ip_input.c revision 1.130
1/* $NetBSD: ip_input.c,v 1.130 2001/03/02 04:26:10 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; 214 215static __inline int ipq_lock_try __P((void)); 216static __inline void ipq_unlock __P((void)); 217 218static __inline int 219ipq_lock_try() 220{ 221 int s; 222 223 s = splimp(); 224 if (ipq_locked) { 225 splx(s); 226 return (0); 227 } 228 ipq_locked = 1; 229 splx(s); 230 return (1); 231} 232 233static __inline void 234ipq_unlock() 235{ 236 int s; 237 238 s = splimp(); 239 ipq_locked = 0; 240 splx(s); 241} 242 243#ifdef DIAGNOSTIC 244#define IPQ_LOCK() \ 245do { \ 246 if (ipq_lock_try() == 0) { \ 247 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 248 panic("ipq_lock"); \ 249 } \ 250} while (0) 251#define IPQ_LOCK_CHECK() \ 252do { \ 253 if (ipq_locked == 0) { \ 254 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 255 panic("ipq lock check"); \ 256 } \ 257} while (0) 258#else 259#define IPQ_LOCK() (void) ipq_lock_try() 260#define IPQ_LOCK_CHECK() /* nothing */ 261#endif 262 263#define IPQ_UNLOCK() ipq_unlock() 264 265struct pool ipqent_pool; 266 267/* 268 * We need to save the IP options in case a protocol wants to respond 269 * to an incoming packet over the same route if the packet got here 270 * using IP source routing. This allows connection establishment and 271 * maintenance when the remote end is on a network that is not known 272 * to us. 273 */ 274int ip_nhops = 0; 275static struct ip_srcrt { 276 struct in_addr dst; /* final destination */ 277 char nop; /* one NOP to align */ 278 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 279 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 280} ip_srcrt; 281 282static void save_rte __P((u_char *, struct in_addr)); 283 284/* 285 * IP initialization: fill in IP protocol switch table. 286 * All protocols not implemented in kernel go to raw IP protocol handler. 287 */ 288void 289ip_init() 290{ 291 struct protosw *pr; 292 int i; 293 294 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 295 0, NULL, NULL, M_IPQ); 296 297 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 298 if (pr == 0) 299 panic("ip_init"); 300 for (i = 0; i < IPPROTO_MAX; i++) 301 ip_protox[i] = pr - inetsw; 302 for (pr = inetdomain.dom_protosw; 303 pr < inetdomain.dom_protoswNPROTOSW; pr++) 304 if (pr->pr_domain->dom_family == PF_INET && 305 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 306 ip_protox[pr->pr_protocol] = pr - inetsw; 307 LIST_INIT(&ipq); 308 ip_id = time.tv_sec & 0xffff; 309 ipintrq.ifq_maxlen = ipqmaxlen; 310 TAILQ_INIT(&in_ifaddr); 311 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 312 M_WAITOK, &in_ifaddrhash); 313 if (ip_mtudisc != 0) 314 ip_mtudisc_timeout_q = 315 rt_timer_queue_create(ip_mtudisc_timeout); 316#ifdef GATEWAY 317 ipflow_init(); 318#endif 319 320#ifdef PFIL_HOOKS 321 /* Register our Packet Filter hook. */ 322 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 323 inet_pfil_hook.ph_af = AF_INET; 324 i = pfil_head_register(&inet_pfil_hook); 325 if (i != 0) 326 printf("ip_init: WARNING: unable to register pfil hook, " 327 "error %d\n", i); 328#endif /* PFIL_HOOKS */ 329} 330 331struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 332struct route ipforward_rt; 333 334/* 335 * IP software interrupt routine 336 */ 337void 338ipintr() 339{ 340 int s; 341 struct mbuf *m; 342 343 while (1) { 344 s = splimp(); 345 IF_DEQUEUE(&ipintrq, m); 346 splx(s); 347 if (m == 0) 348 return; 349 ip_input(m); 350 } 351} 352 353/* 354 * Ip input routine. Checksum and byte swap header. If fragmented 355 * try to reassemble. Process options. Pass to next level. 356 */ 357void 358ip_input(struct mbuf *m) 359{ 360 struct ip *ip = NULL; 361 struct ipq *fp; 362 struct in_ifaddr *ia; 363 struct ifaddr *ifa; 364 struct ipqent *ipqe; 365 int hlen = 0, mff, len; 366 int downmatch; 367 368#ifdef DIAGNOSTIC 369 if ((m->m_flags & M_PKTHDR) == 0) 370 panic("ipintr no HDR"); 371#endif 372#ifdef IPSEC 373 /* 374 * should the inner packet be considered authentic? 375 * see comment in ah4_input(). 376 */ 377 if (m) { 378 m->m_flags &= ~M_AUTHIPHDR; 379 m->m_flags &= ~M_AUTHIPDGM; 380 } 381#endif 382 /* 383 * If no IP addresses have been set yet but the interfaces 384 * are receiving, can't do anything with incoming packets yet. 385 */ 386 if (in_ifaddr.tqh_first == 0) 387 goto bad; 388 ipstat.ips_total++; 389 if (m->m_len < sizeof (struct ip) && 390 (m = m_pullup(m, sizeof (struct ip))) == 0) { 391 ipstat.ips_toosmall++; 392 return; 393 } 394 ip = mtod(m, struct ip *); 395 if (ip->ip_v != IPVERSION) { 396 ipstat.ips_badvers++; 397 goto bad; 398 } 399 hlen = ip->ip_hl << 2; 400 if (hlen < sizeof(struct ip)) { /* minimum header length */ 401 ipstat.ips_badhlen++; 402 goto bad; 403 } 404 if (hlen > m->m_len) { 405 if ((m = m_pullup(m, hlen)) == 0) { 406 ipstat.ips_badhlen++; 407 return; 408 } 409 ip = mtod(m, struct ip *); 410 } 411 412 /* 413 * RFC1122: packets with a multicast source address are 414 * not allowed. 415 */ 416 if (IN_MULTICAST(ip->ip_src.s_addr)) { 417 ipstat.ips_badaddr++; 418 goto bad; 419 } 420 421 /* 127/8 must not appear on wire - RFC1122 */ 422 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 423 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 424 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 425 ipstat.ips_badaddr++; 426 goto bad; 427 } 428 } 429 430 if (in_cksum(m, hlen) != 0) { 431 ipstat.ips_badsum++; 432 goto bad; 433 } 434 435 /* Retrieve the packet length. */ 436 len = ntohs(ip->ip_len); 437 438 /* 439 * Check for additional length bogosity 440 */ 441 if (len < hlen) { 442 ipstat.ips_badlen++; 443 goto bad; 444 } 445 446 /* 447 * Check that the amount of data in the buffers 448 * is as at least much as the IP header would have us expect. 449 * Trim mbufs if longer than we expect. 450 * Drop packet if shorter than we expect. 451 */ 452 if (m->m_pkthdr.len < len) { 453 ipstat.ips_tooshort++; 454 goto bad; 455 } 456 if (m->m_pkthdr.len > len) { 457 if (m->m_len == m->m_pkthdr.len) { 458 m->m_len = len; 459 m->m_pkthdr.len = len; 460 } else 461 m_adj(m, len - m->m_pkthdr.len); 462 } 463 464#ifdef IPSEC 465 /* ipflow (IP fast fowarding) is not compatible with IPsec. */ 466 m->m_flags &= ~M_CANFASTFWD; 467#else 468 /* 469 * Assume that we can create a fast-forward IP flow entry 470 * based on this packet. 471 */ 472 m->m_flags |= M_CANFASTFWD; 473#endif 474 475#ifdef PFIL_HOOKS 476 /* 477 * Run through list of hooks for input packets. If there are any 478 * filters which require that additional packets in the flow are 479 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 480 * Note that filters must _never_ set this flag, as another filter 481 * in the list may have previously cleared it. 482 */ 483 /* 484 * let ipfilter look at packet on the wire, 485 * not the decapsulated packet. 486 */ 487#ifdef IPSEC 488 if (!ipsec_gethist(m, NULL)) 489#else 490 if (1) 491#endif 492 { 493 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 494 PFIL_IN) != 0) 495 return; 496 if (m == NULL) 497 return; 498 ip = mtod(m, struct ip *); 499 } 500#endif /* PFIL_HOOKS */ 501 502#ifdef ALTQ 503 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 504 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 505 /* packet dropped by traffic conditioner */ 506 return; 507 } 508#endif 509 510 /* 511 * Convert fields to host representation. 512 */ 513 NTOHS(ip->ip_len); 514 NTOHS(ip->ip_off); 515 516 /* 517 * Process options and, if not destined for us, 518 * ship it on. ip_dooptions returns 1 when an 519 * error was detected (causing an icmp message 520 * to be sent and the original packet to be freed). 521 */ 522 ip_nhops = 0; /* for source routed packets */ 523 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 524 return; 525 526 /* 527 * Check our list of addresses, to see if the packet is for us. 528 * 529 * Traditional 4.4BSD did not consult IFF_UP at all. 530 * The behavior here is to treat addresses on !IFF_UP interface 531 * as not mine. 532 */ 533 downmatch = 0; 534 for (ia = IN_IFADDR_HASH(ip->ip_dst.s_addr).lh_first; 535 ia != NULL; 536 ia = ia->ia_hash.le_next) { 537 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 538 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 539 break; 540 else 541 downmatch++; 542 } 543 } 544 if (ia != NULL) 545 goto ours; 546 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 547 for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first; 548 ifa != NULL; ifa = ifa->ifa_list.tqe_next) { 549 if (ifa->ifa_addr->sa_family != AF_INET) continue; 550 ia = ifatoia(ifa); 551 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 552 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 553 /* 554 * Look for all-0's host part (old broadcast addr), 555 * either for subnet or net. 556 */ 557 ip->ip_dst.s_addr == ia->ia_subnet || 558 ip->ip_dst.s_addr == ia->ia_net) 559 goto ours; 560 /* 561 * An interface with IP address zero accepts 562 * all packets that arrive on that interface. 563 */ 564 if (in_nullhost(ia->ia_addr.sin_addr)) 565 goto ours; 566 } 567 } 568 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 569 struct in_multi *inm; 570#ifdef MROUTING 571 extern struct socket *ip_mrouter; 572 573 if (m->m_flags & M_EXT) { 574 if ((m = m_pullup(m, hlen)) == 0) { 575 ipstat.ips_toosmall++; 576 return; 577 } 578 ip = mtod(m, struct ip *); 579 } 580 581 if (ip_mrouter) { 582 /* 583 * If we are acting as a multicast router, all 584 * incoming multicast packets are passed to the 585 * kernel-level multicast forwarding function. 586 * The packet is returned (relatively) intact; if 587 * ip_mforward() returns a non-zero value, the packet 588 * must be discarded, else it may be accepted below. 589 * 590 * (The IP ident field is put in the same byte order 591 * as expected when ip_mforward() is called from 592 * ip_output().) 593 */ 594 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 595 ipstat.ips_cantforward++; 596 m_freem(m); 597 return; 598 } 599 600 /* 601 * The process-level routing demon needs to receive 602 * all multicast IGMP packets, whether or not this 603 * host belongs to their destination groups. 604 */ 605 if (ip->ip_p == IPPROTO_IGMP) 606 goto ours; 607 ipstat.ips_forward++; 608 } 609#endif 610 /* 611 * See if we belong to the destination multicast group on the 612 * arrival interface. 613 */ 614 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 615 if (inm == NULL) { 616 ipstat.ips_cantforward++; 617 m_freem(m); 618 return; 619 } 620 goto ours; 621 } 622 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 623 in_nullhost(ip->ip_dst)) 624 goto ours; 625 626 /* 627 * Not for us; forward if possible and desirable. 628 */ 629 if (ipforwarding == 0) { 630 ipstat.ips_cantforward++; 631 m_freem(m); 632 } else { 633 /* 634 * If ip_dst matched any of my address on !IFF_UP interface, 635 * and there's no IFF_UP interface that matches ip_dst, 636 * send icmp unreach. Forwarding it will result in in-kernel 637 * forwarding loop till TTL goes to 0. 638 */ 639 if (downmatch) { 640 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 641 ipstat.ips_cantforward++; 642 return; 643 } 644 ip_forward(m, 0); 645 } 646 return; 647 648ours: 649 /* 650 * If offset or IP_MF are set, must reassemble. 651 * Otherwise, nothing need be done. 652 * (We could look in the reassembly queue to see 653 * if the packet was previously fragmented, 654 * but it's not worth the time; just let them time out.) 655 */ 656 if (ip->ip_off & ~(IP_DF|IP_RF)) { 657 /* 658 * Look for queue of fragments 659 * of this datagram. 660 */ 661 IPQ_LOCK(); 662 for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next) 663 if (ip->ip_id == fp->ipq_id && 664 in_hosteq(ip->ip_src, fp->ipq_src) && 665 in_hosteq(ip->ip_dst, fp->ipq_dst) && 666 ip->ip_p == fp->ipq_p) 667 goto found; 668 fp = 0; 669found: 670 671 /* 672 * Adjust ip_len to not reflect header, 673 * set ipqe_mff if more fragments are expected, 674 * convert offset of this to bytes. 675 */ 676 ip->ip_len -= hlen; 677 mff = (ip->ip_off & IP_MF) != 0; 678 if (mff) { 679 /* 680 * Make sure that fragments have a data length 681 * that's a non-zero multiple of 8 bytes. 682 */ 683 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 684 ipstat.ips_badfrags++; 685 IPQ_UNLOCK(); 686 goto bad; 687 } 688 } 689 ip->ip_off <<= 3; 690 691 /* 692 * If datagram marked as having more fragments 693 * or if this is not the first fragment, 694 * attempt reassembly; if it succeeds, proceed. 695 */ 696 if (mff || ip->ip_off) { 697 ipstat.ips_fragments++; 698 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 699 if (ipqe == NULL) { 700 ipstat.ips_rcvmemdrop++; 701 IPQ_UNLOCK(); 702 goto bad; 703 } 704 ipqe->ipqe_mff = mff; 705 ipqe->ipqe_m = m; 706 ipqe->ipqe_ip = ip; 707 m = ip_reass(ipqe, fp); 708 if (m == 0) { 709 IPQ_UNLOCK(); 710 return; 711 } 712 ipstat.ips_reassembled++; 713 ip = mtod(m, struct ip *); 714 hlen = ip->ip_hl << 2; 715 ip->ip_len += hlen; 716 } else 717 if (fp) 718 ip_freef(fp); 719 IPQ_UNLOCK(); 720 } 721 722#ifdef IPSEC 723 /* 724 * enforce IPsec policy checking if we are seeing last header. 725 * note that we do not visit this with protocols with pcb layer 726 * code - like udp/tcp/raw ip. 727 */ 728 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 729 ipsec4_in_reject(m, NULL)) { 730 ipsecstat.in_polvio++; 731 goto bad; 732 } 733#endif 734 735 /* 736 * Switch out to protocol's input routine. 737 */ 738#if IFA_STATS 739 if (ia && ip) 740 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 741#endif 742 ipstat.ips_delivered++; 743 { 744 int off = hlen, nh = ip->ip_p; 745 746 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 747 return; 748 } 749bad: 750 m_freem(m); 751} 752 753/* 754 * Take incoming datagram fragment and try to 755 * reassemble it into whole datagram. If a chain for 756 * reassembly of this datagram already exists, then it 757 * is given as fp; otherwise have to make a chain. 758 */ 759struct mbuf * 760ip_reass(ipqe, fp) 761 struct ipqent *ipqe; 762 struct ipq *fp; 763{ 764 struct mbuf *m = ipqe->ipqe_m; 765 struct ipqent *nq, *p, *q; 766 struct ip *ip; 767 struct mbuf *t; 768 int hlen = ipqe->ipqe_ip->ip_hl << 2; 769 int i, next; 770 771 IPQ_LOCK_CHECK(); 772 773 /* 774 * Presence of header sizes in mbufs 775 * would confuse code below. 776 */ 777 m->m_data += hlen; 778 m->m_len -= hlen; 779 780 /* 781 * If first fragment to arrive, create a reassembly queue. 782 */ 783 if (fp == 0) { 784 MALLOC(fp, struct ipq *, sizeof (struct ipq), 785 M_FTABLE, M_NOWAIT); 786 if (fp == NULL) 787 goto dropfrag; 788 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 789 fp->ipq_ttl = IPFRAGTTL; 790 fp->ipq_p = ipqe->ipqe_ip->ip_p; 791 fp->ipq_id = ipqe->ipqe_ip->ip_id; 792 LIST_INIT(&fp->ipq_fragq); 793 fp->ipq_src = ipqe->ipqe_ip->ip_src; 794 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 795 p = NULL; 796 goto insert; 797 } 798 799 /* 800 * Find a segment which begins after this one does. 801 */ 802 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 803 p = q, q = q->ipqe_q.le_next) 804 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 805 break; 806 807 /* 808 * If there is a preceding segment, it may provide some of 809 * our data already. If so, drop the data from the incoming 810 * segment. If it provides all of our data, drop us. 811 */ 812 if (p != NULL) { 813 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 814 ipqe->ipqe_ip->ip_off; 815 if (i > 0) { 816 if (i >= ipqe->ipqe_ip->ip_len) 817 goto dropfrag; 818 m_adj(ipqe->ipqe_m, i); 819 ipqe->ipqe_ip->ip_off += i; 820 ipqe->ipqe_ip->ip_len -= i; 821 } 822 } 823 824 /* 825 * While we overlap succeeding segments trim them or, 826 * if they are completely covered, dequeue them. 827 */ 828 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 829 q->ipqe_ip->ip_off; q = nq) { 830 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 831 q->ipqe_ip->ip_off; 832 if (i < q->ipqe_ip->ip_len) { 833 q->ipqe_ip->ip_len -= i; 834 q->ipqe_ip->ip_off += i; 835 m_adj(q->ipqe_m, i); 836 break; 837 } 838 nq = q->ipqe_q.le_next; 839 m_freem(q->ipqe_m); 840 LIST_REMOVE(q, ipqe_q); 841 pool_put(&ipqent_pool, q); 842 } 843 844insert: 845 /* 846 * Stick new segment in its place; 847 * check for complete reassembly. 848 */ 849 if (p == NULL) { 850 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 851 } else { 852 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 853 } 854 next = 0; 855 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 856 p = q, q = q->ipqe_q.le_next) { 857 if (q->ipqe_ip->ip_off != next) 858 return (0); 859 next += q->ipqe_ip->ip_len; 860 } 861 if (p->ipqe_mff) 862 return (0); 863 864 /* 865 * Reassembly is complete. Check for a bogus message size and 866 * concatenate fragments. 867 */ 868 q = fp->ipq_fragq.lh_first; 869 ip = q->ipqe_ip; 870 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 871 ipstat.ips_toolong++; 872 ip_freef(fp); 873 return (0); 874 } 875 m = q->ipqe_m; 876 t = m->m_next; 877 m->m_next = 0; 878 m_cat(m, t); 879 nq = q->ipqe_q.le_next; 880 pool_put(&ipqent_pool, q); 881 for (q = nq; q != NULL; q = nq) { 882 t = q->ipqe_m; 883 nq = q->ipqe_q.le_next; 884 pool_put(&ipqent_pool, q); 885 m_cat(m, t); 886 } 887 888 /* 889 * Create header for new ip packet by 890 * modifying header of first packet; 891 * dequeue and discard fragment reassembly header. 892 * Make header visible. 893 */ 894 ip->ip_len = next; 895 ip->ip_src = fp->ipq_src; 896 ip->ip_dst = fp->ipq_dst; 897 LIST_REMOVE(fp, ipq_q); 898 FREE(fp, M_FTABLE); 899 m->m_len += (ip->ip_hl << 2); 900 m->m_data -= (ip->ip_hl << 2); 901 /* some debugging cruft by sklower, below, will go away soon */ 902 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 903 int plen = 0; 904 for (t = m; t; t = t->m_next) 905 plen += t->m_len; 906 m->m_pkthdr.len = plen; 907 } 908 return (m); 909 910dropfrag: 911 ipstat.ips_fragdropped++; 912 m_freem(m); 913 pool_put(&ipqent_pool, ipqe); 914 return (0); 915} 916 917/* 918 * Free a fragment reassembly header and all 919 * associated datagrams. 920 */ 921void 922ip_freef(fp) 923 struct ipq *fp; 924{ 925 struct ipqent *q, *p; 926 927 IPQ_LOCK_CHECK(); 928 929 for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { 930 p = q->ipqe_q.le_next; 931 m_freem(q->ipqe_m); 932 LIST_REMOVE(q, ipqe_q); 933 pool_put(&ipqent_pool, q); 934 } 935 LIST_REMOVE(fp, ipq_q); 936 FREE(fp, M_FTABLE); 937} 938 939/* 940 * IP timer processing; 941 * if a timer expires on a reassembly 942 * queue, discard it. 943 */ 944void 945ip_slowtimo() 946{ 947 struct ipq *fp, *nfp; 948 int s = splsoftnet(); 949 950 IPQ_LOCK(); 951 for (fp = ipq.lh_first; fp != NULL; fp = nfp) { 952 nfp = fp->ipq_q.le_next; 953 if (--fp->ipq_ttl == 0) { 954 ipstat.ips_fragtimeout++; 955 ip_freef(fp); 956 } 957 } 958 IPQ_UNLOCK(); 959#ifdef GATEWAY 960 ipflow_slowtimo(); 961#endif 962 splx(s); 963} 964 965/* 966 * Drain off all datagram fragments. 967 */ 968void 969ip_drain() 970{ 971 972 /* 973 * We may be called from a device's interrupt context. If 974 * the ipq is already busy, just bail out now. 975 */ 976 if (ipq_lock_try() == 0) 977 return; 978 979 while (ipq.lh_first != NULL) { 980 ipstat.ips_fragdropped++; 981 ip_freef(ipq.lh_first); 982 } 983 984 IPQ_UNLOCK(); 985} 986 987/* 988 * Do option processing on a datagram, 989 * possibly discarding it if bad options are encountered, 990 * or forwarding it if source-routed. 991 * Returns 1 if packet has been forwarded/freed, 992 * 0 if the packet should be processed further. 993 */ 994int 995ip_dooptions(m) 996 struct mbuf *m; 997{ 998 struct ip *ip = mtod(m, struct ip *); 999 u_char *cp, *cp0; 1000 struct ip_timestamp *ipt; 1001 struct in_ifaddr *ia; 1002 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1003 struct in_addr dst; 1004 n_time ntime; 1005 1006 dst = ip->ip_dst; 1007 cp = (u_char *)(ip + 1); 1008 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1009 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1010 opt = cp[IPOPT_OPTVAL]; 1011 if (opt == IPOPT_EOL) 1012 break; 1013 if (opt == IPOPT_NOP) 1014 optlen = 1; 1015 else { 1016 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1017 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1018 goto bad; 1019 } 1020 optlen = cp[IPOPT_OLEN]; 1021 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1022 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1023 goto bad; 1024 } 1025 } 1026 switch (opt) { 1027 1028 default: 1029 break; 1030 1031 /* 1032 * Source routing with record. 1033 * Find interface with current destination address. 1034 * If none on this machine then drop if strictly routed, 1035 * or do nothing if loosely routed. 1036 * Record interface address and bring up next address 1037 * component. If strictly routed make sure next 1038 * address is on directly accessible net. 1039 */ 1040 case IPOPT_LSRR: 1041 case IPOPT_SSRR: 1042 if (ip_allowsrcrt == 0) { 1043 type = ICMP_UNREACH; 1044 code = ICMP_UNREACH_NET_PROHIB; 1045 goto bad; 1046 } 1047 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1048 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1049 goto bad; 1050 } 1051 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1052 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1053 goto bad; 1054 } 1055 ipaddr.sin_addr = ip->ip_dst; 1056 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1057 if (ia == 0) { 1058 if (opt == IPOPT_SSRR) { 1059 type = ICMP_UNREACH; 1060 code = ICMP_UNREACH_SRCFAIL; 1061 goto bad; 1062 } 1063 /* 1064 * Loose routing, and not at next destination 1065 * yet; nothing to do except forward. 1066 */ 1067 break; 1068 } 1069 off--; /* 0 origin */ 1070 if ((off + sizeof(struct in_addr)) > optlen) { 1071 /* 1072 * End of source route. Should be for us. 1073 */ 1074 save_rte(cp, ip->ip_src); 1075 break; 1076 } 1077 /* 1078 * locate outgoing interface 1079 */ 1080 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1081 sizeof(ipaddr.sin_addr)); 1082 if (opt == IPOPT_SSRR) 1083 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1084 else 1085 ia = ip_rtaddr(ipaddr.sin_addr); 1086 if (ia == 0) { 1087 type = ICMP_UNREACH; 1088 code = ICMP_UNREACH_SRCFAIL; 1089 goto bad; 1090 } 1091 ip->ip_dst = ipaddr.sin_addr; 1092 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1093 (caddr_t)(cp + off), sizeof(struct in_addr)); 1094 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1095 /* 1096 * Let ip_intr's mcast routing check handle mcast pkts 1097 */ 1098 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1099 break; 1100 1101 case IPOPT_RR: 1102 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1103 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1104 goto bad; 1105 } 1106 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1107 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1108 goto bad; 1109 } 1110 /* 1111 * If no space remains, ignore. 1112 */ 1113 off--; /* 0 origin */ 1114 if ((off + sizeof(struct in_addr)) > optlen) 1115 break; 1116 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1117 sizeof(ipaddr.sin_addr)); 1118 /* 1119 * locate outgoing interface; if we're the destination, 1120 * use the incoming interface (should be same). 1121 */ 1122 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1123 == NULL && 1124 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1125 type = ICMP_UNREACH; 1126 code = ICMP_UNREACH_HOST; 1127 goto bad; 1128 } 1129 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1130 (caddr_t)(cp + off), sizeof(struct in_addr)); 1131 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1132 break; 1133 1134 case IPOPT_TS: 1135 code = cp - (u_char *)ip; 1136 ipt = (struct ip_timestamp *)cp; 1137 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1138 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1139 goto bad; 1140 } 1141 if (ipt->ipt_ptr < 5) { 1142 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1143 goto bad; 1144 } 1145 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1146 if (++ipt->ipt_oflw == 0) { 1147 code = (u_char *)&ipt->ipt_ptr - 1148 (u_char *)ip; 1149 goto bad; 1150 } 1151 break; 1152 } 1153 cp0 = (cp + ipt->ipt_ptr - 1); 1154 switch (ipt->ipt_flg) { 1155 1156 case IPOPT_TS_TSONLY: 1157 break; 1158 1159 case IPOPT_TS_TSANDADDR: 1160 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1161 sizeof(struct in_addr) > ipt->ipt_len) { 1162 code = (u_char *)&ipt->ipt_ptr - 1163 (u_char *)ip; 1164 goto bad; 1165 } 1166 ipaddr.sin_addr = dst; 1167 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1168 m->m_pkthdr.rcvif)); 1169 if (ia == 0) 1170 continue; 1171 bcopy(&ia->ia_addr.sin_addr, 1172 cp0, sizeof(struct in_addr)); 1173 ipt->ipt_ptr += sizeof(struct in_addr); 1174 break; 1175 1176 case IPOPT_TS_PRESPEC: 1177 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1178 sizeof(struct in_addr) > ipt->ipt_len) { 1179 code = (u_char *)&ipt->ipt_ptr - 1180 (u_char *)ip; 1181 goto bad; 1182 } 1183 bcopy(cp0, &ipaddr.sin_addr, 1184 sizeof(struct in_addr)); 1185 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1186 == NULL) 1187 continue; 1188 ipt->ipt_ptr += sizeof(struct in_addr); 1189 break; 1190 1191 default: 1192 /* XXX can't take &ipt->ipt_flg */ 1193 code = (u_char *)&ipt->ipt_ptr - 1194 (u_char *)ip + 1; 1195 goto bad; 1196 } 1197 ntime = iptime(); 1198 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1199 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1200 sizeof(n_time)); 1201 ipt->ipt_ptr += sizeof(n_time); 1202 } 1203 } 1204 if (forward) { 1205 if (ip_forwsrcrt == 0) { 1206 type = ICMP_UNREACH; 1207 code = ICMP_UNREACH_SRCFAIL; 1208 goto bad; 1209 } 1210 ip_forward(m, 1); 1211 return (1); 1212 } 1213 return (0); 1214bad: 1215 icmp_error(m, type, code, 0, 0); 1216 ipstat.ips_badoptions++; 1217 return (1); 1218} 1219 1220/* 1221 * Given address of next destination (final or next hop), 1222 * return internet address info of interface to be used to get there. 1223 */ 1224struct in_ifaddr * 1225ip_rtaddr(dst) 1226 struct in_addr dst; 1227{ 1228 struct sockaddr_in *sin; 1229 1230 sin = satosin(&ipforward_rt.ro_dst); 1231 1232 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1233 if (ipforward_rt.ro_rt) { 1234 RTFREE(ipforward_rt.ro_rt); 1235 ipforward_rt.ro_rt = 0; 1236 } 1237 sin->sin_family = AF_INET; 1238 sin->sin_len = sizeof(*sin); 1239 sin->sin_addr = dst; 1240 1241 rtalloc(&ipforward_rt); 1242 } 1243 if (ipforward_rt.ro_rt == 0) 1244 return ((struct in_ifaddr *)0); 1245 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1246} 1247 1248/* 1249 * Save incoming source route for use in replies, 1250 * to be picked up later by ip_srcroute if the receiver is interested. 1251 */ 1252void 1253save_rte(option, dst) 1254 u_char *option; 1255 struct in_addr dst; 1256{ 1257 unsigned olen; 1258 1259 olen = option[IPOPT_OLEN]; 1260#ifdef DIAGNOSTIC 1261 if (ipprintfs) 1262 printf("save_rte: olen %d\n", olen); 1263#endif /* 0 */ 1264 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1265 return; 1266 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1267 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1268 ip_srcrt.dst = dst; 1269} 1270 1271/* 1272 * Retrieve incoming source route for use in replies, 1273 * in the same form used by setsockopt. 1274 * The first hop is placed before the options, will be removed later. 1275 */ 1276struct mbuf * 1277ip_srcroute() 1278{ 1279 struct in_addr *p, *q; 1280 struct mbuf *m; 1281 1282 if (ip_nhops == 0) 1283 return ((struct mbuf *)0); 1284 m = m_get(M_DONTWAIT, MT_SOOPTS); 1285 if (m == 0) 1286 return ((struct mbuf *)0); 1287 1288#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1289 1290 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1291 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1292 OPTSIZ; 1293#ifdef DIAGNOSTIC 1294 if (ipprintfs) 1295 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1296#endif 1297 1298 /* 1299 * First save first hop for return route 1300 */ 1301 p = &ip_srcrt.route[ip_nhops - 1]; 1302 *(mtod(m, struct in_addr *)) = *p--; 1303#ifdef DIAGNOSTIC 1304 if (ipprintfs) 1305 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1306#endif 1307 1308 /* 1309 * Copy option fields and padding (nop) to mbuf. 1310 */ 1311 ip_srcrt.nop = IPOPT_NOP; 1312 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1313 bcopy((caddr_t)&ip_srcrt.nop, 1314 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1315 q = (struct in_addr *)(mtod(m, caddr_t) + 1316 sizeof(struct in_addr) + OPTSIZ); 1317#undef OPTSIZ 1318 /* 1319 * Record return path as an IP source route, 1320 * reversing the path (pointers are now aligned). 1321 */ 1322 while (p >= ip_srcrt.route) { 1323#ifdef DIAGNOSTIC 1324 if (ipprintfs) 1325 printf(" %x", ntohl(q->s_addr)); 1326#endif 1327 *q++ = *p--; 1328 } 1329 /* 1330 * Last hop goes to final destination. 1331 */ 1332 *q = ip_srcrt.dst; 1333#ifdef DIAGNOSTIC 1334 if (ipprintfs) 1335 printf(" %x\n", ntohl(q->s_addr)); 1336#endif 1337 return (m); 1338} 1339 1340/* 1341 * Strip out IP options, at higher 1342 * level protocol in the kernel. 1343 * Second argument is buffer to which options 1344 * will be moved, and return value is their length. 1345 * XXX should be deleted; last arg currently ignored. 1346 */ 1347void 1348ip_stripoptions(m, mopt) 1349 struct mbuf *m; 1350 struct mbuf *mopt; 1351{ 1352 int i; 1353 struct ip *ip = mtod(m, struct ip *); 1354 caddr_t opts; 1355 int olen; 1356 1357 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1358 opts = (caddr_t)(ip + 1); 1359 i = m->m_len - (sizeof (struct ip) + olen); 1360 bcopy(opts + olen, opts, (unsigned)i); 1361 m->m_len -= olen; 1362 if (m->m_flags & M_PKTHDR) 1363 m->m_pkthdr.len -= olen; 1364 ip->ip_len -= olen; 1365 ip->ip_hl = sizeof (struct ip) >> 2; 1366} 1367 1368int inetctlerrmap[PRC_NCMDS] = { 1369 0, 0, 0, 0, 1370 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1371 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1372 EMSGSIZE, EHOSTUNREACH, 0, 0, 1373 0, 0, 0, 0, 1374 ENOPROTOOPT 1375}; 1376 1377/* 1378 * Forward a packet. If some error occurs return the sender 1379 * an icmp packet. Note we can't always generate a meaningful 1380 * icmp message because icmp doesn't have a large enough repertoire 1381 * of codes and types. 1382 * 1383 * If not forwarding, just drop the packet. This could be confusing 1384 * if ipforwarding was zero but some routing protocol was advancing 1385 * us as a gateway to somewhere. However, we must let the routing 1386 * protocol deal with that. 1387 * 1388 * The srcrt parameter indicates whether the packet is being forwarded 1389 * via a source route. 1390 */ 1391void 1392ip_forward(m, srcrt) 1393 struct mbuf *m; 1394 int srcrt; 1395{ 1396 struct ip *ip = mtod(m, struct ip *); 1397 struct sockaddr_in *sin; 1398 struct rtentry *rt; 1399 int error, type = 0, code = 0; 1400 struct mbuf *mcopy; 1401 n_long dest; 1402 struct ifnet *destifp; 1403#ifdef IPSEC 1404 struct ifnet dummyifp; 1405#endif 1406 1407 dest = 0; 1408#ifdef DIAGNOSTIC 1409 if (ipprintfs) 1410 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1411 ntohl(ip->ip_src.s_addr), 1412 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1413#endif 1414 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1415 ipstat.ips_cantforward++; 1416 m_freem(m); 1417 return; 1418 } 1419 if (ip->ip_ttl <= IPTTLDEC) { 1420 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1421 return; 1422 } 1423 ip->ip_ttl -= IPTTLDEC; 1424 1425 sin = satosin(&ipforward_rt.ro_dst); 1426 if ((rt = ipforward_rt.ro_rt) == 0 || 1427 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1428 if (ipforward_rt.ro_rt) { 1429 RTFREE(ipforward_rt.ro_rt); 1430 ipforward_rt.ro_rt = 0; 1431 } 1432 sin->sin_family = AF_INET; 1433 sin->sin_len = sizeof(struct sockaddr_in); 1434 sin->sin_addr = ip->ip_dst; 1435 1436 rtalloc(&ipforward_rt); 1437 if (ipforward_rt.ro_rt == 0) { 1438 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1439 return; 1440 } 1441 rt = ipforward_rt.ro_rt; 1442 } 1443 1444 /* 1445 * Save at most 68 bytes of the packet in case 1446 * we need to generate an ICMP message to the src. 1447 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1448 */ 1449 mcopy = m_copym(m, 0, imin((int)ip->ip_len, 68), M_DONTWAIT); 1450 if (mcopy) 1451 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1452 1453 /* 1454 * If forwarding packet using same interface that it came in on, 1455 * perhaps should send a redirect to sender to shortcut a hop. 1456 * Only send redirect if source is sending directly to us, 1457 * and if packet was not source routed (or has any options). 1458 * Also, don't send redirect if forwarding using a default route 1459 * or a route modified by a redirect. 1460 */ 1461 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1462 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1463 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1464 ipsendredirects && !srcrt) { 1465 if (rt->rt_ifa && 1466 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1467 ifatoia(rt->rt_ifa)->ia_subnet) { 1468 if (rt->rt_flags & RTF_GATEWAY) 1469 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1470 else 1471 dest = ip->ip_dst.s_addr; 1472 /* 1473 * Router requirements says to only send host 1474 * redirects. 1475 */ 1476 type = ICMP_REDIRECT; 1477 code = ICMP_REDIRECT_HOST; 1478#ifdef DIAGNOSTIC 1479 if (ipprintfs) 1480 printf("redirect (%d) to %x\n", code, 1481 (u_int32_t)dest); 1482#endif 1483 } 1484 } 1485 1486#ifdef IPSEC 1487 /* Don't lookup socket in forwading case */ 1488 (void)ipsec_setsocket(m, NULL); 1489#endif 1490 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1491 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1492 if (error) 1493 ipstat.ips_cantforward++; 1494 else { 1495 ipstat.ips_forward++; 1496 if (type) 1497 ipstat.ips_redirectsent++; 1498 else { 1499 if (mcopy) { 1500#ifdef GATEWAY 1501 if (mcopy->m_flags & M_CANFASTFWD) 1502 ipflow_create(&ipforward_rt, mcopy); 1503#endif 1504 m_freem(mcopy); 1505 } 1506 return; 1507 } 1508 } 1509 if (mcopy == NULL) 1510 return; 1511 destifp = NULL; 1512 1513 switch (error) { 1514 1515 case 0: /* forwarded, but need redirect */ 1516 /* type, code set above */ 1517 break; 1518 1519 case ENETUNREACH: /* shouldn't happen, checked above */ 1520 case EHOSTUNREACH: 1521 case ENETDOWN: 1522 case EHOSTDOWN: 1523 default: 1524 type = ICMP_UNREACH; 1525 code = ICMP_UNREACH_HOST; 1526 break; 1527 1528 case EMSGSIZE: 1529 type = ICMP_UNREACH; 1530 code = ICMP_UNREACH_NEEDFRAG; 1531#ifndef IPSEC 1532 if (ipforward_rt.ro_rt) 1533 destifp = ipforward_rt.ro_rt->rt_ifp; 1534#else 1535 /* 1536 * If the packet is routed over IPsec tunnel, tell the 1537 * originator the tunnel MTU. 1538 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1539 * XXX quickhack!!! 1540 */ 1541 if (ipforward_rt.ro_rt) { 1542 struct secpolicy *sp; 1543 int ipsecerror; 1544 size_t ipsechdr; 1545 struct route *ro; 1546 1547 sp = ipsec4_getpolicybyaddr(mcopy, 1548 IPSEC_DIR_OUTBOUND, 1549 IP_FORWARDING, 1550 &ipsecerror); 1551 1552 if (sp == NULL) 1553 destifp = ipforward_rt.ro_rt->rt_ifp; 1554 else { 1555 /* count IPsec header size */ 1556 ipsechdr = ipsec4_hdrsiz(mcopy, 1557 IPSEC_DIR_OUTBOUND, 1558 NULL); 1559 1560 /* 1561 * find the correct route for outer IPv4 1562 * header, compute tunnel MTU. 1563 * 1564 * XXX BUG ALERT 1565 * The "dummyifp" code relies upon the fact 1566 * that icmp_error() touches only ifp->if_mtu. 1567 */ 1568 /*XXX*/ 1569 destifp = NULL; 1570 if (sp->req != NULL 1571 && sp->req->sav != NULL 1572 && sp->req->sav->sah != NULL) { 1573 ro = &sp->req->sav->sah->sa_route; 1574 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1575 dummyifp.if_mtu = 1576 ro->ro_rt->rt_ifp->if_mtu; 1577 dummyifp.if_mtu -= ipsechdr; 1578 destifp = &dummyifp; 1579 } 1580 } 1581 1582 key_freesp(sp); 1583 } 1584 } 1585#endif /*IPSEC*/ 1586 ipstat.ips_cantfrag++; 1587 break; 1588 1589 case ENOBUFS: 1590 type = ICMP_SOURCEQUENCH; 1591 code = 0; 1592 break; 1593 } 1594 icmp_error(mcopy, type, code, dest, destifp); 1595} 1596 1597void 1598ip_savecontrol(inp, mp, ip, m) 1599 struct inpcb *inp; 1600 struct mbuf **mp; 1601 struct ip *ip; 1602 struct mbuf *m; 1603{ 1604 1605 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1606 struct timeval tv; 1607 1608 microtime(&tv); 1609 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1610 SCM_TIMESTAMP, SOL_SOCKET); 1611 if (*mp) 1612 mp = &(*mp)->m_next; 1613 } 1614 if (inp->inp_flags & INP_RECVDSTADDR) { 1615 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1616 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1617 if (*mp) 1618 mp = &(*mp)->m_next; 1619 } 1620#ifdef notyet 1621 /* 1622 * XXX 1623 * Moving these out of udp_input() made them even more broken 1624 * than they already were. 1625 * - fenner@parc.xerox.com 1626 */ 1627 /* options were tossed already */ 1628 if (inp->inp_flags & INP_RECVOPTS) { 1629 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1630 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1631 if (*mp) 1632 mp = &(*mp)->m_next; 1633 } 1634 /* ip_srcroute doesn't do what we want here, need to fix */ 1635 if (inp->inp_flags & INP_RECVRETOPTS) { 1636 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1637 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1638 if (*mp) 1639 mp = &(*mp)->m_next; 1640 } 1641#endif 1642 if (inp->inp_flags & INP_RECVIF) { 1643 struct sockaddr_dl sdl; 1644 1645 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1646 sdl.sdl_family = AF_LINK; 1647 sdl.sdl_index = m->m_pkthdr.rcvif ? 1648 m->m_pkthdr.rcvif->if_index : 0; 1649 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1650 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1651 IP_RECVIF, IPPROTO_IP); 1652 if (*mp) 1653 mp = &(*mp)->m_next; 1654 } 1655} 1656 1657int 1658ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1659 int *name; 1660 u_int namelen; 1661 void *oldp; 1662 size_t *oldlenp; 1663 void *newp; 1664 size_t newlen; 1665{ 1666 extern int subnetsarelocal, hostzeroisbroadcast; 1667 1668 int error, old; 1669 1670 /* All sysctl names at this level are terminal. */ 1671 if (namelen != 1) 1672 return (ENOTDIR); 1673 1674 switch (name[0]) { 1675 case IPCTL_FORWARDING: 1676 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1677 case IPCTL_SENDREDIRECTS: 1678 return (sysctl_int(oldp, oldlenp, newp, newlen, 1679 &ipsendredirects)); 1680 case IPCTL_DEFTTL: 1681 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1682#ifdef notyet 1683 case IPCTL_DEFMTU: 1684 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1685#endif 1686 case IPCTL_FORWSRCRT: 1687 /* Don't allow this to change in a secure environment. */ 1688 if (securelevel > 0) 1689 return (sysctl_rdint(oldp, oldlenp, newp, 1690 ip_forwsrcrt)); 1691 else 1692 return (sysctl_int(oldp, oldlenp, newp, newlen, 1693 &ip_forwsrcrt)); 1694 case IPCTL_DIRECTEDBCAST: 1695 return (sysctl_int(oldp, oldlenp, newp, newlen, 1696 &ip_directedbcast)); 1697 case IPCTL_ALLOWSRCRT: 1698 return (sysctl_int(oldp, oldlenp, newp, newlen, 1699 &ip_allowsrcrt)); 1700 case IPCTL_SUBNETSARELOCAL: 1701 return (sysctl_int(oldp, oldlenp, newp, newlen, 1702 &subnetsarelocal)); 1703 case IPCTL_MTUDISC: 1704 error = sysctl_int(oldp, oldlenp, newp, newlen, 1705 &ip_mtudisc); 1706 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1707 ip_mtudisc_timeout_q = 1708 rt_timer_queue_create(ip_mtudisc_timeout); 1709 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1710 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1711 ip_mtudisc_timeout_q = NULL; 1712 } 1713 return error; 1714 case IPCTL_ANONPORTMIN: 1715 old = anonportmin; 1716 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1717 if (anonportmin >= anonportmax || anonportmin < 0 1718 || anonportmin > 65535 1719#ifndef IPNOPRIVPORTS 1720 || anonportmin < IPPORT_RESERVED 1721#endif 1722 ) { 1723 anonportmin = old; 1724 return (EINVAL); 1725 } 1726 return (error); 1727 case IPCTL_ANONPORTMAX: 1728 old = anonportmax; 1729 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1730 if (anonportmin >= anonportmax || anonportmax < 0 1731 || anonportmax > 65535 1732#ifndef IPNOPRIVPORTS 1733 || anonportmax < IPPORT_RESERVED 1734#endif 1735 ) { 1736 anonportmax = old; 1737 return (EINVAL); 1738 } 1739 return (error); 1740 case IPCTL_MTUDISCTIMEOUT: 1741 error = sysctl_int(oldp, oldlenp, newp, newlen, 1742 &ip_mtudisc_timeout); 1743 if (ip_mtudisc_timeout_q != NULL) 1744 rt_timer_queue_change(ip_mtudisc_timeout_q, 1745 ip_mtudisc_timeout); 1746 return (error); 1747#ifdef GATEWAY 1748 case IPCTL_MAXFLOWS: 1749 { 1750 int s; 1751 1752 error = sysctl_int(oldp, oldlenp, newp, newlen, 1753 &ip_maxflows); 1754 s = splsoftnet(); 1755 ipflow_reap(0); 1756 splx(s); 1757 return (error); 1758 } 1759#endif 1760 case IPCTL_HOSTZEROBROADCAST: 1761 return (sysctl_int(oldp, oldlenp, newp, newlen, 1762 &hostzeroisbroadcast)); 1763#if NGIF > 0 1764 case IPCTL_GIF_TTL: 1765 return(sysctl_int(oldp, oldlenp, newp, newlen, 1766 &ip_gif_ttl)); 1767#endif 1768 1769#ifndef IPNOPRIVPORTS 1770 case IPCTL_LOWPORTMIN: 1771 old = lowportmin; 1772 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmin); 1773 if (lowportmin >= lowportmax 1774 || lowportmin > IPPORT_RESERVEDMAX 1775 || lowportmin < IPPORT_RESERVEDMIN 1776 ) { 1777 lowportmin = old; 1778 return (EINVAL); 1779 } 1780 return (error); 1781 case IPCTL_LOWPORTMAX: 1782 old = lowportmax; 1783 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmax); 1784 if (lowportmin >= lowportmax 1785 || lowportmax > IPPORT_RESERVEDMAX 1786 || lowportmax < IPPORT_RESERVEDMIN 1787 ) { 1788 lowportmax = old; 1789 return (EINVAL); 1790 } 1791 return (error); 1792#endif 1793 1794 default: 1795 return (EOPNOTSUPP); 1796 } 1797 /* NOTREACHED */ 1798} 1799