1 /* $NetBSD: ip_output.c,v 1.144 2005/02/26 22:45:12 perry 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 1.144 2005/02/26 22:45:12 perry 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 #ifdef IPSEC_NAT_T 143 #include <netinet/udp.h> 144 #endif 145 #endif /*IPSEC*/ 146 147 #ifdef FAST_IPSEC 148 #include <netipsec/ipsec.h> 149 #include <netipsec/key.h> 150 #include <netipsec/xform.h> 151 #endif /* FAST_IPSEC*/ 152 153 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 154 static struct ifnet *ip_multicast_if(struct in_addr *, int *); 155 static void ip_mloopback(struct ifnet *, struct mbuf *, struct sockaddr_in *); 156 157 #ifdef PFIL_HOOKS 158 extern struct pfil_head inet_pfil_hook; /* XXX */ 159 #endif 160 161 /* 162 * IP output. The packet in mbuf chain m contains a skeletal IP 163 * header (with len, off, ttl, proto, tos, src, dst). 164 * The mbuf chain containing the packet will be freed. 165 * The mbuf opt, if present, will not be freed. 166 */ 167 int 168 ip_output(struct mbuf *m0, ...) 169 { 170 struct ip *ip; 171 struct ifnet *ifp; 172 struct mbuf *m = m0; 173 int hlen = sizeof (struct ip); 174 int len, error = 0; 175 struct route iproute; 176 struct sockaddr_in *dst; 177 struct in_ifaddr *ia; 178 struct mbuf *opt; 179 struct route *ro; 180 int flags, sw_csum; 181 int *mtu_p; 182 u_long mtu; 183 struct ip_moptions *imo; 184 struct socket *so; 185 va_list ap; 186 #ifdef IPSEC 187 struct secpolicy *sp = NULL; 188 #ifdef IPSEC_NAT_T 189 int natt_frag = 0; 190 #endif 191 #endif /*IPSEC*/ 192 #ifdef FAST_IPSEC 193 struct inpcb *inp; 194 struct m_tag *mtag; 195 struct secpolicy *sp = NULL; 196 struct tdb_ident *tdbi; 197 int s; 198 #endif 199 u_int16_t ip_len; 200 201 len = 0; 202 va_start(ap, m0); 203 opt = va_arg(ap, struct mbuf *); 204 ro = va_arg(ap, struct route *); 205 flags = va_arg(ap, int); 206 imo = va_arg(ap, struct ip_moptions *); 207 so = va_arg(ap, struct socket *); 208 if (flags & IP_RETURNMTU) 209 mtu_p = va_arg(ap, int *); 210 else 211 mtu_p = NULL; 212 va_end(ap); 213 214 MCLAIM(m, &ip_tx_mowner); 215 #ifdef FAST_IPSEC 216 if (so != NULL && so->so_proto->pr_domain->dom_family == AF_INET) 217 inp = (struct inpcb *)so->so_pcb; 218 else 219 inp = NULL; 220 #endif /* FAST_IPSEC */ 221 222 #ifdef DIAGNOSTIC 223 if ((m->m_flags & M_PKTHDR) == 0) 224 panic("ip_output no HDR"); 225 #endif 226 if (opt) { 227 m = ip_insertoptions(m, opt, &len); 228 if (len >= sizeof(struct ip)) 229 hlen = len; 230 } 231 ip = mtod(m, struct ip *); 232 /* 233 * Fill in IP header. 234 */ 235 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 236 ip->ip_v = IPVERSION; 237 ip->ip_off = htons(0); 238 ip->ip_id = ip_newid(); 239 ip->ip_hl = hlen >> 2; 240 ipstat.ips_localout++; 241 } else { 242 hlen = ip->ip_hl << 2; 243 } 244 /* 245 * Route packet. 246 */ 247 if (ro == 0) { 248 ro = &iproute; 249 bzero((caddr_t)ro, sizeof (*ro)); 250 } 251 dst = satosin(&ro->ro_dst); 252 /* 253 * If there is a cached route, 254 * check that it is to the same destination 255 * and is still up. If not, free it and try again. 256 * The address family should also be checked in case of sharing the 257 * cache with IPv6. 258 */ 259 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 260 dst->sin_family != AF_INET || 261 !in_hosteq(dst->sin_addr, ip->ip_dst))) { 262 RTFREE(ro->ro_rt); 263 ro->ro_rt = (struct rtentry *)0; 264 } 265 if (ro->ro_rt == 0) { 266 bzero(dst, sizeof(*dst)); 267 dst->sin_family = AF_INET; 268 dst->sin_len = sizeof(*dst); 269 dst->sin_addr = ip->ip_dst; 270 } 271 /* 272 * If routing to interface only, 273 * short circuit routing lookup. 274 */ 275 if (flags & IP_ROUTETOIF) { 276 if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) { 277 ipstat.ips_noroute++; 278 error = ENETUNREACH; 279 goto bad; 280 } 281 ifp = ia->ia_ifp; 282 mtu = ifp->if_mtu; 283 ip->ip_ttl = 1; 284 } else if ((IN_MULTICAST(ip->ip_dst.s_addr) || 285 ip->ip_dst.s_addr == INADDR_BROADCAST) && 286 imo != NULL && imo->imo_multicast_ifp != NULL) { 287 ifp = imo->imo_multicast_ifp; 288 mtu = ifp->if_mtu; 289 IFP_TO_IA(ifp, ia); 290 } else { 291 if (ro->ro_rt == 0) 292 rtalloc(ro); 293 if (ro->ro_rt == 0) { 294 ipstat.ips_noroute++; 295 error = EHOSTUNREACH; 296 goto bad; 297 } 298 ia = ifatoia(ro->ro_rt->rt_ifa); 299 ifp = ro->ro_rt->rt_ifp; 300 if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0) 301 mtu = ifp->if_mtu; 302 ro->ro_rt->rt_use++; 303 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 304 dst = satosin(ro->ro_rt->rt_gateway); 305 } 306 if (IN_MULTICAST(ip->ip_dst.s_addr) || 307 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 308 struct in_multi *inm; 309 310 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 311 M_BCAST : M_MCAST; 312 /* 313 * IP destination address is multicast. Make sure "dst" 314 * still points to the address in "ro". (It may have been 315 * changed to point to a gateway address, above.) 316 */ 317 dst = satosin(&ro->ro_dst); 318 /* 319 * See if the caller provided any multicast options 320 */ 321 if (imo != NULL) 322 ip->ip_ttl = imo->imo_multicast_ttl; 323 else 324 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 325 326 /* 327 * if we don't know the outgoing ifp yet, we can't generate 328 * output 329 */ 330 if (!ifp) { 331 ipstat.ips_noroute++; 332 error = ENETUNREACH; 333 goto bad; 334 } 335 336 /* 337 * If the packet is multicast or broadcast, confirm that 338 * the outgoing interface can transmit it. 339 */ 340 if (((m->m_flags & M_MCAST) && 341 (ifp->if_flags & IFF_MULTICAST) == 0) || 342 ((m->m_flags & M_BCAST) && 343 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 344 ipstat.ips_noroute++; 345 error = ENETUNREACH; 346 goto bad; 347 } 348 /* 349 * If source address not specified yet, use an address 350 * of outgoing interface. 351 */ 352 if (in_nullhost(ip->ip_src)) { 353 struct in_ifaddr *ia; 354 355 IFP_TO_IA(ifp, ia); 356 if (!ia) { 357 error = EADDRNOTAVAIL; 358 goto bad; 359 } 360 ip->ip_src = ia->ia_addr.sin_addr; 361 } 362 363 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 364 if (inm != NULL && 365 (imo == NULL || imo->imo_multicast_loop)) { 366 /* 367 * If we belong to the destination multicast group 368 * on the outgoing interface, and the caller did not 369 * forbid loopback, loop back a copy. 370 */ 371 ip_mloopback(ifp, m, dst); 372 } 373 #ifdef MROUTING 374 else { 375 /* 376 * If we are acting as a multicast router, perform 377 * multicast forwarding as if the packet had just 378 * arrived on the interface to which we are about 379 * to send. The multicast forwarding function 380 * recursively calls this function, using the 381 * IP_FORWARDING flag to prevent infinite recursion. 382 * 383 * Multicasts that are looped back by ip_mloopback(), 384 * above, will be forwarded by the ip_input() routine, 385 * if necessary. 386 */ 387 extern struct socket *ip_mrouter; 388 389 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 390 if (ip_mforward(m, ifp) != 0) { 391 m_freem(m); 392 goto done; 393 } 394 } 395 } 396 #endif 397 /* 398 * Multicasts with a time-to-live of zero may be looped- 399 * back, above, but must not be transmitted on a network. 400 * Also, multicasts addressed to the loopback interface 401 * are not sent -- the above call to ip_mloopback() will 402 * loop back a copy if this host actually belongs to the 403 * destination group on the loopback interface. 404 */ 405 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 406 m_freem(m); 407 goto done; 408 } 409 410 goto sendit; 411 } 412 #ifndef notdef 413 /* 414 * If source address not specified yet, use address 415 * of outgoing interface. 416 */ 417 if (in_nullhost(ip->ip_src)) 418 ip->ip_src = ia->ia_addr.sin_addr; 419 #endif 420 421 /* 422 * packets with Class-D address as source are not valid per 423 * RFC 1112 424 */ 425 if (IN_MULTICAST(ip->ip_src.s_addr)) { 426 ipstat.ips_odropped++; 427 error = EADDRNOTAVAIL; 428 goto bad; 429 } 430 431 /* 432 * Look for broadcast address and 433 * and verify user is allowed to send 434 * such a packet. 435 */ 436 if (in_broadcast(dst->sin_addr, ifp)) { 437 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 438 error = EADDRNOTAVAIL; 439 goto bad; 440 } 441 if ((flags & IP_ALLOWBROADCAST) == 0) { 442 error = EACCES; 443 goto bad; 444 } 445 /* don't allow broadcast messages to be fragmented */ 446 if (ntohs(ip->ip_len) > ifp->if_mtu) { 447 error = EMSGSIZE; 448 goto bad; 449 } 450 m->m_flags |= M_BCAST; 451 } else 452 m->m_flags &= ~M_BCAST; 453 454 sendit: 455 /* 456 * If we're doing Path MTU Discovery, we need to set DF unless 457 * the route's MTU is locked. 458 */ 459 if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL && 460 (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0) 461 ip->ip_off |= htons(IP_DF); 462 463 /* Remember the current ip_len */ 464 ip_len = ntohs(ip->ip_len); 465 466 #ifdef IPSEC 467 /* get SP for this packet */ 468 if (so == NULL) 469 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 470 flags, &error); 471 else { 472 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 473 IPSEC_DIR_OUTBOUND)) 474 goto skip_ipsec; 475 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 476 } 477 478 if (sp == NULL) { 479 ipsecstat.out_inval++; 480 goto bad; 481 } 482 483 error = 0; 484 485 /* check policy */ 486 switch (sp->policy) { 487 case IPSEC_POLICY_DISCARD: 488 /* 489 * This packet is just discarded. 490 */ 491 ipsecstat.out_polvio++; 492 goto bad; 493 494 case IPSEC_POLICY_BYPASS: 495 case IPSEC_POLICY_NONE: 496 /* no need to do IPsec. */ 497 goto skip_ipsec; 498 499 case IPSEC_POLICY_IPSEC: 500 if (sp->req == NULL) { 501 /* XXX should be panic ? */ 502 printf("ip_output: No IPsec request specified.\n"); 503 error = EINVAL; 504 goto bad; 505 } 506 break; 507 508 case IPSEC_POLICY_ENTRUST: 509 default: 510 printf("ip_output: Invalid policy found. %d\n", sp->policy); 511 } 512 513 #ifdef IPSEC_NAT_T 514 /* 515 * NAT-T ESP fragmentation: don't do IPSec processing now, 516 * we'll do it on each fragmented packet. 517 */ 518 if (sp->req->sav && 519 ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) || 520 (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) { 521 if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) { 522 natt_frag = 1; 523 mtu = sp->req->sav->esp_frag; 524 goto skip_ipsec; 525 } 526 } 527 #endif /* IPSEC_NAT_T */ 528 529 /* 530 * ipsec4_output() expects ip_len and ip_off in network 531 * order. They have been set to network order above. 532 */ 533 534 { 535 struct ipsec_output_state state; 536 bzero(&state, sizeof(state)); 537 state.m = m; 538 if (flags & IP_ROUTETOIF) { 539 state.ro = &iproute; 540 bzero(&iproute, sizeof(iproute)); 541 } else 542 state.ro = ro; 543 state.dst = (struct sockaddr *)dst; 544 545 /* 546 * We can't defer the checksum of payload data if 547 * we're about to encrypt/authenticate it. 548 * 549 * XXX When we support crypto offloading functions of 550 * XXX network interfaces, we need to reconsider this, 551 * XXX since it's likely that they'll support checksumming, 552 * XXX as well. 553 */ 554 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 555 in_delayed_cksum(m); 556 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 557 } 558 559 error = ipsec4_output(&state, sp, flags); 560 561 m = state.m; 562 if (flags & IP_ROUTETOIF) { 563 /* 564 * if we have tunnel mode SA, we may need to ignore 565 * IP_ROUTETOIF. 566 */ 567 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 568 flags &= ~IP_ROUTETOIF; 569 ro = state.ro; 570 } 571 } else 572 ro = state.ro; 573 dst = (struct sockaddr_in *)state.dst; 574 if (error) { 575 /* mbuf is already reclaimed in ipsec4_output. */ 576 m0 = NULL; 577 switch (error) { 578 case EHOSTUNREACH: 579 case ENETUNREACH: 580 case EMSGSIZE: 581 case ENOBUFS: 582 case ENOMEM: 583 break; 584 default: 585 printf("ip4_output (ipsec): error code %d\n", error); 586 /*fall through*/ 587 case ENOENT: 588 /* don't show these error codes to the user */ 589 error = 0; 590 break; 591 } 592 goto bad; 593 } 594 595 /* be sure to update variables that are affected by ipsec4_output() */ 596 ip = mtod(m, struct ip *); 597 hlen = ip->ip_hl << 2; 598 ip_len = ntohs(ip->ip_len); 599 600 if (ro->ro_rt == NULL) { 601 if ((flags & IP_ROUTETOIF) == 0) { 602 printf("ip_output: " 603 "can't update route after IPsec processing\n"); 604 error = EHOSTUNREACH; /*XXX*/ 605 goto bad; 606 } 607 } else { 608 /* nobody uses ia beyond here */ 609 if (state.encap) { 610 ifp = ro->ro_rt->rt_ifp; 611 if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0) 612 mtu = ifp->if_mtu; 613 } 614 } 615 } 616 skip_ipsec: 617 #endif /*IPSEC*/ 618 #ifdef FAST_IPSEC 619 /* 620 * Check the security policy (SP) for the packet and, if 621 * required, do IPsec-related processing. There are two 622 * cases here; the first time a packet is sent through 623 * it will be untagged and handled by ipsec4_checkpolicy. 624 * If the packet is resubmitted to ip_output (e.g. after 625 * AH, ESP, etc. processing), there will be a tag to bypass 626 * the lookup and related policy checking. 627 */ 628 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 629 s = splsoftnet(); 630 if (mtag != NULL) { 631 tdbi = (struct tdb_ident *)(mtag + 1); 632 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 633 if (sp == NULL) 634 error = -EINVAL; /* force silent drop */ 635 m_tag_delete(m, mtag); 636 } else { 637 if (inp != NULL && 638 IPSEC_PCB_SKIP_IPSEC(inp->inp_sp, IPSEC_DIR_OUTBOUND)) 639 goto spd_done; 640 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 641 &error, inp); 642 } 643 /* 644 * There are four return cases: 645 * sp != NULL apply IPsec policy 646 * sp == NULL, error == 0 no IPsec handling needed 647 * sp == NULL, error == -EINVAL discard packet w/o error 648 * sp == NULL, error != 0 discard packet, report error 649 */ 650 if (sp != NULL) { 651 /* Loop detection, check if ipsec processing already done */ 652 IPSEC_ASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 653 for (mtag = m_tag_first(m); mtag != NULL; 654 mtag = m_tag_next(m, mtag)) { 655 #ifdef MTAG_ABI_COMPAT 656 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 657 continue; 658 #endif 659 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 660 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 661 continue; 662 /* 663 * Check if policy has an SA associated with it. 664 * This can happen when an SP has yet to acquire 665 * an SA; e.g. on first reference. If it occurs, 666 * then we let ipsec4_process_packet do its thing. 667 */ 668 if (sp->req->sav == NULL) 669 break; 670 tdbi = (struct tdb_ident *)(mtag + 1); 671 if (tdbi->spi == sp->req->sav->spi && 672 tdbi->proto == sp->req->sav->sah->saidx.proto && 673 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 674 sizeof (union sockaddr_union)) == 0) { 675 /* 676 * No IPsec processing is needed, free 677 * reference to SP. 678 * 679 * NB: null pointer to avoid free at 680 * done: below. 681 */ 682 KEY_FREESP(&sp), sp = NULL; 683 splx(s); 684 goto spd_done; 685 } 686 } 687 688 /* 689 * Do delayed checksums now because we send before 690 * this is done in the normal processing path. 691 */ 692 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 693 in_delayed_cksum(m); 694 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 695 } 696 697 #ifdef __FreeBSD__ 698 ip->ip_len = htons(ip->ip_len); 699 ip->ip_off = htons(ip->ip_off); 700 #endif 701 702 /* NB: callee frees mbuf */ 703 error = ipsec4_process_packet(m, sp->req, flags, 0); 704 /* 705 * Preserve KAME behaviour: ENOENT can be returned 706 * when an SA acquire is in progress. Don't propagate 707 * this to user-level; it confuses applications. 708 * 709 * XXX this will go away when the SADB is redone. 710 */ 711 if (error == ENOENT) 712 error = 0; 713 splx(s); 714 goto done; 715 } else { 716 splx(s); 717 718 if (error != 0) { 719 /* 720 * Hack: -EINVAL is used to signal that a packet 721 * should be silently discarded. This is typically 722 * because we asked key management for an SA and 723 * it was delayed (e.g. kicked up to IKE). 724 */ 725 if (error == -EINVAL) 726 error = 0; 727 goto bad; 728 } else { 729 /* No IPsec processing for this packet. */ 730 } 731 #ifdef notyet 732 /* 733 * If deferred crypto processing is needed, check that 734 * the interface supports it. 735 */ 736 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 737 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 738 /* notify IPsec to do its own crypto */ 739 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 740 error = EHOSTUNREACH; 741 goto bad; 742 } 743 #endif 744 } 745 spd_done: 746 #endif /* FAST_IPSEC */ 747 748 #ifdef PFIL_HOOKS 749 /* 750 * Run through list of hooks for output packets. 751 */ 752 if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 753 goto done; 754 if (m == NULL) 755 goto done; 756 757 ip = mtod(m, struct ip *); 758 hlen = ip->ip_hl << 2; 759 #endif /* PFIL_HOOKS */ 760 761 #if IFA_STATS 762 /* 763 * search for the source address structure to 764 * maintain output statistics. 765 */ 766 INADDR_TO_IA(ip->ip_src, ia); 767 #endif 768 769 /* Maybe skip checksums on loopback interfaces. */ 770 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 771 ip_do_loopback_cksum)) 772 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 773 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 774 /* 775 * If small enough for mtu of path, can just send directly. 776 */ 777 if (ip_len <= mtu) { 778 #if IFA_STATS 779 if (ia) 780 ia->ia_ifa.ifa_data.ifad_outbytes += ip_len; 781 #endif 782 /* 783 * Always initialize the sum to 0! Some HW assisted 784 * checksumming requires this. 785 */ 786 ip->ip_sum = 0; 787 788 /* 789 * Perform any checksums that the hardware can't do 790 * for us. 791 * 792 * XXX Does any hardware require the {th,uh}_sum 793 * XXX fields to be 0? 794 */ 795 if (sw_csum & M_CSUM_IPv4) { 796 ip->ip_sum = in_cksum(m, hlen); 797 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 798 } 799 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 800 in_delayed_cksum(m); 801 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 802 } else 803 m->m_pkthdr.csum_data |= hlen << 16; 804 805 #ifdef IPSEC 806 /* clean ipsec history once it goes out of the node */ 807 ipsec_delaux(m); 808 #endif 809 error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt); 810 goto done; 811 } 812 813 /* 814 * We can't use HW checksumming if we're about to 815 * to fragment the packet. 816 * 817 * XXX Some hardware can do this. 818 */ 819 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 820 in_delayed_cksum(m); 821 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 822 } 823 824 /* 825 * Too large for interface; fragment if possible. 826 * Must be able to put at least 8 bytes per fragment. 827 */ 828 if (ntohs(ip->ip_off) & IP_DF) { 829 if (flags & IP_RETURNMTU) 830 *mtu_p = mtu; 831 error = EMSGSIZE; 832 ipstat.ips_cantfrag++; 833 goto bad; 834 } 835 836 error = ip_fragment(m, ifp, mtu); 837 if (error) { 838 m = NULL; 839 goto bad; 840 } 841 842 for (; m; m = m0) { 843 m0 = m->m_nextpkt; 844 m->m_nextpkt = 0; 845 if (error == 0) { 846 #if IFA_STATS 847 if (ia) 848 ia->ia_ifa.ifa_data.ifad_outbytes += 849 ntohs(ip->ip_len); 850 #endif 851 #ifdef IPSEC 852 /* clean ipsec history once it goes out of the node */ 853 ipsec_delaux(m); 854 855 #ifdef IPSEC_NAT_T 856 /* 857 * If we get there, the packet has not been handeld by 858 * IPSec whereas it should have. Now that it has been 859 * fragmented, re-inject it in ip_output so that IPsec 860 * processing can occur. 861 */ 862 if (natt_frag) { 863 error = ip_output(m, opt, 864 ro, flags, imo, so, mtu_p); 865 } else 866 #endif /* IPSEC_NAT_T */ 867 #endif /* IPSEC */ 868 { 869 KASSERT((m->m_pkthdr.csum_flags & 870 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 871 error = (*ifp->if_output)(ifp, m, sintosa(dst), 872 ro->ro_rt); 873 } 874 } else 875 m_freem(m); 876 } 877 878 if (error == 0) 879 ipstat.ips_fragmented++; 880 done: 881 if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) { 882 RTFREE(ro->ro_rt); 883 ro->ro_rt = 0; 884 } 885 886 #ifdef IPSEC 887 if (sp != NULL) { 888 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 889 printf("DP ip_output call free SP:%p\n", sp)); 890 key_freesp(sp); 891 } 892 #endif /* IPSEC */ 893 #ifdef FAST_IPSEC 894 if (sp != NULL) 895 KEY_FREESP(&sp); 896 #endif /* FAST_IPSEC */ 897 898 return (error); 899 bad: 900 m_freem(m); 901 goto done; 902 } 903 904 int 905 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 906 { 907 struct ip *ip, *mhip; 908 struct mbuf *m0; 909 int len, hlen, off; 910 int mhlen, firstlen; 911 struct mbuf **mnext; 912 int sw_csum = m->m_pkthdr.csum_flags; 913 int fragments = 0; 914 int s; 915 int error = 0; 916 917 ip = mtod(m, struct ip *); 918 hlen = ip->ip_hl << 2; 919 if (ifp != NULL) 920 sw_csum &= ~ifp->if_csum_flags_tx; 921 922 len = (mtu - hlen) &~ 7; 923 if (len < 8) { 924 m_freem(m); 925 return (EMSGSIZE); 926 } 927 928 firstlen = len; 929 mnext = &m->m_nextpkt; 930 931 /* 932 * Loop through length of segment after first fragment, 933 * make new header and copy data of each part and link onto chain. 934 */ 935 m0 = m; 936 mhlen = sizeof (struct ip); 937 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 938 MGETHDR(m, M_DONTWAIT, MT_HEADER); 939 if (m == 0) { 940 error = ENOBUFS; 941 ipstat.ips_odropped++; 942 goto sendorfree; 943 } 944 MCLAIM(m, m0->m_owner); 945 *mnext = m; 946 mnext = &m->m_nextpkt; 947 m->m_data += max_linkhdr; 948 mhip = mtod(m, struct ip *); 949 *mhip = *ip; 950 /* we must inherit MCAST and BCAST flags */ 951 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 952 if (hlen > sizeof (struct ip)) { 953 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 954 mhip->ip_hl = mhlen >> 2; 955 } 956 m->m_len = mhlen; 957 mhip->ip_off = ((off - hlen) >> 3) + 958 (ntohs(ip->ip_off) & ~IP_MF); 959 if (ip->ip_off & htons(IP_MF)) 960 mhip->ip_off |= IP_MF; 961 if (off + len >= ntohs(ip->ip_len)) 962 len = ntohs(ip->ip_len) - off; 963 else 964 mhip->ip_off |= IP_MF; 965 HTONS(mhip->ip_off); 966 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 967 m->m_next = m_copy(m0, off, len); 968 if (m->m_next == 0) { 969 error = ENOBUFS; /* ??? */ 970 ipstat.ips_odropped++; 971 goto sendorfree; 972 } 973 m->m_pkthdr.len = mhlen + len; 974 m->m_pkthdr.rcvif = (struct ifnet *)0; 975 mhip->ip_sum = 0; 976 if (sw_csum & M_CSUM_IPv4) { 977 mhip->ip_sum = in_cksum(m, mhlen); 978 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 979 } else { 980 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 981 } 982 ipstat.ips_ofragments++; 983 fragments++; 984 } 985 /* 986 * Update first fragment by trimming what's been copied out 987 * and updating header, then send each fragment (in order). 988 */ 989 m = m0; 990 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 991 m->m_pkthdr.len = hlen + firstlen; 992 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 993 ip->ip_off |= htons(IP_MF); 994 ip->ip_sum = 0; 995 if (sw_csum & M_CSUM_IPv4) { 996 ip->ip_sum = in_cksum(m, hlen); 997 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 998 } else { 999 KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4); 1000 } 1001 sendorfree: 1002 /* 1003 * If there is no room for all the fragments, don't queue 1004 * any of them. 1005 */ 1006 if (ifp != NULL) { 1007 s = splnet(); 1008 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 1009 error == 0) { 1010 error = ENOBUFS; 1011 ipstat.ips_odropped++; 1012 IFQ_INC_DROPS(&ifp->if_snd); 1013 } 1014 splx(s); 1015 } 1016 if (error) { 1017 for (m = m0; m; m = m0) { 1018 m0 = m->m_nextpkt; 1019 m->m_nextpkt = NULL; 1020 m_freem(m); 1021 } 1022 } 1023 return (error); 1024 } 1025 1026 /* 1027 * Process a delayed payload checksum calculation. 1028 */ 1029 void 1030 in_delayed_cksum(struct mbuf *m) 1031 { 1032 struct ip *ip; 1033 u_int16_t csum, offset; 1034 1035 ip = mtod(m, struct ip *); 1036 offset = ip->ip_hl << 2; 1037 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 1038 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 1039 csum = 0xffff; 1040 1041 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1042 1043 if ((offset + sizeof(u_int16_t)) > m->m_len) { 1044 /* This happen when ip options were inserted 1045 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 1046 m->m_len, offset, ip->ip_p); 1047 */ 1048 m_copyback(m, offset, sizeof(csum), (caddr_t) &csum); 1049 } else 1050 *(u_int16_t *)(mtod(m, caddr_t) + offset) = csum; 1051 } 1052 1053 /* 1054 * Determine the maximum length of the options to be inserted; 1055 * we would far rather allocate too much space rather than too little. 1056 */ 1057 1058 u_int 1059 ip_optlen(struct inpcb *inp) 1060 { 1061 struct mbuf *m = inp->inp_options; 1062 1063 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) 1064 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 1065 else 1066 return 0; 1067 } 1068 1069 1070 /* 1071 * Insert IP options into preformed packet. 1072 * Adjust IP destination as required for IP source routing, 1073 * as indicated by a non-zero in_addr at the start of the options. 1074 */ 1075 static struct mbuf * 1076 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 1077 { 1078 struct ipoption *p = mtod(opt, struct ipoption *); 1079 struct mbuf *n; 1080 struct ip *ip = mtod(m, struct ip *); 1081 unsigned optlen; 1082 1083 optlen = opt->m_len - sizeof(p->ipopt_dst); 1084 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 1085 return (m); /* XXX should fail */ 1086 if (!in_nullhost(p->ipopt_dst)) 1087 ip->ip_dst = p->ipopt_dst; 1088 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 1089 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1090 if (n == 0) 1091 return (m); 1092 MCLAIM(n, m->m_owner); 1093 M_COPY_PKTHDR(n, m); 1094 m_tag_delete_chain(m, NULL); 1095 m->m_flags &= ~M_PKTHDR; 1096 m->m_len -= sizeof(struct ip); 1097 m->m_data += sizeof(struct ip); 1098 n->m_next = m; 1099 m = n; 1100 m->m_len = optlen + sizeof(struct ip); 1101 m->m_data += max_linkhdr; 1102 bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1103 } else { 1104 m->m_data -= optlen; 1105 m->m_len += optlen; 1106 memmove(mtod(m, caddr_t), ip, sizeof(struct ip)); 1107 } 1108 m->m_pkthdr.len += optlen; 1109 ip = mtod(m, struct ip *); 1110 bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen); 1111 *phlen = sizeof(struct ip) + optlen; 1112 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 1113 return (m); 1114 } 1115 1116 /* 1117 * Copy options from ip to jp, 1118 * omitting those not copied during fragmentation. 1119 */ 1120 int 1121 ip_optcopy(struct ip *ip, struct ip *jp) 1122 { 1123 u_char *cp, *dp; 1124 int opt, optlen, cnt; 1125 1126 cp = (u_char *)(ip + 1); 1127 dp = (u_char *)(jp + 1); 1128 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1129 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1130 opt = cp[0]; 1131 if (opt == IPOPT_EOL) 1132 break; 1133 if (opt == IPOPT_NOP) { 1134 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1135 *dp++ = IPOPT_NOP; 1136 optlen = 1; 1137 continue; 1138 } 1139 #ifdef DIAGNOSTIC 1140 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1141 panic("malformed IPv4 option passed to ip_optcopy"); 1142 #endif 1143 optlen = cp[IPOPT_OLEN]; 1144 #ifdef DIAGNOSTIC 1145 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1146 panic("malformed IPv4 option passed to ip_optcopy"); 1147 #endif 1148 /* bogus lengths should have been caught by ip_dooptions */ 1149 if (optlen > cnt) 1150 optlen = cnt; 1151 if (IPOPT_COPIED(opt)) { 1152 bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen); 1153 dp += optlen; 1154 } 1155 } 1156 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1157 *dp++ = IPOPT_EOL; 1158 return (optlen); 1159 } 1160 1161 /* 1162 * IP socket option processing. 1163 */ 1164 int 1165 ip_ctloutput(int op, struct socket *so, int level, int optname, 1166 struct mbuf **mp) 1167 { 1168 struct inpcb *inp = sotoinpcb(so); 1169 struct mbuf *m = *mp; 1170 int optval = 0; 1171 int error = 0; 1172 #if defined(IPSEC) || defined(FAST_IPSEC) 1173 struct proc *p = curproc; /*XXX*/ 1174 #endif 1175 1176 if (level != IPPROTO_IP) { 1177 error = EINVAL; 1178 if (op == PRCO_SETOPT && *mp) 1179 (void) m_free(*mp); 1180 } else switch (op) { 1181 1182 case PRCO_SETOPT: 1183 switch (optname) { 1184 case IP_OPTIONS: 1185 #ifdef notyet 1186 case IP_RETOPTS: 1187 return (ip_pcbopts(optname, &inp->inp_options, m)); 1188 #else 1189 return (ip_pcbopts(&inp->inp_options, m)); 1190 #endif 1191 1192 case IP_TOS: 1193 case IP_TTL: 1194 case IP_RECVOPTS: 1195 case IP_RECVRETOPTS: 1196 case IP_RECVDSTADDR: 1197 case IP_RECVIF: 1198 if (m == NULL || m->m_len != sizeof(int)) 1199 error = EINVAL; 1200 else { 1201 optval = *mtod(m, int *); 1202 switch (optname) { 1203 1204 case IP_TOS: 1205 inp->inp_ip.ip_tos = optval; 1206 break; 1207 1208 case IP_TTL: 1209 inp->inp_ip.ip_ttl = optval; 1210 break; 1211 #define OPTSET(bit) \ 1212 if (optval) \ 1213 inp->inp_flags |= bit; \ 1214 else \ 1215 inp->inp_flags &= ~bit; 1216 1217 case IP_RECVOPTS: 1218 OPTSET(INP_RECVOPTS); 1219 break; 1220 1221 case IP_RECVRETOPTS: 1222 OPTSET(INP_RECVRETOPTS); 1223 break; 1224 1225 case IP_RECVDSTADDR: 1226 OPTSET(INP_RECVDSTADDR); 1227 break; 1228 1229 case IP_RECVIF: 1230 OPTSET(INP_RECVIF); 1231 break; 1232 } 1233 } 1234 break; 1235 #undef OPTSET 1236 1237 case IP_MULTICAST_IF: 1238 case IP_MULTICAST_TTL: 1239 case IP_MULTICAST_LOOP: 1240 case IP_ADD_MEMBERSHIP: 1241 case IP_DROP_MEMBERSHIP: 1242 error = ip_setmoptions(optname, &inp->inp_moptions, m); 1243 break; 1244 1245 case IP_PORTRANGE: 1246 if (m == 0 || m->m_len != sizeof(int)) 1247 error = EINVAL; 1248 else { 1249 optval = *mtod(m, int *); 1250 1251 switch (optval) { 1252 1253 case IP_PORTRANGE_DEFAULT: 1254 case IP_PORTRANGE_HIGH: 1255 inp->inp_flags &= ~(INP_LOWPORT); 1256 break; 1257 1258 case IP_PORTRANGE_LOW: 1259 inp->inp_flags |= INP_LOWPORT; 1260 break; 1261 1262 default: 1263 error = EINVAL; 1264 break; 1265 } 1266 } 1267 break; 1268 1269 #if defined(IPSEC) || defined(FAST_IPSEC) 1270 case IP_IPSEC_POLICY: 1271 { 1272 caddr_t req = NULL; 1273 size_t len = 0; 1274 int priv = 0; 1275 1276 #ifdef __NetBSD__ 1277 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1278 priv = 0; 1279 else 1280 priv = 1; 1281 #else 1282 priv = (in6p->in6p_socket->so_state & SS_PRIV); 1283 #endif 1284 if (m) { 1285 req = mtod(m, caddr_t); 1286 len = m->m_len; 1287 } 1288 error = ipsec4_set_policy(inp, optname, req, len, priv); 1289 break; 1290 } 1291 #endif /*IPSEC*/ 1292 1293 default: 1294 error = ENOPROTOOPT; 1295 break; 1296 } 1297 if (m) 1298 (void)m_free(m); 1299 break; 1300 1301 case PRCO_GETOPT: 1302 switch (optname) { 1303 case IP_OPTIONS: 1304 case IP_RETOPTS: 1305 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1306 MCLAIM(m, so->so_mowner); 1307 if (inp->inp_options) { 1308 m->m_len = inp->inp_options->m_len; 1309 bcopy(mtod(inp->inp_options, caddr_t), 1310 mtod(m, caddr_t), (unsigned)m->m_len); 1311 } else 1312 m->m_len = 0; 1313 break; 1314 1315 case IP_TOS: 1316 case IP_TTL: 1317 case IP_RECVOPTS: 1318 case IP_RECVRETOPTS: 1319 case IP_RECVDSTADDR: 1320 case IP_RECVIF: 1321 case IP_ERRORMTU: 1322 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1323 MCLAIM(m, so->so_mowner); 1324 m->m_len = sizeof(int); 1325 switch (optname) { 1326 1327 case IP_TOS: 1328 optval = inp->inp_ip.ip_tos; 1329 break; 1330 1331 case IP_TTL: 1332 optval = inp->inp_ip.ip_ttl; 1333 break; 1334 1335 case IP_ERRORMTU: 1336 optval = inp->inp_errormtu; 1337 break; 1338 1339 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1340 1341 case IP_RECVOPTS: 1342 optval = OPTBIT(INP_RECVOPTS); 1343 break; 1344 1345 case IP_RECVRETOPTS: 1346 optval = OPTBIT(INP_RECVRETOPTS); 1347 break; 1348 1349 case IP_RECVDSTADDR: 1350 optval = OPTBIT(INP_RECVDSTADDR); 1351 break; 1352 1353 case IP_RECVIF: 1354 optval = OPTBIT(INP_RECVIF); 1355 break; 1356 } 1357 *mtod(m, int *) = optval; 1358 break; 1359 1360 #if 0 /* defined(IPSEC) || defined(FAST_IPSEC) */ 1361 /* XXX: code broken */ 1362 case IP_IPSEC_POLICY: 1363 { 1364 caddr_t req = NULL; 1365 size_t len = 0; 1366 1367 if (m) { 1368 req = mtod(m, caddr_t); 1369 len = m->m_len; 1370 } 1371 error = ipsec4_get_policy(inp, req, len, mp); 1372 break; 1373 } 1374 #endif /*IPSEC*/ 1375 1376 case IP_MULTICAST_IF: 1377 case IP_MULTICAST_TTL: 1378 case IP_MULTICAST_LOOP: 1379 case IP_ADD_MEMBERSHIP: 1380 case IP_DROP_MEMBERSHIP: 1381 error = ip_getmoptions(optname, inp->inp_moptions, mp); 1382 if (*mp) 1383 MCLAIM(*mp, so->so_mowner); 1384 break; 1385 1386 case IP_PORTRANGE: 1387 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1388 MCLAIM(m, so->so_mowner); 1389 m->m_len = sizeof(int); 1390 1391 if (inp->inp_flags & INP_LOWPORT) 1392 optval = IP_PORTRANGE_LOW; 1393 else 1394 optval = IP_PORTRANGE_DEFAULT; 1395 1396 *mtod(m, int *) = optval; 1397 break; 1398 1399 default: 1400 error = ENOPROTOOPT; 1401 break; 1402 } 1403 break; 1404 } 1405 return (error); 1406 } 1407 1408 /* 1409 * Set up IP options in pcb for insertion in output packets. 1410 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1411 * with destination address if source routed. 1412 */ 1413 int 1414 #ifdef notyet 1415 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m) 1416 #else 1417 ip_pcbopts(struct mbuf **pcbopt, struct mbuf *m) 1418 #endif 1419 { 1420 int cnt, optlen; 1421 u_char *cp; 1422 u_char opt; 1423 1424 /* turn off any old options */ 1425 if (*pcbopt) 1426 (void)m_free(*pcbopt); 1427 *pcbopt = 0; 1428 if (m == (struct mbuf *)0 || m->m_len == 0) { 1429 /* 1430 * Only turning off any previous options. 1431 */ 1432 if (m) 1433 (void)m_free(m); 1434 return (0); 1435 } 1436 1437 #ifndef __vax__ 1438 if (m->m_len % sizeof(int32_t)) 1439 goto bad; 1440 #endif 1441 /* 1442 * IP first-hop destination address will be stored before 1443 * actual options; move other options back 1444 * and clear it when none present. 1445 */ 1446 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1447 goto bad; 1448 cnt = m->m_len; 1449 m->m_len += sizeof(struct in_addr); 1450 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1451 memmove(cp, mtod(m, caddr_t), (unsigned)cnt); 1452 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1453 1454 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1455 opt = cp[IPOPT_OPTVAL]; 1456 if (opt == IPOPT_EOL) 1457 break; 1458 if (opt == IPOPT_NOP) 1459 optlen = 1; 1460 else { 1461 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1462 goto bad; 1463 optlen = cp[IPOPT_OLEN]; 1464 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1465 goto bad; 1466 } 1467 switch (opt) { 1468 1469 default: 1470 break; 1471 1472 case IPOPT_LSRR: 1473 case IPOPT_SSRR: 1474 /* 1475 * user process specifies route as: 1476 * ->A->B->C->D 1477 * D must be our final destination (but we can't 1478 * check that since we may not have connected yet). 1479 * A is first hop destination, which doesn't appear in 1480 * actual IP option, but is stored before the options. 1481 */ 1482 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1483 goto bad; 1484 m->m_len -= sizeof(struct in_addr); 1485 cnt -= sizeof(struct in_addr); 1486 optlen -= sizeof(struct in_addr); 1487 cp[IPOPT_OLEN] = optlen; 1488 /* 1489 * Move first hop before start of options. 1490 */ 1491 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1492 sizeof(struct in_addr)); 1493 /* 1494 * Then copy rest of options back 1495 * to close up the deleted entry. 1496 */ 1497 (void)memmove(&cp[IPOPT_OFFSET+1], 1498 &cp[IPOPT_OFFSET+1] + sizeof(struct in_addr), 1499 (unsigned)cnt - (IPOPT_MINOFF - 1)); 1500 break; 1501 } 1502 } 1503 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1504 goto bad; 1505 *pcbopt = m; 1506 return (0); 1507 1508 bad: 1509 (void)m_free(m); 1510 return (EINVAL); 1511 } 1512 1513 /* 1514 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1515 */ 1516 static struct ifnet * 1517 ip_multicast_if(struct in_addr *a, int *ifindexp) 1518 { 1519 int ifindex; 1520 struct ifnet *ifp = NULL; 1521 struct in_ifaddr *ia; 1522 1523 if (ifindexp) 1524 *ifindexp = 0; 1525 if (ntohl(a->s_addr) >> 24 == 0) { 1526 ifindex = ntohl(a->s_addr) & 0xffffff; 1527 if (ifindex < 0 || if_indexlim <= ifindex) 1528 return NULL; 1529 ifp = ifindex2ifnet[ifindex]; 1530 if (!ifp) 1531 return NULL; 1532 if (ifindexp) 1533 *ifindexp = ifindex; 1534 } else { 1535 LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) { 1536 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1537 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1538 ifp = ia->ia_ifp; 1539 break; 1540 } 1541 } 1542 } 1543 return ifp; 1544 } 1545 1546 /* 1547 * Set the IP multicast options in response to user setsockopt(). 1548 */ 1549 int 1550 ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m) 1551 { 1552 int error = 0; 1553 u_char loop; 1554 int i; 1555 struct in_addr addr; 1556 struct ip_mreq *mreq; 1557 struct ifnet *ifp; 1558 struct ip_moptions *imo = *imop; 1559 struct route ro; 1560 struct sockaddr_in *dst; 1561 int ifindex; 1562 1563 if (imo == NULL) { 1564 /* 1565 * No multicast option buffer attached to the pcb; 1566 * allocate one and initialize to default values. 1567 */ 1568 imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS, 1569 M_WAITOK); 1570 1571 if (imo == NULL) 1572 return (ENOBUFS); 1573 *imop = imo; 1574 imo->imo_multicast_ifp = NULL; 1575 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1576 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1577 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1578 imo->imo_num_memberships = 0; 1579 } 1580 1581 switch (optname) { 1582 1583 case IP_MULTICAST_IF: 1584 /* 1585 * Select the interface for outgoing multicast packets. 1586 */ 1587 if (m == NULL || m->m_len != sizeof(struct in_addr)) { 1588 error = EINVAL; 1589 break; 1590 } 1591 addr = *(mtod(m, struct in_addr *)); 1592 /* 1593 * INADDR_ANY is used to remove a previous selection. 1594 * When no interface is selected, a default one is 1595 * chosen every time a multicast packet is sent. 1596 */ 1597 if (in_nullhost(addr)) { 1598 imo->imo_multicast_ifp = NULL; 1599 break; 1600 } 1601 /* 1602 * The selected interface is identified by its local 1603 * IP address. Find the interface and confirm that 1604 * it supports multicasting. 1605 */ 1606 ifp = ip_multicast_if(&addr, &ifindex); 1607 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1608 error = EADDRNOTAVAIL; 1609 break; 1610 } 1611 imo->imo_multicast_ifp = ifp; 1612 if (ifindex) 1613 imo->imo_multicast_addr = addr; 1614 else 1615 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1616 break; 1617 1618 case IP_MULTICAST_TTL: 1619 /* 1620 * Set the IP time-to-live for outgoing multicast packets. 1621 */ 1622 if (m == NULL || m->m_len != 1) { 1623 error = EINVAL; 1624 break; 1625 } 1626 imo->imo_multicast_ttl = *(mtod(m, u_char *)); 1627 break; 1628 1629 case IP_MULTICAST_LOOP: 1630 /* 1631 * Set the loopback flag for outgoing multicast packets. 1632 * Must be zero or one. 1633 */ 1634 if (m == NULL || m->m_len != 1 || 1635 (loop = *(mtod(m, u_char *))) > 1) { 1636 error = EINVAL; 1637 break; 1638 } 1639 imo->imo_multicast_loop = loop; 1640 break; 1641 1642 case IP_ADD_MEMBERSHIP: 1643 /* 1644 * Add a multicast group membership. 1645 * Group must be a valid IP multicast address. 1646 */ 1647 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1648 error = EINVAL; 1649 break; 1650 } 1651 mreq = mtod(m, struct ip_mreq *); 1652 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1653 error = EINVAL; 1654 break; 1655 } 1656 /* 1657 * If no interface address was provided, use the interface of 1658 * the route to the given multicast address. 1659 */ 1660 if (in_nullhost(mreq->imr_interface)) { 1661 bzero((caddr_t)&ro, sizeof(ro)); 1662 ro.ro_rt = NULL; 1663 dst = satosin(&ro.ro_dst); 1664 dst->sin_len = sizeof(*dst); 1665 dst->sin_family = AF_INET; 1666 dst->sin_addr = mreq->imr_multiaddr; 1667 rtalloc(&ro); 1668 if (ro.ro_rt == NULL) { 1669 error = EADDRNOTAVAIL; 1670 break; 1671 } 1672 ifp = ro.ro_rt->rt_ifp; 1673 rtfree(ro.ro_rt); 1674 } else { 1675 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1676 } 1677 /* 1678 * See if we found an interface, and confirm that it 1679 * supports multicast. 1680 */ 1681 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1682 error = EADDRNOTAVAIL; 1683 break; 1684 } 1685 /* 1686 * See if the membership already exists or if all the 1687 * membership slots are full. 1688 */ 1689 for (i = 0; i < imo->imo_num_memberships; ++i) { 1690 if (imo->imo_membership[i]->inm_ifp == ifp && 1691 in_hosteq(imo->imo_membership[i]->inm_addr, 1692 mreq->imr_multiaddr)) 1693 break; 1694 } 1695 if (i < imo->imo_num_memberships) { 1696 error = EADDRINUSE; 1697 break; 1698 } 1699 if (i == IP_MAX_MEMBERSHIPS) { 1700 error = ETOOMANYREFS; 1701 break; 1702 } 1703 /* 1704 * Everything looks good; add a new record to the multicast 1705 * address list for the given interface. 1706 */ 1707 if ((imo->imo_membership[i] = 1708 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) { 1709 error = ENOBUFS; 1710 break; 1711 } 1712 ++imo->imo_num_memberships; 1713 break; 1714 1715 case IP_DROP_MEMBERSHIP: 1716 /* 1717 * Drop a multicast group membership. 1718 * Group must be a valid IP multicast address. 1719 */ 1720 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1721 error = EINVAL; 1722 break; 1723 } 1724 mreq = mtod(m, struct ip_mreq *); 1725 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1726 error = EINVAL; 1727 break; 1728 } 1729 /* 1730 * If an interface address was specified, get a pointer 1731 * to its ifnet structure. 1732 */ 1733 if (in_nullhost(mreq->imr_interface)) 1734 ifp = NULL; 1735 else { 1736 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1737 if (ifp == NULL) { 1738 error = EADDRNOTAVAIL; 1739 break; 1740 } 1741 } 1742 /* 1743 * Find the membership in the membership array. 1744 */ 1745 for (i = 0; i < imo->imo_num_memberships; ++i) { 1746 if ((ifp == NULL || 1747 imo->imo_membership[i]->inm_ifp == ifp) && 1748 in_hosteq(imo->imo_membership[i]->inm_addr, 1749 mreq->imr_multiaddr)) 1750 break; 1751 } 1752 if (i == imo->imo_num_memberships) { 1753 error = EADDRNOTAVAIL; 1754 break; 1755 } 1756 /* 1757 * Give up the multicast address record to which the 1758 * membership points. 1759 */ 1760 in_delmulti(imo->imo_membership[i]); 1761 /* 1762 * Remove the gap in the membership array. 1763 */ 1764 for (++i; i < imo->imo_num_memberships; ++i) 1765 imo->imo_membership[i-1] = imo->imo_membership[i]; 1766 --imo->imo_num_memberships; 1767 break; 1768 1769 default: 1770 error = EOPNOTSUPP; 1771 break; 1772 } 1773 1774 /* 1775 * If all options have default values, no need to keep the mbuf. 1776 */ 1777 if (imo->imo_multicast_ifp == NULL && 1778 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1779 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1780 imo->imo_num_memberships == 0) { 1781 free(*imop, M_IPMOPTS); 1782 *imop = NULL; 1783 } 1784 1785 return (error); 1786 } 1787 1788 /* 1789 * Return the IP multicast options in response to user getsockopt(). 1790 */ 1791 int 1792 ip_getmoptions(int optname, struct ip_moptions *imo, struct mbuf **mp) 1793 { 1794 u_char *ttl; 1795 u_char *loop; 1796 struct in_addr *addr; 1797 struct in_ifaddr *ia; 1798 1799 *mp = m_get(M_WAIT, MT_SOOPTS); 1800 1801 switch (optname) { 1802 1803 case IP_MULTICAST_IF: 1804 addr = mtod(*mp, struct in_addr *); 1805 (*mp)->m_len = sizeof(struct in_addr); 1806 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1807 *addr = zeroin_addr; 1808 else if (imo->imo_multicast_addr.s_addr) { 1809 /* return the value user has set */ 1810 *addr = imo->imo_multicast_addr; 1811 } else { 1812 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1813 *addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1814 } 1815 return (0); 1816 1817 case IP_MULTICAST_TTL: 1818 ttl = mtod(*mp, u_char *); 1819 (*mp)->m_len = 1; 1820 *ttl = imo ? imo->imo_multicast_ttl 1821 : IP_DEFAULT_MULTICAST_TTL; 1822 return (0); 1823 1824 case IP_MULTICAST_LOOP: 1825 loop = mtod(*mp, u_char *); 1826 (*mp)->m_len = 1; 1827 *loop = imo ? imo->imo_multicast_loop 1828 : IP_DEFAULT_MULTICAST_LOOP; 1829 return (0); 1830 1831 default: 1832 return (EOPNOTSUPP); 1833 } 1834 } 1835 1836 /* 1837 * Discard the IP multicast options. 1838 */ 1839 void 1840 ip_freemoptions(struct ip_moptions *imo) 1841 { 1842 int i; 1843 1844 if (imo != NULL) { 1845 for (i = 0; i < imo->imo_num_memberships; ++i) 1846 in_delmulti(imo->imo_membership[i]); 1847 free(imo, M_IPMOPTS); 1848 } 1849 } 1850 1851 /* 1852 * Routine called from ip_output() to loop back a copy of an IP multicast 1853 * packet to the input queue of a specified interface. Note that this 1854 * calls the output routine of the loopback "driver", but with an interface 1855 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 1856 */ 1857 static void 1858 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst) 1859 { 1860 struct ip *ip; 1861 struct mbuf *copym; 1862 1863 copym = m_copy(m, 0, M_COPYALL); 1864 if (copym != NULL 1865 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1866 copym = m_pullup(copym, sizeof(struct ip)); 1867 if (copym != NULL) { 1868 /* 1869 * We don't bother to fragment if the IP length is greater 1870 * than the interface's MTU. Can this possibly matter? 1871 */ 1872 ip = mtod(copym, struct ip *); 1873 1874 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1875 in_delayed_cksum(copym); 1876 copym->m_pkthdr.csum_flags &= 1877 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1878 } 1879 1880 ip->ip_sum = 0; 1881 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1882 (void) looutput(ifp, copym, sintosa(dst), NULL); 1883 } 1884 } 1885