ip_input.c revision 1.138
1/* $NetBSD: ip_input.c,v 1.138 2001/10/29 07:02:33 simonb 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 "opt_gateway.h" 105#include "opt_pfil_hooks.h" 106#include "opt_ipsec.h" 107#include "opt_mrouting.h" 108#include "opt_inet_csum.h" 109 110#include <sys/param.h> 111#include <sys/systm.h> 112#include <sys/malloc.h> 113#include <sys/mbuf.h> 114#include <sys/domain.h> 115#include <sys/protosw.h> 116#include <sys/socket.h> 117#include <sys/socketvar.h> 118#include <sys/errno.h> 119#include <sys/time.h> 120#include <sys/kernel.h> 121#include <sys/pool.h> 122#include <sys/sysctl.h> 123 124#include <net/if.h> 125#include <net/if_dl.h> 126#include <net/route.h> 127#include <net/pfil.h> 128 129#include <netinet/in.h> 130#include <netinet/in_systm.h> 131#include <netinet/ip.h> 132#include <netinet/in_pcb.h> 133#include <netinet/in_var.h> 134#include <netinet/ip_var.h> 135#include <netinet/ip_icmp.h> 136/* just for gif_ttl */ 137#include <netinet/in_gif.h> 138#include "gif.h" 139 140#ifdef MROUTING 141#include <netinet/ip_mroute.h> 142#endif 143 144#ifdef IPSEC 145#include <netinet6/ipsec.h> 146#include <netkey/key.h> 147#endif 148 149#ifndef IPFORWARDING 150#ifdef GATEWAY 151#define IPFORWARDING 1 /* forward IP packets not for us */ 152#else /* GATEWAY */ 153#define IPFORWARDING 0 /* don't forward IP packets not for us */ 154#endif /* GATEWAY */ 155#endif /* IPFORWARDING */ 156#ifndef IPSENDREDIRECTS 157#define IPSENDREDIRECTS 1 158#endif 159#ifndef IPFORWSRCRT 160#define IPFORWSRCRT 1 /* forward source-routed packets */ 161#endif 162#ifndef IPALLOWSRCRT 163#define IPALLOWSRCRT 1 /* allow source-routed packets */ 164#endif 165#ifndef IPMTUDISC 166#define IPMTUDISC 0 167#endif 168#ifndef IPMTUDISCTIMEOUT 169#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 170#endif 171 172/* 173 * Note: DIRECTED_BROADCAST is handled this way so that previous 174 * configuration using this option will Just Work. 175 */ 176#ifndef IPDIRECTEDBCAST 177#ifdef DIRECTED_BROADCAST 178#define IPDIRECTEDBCAST 1 179#else 180#define IPDIRECTEDBCAST 0 181#endif /* DIRECTED_BROADCAST */ 182#endif /* IPDIRECTEDBCAST */ 183int ipforwarding = IPFORWARDING; 184int ipsendredirects = IPSENDREDIRECTS; 185int ip_defttl = IPDEFTTL; 186int ip_forwsrcrt = IPFORWSRCRT; 187int ip_directedbcast = IPDIRECTEDBCAST; 188int ip_allowsrcrt = IPALLOWSRCRT; 189int ip_mtudisc = IPMTUDISC; 190u_int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 191#ifdef DIAGNOSTIC 192int ipprintfs = 0; 193#endif 194 195struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 196 197extern struct domain inetdomain; 198int ipqmaxlen = IFQ_MAXLEN; 199struct in_ifaddrhead in_ifaddr; 200struct in_ifaddrhashhead *in_ifaddrhashtbl; 201struct ifqueue ipintrq; 202struct ipstat ipstat; 203u_int16_t ip_id; 204 205#ifdef PFIL_HOOKS 206struct pfil_head inet_pfil_hook; 207#endif 208 209struct ipqhead ipq; 210int ipq_locked; 211int ip_nfragpackets = 0; 212int ip_maxfragpackets = 200; 213 214static __inline int ipq_lock_try __P((void)); 215static __inline void ipq_unlock __P((void)); 216 217static __inline int 218ipq_lock_try() 219{ 220 int s; 221 222 /* 223 * Use splvm() -- we're bloking things that would cause 224 * mbuf allocation. 225 */ 226 s = splvm(); 227 if (ipq_locked) { 228 splx(s); 229 return (0); 230 } 231 ipq_locked = 1; 232 splx(s); 233 return (1); 234} 235 236static __inline void 237ipq_unlock() 238{ 239 int s; 240 241 s = splvm(); 242 ipq_locked = 0; 243 splx(s); 244} 245 246#ifdef DIAGNOSTIC 247#define IPQ_LOCK() \ 248do { \ 249 if (ipq_lock_try() == 0) { \ 250 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 251 panic("ipq_lock"); \ 252 } \ 253} while (0) 254#define IPQ_LOCK_CHECK() \ 255do { \ 256 if (ipq_locked == 0) { \ 257 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 258 panic("ipq lock check"); \ 259 } \ 260} while (0) 261#else 262#define IPQ_LOCK() (void) ipq_lock_try() 263#define IPQ_LOCK_CHECK() /* nothing */ 264#endif 265 266#define IPQ_UNLOCK() ipq_unlock() 267 268struct pool ipqent_pool; 269 270#ifdef INET_CSUM_COUNTERS 271#include <sys/device.h> 272 273struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 274 NULL, "inet", "hwcsum bad"); 275struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 276 NULL, "inet", "hwcsum ok"); 277struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 278 NULL, "inet", "swcsum"); 279 280#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 281 282#else 283 284#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 285 286#endif /* INET_CSUM_COUNTERS */ 287 288/* 289 * We need to save the IP options in case a protocol wants to respond 290 * to an incoming packet over the same route if the packet got here 291 * using IP source routing. This allows connection establishment and 292 * maintenance when the remote end is on a network that is not known 293 * to us. 294 */ 295int ip_nhops = 0; 296static struct ip_srcrt { 297 struct in_addr dst; /* final destination */ 298 char nop; /* one NOP to align */ 299 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 300 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 301} ip_srcrt; 302 303static void save_rte __P((u_char *, struct in_addr)); 304 305/* 306 * IP initialization: fill in IP protocol switch table. 307 * All protocols not implemented in kernel go to raw IP protocol handler. 308 */ 309void 310ip_init() 311{ 312 struct protosw *pr; 313 int i; 314 315 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 316 0, NULL, NULL, M_IPQ); 317 318 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 319 if (pr == 0) 320 panic("ip_init"); 321 for (i = 0; i < IPPROTO_MAX; i++) 322 ip_protox[i] = pr - inetsw; 323 for (pr = inetdomain.dom_protosw; 324 pr < inetdomain.dom_protoswNPROTOSW; pr++) 325 if (pr->pr_domain->dom_family == PF_INET && 326 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 327 ip_protox[pr->pr_protocol] = pr - inetsw; 328 LIST_INIT(&ipq); 329 ip_id = time.tv_sec & 0xffff; 330 ipintrq.ifq_maxlen = ipqmaxlen; 331 TAILQ_INIT(&in_ifaddr); 332 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 333 M_WAITOK, &in_ifaddrhash); 334 if (ip_mtudisc != 0) 335 ip_mtudisc_timeout_q = 336 rt_timer_queue_create(ip_mtudisc_timeout); 337#ifdef GATEWAY 338 ipflow_init(); 339#endif 340 341#ifdef PFIL_HOOKS 342 /* Register our Packet Filter hook. */ 343 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 344 inet_pfil_hook.ph_af = AF_INET; 345 i = pfil_head_register(&inet_pfil_hook); 346 if (i != 0) 347 printf("ip_init: WARNING: unable to register pfil hook, " 348 "error %d\n", i); 349#endif /* PFIL_HOOKS */ 350 351#ifdef INET_CSUM_COUNTERS 352 evcnt_attach_static(&ip_hwcsum_bad); 353 evcnt_attach_static(&ip_hwcsum_ok); 354 evcnt_attach_static(&ip_swcsum); 355#endif /* INET_CSUM_COUNTERS */ 356} 357 358struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 359struct route ipforward_rt; 360 361/* 362 * IP software interrupt routine 363 */ 364void 365ipintr() 366{ 367 int s; 368 struct mbuf *m; 369 370 while (1) { 371 s = splnet(); 372 IF_DEQUEUE(&ipintrq, m); 373 splx(s); 374 if (m == 0) 375 return; 376 ip_input(m); 377 } 378} 379 380/* 381 * Ip input routine. Checksum and byte swap header. If fragmented 382 * try to reassemble. Process options. Pass to next level. 383 */ 384void 385ip_input(struct mbuf *m) 386{ 387 struct ip *ip = NULL; 388 struct ipq *fp; 389 struct in_ifaddr *ia; 390 struct ifaddr *ifa; 391 struct ipqent *ipqe; 392 int hlen = 0, mff, len; 393 int downmatch; 394 395#ifdef DIAGNOSTIC 396 if ((m->m_flags & M_PKTHDR) == 0) 397 panic("ipintr no HDR"); 398#endif 399#ifdef IPSEC 400 /* 401 * should the inner packet be considered authentic? 402 * see comment in ah4_input(). 403 */ 404 if (m) { 405 m->m_flags &= ~M_AUTHIPHDR; 406 m->m_flags &= ~M_AUTHIPDGM; 407 } 408#endif 409 /* 410 * If no IP addresses have been set yet but the interfaces 411 * are receiving, can't do anything with incoming packets yet. 412 */ 413 if (in_ifaddr.tqh_first == 0) 414 goto bad; 415 ipstat.ips_total++; 416 if (m->m_len < sizeof (struct ip) && 417 (m = m_pullup(m, sizeof (struct ip))) == 0) { 418 ipstat.ips_toosmall++; 419 return; 420 } 421 ip = mtod(m, struct ip *); 422 if (ip->ip_v != IPVERSION) { 423 ipstat.ips_badvers++; 424 goto bad; 425 } 426 hlen = ip->ip_hl << 2; 427 if (hlen < sizeof(struct ip)) { /* minimum header length */ 428 ipstat.ips_badhlen++; 429 goto bad; 430 } 431 if (hlen > m->m_len) { 432 if ((m = m_pullup(m, hlen)) == 0) { 433 ipstat.ips_badhlen++; 434 return; 435 } 436 ip = mtod(m, struct ip *); 437 } 438 439 /* 440 * RFC1122: packets with a multicast source address are 441 * not allowed. 442 */ 443 if (IN_MULTICAST(ip->ip_src.s_addr)) { 444 ipstat.ips_badaddr++; 445 goto bad; 446 } 447 448 /* 127/8 must not appear on wire - RFC1122 */ 449 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 450 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 451 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 452 ipstat.ips_badaddr++; 453 goto bad; 454 } 455 } 456 457 switch (m->m_pkthdr.csum_flags & 458 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) | 459 M_CSUM_IPv4_BAD)) { 460 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 461 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 462 goto badcsum; 463 464 case M_CSUM_IPv4: 465 /* Checksum was okay. */ 466 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 467 break; 468 469 default: 470 /* Must compute it ourselves. */ 471 INET_CSUM_COUNTER_INCR(&ip_swcsum); 472 if (in_cksum(m, hlen) != 0) 473 goto bad; 474 break; 475 } 476 477 /* Retrieve the packet length. */ 478 len = ntohs(ip->ip_len); 479 480 /* 481 * Check for additional length bogosity 482 */ 483 if (len < hlen) { 484 ipstat.ips_badlen++; 485 goto bad; 486 } 487 488 /* 489 * Check that the amount of data in the buffers 490 * is as at least much as the IP header would have us expect. 491 * Trim mbufs if longer than we expect. 492 * Drop packet if shorter than we expect. 493 */ 494 if (m->m_pkthdr.len < len) { 495 ipstat.ips_tooshort++; 496 goto bad; 497 } 498 if (m->m_pkthdr.len > len) { 499 if (m->m_len == m->m_pkthdr.len) { 500 m->m_len = len; 501 m->m_pkthdr.len = len; 502 } else 503 m_adj(m, len - m->m_pkthdr.len); 504 } 505 506#ifdef IPSEC 507 /* ipflow (IP fast fowarding) is not compatible with IPsec. */ 508 m->m_flags &= ~M_CANFASTFWD; 509#else 510 /* 511 * Assume that we can create a fast-forward IP flow entry 512 * based on this packet. 513 */ 514 m->m_flags |= M_CANFASTFWD; 515#endif 516 517#ifdef PFIL_HOOKS 518 /* 519 * Run through list of hooks for input packets. If there are any 520 * filters which require that additional packets in the flow are 521 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 522 * Note that filters must _never_ set this flag, as another filter 523 * in the list may have previously cleared it. 524 */ 525 /* 526 * let ipfilter look at packet on the wire, 527 * not the decapsulated packet. 528 */ 529#ifdef IPSEC 530 if (!ipsec_getnhist(m)) 531#else 532 if (1) 533#endif 534 { 535 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 536 PFIL_IN) != 0) 537 return; 538 if (m == NULL) 539 return; 540 ip = mtod(m, struct ip *); 541 } 542#endif /* PFIL_HOOKS */ 543 544#ifdef ALTQ 545 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 546 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 547 /* packet dropped by traffic conditioner */ 548 return; 549 } 550#endif 551 552 /* 553 * Convert fields to host representation. 554 */ 555 NTOHS(ip->ip_len); 556 NTOHS(ip->ip_off); 557 558 /* 559 * Process options and, if not destined for us, 560 * ship it on. ip_dooptions returns 1 when an 561 * error was detected (causing an icmp message 562 * to be sent and the original packet to be freed). 563 */ 564 ip_nhops = 0; /* for source routed packets */ 565 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 566 return; 567 568 /* 569 * Check our list of addresses, to see if the packet is for us. 570 * 571 * Traditional 4.4BSD did not consult IFF_UP at all. 572 * The behavior here is to treat addresses on !IFF_UP interface 573 * as not mine. 574 */ 575 downmatch = 0; 576 for (ia = IN_IFADDR_HASH(ip->ip_dst.s_addr).lh_first; 577 ia != NULL; 578 ia = ia->ia_hash.le_next) { 579 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 580 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 581 break; 582 else 583 downmatch++; 584 } 585 } 586 if (ia != NULL) 587 goto ours; 588 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 589 for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first; 590 ifa != NULL; ifa = ifa->ifa_list.tqe_next) { 591 if (ifa->ifa_addr->sa_family != AF_INET) continue; 592 ia = ifatoia(ifa); 593 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 594 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 595 /* 596 * Look for all-0's host part (old broadcast addr), 597 * either for subnet or net. 598 */ 599 ip->ip_dst.s_addr == ia->ia_subnet || 600 ip->ip_dst.s_addr == ia->ia_net) 601 goto ours; 602 /* 603 * An interface with IP address zero accepts 604 * all packets that arrive on that interface. 605 */ 606 if (in_nullhost(ia->ia_addr.sin_addr)) 607 goto ours; 608 } 609 } 610 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 611 struct in_multi *inm; 612#ifdef MROUTING 613 extern struct socket *ip_mrouter; 614 615 if (m->m_flags & M_EXT) { 616 if ((m = m_pullup(m, hlen)) == 0) { 617 ipstat.ips_toosmall++; 618 return; 619 } 620 ip = mtod(m, struct ip *); 621 } 622 623 if (ip_mrouter) { 624 /* 625 * If we are acting as a multicast router, all 626 * incoming multicast packets are passed to the 627 * kernel-level multicast forwarding function. 628 * The packet is returned (relatively) intact; if 629 * ip_mforward() returns a non-zero value, the packet 630 * must be discarded, else it may be accepted below. 631 * 632 * (The IP ident field is put in the same byte order 633 * as expected when ip_mforward() is called from 634 * ip_output().) 635 */ 636 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 637 ipstat.ips_cantforward++; 638 m_freem(m); 639 return; 640 } 641 642 /* 643 * The process-level routing demon needs to receive 644 * all multicast IGMP packets, whether or not this 645 * host belongs to their destination groups. 646 */ 647 if (ip->ip_p == IPPROTO_IGMP) 648 goto ours; 649 ipstat.ips_forward++; 650 } 651#endif 652 /* 653 * See if we belong to the destination multicast group on the 654 * arrival interface. 655 */ 656 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 657 if (inm == NULL) { 658 ipstat.ips_cantforward++; 659 m_freem(m); 660 return; 661 } 662 goto ours; 663 } 664 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 665 in_nullhost(ip->ip_dst)) 666 goto ours; 667 668 /* 669 * Not for us; forward if possible and desirable. 670 */ 671 if (ipforwarding == 0) { 672 ipstat.ips_cantforward++; 673 m_freem(m); 674 } else { 675 /* 676 * If ip_dst matched any of my address on !IFF_UP interface, 677 * and there's no IFF_UP interface that matches ip_dst, 678 * send icmp unreach. Forwarding it will result in in-kernel 679 * forwarding loop till TTL goes to 0. 680 */ 681 if (downmatch) { 682 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 683 ipstat.ips_cantforward++; 684 return; 685 } 686 ip_forward(m, 0); 687 } 688 return; 689 690ours: 691 /* 692 * If offset or IP_MF are set, must reassemble. 693 * Otherwise, nothing need be done. 694 * (We could look in the reassembly queue to see 695 * if the packet was previously fragmented, 696 * but it's not worth the time; just let them time out.) 697 */ 698 if (ip->ip_off & ~(IP_DF|IP_RF)) { 699 /* 700 * Look for queue of fragments 701 * of this datagram. 702 */ 703 IPQ_LOCK(); 704 for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next) 705 if (ip->ip_id == fp->ipq_id && 706 in_hosteq(ip->ip_src, fp->ipq_src) && 707 in_hosteq(ip->ip_dst, fp->ipq_dst) && 708 ip->ip_p == fp->ipq_p) 709 goto found; 710 fp = 0; 711found: 712 713 /* 714 * Adjust ip_len to not reflect header, 715 * set ipqe_mff if more fragments are expected, 716 * convert offset of this to bytes. 717 */ 718 ip->ip_len -= hlen; 719 mff = (ip->ip_off & IP_MF) != 0; 720 if (mff) { 721 /* 722 * Make sure that fragments have a data length 723 * that's a non-zero multiple of 8 bytes. 724 */ 725 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 726 ipstat.ips_badfrags++; 727 IPQ_UNLOCK(); 728 goto bad; 729 } 730 } 731 ip->ip_off <<= 3; 732 733 /* 734 * If datagram marked as having more fragments 735 * or if this is not the first fragment, 736 * attempt reassembly; if it succeeds, proceed. 737 */ 738 if (mff || ip->ip_off) { 739 ipstat.ips_fragments++; 740 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 741 if (ipqe == NULL) { 742 ipstat.ips_rcvmemdrop++; 743 IPQ_UNLOCK(); 744 goto bad; 745 } 746 ipqe->ipqe_mff = mff; 747 ipqe->ipqe_m = m; 748 ipqe->ipqe_ip = ip; 749 m = ip_reass(ipqe, fp); 750 if (m == 0) { 751 IPQ_UNLOCK(); 752 return; 753 } 754 ipstat.ips_reassembled++; 755 ip = mtod(m, struct ip *); 756 hlen = ip->ip_hl << 2; 757 ip->ip_len += hlen; 758 } else 759 if (fp) 760 ip_freef(fp); 761 IPQ_UNLOCK(); 762 } 763 764#ifdef IPSEC 765 /* 766 * enforce IPsec policy checking if we are seeing last header. 767 * note that we do not visit this with protocols with pcb layer 768 * code - like udp/tcp/raw ip. 769 */ 770 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 771 ipsec4_in_reject(m, NULL)) { 772 ipsecstat.in_polvio++; 773 goto bad; 774 } 775#endif 776 777 /* 778 * Switch out to protocol's input routine. 779 */ 780#if IFA_STATS 781 if (ia && ip) 782 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 783#endif 784 ipstat.ips_delivered++; 785 { 786 int off = hlen, nh = ip->ip_p; 787 788 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 789 return; 790 } 791bad: 792 m_freem(m); 793 return; 794 795badcsum: 796 ipstat.ips_badsum++; 797 m_freem(m); 798} 799 800/* 801 * Take incoming datagram fragment and try to 802 * reassemble it into whole datagram. If a chain for 803 * reassembly of this datagram already exists, then it 804 * is given as fp; otherwise have to make a chain. 805 */ 806struct mbuf * 807ip_reass(ipqe, fp) 808 struct ipqent *ipqe; 809 struct ipq *fp; 810{ 811 struct mbuf *m = ipqe->ipqe_m; 812 struct ipqent *nq, *p, *q; 813 struct ip *ip; 814 struct mbuf *t; 815 int hlen = ipqe->ipqe_ip->ip_hl << 2; 816 int i, next; 817 818 IPQ_LOCK_CHECK(); 819 820 /* 821 * Presence of header sizes in mbufs 822 * would confuse code below. 823 */ 824 m->m_data += hlen; 825 m->m_len -= hlen; 826 827 /* 828 * If first fragment to arrive, create a reassembly queue. 829 */ 830 if (fp == 0) { 831 /* 832 * Enforce upper bound on number of fragmented packets 833 * for which we attempt reassembly; 834 * If maxfrag is 0, never accept fragments. 835 * If maxfrag is -1, accept all fragments without limitation. 836 */ 837 if (ip_maxfragpackets < 0) 838 ; 839 else if (ip_nfragpackets >= ip_maxfragpackets) 840 goto dropfrag; 841 ip_nfragpackets++; 842 MALLOC(fp, struct ipq *, sizeof (struct ipq), 843 M_FTABLE, M_NOWAIT); 844 if (fp == NULL) 845 goto dropfrag; 846 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 847 fp->ipq_ttl = IPFRAGTTL; 848 fp->ipq_p = ipqe->ipqe_ip->ip_p; 849 fp->ipq_id = ipqe->ipqe_ip->ip_id; 850 LIST_INIT(&fp->ipq_fragq); 851 fp->ipq_src = ipqe->ipqe_ip->ip_src; 852 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 853 p = NULL; 854 goto insert; 855 } 856 857 /* 858 * Find a segment which begins after this one does. 859 */ 860 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 861 p = q, q = q->ipqe_q.le_next) 862 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 863 break; 864 865 /* 866 * If there is a preceding segment, it may provide some of 867 * our data already. If so, drop the data from the incoming 868 * segment. If it provides all of our data, drop us. 869 */ 870 if (p != NULL) { 871 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 872 ipqe->ipqe_ip->ip_off; 873 if (i > 0) { 874 if (i >= ipqe->ipqe_ip->ip_len) 875 goto dropfrag; 876 m_adj(ipqe->ipqe_m, i); 877 ipqe->ipqe_ip->ip_off += i; 878 ipqe->ipqe_ip->ip_len -= i; 879 } 880 } 881 882 /* 883 * While we overlap succeeding segments trim them or, 884 * if they are completely covered, dequeue them. 885 */ 886 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 887 q->ipqe_ip->ip_off; q = nq) { 888 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 889 q->ipqe_ip->ip_off; 890 if (i < q->ipqe_ip->ip_len) { 891 q->ipqe_ip->ip_len -= i; 892 q->ipqe_ip->ip_off += i; 893 m_adj(q->ipqe_m, i); 894 break; 895 } 896 nq = q->ipqe_q.le_next; 897 m_freem(q->ipqe_m); 898 LIST_REMOVE(q, ipqe_q); 899 pool_put(&ipqent_pool, q); 900 } 901 902insert: 903 /* 904 * Stick new segment in its place; 905 * check for complete reassembly. 906 */ 907 if (p == NULL) { 908 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 909 } else { 910 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 911 } 912 next = 0; 913 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 914 p = q, q = q->ipqe_q.le_next) { 915 if (q->ipqe_ip->ip_off != next) 916 return (0); 917 next += q->ipqe_ip->ip_len; 918 } 919 if (p->ipqe_mff) 920 return (0); 921 922 /* 923 * Reassembly is complete. Check for a bogus message size and 924 * concatenate fragments. 925 */ 926 q = fp->ipq_fragq.lh_first; 927 ip = q->ipqe_ip; 928 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 929 ipstat.ips_toolong++; 930 ip_freef(fp); 931 return (0); 932 } 933 m = q->ipqe_m; 934 t = m->m_next; 935 m->m_next = 0; 936 m_cat(m, t); 937 nq = q->ipqe_q.le_next; 938 pool_put(&ipqent_pool, q); 939 for (q = nq; q != NULL; q = nq) { 940 t = q->ipqe_m; 941 nq = q->ipqe_q.le_next; 942 pool_put(&ipqent_pool, q); 943 m_cat(m, t); 944 } 945 946 /* 947 * Create header for new ip packet by 948 * modifying header of first packet; 949 * dequeue and discard fragment reassembly header. 950 * Make header visible. 951 */ 952 ip->ip_len = next; 953 ip->ip_src = fp->ipq_src; 954 ip->ip_dst = fp->ipq_dst; 955 LIST_REMOVE(fp, ipq_q); 956 FREE(fp, M_FTABLE); 957 ip_nfragpackets--; 958 m->m_len += (ip->ip_hl << 2); 959 m->m_data -= (ip->ip_hl << 2); 960 /* some debugging cruft by sklower, below, will go away soon */ 961 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 962 int plen = 0; 963 for (t = m; t; t = t->m_next) 964 plen += t->m_len; 965 m->m_pkthdr.len = plen; 966 } 967 return (m); 968 969dropfrag: 970 ipstat.ips_fragdropped++; 971 m_freem(m); 972 pool_put(&ipqent_pool, ipqe); 973 return (0); 974} 975 976/* 977 * Free a fragment reassembly header and all 978 * associated datagrams. 979 */ 980void 981ip_freef(fp) 982 struct ipq *fp; 983{ 984 struct ipqent *q, *p; 985 986 IPQ_LOCK_CHECK(); 987 988 for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { 989 p = q->ipqe_q.le_next; 990 m_freem(q->ipqe_m); 991 LIST_REMOVE(q, ipqe_q); 992 pool_put(&ipqent_pool, q); 993 } 994 LIST_REMOVE(fp, ipq_q); 995 FREE(fp, M_FTABLE); 996 ip_nfragpackets--; 997} 998 999/* 1000 * IP timer processing; 1001 * if a timer expires on a reassembly 1002 * queue, discard it. 1003 */ 1004void 1005ip_slowtimo() 1006{ 1007 struct ipq *fp, *nfp; 1008 int s = splsoftnet(); 1009 1010 IPQ_LOCK(); 1011 for (fp = ipq.lh_first; fp != NULL; fp = nfp) { 1012 nfp = fp->ipq_q.le_next; 1013 if (--fp->ipq_ttl == 0) { 1014 ipstat.ips_fragtimeout++; 1015 ip_freef(fp); 1016 } 1017 } 1018 /* 1019 * If we are over the maximum number of fragments 1020 * (due to the limit being lowered), drain off 1021 * enough to get down to the new limit. 1022 */ 1023 if (ip_maxfragpackets < 0) 1024 ; 1025 else { 1026 while (ip_nfragpackets > ip_maxfragpackets && ipq.lh_first) 1027 ip_freef(ipq.lh_first); 1028 } 1029 IPQ_UNLOCK(); 1030#ifdef GATEWAY 1031 ipflow_slowtimo(); 1032#endif 1033 splx(s); 1034} 1035 1036/* 1037 * Drain off all datagram fragments. 1038 */ 1039void 1040ip_drain() 1041{ 1042 1043 /* 1044 * We may be called from a device's interrupt context. If 1045 * the ipq is already busy, just bail out now. 1046 */ 1047 if (ipq_lock_try() == 0) 1048 return; 1049 1050 while (ipq.lh_first != NULL) { 1051 ipstat.ips_fragdropped++; 1052 ip_freef(ipq.lh_first); 1053 } 1054 1055 IPQ_UNLOCK(); 1056} 1057 1058/* 1059 * Do option processing on a datagram, 1060 * possibly discarding it if bad options are encountered, 1061 * or forwarding it if source-routed. 1062 * Returns 1 if packet has been forwarded/freed, 1063 * 0 if the packet should be processed further. 1064 */ 1065int 1066ip_dooptions(m) 1067 struct mbuf *m; 1068{ 1069 struct ip *ip = mtod(m, struct ip *); 1070 u_char *cp, *cp0; 1071 struct ip_timestamp *ipt; 1072 struct in_ifaddr *ia; 1073 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1074 struct in_addr dst; 1075 n_time ntime; 1076 1077 dst = ip->ip_dst; 1078 cp = (u_char *)(ip + 1); 1079 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1080 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1081 opt = cp[IPOPT_OPTVAL]; 1082 if (opt == IPOPT_EOL) 1083 break; 1084 if (opt == IPOPT_NOP) 1085 optlen = 1; 1086 else { 1087 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1088 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1089 goto bad; 1090 } 1091 optlen = cp[IPOPT_OLEN]; 1092 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1093 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1094 goto bad; 1095 } 1096 } 1097 switch (opt) { 1098 1099 default: 1100 break; 1101 1102 /* 1103 * Source routing with record. 1104 * Find interface with current destination address. 1105 * If none on this machine then drop if strictly routed, 1106 * or do nothing if loosely routed. 1107 * Record interface address and bring up next address 1108 * component. If strictly routed make sure next 1109 * address is on directly accessible net. 1110 */ 1111 case IPOPT_LSRR: 1112 case IPOPT_SSRR: 1113 if (ip_allowsrcrt == 0) { 1114 type = ICMP_UNREACH; 1115 code = ICMP_UNREACH_NET_PROHIB; 1116 goto bad; 1117 } 1118 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1119 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1120 goto bad; 1121 } 1122 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1123 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1124 goto bad; 1125 } 1126 ipaddr.sin_addr = ip->ip_dst; 1127 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1128 if (ia == 0) { 1129 if (opt == IPOPT_SSRR) { 1130 type = ICMP_UNREACH; 1131 code = ICMP_UNREACH_SRCFAIL; 1132 goto bad; 1133 } 1134 /* 1135 * Loose routing, and not at next destination 1136 * yet; nothing to do except forward. 1137 */ 1138 break; 1139 } 1140 off--; /* 0 origin */ 1141 if ((off + sizeof(struct in_addr)) > optlen) { 1142 /* 1143 * End of source route. Should be for us. 1144 */ 1145 save_rte(cp, ip->ip_src); 1146 break; 1147 } 1148 /* 1149 * locate outgoing interface 1150 */ 1151 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1152 sizeof(ipaddr.sin_addr)); 1153 if (opt == IPOPT_SSRR) 1154 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1155 else 1156 ia = ip_rtaddr(ipaddr.sin_addr); 1157 if (ia == 0) { 1158 type = ICMP_UNREACH; 1159 code = ICMP_UNREACH_SRCFAIL; 1160 goto bad; 1161 } 1162 ip->ip_dst = ipaddr.sin_addr; 1163 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1164 (caddr_t)(cp + off), sizeof(struct in_addr)); 1165 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1166 /* 1167 * Let ip_intr's mcast routing check handle mcast pkts 1168 */ 1169 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1170 break; 1171 1172 case IPOPT_RR: 1173 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1174 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1175 goto bad; 1176 } 1177 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1178 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1179 goto bad; 1180 } 1181 /* 1182 * If no space remains, ignore. 1183 */ 1184 off--; /* 0 origin */ 1185 if ((off + sizeof(struct in_addr)) > optlen) 1186 break; 1187 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1188 sizeof(ipaddr.sin_addr)); 1189 /* 1190 * locate outgoing interface; if we're the destination, 1191 * use the incoming interface (should be same). 1192 */ 1193 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1194 == NULL && 1195 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1196 type = ICMP_UNREACH; 1197 code = ICMP_UNREACH_HOST; 1198 goto bad; 1199 } 1200 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1201 (caddr_t)(cp + off), sizeof(struct in_addr)); 1202 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1203 break; 1204 1205 case IPOPT_TS: 1206 code = cp - (u_char *)ip; 1207 ipt = (struct ip_timestamp *)cp; 1208 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1209 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1210 goto bad; 1211 } 1212 if (ipt->ipt_ptr < 5) { 1213 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1214 goto bad; 1215 } 1216 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1217 if (++ipt->ipt_oflw == 0) { 1218 code = (u_char *)&ipt->ipt_ptr - 1219 (u_char *)ip; 1220 goto bad; 1221 } 1222 break; 1223 } 1224 cp0 = (cp + ipt->ipt_ptr - 1); 1225 switch (ipt->ipt_flg) { 1226 1227 case IPOPT_TS_TSONLY: 1228 break; 1229 1230 case IPOPT_TS_TSANDADDR: 1231 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1232 sizeof(struct in_addr) > ipt->ipt_len) { 1233 code = (u_char *)&ipt->ipt_ptr - 1234 (u_char *)ip; 1235 goto bad; 1236 } 1237 ipaddr.sin_addr = dst; 1238 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1239 m->m_pkthdr.rcvif)); 1240 if (ia == 0) 1241 continue; 1242 bcopy(&ia->ia_addr.sin_addr, 1243 cp0, sizeof(struct in_addr)); 1244 ipt->ipt_ptr += sizeof(struct in_addr); 1245 break; 1246 1247 case IPOPT_TS_PRESPEC: 1248 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1249 sizeof(struct in_addr) > ipt->ipt_len) { 1250 code = (u_char *)&ipt->ipt_ptr - 1251 (u_char *)ip; 1252 goto bad; 1253 } 1254 bcopy(cp0, &ipaddr.sin_addr, 1255 sizeof(struct in_addr)); 1256 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1257 == NULL) 1258 continue; 1259 ipt->ipt_ptr += sizeof(struct in_addr); 1260 break; 1261 1262 default: 1263 /* XXX can't take &ipt->ipt_flg */ 1264 code = (u_char *)&ipt->ipt_ptr - 1265 (u_char *)ip + 1; 1266 goto bad; 1267 } 1268 ntime = iptime(); 1269 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1270 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1271 sizeof(n_time)); 1272 ipt->ipt_ptr += sizeof(n_time); 1273 } 1274 } 1275 if (forward) { 1276 if (ip_forwsrcrt == 0) { 1277 type = ICMP_UNREACH; 1278 code = ICMP_UNREACH_SRCFAIL; 1279 goto bad; 1280 } 1281 ip_forward(m, 1); 1282 return (1); 1283 } 1284 return (0); 1285bad: 1286 icmp_error(m, type, code, 0, 0); 1287 ipstat.ips_badoptions++; 1288 return (1); 1289} 1290 1291/* 1292 * Given address of next destination (final or next hop), 1293 * return internet address info of interface to be used to get there. 1294 */ 1295struct in_ifaddr * 1296ip_rtaddr(dst) 1297 struct in_addr dst; 1298{ 1299 struct sockaddr_in *sin; 1300 1301 sin = satosin(&ipforward_rt.ro_dst); 1302 1303 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1304 if (ipforward_rt.ro_rt) { 1305 RTFREE(ipforward_rt.ro_rt); 1306 ipforward_rt.ro_rt = 0; 1307 } 1308 sin->sin_family = AF_INET; 1309 sin->sin_len = sizeof(*sin); 1310 sin->sin_addr = dst; 1311 1312 rtalloc(&ipforward_rt); 1313 } 1314 if (ipforward_rt.ro_rt == 0) 1315 return ((struct in_ifaddr *)0); 1316 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1317} 1318 1319/* 1320 * Save incoming source route for use in replies, 1321 * to be picked up later by ip_srcroute if the receiver is interested. 1322 */ 1323void 1324save_rte(option, dst) 1325 u_char *option; 1326 struct in_addr dst; 1327{ 1328 unsigned olen; 1329 1330 olen = option[IPOPT_OLEN]; 1331#ifdef DIAGNOSTIC 1332 if (ipprintfs) 1333 printf("save_rte: olen %d\n", olen); 1334#endif /* 0 */ 1335 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1336 return; 1337 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1338 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1339 ip_srcrt.dst = dst; 1340} 1341 1342/* 1343 * Retrieve incoming source route for use in replies, 1344 * in the same form used by setsockopt. 1345 * The first hop is placed before the options, will be removed later. 1346 */ 1347struct mbuf * 1348ip_srcroute() 1349{ 1350 struct in_addr *p, *q; 1351 struct mbuf *m; 1352 1353 if (ip_nhops == 0) 1354 return ((struct mbuf *)0); 1355 m = m_get(M_DONTWAIT, MT_SOOPTS); 1356 if (m == 0) 1357 return ((struct mbuf *)0); 1358 1359#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1360 1361 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1362 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1363 OPTSIZ; 1364#ifdef DIAGNOSTIC 1365 if (ipprintfs) 1366 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1367#endif 1368 1369 /* 1370 * First save first hop for return route 1371 */ 1372 p = &ip_srcrt.route[ip_nhops - 1]; 1373 *(mtod(m, struct in_addr *)) = *p--; 1374#ifdef DIAGNOSTIC 1375 if (ipprintfs) 1376 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1377#endif 1378 1379 /* 1380 * Copy option fields and padding (nop) to mbuf. 1381 */ 1382 ip_srcrt.nop = IPOPT_NOP; 1383 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1384 bcopy((caddr_t)&ip_srcrt.nop, 1385 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1386 q = (struct in_addr *)(mtod(m, caddr_t) + 1387 sizeof(struct in_addr) + OPTSIZ); 1388#undef OPTSIZ 1389 /* 1390 * Record return path as an IP source route, 1391 * reversing the path (pointers are now aligned). 1392 */ 1393 while (p >= ip_srcrt.route) { 1394#ifdef DIAGNOSTIC 1395 if (ipprintfs) 1396 printf(" %x", ntohl(q->s_addr)); 1397#endif 1398 *q++ = *p--; 1399 } 1400 /* 1401 * Last hop goes to final destination. 1402 */ 1403 *q = ip_srcrt.dst; 1404#ifdef DIAGNOSTIC 1405 if (ipprintfs) 1406 printf(" %x\n", ntohl(q->s_addr)); 1407#endif 1408 return (m); 1409} 1410 1411/* 1412 * Strip out IP options, at higher 1413 * level protocol in the kernel. 1414 * Second argument is buffer to which options 1415 * will be moved, and return value is their length. 1416 * XXX should be deleted; last arg currently ignored. 1417 */ 1418void 1419ip_stripoptions(m, mopt) 1420 struct mbuf *m; 1421 struct mbuf *mopt; 1422{ 1423 int i; 1424 struct ip *ip = mtod(m, struct ip *); 1425 caddr_t opts; 1426 int olen; 1427 1428 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1429 opts = (caddr_t)(ip + 1); 1430 i = m->m_len - (sizeof (struct ip) + olen); 1431 bcopy(opts + olen, opts, (unsigned)i); 1432 m->m_len -= olen; 1433 if (m->m_flags & M_PKTHDR) 1434 m->m_pkthdr.len -= olen; 1435 ip->ip_len -= olen; 1436 ip->ip_hl = sizeof (struct ip) >> 2; 1437} 1438 1439int inetctlerrmap[PRC_NCMDS] = { 1440 0, 0, 0, 0, 1441 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1442 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1443 EMSGSIZE, EHOSTUNREACH, 0, 0, 1444 0, 0, 0, 0, 1445 ENOPROTOOPT 1446}; 1447 1448/* 1449 * Forward a packet. If some error occurs return the sender 1450 * an icmp packet. Note we can't always generate a meaningful 1451 * icmp message because icmp doesn't have a large enough repertoire 1452 * of codes and types. 1453 * 1454 * If not forwarding, just drop the packet. This could be confusing 1455 * if ipforwarding was zero but some routing protocol was advancing 1456 * us as a gateway to somewhere. However, we must let the routing 1457 * protocol deal with that. 1458 * 1459 * The srcrt parameter indicates whether the packet is being forwarded 1460 * via a source route. 1461 */ 1462void 1463ip_forward(m, srcrt) 1464 struct mbuf *m; 1465 int srcrt; 1466{ 1467 struct ip *ip = mtod(m, struct ip *); 1468 struct sockaddr_in *sin; 1469 struct rtentry *rt; 1470 int error, type = 0, code = 0; 1471 struct mbuf *mcopy; 1472 n_long dest; 1473 struct ifnet *destifp; 1474#ifdef IPSEC 1475 struct ifnet dummyifp; 1476#endif 1477 1478 /* 1479 * Clear any in-bound checksum flags for this packet. 1480 */ 1481 m->m_pkthdr.csum_flags = 0; 1482 1483 dest = 0; 1484#ifdef DIAGNOSTIC 1485 if (ipprintfs) 1486 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1487 ntohl(ip->ip_src.s_addr), 1488 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1489#endif 1490 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1491 ipstat.ips_cantforward++; 1492 m_freem(m); 1493 return; 1494 } 1495 if (ip->ip_ttl <= IPTTLDEC) { 1496 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1497 return; 1498 } 1499 ip->ip_ttl -= IPTTLDEC; 1500 1501 sin = satosin(&ipforward_rt.ro_dst); 1502 if ((rt = ipforward_rt.ro_rt) == 0 || 1503 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1504 if (ipforward_rt.ro_rt) { 1505 RTFREE(ipforward_rt.ro_rt); 1506 ipforward_rt.ro_rt = 0; 1507 } 1508 sin->sin_family = AF_INET; 1509 sin->sin_len = sizeof(struct sockaddr_in); 1510 sin->sin_addr = ip->ip_dst; 1511 1512 rtalloc(&ipforward_rt); 1513 if (ipforward_rt.ro_rt == 0) { 1514 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1515 return; 1516 } 1517 rt = ipforward_rt.ro_rt; 1518 } 1519 1520 /* 1521 * Save at most 68 bytes of the packet in case 1522 * we need to generate an ICMP message to the src. 1523 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1524 */ 1525 mcopy = m_copym(m, 0, imin((int)ip->ip_len, 68), M_DONTWAIT); 1526 if (mcopy) 1527 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1528 1529 /* 1530 * If forwarding packet using same interface that it came in on, 1531 * perhaps should send a redirect to sender to shortcut a hop. 1532 * Only send redirect if source is sending directly to us, 1533 * and if packet was not source routed (or has any options). 1534 * Also, don't send redirect if forwarding using a default route 1535 * or a route modified by a redirect. 1536 */ 1537 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1538 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1539 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1540 ipsendredirects && !srcrt) { 1541 if (rt->rt_ifa && 1542 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1543 ifatoia(rt->rt_ifa)->ia_subnet) { 1544 if (rt->rt_flags & RTF_GATEWAY) 1545 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1546 else 1547 dest = ip->ip_dst.s_addr; 1548 /* 1549 * Router requirements says to only send host 1550 * redirects. 1551 */ 1552 type = ICMP_REDIRECT; 1553 code = ICMP_REDIRECT_HOST; 1554#ifdef DIAGNOSTIC 1555 if (ipprintfs) 1556 printf("redirect (%d) to %x\n", code, 1557 (u_int32_t)dest); 1558#endif 1559 } 1560 } 1561 1562#ifdef IPSEC 1563 /* Don't lookup socket in forwarding case */ 1564 (void)ipsec_setsocket(m, NULL); 1565#endif 1566 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1567 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1568 if (error) 1569 ipstat.ips_cantforward++; 1570 else { 1571 ipstat.ips_forward++; 1572 if (type) 1573 ipstat.ips_redirectsent++; 1574 else { 1575 if (mcopy) { 1576#ifdef GATEWAY 1577 if (mcopy->m_flags & M_CANFASTFWD) 1578 ipflow_create(&ipforward_rt, mcopy); 1579#endif 1580 m_freem(mcopy); 1581 } 1582 return; 1583 } 1584 } 1585 if (mcopy == NULL) 1586 return; 1587 destifp = NULL; 1588 1589 switch (error) { 1590 1591 case 0: /* forwarded, but need redirect */ 1592 /* type, code set above */ 1593 break; 1594 1595 case ENETUNREACH: /* shouldn't happen, checked above */ 1596 case EHOSTUNREACH: 1597 case ENETDOWN: 1598 case EHOSTDOWN: 1599 default: 1600 type = ICMP_UNREACH; 1601 code = ICMP_UNREACH_HOST; 1602 break; 1603 1604 case EMSGSIZE: 1605 type = ICMP_UNREACH; 1606 code = ICMP_UNREACH_NEEDFRAG; 1607#ifndef IPSEC 1608 if (ipforward_rt.ro_rt) 1609 destifp = ipforward_rt.ro_rt->rt_ifp; 1610#else 1611 /* 1612 * If the packet is routed over IPsec tunnel, tell the 1613 * originator the tunnel MTU. 1614 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1615 * XXX quickhack!!! 1616 */ 1617 if (ipforward_rt.ro_rt) { 1618 struct secpolicy *sp; 1619 int ipsecerror; 1620 size_t ipsechdr; 1621 struct route *ro; 1622 1623 sp = ipsec4_getpolicybyaddr(mcopy, 1624 IPSEC_DIR_OUTBOUND, 1625 IP_FORWARDING, 1626 &ipsecerror); 1627 1628 if (sp == NULL) 1629 destifp = ipforward_rt.ro_rt->rt_ifp; 1630 else { 1631 /* count IPsec header size */ 1632 ipsechdr = ipsec4_hdrsiz(mcopy, 1633 IPSEC_DIR_OUTBOUND, 1634 NULL); 1635 1636 /* 1637 * find the correct route for outer IPv4 1638 * header, compute tunnel MTU. 1639 * 1640 * XXX BUG ALERT 1641 * The "dummyifp" code relies upon the fact 1642 * that icmp_error() touches only ifp->if_mtu. 1643 */ 1644 /*XXX*/ 1645 destifp = NULL; 1646 if (sp->req != NULL 1647 && sp->req->sav != NULL 1648 && sp->req->sav->sah != NULL) { 1649 ro = &sp->req->sav->sah->sa_route; 1650 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1651 dummyifp.if_mtu = 1652 ro->ro_rt->rt_ifp->if_mtu; 1653 dummyifp.if_mtu -= ipsechdr; 1654 destifp = &dummyifp; 1655 } 1656 } 1657 1658 key_freesp(sp); 1659 } 1660 } 1661#endif /*IPSEC*/ 1662 ipstat.ips_cantfrag++; 1663 break; 1664 1665 case ENOBUFS: 1666 type = ICMP_SOURCEQUENCH; 1667 code = 0; 1668 break; 1669 } 1670 icmp_error(mcopy, type, code, dest, destifp); 1671} 1672 1673void 1674ip_savecontrol(inp, mp, ip, m) 1675 struct inpcb *inp; 1676 struct mbuf **mp; 1677 struct ip *ip; 1678 struct mbuf *m; 1679{ 1680 1681 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1682 struct timeval tv; 1683 1684 microtime(&tv); 1685 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1686 SCM_TIMESTAMP, SOL_SOCKET); 1687 if (*mp) 1688 mp = &(*mp)->m_next; 1689 } 1690 if (inp->inp_flags & INP_RECVDSTADDR) { 1691 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1692 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1693 if (*mp) 1694 mp = &(*mp)->m_next; 1695 } 1696#ifdef notyet 1697 /* 1698 * XXX 1699 * Moving these out of udp_input() made them even more broken 1700 * than they already were. 1701 * - fenner@parc.xerox.com 1702 */ 1703 /* options were tossed already */ 1704 if (inp->inp_flags & INP_RECVOPTS) { 1705 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1706 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1707 if (*mp) 1708 mp = &(*mp)->m_next; 1709 } 1710 /* ip_srcroute doesn't do what we want here, need to fix */ 1711 if (inp->inp_flags & INP_RECVRETOPTS) { 1712 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1713 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1714 if (*mp) 1715 mp = &(*mp)->m_next; 1716 } 1717#endif 1718 if (inp->inp_flags & INP_RECVIF) { 1719 struct sockaddr_dl sdl; 1720 1721 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1722 sdl.sdl_family = AF_LINK; 1723 sdl.sdl_index = m->m_pkthdr.rcvif ? 1724 m->m_pkthdr.rcvif->if_index : 0; 1725 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1726 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1727 IP_RECVIF, IPPROTO_IP); 1728 if (*mp) 1729 mp = &(*mp)->m_next; 1730 } 1731} 1732 1733int 1734ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1735 int *name; 1736 u_int namelen; 1737 void *oldp; 1738 size_t *oldlenp; 1739 void *newp; 1740 size_t newlen; 1741{ 1742 extern int subnetsarelocal, hostzeroisbroadcast; 1743 1744 int error, old; 1745 1746 /* All sysctl names at this level are terminal. */ 1747 if (namelen != 1) 1748 return (ENOTDIR); 1749 1750 switch (name[0]) { 1751 case IPCTL_FORWARDING: 1752 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1753 case IPCTL_SENDREDIRECTS: 1754 return (sysctl_int(oldp, oldlenp, newp, newlen, 1755 &ipsendredirects)); 1756 case IPCTL_DEFTTL: 1757 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1758#ifdef notyet 1759 case IPCTL_DEFMTU: 1760 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1761#endif 1762 case IPCTL_FORWSRCRT: 1763 /* Don't allow this to change in a secure environment. */ 1764 if (securelevel > 0) 1765 return (sysctl_rdint(oldp, oldlenp, newp, 1766 ip_forwsrcrt)); 1767 else 1768 return (sysctl_int(oldp, oldlenp, newp, newlen, 1769 &ip_forwsrcrt)); 1770 case IPCTL_DIRECTEDBCAST: 1771 return (sysctl_int(oldp, oldlenp, newp, newlen, 1772 &ip_directedbcast)); 1773 case IPCTL_ALLOWSRCRT: 1774 return (sysctl_int(oldp, oldlenp, newp, newlen, 1775 &ip_allowsrcrt)); 1776 case IPCTL_SUBNETSARELOCAL: 1777 return (sysctl_int(oldp, oldlenp, newp, newlen, 1778 &subnetsarelocal)); 1779 case IPCTL_MTUDISC: 1780 error = sysctl_int(oldp, oldlenp, newp, newlen, 1781 &ip_mtudisc); 1782 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1783 ip_mtudisc_timeout_q = 1784 rt_timer_queue_create(ip_mtudisc_timeout); 1785 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1786 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1787 ip_mtudisc_timeout_q = NULL; 1788 } 1789 return error; 1790 case IPCTL_ANONPORTMIN: 1791 old = anonportmin; 1792 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1793 if (anonportmin >= anonportmax || anonportmin < 0 1794 || anonportmin > 65535 1795#ifndef IPNOPRIVPORTS 1796 || anonportmin < IPPORT_RESERVED 1797#endif 1798 ) { 1799 anonportmin = old; 1800 return (EINVAL); 1801 } 1802 return (error); 1803 case IPCTL_ANONPORTMAX: 1804 old = anonportmax; 1805 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1806 if (anonportmin >= anonportmax || anonportmax < 0 1807 || anonportmax > 65535 1808#ifndef IPNOPRIVPORTS 1809 || anonportmax < IPPORT_RESERVED 1810#endif 1811 ) { 1812 anonportmax = old; 1813 return (EINVAL); 1814 } 1815 return (error); 1816 case IPCTL_MTUDISCTIMEOUT: 1817 error = sysctl_int(oldp, oldlenp, newp, newlen, 1818 &ip_mtudisc_timeout); 1819 if (ip_mtudisc_timeout_q != NULL) 1820 rt_timer_queue_change(ip_mtudisc_timeout_q, 1821 ip_mtudisc_timeout); 1822 return (error); 1823#ifdef GATEWAY 1824 case IPCTL_MAXFLOWS: 1825 { 1826 int s; 1827 1828 error = sysctl_int(oldp, oldlenp, newp, newlen, 1829 &ip_maxflows); 1830 s = splsoftnet(); 1831 ipflow_reap(0); 1832 splx(s); 1833 return (error); 1834 } 1835#endif 1836 case IPCTL_HOSTZEROBROADCAST: 1837 return (sysctl_int(oldp, oldlenp, newp, newlen, 1838 &hostzeroisbroadcast)); 1839#if NGIF > 0 1840 case IPCTL_GIF_TTL: 1841 return(sysctl_int(oldp, oldlenp, newp, newlen, 1842 &ip_gif_ttl)); 1843#endif 1844 1845#ifndef IPNOPRIVPORTS 1846 case IPCTL_LOWPORTMIN: 1847 old = lowportmin; 1848 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmin); 1849 if (lowportmin >= lowportmax 1850 || lowportmin > IPPORT_RESERVEDMAX 1851 || lowportmin < IPPORT_RESERVEDMIN 1852 ) { 1853 lowportmin = old; 1854 return (EINVAL); 1855 } 1856 return (error); 1857 case IPCTL_LOWPORTMAX: 1858 old = lowportmax; 1859 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmax); 1860 if (lowportmin >= lowportmax 1861 || lowportmax > IPPORT_RESERVEDMAX 1862 || lowportmax < IPPORT_RESERVEDMIN 1863 ) { 1864 lowportmax = old; 1865 return (EINVAL); 1866 } 1867 return (error); 1868#endif 1869 1870 case IPCTL_MAXFRAGPACKETS: 1871 return (sysctl_int(oldp, oldlenp, newp, newlen, 1872 &ip_maxfragpackets)); 1873 1874 default: 1875 return (EOPNOTSUPP); 1876 } 1877 /* NOTREACHED */ 1878} 1879