1 /* $NetBSD: ip_output.c,v 1.135 2004/09/04 23:30:07 manu Exp $ */ 2 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 32 /*- 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 3. All advertising materials mentioning features or use of this software 49 * must display the following acknowledgement: 50 * This product includes software developed by the NetBSD 51 * Foundation, Inc. and its contributors. 52 * 4. Neither the name of The NetBSD Foundation nor the names of its 53 * contributors may be used to endorse or promote products derived 54 * from this software without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 57 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 58 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 59 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 60 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 61 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 62 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 63 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 64 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 65 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 66 * POSSIBILITY OF SUCH DAMAGE. 67 */ 68 69 /* 70 * Copyright (c) 1982, 1986, 1988, 1990, 1993 71 * The Regents of the University of California. All rights reserved. 72 * 73 * Redistribution and use in source and binary forms, with or without 74 * modification, are permitted provided that the following conditions 75 * are met: 76 * 1. Redistributions of source code must retain the above copyright 77 * notice, this list of conditions and the following disclaimer. 78 * 2. Redistributions in binary form must reproduce the above copyright 79 * notice, this list of conditions and the following disclaimer in the 80 * documentation and/or other materials provided with the distribution. 81 * 3. Neither the name of the University nor the names of its contributors 82 * may be used to endorse or promote products derived from this software 83 * without specific prior written permission. 84 * 85 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 86 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 87 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 88 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 89 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 90 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 91 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 92 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 93 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 94 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 95 * SUCH DAMAGE. 96 * 97 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 98 */ 99 100 #include <sys/cdefs.h> 101 __KERNEL_RCSID(0, "$NetBSD: ip_output.c,v 1.135 2004/09/04 23:30:07 manu Exp $"); 102 103 #include "opt_pfil_hooks.h" 104 #include "opt_inet.h" 105 #include "opt_ipsec.h" 106 #include "opt_mrouting.h" 107 108 #include <sys/param.h> 109 #include <sys/malloc.h> 110 #include <sys/mbuf.h> 111 #include <sys/errno.h> 112 #include <sys/protosw.h> 113 #include <sys/socket.h> 114 #include <sys/socketvar.h> 115 #ifdef FAST_IPSEC 116 #include <sys/domain.h> 117 #endif 118 #include <sys/systm.h> 119 #include <sys/proc.h> 120 121 #include <net/if.h> 122 #include <net/route.h> 123 #include <net/pfil.h> 124 125 #include <netinet/in.h> 126 #include <netinet/in_systm.h> 127 #include <netinet/ip.h> 128 #include <netinet/in_pcb.h> 129 #include <netinet/in_var.h> 130 #include <netinet/ip_var.h> 131 132 #ifdef MROUTING 133 #include <netinet/ip_mroute.h> 134 #endif 135 136 #include <machine/stdarg.h> 137 138 #ifdef IPSEC 139 #include <netinet6/ipsec.h> 140 #include <netkey/key.h> 141 #include <netkey/key_debug.h> 142 #endif /*IPSEC*/ 143 144 #ifdef FAST_IPSEC 145 #include <netipsec/ipsec.h> 146 #include <netipsec/key.h> 147 #include <netipsec/xform.h> 148 #endif /* FAST_IPSEC*/ 149 150 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *)); 151 static struct ifnet *ip_multicast_if __P((struct in_addr *, int *)); 152 static void ip_mloopback 153 __P((struct ifnet *, struct mbuf *, struct sockaddr_in *)); 154 155 #ifdef PFIL_HOOKS 156 extern struct pfil_head inet_pfil_hook; /* XXX */ 157 #endif 158 159 /* 160 * IP output. The packet in mbuf chain m contains a skeletal IP 161 * header (with len, off, ttl, proto, tos, src, dst). 162 * The mbuf chain containing the packet will be freed. 163 * The mbuf opt, if present, will not be freed. 164 */ 165 int 166 ip_output(struct mbuf *m0, ...) 167 { 168 struct ip *ip; 169 struct ifnet *ifp; 170 struct mbuf *m = m0; 171 int hlen = sizeof (struct ip); 172 int len, error = 0; 173 struct route iproute; 174 struct sockaddr_in *dst; 175 struct in_ifaddr *ia; 176 struct mbuf *opt; 177 struct route *ro; 178 int flags, sw_csum; 179 int *mtu_p; 180 u_long mtu; 181 struct ip_moptions *imo; 182 struct socket *so; 183 va_list ap; 184 #ifdef IPSEC 185 struct secpolicy *sp = NULL; 186 #endif /*IPSEC*/ 187 #ifdef FAST_IPSEC 188 struct inpcb *inp; 189 struct m_tag *mtag; 190 struct secpolicy *sp = NULL; 191 struct tdb_ident *tdbi; 192 int s; 193 #endif 194 u_int16_t ip_len; 195 196 len = 0; 197 va_start(ap, m0); 198 opt = va_arg(ap, struct mbuf *); 199 ro = va_arg(ap, struct route *); 200 flags = va_arg(ap, int); 201 imo = va_arg(ap, struct ip_moptions *); 202 so = va_arg(ap, struct socket *); 203 if (flags & IP_RETURNMTU) 204 mtu_p = va_arg(ap, int *); 205 else 206 mtu_p = NULL; 207 va_end(ap); 208 209 MCLAIM(m, &ip_tx_mowner); 210 #ifdef FAST_IPSEC 211 if (so != NULL && so->so_proto->pr_domain->dom_family == AF_INET) 212 inp = (struct inpcb *)so->so_pcb; 213 else 214 inp = NULL; 215 #endif /* FAST_IPSEC */ 216 217 #ifdef DIAGNOSTIC 218 if ((m->m_flags & M_PKTHDR) == 0) 219 panic("ip_output no HDR"); 220 #endif 221 if (opt) { 222 m = ip_insertoptions(m, opt, &len); 223 if (len >= sizeof(struct ip)) 224 hlen = len; 225 } 226 ip = mtod(m, struct ip *); 227 /* 228 * Fill in IP header. 229 */ 230 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 231 ip->ip_v = IPVERSION; 232 ip->ip_off = htons(0); 233 ip->ip_id = ip_newid(); 234 ip->ip_hl = hlen >> 2; 235 ipstat.ips_localout++; 236 } else { 237 hlen = ip->ip_hl << 2; 238 } 239 /* 240 * Route packet. 241 */ 242 if (ro == 0) { 243 ro = &iproute; 244 bzero((caddr_t)ro, sizeof (*ro)); 245 } 246 dst = satosin(&ro->ro_dst); 247 /* 248 * If there is a cached route, 249 * check that it is to the same destination 250 * and is still up. If not, free it and try again. 251 * The address family should also be checked in case of sharing the 252 * cache with IPv6. 253 */ 254 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 255 dst->sin_family != AF_INET || 256 !in_hosteq(dst->sin_addr, ip->ip_dst))) { 257 RTFREE(ro->ro_rt); 258 ro->ro_rt = (struct rtentry *)0; 259 } 260 if (ro->ro_rt == 0) { 261 bzero(dst, sizeof(*dst)); 262 dst->sin_family = AF_INET; 263 dst->sin_len = sizeof(*dst); 264 dst->sin_addr = ip->ip_dst; 265 } 266 /* 267 * If routing to interface only, 268 * short circuit routing lookup. 269 */ 270 if (flags & IP_ROUTETOIF) { 271 if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) { 272 ipstat.ips_noroute++; 273 error = ENETUNREACH; 274 goto bad; 275 } 276 ifp = ia->ia_ifp; 277 mtu = ifp->if_mtu; 278 ip->ip_ttl = 1; 279 } else if ((IN_MULTICAST(ip->ip_dst.s_addr) || 280 ip->ip_dst.s_addr == INADDR_BROADCAST) && 281 imo != NULL && imo->imo_multicast_ifp != NULL) { 282 ifp = imo->imo_multicast_ifp; 283 mtu = ifp->if_mtu; 284 IFP_TO_IA(ifp, ia); 285 } else { 286 if (ro->ro_rt == 0) 287 rtalloc(ro); 288 if (ro->ro_rt == 0) { 289 ipstat.ips_noroute++; 290 error = EHOSTUNREACH; 291 goto bad; 292 } 293 ia = ifatoia(ro->ro_rt->rt_ifa); 294 ifp = ro->ro_rt->rt_ifp; 295 if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0) 296 mtu = ifp->if_mtu; 297 ro->ro_rt->rt_use++; 298 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 299 dst = satosin(ro->ro_rt->rt_gateway); 300 } 301 if (IN_MULTICAST(ip->ip_dst.s_addr) || 302 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 303 struct in_multi *inm; 304 305 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 306 M_BCAST : M_MCAST; 307 /* 308 * IP destination address is multicast. Make sure "dst" 309 * still points to the address in "ro". (It may have been 310 * changed to point to a gateway address, above.) 311 */ 312 dst = satosin(&ro->ro_dst); 313 /* 314 * See if the caller provided any multicast options 315 */ 316 if (imo != NULL) 317 ip->ip_ttl = imo->imo_multicast_ttl; 318 else 319 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 320 321 /* 322 * if we don't know the outgoing ifp yet, we can't generate 323 * output 324 */ 325 if (!ifp) { 326 ipstat.ips_noroute++; 327 error = ENETUNREACH; 328 goto bad; 329 } 330 331 /* 332 * If the packet is multicast or broadcast, confirm that 333 * the outgoing interface can transmit it. 334 */ 335 if (((m->m_flags & M_MCAST) && 336 (ifp->if_flags & IFF_MULTICAST) == 0) || 337 ((m->m_flags & M_BCAST) && 338 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 339 ipstat.ips_noroute++; 340 error = ENETUNREACH; 341 goto bad; 342 } 343 /* 344 * If source address not specified yet, use an address 345 * of outgoing interface. 346 */ 347 if (in_nullhost(ip->ip_src)) { 348 struct in_ifaddr *ia; 349 350 IFP_TO_IA(ifp, ia); 351 if (!ia) { 352 error = EADDRNOTAVAIL; 353 goto bad; 354 } 355 ip->ip_src = ia->ia_addr.sin_addr; 356 } 357 358 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 359 if (inm != NULL && 360 (imo == NULL || imo->imo_multicast_loop)) { 361 /* 362 * If we belong to the destination multicast group 363 * on the outgoing interface, and the caller did not 364 * forbid loopback, loop back a copy. 365 */ 366 ip_mloopback(ifp, m, dst); 367 } 368 #ifdef MROUTING 369 else { 370 /* 371 * If we are acting as a multicast router, perform 372 * multicast forwarding as if the packet had just 373 * arrived on the interface to which we are about 374 * to send. The multicast forwarding function 375 * recursively calls this function, using the 376 * IP_FORWARDING flag to prevent infinite recursion. 377 * 378 * Multicasts that are looped back by ip_mloopback(), 379 * above, will be forwarded by the ip_input() routine, 380 * if necessary. 381 */ 382 extern struct socket *ip_mrouter; 383 384 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 385 if (ip_mforward(m, ifp) != 0) { 386 m_freem(m); 387 goto done; 388 } 389 } 390 } 391 #endif 392 /* 393 * Multicasts with a time-to-live of zero may be looped- 394 * back, above, but must not be transmitted on a network. 395 * Also, multicasts addressed to the loopback interface 396 * are not sent -- the above call to ip_mloopback() will 397 * loop back a copy if this host actually belongs to the 398 * destination group on the loopback interface. 399 */ 400 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 401 m_freem(m); 402 goto done; 403 } 404 405 goto sendit; 406 } 407 #ifndef notdef 408 /* 409 * If source address not specified yet, use address 410 * of outgoing interface. 411 */ 412 if (in_nullhost(ip->ip_src)) 413 ip->ip_src = ia->ia_addr.sin_addr; 414 #endif 415 416 /* 417 * packets with Class-D address as source are not valid per 418 * RFC 1112 419 */ 420 if (IN_MULTICAST(ip->ip_src.s_addr)) { 421 ipstat.ips_odropped++; 422 error = EADDRNOTAVAIL; 423 goto bad; 424 } 425 426 /* 427 * Look for broadcast address and 428 * and verify user is allowed to send 429 * such a packet. 430 */ 431 if (in_broadcast(dst->sin_addr, ifp)) { 432 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 433 error = EADDRNOTAVAIL; 434 goto bad; 435 } 436 if ((flags & IP_ALLOWBROADCAST) == 0) { 437 error = EACCES; 438 goto bad; 439 } 440 /* don't allow broadcast messages to be fragmented */ 441 if (ntohs(ip->ip_len) > ifp->if_mtu) { 442 error = EMSGSIZE; 443 goto bad; 444 } 445 m->m_flags |= M_BCAST; 446 } else 447 m->m_flags &= ~M_BCAST; 448 449 sendit: 450 /* 451 * If we're doing Path MTU Discovery, we need to set DF unless 452 * the route's MTU is locked. 453 */ 454 if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL && 455 (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0) 456 ip->ip_off |= htons(IP_DF); 457 458 /* Remember the current ip_len */ 459 ip_len = ntohs(ip->ip_len); 460 461 #ifdef IPSEC 462 /* get SP for this packet */ 463 if (so == NULL) 464 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 465 flags, &error); 466 else { 467 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 468 IPSEC_DIR_OUTBOUND)) 469 goto skip_ipsec; 470 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 471 } 472 473 if (sp == NULL) { 474 ipsecstat.out_inval++; 475 goto bad; 476 } 477 478 error = 0; 479 480 /* check policy */ 481 switch (sp->policy) { 482 case IPSEC_POLICY_DISCARD: 483 /* 484 * This packet is just discarded. 485 */ 486 ipsecstat.out_polvio++; 487 goto bad; 488 489 case IPSEC_POLICY_BYPASS: 490 case IPSEC_POLICY_NONE: 491 /* no need to do IPsec. */ 492 goto skip_ipsec; 493 494 case IPSEC_POLICY_IPSEC: 495 if (sp->req == NULL) { 496 /* XXX should be panic ? */ 497 printf("ip_output: No IPsec request specified.\n"); 498 error = EINVAL; 499 goto bad; 500 } 501 break; 502 503 case IPSEC_POLICY_ENTRUST: 504 default: 505 printf("ip_output: Invalid policy found. %d\n", sp->policy); 506 } 507 508 /* 509 * ipsec4_output() expects ip_len and ip_off in network 510 * order. They have been set to network order above. 511 */ 512 513 { 514 struct ipsec_output_state state; 515 bzero(&state, sizeof(state)); 516 state.m = m; 517 if (flags & IP_ROUTETOIF) { 518 state.ro = &iproute; 519 bzero(&iproute, sizeof(iproute)); 520 } else 521 state.ro = ro; 522 state.dst = (struct sockaddr *)dst; 523 524 /* 525 * We can't defer the checksum of payload data if 526 * we're about to encrypt/authenticate it. 527 * 528 * XXX When we support crypto offloading functions of 529 * XXX network interfaces, we need to reconsider this, 530 * XXX since it's likely that they'll support checksumming, 531 * XXX as well. 532 */ 533 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 534 in_delayed_cksum(m); 535 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 536 } 537 538 error = ipsec4_output(&state, sp, flags); 539 540 m = state.m; 541 if (flags & IP_ROUTETOIF) { 542 /* 543 * if we have tunnel mode SA, we may need to ignore 544 * IP_ROUTETOIF. 545 */ 546 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 547 flags &= ~IP_ROUTETOIF; 548 ro = state.ro; 549 } 550 } else 551 ro = state.ro; 552 dst = (struct sockaddr_in *)state.dst; 553 if (error) { 554 /* mbuf is already reclaimed in ipsec4_output. */ 555 m0 = NULL; 556 switch (error) { 557 case EHOSTUNREACH: 558 case ENETUNREACH: 559 case EMSGSIZE: 560 case ENOBUFS: 561 case ENOMEM: 562 break; 563 default: 564 printf("ip4_output (ipsec): error code %d\n", error); 565 /*fall through*/ 566 case ENOENT: 567 /* don't show these error codes to the user */ 568 error = 0; 569 break; 570 } 571 goto bad; 572 } 573 574 /* be sure to update variables that are affected by ipsec4_output() */ 575 ip = mtod(m, struct ip *); 576 hlen = ip->ip_hl << 2; 577 ip_len = ntohs(ip->ip_len); 578 579 if (ro->ro_rt == NULL) { 580 if ((flags & IP_ROUTETOIF) == 0) { 581 printf("ip_output: " 582 "can't update route after IPsec processing\n"); 583 error = EHOSTUNREACH; /*XXX*/ 584 goto bad; 585 } 586 } else { 587 /* nobody uses ia beyond here */ 588 if (state.encap) { 589 ifp = ro->ro_rt->rt_ifp; 590 if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0) 591 mtu = ifp->if_mtu; 592 } 593 } 594 } 595 skip_ipsec: 596 #endif /*IPSEC*/ 597 #ifdef FAST_IPSEC 598 /* 599 * Check the security policy (SP) for the packet and, if 600 * required, do IPsec-related processing. There are two 601 * cases here; the first time a packet is sent through 602 * it will be untagged and handled by ipsec4_checkpolicy. 603 * If the packet is resubmitted to ip_output (e.g. after 604 * AH, ESP, etc. processing), there will be a tag to bypass 605 * the lookup and related policy checking. 606 */ 607 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 608 s = splsoftnet(); 609 if (mtag != NULL) { 610 tdbi = (struct tdb_ident *)(mtag + 1); 611 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 612 if (sp == NULL) 613 error = -EINVAL; /* force silent drop */ 614 m_tag_delete(m, mtag); 615 } else { 616 if (inp != NULL && 617 IPSEC_PCB_SKIP_IPSEC(inp->inp_sp, IPSEC_DIR_OUTBOUND)) 618 goto spd_done; 619 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 620 &error, inp); 621 } 622 /* 623 * There are four return cases: 624 * sp != NULL apply IPsec policy 625 * sp == NULL, error == 0 no IPsec handling needed 626 * sp == NULL, error == -EINVAL discard packet w/o error 627 * sp == NULL, error != 0 discard packet, report error 628 */ 629 if (sp != NULL) { 630 /* Loop detection, check if ipsec processing already done */ 631 IPSEC_ASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 632 for (mtag = m_tag_first(m); mtag != NULL; 633 mtag = m_tag_next(m, mtag)) { 634 #ifdef MTAG_ABI_COMPAT 635 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 636 continue; 637 #endif 638 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 639 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 640 continue; 641 /* 642 * Check if policy has an SA associated with it. 643 * This can happen when an SP has yet to acquire 644 * an SA; e.g. on first reference. If it occurs, 645 * then we let ipsec4_process_packet do its thing. 646 */ 647 if (sp->req->sav == NULL) 648 break; 649 tdbi = (struct tdb_ident *)(mtag + 1); 650 if (tdbi->spi == sp->req->sav->spi && 651 tdbi->proto == sp->req->sav->sah->saidx.proto && 652 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 653 sizeof (union sockaddr_union)) == 0) { 654 /* 655 * No IPsec processing is needed, free 656 * reference to SP. 657 * 658 * NB: null pointer to avoid free at 659 * done: below. 660 */ 661 KEY_FREESP(&sp), sp = NULL; 662 splx(s); 663 goto spd_done; 664 } 665 } 666 667 /* 668 * Do delayed checksums now because we send before 669 * this is done in the normal processing path. 670 */ 671 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 672 in_delayed_cksum(m); 673 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 674 } 675 676 #ifdef __FreeBSD__ 677 ip->ip_len = htons(ip->ip_len); 678 ip->ip_off = htons(ip->ip_off); 679 #endif 680 681 /* NB: callee frees mbuf */ 682 error = ipsec4_process_packet(m, sp->req, flags, 0); 683 /* 684 * Preserve KAME behaviour: ENOENT can be returned 685 * when an SA acquire is in progress. Don't propagate 686 * this to user-level; it confuses applications. 687 * 688 * XXX this will go away when the SADB is redone. 689 */ 690 if (error == ENOENT) 691 error = 0; 692 splx(s); 693 goto done; 694 } else { 695 splx(s); 696 697 if (error != 0) { 698 /* 699 * Hack: -EINVAL is used to signal that a packet 700 * should be silently discarded. This is typically 701 * because we asked key management for an SA and 702 * it was delayed (e.g. kicked up to IKE). 703 */ 704 if (error == -EINVAL) 705 error = 0; 706 goto bad; 707 } else { 708 /* No IPsec processing for this packet. */ 709 } 710 #ifdef notyet 711 /* 712 * If deferred crypto processing is needed, check that 713 * the interface supports it. 714 */ 715 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 716 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 717 /* notify IPsec to do its own crypto */ 718 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 719 error = EHOSTUNREACH; 720 goto bad; 721 } 722 #endif 723 } 724 spd_done: 725 #endif /* FAST_IPSEC */ 726 727 #ifdef PFIL_HOOKS 728 /* 729 * Run through list of hooks for output packets. 730 */ 731 if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 732 goto done; 733 if (m == NULL) 734 goto done; 735 736 ip = mtod(m, struct ip *); 737 hlen = ip->ip_hl << 2; 738 #endif /* PFIL_HOOKS */ 739 740 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 741 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 742 /* 743 * If small enough for mtu of path, can just send directly. 744 */ 745 if (ip_len <= mtu) { 746 #if IFA_STATS 747 /* 748 * search for the source address structure to 749 * maintain output statistics. 750 */ 751 INADDR_TO_IA(ip->ip_src, ia); 752 if (ia) 753 ia->ia_ifa.ifa_data.ifad_outbytes += ip_len; 754 #endif 755 /* 756 * Always initialize the sum to 0! Some HW assisted 757 * checksumming requires this. 758 */ 759 ip->ip_sum = 0; 760 761 /* 762 * Perform any checksums that the hardware can't do 763 * for us. 764 * 765 * XXX Does any hardware require the {th,uh}_sum 766 * XXX fields to be 0? 767 */ 768 if (sw_csum & M_CSUM_IPv4) { 769 ip->ip_sum = in_cksum(m, hlen); 770 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 771 } 772 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 773 in_delayed_cksum(m); 774 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 775 } 776 777 #ifdef IPSEC 778 /* clean ipsec history once it goes out of the node */ 779 ipsec_delaux(m); 780 #endif 781 error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt); 782 goto done; 783 } 784 785 /* 786 * We can't use HW checksumming if we're about to 787 * to fragment the packet. 788 * 789 * XXX Some hardware can do this. 790 */ 791 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 792 in_delayed_cksum(m); 793 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 794 } 795 796 /* 797 * Too large for interface; fragment if possible. 798 * Must be able to put at least 8 bytes per fragment. 799 */ 800 if (ntohs(ip->ip_off) & IP_DF) { 801 if (flags & IP_RETURNMTU) 802 *mtu_p = mtu; 803 error = EMSGSIZE; 804 ipstat.ips_cantfrag++; 805 goto bad; 806 } 807 808 error = ip_fragment(m, ifp, mtu); 809 if (error) { 810 m = NULL; 811 goto bad; 812 } 813 814 for (; m; m = m0) { 815 m0 = m->m_nextpkt; 816 m->m_nextpkt = 0; 817 if (error == 0) { 818 #if IFA_STATS 819 /* 820 * search for the source address structure to 821 * maintain output statistics. 822 */ 823 INADDR_TO_IA(ip->ip_src, ia); 824 if (ia) { 825 ia->ia_ifa.ifa_data.ifad_outbytes += 826 ntohs(ip->ip_len); 827 } 828 #endif 829 #ifdef IPSEC 830 /* clean ipsec history once it goes out of the node */ 831 ipsec_delaux(m); 832 #endif 833 KASSERT((m->m_pkthdr.csum_flags & 834 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 835 error = (*ifp->if_output)(ifp, m, sintosa(dst), 836 ro->ro_rt); 837 } else 838 m_freem(m); 839 } 840 841 if (error == 0) 842 ipstat.ips_fragmented++; 843 done: 844 if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) { 845 RTFREE(ro->ro_rt); 846 ro->ro_rt = 0; 847 } 848 849 #ifdef IPSEC 850 if (sp != NULL) { 851 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 852 printf("DP ip_output call free SP:%p\n", sp)); 853 key_freesp(sp); 854 } 855 #endif /* IPSEC */ 856 #ifdef FAST_IPSEC 857 if (sp != NULL) 858 KEY_FREESP(&sp); 859 #endif /* FAST_IPSEC */ 860 861 return (error); 862 bad: 863 m_freem(m); 864 goto done; 865 } 866 867 int 868 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 869 { 870 struct ip *ip, *mhip; 871 struct mbuf *m0; 872 int len, hlen, off; 873 int mhlen, firstlen; 874 struct mbuf **mnext; 875 int sw_csum = m->m_pkthdr.csum_flags; 876 int fragments = 0; 877 int s; 878 int error = 0; 879 880 ip = mtod(m, struct ip *); 881 hlen = ip->ip_hl << 2; 882 if (ifp != NULL) 883 sw_csum &= ~ifp->if_csum_flags_tx; 884 885 len = (mtu - hlen) &~ 7; 886 if (len < 8) { 887 m_freem(m); 888 return (EMSGSIZE); 889 } 890 891 firstlen = len; 892 mnext = &m->m_nextpkt; 893 894 /* 895 * Loop through length of segment after first fragment, 896 * make new header and copy data of each part and link onto chain. 897 */ 898 m0 = m; 899 mhlen = sizeof (struct ip); 900 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 901 MGETHDR(m, M_DONTWAIT, MT_HEADER); 902 if (m == 0) { 903 error = ENOBUFS; 904 ipstat.ips_odropped++; 905 goto sendorfree; 906 } 907 MCLAIM(m, m0->m_owner); 908 *mnext = m; 909 mnext = &m->m_nextpkt; 910 m->m_data += max_linkhdr; 911 mhip = mtod(m, struct ip *); 912 *mhip = *ip; 913 /* we must inherit MCAST and BCAST flags */ 914 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 915 if (hlen > sizeof (struct ip)) { 916 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 917 mhip->ip_hl = mhlen >> 2; 918 } 919 m->m_len = mhlen; 920 mhip->ip_off = ((off - hlen) >> 3) + 921 (ntohs(ip->ip_off) & ~IP_MF); 922 if (ip->ip_off & htons(IP_MF)) 923 mhip->ip_off |= IP_MF; 924 if (off + len >= ntohs(ip->ip_len)) 925 len = ntohs(ip->ip_len) - off; 926 else 927 mhip->ip_off |= IP_MF; 928 HTONS(mhip->ip_off); 929 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 930 m->m_next = m_copy(m0, off, len); 931 if (m->m_next == 0) { 932 error = ENOBUFS; /* ??? */ 933 ipstat.ips_odropped++; 934 goto sendorfree; 935 } 936 m->m_pkthdr.len = mhlen + len; 937 m->m_pkthdr.rcvif = (struct ifnet *)0; 938 mhip->ip_sum = 0; 939 if (sw_csum & M_CSUM_IPv4) { 940 mhip->ip_sum = in_cksum(m, mhlen); 941 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 942 } else { 943 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 944 } 945 ipstat.ips_ofragments++; 946 fragments++; 947 } 948 /* 949 * Update first fragment by trimming what's been copied out 950 * and updating header, then send each fragment (in order). 951 */ 952 m = m0; 953 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 954 m->m_pkthdr.len = hlen + firstlen; 955 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 956 ip->ip_off |= htons(IP_MF); 957 ip->ip_sum = 0; 958 if (sw_csum & M_CSUM_IPv4) { 959 ip->ip_sum = in_cksum(m, hlen); 960 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 961 } else { 962 KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4); 963 } 964 sendorfree: 965 /* 966 * If there is no room for all the fragments, don't queue 967 * any of them. 968 */ 969 if (ifp != NULL) { 970 s = splnet(); 971 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 972 error == 0) { 973 error = ENOBUFS; 974 ipstat.ips_odropped++; 975 IFQ_INC_DROPS(&ifp->if_snd); 976 } 977 splx(s); 978 } 979 if (error) { 980 for (m = m0; m; m = m0) { 981 m0 = m->m_nextpkt; 982 m->m_nextpkt = NULL; 983 m_freem(m); 984 } 985 } 986 return (error); 987 } 988 989 /* 990 * Process a delayed payload checksum calculation. 991 */ 992 void 993 in_delayed_cksum(struct mbuf *m) 994 { 995 struct ip *ip; 996 u_int16_t csum, offset; 997 998 ip = mtod(m, struct ip *); 999 offset = ip->ip_hl << 2; 1000 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 1001 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 1002 csum = 0xffff; 1003 1004 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1005 1006 if ((offset + sizeof(u_int16_t)) > m->m_len) { 1007 /* This happen when ip options were inserted 1008 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 1009 m->m_len, offset, ip->ip_p); 1010 */ 1011 m_copyback(m, offset, sizeof(csum), (caddr_t) &csum); 1012 } else 1013 *(u_int16_t *)(mtod(m, caddr_t) + offset) = csum; 1014 } 1015 1016 /* 1017 * Determine the maximum length of the options to be inserted; 1018 * we would far rather allocate too much space rather than too little. 1019 */ 1020 1021 u_int 1022 ip_optlen(inp) 1023 struct inpcb *inp; 1024 { 1025 struct mbuf *m = inp->inp_options; 1026 1027 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) 1028 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 1029 else 1030 return 0; 1031 } 1032 1033 1034 /* 1035 * Insert IP options into preformed packet. 1036 * Adjust IP destination as required for IP source routing, 1037 * as indicated by a non-zero in_addr at the start of the options. 1038 */ 1039 static struct mbuf * 1040 ip_insertoptions(m, opt, phlen) 1041 struct mbuf *m; 1042 struct mbuf *opt; 1043 int *phlen; 1044 { 1045 struct ipoption *p = mtod(opt, struct ipoption *); 1046 struct mbuf *n; 1047 struct ip *ip = mtod(m, struct ip *); 1048 unsigned optlen; 1049 1050 optlen = opt->m_len - sizeof(p->ipopt_dst); 1051 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 1052 return (m); /* XXX should fail */ 1053 if (!in_nullhost(p->ipopt_dst)) 1054 ip->ip_dst = p->ipopt_dst; 1055 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 1056 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1057 if (n == 0) 1058 return (m); 1059 MCLAIM(n, m->m_owner); 1060 M_COPY_PKTHDR(n, m); 1061 m_tag_delete_chain(m, NULL); 1062 m->m_flags &= ~M_PKTHDR; 1063 m->m_len -= sizeof(struct ip); 1064 m->m_data += sizeof(struct ip); 1065 n->m_next = m; 1066 m = n; 1067 m->m_len = optlen + sizeof(struct ip); 1068 m->m_data += max_linkhdr; 1069 bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1070 } else { 1071 m->m_data -= optlen; 1072 m->m_len += optlen; 1073 memmove(mtod(m, caddr_t), ip, sizeof(struct ip)); 1074 } 1075 m->m_pkthdr.len += optlen; 1076 ip = mtod(m, struct ip *); 1077 bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen); 1078 *phlen = sizeof(struct ip) + optlen; 1079 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 1080 return (m); 1081 } 1082 1083 /* 1084 * Copy options from ip to jp, 1085 * omitting those not copied during fragmentation. 1086 */ 1087 int 1088 ip_optcopy(ip, jp) 1089 struct ip *ip, *jp; 1090 { 1091 u_char *cp, *dp; 1092 int opt, optlen, cnt; 1093 1094 cp = (u_char *)(ip + 1); 1095 dp = (u_char *)(jp + 1); 1096 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1097 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1098 opt = cp[0]; 1099 if (opt == IPOPT_EOL) 1100 break; 1101 if (opt == IPOPT_NOP) { 1102 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1103 *dp++ = IPOPT_NOP; 1104 optlen = 1; 1105 continue; 1106 } 1107 #ifdef DIAGNOSTIC 1108 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1109 panic("malformed IPv4 option passed to ip_optcopy"); 1110 #endif 1111 optlen = cp[IPOPT_OLEN]; 1112 #ifdef DIAGNOSTIC 1113 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1114 panic("malformed IPv4 option passed to ip_optcopy"); 1115 #endif 1116 /* bogus lengths should have been caught by ip_dooptions */ 1117 if (optlen > cnt) 1118 optlen = cnt; 1119 if (IPOPT_COPIED(opt)) { 1120 bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen); 1121 dp += optlen; 1122 } 1123 } 1124 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1125 *dp++ = IPOPT_EOL; 1126 return (optlen); 1127 } 1128 1129 /* 1130 * IP socket option processing. 1131 */ 1132 int 1133 ip_ctloutput(op, so, level, optname, mp) 1134 int op; 1135 struct socket *so; 1136 int level, optname; 1137 struct mbuf **mp; 1138 { 1139 struct inpcb *inp = sotoinpcb(so); 1140 struct mbuf *m = *mp; 1141 int optval = 0; 1142 int error = 0; 1143 #if defined(IPSEC) || defined(FAST_IPSEC) 1144 struct proc *p = curproc; /*XXX*/ 1145 #endif 1146 1147 if (level != IPPROTO_IP) { 1148 error = EINVAL; 1149 if (op == PRCO_SETOPT && *mp) 1150 (void) m_free(*mp); 1151 } else switch (op) { 1152 1153 case PRCO_SETOPT: 1154 switch (optname) { 1155 case IP_OPTIONS: 1156 #ifdef notyet 1157 case IP_RETOPTS: 1158 return (ip_pcbopts(optname, &inp->inp_options, m)); 1159 #else 1160 return (ip_pcbopts(&inp->inp_options, m)); 1161 #endif 1162 1163 case IP_TOS: 1164 case IP_TTL: 1165 case IP_RECVOPTS: 1166 case IP_RECVRETOPTS: 1167 case IP_RECVDSTADDR: 1168 case IP_RECVIF: 1169 if (m == NULL || m->m_len != sizeof(int)) 1170 error = EINVAL; 1171 else { 1172 optval = *mtod(m, int *); 1173 switch (optname) { 1174 1175 case IP_TOS: 1176 inp->inp_ip.ip_tos = optval; 1177 break; 1178 1179 case IP_TTL: 1180 inp->inp_ip.ip_ttl = optval; 1181 break; 1182 #define OPTSET(bit) \ 1183 if (optval) \ 1184 inp->inp_flags |= bit; \ 1185 else \ 1186 inp->inp_flags &= ~bit; 1187 1188 case IP_RECVOPTS: 1189 OPTSET(INP_RECVOPTS); 1190 break; 1191 1192 case IP_RECVRETOPTS: 1193 OPTSET(INP_RECVRETOPTS); 1194 break; 1195 1196 case IP_RECVDSTADDR: 1197 OPTSET(INP_RECVDSTADDR); 1198 break; 1199 1200 case IP_RECVIF: 1201 OPTSET(INP_RECVIF); 1202 break; 1203 } 1204 } 1205 break; 1206 #undef OPTSET 1207 1208 case IP_MULTICAST_IF: 1209 case IP_MULTICAST_TTL: 1210 case IP_MULTICAST_LOOP: 1211 case IP_ADD_MEMBERSHIP: 1212 case IP_DROP_MEMBERSHIP: 1213 error = ip_setmoptions(optname, &inp->inp_moptions, m); 1214 break; 1215 1216 case IP_PORTRANGE: 1217 if (m == 0 || m->m_len != sizeof(int)) 1218 error = EINVAL; 1219 else { 1220 optval = *mtod(m, int *); 1221 1222 switch (optval) { 1223 1224 case IP_PORTRANGE_DEFAULT: 1225 case IP_PORTRANGE_HIGH: 1226 inp->inp_flags &= ~(INP_LOWPORT); 1227 break; 1228 1229 case IP_PORTRANGE_LOW: 1230 inp->inp_flags |= INP_LOWPORT; 1231 break; 1232 1233 default: 1234 error = EINVAL; 1235 break; 1236 } 1237 } 1238 break; 1239 1240 #if defined(IPSEC) || defined(FAST_IPSEC) 1241 case IP_IPSEC_POLICY: 1242 { 1243 caddr_t req = NULL; 1244 size_t len = 0; 1245 int priv = 0; 1246 1247 #ifdef __NetBSD__ 1248 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1249 priv = 0; 1250 else 1251 priv = 1; 1252 #else 1253 priv = (in6p->in6p_socket->so_state & SS_PRIV); 1254 #endif 1255 if (m) { 1256 req = mtod(m, caddr_t); 1257 len = m->m_len; 1258 } 1259 error = ipsec4_set_policy(inp, optname, req, len, priv); 1260 break; 1261 } 1262 #endif /*IPSEC*/ 1263 1264 default: 1265 error = ENOPROTOOPT; 1266 break; 1267 } 1268 if (m) 1269 (void)m_free(m); 1270 break; 1271 1272 case PRCO_GETOPT: 1273 switch (optname) { 1274 case IP_OPTIONS: 1275 case IP_RETOPTS: 1276 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1277 MCLAIM(m, so->so_mowner); 1278 if (inp->inp_options) { 1279 m->m_len = inp->inp_options->m_len; 1280 bcopy(mtod(inp->inp_options, caddr_t), 1281 mtod(m, caddr_t), (unsigned)m->m_len); 1282 } else 1283 m->m_len = 0; 1284 break; 1285 1286 case IP_TOS: 1287 case IP_TTL: 1288 case IP_RECVOPTS: 1289 case IP_RECVRETOPTS: 1290 case IP_RECVDSTADDR: 1291 case IP_RECVIF: 1292 case IP_ERRORMTU: 1293 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1294 MCLAIM(m, so->so_mowner); 1295 m->m_len = sizeof(int); 1296 switch (optname) { 1297 1298 case IP_TOS: 1299 optval = inp->inp_ip.ip_tos; 1300 break; 1301 1302 case IP_TTL: 1303 optval = inp->inp_ip.ip_ttl; 1304 break; 1305 1306 case IP_ERRORMTU: 1307 optval = inp->inp_errormtu; 1308 break; 1309 1310 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1311 1312 case IP_RECVOPTS: 1313 optval = OPTBIT(INP_RECVOPTS); 1314 break; 1315 1316 case IP_RECVRETOPTS: 1317 optval = OPTBIT(INP_RECVRETOPTS); 1318 break; 1319 1320 case IP_RECVDSTADDR: 1321 optval = OPTBIT(INP_RECVDSTADDR); 1322 break; 1323 1324 case IP_RECVIF: 1325 optval = OPTBIT(INP_RECVIF); 1326 break; 1327 } 1328 *mtod(m, int *) = optval; 1329 break; 1330 1331 #if 0 /* defined(IPSEC) || defined(FAST_IPSEC) */ 1332 /* XXX: code broken */ 1333 case IP_IPSEC_POLICY: 1334 { 1335 caddr_t req = NULL; 1336 size_t len = 0; 1337 1338 if (m) { 1339 req = mtod(m, caddr_t); 1340 len = m->m_len; 1341 } 1342 error = ipsec4_get_policy(inp, req, len, mp); 1343 break; 1344 } 1345 #endif /*IPSEC*/ 1346 1347 case IP_MULTICAST_IF: 1348 case IP_MULTICAST_TTL: 1349 case IP_MULTICAST_LOOP: 1350 case IP_ADD_MEMBERSHIP: 1351 case IP_DROP_MEMBERSHIP: 1352 error = ip_getmoptions(optname, inp->inp_moptions, mp); 1353 if (*mp) 1354 MCLAIM(*mp, so->so_mowner); 1355 break; 1356 1357 case IP_PORTRANGE: 1358 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1359 MCLAIM(m, so->so_mowner); 1360 m->m_len = sizeof(int); 1361 1362 if (inp->inp_flags & INP_LOWPORT) 1363 optval = IP_PORTRANGE_LOW; 1364 else 1365 optval = IP_PORTRANGE_DEFAULT; 1366 1367 *mtod(m, int *) = optval; 1368 break; 1369 1370 default: 1371 error = ENOPROTOOPT; 1372 break; 1373 } 1374 break; 1375 } 1376 return (error); 1377 } 1378 1379 /* 1380 * Set up IP options in pcb for insertion in output packets. 1381 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1382 * with destination address if source routed. 1383 */ 1384 int 1385 #ifdef notyet 1386 ip_pcbopts(optname, pcbopt, m) 1387 int optname; 1388 #else 1389 ip_pcbopts(pcbopt, m) 1390 #endif 1391 struct mbuf **pcbopt; 1392 struct mbuf *m; 1393 { 1394 int cnt, optlen; 1395 u_char *cp; 1396 u_char opt; 1397 1398 /* turn off any old options */ 1399 if (*pcbopt) 1400 (void)m_free(*pcbopt); 1401 *pcbopt = 0; 1402 if (m == (struct mbuf *)0 || m->m_len == 0) { 1403 /* 1404 * Only turning off any previous options. 1405 */ 1406 if (m) 1407 (void)m_free(m); 1408 return (0); 1409 } 1410 1411 #ifndef __vax__ 1412 if (m->m_len % sizeof(int32_t)) 1413 goto bad; 1414 #endif 1415 /* 1416 * IP first-hop destination address will be stored before 1417 * actual options; move other options back 1418 * and clear it when none present. 1419 */ 1420 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1421 goto bad; 1422 cnt = m->m_len; 1423 m->m_len += sizeof(struct in_addr); 1424 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1425 memmove(cp, mtod(m, caddr_t), (unsigned)cnt); 1426 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1427 1428 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1429 opt = cp[IPOPT_OPTVAL]; 1430 if (opt == IPOPT_EOL) 1431 break; 1432 if (opt == IPOPT_NOP) 1433 optlen = 1; 1434 else { 1435 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1436 goto bad; 1437 optlen = cp[IPOPT_OLEN]; 1438 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1439 goto bad; 1440 } 1441 switch (opt) { 1442 1443 default: 1444 break; 1445 1446 case IPOPT_LSRR: 1447 case IPOPT_SSRR: 1448 /* 1449 * user process specifies route as: 1450 * ->A->B->C->D 1451 * D must be our final destination (but we can't 1452 * check that since we may not have connected yet). 1453 * A is first hop destination, which doesn't appear in 1454 * actual IP option, but is stored before the options. 1455 */ 1456 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1457 goto bad; 1458 m->m_len -= sizeof(struct in_addr); 1459 cnt -= sizeof(struct in_addr); 1460 optlen -= sizeof(struct in_addr); 1461 cp[IPOPT_OLEN] = optlen; 1462 /* 1463 * Move first hop before start of options. 1464 */ 1465 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1466 sizeof(struct in_addr)); 1467 /* 1468 * Then copy rest of options back 1469 * to close up the deleted entry. 1470 */ 1471 (void)memmove(&cp[IPOPT_OFFSET+1], 1472 &cp[IPOPT_OFFSET+1] + sizeof(struct in_addr), 1473 (unsigned)cnt - (IPOPT_MINOFF - 1)); 1474 break; 1475 } 1476 } 1477 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1478 goto bad; 1479 *pcbopt = m; 1480 return (0); 1481 1482 bad: 1483 (void)m_free(m); 1484 return (EINVAL); 1485 } 1486 1487 /* 1488 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1489 */ 1490 static struct ifnet * 1491 ip_multicast_if(a, ifindexp) 1492 struct in_addr *a; 1493 int *ifindexp; 1494 { 1495 int ifindex; 1496 struct ifnet *ifp = NULL; 1497 struct in_ifaddr *ia; 1498 1499 if (ifindexp) 1500 *ifindexp = 0; 1501 if (ntohl(a->s_addr) >> 24 == 0) { 1502 ifindex = ntohl(a->s_addr) & 0xffffff; 1503 if (ifindex < 0 || if_indexlim <= ifindex) 1504 return NULL; 1505 ifp = ifindex2ifnet[ifindex]; 1506 if (!ifp) 1507 return NULL; 1508 if (ifindexp) 1509 *ifindexp = ifindex; 1510 } else { 1511 LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) { 1512 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1513 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1514 ifp = ia->ia_ifp; 1515 break; 1516 } 1517 } 1518 } 1519 return ifp; 1520 } 1521 1522 /* 1523 * Set the IP multicast options in response to user setsockopt(). 1524 */ 1525 int 1526 ip_setmoptions(optname, imop, m) 1527 int optname; 1528 struct ip_moptions **imop; 1529 struct mbuf *m; 1530 { 1531 int error = 0; 1532 u_char loop; 1533 int i; 1534 struct in_addr addr; 1535 struct ip_mreq *mreq; 1536 struct ifnet *ifp; 1537 struct ip_moptions *imo = *imop; 1538 struct route ro; 1539 struct sockaddr_in *dst; 1540 int ifindex; 1541 1542 if (imo == NULL) { 1543 /* 1544 * No multicast option buffer attached to the pcb; 1545 * allocate one and initialize to default values. 1546 */ 1547 imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS, 1548 M_WAITOK); 1549 1550 if (imo == NULL) 1551 return (ENOBUFS); 1552 *imop = imo; 1553 imo->imo_multicast_ifp = NULL; 1554 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1555 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1556 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1557 imo->imo_num_memberships = 0; 1558 } 1559 1560 switch (optname) { 1561 1562 case IP_MULTICAST_IF: 1563 /* 1564 * Select the interface for outgoing multicast packets. 1565 */ 1566 if (m == NULL || m->m_len != sizeof(struct in_addr)) { 1567 error = EINVAL; 1568 break; 1569 } 1570 addr = *(mtod(m, struct in_addr *)); 1571 /* 1572 * INADDR_ANY is used to remove a previous selection. 1573 * When no interface is selected, a default one is 1574 * chosen every time a multicast packet is sent. 1575 */ 1576 if (in_nullhost(addr)) { 1577 imo->imo_multicast_ifp = NULL; 1578 break; 1579 } 1580 /* 1581 * The selected interface is identified by its local 1582 * IP address. Find the interface and confirm that 1583 * it supports multicasting. 1584 */ 1585 ifp = ip_multicast_if(&addr, &ifindex); 1586 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1587 error = EADDRNOTAVAIL; 1588 break; 1589 } 1590 imo->imo_multicast_ifp = ifp; 1591 if (ifindex) 1592 imo->imo_multicast_addr = addr; 1593 else 1594 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1595 break; 1596 1597 case IP_MULTICAST_TTL: 1598 /* 1599 * Set the IP time-to-live for outgoing multicast packets. 1600 */ 1601 if (m == NULL || m->m_len != 1) { 1602 error = EINVAL; 1603 break; 1604 } 1605 imo->imo_multicast_ttl = *(mtod(m, u_char *)); 1606 break; 1607 1608 case IP_MULTICAST_LOOP: 1609 /* 1610 * Set the loopback flag for outgoing multicast packets. 1611 * Must be zero or one. 1612 */ 1613 if (m == NULL || m->m_len != 1 || 1614 (loop = *(mtod(m, u_char *))) > 1) { 1615 error = EINVAL; 1616 break; 1617 } 1618 imo->imo_multicast_loop = loop; 1619 break; 1620 1621 case IP_ADD_MEMBERSHIP: 1622 /* 1623 * Add a multicast group membership. 1624 * Group must be a valid IP multicast address. 1625 */ 1626 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1627 error = EINVAL; 1628 break; 1629 } 1630 mreq = mtod(m, struct ip_mreq *); 1631 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1632 error = EINVAL; 1633 break; 1634 } 1635 /* 1636 * If no interface address was provided, use the interface of 1637 * the route to the given multicast address. 1638 */ 1639 if (in_nullhost(mreq->imr_interface)) { 1640 bzero((caddr_t)&ro, sizeof(ro)); 1641 ro.ro_rt = NULL; 1642 dst = satosin(&ro.ro_dst); 1643 dst->sin_len = sizeof(*dst); 1644 dst->sin_family = AF_INET; 1645 dst->sin_addr = mreq->imr_multiaddr; 1646 rtalloc(&ro); 1647 if (ro.ro_rt == NULL) { 1648 error = EADDRNOTAVAIL; 1649 break; 1650 } 1651 ifp = ro.ro_rt->rt_ifp; 1652 rtfree(ro.ro_rt); 1653 } else { 1654 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1655 } 1656 /* 1657 * See if we found an interface, and confirm that it 1658 * supports multicast. 1659 */ 1660 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1661 error = EADDRNOTAVAIL; 1662 break; 1663 } 1664 /* 1665 * See if the membership already exists or if all the 1666 * membership slots are full. 1667 */ 1668 for (i = 0; i < imo->imo_num_memberships; ++i) { 1669 if (imo->imo_membership[i]->inm_ifp == ifp && 1670 in_hosteq(imo->imo_membership[i]->inm_addr, 1671 mreq->imr_multiaddr)) 1672 break; 1673 } 1674 if (i < imo->imo_num_memberships) { 1675 error = EADDRINUSE; 1676 break; 1677 } 1678 if (i == IP_MAX_MEMBERSHIPS) { 1679 error = ETOOMANYREFS; 1680 break; 1681 } 1682 /* 1683 * Everything looks good; add a new record to the multicast 1684 * address list for the given interface. 1685 */ 1686 if ((imo->imo_membership[i] = 1687 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) { 1688 error = ENOBUFS; 1689 break; 1690 } 1691 ++imo->imo_num_memberships; 1692 break; 1693 1694 case IP_DROP_MEMBERSHIP: 1695 /* 1696 * Drop a multicast group membership. 1697 * Group must be a valid IP multicast address. 1698 */ 1699 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1700 error = EINVAL; 1701 break; 1702 } 1703 mreq = mtod(m, struct ip_mreq *); 1704 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1705 error = EINVAL; 1706 break; 1707 } 1708 /* 1709 * If an interface address was specified, get a pointer 1710 * to its ifnet structure. 1711 */ 1712 if (in_nullhost(mreq->imr_interface)) 1713 ifp = NULL; 1714 else { 1715 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1716 if (ifp == NULL) { 1717 error = EADDRNOTAVAIL; 1718 break; 1719 } 1720 } 1721 /* 1722 * Find the membership in the membership array. 1723 */ 1724 for (i = 0; i < imo->imo_num_memberships; ++i) { 1725 if ((ifp == NULL || 1726 imo->imo_membership[i]->inm_ifp == ifp) && 1727 in_hosteq(imo->imo_membership[i]->inm_addr, 1728 mreq->imr_multiaddr)) 1729 break; 1730 } 1731 if (i == imo->imo_num_memberships) { 1732 error = EADDRNOTAVAIL; 1733 break; 1734 } 1735 /* 1736 * Give up the multicast address record to which the 1737 * membership points. 1738 */ 1739 in_delmulti(imo->imo_membership[i]); 1740 /* 1741 * Remove the gap in the membership array. 1742 */ 1743 for (++i; i < imo->imo_num_memberships; ++i) 1744 imo->imo_membership[i-1] = imo->imo_membership[i]; 1745 --imo->imo_num_memberships; 1746 break; 1747 1748 default: 1749 error = EOPNOTSUPP; 1750 break; 1751 } 1752 1753 /* 1754 * If all options have default values, no need to keep the mbuf. 1755 */ 1756 if (imo->imo_multicast_ifp == NULL && 1757 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1758 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1759 imo->imo_num_memberships == 0) { 1760 free(*imop, M_IPMOPTS); 1761 *imop = NULL; 1762 } 1763 1764 return (error); 1765 } 1766 1767 /* 1768 * Return the IP multicast options in response to user getsockopt(). 1769 */ 1770 int 1771 ip_getmoptions(optname, imo, mp) 1772 int optname; 1773 struct ip_moptions *imo; 1774 struct mbuf **mp; 1775 { 1776 u_char *ttl; 1777 u_char *loop; 1778 struct in_addr *addr; 1779 struct in_ifaddr *ia; 1780 1781 *mp = m_get(M_WAIT, MT_SOOPTS); 1782 1783 switch (optname) { 1784 1785 case IP_MULTICAST_IF: 1786 addr = mtod(*mp, struct in_addr *); 1787 (*mp)->m_len = sizeof(struct in_addr); 1788 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1789 *addr = zeroin_addr; 1790 else if (imo->imo_multicast_addr.s_addr) { 1791 /* return the value user has set */ 1792 *addr = imo->imo_multicast_addr; 1793 } else { 1794 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1795 *addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1796 } 1797 return (0); 1798 1799 case IP_MULTICAST_TTL: 1800 ttl = mtod(*mp, u_char *); 1801 (*mp)->m_len = 1; 1802 *ttl = imo ? imo->imo_multicast_ttl 1803 : IP_DEFAULT_MULTICAST_TTL; 1804 return (0); 1805 1806 case IP_MULTICAST_LOOP: 1807 loop = mtod(*mp, u_char *); 1808 (*mp)->m_len = 1; 1809 *loop = imo ? imo->imo_multicast_loop 1810 : IP_DEFAULT_MULTICAST_LOOP; 1811 return (0); 1812 1813 default: 1814 return (EOPNOTSUPP); 1815 } 1816 } 1817 1818 /* 1819 * Discard the IP multicast options. 1820 */ 1821 void 1822 ip_freemoptions(imo) 1823 struct ip_moptions *imo; 1824 { 1825 int i; 1826 1827 if (imo != NULL) { 1828 for (i = 0; i < imo->imo_num_memberships; ++i) 1829 in_delmulti(imo->imo_membership[i]); 1830 free(imo, M_IPMOPTS); 1831 } 1832 } 1833 1834 /* 1835 * Routine called from ip_output() to loop back a copy of an IP multicast 1836 * packet to the input queue of a specified interface. Note that this 1837 * calls the output routine of the loopback "driver", but with an interface 1838 * pointer that might NOT be &loif -- easier than replicating that code here. 1839 */ 1840 static void 1841 ip_mloopback(ifp, m, dst) 1842 struct ifnet *ifp; 1843 struct mbuf *m; 1844 struct sockaddr_in *dst; 1845 { 1846 struct ip *ip; 1847 struct mbuf *copym; 1848 1849 copym = m_copy(m, 0, M_COPYALL); 1850 if (copym != NULL 1851 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1852 copym = m_pullup(copym, sizeof(struct ip)); 1853 if (copym != NULL) { 1854 /* 1855 * We don't bother to fragment if the IP length is greater 1856 * than the interface's MTU. Can this possibly matter? 1857 */ 1858 ip = mtod(copym, struct ip *); 1859 1860 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1861 in_delayed_cksum(copym); 1862 copym->m_pkthdr.csum_flags &= 1863 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1864 } 1865 1866 ip->ip_sum = 0; 1867 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1868 (void) looutput(ifp, copym, sintosa(dst), NULL); 1869 } 1870 } 1871