_IPV6; 900 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr)); 901 } 902 #ifdef SCTP 903 if (sw_csum & CSUM_SCTP_IPV6) { 904 sw_csum &= ~CSUM_SCTP_IPV6; 905 sctp_delayed_cksum(m, sizeof(struct ip6_hdr)); 906 } 907 #endif 908 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 909 tlen = m->m_pkthdr.len; 910 911 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso) 912 dontfrag = 1; 913 else 914 dontfrag = 0; 915 if (dontfrag && alwaysfrag) { /* case 4 */ 916 /* conflicting request - can't transmit */ 917 error = EMSGSIZE; 918 goto bad; 919 } 920 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* case 2-b */ 921 /* 922 * Even if the DONTFRAG option is specified, we cannot send the 923 * packet when the data length is larger than the MTU of the 924 * outgoing interface. 925 * Notify the error by sending IPV6_PATHMTU ancillary data if 926 * application wanted to know the MTU value. Also return an 927 * error code (this is not described in the API spec). 928 */ 929 if (inp != NULL) 930 ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu); 931 error = EMSGSIZE; 932 goto bad; 933 } 934 935 /* 936 * transmit packet without fragmentation 937 */ 938 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */ 939 struct in6_ifaddr *ia6; 940 941 ip6 = mtod(m, struct ip6_hdr *); 942 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 943 if (ia6) { 944 /* Record statistics for this interface address. */ 945 counter_u64_add(ia6->ia_ifa.ifa_opackets, 1); 946 counter_u64_add(ia6->ia_ifa.ifa_obytes, 947 m->m_pkthdr.len); 948 ifa_free(&ia6->ia_ifa); 949 } 950 #ifdef RATELIMIT 951 if (inp != NULL) { 952 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) 953 in_pcboutput_txrtlmt(inp, ifp, m); 954 /* stamp send tag on mbuf */ 955 m->m_pkthdr.snd_tag = inp->inp_snd_tag; 956 } else { 957 m->m_pkthdr.snd_tag = NULL; 958 } 959 #endif 960 error = nd6_output_ifp(ifp, origifp, m, dst, 961 (struct route *)ro); 962 #ifdef RATELIMIT 963 /* check for route change */ 964 if (error == EAGAIN) 965 in_pcboutput_eagain(inp); 966 #endif 967 goto done; 968 } 969 970 /* 971 * try to fragment the packet. case 1-b and 3 972 */ 973 if (mtu < IPV6_MMTU) { 974 /* path MTU cannot be less than IPV6_MMTU */ 975 error = EMSGSIZE; 976 in6_ifstat_inc(ifp, ifs6_out_fragfail); 977 goto bad; 978 } else if (ip6->ip6_plen == 0) { 979 /* jumbo payload cannot be fragmented */ 980 error = EMSGSIZE; 981 in6_ifstat_inc(ifp, ifs6_out_fragfail); 982 goto bad; 983 } else { 984 u_char nextproto; 985 986 /* 987 * Too large for the destination or interface; 988 * fragment if possible. 989 * Must be able to put at least 8 bytes per fragment. 990 */ 991 hlen = unfragpartlen; 992 if (mtu > IPV6_MAXPACKET) 993 mtu = IPV6_MAXPACKET; 994 995 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 996 if (len < 8) { 997 error = EMSGSIZE; 998 in6_ifstat_inc(ifp, ifs6_out_fragfail); 999 goto bad; 1000 } 1001 1002 /* 1003 * If the interface will not calculate checksums on 1004 * fragmented packets, then do it here. 1005 * XXX-BZ handle the hw offloading case. Need flags. 1006 */ 1007 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 1008 in6_delayed_cksum(m, plen, hlen); 1009 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6; 1010 } 1011 #ifdef SCTP 1012 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) { 1013 sctp_delayed_cksum(m, hlen); 1014 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6; 1015 } 1016 #endif 1017 /* 1018 * Change the next header field of the last header in the 1019 * unfragmentable part. 1020 */ 1021 if (exthdrs.ip6e_rthdr) { 1022 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 1023 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 1024 } else if (exthdrs.ip6e_dest1) { 1025 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 1026 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 1027 } else if (exthdrs.ip6e_hbh) { 1028 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 1029 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 1030 } else { 1031 nextproto = ip6->ip6_nxt; 1032 ip6->ip6_nxt = IPPROTO_FRAGMENT; 1033 } 1034 1035 /* 1036 * Loop through length of segment after first fragment, 1037 * make new header and copy data of each part and link onto 1038 * chain. 1039 */ 1040 m0 = m; 1041 id = htonl(ip6_randomid()); 1042 if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id))) 1043 goto sendorfree; 1044 1045 in6_ifstat_inc(ifp, ifs6_out_fragok); 1046 } 1047 1048 /* 1049 * Remove leading garbages. 1050 */ 1051 sendorfree: 1052 m = m0->m_nextpkt; 1053 m0->m_nextpkt = 0; 1054 m_freem(m0); 1055 for (m0 = m; m; m = m0) { 1056 m0 = m->m_nextpkt; 1057 m->m_nextpkt = 0; 1058 if (error == 0) { 1059 /* Record statistics for this interface address. */ 1060 if (ia) { 1061 counter_u64_add(ia->ia_ifa.ifa_opackets, 1); 1062 counter_u64_add(ia->ia_ifa.ifa_obytes, 1063 m->m_pkthdr.len); 1064 } 1065 #ifdef RATELIMIT 1066 if (inp != NULL) { 1067 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) 1068 in_pcboutput_txrtlmt(inp, ifp, m); 1069 /* stamp send tag on mbuf */ 1070 m->m_pkthdr.snd_tag = inp->inp_snd_tag; 1071 } else { 1072 m->m_pkthdr.snd_tag = NULL; 1073 } 1074 #endif 1075 error = nd6_output_ifp(ifp, origifp, m, dst, 1076 (struct route *)ro); 1077 #ifdef RATELIMIT 1078 /* check for route change */ 1079 if (error == EAGAIN) 1080 in_pcboutput_eagain(inp); 1081 #endif 1082 } else 1083 m_freem(m); 1084 } 1085 1086 if (error == 0) 1087 IP6STAT_INC(ip6s_fragmented); 1088 1089 done: 1090 if (ro == &ip6route) 1091 RO_RTFREE(ro); 1092 return (error); 1093 1094 freehdrs: 1095 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1096 m_freem(exthdrs.ip6e_dest1); 1097 m_freem(exthdrs.ip6e_rthdr); 1098 m_freem(exthdrs.ip6e_dest2); 1099 /* FALLTHROUGH */ 1100 bad: 1101 if (m) 1102 m_freem(m); 1103 goto done; 1104 } 1105 1106 static int 1107 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen) 1108 { 1109 struct mbuf *m; 1110 1111 if (hlen > MCLBYTES) 1112 return (ENOBUFS); /* XXX */ 1113 1114 if (hlen > MLEN) 1115 m = m_getcl(M_NOWAIT, MT_DATA, 0); 1116 else 1117 m = m_get(M_NOWAIT, MT_DATA); 1118 if (m == NULL) 1119 return (ENOBUFS); 1120 m->m_len = hlen; 1121 if (hdr) 1122 bcopy(hdr, mtod(m, caddr_t), hlen); 1123 1124 *mp = m; 1125 return (0); 1126 } 1127 1128 /* 1129 * Insert jumbo payload option. 1130 */ 1131 static int 1132 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1133 { 1134 struct mbuf *mopt; 1135 u_char *optbuf; 1136 u_int32_t v; 1137 1138 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1139 1140 /* 1141 * If there is no hop-by-hop options header, allocate new one. 1142 * If there is one but it doesn't have enough space to store the 1143 * jumbo payload option, allocate a cluster to store the whole options. 1144 * Otherwise, use it to store the options. 1145 */ 1146 if (exthdrs->ip6e_hbh == NULL) { 1147 mopt = m_get(M_NOWAIT, MT_DATA); 1148 if (mopt == NULL) 1149 return (ENOBUFS); 1150 mopt->m_len = JUMBOOPTLEN; 1151 optbuf = mtod(mopt, u_char *); 1152 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1153 exthdrs->ip6e_hbh = mopt; 1154 } else { 1155 struct ip6_hbh *hbh; 1156 1157 mopt = exthdrs->ip6e_hbh; 1158 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1159 /* 1160 * XXX assumption: 1161 * - exthdrs->ip6e_hbh is not referenced from places 1162 * other than exthdrs. 1163 * - exthdrs->ip6e_hbh is not an mbuf chain. 1164 */ 1165 int oldoptlen = mopt->m_len; 1166 struct mbuf *n; 1167 1168 /* 1169 * XXX: give up if the whole (new) hbh header does 1170 * not fit even in an mbuf cluster. 1171 */ 1172 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1173 return (ENOBUFS); 1174 1175 /* 1176 * As a consequence, we must always prepare a cluster 1177 * at this point. 1178 */ 1179 n = m_getcl(M_NOWAIT, MT_DATA, 0); 1180 if (n == NULL) 1181 return (ENOBUFS); 1182 n->m_len = oldoptlen + JUMBOOPTLEN; 1183 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1184 oldoptlen); 1185 optbuf = mtod(n, caddr_t) + oldoptlen; 1186 m_freem(mopt); 1187 mopt = exthdrs->ip6e_hbh = n; 1188 } else { 1189 optbuf = mtod(mopt, u_char *) + mopt->m_len; 1190 mopt->m_len += JUMBOOPTLEN; 1191 } 1192 optbuf[0] = IP6OPT_PADN; 1193 optbuf[1] = 1; 1194 1195 /* 1196 * Adjust the header length according to the pad and 1197 * the jumbo payload option. 1198 */ 1199 hbh = mtod(mopt, struct ip6_hbh *); 1200 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1201 } 1202 1203 /* fill in the option. */ 1204 optbuf[2] = IP6OPT_JUMBO; 1205 optbuf[3] = 4; 1206 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1207 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1208 1209 /* finally, adjust the packet header length */ 1210 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1211 1212 return (0); 1213 #undef JUMBOOPTLEN 1214 } 1215 1216 /* 1217 * Insert fragment header and copy unfragmentable header portions. 1218 */ 1219 static int 1220 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1221 struct ip6_frag **frghdrp) 1222 { 1223 struct mbuf *n, *mlast; 1224 1225 if (hlen > sizeof(struct ip6_hdr)) { 1226 n = m_copym(m0, sizeof(struct ip6_hdr), 1227 hlen - sizeof(struct ip6_hdr), M_NOWAIT); 1228 if (n == NULL) 1229 return (ENOBUFS); 1230 m->m_next = n; 1231 } else 1232 n = m; 1233 1234 /* Search for the last mbuf of unfragmentable part. */ 1235 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1236 ; 1237 1238 if (M_WRITABLE(mlast) && 1239 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1240 /* use the trailing space of the last mbuf for the fragment hdr */ 1241 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1242 mlast->m_len); 1243 mlast->m_len += sizeof(struct ip6_frag); 1244 m->m_pkthdr.len += sizeof(struct ip6_frag); 1245 } else { 1246 /* allocate a new mbuf for the fragment header */ 1247 struct mbuf *mfrg; 1248 1249 mfrg = m_get(M_NOWAIT, MT_DATA); 1250 if (mfrg == NULL) 1251 return (ENOBUFS); 1252 mfrg->m_len = sizeof(struct ip6_frag); 1253 *frghdrp = mtod(mfrg, struct ip6_frag *); 1254 mlast->m_next = mfrg; 1255 } 1256 1257 return (0); 1258 } 1259 1260 /* 1261 * Calculates IPv6 path mtu for destination @dst. 1262 * Resulting MTU is stored in @mtup. 1263 * 1264 * Returns 0 on success. 1265 */ 1266 static int 1267 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup) 1268 { 1269 struct nhop6_extended nh6; 1270 struct in6_addr kdst; 1271 uint32_t scopeid; 1272 struct ifnet *ifp; 1273 u_long mtu; 1274 int error; 1275 1276 in6_splitscope(dst, &kdst, &scopeid); 1277 if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0) 1278 return (EHOSTUNREACH); 1279 1280 ifp = nh6.nh_ifp; 1281 mtu = nh6.nh_mtu; 1282 1283 error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0); 1284 fib6_free_nh_ext(fibnum, &nh6); 1285 1286 return (error); 1287 } 1288 1289 /* 1290 * Calculates IPv6 path MTU for @dst based on transmit @ifp, 1291 * and cached data in @ro_pmtu. 1292 * MTU from (successful) route lookup is saved (along with dst) 1293 * inside @ro_pmtu to avoid subsequent route lookups after packet 1294 * filter processing. 1295 * 1296 * Stores mtu and always-frag value into @mtup and @alwaysfragp. 1297 * Returns 0 on success. 1298 */ 1299 static int 1300 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup, 1301 struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup, 1302 int *alwaysfragp, u_int fibnum, u_int proto) 1303 { 1304 struct nhop6_basic nh6; 1305 struct in6_addr kdst; 1306 uint32_t scopeid; 1307 struct sockaddr_in6 *sa6_dst; 1308 u_long mtu; 1309 1310 mtu = 0; 1311 if (do_lookup) { 1312 1313 /* 1314 * Here ro_pmtu has final destination address, while 1315 * ro might represent immediate destination. 1316 * Use ro_pmtu destination since mtu might differ. 1317 */ 1318 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1319 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst)) 1320 ro_pmtu->ro_mtu = 0; 1321 1322 if (ro_pmtu->ro_mtu == 0) { 1323 bzero(sa6_dst, sizeof(*sa6_dst)); 1324 sa6_dst->sin6_family = AF_INET6; 1325 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1326 sa6_dst->sin6_addr = *dst; 1327 1328 in6_splitscope(dst, &kdst, &scopeid); 1329 if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0, 1330 &nh6) == 0) 1331 ro_pmtu->ro_mtu = nh6.nh_mtu; 1332 } 1333 1334 mtu = ro_pmtu->ro_mtu; 1335 } 1336 1337 if (ro_pmtu->ro_rt) 1338 mtu = ro_pmtu->ro_rt->rt_mtu; 1339 1340 return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto)); 1341 } 1342 1343 /* 1344 * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and 1345 * hostcache data for @dst. 1346 * Stores mtu and always-frag value into @mtup and @alwaysfragp. 1347 * 1348 * Returns 0 on success. 1349 */ 1350 static int 1351 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu, 1352 u_long *mtup, int *alwaysfragp, u_int proto) 1353 { 1354 u_long mtu = 0; 1355 int alwaysfrag = 0; 1356 int error = 0; 1357 1358 if (rt_mtu > 0) { 1359 u_int32_t ifmtu; 1360 struct in_conninfo inc; 1361 1362 bzero(&inc, sizeof(inc)); 1363 inc.inc_flags |= INC_ISIPV6; 1364 inc.inc6_faddr = *dst; 1365 1366 ifmtu = IN6_LINKMTU(ifp); 1367 1368 /* TCP is known to react to pmtu changes so skip hc */ 1369 if (proto != IPPROTO_TCP) 1370 mtu = tcp_hc_getmtu(&inc); 1371 1372 if (mtu) 1373 mtu = min(mtu, rt_mtu); 1374 else 1375 mtu = rt_mtu; 1376 if (mtu == 0) 1377 mtu = ifmtu; 1378 else if (mtu < IPV6_MMTU) { 1379 /* 1380 * RFC2460 section 5, last paragraph: 1381 * if we record ICMPv6 too big message with 1382 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1383 * or smaller, with framgent header attached. 1384 * (fragment header is needed regardless from the 1385 * packet size, for translators to identify packets) 1386 */ 1387 alwaysfrag = 1; 1388 mtu = IPV6_MMTU; 1389 } 1390 } else if (ifp) { 1391 mtu = IN6_LINKMTU(ifp); 1392 } else 1393 error = EHOSTUNREACH; /* XXX */ 1394 1395 *mtup = mtu; 1396 if (alwaysfragp) 1397 *alwaysfragp = alwaysfrag; 1398 return (error); 1399 } 1400 1401 /* 1402 * IP6 socket option processing. 1403 */ 1404 int 1405 ip6_ctloutput(struct socket *so, struct sockopt *sopt) 1406 { 1407 int optdatalen, uproto; 1408 void *optdata; 1409 struct inpcb *in6p = sotoinpcb(so); 1410 int error, optval; 1411 int level, op, optname; 1412 int optlen; 1413 struct thread *td; 1414 #ifdef RSS 1415 uint32_t rss_bucket; 1416 int retval; 1417 #endif 1418 1419 /* 1420 * Don't use more than a quarter of mbuf clusters. N.B.: 1421 * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow 1422 * on LP64 architectures, so cast to u_long to avoid undefined 1423 * behavior. ILP32 architectures cannot have nmbclusters 1424 * large enough to overflow for other reasons. 1425 */ 1426 #define IPV6_PKTOPTIONS_MBUF_LIMIT ((u_long)nmbclusters * MCLBYTES / 4) 1427 1428 level = sopt->sopt_level; 1429 op = sopt->sopt_dir; 1430 optname = sopt->sopt_name; 1431 optlen = sopt->sopt_valsize; 1432 td = sopt->sopt_td; 1433 error = 0; 1434 optval = 0; 1435 uproto = (int)so->so_proto->pr_protocol; 1436 1437 if (level != IPPROTO_IPV6) { 1438 error = EINVAL; 1439 1440 if (sopt->sopt_level == SOL_SOCKET && 1441 sopt->sopt_dir == SOPT_SET) { 1442 switch (sopt->sopt_name) { 1443 case SO_REUSEADDR: 1444 INP_WLOCK(in6p); 1445 if ((so->so_options & SO_REUSEADDR) != 0) 1446 in6p->inp_flags2 |= INP_REUSEADDR; 1447 else 1448 in6p->inp_flags2 &= ~INP_REUSEADDR; 1449 INP_WUNLOCK(in6p); 1450 error = 0; 1451 break; 1452 case SO_REUSEPORT: 1453 INP_WLOCK(in6p); 1454 if ((so->so_options & SO_REUSEPORT) != 0) 1455 in6p->inp_flags2 |= INP_REUSEPORT; 1456 else 1457 in6p->inp_flags2 &= ~INP_REUSEPORT; 1458 INP_WUNLOCK(in6p); 1459 error = 0; 1460 break; 1461 case SO_SETFIB: 1462 INP_WLOCK(in6p); 1463 in6p->inp_inc.inc_fibnum = so->so_fibnum; 1464 INP_WUNLOCK(in6p); 1465 error = 0; 1466 break; 1467 case SO_MAX_PACING_RATE: 1468 #ifdef RATELIMIT 1469 INP_WLOCK(in6p); 1470 in6p->inp_flags2 |= INP_RATE_LIMIT_CHANGED; 1471 INP_WUNLOCK(in6p); 1472 error = 0; 1473 #else 1474 error = EOPNOTSUPP; 1475 #endif 1476 break; 1477 default: 1478 break; 1479 } 1480 } 1481 } else { /* level == IPPROTO_IPV6 */ 1482 switch (op) { 1483 1484 case SOPT_SET: 1485 switch (optname) { 1486 case IPV6_2292PKTOPTIONS: 1487 #ifdef IPV6_PKTOPTIONS 1488 case IPV6_PKTOPTIONS: 1489 #endif 1490 { 1491 struct mbuf *m; 1492 1493 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) { 1494 printf("ip6_ctloutput: mbuf limit hit\n"); 1495 error = ENOBUFS; 1496 break; 1497 } 1498 1499 error = soopt_getm(sopt, &m); /* XXX */ 1500 if (error != 0) 1501 break; 1502 error = soopt_mcopyin(sopt, m); /* XXX */ 1503 if (error != 0) 1504 break; 1505 error = ip6_pcbopts(&in6p->in6p_outputopts, 1506 m, so, sopt); 1507 m_freem(m); /* XXX */ 1508 break; 1509 } 1510 1511 /* 1512 * Use of some Hop-by-Hop options or some 1513 * Destination options, might require special 1514 * privilege. That is, normal applications 1515 * (without special privilege) might be forbidden 1516 * from setting certain options in outgoing packets, 1517 * and might never see certain options in received 1518 * packets. [RFC 2292 Section 6] 1519 * KAME specific note: 1520 * KAME prevents non-privileged users from sending or 1521 * receiving ANY hbh/dst options in order to avoid 1522 * overhead of parsing options in the kernel. 1523 */ 1524 case IPV6_RECVHOPOPTS: 1525 case IPV6_RECVDSTOPTS: 1526 case IPV6_RECVRTHDRDSTOPTS: 1527 if (td != NULL) { 1528 error = priv_check(td, 1529 PRIV_NETINET_SETHDROPTS); 1530 if (error) 1531 break; 1532 } 1533 /* FALLTHROUGH */ 1534 case IPV6_UNICAST_HOPS: 1535 case IPV6_HOPLIMIT: 1536 1537 case IPV6_RECVPKTINFO: 1538 case IPV6_RECVHOPLIMIT: 1539 case IPV6_RECVRTHDR: 1540 case IPV6_RECVPATHMTU: 1541 case IPV6_RECVTCLASS: 1542 case IPV6_RECVFLOWID: 1543 #ifdef RSS 1544 case IPV6_RECVRSSBUCKETID: 1545 #endif 1546 case IPV6_V6ONLY: 1547 case IPV6_AUTOFLOWLABEL: 1548 case IPV6_BINDANY: 1549 case IPV6_BINDMULTI: 1550 #ifdef RSS 1551 case IPV6_RSS_LISTEN_BUCKET: 1552 #endif 1553 if (optname == IPV6_BINDANY && td != NULL) { 1554 error = priv_check(td, 1555 PRIV_NETINET_BINDANY); 1556 if (error) 1557 break; 1558 } 1559 1560 if (optlen != sizeof(int)) { 1561 error = EINVAL; 1562 break; 1563 } 1564 error = sooptcopyin(sopt, &optval, 1565 sizeof optval, sizeof optval); 1566 if (error) 1567 break; 1568 switch (optname) { 1569 1570 case IPV6_UNICAST_HOPS: 1571 if (optval < -1 || optval >= 256) 1572 error = EINVAL; 1573 else { 1574 /* -1 = kernel default */ 1575 in6p->in6p_hops = optval; 1576 if ((in6p->inp_vflag & 1577 INP_IPV4) != 0) 1578 in6p->inp_ip_ttl = optval; 1579 } 1580 break; 1581 #define OPTSET(bit) \ 1582 do { \ 1583 INP_WLOCK(in6p); \ 1584 if (optval) \ 1585 in6p->inp_flags |= (bit); \ 1586 else \ 1587 in6p->inp_flags &= ~(bit); \ 1588 INP_WUNLOCK(in6p); \ 1589 } while (/*CONSTCOND*/ 0) 1590 #define OPTSET2292(bit) \ 1591 do { \ 1592 INP_WLOCK(in6p); \ 1593 in6p->inp_flags |= IN6P_RFC2292; \ 1594 if (optval) \ 1595 in6p->inp_flags |= (bit); \ 1596 else \ 1597 in6p->inp_flags &= ~(bit); \ 1598 INP_WUNLOCK(in6p); \ 1599 } while (/*CONSTCOND*/ 0) 1600 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0) 1601 1602 #define OPTSET2(bit, val) do { \ 1603 INP_WLOCK(in6p); \ 1604 if (val) \ 1605 in6p->inp_flags2 |= bit; \ 1606 else \ 1607 in6p->inp_flags2 &= ~bit; \ 1608 INP_WUNLOCK(in6p); \ 1609 } while (0) 1610 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0) 1611 1612 case IPV6_RECVPKTINFO: 1613 /* cannot mix with RFC2292 */ 1614 if (OPTBIT(IN6P_RFC2292)) { 1615 error = EINVAL; 1616 break; 1617 } 1618 OPTSET(IN6P_PKTINFO); 1619 break; 1620 1621 case IPV6_HOPLIMIT: 1622 { 1623 struct ip6_pktopts **optp; 1624 1625 /* cannot mix with RFC2292 */ 1626 if (OPTBIT(IN6P_RFC2292)) { 1627 error = EINVAL; 1628 break; 1629 } 1630 optp = &in6p->in6p_outputopts; 1631 error = ip6_pcbopt(IPV6_HOPLIMIT, 1632 (u_char *)&optval, sizeof(optval), 1633 optp, (td != NULL) ? td->td_ucred : 1634 NULL, uproto); 1635 break; 1636 } 1637 1638 case IPV6_RECVHOPLIMIT: 1639 /* cannot mix with RFC2292 */ 1640 if (OPTBIT(IN6P_RFC2292)) { 1641 error = EINVAL; 1642 break; 1643 } 1644 OPTSET(IN6P_HOPLIMIT); 1645 break; 1646 1647 case IPV6_RECVHOPOPTS: 1648 /* cannot mix with RFC2292 */ 1649 if (OPTBIT(IN6P_RFC2292)) { 1650 error = EINVAL; 1651 break; 1652 } 1653 OPTSET(IN6P_HOPOPTS); 1654 break; 1655 1656 case IPV6_RECVDSTOPTS: 1657 /* cannot mix with RFC2292 */ 1658 if (OPTBIT(IN6P_RFC2292)) { 1659 error = EINVAL; 1660 break; 1661 } 1662 OPTSET(IN6P_DSTOPTS); 1663 break; 1664 1665 case IPV6_RECVRTHDRDSTOPTS: 1666 /* cannot mix with RFC2292 */ 1667 if (OPTBIT(IN6P_RFC2292)) { 1668 error = EINVAL; 1669 break; 1670 } 1671 OPTSET(IN6P_RTHDRDSTOPTS); 1672 break; 1673 1674 case IPV6_RECVRTHDR: 1675 /* cannot mix with RFC2292 */ 1676 if (OPTBIT(IN6P_RFC2292)) { 1677 error = EINVAL; 1678 break; 1679 } 1680 OPTSET(IN6P_RTHDR); 1681 break; 1682 1683 case IPV6_RECVPATHMTU: 1684 /* 1685 * We ignore this option for TCP 1686 * sockets. 1687 * (RFC3542 leaves this case 1688 * unspecified.) 1689 */ 1690 if (uproto != IPPROTO_TCP) 1691 OPTSET(IN6P_MTU); 1692 break; 1693 1694 case IPV6_RECVFLOWID: 1695 OPTSET2(INP_RECVFLOWID, optval); 1696 break; 1697 1698 #ifdef RSS 1699 case IPV6_RECVRSSBUCKETID: 1700 OPTSET2(INP_RECVRSSBUCKETID, optval); 1701 break; 1702 #endif 1703 1704 case IPV6_V6ONLY: 1705 /* 1706 * make setsockopt(IPV6_V6ONLY) 1707 * available only prior to bind(2). 1708 * see ipng mailing list, Jun 22 2001. 1709 */ 1710 if (in6p->inp_lport || 1711 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1712 error = EINVAL; 1713 break; 1714 } 1715 OPTSET(IN6P_IPV6_V6ONLY); 1716 if (optval) 1717 in6p->inp_vflag &= ~INP_IPV4; 1718 else 1719 in6p->inp_vflag |= INP_IPV4; 1720 break; 1721 case IPV6_RECVTCLASS: 1722 /* cannot mix with RFC2292 XXX */ 1723 if (OPTBIT(IN6P_RFC2292)) { 1724 error = EINVAL; 1725 break; 1726 } 1727 OPTSET(IN6P_TCLASS); 1728 break; 1729 case IPV6_AUTOFLOWLABEL: 1730 OPTSET(IN6P_AUTOFLOWLABEL); 1731 break; 1732 1733 case IPV6_BINDANY: 1734 OPTSET(INP_BINDANY); 1735 break; 1736 1737 case IPV6_BINDMULTI: 1738 OPTSET2(INP_BINDMULTI, optval); 1739 break; 1740 #ifdef RSS 1741 case IPV6_RSS_LISTEN_BUCKET: 1742 if ((optval >= 0) && 1743 (optval < rss_getnumbuckets())) { 1744 in6p->inp_rss_listen_bucket = optval; 1745 OPTSET2(INP_RSS_BUCKET_SET, 1); 1746 } else { 1747 error = EINVAL; 1748 } 1749 break; 1750 #endif 1751 } 1752 break; 1753 1754 case IPV6_TCLASS: 1755 case IPV6_DONTFRAG: 1756 case IPV6_USE_MIN_MTU: 1757 case IPV6_PREFER_TEMPADDR: 1758 if (optlen != sizeof(optval)) { 1759 error = EINVAL; 1760 break; 1761 } 1762 error = sooptcopyin(sopt, &optval, 1763 sizeof optval, sizeof optval); 1764 if (error) 1765 break; 1766 { 1767 struct ip6_pktopts **optp; 1768 optp = &in6p->in6p_outputopts; 1769 error = ip6_pcbopt(optname, 1770 (u_char *)&optval, sizeof(optval), 1771 optp, (td != NULL) ? td->td_ucred : 1772 NULL, uproto); 1773 break; 1774 } 1775 1776 case IPV6_2292PKTINFO: 1777 case IPV6_2292HOPLIMIT: 1778 case IPV6_2292HOPOPTS: 1779 case IPV6_2292DSTOPTS: 1780 case IPV6_2292RTHDR: 1781 /* RFC 2292 */ 1782 if (optlen != sizeof(int)) { 1783 error = EINVAL; 1784 break; 1785 } 1786 error = sooptcopyin(sopt, &optval, 1787 sizeof optval, sizeof optval); 1788 if (error) 1789 break; 1790 switch (optname) { 1791 case IPV6_2292PKTINFO: 1792 OPTSET2292(IN6P_PKTINFO); 1793 break; 1794 case IPV6_2292HOPLIMIT: 1795 OPTSET2292(IN6P_HOPLIMIT); 1796 break; 1797 case IPV6_2292HOPOPTS: 1798 /* 1799 * Check super-user privilege. 1800 * See comments for IPV6_RECVHOPOPTS. 1801 */ 1802 if (td != NULL) { 1803 error = priv_check(td, 1804 PRIV_NETINET_SETHDROPTS); 1805 if (error) 1806 return (error); 1807 } 1808 OPTSET2292(IN6P_HOPOPTS); 1809 break; 1810 case IPV6_2292DSTOPTS: 1811 if (td != NULL) { 1812 error = priv_check(td, 1813 PRIV_NETINET_SETHDROPTS); 1814 if (error) 1815 return (error); 1816 } 1817 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1818 break; 1819 case IPV6_2292RTHDR: 1820 OPTSET2292(IN6P_RTHDR); 1821 break; 1822 } 1823 break; 1824 case IPV6_PKTINFO: 1825 case IPV6_HOPOPTS: 1826 case IPV6_RTHDR: 1827 case IPV6_DSTOPTS: 1828 case IPV6_RTHDRDSTOPTS: 1829 case IPV6_NEXTHOP: 1830 { 1831 /* new advanced API (RFC3542) */ 1832 u_char *optbuf; 1833 u_char optbuf_storage[MCLBYTES]; 1834 int optlen; 1835 struct ip6_pktopts **optp; 1836 1837 /* cannot mix with RFC2292 */ 1838 if (OPTBIT(IN6P_RFC2292)) { 1839 error = EINVAL; 1840 break; 1841 } 1842 1843 /* 1844 * We only ensure valsize is not too large 1845 * here. Further validation will be done 1846 * later. 1847 */ 1848 error = sooptcopyin(sopt, optbuf_storage, 1849 sizeof(optbuf_storage), 0); 1850 if (error) 1851 break; 1852 optlen = sopt->sopt_valsize; 1853 optbuf = optbuf_storage; 1854 optp = &in6p->in6p_outputopts; 1855 error = ip6_pcbopt(optname, optbuf, optlen, 1856 optp, (td != NULL) ? td->td_ucred : NULL, 1857 uproto); 1858 break; 1859 } 1860 #undef OPTSET 1861 1862 case IPV6_MULTICAST_IF: 1863 case IPV6_MULTICAST_HOPS: 1864 case IPV6_MULTICAST_LOOP: 1865 case IPV6_JOIN_GROUP: 1866 case IPV6_LEAVE_GROUP: 1867 case IPV6_MSFILTER: 1868 case MCAST_BLOCK_SOURCE: 1869 case MCAST_UNBLOCK_SOURCE: 1870 case MCAST_JOIN_GROUP: 1871 case MCAST_LEAVE_GROUP: 1872 case MCAST_JOIN_SOURCE_GROUP: 1873 case MCAST_LEAVE_SOURCE_GROUP: 1874 error = ip6_setmoptions(in6p, sopt); 1875 break; 1876 1877 case IPV6_PORTRANGE: 1878 error = sooptcopyin(sopt, &optval, 1879 sizeof optval, sizeof optval); 1880 if (error) 1881 break; 1882 1883 INP_WLOCK(in6p); 1884 switch (optval) { 1885 case IPV6_PORTRANGE_DEFAULT: 1886 in6p->inp_flags &= ~(INP_LOWPORT); 1887 in6p->inp_flags &= ~(INP_HIGHPORT); 1888 break; 1889 1890 case IPV6_PORTRANGE_HIGH: 1891 in6p->inp_flags &= ~(INP_LOWPORT); 1892 in6p->inp_flags |= INP_HIGHPORT; 1893 break; 1894 1895 case IPV6_PORTRANGE_LOW: 1896 in6p->inp_flags &= ~(INP_HIGHPORT); 1897 in6p->inp_flags |= INP_LOWPORT; 1898 break; 1899 1900 default: 1901 error = EINVAL; 1902 break; 1903 } 1904 INP_WUNLOCK(in6p); 1905 break; 1906 1907 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 1908 case IPV6_IPSEC_POLICY: 1909 if (IPSEC_ENABLED(ipv6)) { 1910 error = IPSEC_PCBCTL(ipv6, in6p, sopt); 1911 break; 1912 } 1913 /* FALLTHROUGH */ 1914 #endif /* IPSEC */ 1915 1916 default: 1917 error = ENOPROTOOPT; 1918 break; 1919 } 1920 break; 1921 1922 case SOPT_GET: 1923 switch (optname) { 1924 1925 case IPV6_2292PKTOPTIONS: 1926 #ifdef IPV6_PKTOPTIONS 1927 case IPV6_PKTOPTIONS: 1928 #endif 1929 /* 1930 * RFC3542 (effectively) deprecated the 1931 * semantics of the 2292-style pktoptions. 1932 * Since it was not reliable in nature (i.e., 1933 * applications had to expect the lack of some 1934 * information after all), it would make sense 1935 * to simplify this part by always returning 1936 * empty data. 1937 */ 1938 sopt->sopt_valsize = 0; 1939 break; 1940 1941 case IPV6_RECVHOPOPTS: 1942 case IPV6_RECVDSTOPTS: 1943 case IPV6_RECVRTHDRDSTOPTS: 1944 case IPV6_UNICAST_HOPS: 1945 case IPV6_RECVPKTINFO: 1946 case IPV6_RECVHOPLIMIT: 1947 case IPV6_RECVRTHDR: 1948 case IPV6_RECVPATHMTU: 1949 1950 case IPV6_V6ONLY: 1951 case IPV6_PORTRANGE: 1952 case IPV6_RECVTCLASS: 1953 case IPV6_AUTOFLOWLABEL: 1954 case IPV6_BINDANY: 1955 case IPV6_FLOWID: 1956 case IPV6_FLOWTYPE: 1957 case IPV6_RECVFLOWID: 1958 #ifdef RSS 1959 case IPV6_RSSBUCKETID: 1960 case IPV6_RECVRSSBUCKETID: 1961 #endif 1962 case IPV6_BINDMULTI: 1963 switch (optname) { 1964 1965 case IPV6_RECVHOPOPTS: 1966 optval = OPTBIT(IN6P_HOPOPTS); 1967 break; 1968 1969 case IPV6_RECVDSTOPTS: 1970 optval = OPTBIT(IN6P_DSTOPTS); 1971 break; 1972 1973 case IPV6_RECVRTHDRDSTOPTS: 1974 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1975 break; 1976 1977 case IPV6_UNICAST_HOPS: 1978 optval = in6p->in6p_hops; 1979 break; 1980 1981 case IPV6_RECVPKTINFO: 1982 optval = OPTBIT(IN6P_PKTINFO); 1983 break; 1984 1985 case IPV6_RECVHOPLIMIT: 1986 optval = OPTBIT(IN6P_HOPLIMIT); 1987 break; 1988 1989 case IPV6_RECVRTHDR: 1990 optval = OPTBIT(IN6P_RTHDR); 1991 break; 1992 1993 case IPV6_RECVPATHMTU: 1994 optval = OPTBIT(IN6P_MTU); 1995 break; 1996 1997 case IPV6_V6ONLY: 1998 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1999 break; 2000 2001 case IPV6_PORTRANGE: 2002 { 2003 int flags; 2004 flags = in6p->inp_flags; 2005 if (flags & INP_HIGHPORT) 2006 optval = IPV6_PORTRANGE_HIGH; 2007 else if (flags & INP_LOWPORT) 2008 optval = IPV6_PORTRANGE_LOW; 2009 else 2010 optval = 0; 2011 break; 2012 } 2013 case IPV6_RECVTCLASS: 2014 optval = OPTBIT(IN6P_TCLASS); 2015 break; 2016 2017 case IPV6_AUTOFLOWLABEL: 2018 optval = OPTBIT(IN6P_AUTOFLOWLABEL); 2019 break; 2020 2021 case IPV6_BINDANY: 2022 optval = OPTBIT(INP_BINDANY); 2023 break; 2024 2025 case IPV6_FLOWID: 2026 optval = in6p->inp_flowid; 2027 break; 2028 2029 case IPV6_FLOWTYPE: 2030 optval = in6p->inp_flowtype; 2031 break; 2032 2033 case IPV6_RECVFLOWID: 2034 optval = OPTBIT2(INP_RECVFLOWID); 2035 break; 2036 #ifdef RSS 2037 case IPV6_RSSBUCKETID: 2038 retval = 2039 rss_hash2bucket(in6p->inp_flowid, 2040 in6p->inp_flowtype, 2041 &rss_bucket); 2042 if (retval == 0) 2043 optval = rss_bucket; 2044 else 2045 error = EINVAL; 2046 break; 2047 2048 case IPV6_RECVRSSBUCKETID: 2049 optval = OPTBIT2(INP_RECVRSSBUCKETID); 2050 break; 2051 #endif 2052 2053 case IPV6_BINDMULTI: 2054 optval = OPTBIT2(INP_BINDMULTI); 2055 break; 2056 2057 } 2058 if (error) 2059 break; 2060 error = sooptcopyout(sopt, &optval, 2061 sizeof optval); 2062 break; 2063 2064 case IPV6_PATHMTU: 2065 { 2066 u_long pmtu = 0; 2067 struct ip6_mtuinfo mtuinfo; 2068 2069 if (!(so->so_state & SS_ISCONNECTED)) 2070 return (ENOTCONN); 2071 /* 2072 * XXX: we dot not consider the case of source 2073 * routing, or optional information to specify 2074 * the outgoing interface. 2075 */ 2076 error = ip6_getpmtu_ctl(so->so_fibnum, 2077 &in6p->in6p_faddr, &pmtu); 2078 if (error) 2079 break; 2080 if (pmtu > IPV6_MAXPACKET) 2081 pmtu = IPV6_MAXPACKET; 2082 2083 bzero(&mtuinfo, sizeof(mtuinfo)); 2084 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 2085 optdata = (void *)&mtuinfo; 2086 optdatalen = sizeof(mtuinfo); 2087 error = sooptcopyout(sopt, optdata, 2088 optdatalen); 2089 break; 2090 } 2091 2092 case IPV6_2292PKTINFO: 2093 case IPV6_2292HOPLIMIT: 2094 case IPV6_2292HOPOPTS: 2095 case IPV6_2292RTHDR: 2096 case IPV6_2292DSTOPTS: 2097 switch (optname) { 2098 case IPV6_2292PKTINFO: 2099 optval = OPTBIT(IN6P_PKTINFO); 2100 break; 2101 case IPV6_2292HOPLIMIT: 2102 optval = OPTBIT(IN6P_HOPLIMIT); 2103 break; 2104 case IPV6_2292HOPOPTS: 2105 optval = OPTBIT(IN6P_HOPOPTS); 2106 break; 2107 case IPV6_2292RTHDR: 2108 optval = OPTBIT(IN6P_RTHDR); 2109 break; 2110 case IPV6_2292DSTOPTS: 2111 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 2112 break; 2113 } 2114 error = sooptcopyout(sopt, &optval, 2115 sizeof optval); 2116 break; 2117 case IPV6_PKTINFO: 2118 case IPV6_HOPOPTS: 2119 case IPV6_RTHDR: 2120 case IPV6_DSTOPTS: 2121 case IPV6_RTHDRDSTOPTS: 2122 case IPV6_NEXTHOP: 2123 case IPV6_TCLASS: 2124 case IPV6_DONTFRAG: 2125 case IPV6_USE_MIN_MTU: 2126 case IPV6_PREFER_TEMPADDR: 2127 error = ip6_getpcbopt(in6p->in6p_outputopts, 2128 optname, sopt); 2129 break; 2130 2131 case IPV6_MULTICAST_IF: 2132 case IPV6_MULTICAST_HOPS: 2133 case IPV6_MULTICAST_LOOP: 2134 case IPV6_MSFILTER: 2135 error = ip6_getmoptions(in6p, sopt); 2136 break; 2137 2138 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 2139 case IPV6_IPSEC_POLICY: 2140 if (IPSEC_ENABLED(ipv6)) { 2141 error = IPSEC_PCBCTL(ipv6, in6p, sopt); 2142 break; 2143 } 2144 /* FALLTHROUGH */ 2145 #endif /* IPSEC */ 2146 default: 2147 error = ENOPROTOOPT; 2148 break; 2149 } 2150 break; 2151 } 2152 } 2153 return (error); 2154 } 2155 2156 int 2157 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt) 2158 { 2159 int error = 0, optval, optlen; 2160 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2161 struct inpcb *in6p = sotoinpcb(so); 2162 int level, op, optname; 2163 2164 level = sopt->sopt_level; 2165 op = sopt->sopt_dir; 2166 optname = sopt->sopt_name; 2167 optlen = sopt->sopt_valsize; 2168 2169 if (level != IPPROTO_IPV6) { 2170 return (EINVAL); 2171 } 2172 2173 switch (optname) { 2174 case IPV6_CHECKSUM: 2175 /* 2176 * For ICMPv6 sockets, no modification allowed for checksum 2177 * offset, permit "no change" values to help existing apps. 2178 * 2179 * RFC3542 says: "An attempt to set IPV6_CHECKSUM 2180 * for an ICMPv6 socket will fail." 2181 * The current behavior does not meet RFC3542. 2182 */ 2183 switch (op) { 2184 case SOPT_SET: 2185 if (optlen != sizeof(int)) { 2186 error = EINVAL; 2187 break; 2188 } 2189 error = sooptcopyin(sopt, &optval, sizeof(optval), 2190 sizeof(optval)); 2191 if (error) 2192 break; 2193 if ((optval % 2) != 0) { 2194 /* the API assumes even offset values */ 2195 error = EINVAL; 2196 } else if (so->so_proto->pr_protocol == 2197 IPPROTO_ICMPV6) { 2198 if (optval != icmp6off) 2199 error = EINVAL; 2200 } else 2201 in6p->in6p_cksum = optval; 2202 break; 2203 2204 case SOPT_GET: 2205 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2206 optval = icmp6off; 2207 else 2208 optval = in6p->in6p_cksum; 2209 2210 error = sooptcopyout(sopt, &optval, sizeof(optval)); 2211 break; 2212 2213 default: 2214 error = EINVAL; 2215 break; 2216 } 2217 break; 2218 2219 default: 2220 error = ENOPROTOOPT; 2221 break; 2222 } 2223 2224 return (error); 2225 } 2226 2227 /* 2228 * Set up IP6 options in pcb for insertion in output packets or 2229 * specifying behavior of outgoing packets. 2230 */ 2231 static int 2232 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, 2233 struct socket *so, struct sockopt *sopt) 2234 { 2235 struct ip6_pktopts *opt = *pktopt; 2236 int error = 0; 2237 struct thread *td = sopt->sopt_td; 2238 2239 /* turn off any old options. */ 2240 if (opt) { 2241 #ifdef DIAGNOSTIC 2242 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2243 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2244 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2245 printf("ip6_pcbopts: all specified options are cleared.\n"); 2246 #endif 2247 ip6_clearpktopts(opt, -1); 2248 } else 2249 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2250 *pktopt = NULL; 2251 2252 if (!m || m->m_len == 0) { 2253 /* 2254 * Only turning off any previous options, regardless of 2255 * whether the opt is just created or given. 2256 */ 2257 free(opt, M_IP6OPT); 2258 return (0); 2259 } 2260 2261 /* set options specified by user. */ 2262 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ? 2263 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) { 2264 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2265 free(opt, M_IP6OPT); 2266 return (error); 2267 } 2268 *pktopt = opt; 2269 return (0); 2270 } 2271 2272 /* 2273 * initialize ip6_pktopts. beware that there are non-zero default values in 2274 * the struct. 2275 */ 2276 void 2277 ip6_initpktopts(struct ip6_pktopts *opt) 2278 { 2279 2280 bzero(opt, sizeof(*opt)); 2281 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2282 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2283 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2284 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; 2285 } 2286 2287 static int 2288 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2289 struct ucred *cred, int uproto) 2290 { 2291 struct ip6_pktopts *opt; 2292 2293 if (*pktopt == NULL) { 2294 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2295 M_WAITOK); 2296 ip6_initpktopts(*pktopt); 2297 } 2298 opt = *pktopt; 2299 2300 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); 2301 } 2302 2303 static int 2304 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 2305 { 2306 void *optdata = NULL; 2307 int optdatalen = 0; 2308 struct ip6_ext *ip6e; 2309 int error = 0; 2310 struct in6_pktinfo null_pktinfo; 2311 int deftclass = 0, on; 2312 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2313 int defpreftemp = IP6PO_TEMPADDR_SYSTEM; 2314 2315 switch (optname) { 2316 case IPV6_PKTINFO: 2317 optdata = (void *)&null_pktinfo; 2318 if (pktopt && pktopt->ip6po_pktinfo) { 2319 bcopy(pktopt->ip6po_pktinfo, &null_pktinfo, 2320 sizeof(null_pktinfo)); 2321 in6_clearscope(&null_pktinfo.ipi6_addr); 2322 } else { 2323 /* XXX: we don't have to do this every time... */ 2324 bzero(&null_pktinfo, sizeof(null_pktinfo)); 2325 } 2326 optdatalen = sizeof(struct in6_pktinfo); 2327 break; 2328 case IPV6_TCLASS: 2329 if (pktopt && pktopt->ip6po_tclass >= 0) 2330 optdata = (void *)&pktopt->ip6po_tclass; 2331 else 2332 optdata = (void *)&deftclass; 2333 optdatalen = sizeof(int); 2334 break; 2335 case IPV6_HOPOPTS: 2336 if (pktopt && pktopt->ip6po_hbh) { 2337 optdata = (void *)pktopt->ip6po_hbh; 2338 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2339 optdatalen = (ip6e->ip6e_len + 1) << 3; 2340 } 2341 break; 2342 case IPV6_RTHDR: 2343 if (pktopt && pktopt->ip6po_rthdr) { 2344 optdata = (void *)pktopt->ip6po_rthdr; 2345 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2346 optdatalen = (ip6e->ip6e_len + 1) << 3; 2347 } 2348 break; 2349 case IPV6_RTHDRDSTOPTS: 2350 if (pktopt && pktopt->ip6po_dest1) { 2351 optdata = (void *)pktopt->ip6po_dest1; 2352 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2353 optdatalen = (ip6e->ip6e_len + 1) << 3; 2354 } 2355 break; 2356 case IPV6_DSTOPTS: 2357 if (pktopt && pktopt->ip6po_dest2) { 2358 optdata = (void *)pktopt->ip6po_dest2; 2359 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2360 optdatalen = (ip6e->ip6e_len + 1) << 3; 2361 } 2362 break; 2363 case IPV6_NEXTHOP: 2364 if (pktopt && pktopt->ip6po_nexthop) { 2365 optdata = (void *)pktopt->ip6po_nexthop; 2366 optdatalen = pktopt->ip6po_nexthop->sa_len; 2367 } 2368 break; 2369 case IPV6_USE_MIN_MTU: 2370 if (pktopt) 2371 optdata = (void *)&pktopt->ip6po_minmtu; 2372 else 2373 optdata = (void *)&defminmtu; 2374 optdatalen = sizeof(int); 2375 break; 2376 case IPV6_DONTFRAG: 2377 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2378 on = 1; 2379 else 2380 on = 0; 2381 optdata = (void *)&on; 2382 optdatalen = sizeof(on); 2383 break; 2384 case IPV6_PREFER_TEMPADDR: 2385 if (pktopt) 2386 optdata = (void *)&pktopt->ip6po_prefer_tempaddr; 2387 else 2388 optdata = (void *)&defpreftemp; 2389 optdatalen = sizeof(int); 2390 break; 2391 default: /* should not happen */ 2392 #ifdef DIAGNOSTIC 2393 panic("ip6_getpcbopt: unexpected option\n"); 2394 #endif 2395 return (ENOPROTOOPT); 2396 } 2397 2398 error = sooptcopyout(sopt, optdata, optdatalen); 2399 2400 return (error); 2401 } 2402 2403 void 2404 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2405 { 2406 if (pktopt == NULL) 2407 return; 2408 2409 if (optname == -1 || optname == IPV6_PKTINFO) { 2410 if (pktopt->ip6po_pktinfo) 2411 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2412 pktopt->ip6po_pktinfo = NULL; 2413 } 2414 if (optname == -1 || optname == IPV6_HOPLIMIT) 2415 pktopt->ip6po_hlim = -1; 2416 if (optname == -1 || optname == IPV6_TCLASS) 2417 pktopt->ip6po_tclass = -1; 2418 if (optname == -1 || optname == IPV6_NEXTHOP) { 2419 if (pktopt->ip6po_nextroute.ro_rt) { 2420 RTFREE(pktopt->ip6po_nextroute.ro_rt); 2421 pktopt->ip6po_nextroute.ro_rt = NULL; 2422 } 2423 if (pktopt->ip6po_nexthop) 2424 free(pktopt->ip6po_nexthop, M_IP6OPT); 2425 pktopt->ip6po_nexthop = NULL; 2426 } 2427 if (optname == -1 || optname == IPV6_HOPOPTS) { 2428 if (pktopt->ip6po_hbh) 2429 free(pktopt->ip6po_hbh, M_IP6OPT); 2430 pktopt->ip6po_hbh = NULL; 2431 } 2432 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2433 if (pktopt->ip6po_dest1) 2434 free(pktopt->ip6po_dest1, M_IP6OPT); 2435 pktopt->ip6po_dest1 = NULL; 2436 } 2437 if (optname == -1 || optname == IPV6_RTHDR) { 2438 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2439 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2440 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2441 if (pktopt->ip6po_route.ro_rt) { 2442 RTFREE(pktopt->ip6po_route.ro_rt); 2443 pktopt->ip6po_route.ro_rt = NULL; 2444 } 2445 } 2446 if (optname == -1 || optname == IPV6_DSTOPTS) { 2447 if (pktopt->ip6po_dest2) 2448 free(pktopt->ip6po_dest2, M_IP6OPT); 2449 pktopt->ip6po_dest2 = NULL; 2450 } 2451 } 2452 2453 #define PKTOPT_EXTHDRCPY(type) \ 2454 do {\ 2455 if (src->type) {\ 2456 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2457 dst->type = malloc(hlen, M_IP6OPT, canwait);\ 2458 if (dst->type == NULL && canwait == M_NOWAIT)\ 2459 goto bad;\ 2460 bcopy(src->type, dst->type, hlen);\ 2461 }\ 2462 } while (/*CONSTCOND*/ 0) 2463 2464 static int 2465 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2466 { 2467 if (dst == NULL || src == NULL) { 2468 printf("ip6_clearpktopts: invalid argument\n"); 2469 return (EINVAL); 2470 } 2471 2472 dst->ip6po_hlim = src->ip6po_hlim; 2473 dst->ip6po_tclass = src->ip6po_tclass; 2474 dst->ip6po_flags = src->ip6po_flags; 2475 dst->ip6po_minmtu = src->ip6po_minmtu; 2476 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr; 2477 if (src->ip6po_pktinfo) { 2478 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2479 M_IP6OPT, canwait); 2480 if (dst->ip6po_pktinfo == NULL) 2481 goto bad; 2482 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2483 } 2484 if (src->ip6po_nexthop) { 2485 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2486 M_IP6OPT, canwait); 2487 if (dst->ip6po_nexthop == NULL) 2488 goto bad; 2489 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop, 2490 src->ip6po_nexthop->sa_len); 2491 } 2492 PKTOPT_EXTHDRCPY(ip6po_hbh); 2493 PKTOPT_EXTHDRCPY(ip6po_dest1); 2494 PKTOPT_EXTHDRCPY(ip6po_dest2); 2495 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2496 return (0); 2497 2498 bad: 2499 ip6_clearpktopts(dst, -1); 2500 return (ENOBUFS); 2501 } 2502 #undef PKTOPT_EXTHDRCPY 2503 2504 struct ip6_pktopts * 2505 ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2506 { 2507 int error; 2508 struct ip6_pktopts *dst; 2509 2510 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2511 if (dst == NULL) 2512 return (NULL); 2513 ip6_initpktopts(dst); 2514 2515 if ((error = copypktopts(dst, src, canwait)) != 0) { 2516 free(dst, M_IP6OPT); 2517 return (NULL); 2518 } 2519 2520 return (dst); 2521 } 2522 2523 void 2524 ip6_freepcbopts(struct ip6_pktopts *pktopt) 2525 { 2526 if (pktopt == NULL) 2527 return; 2528 2529 ip6_clearpktopts(pktopt, -1); 2530 2531 free(pktopt, M_IP6OPT); 2532 } 2533 2534 /* 2535 * Set IPv6 outgoing packet options based on advanced API. 2536 */ 2537 int 2538 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2539 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto) 2540 { 2541 struct cmsghdr *cm = NULL; 2542 2543 if (control == NULL || opt == NULL) 2544 return (EINVAL); 2545 2546 ip6_initpktopts(opt); 2547 if (stickyopt) { 2548 int error; 2549 2550 /* 2551 * If stickyopt is provided, make a local copy of the options 2552 * for this particular packet, then override them by ancillary 2553 * objects. 2554 * XXX: copypktopts() does not copy the cached route to a next 2555 * hop (if any). This is not very good in terms of efficiency, 2556 * but we can allow this since this option should be rarely 2557 * used. 2558 */ 2559 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2560 return (error); 2561 } 2562 2563 /* 2564 * XXX: Currently, we assume all the optional information is stored 2565 * in a single mbuf. 2566 */ 2567 if (control->m_next) 2568 return (EINVAL); 2569 2570 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2571 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2572 int error; 2573 2574 if (control->m_len < CMSG_LEN(0)) 2575 return (EINVAL); 2576 2577 cm = mtod(control, struct cmsghdr *); 2578 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2579 return (EINVAL); 2580 if (cm->cmsg_level != IPPROTO_IPV6) 2581 continue; 2582 2583 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2584 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); 2585 if (error) 2586 return (error); 2587 } 2588 2589 return (0); 2590 } 2591 2592 /* 2593 * Set a particular packet option, as a sticky option or an ancillary data 2594 * item. "len" can be 0 only when it's a sticky option. 2595 * We have 4 cases of combination of "sticky" and "cmsg": 2596 * "sticky=0, cmsg=0": impossible 2597 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2598 * "sticky=1, cmsg=0": RFC3542 socket option 2599 * "sticky=1, cmsg=1": RFC2292 socket option 2600 */ 2601 static int 2602 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2603 struct ucred *cred, int sticky, int cmsg, int uproto) 2604 { 2605 int minmtupolicy, preftemp; 2606 int error; 2607 2608 if (!sticky && !cmsg) { 2609 #ifdef DIAGNOSTIC 2610 printf("ip6_setpktopt: impossible case\n"); 2611 #endif 2612 return (EINVAL); 2613 } 2614 2615 /* 2616 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2617 * not be specified in the context of RFC3542. Conversely, 2618 * RFC3542 types should not be specified in the context of RFC2292. 2619 */ 2620 if (!cmsg) { 2621 switch (optname) { 2622 case IPV6_2292PKTINFO: 2623 case IPV6_2292HOPLIMIT: 2624 case IPV6_2292NEXTHOP: 2625 case IPV6_2292HOPOPTS: 2626 case IPV6_2292DSTOPTS: 2627 case IPV6_2292RTHDR: 2628 case IPV6_2292PKTOPTIONS: 2629 return (ENOPROTOOPT); 2630 } 2631 } 2632 if (sticky && cmsg) { 2633 switch (optname) { 2634 case IPV6_PKTINFO: 2635 case IPV6_HOPLIMIT: 2636 case IPV6_NEXTHOP: 2637 case IPV6_HOPOPTS: 2638 case IPV6_DSTOPTS: 2639 case IPV6_RTHDRDSTOPTS: 2640 case IPV6_RTHDR: 2641 case IPV6_USE_MIN_MTU: 2642 case IPV6_DONTFRAG: 2643 case IPV6_TCLASS: 2644 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */ 2645 return (ENOPROTOOPT); 2646 } 2647 } 2648 2649 switch (optname) { 2650 case IPV6_2292PKTINFO: 2651 case IPV6_PKTINFO: 2652 { 2653 struct ifnet *ifp = NULL; 2654 struct in6_pktinfo *pktinfo; 2655 2656 if (len != sizeof(struct in6_pktinfo)) 2657 return (EINVAL); 2658 2659 pktinfo = (struct in6_pktinfo *)buf; 2660 2661 /* 2662 * An application can clear any sticky IPV6_PKTINFO option by 2663 * doing a "regular" setsockopt with ipi6_addr being 2664 * in6addr_any and ipi6_ifindex being zero. 2665 * [RFC 3542, Section 6] 2666 */ 2667 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2668 pktinfo->ipi6_ifindex == 0 && 2669 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2670 ip6_clearpktopts(opt, optname); 2671 break; 2672 } 2673 2674 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2675 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2676 return (EINVAL); 2677 } 2678 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr)) 2679 return (EINVAL); 2680 /* validate the interface index if specified. */ 2681 if (pktinfo->ipi6_ifindex > V_if_index) 2682 return (ENXIO); 2683 if (pktinfo->ipi6_ifindex) { 2684 ifp = ifnet_byindex(pktinfo->ipi6_ifindex); 2685 if (ifp == NULL) 2686 return (ENXIO); 2687 } 2688 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL || 2689 (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0)) 2690 return (ENETDOWN); 2691 2692 if (ifp != NULL && 2693 !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2694 struct in6_ifaddr *ia; 2695 2696 in6_setscope(&pktinfo->ipi6_addr, ifp, NULL); 2697 ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr); 2698 if (ia == NULL) 2699 return (EADDRNOTAVAIL); 2700 ifa_free(&ia->ia_ifa); 2701 } 2702 /* 2703 * We store the address anyway, and let in6_selectsrc() 2704 * validate the specified address. This is because ipi6_addr 2705 * may not have enough information about its scope zone, and 2706 * we may need additional information (such as outgoing 2707 * interface or the scope zone of a destination address) to 2708 * disambiguate the scope. 2709 * XXX: the delay of the validation may confuse the 2710 * application when it is used as a sticky option. 2711 */ 2712 if (opt->ip6po_pktinfo == NULL) { 2713 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2714 M_IP6OPT, M_NOWAIT); 2715 if (opt->ip6po_pktinfo == NULL) 2716 return (ENOBUFS); 2717 } 2718 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo)); 2719 break; 2720 } 2721 2722 case IPV6_2292HOPLIMIT: 2723 case IPV6_HOPLIMIT: 2724 { 2725 int *hlimp; 2726 2727 /* 2728 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2729 * to simplify the ordering among hoplimit options. 2730 */ 2731 if (optname == IPV6_HOPLIMIT && sticky) 2732 return (ENOPROTOOPT); 2733 2734 if (len != sizeof(int)) 2735 return (EINVAL); 2736 hlimp = (int *)buf; 2737 if (*hlimp < -1 || *hlimp > 255) 2738 return (EINVAL); 2739 2740 opt->ip6po_hlim = *hlimp; 2741 break; 2742 } 2743 2744 case IPV6_TCLASS: 2745 { 2746 int tclass; 2747 2748 if (len != sizeof(int)) 2749 return (EINVAL); 2750 tclass = *(int *)buf; 2751 if (tclass < -1 || tclass > 255) 2752 return (EINVAL); 2753 2754 opt->ip6po_tclass = tclass; 2755 break; 2756 } 2757 2758 case IPV6_2292NEXTHOP: 2759 case IPV6_NEXTHOP: 2760 if (cred != NULL) { 2761 error = priv_check_cred(cred, 2762 PRIV_NETINET_SETHDROPTS, 0); 2763 if (error) 2764 return (error); 2765 } 2766 2767 if (len == 0) { /* just remove the option */ 2768 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2769 break; 2770 } 2771 2772 /* check if cmsg_len is large enough for sa_len */ 2773 if (len < sizeof(struct sockaddr) || len < *buf) 2774 return (EINVAL); 2775 2776 switch (((struct sockaddr *)buf)->sa_family) { 2777 case AF_INET6: 2778 { 2779 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 2780 int error; 2781 2782 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 2783 return (EINVAL); 2784 2785 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 2786 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 2787 return (EINVAL); 2788 } 2789 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone)) 2790 != 0) { 2791 return (error); 2792 } 2793 break; 2794 } 2795 case AF_LINK: /* should eventually be supported */ 2796 default: 2797 return (EAFNOSUPPORT); 2798 } 2799 2800 /* turn off the previous option, then set the new option. */ 2801 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2802 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 2803 if (opt->ip6po_nexthop == NULL) 2804 return (ENOBUFS); 2805 bcopy(buf, opt->ip6po_nexthop, *buf); 2806 break; 2807 2808 case IPV6_2292HOPOPTS: 2809 case IPV6_HOPOPTS: 2810 { 2811 struct ip6_hbh *hbh; 2812 int hbhlen; 2813 2814 /* 2815 * XXX: We don't allow a non-privileged user to set ANY HbH 2816 * options, since per-option restriction has too much 2817 * overhead. 2818 */ 2819 if (cred != NULL) { 2820 error = priv_check_cred(cred, 2821 PRIV_NETINET_SETHDROPTS, 0); 2822 if (error) 2823 return (error); 2824 } 2825 2826 if (len == 0) { 2827 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2828 break; /* just remove the option */ 2829 } 2830 2831 /* message length validation */ 2832 if (len < sizeof(struct ip6_hbh)) 2833 return (EINVAL); 2834 hbh = (struct ip6_hbh *)buf; 2835 hbhlen = (hbh->ip6h_len + 1) << 3; 2836 if (len != hbhlen) 2837 return (EINVAL); 2838 2839 /* turn off the previous option, then set the new option. */ 2840 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2841 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 2842 if (opt->ip6po_hbh == NULL) 2843 return (ENOBUFS); 2844 bcopy(hbh, opt->ip6po_hbh, hbhlen); 2845 2846 break; 2847 } 2848 2849 case IPV6_2292DSTOPTS: 2850 case IPV6_DSTOPTS: 2851 case IPV6_RTHDRDSTOPTS: 2852 { 2853 struct ip6_dest *dest, **newdest = NULL; 2854 int destlen; 2855 2856 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */ 2857 error = priv_check_cred(cred, 2858 PRIV_NETINET_SETHDROPTS, 0); 2859 if (error) 2860 return (error); 2861 } 2862 2863 if (len == 0) { 2864 ip6_clearpktopts(opt, optname); 2865 break; /* just remove the option */ 2866 } 2867 2868 /* message length validation */ 2869 if (len < sizeof(struct ip6_dest)) 2870 return (EINVAL); 2871 dest = (struct ip6_dest *)buf; 2872 destlen = (dest->ip6d_len + 1) << 3; 2873 if (len != destlen) 2874 return (EINVAL); 2875 2876 /* 2877 * Determine the position that the destination options header 2878 * should be inserted; before or after the routing header. 2879 */ 2880 switch (optname) { 2881 case IPV6_2292DSTOPTS: 2882 /* 2883 * The old advacned API is ambiguous on this point. 2884 * Our approach is to determine the position based 2885 * according to the existence of a routing header. 2886 * Note, however, that this depends on the order of the 2887 * extension headers in the ancillary data; the 1st 2888 * part of the destination options header must appear 2889 * before the routing header in the ancillary data, 2890 * too. 2891 * RFC3542 solved the ambiguity by introducing 2892 * separate ancillary data or option types. 2893 */ 2894 if (opt->ip6po_rthdr == NULL) 2895 newdest = &opt->ip6po_dest1; 2896 else 2897 newdest = &opt->ip6po_dest2; 2898 break; 2899 case IPV6_RTHDRDSTOPTS: 2900 newdest = &opt->ip6po_dest1; 2901 break; 2902 case IPV6_DSTOPTS: 2903 newdest = &opt->ip6po_dest2; 2904 break; 2905 } 2906 2907 /* turn off the previous option, then set the new option. */ 2908 ip6_clearpktopts(opt, optname); 2909 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 2910 if (*newdest == NULL) 2911 return (ENOBUFS); 2912 bcopy(dest, *newdest, destlen); 2913 2914 break; 2915 } 2916 2917 case IPV6_2292RTHDR: 2918 case IPV6_RTHDR: 2919 { 2920 struct ip6_rthdr *rth; 2921 int rthlen; 2922 2923 if (len == 0) { 2924 ip6_clearpktopts(opt, IPV6_RTHDR); 2925 break; /* just remove the option */ 2926 } 2927 2928 /* message length validation */ 2929 if (len < sizeof(struct ip6_rthdr)) 2930 return (EINVAL); 2931 rth = (struct ip6_rthdr *)buf; 2932 rthlen = (rth->ip6r_len + 1) << 3; 2933 if (len != rthlen) 2934 return (EINVAL); 2935 2936 switch (rth->ip6r_type) { 2937 case IPV6_RTHDR_TYPE_0: 2938 if (rth->ip6r_len == 0) /* must contain one addr */ 2939 return (EINVAL); 2940 if (rth->ip6r_len % 2) /* length must be even */ 2941 return (EINVAL); 2942 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 2943 return (EINVAL); 2944 break; 2945 default: 2946 return (EINVAL); /* not supported */ 2947 } 2948 2949 /* turn off the previous option */ 2950 ip6_clearpktopts(opt, IPV6_RTHDR); 2951 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 2952 if (opt->ip6po_rthdr == NULL) 2953 return (ENOBUFS); 2954 bcopy(rth, opt->ip6po_rthdr, rthlen); 2955 2956 break; 2957 } 2958 2959 case IPV6_USE_MIN_MTU: 2960 if (len != sizeof(int)) 2961 return (EINVAL); 2962 minmtupolicy = *(int *)buf; 2963 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 2964 minmtupolicy != IP6PO_MINMTU_DISABLE && 2965 minmtupolicy != IP6PO_MINMTU_ALL) { 2966 return (EINVAL); 2967 } 2968 opt->ip6po_minmtu = minmtupolicy; 2969 break; 2970 2971 case IPV6_DONTFRAG: 2972 if (len != sizeof(int)) 2973 return (EINVAL); 2974 2975 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 2976 /* 2977 * we ignore this option for TCP sockets. 2978 * (RFC3542 leaves this case unspecified.) 2979 */ 2980 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 2981 } else 2982 opt->ip6po_flags |= IP6PO_DONTFRAG; 2983 break; 2984 2985 case IPV6_PREFER_TEMPADDR: 2986 if (len != sizeof(int)) 2987 return (EINVAL); 2988 preftemp = *(int *)buf; 2989 if (preftemp != IP6PO_TEMPADDR_SYSTEM && 2990 preftemp != IP6PO_TEMPADDR_NOTPREFER && 2991 preftemp != IP6PO_TEMPADDR_PREFER) { 2992 return (EINVAL); 2993 } 2994 opt->ip6po_prefer_tempaddr = preftemp; 2995 break; 2996 2997 default: 2998 return (ENOPROTOOPT); 2999 } /* end of switch */ 3000 3001 return (0); 3002 } 3003 3004 /* 3005 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 3006 * packet to the input queue of a specified interface. Note that this 3007 * calls the output routine of the loopback "driver", but with an interface 3008 * pointer that might NOT be &loif -- easier than replicating that code here. 3009 */ 3010 void 3011 ip6_mloopback(struct ifnet *ifp, struct mbuf *m) 3012 { 3013 struct mbuf *copym; 3014 struct ip6_hdr *ip6; 3015 3016 copym = m_copym(m, 0, M_COPYALL, M_NOWAIT); 3017 if (copym == NULL) 3018 return; 3019 3020 /* 3021 * Make sure to deep-copy IPv6 header portion in case the data 3022 * is in an mbuf cluster, so that we can safely override the IPv6 3023 * header portion later. 3024 */ 3025 if (!M_WRITABLE(copym) || 3026 copym->m_len < sizeof(struct ip6_hdr)) { 3027 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 3028 if (copym == NULL) 3029 return; 3030 } 3031 ip6 = mtod(copym, struct ip6_hdr *); 3032 /* 3033 * clear embedded scope identifiers if necessary. 3034 * in6_clearscope will touch the addresses only when necessary. 3035 */ 3036 in6_clearscope(&ip6->ip6_src); 3037 in6_clearscope(&ip6->ip6_dst); 3038 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 3039 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 | 3040 CSUM_PSEUDO_HDR; 3041 copym->m_pkthdr.csum_data = 0xffff; 3042 } 3043 if_simloop(ifp, copym, AF_INET6, 0); 3044 } 3045 3046 /* 3047 * Chop IPv6 header off from the payload. 3048 */ 3049 static int 3050 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3051 { 3052 struct mbuf *mh; 3053 struct ip6_hdr *ip6; 3054 3055 ip6 = mtod(m, struct ip6_hdr *); 3056 if (m->m_len > sizeof(*ip6)) { 3057 mh = m_gethdr(M_NOWAIT, MT_DATA); 3058 if (mh == NULL) { 3059 m_freem(m); 3060 return ENOBUFS; 3061 } 3062 m_move_pkthdr(mh, m); 3063 M_ALIGN(mh, sizeof(*ip6)); 3064 m->m_len -= sizeof(*ip6); 3065 m->m_data += sizeof(*ip6); 3066 mh->m_next = m; 3067 m = mh; 3068 m->m_len = sizeof(*ip6); 3069 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 3070 } 3071 exthdrs->ip6e_ip6 = m; 3072 return 0; 3073 } 3074 3075 /* 3076 * Compute IPv6 extension header length. 3077 */ 3078 int 3079 ip6_optlen(struct inpcb *in6p) 3080 { 3081 int len; 3082 3083 if (!in6p->in6p_outputopts) 3084 return 0; 3085 3086 len = 0; 3087 #define elen(x) \ 3088 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3089 3090 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3091 if (in6p->in6p_outputopts->ip6po_rthdr) 3092 /* dest1 is valid with rthdr only */ 3093 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3094 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3095 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3096 return len; 3097 #undef elen 3098 } 3099