ip_output.c revision 1.107.2.9
1 /* $NetBSD: ip_output.c,v 1.107.2.9 2005/03/08 13:53:12 skrll 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 1.107.2.9 2005/03/08 13:53:12 skrll 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 m->m_pkthdr.csum_data |= hlen << 16; 762 763 #if IFA_STATS 764 /* 765 * search for the source address structure to 766 * maintain output statistics. 767 */ 768 INADDR_TO_IA(ip->ip_src, ia); 769 #endif 770 771 if (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) { 772 #if IFA_STATS 773 if (ia) 774 ia->ia_ifa.ifa_data.ifad_outbytes += ip_len; 775 #endif 776 #ifdef IPSEC 777 /* clean ipsec history once it goes out of the node */ 778 ipsec_delaux(m); 779 #endif 780 error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt); 781 goto done; 782 } 783 784 /* Maybe skip checksums on loopback interfaces. */ 785 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 786 ip_do_loopback_cksum)) 787 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 788 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 789 /* 790 * If small enough for mtu of path, can just send directly. 791 */ 792 if (ip_len <= mtu) { 793 #if IFA_STATS 794 if (ia) 795 ia->ia_ifa.ifa_data.ifad_outbytes += ip_len; 796 #endif 797 /* 798 * Always initialize the sum to 0! Some HW assisted 799 * checksumming requires this. 800 */ 801 ip->ip_sum = 0; 802 803 /* 804 * Perform any checksums that the hardware can't do 805 * for us. 806 * 807 * XXX Does any hardware require the {th,uh}_sum 808 * XXX fields to be 0? 809 */ 810 if (sw_csum & M_CSUM_IPv4) { 811 ip->ip_sum = in_cksum(m, hlen); 812 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 813 } 814 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 815 in_delayed_cksum(m); 816 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 817 } 818 819 #ifdef IPSEC 820 /* clean ipsec history once it goes out of the node */ 821 ipsec_delaux(m); 822 #endif 823 error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt); 824 goto done; 825 } 826 827 /* 828 * We can't use HW checksumming if we're about to 829 * to fragment the packet. 830 * 831 * XXX Some hardware can do this. 832 */ 833 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 834 in_delayed_cksum(m); 835 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 836 } 837 838 /* 839 * Too large for interface; fragment if possible. 840 * Must be able to put at least 8 bytes per fragment. 841 */ 842 if (ntohs(ip->ip_off) & IP_DF) { 843 if (flags & IP_RETURNMTU) 844 *mtu_p = mtu; 845 error = EMSGSIZE; 846 ipstat.ips_cantfrag++; 847 goto bad; 848 } 849 850 error = ip_fragment(m, ifp, mtu); 851 if (error) { 852 m = NULL; 853 goto bad; 854 } 855 856 for (; m; m = m0) { 857 m0 = m->m_nextpkt; 858 m->m_nextpkt = 0; 859 if (error == 0) { 860 #if IFA_STATS 861 if (ia) 862 ia->ia_ifa.ifa_data.ifad_outbytes += 863 ntohs(ip->ip_len); 864 #endif 865 #ifdef IPSEC 866 /* clean ipsec history once it goes out of the node */ 867 ipsec_delaux(m); 868 869 #ifdef IPSEC_NAT_T 870 /* 871 * If we get there, the packet has not been handeld by 872 * IPSec whereas it should have. Now that it has been 873 * fragmented, re-inject it in ip_output so that IPsec 874 * processing can occur. 875 */ 876 if (natt_frag) { 877 error = ip_output(m, opt, 878 ro, flags, imo, so, mtu_p); 879 } else 880 #endif /* IPSEC_NAT_T */ 881 #endif /* IPSEC */ 882 { 883 KASSERT((m->m_pkthdr.csum_flags & 884 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 885 error = (*ifp->if_output)(ifp, m, sintosa(dst), 886 ro->ro_rt); 887 } 888 } else 889 m_freem(m); 890 } 891 892 if (error == 0) 893 ipstat.ips_fragmented++; 894 done: 895 if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) { 896 RTFREE(ro->ro_rt); 897 ro->ro_rt = 0; 898 } 899 900 #ifdef IPSEC 901 if (sp != NULL) { 902 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 903 printf("DP ip_output call free SP:%p\n", sp)); 904 key_freesp(sp); 905 } 906 #endif /* IPSEC */ 907 #ifdef FAST_IPSEC 908 if (sp != NULL) 909 KEY_FREESP(&sp); 910 #endif /* FAST_IPSEC */ 911 912 return (error); 913 bad: 914 m_freem(m); 915 goto done; 916 } 917 918 int 919 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 920 { 921 struct ip *ip, *mhip; 922 struct mbuf *m0; 923 int len, hlen, off; 924 int mhlen, firstlen; 925 struct mbuf **mnext; 926 int sw_csum = m->m_pkthdr.csum_flags; 927 int fragments = 0; 928 int s; 929 int error = 0; 930 931 ip = mtod(m, struct ip *); 932 hlen = ip->ip_hl << 2; 933 if (ifp != NULL) 934 sw_csum &= ~ifp->if_csum_flags_tx; 935 936 len = (mtu - hlen) &~ 7; 937 if (len < 8) { 938 m_freem(m); 939 return (EMSGSIZE); 940 } 941 942 firstlen = len; 943 mnext = &m->m_nextpkt; 944 945 /* 946 * Loop through length of segment after first fragment, 947 * make new header and copy data of each part and link onto chain. 948 */ 949 m0 = m; 950 mhlen = sizeof (struct ip); 951 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 952 MGETHDR(m, M_DONTWAIT, MT_HEADER); 953 if (m == 0) { 954 error = ENOBUFS; 955 ipstat.ips_odropped++; 956 goto sendorfree; 957 } 958 MCLAIM(m, m0->m_owner); 959 *mnext = m; 960 mnext = &m->m_nextpkt; 961 m->m_data += max_linkhdr; 962 mhip = mtod(m, struct ip *); 963 *mhip = *ip; 964 /* we must inherit MCAST and BCAST flags */ 965 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 966 if (hlen > sizeof (struct ip)) { 967 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 968 mhip->ip_hl = mhlen >> 2; 969 } 970 m->m_len = mhlen; 971 mhip->ip_off = ((off - hlen) >> 3) + 972 (ntohs(ip->ip_off) & ~IP_MF); 973 if (ip->ip_off & htons(IP_MF)) 974 mhip->ip_off |= IP_MF; 975 if (off + len >= ntohs(ip->ip_len)) 976 len = ntohs(ip->ip_len) - off; 977 else 978 mhip->ip_off |= IP_MF; 979 HTONS(mhip->ip_off); 980 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 981 m->m_next = m_copy(m0, off, len); 982 if (m->m_next == 0) { 983 error = ENOBUFS; /* ??? */ 984 ipstat.ips_odropped++; 985 goto sendorfree; 986 } 987 m->m_pkthdr.len = mhlen + len; 988 m->m_pkthdr.rcvif = (struct ifnet *)0; 989 mhip->ip_sum = 0; 990 if (sw_csum & M_CSUM_IPv4) { 991 mhip->ip_sum = in_cksum(m, mhlen); 992 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 993 } else { 994 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 995 m->m_pkthdr.csum_data |= hlen << 16; 996 } 997 ipstat.ips_ofragments++; 998 fragments++; 999 } 1000 /* 1001 * Update first fragment by trimming what's been copied out 1002 * and updating header, then send each fragment (in order). 1003 */ 1004 m = m0; 1005 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 1006 m->m_pkthdr.len = hlen + firstlen; 1007 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 1008 ip->ip_off |= htons(IP_MF); 1009 ip->ip_sum = 0; 1010 if (sw_csum & M_CSUM_IPv4) { 1011 ip->ip_sum = in_cksum(m, hlen); 1012 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 1013 } else { 1014 KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4); 1015 m->m_pkthdr.csum_data |= hlen << 16; 1016 } 1017 sendorfree: 1018 /* 1019 * If there is no room for all the fragments, don't queue 1020 * any of them. 1021 */ 1022 if (ifp != NULL) { 1023 s = splnet(); 1024 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 1025 error == 0) { 1026 error = ENOBUFS; 1027 ipstat.ips_odropped++; 1028 IFQ_INC_DROPS(&ifp->if_snd); 1029 } 1030 splx(s); 1031 } 1032 if (error) { 1033 for (m = m0; m; m = m0) { 1034 m0 = m->m_nextpkt; 1035 m->m_nextpkt = NULL; 1036 m_freem(m); 1037 } 1038 } 1039 return (error); 1040 } 1041 1042 /* 1043 * Process a delayed payload checksum calculation. 1044 */ 1045 void 1046 in_delayed_cksum(struct mbuf *m) 1047 { 1048 struct ip *ip; 1049 u_int16_t csum, offset; 1050 1051 ip = mtod(m, struct ip *); 1052 offset = ip->ip_hl << 2; 1053 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 1054 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 1055 csum = 0xffff; 1056 1057 offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data); 1058 1059 if ((offset + sizeof(u_int16_t)) > m->m_len) { 1060 /* This happen when ip options were inserted 1061 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 1062 m->m_len, offset, ip->ip_p); 1063 */ 1064 m_copyback(m, offset, sizeof(csum), (caddr_t) &csum); 1065 } else 1066 *(u_int16_t *)(mtod(m, caddr_t) + offset) = csum; 1067 } 1068 1069 /* 1070 * Determine the maximum length of the options to be inserted; 1071 * we would far rather allocate too much space rather than too little. 1072 */ 1073 1074 u_int 1075 ip_optlen(struct inpcb *inp) 1076 { 1077 struct mbuf *m = inp->inp_options; 1078 1079 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) 1080 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 1081 else 1082 return 0; 1083 } 1084 1085 1086 /* 1087 * Insert IP options into preformed packet. 1088 * Adjust IP destination as required for IP source routing, 1089 * as indicated by a non-zero in_addr at the start of the options. 1090 */ 1091 static struct mbuf * 1092 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 1093 { 1094 struct ipoption *p = mtod(opt, struct ipoption *); 1095 struct mbuf *n; 1096 struct ip *ip = mtod(m, struct ip *); 1097 unsigned optlen; 1098 1099 optlen = opt->m_len - sizeof(p->ipopt_dst); 1100 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 1101 return (m); /* XXX should fail */ 1102 if (!in_nullhost(p->ipopt_dst)) 1103 ip->ip_dst = p->ipopt_dst; 1104 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 1105 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1106 if (n == 0) 1107 return (m); 1108 MCLAIM(n, m->m_owner); 1109 M_COPY_PKTHDR(n, m); 1110 m_tag_delete_chain(m, NULL); 1111 m->m_flags &= ~M_PKTHDR; 1112 m->m_len -= sizeof(struct ip); 1113 m->m_data += sizeof(struct ip); 1114 n->m_next = m; 1115 m = n; 1116 m->m_len = optlen + sizeof(struct ip); 1117 m->m_data += max_linkhdr; 1118 bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1119 } else { 1120 m->m_data -= optlen; 1121 m->m_len += optlen; 1122 memmove(mtod(m, caddr_t), ip, sizeof(struct ip)); 1123 } 1124 m->m_pkthdr.len += optlen; 1125 ip = mtod(m, struct ip *); 1126 bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen); 1127 *phlen = sizeof(struct ip) + optlen; 1128 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 1129 return (m); 1130 } 1131 1132 /* 1133 * Copy options from ip to jp, 1134 * omitting those not copied during fragmentation. 1135 */ 1136 int 1137 ip_optcopy(struct ip *ip, struct ip *jp) 1138 { 1139 u_char *cp, *dp; 1140 int opt, optlen, cnt; 1141 1142 cp = (u_char *)(ip + 1); 1143 dp = (u_char *)(jp + 1); 1144 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1145 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1146 opt = cp[0]; 1147 if (opt == IPOPT_EOL) 1148 break; 1149 if (opt == IPOPT_NOP) { 1150 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1151 *dp++ = IPOPT_NOP; 1152 optlen = 1; 1153 continue; 1154 } 1155 #ifdef DIAGNOSTIC 1156 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1157 panic("malformed IPv4 option passed to ip_optcopy"); 1158 #endif 1159 optlen = cp[IPOPT_OLEN]; 1160 #ifdef DIAGNOSTIC 1161 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1162 panic("malformed IPv4 option passed to ip_optcopy"); 1163 #endif 1164 /* bogus lengths should have been caught by ip_dooptions */ 1165 if (optlen > cnt) 1166 optlen = cnt; 1167 if (IPOPT_COPIED(opt)) { 1168 bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen); 1169 dp += optlen; 1170 } 1171 } 1172 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1173 *dp++ = IPOPT_EOL; 1174 return (optlen); 1175 } 1176 1177 /* 1178 * IP socket option processing. 1179 */ 1180 int 1181 ip_ctloutput(int op, struct socket *so, int level, int optname, 1182 struct mbuf **mp) 1183 { 1184 struct inpcb *inp = sotoinpcb(so); 1185 struct mbuf *m = *mp; 1186 int optval = 0; 1187 int error = 0; 1188 #if defined(IPSEC) || defined(FAST_IPSEC) 1189 struct proc *p = curproc; /*XXX*/ 1190 #endif 1191 1192 if (level != IPPROTO_IP) { 1193 error = EINVAL; 1194 if (op == PRCO_SETOPT && *mp) 1195 (void) m_free(*mp); 1196 } else switch (op) { 1197 1198 case PRCO_SETOPT: 1199 switch (optname) { 1200 case IP_OPTIONS: 1201 #ifdef notyet 1202 case IP_RETOPTS: 1203 return (ip_pcbopts(optname, &inp->inp_options, m)); 1204 #else 1205 return (ip_pcbopts(&inp->inp_options, m)); 1206 #endif 1207 1208 case IP_TOS: 1209 case IP_TTL: 1210 case IP_RECVOPTS: 1211 case IP_RECVRETOPTS: 1212 case IP_RECVDSTADDR: 1213 case IP_RECVIF: 1214 if (m == NULL || m->m_len != sizeof(int)) 1215 error = EINVAL; 1216 else { 1217 optval = *mtod(m, int *); 1218 switch (optname) { 1219 1220 case IP_TOS: 1221 inp->inp_ip.ip_tos = optval; 1222 break; 1223 1224 case IP_TTL: 1225 inp->inp_ip.ip_ttl = optval; 1226 break; 1227 #define OPTSET(bit) \ 1228 if (optval) \ 1229 inp->inp_flags |= bit; \ 1230 else \ 1231 inp->inp_flags &= ~bit; 1232 1233 case IP_RECVOPTS: 1234 OPTSET(INP_RECVOPTS); 1235 break; 1236 1237 case IP_RECVRETOPTS: 1238 OPTSET(INP_RECVRETOPTS); 1239 break; 1240 1241 case IP_RECVDSTADDR: 1242 OPTSET(INP_RECVDSTADDR); 1243 break; 1244 1245 case IP_RECVIF: 1246 OPTSET(INP_RECVIF); 1247 break; 1248 } 1249 } 1250 break; 1251 #undef OPTSET 1252 1253 case IP_MULTICAST_IF: 1254 case IP_MULTICAST_TTL: 1255 case IP_MULTICAST_LOOP: 1256 case IP_ADD_MEMBERSHIP: 1257 case IP_DROP_MEMBERSHIP: 1258 error = ip_setmoptions(optname, &inp->inp_moptions, m); 1259 break; 1260 1261 case IP_PORTRANGE: 1262 if (m == 0 || m->m_len != sizeof(int)) 1263 error = EINVAL; 1264 else { 1265 optval = *mtod(m, int *); 1266 1267 switch (optval) { 1268 1269 case IP_PORTRANGE_DEFAULT: 1270 case IP_PORTRANGE_HIGH: 1271 inp->inp_flags &= ~(INP_LOWPORT); 1272 break; 1273 1274 case IP_PORTRANGE_LOW: 1275 inp->inp_flags |= INP_LOWPORT; 1276 break; 1277 1278 default: 1279 error = EINVAL; 1280 break; 1281 } 1282 } 1283 break; 1284 1285 #if defined(IPSEC) || defined(FAST_IPSEC) 1286 case IP_IPSEC_POLICY: 1287 { 1288 caddr_t req = NULL; 1289 size_t len = 0; 1290 int priv = 0; 1291 1292 #ifdef __NetBSD__ 1293 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1294 priv = 0; 1295 else 1296 priv = 1; 1297 #else 1298 priv = (in6p->in6p_socket->so_state & SS_PRIV); 1299 #endif 1300 if (m) { 1301 req = mtod(m, caddr_t); 1302 len = m->m_len; 1303 } 1304 error = ipsec4_set_policy(inp, optname, req, len, priv); 1305 break; 1306 } 1307 #endif /*IPSEC*/ 1308 1309 default: 1310 error = ENOPROTOOPT; 1311 break; 1312 } 1313 if (m) 1314 (void)m_free(m); 1315 break; 1316 1317 case PRCO_GETOPT: 1318 switch (optname) { 1319 case IP_OPTIONS: 1320 case IP_RETOPTS: 1321 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1322 MCLAIM(m, so->so_mowner); 1323 if (inp->inp_options) { 1324 m->m_len = inp->inp_options->m_len; 1325 bcopy(mtod(inp->inp_options, caddr_t), 1326 mtod(m, caddr_t), (unsigned)m->m_len); 1327 } else 1328 m->m_len = 0; 1329 break; 1330 1331 case IP_TOS: 1332 case IP_TTL: 1333 case IP_RECVOPTS: 1334 case IP_RECVRETOPTS: 1335 case IP_RECVDSTADDR: 1336 case IP_RECVIF: 1337 case IP_ERRORMTU: 1338 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1339 MCLAIM(m, so->so_mowner); 1340 m->m_len = sizeof(int); 1341 switch (optname) { 1342 1343 case IP_TOS: 1344 optval = inp->inp_ip.ip_tos; 1345 break; 1346 1347 case IP_TTL: 1348 optval = inp->inp_ip.ip_ttl; 1349 break; 1350 1351 case IP_ERRORMTU: 1352 optval = inp->inp_errormtu; 1353 break; 1354 1355 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1356 1357 case IP_RECVOPTS: 1358 optval = OPTBIT(INP_RECVOPTS); 1359 break; 1360 1361 case IP_RECVRETOPTS: 1362 optval = OPTBIT(INP_RECVRETOPTS); 1363 break; 1364 1365 case IP_RECVDSTADDR: 1366 optval = OPTBIT(INP_RECVDSTADDR); 1367 break; 1368 1369 case IP_RECVIF: 1370 optval = OPTBIT(INP_RECVIF); 1371 break; 1372 } 1373 *mtod(m, int *) = optval; 1374 break; 1375 1376 #if 0 /* defined(IPSEC) || defined(FAST_IPSEC) */ 1377 /* XXX: code broken */ 1378 case IP_IPSEC_POLICY: 1379 { 1380 caddr_t req = NULL; 1381 size_t len = 0; 1382 1383 if (m) { 1384 req = mtod(m, caddr_t); 1385 len = m->m_len; 1386 } 1387 error = ipsec4_get_policy(inp, req, len, mp); 1388 break; 1389 } 1390 #endif /*IPSEC*/ 1391 1392 case IP_MULTICAST_IF: 1393 case IP_MULTICAST_TTL: 1394 case IP_MULTICAST_LOOP: 1395 case IP_ADD_MEMBERSHIP: 1396 case IP_DROP_MEMBERSHIP: 1397 error = ip_getmoptions(optname, inp->inp_moptions, mp); 1398 if (*mp) 1399 MCLAIM(*mp, so->so_mowner); 1400 break; 1401 1402 case IP_PORTRANGE: 1403 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1404 MCLAIM(m, so->so_mowner); 1405 m->m_len = sizeof(int); 1406 1407 if (inp->inp_flags & INP_LOWPORT) 1408 optval = IP_PORTRANGE_LOW; 1409 else 1410 optval = IP_PORTRANGE_DEFAULT; 1411 1412 *mtod(m, int *) = optval; 1413 break; 1414 1415 default: 1416 error = ENOPROTOOPT; 1417 break; 1418 } 1419 break; 1420 } 1421 return (error); 1422 } 1423 1424 /* 1425 * Set up IP options in pcb for insertion in output packets. 1426 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1427 * with destination address if source routed. 1428 */ 1429 int 1430 #ifdef notyet 1431 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m) 1432 #else 1433 ip_pcbopts(struct mbuf **pcbopt, struct mbuf *m) 1434 #endif 1435 { 1436 int cnt, optlen; 1437 u_char *cp; 1438 u_char opt; 1439 1440 /* turn off any old options */ 1441 if (*pcbopt) 1442 (void)m_free(*pcbopt); 1443 *pcbopt = 0; 1444 if (m == (struct mbuf *)0 || m->m_len == 0) { 1445 /* 1446 * Only turning off any previous options. 1447 */ 1448 if (m) 1449 (void)m_free(m); 1450 return (0); 1451 } 1452 1453 #ifndef __vax__ 1454 if (m->m_len % sizeof(int32_t)) 1455 goto bad; 1456 #endif 1457 /* 1458 * IP first-hop destination address will be stored before 1459 * actual options; move other options back 1460 * and clear it when none present. 1461 */ 1462 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1463 goto bad; 1464 cnt = m->m_len; 1465 m->m_len += sizeof(struct in_addr); 1466 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1467 memmove(cp, mtod(m, caddr_t), (unsigned)cnt); 1468 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1469 1470 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1471 opt = cp[IPOPT_OPTVAL]; 1472 if (opt == IPOPT_EOL) 1473 break; 1474 if (opt == IPOPT_NOP) 1475 optlen = 1; 1476 else { 1477 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1478 goto bad; 1479 optlen = cp[IPOPT_OLEN]; 1480 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1481 goto bad; 1482 } 1483 switch (opt) { 1484 1485 default: 1486 break; 1487 1488 case IPOPT_LSRR: 1489 case IPOPT_SSRR: 1490 /* 1491 * user process specifies route as: 1492 * ->A->B->C->D 1493 * D must be our final destination (but we can't 1494 * check that since we may not have connected yet). 1495 * A is first hop destination, which doesn't appear in 1496 * actual IP option, but is stored before the options. 1497 */ 1498 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1499 goto bad; 1500 m->m_len -= sizeof(struct in_addr); 1501 cnt -= sizeof(struct in_addr); 1502 optlen -= sizeof(struct in_addr); 1503 cp[IPOPT_OLEN] = optlen; 1504 /* 1505 * Move first hop before start of options. 1506 */ 1507 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1508 sizeof(struct in_addr)); 1509 /* 1510 * Then copy rest of options back 1511 * to close up the deleted entry. 1512 */ 1513 (void)memmove(&cp[IPOPT_OFFSET+1], 1514 &cp[IPOPT_OFFSET+1] + sizeof(struct in_addr), 1515 (unsigned)cnt - (IPOPT_MINOFF - 1)); 1516 break; 1517 } 1518 } 1519 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1520 goto bad; 1521 *pcbopt = m; 1522 return (0); 1523 1524 bad: 1525 (void)m_free(m); 1526 return (EINVAL); 1527 } 1528 1529 /* 1530 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1531 */ 1532 static struct ifnet * 1533 ip_multicast_if(struct in_addr *a, int *ifindexp) 1534 { 1535 int ifindex; 1536 struct ifnet *ifp = NULL; 1537 struct in_ifaddr *ia; 1538 1539 if (ifindexp) 1540 *ifindexp = 0; 1541 if (ntohl(a->s_addr) >> 24 == 0) { 1542 ifindex = ntohl(a->s_addr) & 0xffffff; 1543 if (ifindex < 0 || if_indexlim <= ifindex) 1544 return NULL; 1545 ifp = ifindex2ifnet[ifindex]; 1546 if (!ifp) 1547 return NULL; 1548 if (ifindexp) 1549 *ifindexp = ifindex; 1550 } else { 1551 LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) { 1552 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1553 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1554 ifp = ia->ia_ifp; 1555 break; 1556 } 1557 } 1558 } 1559 return ifp; 1560 } 1561 1562 /* 1563 * Set the IP multicast options in response to user setsockopt(). 1564 */ 1565 int 1566 ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m) 1567 { 1568 int error = 0; 1569 u_char loop; 1570 int i; 1571 struct in_addr addr; 1572 struct ip_mreq *mreq; 1573 struct ifnet *ifp; 1574 struct ip_moptions *imo = *imop; 1575 struct route ro; 1576 struct sockaddr_in *dst; 1577 int ifindex; 1578 1579 if (imo == NULL) { 1580 /* 1581 * No multicast option buffer attached to the pcb; 1582 * allocate one and initialize to default values. 1583 */ 1584 imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS, 1585 M_WAITOK); 1586 1587 if (imo == NULL) 1588 return (ENOBUFS); 1589 *imop = imo; 1590 imo->imo_multicast_ifp = NULL; 1591 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1592 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1593 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1594 imo->imo_num_memberships = 0; 1595 } 1596 1597 switch (optname) { 1598 1599 case IP_MULTICAST_IF: 1600 /* 1601 * Select the interface for outgoing multicast packets. 1602 */ 1603 if (m == NULL || m->m_len != sizeof(struct in_addr)) { 1604 error = EINVAL; 1605 break; 1606 } 1607 addr = *(mtod(m, struct in_addr *)); 1608 /* 1609 * INADDR_ANY is used to remove a previous selection. 1610 * When no interface is selected, a default one is 1611 * chosen every time a multicast packet is sent. 1612 */ 1613 if (in_nullhost(addr)) { 1614 imo->imo_multicast_ifp = NULL; 1615 break; 1616 } 1617 /* 1618 * The selected interface is identified by its local 1619 * IP address. Find the interface and confirm that 1620 * it supports multicasting. 1621 */ 1622 ifp = ip_multicast_if(&addr, &ifindex); 1623 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1624 error = EADDRNOTAVAIL; 1625 break; 1626 } 1627 imo->imo_multicast_ifp = ifp; 1628 if (ifindex) 1629 imo->imo_multicast_addr = addr; 1630 else 1631 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1632 break; 1633 1634 case IP_MULTICAST_TTL: 1635 /* 1636 * Set the IP time-to-live for outgoing multicast packets. 1637 */ 1638 if (m == NULL || m->m_len != 1) { 1639 error = EINVAL; 1640 break; 1641 } 1642 imo->imo_multicast_ttl = *(mtod(m, u_char *)); 1643 break; 1644 1645 case IP_MULTICAST_LOOP: 1646 /* 1647 * Set the loopback flag for outgoing multicast packets. 1648 * Must be zero or one. 1649 */ 1650 if (m == NULL || m->m_len != 1 || 1651 (loop = *(mtod(m, u_char *))) > 1) { 1652 error = EINVAL; 1653 break; 1654 } 1655 imo->imo_multicast_loop = loop; 1656 break; 1657 1658 case IP_ADD_MEMBERSHIP: 1659 /* 1660 * Add a multicast group membership. 1661 * Group must be a valid IP multicast address. 1662 */ 1663 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1664 error = EINVAL; 1665 break; 1666 } 1667 mreq = mtod(m, struct ip_mreq *); 1668 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1669 error = EINVAL; 1670 break; 1671 } 1672 /* 1673 * If no interface address was provided, use the interface of 1674 * the route to the given multicast address. 1675 */ 1676 if (in_nullhost(mreq->imr_interface)) { 1677 bzero((caddr_t)&ro, sizeof(ro)); 1678 ro.ro_rt = NULL; 1679 dst = satosin(&ro.ro_dst); 1680 dst->sin_len = sizeof(*dst); 1681 dst->sin_family = AF_INET; 1682 dst->sin_addr = mreq->imr_multiaddr; 1683 rtalloc(&ro); 1684 if (ro.ro_rt == NULL) { 1685 error = EADDRNOTAVAIL; 1686 break; 1687 } 1688 ifp = ro.ro_rt->rt_ifp; 1689 rtfree(ro.ro_rt); 1690 } else { 1691 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1692 } 1693 /* 1694 * See if we found an interface, and confirm that it 1695 * supports multicast. 1696 */ 1697 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1698 error = EADDRNOTAVAIL; 1699 break; 1700 } 1701 /* 1702 * See if the membership already exists or if all the 1703 * membership slots are full. 1704 */ 1705 for (i = 0; i < imo->imo_num_memberships; ++i) { 1706 if (imo->imo_membership[i]->inm_ifp == ifp && 1707 in_hosteq(imo->imo_membership[i]->inm_addr, 1708 mreq->imr_multiaddr)) 1709 break; 1710 } 1711 if (i < imo->imo_num_memberships) { 1712 error = EADDRINUSE; 1713 break; 1714 } 1715 if (i == IP_MAX_MEMBERSHIPS) { 1716 error = ETOOMANYREFS; 1717 break; 1718 } 1719 /* 1720 * Everything looks good; add a new record to the multicast 1721 * address list for the given interface. 1722 */ 1723 if ((imo->imo_membership[i] = 1724 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) { 1725 error = ENOBUFS; 1726 break; 1727 } 1728 ++imo->imo_num_memberships; 1729 break; 1730 1731 case IP_DROP_MEMBERSHIP: 1732 /* 1733 * Drop a multicast group membership. 1734 * Group must be a valid IP multicast address. 1735 */ 1736 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1737 error = EINVAL; 1738 break; 1739 } 1740 mreq = mtod(m, struct ip_mreq *); 1741 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1742 error = EINVAL; 1743 break; 1744 } 1745 /* 1746 * If an interface address was specified, get a pointer 1747 * to its ifnet structure. 1748 */ 1749 if (in_nullhost(mreq->imr_interface)) 1750 ifp = NULL; 1751 else { 1752 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1753 if (ifp == NULL) { 1754 error = EADDRNOTAVAIL; 1755 break; 1756 } 1757 } 1758 /* 1759 * Find the membership in the membership array. 1760 */ 1761 for (i = 0; i < imo->imo_num_memberships; ++i) { 1762 if ((ifp == NULL || 1763 imo->imo_membership[i]->inm_ifp == ifp) && 1764 in_hosteq(imo->imo_membership[i]->inm_addr, 1765 mreq->imr_multiaddr)) 1766 break; 1767 } 1768 if (i == imo->imo_num_memberships) { 1769 error = EADDRNOTAVAIL; 1770 break; 1771 } 1772 /* 1773 * Give up the multicast address record to which the 1774 * membership points. 1775 */ 1776 in_delmulti(imo->imo_membership[i]); 1777 /* 1778 * Remove the gap in the membership array. 1779 */ 1780 for (++i; i < imo->imo_num_memberships; ++i) 1781 imo->imo_membership[i-1] = imo->imo_membership[i]; 1782 --imo->imo_num_memberships; 1783 break; 1784 1785 default: 1786 error = EOPNOTSUPP; 1787 break; 1788 } 1789 1790 /* 1791 * If all options have default values, no need to keep the mbuf. 1792 */ 1793 if (imo->imo_multicast_ifp == NULL && 1794 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1795 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1796 imo->imo_num_memberships == 0) { 1797 free(*imop, M_IPMOPTS); 1798 *imop = NULL; 1799 } 1800 1801 return (error); 1802 } 1803 1804 /* 1805 * Return the IP multicast options in response to user getsockopt(). 1806 */ 1807 int 1808 ip_getmoptions(int optname, struct ip_moptions *imo, struct mbuf **mp) 1809 { 1810 u_char *ttl; 1811 u_char *loop; 1812 struct in_addr *addr; 1813 struct in_ifaddr *ia; 1814 1815 *mp = m_get(M_WAIT, MT_SOOPTS); 1816 1817 switch (optname) { 1818 1819 case IP_MULTICAST_IF: 1820 addr = mtod(*mp, struct in_addr *); 1821 (*mp)->m_len = sizeof(struct in_addr); 1822 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1823 *addr = zeroin_addr; 1824 else if (imo->imo_multicast_addr.s_addr) { 1825 /* return the value user has set */ 1826 *addr = imo->imo_multicast_addr; 1827 } else { 1828 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1829 *addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1830 } 1831 return (0); 1832 1833 case IP_MULTICAST_TTL: 1834 ttl = mtod(*mp, u_char *); 1835 (*mp)->m_len = 1; 1836 *ttl = imo ? imo->imo_multicast_ttl 1837 : IP_DEFAULT_MULTICAST_TTL; 1838 return (0); 1839 1840 case IP_MULTICAST_LOOP: 1841 loop = mtod(*mp, u_char *); 1842 (*mp)->m_len = 1; 1843 *loop = imo ? imo->imo_multicast_loop 1844 : IP_DEFAULT_MULTICAST_LOOP; 1845 return (0); 1846 1847 default: 1848 return (EOPNOTSUPP); 1849 } 1850 } 1851 1852 /* 1853 * Discard the IP multicast options. 1854 */ 1855 void 1856 ip_freemoptions(struct ip_moptions *imo) 1857 { 1858 int i; 1859 1860 if (imo != NULL) { 1861 for (i = 0; i < imo->imo_num_memberships; ++i) 1862 in_delmulti(imo->imo_membership[i]); 1863 free(imo, M_IPMOPTS); 1864 } 1865 } 1866 1867 /* 1868 * Routine called from ip_output() to loop back a copy of an IP multicast 1869 * packet to the input queue of a specified interface. Note that this 1870 * calls the output routine of the loopback "driver", but with an interface 1871 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 1872 */ 1873 static void 1874 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst) 1875 { 1876 struct ip *ip; 1877 struct mbuf *copym; 1878 1879 copym = m_copy(m, 0, M_COPYALL); 1880 if (copym != NULL 1881 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1882 copym = m_pullup(copym, sizeof(struct ip)); 1883 if (copym != NULL) { 1884 /* 1885 * We don't bother to fragment if the IP length is greater 1886 * than the interface's MTU. Can this possibly matter? 1887 */ 1888 ip = mtod(copym, struct ip *); 1889 1890 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1891 in_delayed_cksum(copym); 1892 copym->m_pkthdr.csum_flags &= 1893 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1894 } 1895 1896 ip->ip_sum = 0; 1897 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1898 (void) looutput(ifp, copym, sintosa(dst), NULL); 1899 } 1900 } 1901
Indexes created Wed Oct 01 13:09:50 GMT 2025