ip_input.c revision 1.131
1/* $NetBSD: ip_input.c,v 1.131 2001/03/27 02:24:38 itojun 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" 108 109#include <sys/param.h> 110#include <sys/systm.h> 111#include <sys/malloc.h> 112#include <sys/mbuf.h> 113#include <sys/domain.h> 114#include <sys/protosw.h> 115#include <sys/socket.h> 116#include <sys/socketvar.h> 117#include <sys/errno.h> 118#include <sys/time.h> 119#include <sys/kernel.h> 120#include <sys/proc.h> 121#include <sys/pool.h> 122 123#include <uvm/uvm_extern.h> 124 125#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 = -1; 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 s = splimp(); 226 if (ipq_locked) { 227 splx(s); 228 return (0); 229 } 230 ipq_locked = 1; 231 splx(s); 232 return (1); 233} 234 235static __inline void 236ipq_unlock() 237{ 238 int s; 239 240 s = splimp(); 241 ipq_locked = 0; 242 splx(s); 243} 244 245#ifdef DIAGNOSTIC 246#define IPQ_LOCK() \ 247do { \ 248 if (ipq_lock_try() == 0) { \ 249 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 250 panic("ipq_lock"); \ 251 } \ 252} while (0) 253#define IPQ_LOCK_CHECK() \ 254do { \ 255 if (ipq_locked == 0) { \ 256 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 257 panic("ipq lock check"); \ 258 } \ 259} while (0) 260#else 261#define IPQ_LOCK() (void) ipq_lock_try() 262#define IPQ_LOCK_CHECK() /* nothing */ 263#endif 264 265#define IPQ_UNLOCK() ipq_unlock() 266 267struct pool ipqent_pool; 268 269/* 270 * We need to save the IP options in case a protocol wants to respond 271 * to an incoming packet over the same route if the packet got here 272 * using IP source routing. This allows connection establishment and 273 * maintenance when the remote end is on a network that is not known 274 * to us. 275 */ 276int ip_nhops = 0; 277static struct ip_srcrt { 278 struct in_addr dst; /* final destination */ 279 char nop; /* one NOP to align */ 280 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 281 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 282} ip_srcrt; 283 284static void save_rte __P((u_char *, struct in_addr)); 285 286/* 287 * IP initialization: fill in IP protocol switch table. 288 * All protocols not implemented in kernel go to raw IP protocol handler. 289 */ 290void 291ip_init() 292{ 293 struct protosw *pr; 294 int i; 295 296 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 297 0, NULL, NULL, M_IPQ); 298 299 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 300 if (pr == 0) 301 panic("ip_init"); 302 for (i = 0; i < IPPROTO_MAX; i++) 303 ip_protox[i] = pr - inetsw; 304 for (pr = inetdomain.dom_protosw; 305 pr < inetdomain.dom_protoswNPROTOSW; pr++) 306 if (pr->pr_domain->dom_family == PF_INET && 307 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 308 ip_protox[pr->pr_protocol] = pr - inetsw; 309 LIST_INIT(&ipq); 310 ip_id = time.tv_sec & 0xffff; 311 ipintrq.ifq_maxlen = ipqmaxlen; 312 TAILQ_INIT(&in_ifaddr); 313 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 314 M_WAITOK, &in_ifaddrhash); 315 if (ip_mtudisc != 0) 316 ip_mtudisc_timeout_q = 317 rt_timer_queue_create(ip_mtudisc_timeout); 318#ifdef GATEWAY 319 ipflow_init(); 320#endif 321 322#ifdef PFIL_HOOKS 323 /* Register our Packet Filter hook. */ 324 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 325 inet_pfil_hook.ph_af = AF_INET; 326 i = pfil_head_register(&inet_pfil_hook); 327 if (i != 0) 328 printf("ip_init: WARNING: unable to register pfil hook, " 329 "error %d\n", i); 330#endif /* PFIL_HOOKS */ 331} 332 333struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 334struct route ipforward_rt; 335 336/* 337 * IP software interrupt routine 338 */ 339void 340ipintr() 341{ 342 int s; 343 struct mbuf *m; 344 345 while (1) { 346 s = splimp(); 347 IF_DEQUEUE(&ipintrq, m); 348 splx(s); 349 if (m == 0) 350 return; 351 ip_input(m); 352 } 353} 354 355/* 356 * Ip input routine. Checksum and byte swap header. If fragmented 357 * try to reassemble. Process options. Pass to next level. 358 */ 359void 360ip_input(struct mbuf *m) 361{ 362 struct ip *ip = NULL; 363 struct ipq *fp; 364 struct in_ifaddr *ia; 365 struct ifaddr *ifa; 366 struct ipqent *ipqe; 367 int hlen = 0, mff, len; 368 int downmatch; 369 370#ifdef DIAGNOSTIC 371 if ((m->m_flags & M_PKTHDR) == 0) 372 panic("ipintr no HDR"); 373#endif 374#ifdef IPSEC 375 /* 376 * should the inner packet be considered authentic? 377 * see comment in ah4_input(). 378 */ 379 if (m) { 380 m->m_flags &= ~M_AUTHIPHDR; 381 m->m_flags &= ~M_AUTHIPDGM; 382 } 383#endif 384 /* 385 * If no IP addresses have been set yet but the interfaces 386 * are receiving, can't do anything with incoming packets yet. 387 */ 388 if (in_ifaddr.tqh_first == 0) 389 goto bad; 390 ipstat.ips_total++; 391 if (m->m_len < sizeof (struct ip) && 392 (m = m_pullup(m, sizeof (struct ip))) == 0) { 393 ipstat.ips_toosmall++; 394 return; 395 } 396 ip = mtod(m, struct ip *); 397 if (ip->ip_v != IPVERSION) { 398 ipstat.ips_badvers++; 399 goto bad; 400 } 401 hlen = ip->ip_hl << 2; 402 if (hlen < sizeof(struct ip)) { /* minimum header length */ 403 ipstat.ips_badhlen++; 404 goto bad; 405 } 406 if (hlen > m->m_len) { 407 if ((m = m_pullup(m, hlen)) == 0) { 408 ipstat.ips_badhlen++; 409 return; 410 } 411 ip = mtod(m, struct ip *); 412 } 413 414 /* 415 * RFC1122: packets with a multicast source address are 416 * not allowed. 417 */ 418 if (IN_MULTICAST(ip->ip_src.s_addr)) { 419 ipstat.ips_badaddr++; 420 goto bad; 421 } 422 423 /* 127/8 must not appear on wire - RFC1122 */ 424 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 425 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 426 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 427 ipstat.ips_badaddr++; 428 goto bad; 429 } 430 } 431 432 if (in_cksum(m, hlen) != 0) { 433 ipstat.ips_badsum++; 434 goto bad; 435 } 436 437 /* Retrieve the packet length. */ 438 len = ntohs(ip->ip_len); 439 440 /* 441 * Check for additional length bogosity 442 */ 443 if (len < hlen) { 444 ipstat.ips_badlen++; 445 goto bad; 446 } 447 448 /* 449 * Check that the amount of data in the buffers 450 * is as at least much as the IP header would have us expect. 451 * Trim mbufs if longer than we expect. 452 * Drop packet if shorter than we expect. 453 */ 454 if (m->m_pkthdr.len < len) { 455 ipstat.ips_tooshort++; 456 goto bad; 457 } 458 if (m->m_pkthdr.len > len) { 459 if (m->m_len == m->m_pkthdr.len) { 460 m->m_len = len; 461 m->m_pkthdr.len = len; 462 } else 463 m_adj(m, len - m->m_pkthdr.len); 464 } 465 466#ifdef IPSEC 467 /* ipflow (IP fast fowarding) is not compatible with IPsec. */ 468 m->m_flags &= ~M_CANFASTFWD; 469#else 470 /* 471 * Assume that we can create a fast-forward IP flow entry 472 * based on this packet. 473 */ 474 m->m_flags |= M_CANFASTFWD; 475#endif 476 477#ifdef PFIL_HOOKS 478 /* 479 * Run through list of hooks for input packets. If there are any 480 * filters which require that additional packets in the flow are 481 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 482 * Note that filters must _never_ set this flag, as another filter 483 * in the list may have previously cleared it. 484 */ 485 /* 486 * let ipfilter look at packet on the wire, 487 * not the decapsulated packet. 488 */ 489#ifdef IPSEC 490 if (!ipsec_gethist(m, NULL)) 491#else 492 if (1) 493#endif 494 { 495 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 496 PFIL_IN) != 0) 497 return; 498 if (m == NULL) 499 return; 500 ip = mtod(m, struct ip *); 501 } 502#endif /* PFIL_HOOKS */ 503 504#ifdef ALTQ 505 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 506 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 507 /* packet dropped by traffic conditioner */ 508 return; 509 } 510#endif 511 512 /* 513 * Convert fields to host representation. 514 */ 515 NTOHS(ip->ip_len); 516 NTOHS(ip->ip_off); 517 518 /* 519 * Process options and, if not destined for us, 520 * ship it on. ip_dooptions returns 1 when an 521 * error was detected (causing an icmp message 522 * to be sent and the original packet to be freed). 523 */ 524 ip_nhops = 0; /* for source routed packets */ 525 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 526 return; 527 528 /* 529 * Check our list of addresses, to see if the packet is for us. 530 * 531 * Traditional 4.4BSD did not consult IFF_UP at all. 532 * The behavior here is to treat addresses on !IFF_UP interface 533 * as not mine. 534 */ 535 downmatch = 0; 536 for (ia = IN_IFADDR_HASH(ip->ip_dst.s_addr).lh_first; 537 ia != NULL; 538 ia = ia->ia_hash.le_next) { 539 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 540 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 541 break; 542 else 543 downmatch++; 544 } 545 } 546 if (ia != NULL) 547 goto ours; 548 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 549 for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first; 550 ifa != NULL; ifa = ifa->ifa_list.tqe_next) { 551 if (ifa->ifa_addr->sa_family != AF_INET) continue; 552 ia = ifatoia(ifa); 553 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 554 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 555 /* 556 * Look for all-0's host part (old broadcast addr), 557 * either for subnet or net. 558 */ 559 ip->ip_dst.s_addr == ia->ia_subnet || 560 ip->ip_dst.s_addr == ia->ia_net) 561 goto ours; 562 /* 563 * An interface with IP address zero accepts 564 * all packets that arrive on that interface. 565 */ 566 if (in_nullhost(ia->ia_addr.sin_addr)) 567 goto ours; 568 } 569 } 570 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 571 struct in_multi *inm; 572#ifdef MROUTING 573 extern struct socket *ip_mrouter; 574 575 if (m->m_flags & M_EXT) { 576 if ((m = m_pullup(m, hlen)) == 0) { 577 ipstat.ips_toosmall++; 578 return; 579 } 580 ip = mtod(m, struct ip *); 581 } 582 583 if (ip_mrouter) { 584 /* 585 * If we are acting as a multicast router, all 586 * incoming multicast packets are passed to the 587 * kernel-level multicast forwarding function. 588 * The packet is returned (relatively) intact; if 589 * ip_mforward() returns a non-zero value, the packet 590 * must be discarded, else it may be accepted below. 591 * 592 * (The IP ident field is put in the same byte order 593 * as expected when ip_mforward() is called from 594 * ip_output().) 595 */ 596 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 597 ipstat.ips_cantforward++; 598 m_freem(m); 599 return; 600 } 601 602 /* 603 * The process-level routing demon needs to receive 604 * all multicast IGMP packets, whether or not this 605 * host belongs to their destination groups. 606 */ 607 if (ip->ip_p == IPPROTO_IGMP) 608 goto ours; 609 ipstat.ips_forward++; 610 } 611#endif 612 /* 613 * See if we belong to the destination multicast group on the 614 * arrival interface. 615 */ 616 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 617 if (inm == NULL) { 618 ipstat.ips_cantforward++; 619 m_freem(m); 620 return; 621 } 622 goto ours; 623 } 624 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 625 in_nullhost(ip->ip_dst)) 626 goto ours; 627 628 /* 629 * Not for us; forward if possible and desirable. 630 */ 631 if (ipforwarding == 0) { 632 ipstat.ips_cantforward++; 633 m_freem(m); 634 } else { 635 /* 636 * If ip_dst matched any of my address on !IFF_UP interface, 637 * and there's no IFF_UP interface that matches ip_dst, 638 * send icmp unreach. Forwarding it will result in in-kernel 639 * forwarding loop till TTL goes to 0. 640 */ 641 if (downmatch) { 642 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 643 ipstat.ips_cantforward++; 644 return; 645 } 646 ip_forward(m, 0); 647 } 648 return; 649 650ours: 651 /* 652 * If offset or IP_MF are set, must reassemble. 653 * Otherwise, nothing need be done. 654 * (We could look in the reassembly queue to see 655 * if the packet was previously fragmented, 656 * but it's not worth the time; just let them time out.) 657 */ 658 if (ip->ip_off & ~(IP_DF|IP_RF)) { 659 /* 660 * Look for queue of fragments 661 * of this datagram. 662 */ 663 IPQ_LOCK(); 664 for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next) 665 if (ip->ip_id == fp->ipq_id && 666 in_hosteq(ip->ip_src, fp->ipq_src) && 667 in_hosteq(ip->ip_dst, fp->ipq_dst) && 668 ip->ip_p == fp->ipq_p) 669 goto found; 670 fp = 0; 671found: 672 673 /* 674 * Adjust ip_len to not reflect header, 675 * set ipqe_mff if more fragments are expected, 676 * convert offset of this to bytes. 677 */ 678 ip->ip_len -= hlen; 679 mff = (ip->ip_off & IP_MF) != 0; 680 if (mff) { 681 /* 682 * Make sure that fragments have a data length 683 * that's a non-zero multiple of 8 bytes. 684 */ 685 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 686 ipstat.ips_badfrags++; 687 IPQ_UNLOCK(); 688 goto bad; 689 } 690 } 691 ip->ip_off <<= 3; 692 693 /* 694 * If datagram marked as having more fragments 695 * or if this is not the first fragment, 696 * attempt reassembly; if it succeeds, proceed. 697 */ 698 if (mff || ip->ip_off) { 699 ipstat.ips_fragments++; 700 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 701 if (ipqe == NULL) { 702 ipstat.ips_rcvmemdrop++; 703 IPQ_UNLOCK(); 704 goto bad; 705 } 706 ipqe->ipqe_mff = mff; 707 ipqe->ipqe_m = m; 708 ipqe->ipqe_ip = ip; 709 m = ip_reass(ipqe, fp); 710 if (m == 0) { 711 IPQ_UNLOCK(); 712 return; 713 } 714 ipstat.ips_reassembled++; 715 ip = mtod(m, struct ip *); 716 hlen = ip->ip_hl << 2; 717 ip->ip_len += hlen; 718 } else 719 if (fp) 720 ip_freef(fp); 721 IPQ_UNLOCK(); 722 } 723 724#ifdef IPSEC 725 /* 726 * enforce IPsec policy checking if we are seeing last header. 727 * note that we do not visit this with protocols with pcb layer 728 * code - like udp/tcp/raw ip. 729 */ 730 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 731 ipsec4_in_reject(m, NULL)) { 732 ipsecstat.in_polvio++; 733 goto bad; 734 } 735#endif 736 737 /* 738 * Switch out to protocol's input routine. 739 */ 740#if IFA_STATS 741 if (ia && ip) 742 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 743#endif 744 ipstat.ips_delivered++; 745 { 746 int off = hlen, nh = ip->ip_p; 747 748 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 749 return; 750 } 751bad: 752 m_freem(m); 753} 754 755/* 756 * Take incoming datagram fragment and try to 757 * reassemble it into whole datagram. If a chain for 758 * reassembly of this datagram already exists, then it 759 * is given as fp; otherwise have to make a chain. 760 */ 761struct mbuf * 762ip_reass(ipqe, fp) 763 struct ipqent *ipqe; 764 struct ipq *fp; 765{ 766 struct mbuf *m = ipqe->ipqe_m; 767 struct ipqent *nq, *p, *q; 768 struct ip *ip; 769 struct mbuf *t; 770 int hlen = ipqe->ipqe_ip->ip_hl << 2; 771 int i, next; 772 773 IPQ_LOCK_CHECK(); 774 775 /* 776 * Presence of header sizes in mbufs 777 * would confuse code below. 778 */ 779 m->m_data += hlen; 780 m->m_len -= hlen; 781 782 /* 783 * If first fragment to arrive, create a reassembly queue. 784 */ 785 if (fp == 0) { 786 /* 787 * Enforce upper bound on number of fragmented packets 788 * for which we attempt reassembly; 789 * If maxfrag is 0, never accept fragments. 790 * If maxfrag is -1, accept all fragments without limitation. 791 */ 792 if (ip_maxfragpackets < 0) 793 ; 794 else if (ip_nfragpackets >= ip_maxfragpackets) 795 goto dropfrag; 796 ip_nfragpackets++; 797 MALLOC(fp, struct ipq *, sizeof (struct ipq), 798 M_FTABLE, M_NOWAIT); 799 if (fp == NULL) 800 goto dropfrag; 801 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 802 fp->ipq_ttl = IPFRAGTTL; 803 fp->ipq_p = ipqe->ipqe_ip->ip_p; 804 fp->ipq_id = ipqe->ipqe_ip->ip_id; 805 LIST_INIT(&fp->ipq_fragq); 806 fp->ipq_src = ipqe->ipqe_ip->ip_src; 807 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 808 p = NULL; 809 goto insert; 810 } 811 812 /* 813 * Find a segment which begins after this one does. 814 */ 815 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 816 p = q, q = q->ipqe_q.le_next) 817 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 818 break; 819 820 /* 821 * If there is a preceding segment, it may provide some of 822 * our data already. If so, drop the data from the incoming 823 * segment. If it provides all of our data, drop us. 824 */ 825 if (p != NULL) { 826 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 827 ipqe->ipqe_ip->ip_off; 828 if (i > 0) { 829 if (i >= ipqe->ipqe_ip->ip_len) 830 goto dropfrag; 831 m_adj(ipqe->ipqe_m, i); 832 ipqe->ipqe_ip->ip_off += i; 833 ipqe->ipqe_ip->ip_len -= i; 834 } 835 } 836 837 /* 838 * While we overlap succeeding segments trim them or, 839 * if they are completely covered, dequeue them. 840 */ 841 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 842 q->ipqe_ip->ip_off; q = nq) { 843 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 844 q->ipqe_ip->ip_off; 845 if (i < q->ipqe_ip->ip_len) { 846 q->ipqe_ip->ip_len -= i; 847 q->ipqe_ip->ip_off += i; 848 m_adj(q->ipqe_m, i); 849 break; 850 } 851 nq = q->ipqe_q.le_next; 852 m_freem(q->ipqe_m); 853 LIST_REMOVE(q, ipqe_q); 854 pool_put(&ipqent_pool, q); 855 } 856 857insert: 858 /* 859 * Stick new segment in its place; 860 * check for complete reassembly. 861 */ 862 if (p == NULL) { 863 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 864 } else { 865 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 866 } 867 next = 0; 868 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 869 p = q, q = q->ipqe_q.le_next) { 870 if (q->ipqe_ip->ip_off != next) 871 return (0); 872 next += q->ipqe_ip->ip_len; 873 } 874 if (p->ipqe_mff) 875 return (0); 876 877 /* 878 * Reassembly is complete. Check for a bogus message size and 879 * concatenate fragments. 880 */ 881 q = fp->ipq_fragq.lh_first; 882 ip = q->ipqe_ip; 883 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 884 ipstat.ips_toolong++; 885 ip_freef(fp); 886 return (0); 887 } 888 m = q->ipqe_m; 889 t = m->m_next; 890 m->m_next = 0; 891 m_cat(m, t); 892 nq = q->ipqe_q.le_next; 893 pool_put(&ipqent_pool, q); 894 for (q = nq; q != NULL; q = nq) { 895 t = q->ipqe_m; 896 nq = q->ipqe_q.le_next; 897 pool_put(&ipqent_pool, q); 898 m_cat(m, t); 899 } 900 901 /* 902 * Create header for new ip packet by 903 * modifying header of first packet; 904 * dequeue and discard fragment reassembly header. 905 * Make header visible. 906 */ 907 ip->ip_len = next; 908 ip->ip_src = fp->ipq_src; 909 ip->ip_dst = fp->ipq_dst; 910 LIST_REMOVE(fp, ipq_q); 911 FREE(fp, M_FTABLE); 912 ip_nfragpackets--; 913 m->m_len += (ip->ip_hl << 2); 914 m->m_data -= (ip->ip_hl << 2); 915 /* some debugging cruft by sklower, below, will go away soon */ 916 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 917 int plen = 0; 918 for (t = m; t; t = t->m_next) 919 plen += t->m_len; 920 m->m_pkthdr.len = plen; 921 } 922 return (m); 923 924dropfrag: 925 ipstat.ips_fragdropped++; 926 m_freem(m); 927 pool_put(&ipqent_pool, ipqe); 928 return (0); 929} 930 931/* 932 * Free a fragment reassembly header and all 933 * associated datagrams. 934 */ 935void 936ip_freef(fp) 937 struct ipq *fp; 938{ 939 struct ipqent *q, *p; 940 941 IPQ_LOCK_CHECK(); 942 943 for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { 944 p = q->ipqe_q.le_next; 945 m_freem(q->ipqe_m); 946 LIST_REMOVE(q, ipqe_q); 947 pool_put(&ipqent_pool, q); 948 } 949 LIST_REMOVE(fp, ipq_q); 950 FREE(fp, M_FTABLE); 951 ip_nfragpackets--; 952} 953 954/* 955 * IP timer processing; 956 * if a timer expires on a reassembly 957 * queue, discard it. 958 */ 959void 960ip_slowtimo() 961{ 962 struct ipq *fp, *nfp; 963 int s = splsoftnet(); 964 965 IPQ_LOCK(); 966 for (fp = ipq.lh_first; fp != NULL; fp = nfp) { 967 nfp = fp->ipq_q.le_next; 968 if (--fp->ipq_ttl == 0) { 969 ipstat.ips_fragtimeout++; 970 ip_freef(fp); 971 } 972 } 973 /* 974 * If we are over the maximum number of fragments 975 * (due to the limit being lowered), drain off 976 * enough to get down to the new limit. 977 */ 978 if (ip_maxfragpackets < 0) 979 ; 980 else { 981 while (ip_nfragpackets > ip_maxfragpackets && ipq.lh_first) 982 ip_freef(ipq.lh_first); 983 } 984 IPQ_UNLOCK(); 985#ifdef GATEWAY 986 ipflow_slowtimo(); 987#endif 988 splx(s); 989} 990 991/* 992 * Drain off all datagram fragments. 993 */ 994void 995ip_drain() 996{ 997 998 /* 999 * We may be called from a device's interrupt context. If 1000 * the ipq is already busy, just bail out now. 1001 */ 1002 if (ipq_lock_try() == 0) 1003 return; 1004 1005 while (ipq.lh_first != NULL) { 1006 ipstat.ips_fragdropped++; 1007 ip_freef(ipq.lh_first); 1008 } 1009 1010 IPQ_UNLOCK(); 1011} 1012 1013/* 1014 * Do option processing on a datagram, 1015 * possibly discarding it if bad options are encountered, 1016 * or forwarding it if source-routed. 1017 * Returns 1 if packet has been forwarded/freed, 1018 * 0 if the packet should be processed further. 1019 */ 1020int 1021ip_dooptions(m) 1022 struct mbuf *m; 1023{ 1024 struct ip *ip = mtod(m, struct ip *); 1025 u_char *cp, *cp0; 1026 struct ip_timestamp *ipt; 1027 struct in_ifaddr *ia; 1028 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1029 struct in_addr dst; 1030 n_time ntime; 1031 1032 dst = ip->ip_dst; 1033 cp = (u_char *)(ip + 1); 1034 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1035 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1036 opt = cp[IPOPT_OPTVAL]; 1037 if (opt == IPOPT_EOL) 1038 break; 1039 if (opt == IPOPT_NOP) 1040 optlen = 1; 1041 else { 1042 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1043 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1044 goto bad; 1045 } 1046 optlen = cp[IPOPT_OLEN]; 1047 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1048 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1049 goto bad; 1050 } 1051 } 1052 switch (opt) { 1053 1054 default: 1055 break; 1056 1057 /* 1058 * Source routing with record. 1059 * Find interface with current destination address. 1060 * If none on this machine then drop if strictly routed, 1061 * or do nothing if loosely routed. 1062 * Record interface address and bring up next address 1063 * component. If strictly routed make sure next 1064 * address is on directly accessible net. 1065 */ 1066 case IPOPT_LSRR: 1067 case IPOPT_SSRR: 1068 if (ip_allowsrcrt == 0) { 1069 type = ICMP_UNREACH; 1070 code = ICMP_UNREACH_NET_PROHIB; 1071 goto bad; 1072 } 1073 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1074 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1075 goto bad; 1076 } 1077 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1078 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1079 goto bad; 1080 } 1081 ipaddr.sin_addr = ip->ip_dst; 1082 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1083 if (ia == 0) { 1084 if (opt == IPOPT_SSRR) { 1085 type = ICMP_UNREACH; 1086 code = ICMP_UNREACH_SRCFAIL; 1087 goto bad; 1088 } 1089 /* 1090 * Loose routing, and not at next destination 1091 * yet; nothing to do except forward. 1092 */ 1093 break; 1094 } 1095 off--; /* 0 origin */ 1096 if ((off + sizeof(struct in_addr)) > optlen) { 1097 /* 1098 * End of source route. Should be for us. 1099 */ 1100 save_rte(cp, ip->ip_src); 1101 break; 1102 } 1103 /* 1104 * locate outgoing interface 1105 */ 1106 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1107 sizeof(ipaddr.sin_addr)); 1108 if (opt == IPOPT_SSRR) 1109 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1110 else 1111 ia = ip_rtaddr(ipaddr.sin_addr); 1112 if (ia == 0) { 1113 type = ICMP_UNREACH; 1114 code = ICMP_UNREACH_SRCFAIL; 1115 goto bad; 1116 } 1117 ip->ip_dst = ipaddr.sin_addr; 1118 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1119 (caddr_t)(cp + off), sizeof(struct in_addr)); 1120 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1121 /* 1122 * Let ip_intr's mcast routing check handle mcast pkts 1123 */ 1124 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1125 break; 1126 1127 case IPOPT_RR: 1128 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1129 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1130 goto bad; 1131 } 1132 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1133 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1134 goto bad; 1135 } 1136 /* 1137 * If no space remains, ignore. 1138 */ 1139 off--; /* 0 origin */ 1140 if ((off + sizeof(struct in_addr)) > optlen) 1141 break; 1142 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1143 sizeof(ipaddr.sin_addr)); 1144 /* 1145 * locate outgoing interface; if we're the destination, 1146 * use the incoming interface (should be same). 1147 */ 1148 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1149 == NULL && 1150 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1151 type = ICMP_UNREACH; 1152 code = ICMP_UNREACH_HOST; 1153 goto bad; 1154 } 1155 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1156 (caddr_t)(cp + off), sizeof(struct in_addr)); 1157 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1158 break; 1159 1160 case IPOPT_TS: 1161 code = cp - (u_char *)ip; 1162 ipt = (struct ip_timestamp *)cp; 1163 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1164 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1165 goto bad; 1166 } 1167 if (ipt->ipt_ptr < 5) { 1168 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1169 goto bad; 1170 } 1171 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1172 if (++ipt->ipt_oflw == 0) { 1173 code = (u_char *)&ipt->ipt_ptr - 1174 (u_char *)ip; 1175 goto bad; 1176 } 1177 break; 1178 } 1179 cp0 = (cp + ipt->ipt_ptr - 1); 1180 switch (ipt->ipt_flg) { 1181 1182 case IPOPT_TS_TSONLY: 1183 break; 1184 1185 case IPOPT_TS_TSANDADDR: 1186 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1187 sizeof(struct in_addr) > ipt->ipt_len) { 1188 code = (u_char *)&ipt->ipt_ptr - 1189 (u_char *)ip; 1190 goto bad; 1191 } 1192 ipaddr.sin_addr = dst; 1193 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1194 m->m_pkthdr.rcvif)); 1195 if (ia == 0) 1196 continue; 1197 bcopy(&ia->ia_addr.sin_addr, 1198 cp0, sizeof(struct in_addr)); 1199 ipt->ipt_ptr += sizeof(struct in_addr); 1200 break; 1201 1202 case IPOPT_TS_PRESPEC: 1203 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1204 sizeof(struct in_addr) > ipt->ipt_len) { 1205 code = (u_char *)&ipt->ipt_ptr - 1206 (u_char *)ip; 1207 goto bad; 1208 } 1209 bcopy(cp0, &ipaddr.sin_addr, 1210 sizeof(struct in_addr)); 1211 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1212 == NULL) 1213 continue; 1214 ipt->ipt_ptr += sizeof(struct in_addr); 1215 break; 1216 1217 default: 1218 /* XXX can't take &ipt->ipt_flg */ 1219 code = (u_char *)&ipt->ipt_ptr - 1220 (u_char *)ip + 1; 1221 goto bad; 1222 } 1223 ntime = iptime(); 1224 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1225 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1226 sizeof(n_time)); 1227 ipt->ipt_ptr += sizeof(n_time); 1228 } 1229 } 1230 if (forward) { 1231 if (ip_forwsrcrt == 0) { 1232 type = ICMP_UNREACH; 1233 code = ICMP_UNREACH_SRCFAIL; 1234 goto bad; 1235 } 1236 ip_forward(m, 1); 1237 return (1); 1238 } 1239 return (0); 1240bad: 1241 icmp_error(m, type, code, 0, 0); 1242 ipstat.ips_badoptions++; 1243 return (1); 1244} 1245 1246/* 1247 * Given address of next destination (final or next hop), 1248 * return internet address info of interface to be used to get there. 1249 */ 1250struct in_ifaddr * 1251ip_rtaddr(dst) 1252 struct in_addr dst; 1253{ 1254 struct sockaddr_in *sin; 1255 1256 sin = satosin(&ipforward_rt.ro_dst); 1257 1258 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1259 if (ipforward_rt.ro_rt) { 1260 RTFREE(ipforward_rt.ro_rt); 1261 ipforward_rt.ro_rt = 0; 1262 } 1263 sin->sin_family = AF_INET; 1264 sin->sin_len = sizeof(*sin); 1265 sin->sin_addr = dst; 1266 1267 rtalloc(&ipforward_rt); 1268 } 1269 if (ipforward_rt.ro_rt == 0) 1270 return ((struct in_ifaddr *)0); 1271 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1272} 1273 1274/* 1275 * Save incoming source route for use in replies, 1276 * to be picked up later by ip_srcroute if the receiver is interested. 1277 */ 1278void 1279save_rte(option, dst) 1280 u_char *option; 1281 struct in_addr dst; 1282{ 1283 unsigned olen; 1284 1285 olen = option[IPOPT_OLEN]; 1286#ifdef DIAGNOSTIC 1287 if (ipprintfs) 1288 printf("save_rte: olen %d\n", olen); 1289#endif /* 0 */ 1290 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1291 return; 1292 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1293 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1294 ip_srcrt.dst = dst; 1295} 1296 1297/* 1298 * Retrieve incoming source route for use in replies, 1299 * in the same form used by setsockopt. 1300 * The first hop is placed before the options, will be removed later. 1301 */ 1302struct mbuf * 1303ip_srcroute() 1304{ 1305 struct in_addr *p, *q; 1306 struct mbuf *m; 1307 1308 if (ip_nhops == 0) 1309 return ((struct mbuf *)0); 1310 m = m_get(M_DONTWAIT, MT_SOOPTS); 1311 if (m == 0) 1312 return ((struct mbuf *)0); 1313 1314#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1315 1316 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1317 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1318 OPTSIZ; 1319#ifdef DIAGNOSTIC 1320 if (ipprintfs) 1321 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1322#endif 1323 1324 /* 1325 * First save first hop for return route 1326 */ 1327 p = &ip_srcrt.route[ip_nhops - 1]; 1328 *(mtod(m, struct in_addr *)) = *p--; 1329#ifdef DIAGNOSTIC 1330 if (ipprintfs) 1331 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1332#endif 1333 1334 /* 1335 * Copy option fields and padding (nop) to mbuf. 1336 */ 1337 ip_srcrt.nop = IPOPT_NOP; 1338 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1339 bcopy((caddr_t)&ip_srcrt.nop, 1340 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1341 q = (struct in_addr *)(mtod(m, caddr_t) + 1342 sizeof(struct in_addr) + OPTSIZ); 1343#undef OPTSIZ 1344 /* 1345 * Record return path as an IP source route, 1346 * reversing the path (pointers are now aligned). 1347 */ 1348 while (p >= ip_srcrt.route) { 1349#ifdef DIAGNOSTIC 1350 if (ipprintfs) 1351 printf(" %x", ntohl(q->s_addr)); 1352#endif 1353 *q++ = *p--; 1354 } 1355 /* 1356 * Last hop goes to final destination. 1357 */ 1358 *q = ip_srcrt.dst; 1359#ifdef DIAGNOSTIC 1360 if (ipprintfs) 1361 printf(" %x\n", ntohl(q->s_addr)); 1362#endif 1363 return (m); 1364} 1365 1366/* 1367 * Strip out IP options, at higher 1368 * level protocol in the kernel. 1369 * Second argument is buffer to which options 1370 * will be moved, and return value is their length. 1371 * XXX should be deleted; last arg currently ignored. 1372 */ 1373void 1374ip_stripoptions(m, mopt) 1375 struct mbuf *m; 1376 struct mbuf *mopt; 1377{ 1378 int i; 1379 struct ip *ip = mtod(m, struct ip *); 1380 caddr_t opts; 1381 int olen; 1382 1383 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1384 opts = (caddr_t)(ip + 1); 1385 i = m->m_len - (sizeof (struct ip) + olen); 1386 bcopy(opts + olen, opts, (unsigned)i); 1387 m->m_len -= olen; 1388 if (m->m_flags & M_PKTHDR) 1389 m->m_pkthdr.len -= olen; 1390 ip->ip_len -= olen; 1391 ip->ip_hl = sizeof (struct ip) >> 2; 1392} 1393 1394int inetctlerrmap[PRC_NCMDS] = { 1395 0, 0, 0, 0, 1396 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1397 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1398 EMSGSIZE, EHOSTUNREACH, 0, 0, 1399 0, 0, 0, 0, 1400 ENOPROTOOPT 1401}; 1402 1403/* 1404 * Forward a packet. If some error occurs return the sender 1405 * an icmp packet. Note we can't always generate a meaningful 1406 * icmp message because icmp doesn't have a large enough repertoire 1407 * of codes and types. 1408 * 1409 * If not forwarding, just drop the packet. This could be confusing 1410 * if ipforwarding was zero but some routing protocol was advancing 1411 * us as a gateway to somewhere. However, we must let the routing 1412 * protocol deal with that. 1413 * 1414 * The srcrt parameter indicates whether the packet is being forwarded 1415 * via a source route. 1416 */ 1417void 1418ip_forward(m, srcrt) 1419 struct mbuf *m; 1420 int srcrt; 1421{ 1422 struct ip *ip = mtod(m, struct ip *); 1423 struct sockaddr_in *sin; 1424 struct rtentry *rt; 1425 int error, type = 0, code = 0; 1426 struct mbuf *mcopy; 1427 n_long dest; 1428 struct ifnet *destifp; 1429#ifdef IPSEC 1430 struct ifnet dummyifp; 1431#endif 1432 1433 dest = 0; 1434#ifdef DIAGNOSTIC 1435 if (ipprintfs) 1436 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1437 ntohl(ip->ip_src.s_addr), 1438 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1439#endif 1440 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1441 ipstat.ips_cantforward++; 1442 m_freem(m); 1443 return; 1444 } 1445 if (ip->ip_ttl <= IPTTLDEC) { 1446 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1447 return; 1448 } 1449 ip->ip_ttl -= IPTTLDEC; 1450 1451 sin = satosin(&ipforward_rt.ro_dst); 1452 if ((rt = ipforward_rt.ro_rt) == 0 || 1453 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1454 if (ipforward_rt.ro_rt) { 1455 RTFREE(ipforward_rt.ro_rt); 1456 ipforward_rt.ro_rt = 0; 1457 } 1458 sin->sin_family = AF_INET; 1459 sin->sin_len = sizeof(struct sockaddr_in); 1460 sin->sin_addr = ip->ip_dst; 1461 1462 rtalloc(&ipforward_rt); 1463 if (ipforward_rt.ro_rt == 0) { 1464 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1465 return; 1466 } 1467 rt = ipforward_rt.ro_rt; 1468 } 1469 1470 /* 1471 * Save at most 68 bytes of the packet in case 1472 * we need to generate an ICMP message to the src. 1473 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1474 */ 1475 mcopy = m_copym(m, 0, imin((int)ip->ip_len, 68), M_DONTWAIT); 1476 if (mcopy) 1477 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1478 1479 /* 1480 * If forwarding packet using same interface that it came in on, 1481 * perhaps should send a redirect to sender to shortcut a hop. 1482 * Only send redirect if source is sending directly to us, 1483 * and if packet was not source routed (or has any options). 1484 * Also, don't send redirect if forwarding using a default route 1485 * or a route modified by a redirect. 1486 */ 1487 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1488 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1489 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1490 ipsendredirects && !srcrt) { 1491 if (rt->rt_ifa && 1492 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1493 ifatoia(rt->rt_ifa)->ia_subnet) { 1494 if (rt->rt_flags & RTF_GATEWAY) 1495 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1496 else 1497 dest = ip->ip_dst.s_addr; 1498 /* 1499 * Router requirements says to only send host 1500 * redirects. 1501 */ 1502 type = ICMP_REDIRECT; 1503 code = ICMP_REDIRECT_HOST; 1504#ifdef DIAGNOSTIC 1505 if (ipprintfs) 1506 printf("redirect (%d) to %x\n", code, 1507 (u_int32_t)dest); 1508#endif 1509 } 1510 } 1511 1512#ifdef IPSEC 1513 /* Don't lookup socket in forwading case */ 1514 (void)ipsec_setsocket(m, NULL); 1515#endif 1516 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1517 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1518 if (error) 1519 ipstat.ips_cantforward++; 1520 else { 1521 ipstat.ips_forward++; 1522 if (type) 1523 ipstat.ips_redirectsent++; 1524 else { 1525 if (mcopy) { 1526#ifdef GATEWAY 1527 if (mcopy->m_flags & M_CANFASTFWD) 1528 ipflow_create(&ipforward_rt, mcopy); 1529#endif 1530 m_freem(mcopy); 1531 } 1532 return; 1533 } 1534 } 1535 if (mcopy == NULL) 1536 return; 1537 destifp = NULL; 1538 1539 switch (error) { 1540 1541 case 0: /* forwarded, but need redirect */ 1542 /* type, code set above */ 1543 break; 1544 1545 case ENETUNREACH: /* shouldn't happen, checked above */ 1546 case EHOSTUNREACH: 1547 case ENETDOWN: 1548 case EHOSTDOWN: 1549 default: 1550 type = ICMP_UNREACH; 1551 code = ICMP_UNREACH_HOST; 1552 break; 1553 1554 case EMSGSIZE: 1555 type = ICMP_UNREACH; 1556 code = ICMP_UNREACH_NEEDFRAG; 1557#ifndef IPSEC 1558 if (ipforward_rt.ro_rt) 1559 destifp = ipforward_rt.ro_rt->rt_ifp; 1560#else 1561 /* 1562 * If the packet is routed over IPsec tunnel, tell the 1563 * originator the tunnel MTU. 1564 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1565 * XXX quickhack!!! 1566 */ 1567 if (ipforward_rt.ro_rt) { 1568 struct secpolicy *sp; 1569 int ipsecerror; 1570 size_t ipsechdr; 1571 struct route *ro; 1572 1573 sp = ipsec4_getpolicybyaddr(mcopy, 1574 IPSEC_DIR_OUTBOUND, 1575 IP_FORWARDING, 1576 &ipsecerror); 1577 1578 if (sp == NULL) 1579 destifp = ipforward_rt.ro_rt->rt_ifp; 1580 else { 1581 /* count IPsec header size */ 1582 ipsechdr = ipsec4_hdrsiz(mcopy, 1583 IPSEC_DIR_OUTBOUND, 1584 NULL); 1585 1586 /* 1587 * find the correct route for outer IPv4 1588 * header, compute tunnel MTU. 1589 * 1590 * XXX BUG ALERT 1591 * The "dummyifp" code relies upon the fact 1592 * that icmp_error() touches only ifp->if_mtu. 1593 */ 1594 /*XXX*/ 1595 destifp = NULL; 1596 if (sp->req != NULL 1597 && sp->req->sav != NULL 1598 && sp->req->sav->sah != NULL) { 1599 ro = &sp->req->sav->sah->sa_route; 1600 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1601 dummyifp.if_mtu = 1602 ro->ro_rt->rt_ifp->if_mtu; 1603 dummyifp.if_mtu -= ipsechdr; 1604 destifp = &dummyifp; 1605 } 1606 } 1607 1608 key_freesp(sp); 1609 } 1610 } 1611#endif /*IPSEC*/ 1612 ipstat.ips_cantfrag++; 1613 break; 1614 1615 case ENOBUFS: 1616 type = ICMP_SOURCEQUENCH; 1617 code = 0; 1618 break; 1619 } 1620 icmp_error(mcopy, type, code, dest, destifp); 1621} 1622 1623void 1624ip_savecontrol(inp, mp, ip, m) 1625 struct inpcb *inp; 1626 struct mbuf **mp; 1627 struct ip *ip; 1628 struct mbuf *m; 1629{ 1630 1631 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1632 struct timeval tv; 1633 1634 microtime(&tv); 1635 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1636 SCM_TIMESTAMP, SOL_SOCKET); 1637 if (*mp) 1638 mp = &(*mp)->m_next; 1639 } 1640 if (inp->inp_flags & INP_RECVDSTADDR) { 1641 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1642 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1643 if (*mp) 1644 mp = &(*mp)->m_next; 1645 } 1646#ifdef notyet 1647 /* 1648 * XXX 1649 * Moving these out of udp_input() made them even more broken 1650 * than they already were. 1651 * - fenner@parc.xerox.com 1652 */ 1653 /* options were tossed already */ 1654 if (inp->inp_flags & INP_RECVOPTS) { 1655 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1656 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1657 if (*mp) 1658 mp = &(*mp)->m_next; 1659 } 1660 /* ip_srcroute doesn't do what we want here, need to fix */ 1661 if (inp->inp_flags & INP_RECVRETOPTS) { 1662 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1663 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1664 if (*mp) 1665 mp = &(*mp)->m_next; 1666 } 1667#endif 1668 if (inp->inp_flags & INP_RECVIF) { 1669 struct sockaddr_dl sdl; 1670 1671 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1672 sdl.sdl_family = AF_LINK; 1673 sdl.sdl_index = m->m_pkthdr.rcvif ? 1674 m->m_pkthdr.rcvif->if_index : 0; 1675 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1676 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1677 IP_RECVIF, IPPROTO_IP); 1678 if (*mp) 1679 mp = &(*mp)->m_next; 1680 } 1681} 1682 1683int 1684ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1685 int *name; 1686 u_int namelen; 1687 void *oldp; 1688 size_t *oldlenp; 1689 void *newp; 1690 size_t newlen; 1691{ 1692 extern int subnetsarelocal, hostzeroisbroadcast; 1693 1694 int error, old; 1695 1696 /* All sysctl names at this level are terminal. */ 1697 if (namelen != 1) 1698 return (ENOTDIR); 1699 1700 switch (name[0]) { 1701 case IPCTL_FORWARDING: 1702 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1703 case IPCTL_SENDREDIRECTS: 1704 return (sysctl_int(oldp, oldlenp, newp, newlen, 1705 &ipsendredirects)); 1706 case IPCTL_DEFTTL: 1707 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1708#ifdef notyet 1709 case IPCTL_DEFMTU: 1710 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1711#endif 1712 case IPCTL_FORWSRCRT: 1713 /* Don't allow this to change in a secure environment. */ 1714 if (securelevel > 0) 1715 return (sysctl_rdint(oldp, oldlenp, newp, 1716 ip_forwsrcrt)); 1717 else 1718 return (sysctl_int(oldp, oldlenp, newp, newlen, 1719 &ip_forwsrcrt)); 1720 case IPCTL_DIRECTEDBCAST: 1721 return (sysctl_int(oldp, oldlenp, newp, newlen, 1722 &ip_directedbcast)); 1723 case IPCTL_ALLOWSRCRT: 1724 return (sysctl_int(oldp, oldlenp, newp, newlen, 1725 &ip_allowsrcrt)); 1726 case IPCTL_SUBNETSARELOCAL: 1727 return (sysctl_int(oldp, oldlenp, newp, newlen, 1728 &subnetsarelocal)); 1729 case IPCTL_MTUDISC: 1730 error = sysctl_int(oldp, oldlenp, newp, newlen, 1731 &ip_mtudisc); 1732 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1733 ip_mtudisc_timeout_q = 1734 rt_timer_queue_create(ip_mtudisc_timeout); 1735 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1736 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1737 ip_mtudisc_timeout_q = NULL; 1738 } 1739 return error; 1740 case IPCTL_ANONPORTMIN: 1741 old = anonportmin; 1742 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1743 if (anonportmin >= anonportmax || anonportmin < 0 1744 || anonportmin > 65535 1745#ifndef IPNOPRIVPORTS 1746 || anonportmin < IPPORT_RESERVED 1747#endif 1748 ) { 1749 anonportmin = old; 1750 return (EINVAL); 1751 } 1752 return (error); 1753 case IPCTL_ANONPORTMAX: 1754 old = anonportmax; 1755 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1756 if (anonportmin >= anonportmax || anonportmax < 0 1757 || anonportmax > 65535 1758#ifndef IPNOPRIVPORTS 1759 || anonportmax < IPPORT_RESERVED 1760#endif 1761 ) { 1762 anonportmax = old; 1763 return (EINVAL); 1764 } 1765 return (error); 1766 case IPCTL_MTUDISCTIMEOUT: 1767 error = sysctl_int(oldp, oldlenp, newp, newlen, 1768 &ip_mtudisc_timeout); 1769 if (ip_mtudisc_timeout_q != NULL) 1770 rt_timer_queue_change(ip_mtudisc_timeout_q, 1771 ip_mtudisc_timeout); 1772 return (error); 1773#ifdef GATEWAY 1774 case IPCTL_MAXFLOWS: 1775 { 1776 int s; 1777 1778 error = sysctl_int(oldp, oldlenp, newp, newlen, 1779 &ip_maxflows); 1780 s = splsoftnet(); 1781 ipflow_reap(0); 1782 splx(s); 1783 return (error); 1784 } 1785#endif 1786 case IPCTL_HOSTZEROBROADCAST: 1787 return (sysctl_int(oldp, oldlenp, newp, newlen, 1788 &hostzeroisbroadcast)); 1789#if NGIF > 0 1790 case IPCTL_GIF_TTL: 1791 return(sysctl_int(oldp, oldlenp, newp, newlen, 1792 &ip_gif_ttl)); 1793#endif 1794 1795#ifndef IPNOPRIVPORTS 1796 case IPCTL_LOWPORTMIN: 1797 old = lowportmin; 1798 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmin); 1799 if (lowportmin >= lowportmax 1800 || lowportmin > IPPORT_RESERVEDMAX 1801 || lowportmin < IPPORT_RESERVEDMIN 1802 ) { 1803 lowportmin = old; 1804 return (EINVAL); 1805 } 1806 return (error); 1807 case IPCTL_LOWPORTMAX: 1808 old = lowportmax; 1809 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmax); 1810 if (lowportmin >= lowportmax 1811 || lowportmax > IPPORT_RESERVEDMAX 1812 || lowportmax < IPPORT_RESERVEDMIN 1813 ) { 1814 lowportmax = old; 1815 return (EINVAL); 1816 } 1817 return (error); 1818#endif 1819 1820 case IPCTL_MAXFRAGPACKETS: 1821 return (sysctl_int(oldp, oldlenp, newp, newlen, 1822 &ip_maxfragpackets)); 1823 1824 default: 1825 return (EOPNOTSUPP); 1826 } 1827 /* NOTREACHED */ 1828} 1829