ip_input.c revision 1.141
1/* $NetBSD: ip_input.c,v 1.141 2001/11/13 00:32:38 lukem 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, 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. All advertising materials mentioning features or use of this software 82 * must display the following acknowledgement: 83 * This product includes software developed by the University of 84 * California, Berkeley and its contributors. 85 * 4. Neither the name of the University nor the names of its contributors 86 * may be used to endorse or promote products derived from this software 87 * without specific prior written permission. 88 * 89 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 90 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 91 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 92 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 93 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 94 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 95 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 96 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 97 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 98 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 99 * SUCH DAMAGE. 100 * 101 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 102 */ 103 104#include <sys/cdefs.h> 105__KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.141 2001/11/13 00:32:38 lukem Exp $"); 106 107#include "opt_gateway.h" 108#include "opt_pfil_hooks.h" 109#include "opt_ipsec.h" 110#include "opt_mrouting.h" 111#include "opt_inet_csum.h" 112 113#include <sys/param.h> 114#include <sys/systm.h> 115#include <sys/malloc.h> 116#include <sys/mbuf.h> 117#include <sys/domain.h> 118#include <sys/protosw.h> 119#include <sys/socket.h> 120#include <sys/socketvar.h> 121#include <sys/errno.h> 122#include <sys/time.h> 123#include <sys/kernel.h> 124#include <sys/pool.h> 125#include <sys/sysctl.h> 126 127#include <net/if.h> 128#include <net/if_dl.h> 129#include <net/route.h> 130#include <net/pfil.h> 131 132#include <netinet/in.h> 133#include <netinet/in_systm.h> 134#include <netinet/ip.h> 135#include <netinet/in_pcb.h> 136#include <netinet/in_var.h> 137#include <netinet/ip_var.h> 138#include <netinet/ip_icmp.h> 139/* just for gif_ttl */ 140#include <netinet/in_gif.h> 141#include "gif.h" 142 143#ifdef MROUTING 144#include <netinet/ip_mroute.h> 145#endif 146 147#ifdef IPSEC 148#include <netinet6/ipsec.h> 149#include <netkey/key.h> 150#endif 151 152#ifndef IPFORWARDING 153#ifdef GATEWAY 154#define IPFORWARDING 1 /* forward IP packets not for us */ 155#else /* GATEWAY */ 156#define IPFORWARDING 0 /* don't forward IP packets not for us */ 157#endif /* GATEWAY */ 158#endif /* IPFORWARDING */ 159#ifndef IPSENDREDIRECTS 160#define IPSENDREDIRECTS 1 161#endif 162#ifndef IPFORWSRCRT 163#define IPFORWSRCRT 1 /* forward source-routed packets */ 164#endif 165#ifndef IPALLOWSRCRT 166#define IPALLOWSRCRT 1 /* allow source-routed packets */ 167#endif 168#ifndef IPMTUDISC 169#define IPMTUDISC 0 170#endif 171#ifndef IPMTUDISCTIMEOUT 172#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 173#endif 174 175/* 176 * Note: DIRECTED_BROADCAST is handled this way so that previous 177 * configuration using this option will Just Work. 178 */ 179#ifndef IPDIRECTEDBCAST 180#ifdef DIRECTED_BROADCAST 181#define IPDIRECTEDBCAST 1 182#else 183#define IPDIRECTEDBCAST 0 184#endif /* DIRECTED_BROADCAST */ 185#endif /* IPDIRECTEDBCAST */ 186int ipforwarding = IPFORWARDING; 187int ipsendredirects = IPSENDREDIRECTS; 188int ip_defttl = IPDEFTTL; 189int ip_forwsrcrt = IPFORWSRCRT; 190int ip_directedbcast = IPDIRECTEDBCAST; 191int ip_allowsrcrt = IPALLOWSRCRT; 192int ip_mtudisc = IPMTUDISC; 193u_int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 194#ifdef DIAGNOSTIC 195int ipprintfs = 0; 196#endif 197 198struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 199 200extern struct domain inetdomain; 201int ipqmaxlen = IFQ_MAXLEN; 202struct in_ifaddrhead in_ifaddr; 203struct in_ifaddrhashhead *in_ifaddrhashtbl; 204struct ifqueue ipintrq; 205struct ipstat ipstat; 206u_int16_t ip_id; 207 208#ifdef PFIL_HOOKS 209struct pfil_head inet_pfil_hook; 210#endif 211 212struct ipqhead ipq; 213int ipq_locked; 214int ip_nfragpackets = 0; 215int ip_maxfragpackets = 200; 216 217static __inline int ipq_lock_try __P((void)); 218static __inline void ipq_unlock __P((void)); 219 220static __inline int 221ipq_lock_try() 222{ 223 int s; 224 225 /* 226 * Use splvm() -- we're bloking things that would cause 227 * mbuf allocation. 228 */ 229 s = splvm(); 230 if (ipq_locked) { 231 splx(s); 232 return (0); 233 } 234 ipq_locked = 1; 235 splx(s); 236 return (1); 237} 238 239static __inline void 240ipq_unlock() 241{ 242 int s; 243 244 s = splvm(); 245 ipq_locked = 0; 246 splx(s); 247} 248 249#ifdef DIAGNOSTIC 250#define IPQ_LOCK() \ 251do { \ 252 if (ipq_lock_try() == 0) { \ 253 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 254 panic("ipq_lock"); \ 255 } \ 256} while (0) 257#define IPQ_LOCK_CHECK() \ 258do { \ 259 if (ipq_locked == 0) { \ 260 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 261 panic("ipq lock check"); \ 262 } \ 263} while (0) 264#else 265#define IPQ_LOCK() (void) ipq_lock_try() 266#define IPQ_LOCK_CHECK() /* nothing */ 267#endif 268 269#define IPQ_UNLOCK() ipq_unlock() 270 271struct pool ipqent_pool; 272 273#ifdef INET_CSUM_COUNTERS 274#include <sys/device.h> 275 276struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 277 NULL, "inet", "hwcsum bad"); 278struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 279 NULL, "inet", "hwcsum ok"); 280struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 281 NULL, "inet", "swcsum"); 282 283#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 284 285#else 286 287#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 288 289#endif /* INET_CSUM_COUNTERS */ 290 291/* 292 * We need to save the IP options in case a protocol wants to respond 293 * to an incoming packet over the same route if the packet got here 294 * using IP source routing. This allows connection establishment and 295 * maintenance when the remote end is on a network that is not known 296 * to us. 297 */ 298int ip_nhops = 0; 299static struct ip_srcrt { 300 struct in_addr dst; /* final destination */ 301 char nop; /* one NOP to align */ 302 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 303 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 304} ip_srcrt; 305 306static void save_rte __P((u_char *, struct in_addr)); 307 308/* 309 * IP initialization: fill in IP protocol switch table. 310 * All protocols not implemented in kernel go to raw IP protocol handler. 311 */ 312void 313ip_init() 314{ 315 struct protosw *pr; 316 int i; 317 318 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 319 0, NULL, NULL, M_IPQ); 320 321 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 322 if (pr == 0) 323 panic("ip_init"); 324 for (i = 0; i < IPPROTO_MAX; i++) 325 ip_protox[i] = pr - inetsw; 326 for (pr = inetdomain.dom_protosw; 327 pr < inetdomain.dom_protoswNPROTOSW; pr++) 328 if (pr->pr_domain->dom_family == PF_INET && 329 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 330 ip_protox[pr->pr_protocol] = pr - inetsw; 331 LIST_INIT(&ipq); 332 ip_id = time.tv_sec & 0xffff; 333 ipintrq.ifq_maxlen = ipqmaxlen; 334 TAILQ_INIT(&in_ifaddr); 335 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 336 M_WAITOK, &in_ifaddrhash); 337 if (ip_mtudisc != 0) 338 ip_mtudisc_timeout_q = 339 rt_timer_queue_create(ip_mtudisc_timeout); 340#ifdef GATEWAY 341 ipflow_init(); 342#endif 343 344#ifdef PFIL_HOOKS 345 /* Register our Packet Filter hook. */ 346 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 347 inet_pfil_hook.ph_af = AF_INET; 348 i = pfil_head_register(&inet_pfil_hook); 349 if (i != 0) 350 printf("ip_init: WARNING: unable to register pfil hook, " 351 "error %d\n", i); 352#endif /* PFIL_HOOKS */ 353 354#ifdef INET_CSUM_COUNTERS 355 evcnt_attach_static(&ip_hwcsum_bad); 356 evcnt_attach_static(&ip_hwcsum_ok); 357 evcnt_attach_static(&ip_swcsum); 358#endif /* INET_CSUM_COUNTERS */ 359} 360 361struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 362struct route ipforward_rt; 363 364/* 365 * IP software interrupt routine 366 */ 367void 368ipintr() 369{ 370 int s; 371 struct mbuf *m; 372 373 while (1) { 374 s = splnet(); 375 IF_DEQUEUE(&ipintrq, m); 376 splx(s); 377 if (m == 0) 378 return; 379 ip_input(m); 380 } 381} 382 383/* 384 * Ip input routine. Checksum and byte swap header. If fragmented 385 * try to reassemble. Process options. Pass to next level. 386 */ 387void 388ip_input(struct mbuf *m) 389{ 390 struct ip *ip = NULL; 391 struct ipq *fp; 392 struct in_ifaddr *ia; 393 struct ifaddr *ifa; 394 struct ipqent *ipqe; 395 int hlen = 0, mff, len; 396 int downmatch; 397 398#ifdef DIAGNOSTIC 399 if ((m->m_flags & M_PKTHDR) == 0) 400 panic("ipintr no HDR"); 401#endif 402#ifdef IPSEC 403 /* 404 * should the inner packet be considered authentic? 405 * see comment in ah4_input(). 406 */ 407 if (m) { 408 m->m_flags &= ~M_AUTHIPHDR; 409 m->m_flags &= ~M_AUTHIPDGM; 410 } 411#endif 412 /* 413 * If no IP addresses have been set yet but the interfaces 414 * are receiving, can't do anything with incoming packets yet. 415 */ 416 if (TAILQ_FIRST(&in_ifaddr) == 0) 417 goto bad; 418 ipstat.ips_total++; 419 if (m->m_len < sizeof (struct ip) && 420 (m = m_pullup(m, sizeof (struct ip))) == 0) { 421 ipstat.ips_toosmall++; 422 return; 423 } 424 ip = mtod(m, struct ip *); 425 if (ip->ip_v != IPVERSION) { 426 ipstat.ips_badvers++; 427 goto bad; 428 } 429 hlen = ip->ip_hl << 2; 430 if (hlen < sizeof(struct ip)) { /* minimum header length */ 431 ipstat.ips_badhlen++; 432 goto bad; 433 } 434 if (hlen > m->m_len) { 435 if ((m = m_pullup(m, hlen)) == 0) { 436 ipstat.ips_badhlen++; 437 return; 438 } 439 ip = mtod(m, struct ip *); 440 } 441 442 /* 443 * RFC1122: packets with a multicast source address are 444 * not allowed. 445 */ 446 if (IN_MULTICAST(ip->ip_src.s_addr)) { 447 ipstat.ips_badaddr++; 448 goto bad; 449 } 450 451 /* 127/8 must not appear on wire - RFC1122 */ 452 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 453 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 454 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 455 ipstat.ips_badaddr++; 456 goto bad; 457 } 458 } 459 460 switch (m->m_pkthdr.csum_flags & 461 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) | 462 M_CSUM_IPv4_BAD)) { 463 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 464 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 465 goto badcsum; 466 467 case M_CSUM_IPv4: 468 /* Checksum was okay. */ 469 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 470 break; 471 472 default: 473 /* Must compute it ourselves. */ 474 INET_CSUM_COUNTER_INCR(&ip_swcsum); 475 if (in_cksum(m, hlen) != 0) 476 goto bad; 477 break; 478 } 479 480 /* Retrieve the packet length. */ 481 len = ntohs(ip->ip_len); 482 483 /* 484 * Check for additional length bogosity 485 */ 486 if (len < hlen) { 487 ipstat.ips_badlen++; 488 goto bad; 489 } 490 491 /* 492 * Check that the amount of data in the buffers 493 * is as at least much as the IP header would have us expect. 494 * Trim mbufs if longer than we expect. 495 * Drop packet if shorter than we expect. 496 */ 497 if (m->m_pkthdr.len < len) { 498 ipstat.ips_tooshort++; 499 goto bad; 500 } 501 if (m->m_pkthdr.len > len) { 502 if (m->m_len == m->m_pkthdr.len) { 503 m->m_len = len; 504 m->m_pkthdr.len = len; 505 } else 506 m_adj(m, len - m->m_pkthdr.len); 507 } 508 509#ifdef IPSEC 510 /* ipflow (IP fast fowarding) is not compatible with IPsec. */ 511 m->m_flags &= ~M_CANFASTFWD; 512#else 513 /* 514 * Assume that we can create a fast-forward IP flow entry 515 * based on this packet. 516 */ 517 m->m_flags |= M_CANFASTFWD; 518#endif 519 520#ifdef PFIL_HOOKS 521 /* 522 * Run through list of hooks for input packets. If there are any 523 * filters which require that additional packets in the flow are 524 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 525 * Note that filters must _never_ set this flag, as another filter 526 * in the list may have previously cleared it. 527 */ 528 /* 529 * let ipfilter look at packet on the wire, 530 * not the decapsulated packet. 531 */ 532#ifdef IPSEC 533 if (!ipsec_getnhist(m)) 534#else 535 if (1) 536#endif 537 { 538 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 539 PFIL_IN) != 0) 540 return; 541 if (m == NULL) 542 return; 543 ip = mtod(m, struct ip *); 544 } 545#endif /* PFIL_HOOKS */ 546 547#ifdef ALTQ 548 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 549 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 550 /* packet dropped by traffic conditioner */ 551 return; 552 } 553#endif 554 555 /* 556 * Convert fields to host representation. 557 */ 558 NTOHS(ip->ip_len); 559 NTOHS(ip->ip_off); 560 561 /* 562 * Process options and, if not destined for us, 563 * ship it on. ip_dooptions returns 1 when an 564 * error was detected (causing an icmp message 565 * to be sent and the original packet to be freed). 566 */ 567 ip_nhops = 0; /* for source routed packets */ 568 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 569 return; 570 571 /* 572 * Check our list of addresses, to see if the packet is for us. 573 * 574 * Traditional 4.4BSD did not consult IFF_UP at all. 575 * The behavior here is to treat addresses on !IFF_UP interface 576 * as not mine. 577 */ 578 downmatch = 0; 579 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 580 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 581 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 582 break; 583 else 584 downmatch++; 585 } 586 } 587 if (ia != NULL) 588 goto ours; 589 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 590 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrlist, ifa_list) { 591 if (ifa->ifa_addr->sa_family != AF_INET) 592 continue; 593 ia = ifatoia(ifa); 594 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 595 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 596 /* 597 * Look for all-0's host part (old broadcast addr), 598 * either for subnet or net. 599 */ 600 ip->ip_dst.s_addr == ia->ia_subnet || 601 ip->ip_dst.s_addr == ia->ia_net) 602 goto ours; 603 /* 604 * An interface with IP address zero accepts 605 * all packets that arrive on that interface. 606 */ 607 if (in_nullhost(ia->ia_addr.sin_addr)) 608 goto ours; 609 } 610 } 611 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 612 struct in_multi *inm; 613#ifdef MROUTING 614 extern struct socket *ip_mrouter; 615 616 if (m->m_flags & M_EXT) { 617 if ((m = m_pullup(m, hlen)) == 0) { 618 ipstat.ips_toosmall++; 619 return; 620 } 621 ip = mtod(m, struct ip *); 622 } 623 624 if (ip_mrouter) { 625 /* 626 * If we are acting as a multicast router, all 627 * incoming multicast packets are passed to the 628 * kernel-level multicast forwarding function. 629 * The packet is returned (relatively) intact; if 630 * ip_mforward() returns a non-zero value, the packet 631 * must be discarded, else it may be accepted below. 632 * 633 * (The IP ident field is put in the same byte order 634 * as expected when ip_mforward() is called from 635 * ip_output().) 636 */ 637 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 638 ipstat.ips_cantforward++; 639 m_freem(m); 640 return; 641 } 642 643 /* 644 * The process-level routing demon needs to receive 645 * all multicast IGMP packets, whether or not this 646 * host belongs to their destination groups. 647 */ 648 if (ip->ip_p == IPPROTO_IGMP) 649 goto ours; 650 ipstat.ips_forward++; 651 } 652#endif 653 /* 654 * See if we belong to the destination multicast group on the 655 * arrival interface. 656 */ 657 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 658 if (inm == NULL) { 659 ipstat.ips_cantforward++; 660 m_freem(m); 661 return; 662 } 663 goto ours; 664 } 665 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 666 in_nullhost(ip->ip_dst)) 667 goto ours; 668 669 /* 670 * Not for us; forward if possible and desirable. 671 */ 672 if (ipforwarding == 0) { 673 ipstat.ips_cantforward++; 674 m_freem(m); 675 } else { 676 /* 677 * If ip_dst matched any of my address on !IFF_UP interface, 678 * and there's no IFF_UP interface that matches ip_dst, 679 * send icmp unreach. Forwarding it will result in in-kernel 680 * forwarding loop till TTL goes to 0. 681 */ 682 if (downmatch) { 683 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 684 ipstat.ips_cantforward++; 685 return; 686 } 687 ip_forward(m, 0); 688 } 689 return; 690 691ours: 692 /* 693 * If offset or IP_MF are set, must reassemble. 694 * Otherwise, nothing need be done. 695 * (We could look in the reassembly queue to see 696 * if the packet was previously fragmented, 697 * but it's not worth the time; just let them time out.) 698 */ 699 if (ip->ip_off & ~(IP_DF|IP_RF)) { 700 /* 701 * Look for queue of fragments 702 * of this datagram. 703 */ 704 IPQ_LOCK(); 705 LIST_FOREACH(fp, &ipq, ipq_q) 706 if (ip->ip_id == fp->ipq_id && 707 in_hosteq(ip->ip_src, fp->ipq_src) && 708 in_hosteq(ip->ip_dst, fp->ipq_dst) && 709 ip->ip_p == fp->ipq_p) 710 goto found; 711 fp = 0; 712found: 713 714 /* 715 * Adjust ip_len to not reflect header, 716 * set ipqe_mff if more fragments are expected, 717 * convert offset of this to bytes. 718 */ 719 ip->ip_len -= hlen; 720 mff = (ip->ip_off & IP_MF) != 0; 721 if (mff) { 722 /* 723 * Make sure that fragments have a data length 724 * that's a non-zero multiple of 8 bytes. 725 */ 726 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 727 ipstat.ips_badfrags++; 728 IPQ_UNLOCK(); 729 goto bad; 730 } 731 } 732 ip->ip_off <<= 3; 733 734 /* 735 * If datagram marked as having more fragments 736 * or if this is not the first fragment, 737 * attempt reassembly; if it succeeds, proceed. 738 */ 739 if (mff || ip->ip_off) { 740 ipstat.ips_fragments++; 741 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 742 if (ipqe == NULL) { 743 ipstat.ips_rcvmemdrop++; 744 IPQ_UNLOCK(); 745 goto bad; 746 } 747 ipqe->ipqe_mff = mff; 748 ipqe->ipqe_m = m; 749 ipqe->ipqe_ip = ip; 750 m = ip_reass(ipqe, fp); 751 if (m == 0) { 752 IPQ_UNLOCK(); 753 return; 754 } 755 ipstat.ips_reassembled++; 756 ip = mtod(m, struct ip *); 757 hlen = ip->ip_hl << 2; 758 ip->ip_len += hlen; 759 } else 760 if (fp) 761 ip_freef(fp); 762 IPQ_UNLOCK(); 763 } 764 765#ifdef IPSEC 766 /* 767 * enforce IPsec policy checking if we are seeing last header. 768 * note that we do not visit this with protocols with pcb layer 769 * code - like udp/tcp/raw ip. 770 */ 771 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 772 ipsec4_in_reject(m, NULL)) { 773 ipsecstat.in_polvio++; 774 goto bad; 775 } 776#endif 777 778 /* 779 * Switch out to protocol's input routine. 780 */ 781#if IFA_STATS 782 if (ia && ip) 783 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 784#endif 785 ipstat.ips_delivered++; 786 { 787 int off = hlen, nh = ip->ip_p; 788 789 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 790 return; 791 } 792bad: 793 m_freem(m); 794 return; 795 796badcsum: 797 ipstat.ips_badsum++; 798 m_freem(m); 799} 800 801/* 802 * Take incoming datagram fragment and try to 803 * reassemble it into whole datagram. If a chain for 804 * reassembly of this datagram already exists, then it 805 * is given as fp; otherwise have to make a chain. 806 */ 807struct mbuf * 808ip_reass(ipqe, fp) 809 struct ipqent *ipqe; 810 struct ipq *fp; 811{ 812 struct mbuf *m = ipqe->ipqe_m; 813 struct ipqent *nq, *p, *q; 814 struct ip *ip; 815 struct mbuf *t; 816 int hlen = ipqe->ipqe_ip->ip_hl << 2; 817 int i, next; 818 819 IPQ_LOCK_CHECK(); 820 821 /* 822 * Presence of header sizes in mbufs 823 * would confuse code below. 824 */ 825 m->m_data += hlen; 826 m->m_len -= hlen; 827 828 /* 829 * If first fragment to arrive, create a reassembly queue. 830 */ 831 if (fp == 0) { 832 /* 833 * Enforce upper bound on number of fragmented packets 834 * for which we attempt reassembly; 835 * If maxfrag is 0, never accept fragments. 836 * If maxfrag is -1, accept all fragments without limitation. 837 */ 838 if (ip_maxfragpackets < 0) 839 ; 840 else if (ip_nfragpackets >= ip_maxfragpackets) 841 goto dropfrag; 842 ip_nfragpackets++; 843 MALLOC(fp, struct ipq *, sizeof (struct ipq), 844 M_FTABLE, M_NOWAIT); 845 if (fp == NULL) 846 goto dropfrag; 847 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 848 fp->ipq_ttl = IPFRAGTTL; 849 fp->ipq_p = ipqe->ipqe_ip->ip_p; 850 fp->ipq_id = ipqe->ipqe_ip->ip_id; 851 LIST_INIT(&fp->ipq_fragq); 852 fp->ipq_src = ipqe->ipqe_ip->ip_src; 853 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 854 p = NULL; 855 goto insert; 856 } 857 858 /* 859 * Find a segment which begins after this one does. 860 */ 861 for (p = NULL, q = LIST_FIRST(&fp->ipq_fragq); q != NULL; 862 p = q, q = LIST_NEXT(q, ipqe_q)) 863 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 864 break; 865 866 /* 867 * If there is a preceding segment, it may provide some of 868 * our data already. If so, drop the data from the incoming 869 * segment. If it provides all of our data, drop us. 870 */ 871 if (p != NULL) { 872 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 873 ipqe->ipqe_ip->ip_off; 874 if (i > 0) { 875 if (i >= ipqe->ipqe_ip->ip_len) 876 goto dropfrag; 877 m_adj(ipqe->ipqe_m, i); 878 ipqe->ipqe_ip->ip_off += i; 879 ipqe->ipqe_ip->ip_len -= i; 880 } 881 } 882 883 /* 884 * While we overlap succeeding segments trim them or, 885 * if they are completely covered, dequeue them. 886 */ 887 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 888 q->ipqe_ip->ip_off; q = nq) { 889 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 890 q->ipqe_ip->ip_off; 891 if (i < q->ipqe_ip->ip_len) { 892 q->ipqe_ip->ip_len -= i; 893 q->ipqe_ip->ip_off += i; 894 m_adj(q->ipqe_m, i); 895 break; 896 } 897 nq = LIST_NEXT(q, ipqe_q); 898 m_freem(q->ipqe_m); 899 LIST_REMOVE(q, ipqe_q); 900 pool_put(&ipqent_pool, q); 901 } 902 903insert: 904 /* 905 * Stick new segment in its place; 906 * check for complete reassembly. 907 */ 908 if (p == NULL) { 909 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 910 } else { 911 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 912 } 913 next = 0; 914 for (p = NULL, q = LIST_FIRST(&fp->ipq_fragq); q != NULL; 915 p = q, q = LIST_NEXT(q, ipqe_q)) { 916 if (q->ipqe_ip->ip_off != next) 917 return (0); 918 next += q->ipqe_ip->ip_len; 919 } 920 if (p->ipqe_mff) 921 return (0); 922 923 /* 924 * Reassembly is complete. Check for a bogus message size and 925 * concatenate fragments. 926 */ 927 q = LIST_FIRST(&fp->ipq_fragq); 928 ip = q->ipqe_ip; 929 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 930 ipstat.ips_toolong++; 931 ip_freef(fp); 932 return (0); 933 } 934 m = q->ipqe_m; 935 t = m->m_next; 936 m->m_next = 0; 937 m_cat(m, t); 938 nq = LIST_NEXT(q, ipqe_q); 939 pool_put(&ipqent_pool, q); 940 for (q = nq; q != NULL; q = nq) { 941 t = q->ipqe_m; 942 nq = LIST_NEXT(q, ipqe_q); 943 pool_put(&ipqent_pool, q); 944 m_cat(m, t); 945 } 946 947 /* 948 * Create header for new ip packet by 949 * modifying header of first packet; 950 * dequeue and discard fragment reassembly header. 951 * Make header visible. 952 */ 953 ip->ip_len = next; 954 ip->ip_src = fp->ipq_src; 955 ip->ip_dst = fp->ipq_dst; 956 LIST_REMOVE(fp, ipq_q); 957 FREE(fp, M_FTABLE); 958 ip_nfragpackets--; 959 m->m_len += (ip->ip_hl << 2); 960 m->m_data -= (ip->ip_hl << 2); 961 /* some debugging cruft by sklower, below, will go away soon */ 962 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 963 int plen = 0; 964 for (t = m; t; t = t->m_next) 965 plen += t->m_len; 966 m->m_pkthdr.len = plen; 967 } 968 return (m); 969 970dropfrag: 971 ipstat.ips_fragdropped++; 972 m_freem(m); 973 pool_put(&ipqent_pool, ipqe); 974 return (0); 975} 976 977/* 978 * Free a fragment reassembly header and all 979 * associated datagrams. 980 */ 981void 982ip_freef(fp) 983 struct ipq *fp; 984{ 985 struct ipqent *q, *p; 986 987 IPQ_LOCK_CHECK(); 988 989 for (q = LIST_FIRST(&fp->ipq_fragq); q != NULL; q = p) { 990 p = LIST_NEXT(q, ipqe_q); 991 m_freem(q->ipqe_m); 992 LIST_REMOVE(q, ipqe_q); 993 pool_put(&ipqent_pool, q); 994 } 995 LIST_REMOVE(fp, ipq_q); 996 FREE(fp, M_FTABLE); 997 ip_nfragpackets--; 998} 999 1000/* 1001 * IP timer processing; 1002 * if a timer expires on a reassembly 1003 * queue, discard it. 1004 */ 1005void 1006ip_slowtimo() 1007{ 1008 struct ipq *fp, *nfp; 1009 int s = splsoftnet(); 1010 1011 IPQ_LOCK(); 1012 for (fp = LIST_FIRST(&ipq); fp != NULL; fp = nfp) { 1013 nfp = LIST_NEXT(fp, ipq_q); 1014 if (--fp->ipq_ttl == 0) { 1015 ipstat.ips_fragtimeout++; 1016 ip_freef(fp); 1017 } 1018 } 1019 /* 1020 * If we are over the maximum number of fragments 1021 * (due to the limit being lowered), drain off 1022 * enough to get down to the new limit. 1023 */ 1024 if (ip_maxfragpackets < 0) 1025 ; 1026 else { 1027 while (ip_nfragpackets > ip_maxfragpackets && LIST_FIRST(&ipq)) 1028 ip_freef(LIST_FIRST(&ipq)); 1029 } 1030 IPQ_UNLOCK(); 1031#ifdef GATEWAY 1032 ipflow_slowtimo(); 1033#endif 1034 splx(s); 1035} 1036 1037/* 1038 * Drain off all datagram fragments. 1039 */ 1040void 1041ip_drain() 1042{ 1043 1044 /* 1045 * We may be called from a device's interrupt context. If 1046 * the ipq is already busy, just bail out now. 1047 */ 1048 if (ipq_lock_try() == 0) 1049 return; 1050 1051 while (LIST_FIRST(&ipq) != NULL) { 1052 ipstat.ips_fragdropped++; 1053 ip_freef(LIST_FIRST(&ipq)); 1054 } 1055 1056 IPQ_UNLOCK(); 1057} 1058 1059/* 1060 * Do option processing on a datagram, 1061 * possibly discarding it if bad options are encountered, 1062 * or forwarding it if source-routed. 1063 * Returns 1 if packet has been forwarded/freed, 1064 * 0 if the packet should be processed further. 1065 */ 1066int 1067ip_dooptions(m) 1068 struct mbuf *m; 1069{ 1070 struct ip *ip = mtod(m, struct ip *); 1071 u_char *cp, *cp0; 1072 struct ip_timestamp *ipt; 1073 struct in_ifaddr *ia; 1074 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1075 struct in_addr dst; 1076 n_time ntime; 1077 1078 dst = ip->ip_dst; 1079 cp = (u_char *)(ip + 1); 1080 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1081 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1082 opt = cp[IPOPT_OPTVAL]; 1083 if (opt == IPOPT_EOL) 1084 break; 1085 if (opt == IPOPT_NOP) 1086 optlen = 1; 1087 else { 1088 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1089 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1090 goto bad; 1091 } 1092 optlen = cp[IPOPT_OLEN]; 1093 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1094 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1095 goto bad; 1096 } 1097 } 1098 switch (opt) { 1099 1100 default: 1101 break; 1102 1103 /* 1104 * Source routing with record. 1105 * Find interface with current destination address. 1106 * If none on this machine then drop if strictly routed, 1107 * or do nothing if loosely routed. 1108 * Record interface address and bring up next address 1109 * component. If strictly routed make sure next 1110 * address is on directly accessible net. 1111 */ 1112 case IPOPT_LSRR: 1113 case IPOPT_SSRR: 1114 if (ip_allowsrcrt == 0) { 1115 type = ICMP_UNREACH; 1116 code = ICMP_UNREACH_NET_PROHIB; 1117 goto bad; 1118 } 1119 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1120 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1121 goto bad; 1122 } 1123 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1124 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1125 goto bad; 1126 } 1127 ipaddr.sin_addr = ip->ip_dst; 1128 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1129 if (ia == 0) { 1130 if (opt == IPOPT_SSRR) { 1131 type = ICMP_UNREACH; 1132 code = ICMP_UNREACH_SRCFAIL; 1133 goto bad; 1134 } 1135 /* 1136 * Loose routing, and not at next destination 1137 * yet; nothing to do except forward. 1138 */ 1139 break; 1140 } 1141 off--; /* 0 origin */ 1142 if ((off + sizeof(struct in_addr)) > optlen) { 1143 /* 1144 * End of source route. Should be for us. 1145 */ 1146 save_rte(cp, ip->ip_src); 1147 break; 1148 } 1149 /* 1150 * locate outgoing interface 1151 */ 1152 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1153 sizeof(ipaddr.sin_addr)); 1154 if (opt == IPOPT_SSRR) 1155 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1156 else 1157 ia = ip_rtaddr(ipaddr.sin_addr); 1158 if (ia == 0) { 1159 type = ICMP_UNREACH; 1160 code = ICMP_UNREACH_SRCFAIL; 1161 goto bad; 1162 } 1163 ip->ip_dst = ipaddr.sin_addr; 1164 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1165 (caddr_t)(cp + off), sizeof(struct in_addr)); 1166 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1167 /* 1168 * Let ip_intr's mcast routing check handle mcast pkts 1169 */ 1170 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1171 break; 1172 1173 case IPOPT_RR: 1174 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1175 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1176 goto bad; 1177 } 1178 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1179 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1180 goto bad; 1181 } 1182 /* 1183 * If no space remains, ignore. 1184 */ 1185 off--; /* 0 origin */ 1186 if ((off + sizeof(struct in_addr)) > optlen) 1187 break; 1188 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1189 sizeof(ipaddr.sin_addr)); 1190 /* 1191 * locate outgoing interface; if we're the destination, 1192 * use the incoming interface (should be same). 1193 */ 1194 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1195 == NULL && 1196 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1197 type = ICMP_UNREACH; 1198 code = ICMP_UNREACH_HOST; 1199 goto bad; 1200 } 1201 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1202 (caddr_t)(cp + off), sizeof(struct in_addr)); 1203 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1204 break; 1205 1206 case IPOPT_TS: 1207 code = cp - (u_char *)ip; 1208 ipt = (struct ip_timestamp *)cp; 1209 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1210 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1211 goto bad; 1212 } 1213 if (ipt->ipt_ptr < 5) { 1214 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1215 goto bad; 1216 } 1217 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1218 if (++ipt->ipt_oflw == 0) { 1219 code = (u_char *)&ipt->ipt_ptr - 1220 (u_char *)ip; 1221 goto bad; 1222 } 1223 break; 1224 } 1225 cp0 = (cp + ipt->ipt_ptr - 1); 1226 switch (ipt->ipt_flg) { 1227 1228 case IPOPT_TS_TSONLY: 1229 break; 1230 1231 case IPOPT_TS_TSANDADDR: 1232 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1233 sizeof(struct in_addr) > ipt->ipt_len) { 1234 code = (u_char *)&ipt->ipt_ptr - 1235 (u_char *)ip; 1236 goto bad; 1237 } 1238 ipaddr.sin_addr = dst; 1239 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1240 m->m_pkthdr.rcvif)); 1241 if (ia == 0) 1242 continue; 1243 bcopy(&ia->ia_addr.sin_addr, 1244 cp0, sizeof(struct in_addr)); 1245 ipt->ipt_ptr += sizeof(struct in_addr); 1246 break; 1247 1248 case IPOPT_TS_PRESPEC: 1249 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1250 sizeof(struct in_addr) > ipt->ipt_len) { 1251 code = (u_char *)&ipt->ipt_ptr - 1252 (u_char *)ip; 1253 goto bad; 1254 } 1255 bcopy(cp0, &ipaddr.sin_addr, 1256 sizeof(struct in_addr)); 1257 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1258 == NULL) 1259 continue; 1260 ipt->ipt_ptr += sizeof(struct in_addr); 1261 break; 1262 1263 default: 1264 /* XXX can't take &ipt->ipt_flg */ 1265 code = (u_char *)&ipt->ipt_ptr - 1266 (u_char *)ip + 1; 1267 goto bad; 1268 } 1269 ntime = iptime(); 1270 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1271 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1272 sizeof(n_time)); 1273 ipt->ipt_ptr += sizeof(n_time); 1274 } 1275 } 1276 if (forward) { 1277 if (ip_forwsrcrt == 0) { 1278 type = ICMP_UNREACH; 1279 code = ICMP_UNREACH_SRCFAIL; 1280 goto bad; 1281 } 1282 ip_forward(m, 1); 1283 return (1); 1284 } 1285 return (0); 1286bad: 1287 icmp_error(m, type, code, 0, 0); 1288 ipstat.ips_badoptions++; 1289 return (1); 1290} 1291 1292/* 1293 * Given address of next destination (final or next hop), 1294 * return internet address info of interface to be used to get there. 1295 */ 1296struct in_ifaddr * 1297ip_rtaddr(dst) 1298 struct in_addr dst; 1299{ 1300 struct sockaddr_in *sin; 1301 1302 sin = satosin(&ipforward_rt.ro_dst); 1303 1304 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1305 if (ipforward_rt.ro_rt) { 1306 RTFREE(ipforward_rt.ro_rt); 1307 ipforward_rt.ro_rt = 0; 1308 } 1309 sin->sin_family = AF_INET; 1310 sin->sin_len = sizeof(*sin); 1311 sin->sin_addr = dst; 1312 1313 rtalloc(&ipforward_rt); 1314 } 1315 if (ipforward_rt.ro_rt == 0) 1316 return ((struct in_ifaddr *)0); 1317 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1318} 1319 1320/* 1321 * Save incoming source route for use in replies, 1322 * to be picked up later by ip_srcroute if the receiver is interested. 1323 */ 1324void 1325save_rte(option, dst) 1326 u_char *option; 1327 struct in_addr dst; 1328{ 1329 unsigned olen; 1330 1331 olen = option[IPOPT_OLEN]; 1332#ifdef DIAGNOSTIC 1333 if (ipprintfs) 1334 printf("save_rte: olen %d\n", olen); 1335#endif /* 0 */ 1336 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1337 return; 1338 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1339 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1340 ip_srcrt.dst = dst; 1341} 1342 1343/* 1344 * Retrieve incoming source route for use in replies, 1345 * in the same form used by setsockopt. 1346 * The first hop is placed before the options, will be removed later. 1347 */ 1348struct mbuf * 1349ip_srcroute() 1350{ 1351 struct in_addr *p, *q; 1352 struct mbuf *m; 1353 1354 if (ip_nhops == 0) 1355 return ((struct mbuf *)0); 1356 m = m_get(M_DONTWAIT, MT_SOOPTS); 1357 if (m == 0) 1358 return ((struct mbuf *)0); 1359 1360#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1361 1362 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1363 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1364 OPTSIZ; 1365#ifdef DIAGNOSTIC 1366 if (ipprintfs) 1367 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1368#endif 1369 1370 /* 1371 * First save first hop for return route 1372 */ 1373 p = &ip_srcrt.route[ip_nhops - 1]; 1374 *(mtod(m, struct in_addr *)) = *p--; 1375#ifdef DIAGNOSTIC 1376 if (ipprintfs) 1377 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1378#endif 1379 1380 /* 1381 * Copy option fields and padding (nop) to mbuf. 1382 */ 1383 ip_srcrt.nop = IPOPT_NOP; 1384 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1385 bcopy((caddr_t)&ip_srcrt.nop, 1386 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1387 q = (struct in_addr *)(mtod(m, caddr_t) + 1388 sizeof(struct in_addr) + OPTSIZ); 1389#undef OPTSIZ 1390 /* 1391 * Record return path as an IP source route, 1392 * reversing the path (pointers are now aligned). 1393 */ 1394 while (p >= ip_srcrt.route) { 1395#ifdef DIAGNOSTIC 1396 if (ipprintfs) 1397 printf(" %x", ntohl(q->s_addr)); 1398#endif 1399 *q++ = *p--; 1400 } 1401 /* 1402 * Last hop goes to final destination. 1403 */ 1404 *q = ip_srcrt.dst; 1405#ifdef DIAGNOSTIC 1406 if (ipprintfs) 1407 printf(" %x\n", ntohl(q->s_addr)); 1408#endif 1409 return (m); 1410} 1411 1412/* 1413 * Strip out IP options, at higher 1414 * level protocol in the kernel. 1415 * Second argument is buffer to which options 1416 * will be moved, and return value is their length. 1417 * XXX should be deleted; last arg currently ignored. 1418 */ 1419void 1420ip_stripoptions(m, mopt) 1421 struct mbuf *m; 1422 struct mbuf *mopt; 1423{ 1424 int i; 1425 struct ip *ip = mtod(m, struct ip *); 1426 caddr_t opts; 1427 int olen; 1428 1429 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1430 opts = (caddr_t)(ip + 1); 1431 i = m->m_len - (sizeof (struct ip) + olen); 1432 bcopy(opts + olen, opts, (unsigned)i); 1433 m->m_len -= olen; 1434 if (m->m_flags & M_PKTHDR) 1435 m->m_pkthdr.len -= olen; 1436 ip->ip_len -= olen; 1437 ip->ip_hl = sizeof (struct ip) >> 2; 1438} 1439 1440const int inetctlerrmap[PRC_NCMDS] = { 1441 0, 0, 0, 0, 1442 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1443 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1444 EMSGSIZE, EHOSTUNREACH, 0, 0, 1445 0, 0, 0, 0, 1446 ENOPROTOOPT 1447}; 1448 1449/* 1450 * Forward a packet. If some error occurs return the sender 1451 * an icmp packet. Note we can't always generate a meaningful 1452 * icmp message because icmp doesn't have a large enough repertoire 1453 * of codes and types. 1454 * 1455 * If not forwarding, just drop the packet. This could be confusing 1456 * if ipforwarding was zero but some routing protocol was advancing 1457 * us as a gateway to somewhere. However, we must let the routing 1458 * protocol deal with that. 1459 * 1460 * The srcrt parameter indicates whether the packet is being forwarded 1461 * via a source route. 1462 */ 1463void 1464ip_forward(m, srcrt) 1465 struct mbuf *m; 1466 int srcrt; 1467{ 1468 struct ip *ip = mtod(m, struct ip *); 1469 struct sockaddr_in *sin; 1470 struct rtentry *rt; 1471 int error, type = 0, code = 0; 1472 struct mbuf *mcopy; 1473 n_long dest; 1474 struct ifnet *destifp; 1475#ifdef IPSEC 1476 struct ifnet dummyifp; 1477#endif 1478 1479 /* 1480 * Clear any in-bound checksum flags for this packet. 1481 */ 1482 m->m_pkthdr.csum_flags = 0; 1483 1484 dest = 0; 1485#ifdef DIAGNOSTIC 1486 if (ipprintfs) 1487 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1488 ntohl(ip->ip_src.s_addr), 1489 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1490#endif 1491 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1492 ipstat.ips_cantforward++; 1493 m_freem(m); 1494 return; 1495 } 1496 if (ip->ip_ttl <= IPTTLDEC) { 1497 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1498 return; 1499 } 1500 ip->ip_ttl -= IPTTLDEC; 1501 1502 sin = satosin(&ipforward_rt.ro_dst); 1503 if ((rt = ipforward_rt.ro_rt) == 0 || 1504 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1505 if (ipforward_rt.ro_rt) { 1506 RTFREE(ipforward_rt.ro_rt); 1507 ipforward_rt.ro_rt = 0; 1508 } 1509 sin->sin_family = AF_INET; 1510 sin->sin_len = sizeof(struct sockaddr_in); 1511 sin->sin_addr = ip->ip_dst; 1512 1513 rtalloc(&ipforward_rt); 1514 if (ipforward_rt.ro_rt == 0) { 1515 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1516 return; 1517 } 1518 rt = ipforward_rt.ro_rt; 1519 } 1520 1521 /* 1522 * Save at most 68 bytes of the packet in case 1523 * we need to generate an ICMP message to the src. 1524 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1525 */ 1526 mcopy = m_copym(m, 0, imin((int)ip->ip_len, 68), M_DONTWAIT); 1527 if (mcopy) 1528 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1529 1530 /* 1531 * If forwarding packet using same interface that it came in on, 1532 * perhaps should send a redirect to sender to shortcut a hop. 1533 * Only send redirect if source is sending directly to us, 1534 * and if packet was not source routed (or has any options). 1535 * Also, don't send redirect if forwarding using a default route 1536 * or a route modified by a redirect. 1537 */ 1538 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1539 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1540 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1541 ipsendredirects && !srcrt) { 1542 if (rt->rt_ifa && 1543 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1544 ifatoia(rt->rt_ifa)->ia_subnet) { 1545 if (rt->rt_flags & RTF_GATEWAY) 1546 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1547 else 1548 dest = ip->ip_dst.s_addr; 1549 /* 1550 * Router requirements says to only send host 1551 * redirects. 1552 */ 1553 type = ICMP_REDIRECT; 1554 code = ICMP_REDIRECT_HOST; 1555#ifdef DIAGNOSTIC 1556 if (ipprintfs) 1557 printf("redirect (%d) to %x\n", code, 1558 (u_int32_t)dest); 1559#endif 1560 } 1561 } 1562 1563#ifdef IPSEC 1564 /* Don't lookup socket in forwarding case */ 1565 (void)ipsec_setsocket(m, NULL); 1566#endif 1567 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1568 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1569 if (error) 1570 ipstat.ips_cantforward++; 1571 else { 1572 ipstat.ips_forward++; 1573 if (type) 1574 ipstat.ips_redirectsent++; 1575 else { 1576 if (mcopy) { 1577#ifdef GATEWAY 1578 if (mcopy->m_flags & M_CANFASTFWD) 1579 ipflow_create(&ipforward_rt, mcopy); 1580#endif 1581 m_freem(mcopy); 1582 } 1583 return; 1584 } 1585 } 1586 if (mcopy == NULL) 1587 return; 1588 destifp = NULL; 1589 1590 switch (error) { 1591 1592 case 0: /* forwarded, but need redirect */ 1593 /* type, code set above */ 1594 break; 1595 1596 case ENETUNREACH: /* shouldn't happen, checked above */ 1597 case EHOSTUNREACH: 1598 case ENETDOWN: 1599 case EHOSTDOWN: 1600 default: 1601 type = ICMP_UNREACH; 1602 code = ICMP_UNREACH_HOST; 1603 break; 1604 1605 case EMSGSIZE: 1606 type = ICMP_UNREACH; 1607 code = ICMP_UNREACH_NEEDFRAG; 1608#ifndef IPSEC 1609 if (ipforward_rt.ro_rt) 1610 destifp = ipforward_rt.ro_rt->rt_ifp; 1611#else 1612 /* 1613 * If the packet is routed over IPsec tunnel, tell the 1614 * originator the tunnel MTU. 1615 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1616 * XXX quickhack!!! 1617 */ 1618 if (ipforward_rt.ro_rt) { 1619 struct secpolicy *sp; 1620 int ipsecerror; 1621 size_t ipsechdr; 1622 struct route *ro; 1623 1624 sp = ipsec4_getpolicybyaddr(mcopy, 1625 IPSEC_DIR_OUTBOUND, 1626 IP_FORWARDING, 1627 &ipsecerror); 1628 1629 if (sp == NULL) 1630 destifp = ipforward_rt.ro_rt->rt_ifp; 1631 else { 1632 /* count IPsec header size */ 1633 ipsechdr = ipsec4_hdrsiz(mcopy, 1634 IPSEC_DIR_OUTBOUND, 1635 NULL); 1636 1637 /* 1638 * find the correct route for outer IPv4 1639 * header, compute tunnel MTU. 1640 * 1641 * XXX BUG ALERT 1642 * The "dummyifp" code relies upon the fact 1643 * that icmp_error() touches only ifp->if_mtu. 1644 */ 1645 /*XXX*/ 1646 destifp = NULL; 1647 if (sp->req != NULL 1648 && sp->req->sav != NULL 1649 && sp->req->sav->sah != NULL) { 1650 ro = &sp->req->sav->sah->sa_route; 1651 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1652 dummyifp.if_mtu = 1653 ro->ro_rt->rt_ifp->if_mtu; 1654 dummyifp.if_mtu -= ipsechdr; 1655 destifp = &dummyifp; 1656 } 1657 } 1658 1659 key_freesp(sp); 1660 } 1661 } 1662#endif /*IPSEC*/ 1663 ipstat.ips_cantfrag++; 1664 break; 1665 1666 case ENOBUFS: 1667 type = ICMP_SOURCEQUENCH; 1668 code = 0; 1669 break; 1670 } 1671 icmp_error(mcopy, type, code, dest, destifp); 1672} 1673 1674void 1675ip_savecontrol(inp, mp, ip, m) 1676 struct inpcb *inp; 1677 struct mbuf **mp; 1678 struct ip *ip; 1679 struct mbuf *m; 1680{ 1681 1682 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1683 struct timeval tv; 1684 1685 microtime(&tv); 1686 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1687 SCM_TIMESTAMP, SOL_SOCKET); 1688 if (*mp) 1689 mp = &(*mp)->m_next; 1690 } 1691 if (inp->inp_flags & INP_RECVDSTADDR) { 1692 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1693 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1694 if (*mp) 1695 mp = &(*mp)->m_next; 1696 } 1697#ifdef notyet 1698 /* 1699 * XXX 1700 * Moving these out of udp_input() made them even more broken 1701 * than they already were. 1702 * - fenner@parc.xerox.com 1703 */ 1704 /* options were tossed already */ 1705 if (inp->inp_flags & INP_RECVOPTS) { 1706 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1707 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1708 if (*mp) 1709 mp = &(*mp)->m_next; 1710 } 1711 /* ip_srcroute doesn't do what we want here, need to fix */ 1712 if (inp->inp_flags & INP_RECVRETOPTS) { 1713 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1714 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1715 if (*mp) 1716 mp = &(*mp)->m_next; 1717 } 1718#endif 1719 if (inp->inp_flags & INP_RECVIF) { 1720 struct sockaddr_dl sdl; 1721 1722 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1723 sdl.sdl_family = AF_LINK; 1724 sdl.sdl_index = m->m_pkthdr.rcvif ? 1725 m->m_pkthdr.rcvif->if_index : 0; 1726 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1727 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1728 IP_RECVIF, IPPROTO_IP); 1729 if (*mp) 1730 mp = &(*mp)->m_next; 1731 } 1732} 1733 1734int 1735ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1736 int *name; 1737 u_int namelen; 1738 void *oldp; 1739 size_t *oldlenp; 1740 void *newp; 1741 size_t newlen; 1742{ 1743 extern int subnetsarelocal, hostzeroisbroadcast; 1744 1745 int error, old; 1746 1747 /* All sysctl names at this level are terminal. */ 1748 if (namelen != 1) 1749 return (ENOTDIR); 1750 1751 switch (name[0]) { 1752 case IPCTL_FORWARDING: 1753 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1754 case IPCTL_SENDREDIRECTS: 1755 return (sysctl_int(oldp, oldlenp, newp, newlen, 1756 &ipsendredirects)); 1757 case IPCTL_DEFTTL: 1758 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1759#ifdef notyet 1760 case IPCTL_DEFMTU: 1761 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1762#endif 1763 case IPCTL_FORWSRCRT: 1764 /* Don't allow this to change in a secure environment. */ 1765 if (securelevel > 0) 1766 return (sysctl_rdint(oldp, oldlenp, newp, 1767 ip_forwsrcrt)); 1768 else 1769 return (sysctl_int(oldp, oldlenp, newp, newlen, 1770 &ip_forwsrcrt)); 1771 case IPCTL_DIRECTEDBCAST: 1772 return (sysctl_int(oldp, oldlenp, newp, newlen, 1773 &ip_directedbcast)); 1774 case IPCTL_ALLOWSRCRT: 1775 return (sysctl_int(oldp, oldlenp, newp, newlen, 1776 &ip_allowsrcrt)); 1777 case IPCTL_SUBNETSARELOCAL: 1778 return (sysctl_int(oldp, oldlenp, newp, newlen, 1779 &subnetsarelocal)); 1780 case IPCTL_MTUDISC: 1781 error = sysctl_int(oldp, oldlenp, newp, newlen, 1782 &ip_mtudisc); 1783 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1784 ip_mtudisc_timeout_q = 1785 rt_timer_queue_create(ip_mtudisc_timeout); 1786 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1787 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1788 ip_mtudisc_timeout_q = NULL; 1789 } 1790 return error; 1791 case IPCTL_ANONPORTMIN: 1792 old = anonportmin; 1793 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1794 if (anonportmin >= anonportmax || anonportmin < 0 1795 || anonportmin > 65535 1796#ifndef IPNOPRIVPORTS 1797 || anonportmin < IPPORT_RESERVED 1798#endif 1799 ) { 1800 anonportmin = old; 1801 return (EINVAL); 1802 } 1803 return (error); 1804 case IPCTL_ANONPORTMAX: 1805 old = anonportmax; 1806 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1807 if (anonportmin >= anonportmax || anonportmax < 0 1808 || anonportmax > 65535 1809#ifndef IPNOPRIVPORTS 1810 || anonportmax < IPPORT_RESERVED 1811#endif 1812 ) { 1813 anonportmax = old; 1814 return (EINVAL); 1815 } 1816 return (error); 1817 case IPCTL_MTUDISCTIMEOUT: 1818 error = sysctl_int(oldp, oldlenp, newp, newlen, 1819 &ip_mtudisc_timeout); 1820 if (ip_mtudisc_timeout_q != NULL) 1821 rt_timer_queue_change(ip_mtudisc_timeout_q, 1822 ip_mtudisc_timeout); 1823 return (error); 1824#ifdef GATEWAY 1825 case IPCTL_MAXFLOWS: 1826 { 1827 int s; 1828 1829 error = sysctl_int(oldp, oldlenp, newp, newlen, 1830 &ip_maxflows); 1831 s = splsoftnet(); 1832 ipflow_reap(0); 1833 splx(s); 1834 return (error); 1835 } 1836#endif 1837 case IPCTL_HOSTZEROBROADCAST: 1838 return (sysctl_int(oldp, oldlenp, newp, newlen, 1839 &hostzeroisbroadcast)); 1840#if NGIF > 0 1841 case IPCTL_GIF_TTL: 1842 return(sysctl_int(oldp, oldlenp, newp, newlen, 1843 &ip_gif_ttl)); 1844#endif 1845 1846#ifndef IPNOPRIVPORTS 1847 case IPCTL_LOWPORTMIN: 1848 old = lowportmin; 1849 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmin); 1850 if (lowportmin >= lowportmax 1851 || lowportmin > IPPORT_RESERVEDMAX 1852 || lowportmin < IPPORT_RESERVEDMIN 1853 ) { 1854 lowportmin = old; 1855 return (EINVAL); 1856 } 1857 return (error); 1858 case IPCTL_LOWPORTMAX: 1859 old = lowportmax; 1860 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmax); 1861 if (lowportmin >= lowportmax 1862 || lowportmax > IPPORT_RESERVEDMAX 1863 || lowportmax < IPPORT_RESERVEDMIN 1864 ) { 1865 lowportmax = old; 1866 return (EINVAL); 1867 } 1868 return (error); 1869#endif 1870 1871 case IPCTL_MAXFRAGPACKETS: 1872 return (sysctl_int(oldp, oldlenp, newp, newlen, 1873 &ip_maxfragpackets)); 1874 1875 default: 1876 return (EOPNOTSUPP); 1877 } 1878 /* NOTREACHED */ 1879} 1880