ip_input.c revision 1.200
1/* $NetBSD: ip_input.c,v 1.200 2004/04/25 16:42:42 simonb Exp $ */ 2 3/* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32/*- 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 3. All advertising materials mentioning features or use of this software 49 * must display the following acknowledgement: 50 * This product includes software developed by the NetBSD 51 * Foundation, Inc. and its contributors. 52 * 4. Neither the name of The NetBSD Foundation nor the names of its 53 * contributors may be used to endorse or promote products derived 54 * from this software without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 57 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 58 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 59 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 60 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 61 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 62 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 63 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 64 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 65 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 66 * POSSIBILITY OF SUCH DAMAGE. 67 */ 68 69/* 70 * Copyright (c) 1982, 1986, 1988, 1993 71 * The Regents of the University of California. All rights reserved. 72 * 73 * Redistribution and use in source and binary forms, with or without 74 * modification, are permitted provided that the following conditions 75 * are met: 76 * 1. Redistributions of source code must retain the above copyright 77 * notice, this list of conditions and the following disclaimer. 78 * 2. Redistributions in binary form must reproduce the above copyright 79 * notice, this list of conditions and the following disclaimer in the 80 * documentation and/or other materials provided with the distribution. 81 * 3. Neither the name of the University nor the names of its contributors 82 * may be used to endorse or promote products derived from this software 83 * without specific prior written permission. 84 * 85 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 86 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 87 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 88 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 89 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 90 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 91 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 92 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 93 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 94 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 95 * SUCH DAMAGE. 96 * 97 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 98 */ 99 100#include <sys/cdefs.h> 101__KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.200 2004/04/25 16:42:42 simonb Exp $"); 102 103#include "opt_inet.h" 104#include "opt_gateway.h" 105#include "opt_pfil_hooks.h" 106#include "opt_ipsec.h" 107#include "opt_mrouting.h" 108#include "opt_mbuftrace.h" 109#include "opt_inet_csum.h" 110 111#include <sys/param.h> 112#include <sys/systm.h> 113#include <sys/malloc.h> 114#include <sys/mbuf.h> 115#include <sys/domain.h> 116#include <sys/protosw.h> 117#include <sys/socket.h> 118#include <sys/socketvar.h> 119#include <sys/errno.h> 120#include <sys/time.h> 121#include <sys/kernel.h> 122#include <sys/pool.h> 123#include <sys/sysctl.h> 124 125#include <net/if.h> 126#include <net/if_dl.h> 127#include <net/route.h> 128#include <net/pfil.h> 129 130#include <netinet/in.h> 131#include <netinet/in_systm.h> 132#include <netinet/ip.h> 133#include <netinet/in_pcb.h> 134#include <netinet/in_var.h> 135#include <netinet/ip_var.h> 136#include <netinet/ip_icmp.h> 137/* just for gif_ttl */ 138#include <netinet/in_gif.h> 139#include "gif.h" 140#include <net/if_gre.h> 141#include "gre.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#ifdef FAST_IPSEC 152#include <netipsec/ipsec.h> 153#include <netipsec/key.h> 154#endif /* FAST_IPSEC*/ 155 156#ifndef IPFORWARDING 157#ifdef GATEWAY 158#define IPFORWARDING 1 /* forward IP packets not for us */ 159#else /* GATEWAY */ 160#define IPFORWARDING 0 /* don't forward IP packets not for us */ 161#endif /* GATEWAY */ 162#endif /* IPFORWARDING */ 163#ifndef IPSENDREDIRECTS 164#define IPSENDREDIRECTS 1 165#endif 166#ifndef IPFORWSRCRT 167#define IPFORWSRCRT 1 /* forward source-routed packets */ 168#endif 169#ifndef IPALLOWSRCRT 170#define IPALLOWSRCRT 1 /* allow source-routed packets */ 171#endif 172#ifndef IPMTUDISC 173#define IPMTUDISC 1 174#endif 175#ifndef IPMTUDISCTIMEOUT 176#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 177#endif 178 179/* 180 * Note: DIRECTED_BROADCAST is handled this way so that previous 181 * configuration using this option will Just Work. 182 */ 183#ifndef IPDIRECTEDBCAST 184#ifdef DIRECTED_BROADCAST 185#define IPDIRECTEDBCAST 1 186#else 187#define IPDIRECTEDBCAST 0 188#endif /* DIRECTED_BROADCAST */ 189#endif /* IPDIRECTEDBCAST */ 190int ipforwarding = IPFORWARDING; 191int ipsendredirects = IPSENDREDIRECTS; 192int ip_defttl = IPDEFTTL; 193int ip_forwsrcrt = IPFORWSRCRT; 194int ip_directedbcast = IPDIRECTEDBCAST; 195int ip_allowsrcrt = IPALLOWSRCRT; 196int ip_mtudisc = IPMTUDISC; 197int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 198#ifdef DIAGNOSTIC 199int ipprintfs = 0; 200#endif 201 202int ip_do_randomid = 0; 203 204/* 205 * XXX - Setting ip_checkinterface mostly implements the receive side of 206 * the Strong ES model described in RFC 1122, but since the routing table 207 * and transmit implementation do not implement the Strong ES model, 208 * setting this to 1 results in an odd hybrid. 209 * 210 * XXX - ip_checkinterface currently must be disabled if you use ipnat 211 * to translate the destination address to another local interface. 212 * 213 * XXX - ip_checkinterface must be disabled if you add IP aliases 214 * to the loopback interface instead of the interface where the 215 * packets for those addresses are received. 216 */ 217int ip_checkinterface = 0; 218 219 220struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 221 222int ipqmaxlen = IFQ_MAXLEN; 223u_long in_ifaddrhash; /* size of hash table - 1 */ 224int in_ifaddrentries; /* total number of addrs */ 225struct in_ifaddrhead in_ifaddrhead; 226struct in_ifaddrhashhead *in_ifaddrhashtbl; 227u_long in_multihash; /* size of hash table - 1 */ 228int in_multientries; /* total number of addrs */ 229struct in_multihashhead *in_multihashtbl; 230struct ifqueue ipintrq; 231struct ipstat ipstat; 232uint16_t ip_id; 233 234#ifdef PFIL_HOOKS 235struct pfil_head inet_pfil_hook; 236#endif 237 238/* 239 * Cached copy of nmbclusters. If nbclusters is different, 240 * recalculate IP parameters derived from nmbclusters. 241 */ 242static int ip_nmbclusters; /* copy of nmbclusters */ 243static void ip_nmbclusters_changed __P((void)); /* recalc limits */ 244 245#define CHECK_NMBCLUSTER_PARAMS() \ 246do { \ 247 if (__predict_false(ip_nmbclusters != nmbclusters)) \ 248 ip_nmbclusters_changed(); \ 249} while (/*CONSTCOND*/0) 250 251/* IP datagram reassembly queues (hashed) */ 252#define IPREASS_NHASH_LOG2 6 253#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 254#define IPREASS_HMASK (IPREASS_NHASH - 1) 255#define IPREASS_HASH(x,y) \ 256 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 257struct ipqhead ipq[IPREASS_NHASH]; 258int ipq_locked; 259static int ip_nfragpackets; /* packets in reass queue */ 260static int ip_nfrags; /* total fragments in reass queues */ 261 262int ip_maxfragpackets = 200; /* limit on packets. XXX sysctl */ 263int ip_maxfrags; /* limit on fragments. XXX sysctl */ 264 265 266/* 267 * Additive-Increase/Multiplicative-Decrease (AIMD) strategy for 268 * IP reassembly queue buffer managment. 269 * 270 * We keep a count of total IP fragments (NB: not fragmented packets!) 271 * awaiting reassembly (ip_nfrags) and a limit (ip_maxfrags) on fragments. 272 * If ip_nfrags exceeds ip_maxfrags the limit, we drop half the 273 * total fragments in reassembly queues.This AIMD policy avoids 274 * repeatedly deleting single packets under heavy fragmentation load 275 * (e.g., from lossy NFS peers). 276 */ 277static u_int ip_reass_ttl_decr __P((u_int ticks)); 278static void ip_reass_drophalf __P((void)); 279 280 281static __inline int ipq_lock_try __P((void)); 282static __inline void ipq_unlock __P((void)); 283 284static __inline int 285ipq_lock_try() 286{ 287 int s; 288 289 /* 290 * Use splvm() -- we're blocking things that would cause 291 * mbuf allocation. 292 */ 293 s = splvm(); 294 if (ipq_locked) { 295 splx(s); 296 return (0); 297 } 298 ipq_locked = 1; 299 splx(s); 300 return (1); 301} 302 303static __inline void 304ipq_unlock() 305{ 306 int s; 307 308 s = splvm(); 309 ipq_locked = 0; 310 splx(s); 311} 312 313#ifdef DIAGNOSTIC 314#define IPQ_LOCK() \ 315do { \ 316 if (ipq_lock_try() == 0) { \ 317 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 318 panic("ipq_lock"); \ 319 } \ 320} while (/*CONSTCOND*/ 0) 321#define IPQ_LOCK_CHECK() \ 322do { \ 323 if (ipq_locked == 0) { \ 324 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 325 panic("ipq lock check"); \ 326 } \ 327} while (/*CONSTCOND*/ 0) 328#else 329#define IPQ_LOCK() (void) ipq_lock_try() 330#define IPQ_LOCK_CHECK() /* nothing */ 331#endif 332 333#define IPQ_UNLOCK() ipq_unlock() 334 335POOL_INIT(inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl", NULL); 336POOL_INIT(ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", NULL); 337 338#ifdef INET_CSUM_COUNTERS 339#include <sys/device.h> 340 341struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 342 NULL, "inet", "hwcsum bad"); 343struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 344 NULL, "inet", "hwcsum ok"); 345struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 346 NULL, "inet", "swcsum"); 347 348#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 349 350#else 351 352#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 353 354#endif /* INET_CSUM_COUNTERS */ 355 356/* 357 * We need to save the IP options in case a protocol wants to respond 358 * to an incoming packet over the same route if the packet got here 359 * using IP source routing. This allows connection establishment and 360 * maintenance when the remote end is on a network that is not known 361 * to us. 362 */ 363int ip_nhops = 0; 364static struct ip_srcrt { 365 struct in_addr dst; /* final destination */ 366 char nop; /* one NOP to align */ 367 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 368 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 369} ip_srcrt; 370 371static void save_rte __P((u_char *, struct in_addr)); 372 373#ifdef MBUFTRACE 374struct mowner ip_rx_mowner = { "internet", "rx" }; 375struct mowner ip_tx_mowner = { "internet", "tx" }; 376#endif 377 378/* 379 * Compute IP limits derived from the value of nmbclusters. 380 */ 381static void 382ip_nmbclusters_changed(void) 383{ 384 ip_maxfrags = nmbclusters / 4; 385 ip_nmbclusters = nmbclusters; 386} 387 388/* 389 * IP initialization: fill in IP protocol switch table. 390 * All protocols not implemented in kernel go to raw IP protocol handler. 391 */ 392void 393ip_init() 394{ 395 const struct protosw *pr; 396 int i; 397 398 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 399 if (pr == 0) 400 panic("ip_init"); 401 for (i = 0; i < IPPROTO_MAX; i++) 402 ip_protox[i] = pr - inetsw; 403 for (pr = inetdomain.dom_protosw; 404 pr < inetdomain.dom_protoswNPROTOSW; pr++) 405 if (pr->pr_domain->dom_family == PF_INET && 406 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 407 ip_protox[pr->pr_protocol] = pr - inetsw; 408 409 for (i = 0; i < IPREASS_NHASH; i++) 410 LIST_INIT(&ipq[i]); 411 412 ip_id = time.tv_sec & 0xfffff; 413 414 ipintrq.ifq_maxlen = ipqmaxlen; 415 ip_nmbclusters_changed(); 416 417 TAILQ_INIT(&in_ifaddrhead); 418 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 419 M_WAITOK, &in_ifaddrhash); 420 in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IPMADDR, 421 M_WAITOK, &in_multihash); 422 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 423#ifdef GATEWAY 424 ipflow_init(); 425#endif 426 427#ifdef PFIL_HOOKS 428 /* Register our Packet Filter hook. */ 429 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 430 inet_pfil_hook.ph_af = AF_INET; 431 i = pfil_head_register(&inet_pfil_hook); 432 if (i != 0) 433 printf("ip_init: WARNING: unable to register pfil hook, " 434 "error %d\n", i); 435#endif /* PFIL_HOOKS */ 436 437#ifdef INET_CSUM_COUNTERS 438 evcnt_attach_static(&ip_hwcsum_bad); 439 evcnt_attach_static(&ip_hwcsum_ok); 440 evcnt_attach_static(&ip_swcsum); 441#endif /* INET_CSUM_COUNTERS */ 442 443#ifdef MBUFTRACE 444 MOWNER_ATTACH(&ip_tx_mowner); 445 MOWNER_ATTACH(&ip_rx_mowner); 446#endif /* MBUFTRACE */ 447} 448 449struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 450struct route ipforward_rt; 451 452/* 453 * IP software interrupt routine 454 */ 455void 456ipintr() 457{ 458 int s; 459 struct mbuf *m; 460 461 while (1) { 462 s = splnet(); 463 IF_DEQUEUE(&ipintrq, m); 464 splx(s); 465 if (m == 0) 466 return; 467 MCLAIM(m, &ip_rx_mowner); 468 ip_input(m); 469 } 470} 471 472/* 473 * Ip input routine. Checksum and byte swap header. If fragmented 474 * try to reassemble. Process options. Pass to next level. 475 */ 476void 477ip_input(struct mbuf *m) 478{ 479 struct ip *ip = NULL; 480 struct ipq *fp; 481 struct in_ifaddr *ia; 482 struct ifaddr *ifa; 483 struct ipqent *ipqe; 484 int hlen = 0, mff, len; 485 int downmatch; 486 int checkif; 487 int srcrt = 0; 488 u_int hash; 489#ifdef FAST_IPSEC 490 struct m_tag *mtag; 491 struct tdb_ident *tdbi; 492 struct secpolicy *sp; 493 int s, error; 494#endif /* FAST_IPSEC */ 495 496 MCLAIM(m, &ip_rx_mowner); 497#ifdef DIAGNOSTIC 498 if ((m->m_flags & M_PKTHDR) == 0) 499 panic("ipintr no HDR"); 500#endif 501 502 /* 503 * If no IP addresses have been set yet but the interfaces 504 * are receiving, can't do anything with incoming packets yet. 505 */ 506 if (TAILQ_FIRST(&in_ifaddrhead) == 0) 507 goto bad; 508 ipstat.ips_total++; 509 /* 510 * If the IP header is not aligned, slurp it up into a new 511 * mbuf with space for link headers, in the event we forward 512 * it. Otherwise, if it is aligned, make sure the entire 513 * base IP header is in the first mbuf of the chain. 514 */ 515 if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) { 516 if ((m = m_copyup(m, sizeof(struct ip), 517 (max_linkhdr + 3) & ~3)) == NULL) { 518 /* XXXJRT new stat, please */ 519 ipstat.ips_toosmall++; 520 return; 521 } 522 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 523 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 524 ipstat.ips_toosmall++; 525 return; 526 } 527 } 528 ip = mtod(m, struct ip *); 529 if (ip->ip_v != IPVERSION) { 530 ipstat.ips_badvers++; 531 goto bad; 532 } 533 hlen = ip->ip_hl << 2; 534 if (hlen < sizeof(struct ip)) { /* minimum header length */ 535 ipstat.ips_badhlen++; 536 goto bad; 537 } 538 if (hlen > m->m_len) { 539 if ((m = m_pullup(m, hlen)) == 0) { 540 ipstat.ips_badhlen++; 541 return; 542 } 543 ip = mtod(m, struct ip *); 544 } 545 546 /* 547 * RFC1122: packets with a multicast source address are 548 * not allowed. 549 */ 550 if (IN_MULTICAST(ip->ip_src.s_addr)) { 551 ipstat.ips_badaddr++; 552 goto bad; 553 } 554 555 /* 127/8 must not appear on wire - RFC1122 */ 556 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 557 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 558 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 559 ipstat.ips_badaddr++; 560 goto bad; 561 } 562 } 563 564 switch (m->m_pkthdr.csum_flags & 565 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) | 566 M_CSUM_IPv4_BAD)) { 567 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 568 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 569 goto badcsum; 570 571 case M_CSUM_IPv4: 572 /* Checksum was okay. */ 573 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 574 break; 575 576 default: 577 /* Must compute it ourselves. */ 578 INET_CSUM_COUNTER_INCR(&ip_swcsum); 579 if (in_cksum(m, hlen) != 0) 580 goto bad; 581 break; 582 } 583 584 /* Retrieve the packet length. */ 585 len = ntohs(ip->ip_len); 586 587 /* 588 * Check for additional length bogosity 589 */ 590 if (len < hlen) { 591 ipstat.ips_badlen++; 592 goto bad; 593 } 594 595 /* 596 * Check that the amount of data in the buffers 597 * is as at least much as the IP header would have us expect. 598 * Trim mbufs if longer than we expect. 599 * Drop packet if shorter than we expect. 600 */ 601 if (m->m_pkthdr.len < len) { 602 ipstat.ips_tooshort++; 603 goto bad; 604 } 605 if (m->m_pkthdr.len > len) { 606 if (m->m_len == m->m_pkthdr.len) { 607 m->m_len = len; 608 m->m_pkthdr.len = len; 609 } else 610 m_adj(m, len - m->m_pkthdr.len); 611 } 612 613#if defined(IPSEC) 614 /* ipflow (IP fast forwarding) is not compatible with IPsec. */ 615 m->m_flags &= ~M_CANFASTFWD; 616#else 617 /* 618 * Assume that we can create a fast-forward IP flow entry 619 * based on this packet. 620 */ 621 m->m_flags |= M_CANFASTFWD; 622#endif 623 624#ifdef PFIL_HOOKS 625 /* 626 * Run through list of hooks for input packets. If there are any 627 * filters which require that additional packets in the flow are 628 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 629 * Note that filters must _never_ set this flag, as another filter 630 * in the list may have previously cleared it. 631 */ 632 /* 633 * let ipfilter look at packet on the wire, 634 * not the decapsulated packet. 635 */ 636#ifdef IPSEC 637 if (!ipsec_getnhist(m)) 638#elif defined(FAST_IPSEC) 639 if (!ipsec_indone(m)) 640#else 641 if (1) 642#endif 643 { 644 struct in_addr odst; 645 646 odst = ip->ip_dst; 647 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 648 PFIL_IN) != 0) 649 return; 650 if (m == NULL) 651 return; 652 ip = mtod(m, struct ip *); 653 hlen = ip->ip_hl << 2; 654 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 655 } 656#endif /* PFIL_HOOKS */ 657 658#ifdef ALTQ 659 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 660 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 661 /* packet dropped by traffic conditioner */ 662 return; 663 } 664#endif 665 666 /* 667 * Process options and, if not destined for us, 668 * ship it on. ip_dooptions returns 1 when an 669 * error was detected (causing an icmp message 670 * to be sent and the original packet to be freed). 671 */ 672 ip_nhops = 0; /* for source routed packets */ 673 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 674 return; 675 676 /* 677 * Enable a consistency check between the destination address 678 * and the arrival interface for a unicast packet (the RFC 1122 679 * strong ES model) if IP forwarding is disabled and the packet 680 * is not locally generated. 681 * 682 * XXX - Checking also should be disabled if the destination 683 * address is ipnat'ed to a different interface. 684 * 685 * XXX - Checking is incompatible with IP aliases added 686 * to the loopback interface instead of the interface where 687 * the packets are received. 688 * 689 * XXX - We need to add a per ifaddr flag for this so that 690 * we get finer grain control. 691 */ 692 checkif = ip_checkinterface && (ipforwarding == 0) && 693 (m->m_pkthdr.rcvif != NULL) && 694 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0); 695 696 /* 697 * Check our list of addresses, to see if the packet is for us. 698 * 699 * Traditional 4.4BSD did not consult IFF_UP at all. 700 * The behavior here is to treat addresses on !IFF_UP interface 701 * as not mine. 702 */ 703 downmatch = 0; 704 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 705 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 706 if (checkif && ia->ia_ifp != m->m_pkthdr.rcvif) 707 continue; 708 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 709 break; 710 else 711 downmatch++; 712 } 713 } 714 if (ia != NULL) 715 goto ours; 716 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 717 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrlist, ifa_list) { 718 if (ifa->ifa_addr->sa_family != AF_INET) 719 continue; 720 ia = ifatoia(ifa); 721 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 722 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 723 /* 724 * Look for all-0's host part (old broadcast addr), 725 * either for subnet or net. 726 */ 727 ip->ip_dst.s_addr == ia->ia_subnet || 728 ip->ip_dst.s_addr == ia->ia_net) 729 goto ours; 730 /* 731 * An interface with IP address zero accepts 732 * all packets that arrive on that interface. 733 */ 734 if (in_nullhost(ia->ia_addr.sin_addr)) 735 goto ours; 736 } 737 } 738 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 739 struct in_multi *inm; 740#ifdef MROUTING 741 extern struct socket *ip_mrouter; 742 743 if (M_READONLY(m)) { 744 if ((m = m_pullup(m, hlen)) == 0) { 745 ipstat.ips_toosmall++; 746 return; 747 } 748 ip = mtod(m, struct ip *); 749 } 750 751 if (ip_mrouter) { 752 /* 753 * If we are acting as a multicast router, all 754 * incoming multicast packets are passed to the 755 * kernel-level multicast forwarding function. 756 * The packet is returned (relatively) intact; if 757 * ip_mforward() returns a non-zero value, the packet 758 * must be discarded, else it may be accepted below. 759 * 760 * (The IP ident field is put in the same byte order 761 * as expected when ip_mforward() is called from 762 * ip_output().) 763 */ 764 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 765 ipstat.ips_cantforward++; 766 m_freem(m); 767 return; 768 } 769 770 /* 771 * The process-level routing demon needs to receive 772 * all multicast IGMP packets, whether or not this 773 * host belongs to their destination groups. 774 */ 775 if (ip->ip_p == IPPROTO_IGMP) 776 goto ours; 777 ipstat.ips_forward++; 778 } 779#endif 780 /* 781 * See if we belong to the destination multicast group on the 782 * arrival interface. 783 */ 784 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 785 if (inm == NULL) { 786 ipstat.ips_cantforward++; 787 m_freem(m); 788 return; 789 } 790 goto ours; 791 } 792 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 793 in_nullhost(ip->ip_dst)) 794 goto ours; 795 796 /* 797 * Not for us; forward if possible and desirable. 798 */ 799 if (ipforwarding == 0) { 800 ipstat.ips_cantforward++; 801 m_freem(m); 802 } else { 803 /* 804 * If ip_dst matched any of my address on !IFF_UP interface, 805 * and there's no IFF_UP interface that matches ip_dst, 806 * send icmp unreach. Forwarding it will result in in-kernel 807 * forwarding loop till TTL goes to 0. 808 */ 809 if (downmatch) { 810 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 811 ipstat.ips_cantforward++; 812 return; 813 } 814#ifdef IPSEC 815 if (ipsec4_in_reject(m, NULL)) { 816 ipsecstat.in_polvio++; 817 goto bad; 818 } 819#endif 820#ifdef FAST_IPSEC 821 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 822 s = splsoftnet(); 823 if (mtag != NULL) { 824 tdbi = (struct tdb_ident *)(mtag + 1); 825 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 826 } else { 827 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 828 IP_FORWARDING, &error); 829 } 830 if (sp == NULL) { /* NB: can happen if error */ 831 splx(s); 832 /*XXX error stat???*/ 833 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 834 goto bad; 835 } 836 837 /* 838 * Check security policy against packet attributes. 839 */ 840 error = ipsec_in_reject(sp, m); 841 KEY_FREESP(&sp); 842 splx(s); 843 if (error) { 844 ipstat.ips_cantforward++; 845 goto bad; 846 } 847 848 /* 849 * Peek at the outbound SP for this packet to determine if 850 * it's a Fast Forward candidate. 851 */ 852 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 853 if (mtag != NULL) 854 m->m_flags &= ~M_CANFASTFWD; 855 else { 856 s = splsoftnet(); 857 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, 858 (IP_FORWARDING | 859 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 860 &error, NULL); 861 if (sp != NULL) { 862 m->m_flags &= ~M_CANFASTFWD; 863 KEY_FREESP(&sp); 864 } 865 splx(s); 866 } 867#endif /* FAST_IPSEC */ 868 869 ip_forward(m, srcrt); 870 } 871 return; 872 873ours: 874 /* 875 * If offset or IP_MF are set, must reassemble. 876 * Otherwise, nothing need be done. 877 * (We could look in the reassembly queue to see 878 * if the packet was previously fragmented, 879 * but it's not worth the time; just let them time out.) 880 */ 881 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 882 if (M_READONLY(m)) { 883 if ((m = m_pullup(m, hlen)) == NULL) { 884 ipstat.ips_toosmall++; 885 goto bad; 886 } 887 ip = mtod(m, struct ip *); 888 } 889 890 /* 891 * Look for queue of fragments 892 * of this datagram. 893 */ 894 IPQ_LOCK(); 895 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 896 /* XXX LIST_FOREACH(fp, &ipq[hash], ipq_q) */ 897 for (fp = LIST_FIRST(&ipq[hash]); fp != NULL; 898 fp = LIST_NEXT(fp, ipq_q)) { 899 if (ip->ip_id == fp->ipq_id && 900 in_hosteq(ip->ip_src, fp->ipq_src) && 901 in_hosteq(ip->ip_dst, fp->ipq_dst) && 902 ip->ip_p == fp->ipq_p) 903 goto found; 904 905 } 906 fp = 0; 907found: 908 909 /* 910 * Adjust ip_len to not reflect header, 911 * set ipqe_mff if more fragments are expected, 912 * convert offset of this to bytes. 913 */ 914 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 915 mff = (ip->ip_off & htons(IP_MF)) != 0; 916 if (mff) { 917 /* 918 * Make sure that fragments have a data length 919 * that's a non-zero multiple of 8 bytes. 920 */ 921 if (ntohs(ip->ip_len) == 0 || 922 (ntohs(ip->ip_len) & 0x7) != 0) { 923 ipstat.ips_badfrags++; 924 IPQ_UNLOCK(); 925 goto bad; 926 } 927 } 928 ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3); 929 930 /* 931 * If datagram marked as having more fragments 932 * or if this is not the first fragment, 933 * attempt reassembly; if it succeeds, proceed. 934 */ 935 if (mff || ip->ip_off != htons(0)) { 936 ipstat.ips_fragments++; 937 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 938 if (ipqe == NULL) { 939 ipstat.ips_rcvmemdrop++; 940 IPQ_UNLOCK(); 941 goto bad; 942 } 943 ipqe->ipqe_mff = mff; 944 ipqe->ipqe_m = m; 945 ipqe->ipqe_ip = ip; 946 m = ip_reass(ipqe, fp, &ipq[hash]); 947 if (m == 0) { 948 IPQ_UNLOCK(); 949 return; 950 } 951 ipstat.ips_reassembled++; 952 ip = mtod(m, struct ip *); 953 hlen = ip->ip_hl << 2; 954 ip->ip_len = htons(ntohs(ip->ip_len) + hlen); 955 } else 956 if (fp) 957 ip_freef(fp); 958 IPQ_UNLOCK(); 959 } 960 961#if defined(IPSEC) 962 /* 963 * enforce IPsec policy checking if we are seeing last header. 964 * note that we do not visit this with protocols with pcb layer 965 * code - like udp/tcp/raw ip. 966 */ 967 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 968 ipsec4_in_reject(m, NULL)) { 969 ipsecstat.in_polvio++; 970 goto bad; 971 } 972#endif 973#if FAST_IPSEC 974 /* 975 * enforce IPsec policy checking if we are seeing last header. 976 * note that we do not visit this with protocols with pcb layer 977 * code - like udp/tcp/raw ip. 978 */ 979 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 980 /* 981 * Check if the packet has already had IPsec processing 982 * done. If so, then just pass it along. This tag gets 983 * set during AH, ESP, etc. input handling, before the 984 * packet is returned to the ip input queue for delivery. 985 */ 986 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 987 s = splsoftnet(); 988 if (mtag != NULL) { 989 tdbi = (struct tdb_ident *)(mtag + 1); 990 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 991 } else { 992 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 993 IP_FORWARDING, &error); 994 } 995 if (sp != NULL) { 996 /* 997 * Check security policy against packet attributes. 998 */ 999 error = ipsec_in_reject(sp, m); 1000 KEY_FREESP(&sp); 1001 } else { 1002 /* XXX error stat??? */ 1003 error = EINVAL; 1004DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 1005 goto bad; 1006 } 1007 splx(s); 1008 if (error) 1009 goto bad; 1010 } 1011#endif /* FAST_IPSEC */ 1012 1013 /* 1014 * Switch out to protocol's input routine. 1015 */ 1016#if IFA_STATS 1017 if (ia && ip) 1018 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 1019#endif 1020 ipstat.ips_delivered++; 1021 { 1022 int off = hlen, nh = ip->ip_p; 1023 1024 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 1025 return; 1026 } 1027bad: 1028 m_freem(m); 1029 return; 1030 1031badcsum: 1032 ipstat.ips_badsum++; 1033 m_freem(m); 1034} 1035 1036/* 1037 * Take incoming datagram fragment and try to 1038 * reassemble it into whole datagram. If a chain for 1039 * reassembly of this datagram already exists, then it 1040 * is given as fp; otherwise have to make a chain. 1041 */ 1042struct mbuf * 1043ip_reass(ipqe, fp, ipqhead) 1044 struct ipqent *ipqe; 1045 struct ipq *fp; 1046 struct ipqhead *ipqhead; 1047{ 1048 struct mbuf *m = ipqe->ipqe_m; 1049 struct ipqent *nq, *p, *q; 1050 struct ip *ip; 1051 struct mbuf *t; 1052 int hlen = ipqe->ipqe_ip->ip_hl << 2; 1053 int i, next; 1054 1055 IPQ_LOCK_CHECK(); 1056 1057 /* 1058 * Presence of header sizes in mbufs 1059 * would confuse code below. 1060 */ 1061 m->m_data += hlen; 1062 m->m_len -= hlen; 1063 1064#ifdef notyet 1065 /* make sure fragment limit is up-to-date */ 1066 CHECK_NMBCLUSTER_PARAMS(); 1067 1068 /* If we have too many fragments, drop the older half. */ 1069 if (ip_nfrags >= ip_maxfrags) 1070 ip_reass_drophalf(void); 1071#endif 1072 1073 /* 1074 * We are about to add a fragment; increment frag count. 1075 */ 1076 ip_nfrags++; 1077 1078 /* 1079 * If first fragment to arrive, create a reassembly queue. 1080 */ 1081 if (fp == 0) { 1082 /* 1083 * Enforce upper bound on number of fragmented packets 1084 * for which we attempt reassembly; 1085 * If maxfrag is 0, never accept fragments. 1086 * If maxfrag is -1, accept all fragments without limitation. 1087 */ 1088 if (ip_maxfragpackets < 0) 1089 ; 1090 else if (ip_nfragpackets >= ip_maxfragpackets) 1091 goto dropfrag; 1092 ip_nfragpackets++; 1093 MALLOC(fp, struct ipq *, sizeof (struct ipq), 1094 M_FTABLE, M_NOWAIT); 1095 if (fp == NULL) 1096 goto dropfrag; 1097 LIST_INSERT_HEAD(ipqhead, fp, ipq_q); 1098 fp->ipq_nfrags = 1; 1099 fp->ipq_ttl = IPFRAGTTL; 1100 fp->ipq_p = ipqe->ipqe_ip->ip_p; 1101 fp->ipq_id = ipqe->ipqe_ip->ip_id; 1102 TAILQ_INIT(&fp->ipq_fragq); 1103 fp->ipq_src = ipqe->ipqe_ip->ip_src; 1104 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 1105 p = NULL; 1106 goto insert; 1107 } else { 1108 fp->ipq_nfrags++; 1109 } 1110 1111 /* 1112 * Find a segment which begins after this one does. 1113 */ 1114 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1115 p = q, q = TAILQ_NEXT(q, ipqe_q)) 1116 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off)) 1117 break; 1118 1119 /* 1120 * If there is a preceding segment, it may provide some of 1121 * our data already. If so, drop the data from the incoming 1122 * segment. If it provides all of our data, drop us. 1123 */ 1124 if (p != NULL) { 1125 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) - 1126 ntohs(ipqe->ipqe_ip->ip_off); 1127 if (i > 0) { 1128 if (i >= ntohs(ipqe->ipqe_ip->ip_len)) 1129 goto dropfrag; 1130 m_adj(ipqe->ipqe_m, i); 1131 ipqe->ipqe_ip->ip_off = 1132 htons(ntohs(ipqe->ipqe_ip->ip_off) + i); 1133 ipqe->ipqe_ip->ip_len = 1134 htons(ntohs(ipqe->ipqe_ip->ip_len) - i); 1135 } 1136 } 1137 1138 /* 1139 * While we overlap succeeding segments trim them or, 1140 * if they are completely covered, dequeue them. 1141 */ 1142 for (; q != NULL && 1143 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) > 1144 ntohs(q->ipqe_ip->ip_off); q = nq) { 1145 i = (ntohs(ipqe->ipqe_ip->ip_off) + 1146 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off); 1147 if (i < ntohs(q->ipqe_ip->ip_len)) { 1148 q->ipqe_ip->ip_len = 1149 htons(ntohs(q->ipqe_ip->ip_len) - i); 1150 q->ipqe_ip->ip_off = 1151 htons(ntohs(q->ipqe_ip->ip_off) + i); 1152 m_adj(q->ipqe_m, i); 1153 break; 1154 } 1155 nq = TAILQ_NEXT(q, ipqe_q); 1156 m_freem(q->ipqe_m); 1157 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1158 pool_put(&ipqent_pool, q); 1159 fp->ipq_nfrags--; 1160 ip_nfrags--; 1161 } 1162 1163insert: 1164 /* 1165 * Stick new segment in its place; 1166 * check for complete reassembly. 1167 */ 1168 if (p == NULL) { 1169 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 1170 } else { 1171 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q); 1172 } 1173 next = 0; 1174 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1175 p = q, q = TAILQ_NEXT(q, ipqe_q)) { 1176 if (ntohs(q->ipqe_ip->ip_off) != next) 1177 return (0); 1178 next += ntohs(q->ipqe_ip->ip_len); 1179 } 1180 if (p->ipqe_mff) 1181 return (0); 1182 1183 /* 1184 * Reassembly is complete. Check for a bogus message size and 1185 * concatenate fragments. 1186 */ 1187 q = TAILQ_FIRST(&fp->ipq_fragq); 1188 ip = q->ipqe_ip; 1189 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 1190 ipstat.ips_toolong++; 1191 ip_freef(fp); 1192 return (0); 1193 } 1194 m = q->ipqe_m; 1195 t = m->m_next; 1196 m->m_next = 0; 1197 m_cat(m, t); 1198 nq = TAILQ_NEXT(q, ipqe_q); 1199 pool_put(&ipqent_pool, q); 1200 for (q = nq; q != NULL; q = nq) { 1201 t = q->ipqe_m; 1202 nq = TAILQ_NEXT(q, ipqe_q); 1203 pool_put(&ipqent_pool, q); 1204 m_cat(m, t); 1205 } 1206 ip_nfrags -= fp->ipq_nfrags; 1207 1208 /* 1209 * Create header for new ip packet by 1210 * modifying header of first packet; 1211 * dequeue and discard fragment reassembly header. 1212 * Make header visible. 1213 */ 1214 ip->ip_len = htons(next); 1215 ip->ip_src = fp->ipq_src; 1216 ip->ip_dst = fp->ipq_dst; 1217 LIST_REMOVE(fp, ipq_q); 1218 FREE(fp, M_FTABLE); 1219 ip_nfragpackets--; 1220 m->m_len += (ip->ip_hl << 2); 1221 m->m_data -= (ip->ip_hl << 2); 1222 /* some debugging cruft by sklower, below, will go away soon */ 1223 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1224 int plen = 0; 1225 for (t = m; t; t = t->m_next) 1226 plen += t->m_len; 1227 m->m_pkthdr.len = plen; 1228 } 1229 return (m); 1230 1231dropfrag: 1232 if (fp != 0) 1233 fp->ipq_nfrags--; 1234 ip_nfrags--; 1235 ipstat.ips_fragdropped++; 1236 m_freem(m); 1237 pool_put(&ipqent_pool, ipqe); 1238 return (0); 1239} 1240 1241/* 1242 * Free a fragment reassembly header and all 1243 * associated datagrams. 1244 */ 1245void 1246ip_freef(fp) 1247 struct ipq *fp; 1248{ 1249 struct ipqent *q, *p; 1250 u_int nfrags = 0; 1251 1252 IPQ_LOCK_CHECK(); 1253 1254 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) { 1255 p = TAILQ_NEXT(q, ipqe_q); 1256 m_freem(q->ipqe_m); 1257 nfrags++; 1258 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1259 pool_put(&ipqent_pool, q); 1260 } 1261 1262 if (nfrags != fp->ipq_nfrags) 1263 printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags); 1264 ip_nfrags -= nfrags; 1265 LIST_REMOVE(fp, ipq_q); 1266 FREE(fp, M_FTABLE); 1267 ip_nfragpackets--; 1268} 1269 1270/* 1271 * IP reassembly TTL machinery for multiplicative drop. 1272 */ 1273static u_int fragttl_histo[(IPFRAGTTL+1)]; 1274 1275 1276/* 1277 * Decrement TTL of all reasembly queue entries by `ticks'. 1278 * Count number of distinct fragments (as opposed to partial, fragmented 1279 * datagrams) in the reassembly queue. While we traverse the entire 1280 * reassembly queue, compute and return the median TTL over all fragments. 1281 */ 1282static u_int 1283ip_reass_ttl_decr(u_int ticks) 1284{ 1285 u_int nfrags, median, dropfraction, keepfraction; 1286 struct ipq *fp, *nfp; 1287 int i; 1288 1289 nfrags = 0; 1290 memset(fragttl_histo, 0, sizeof fragttl_histo); 1291 1292 for (i = 0; i < IPREASS_NHASH; i++) { 1293 for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) { 1294 fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ? 1295 0 : fp->ipq_ttl - ticks); 1296 nfp = LIST_NEXT(fp, ipq_q); 1297 if (fp->ipq_ttl == 0) { 1298 ipstat.ips_fragtimeout++; 1299 ip_freef(fp); 1300 } else { 1301 nfrags += fp->ipq_nfrags; 1302 fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags; 1303 } 1304 } 1305 } 1306 1307 KASSERT(ip_nfrags == nfrags); 1308 1309 /* Find median (or other drop fraction) in histogram. */ 1310 dropfraction = (ip_nfrags / 2); 1311 keepfraction = ip_nfrags - dropfraction; 1312 for (i = IPFRAGTTL, median = 0; i >= 0; i--) { 1313 median += fragttl_histo[i]; 1314 if (median >= keepfraction) 1315 break; 1316 } 1317 1318 /* Return TTL of median (or other fraction). */ 1319 return (u_int)i; 1320} 1321 1322void 1323ip_reass_drophalf(void) 1324{ 1325 1326 u_int median_ticks; 1327 /* 1328 * Compute median TTL of all fragments, and count frags 1329 * with that TTL or lower (roughly half of all fragments). 1330 */ 1331 median_ticks = ip_reass_ttl_decr(0); 1332 1333 /* Drop half. */ 1334 median_ticks = ip_reass_ttl_decr(median_ticks); 1335 1336} 1337 1338/* 1339 * IP timer processing; 1340 * if a timer expires on a reassembly 1341 * queue, discard it. 1342 */ 1343void 1344ip_slowtimo() 1345{ 1346 static u_int dropscanidx = 0; 1347 u_int i; 1348 u_int median_ttl; 1349 int s = splsoftnet(); 1350 1351 IPQ_LOCK(); 1352 1353 /* Age TTL of all fragments by 1 tick .*/ 1354 median_ttl = ip_reass_ttl_decr(1); 1355 1356 /* make sure fragment limit is up-to-date */ 1357 CHECK_NMBCLUSTER_PARAMS(); 1358 1359 /* If we have too many fragments, drop the older half. */ 1360 if (ip_nfrags > ip_maxfrags) 1361 ip_reass_ttl_decr(median_ttl); 1362 1363 /* 1364 * If we are over the maximum number of fragmented packets 1365 * (due to the limit being lowered), drain off 1366 * enough to get down to the new limit. Start draining 1367 * from the reassembly hashqueue most recently drained. 1368 */ 1369 if (ip_maxfragpackets < 0) 1370 ; 1371 else { 1372 int wrapped = 0; 1373 1374 i = dropscanidx; 1375 while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) { 1376 while (LIST_FIRST(&ipq[i]) != NULL) 1377 ip_freef(LIST_FIRST(&ipq[i])); 1378 if (++i >= IPREASS_NHASH) { 1379 i = 0; 1380 } 1381 /* 1382 * Dont scan forever even if fragment counters are 1383 * wrong: stop after scanning entire reassembly queue. 1384 */ 1385 if (i == dropscanidx) 1386 wrapped = 1; 1387 } 1388 dropscanidx = i; 1389 } 1390 IPQ_UNLOCK(); 1391#ifdef GATEWAY 1392 ipflow_slowtimo(); 1393#endif 1394 splx(s); 1395} 1396 1397/* 1398 * Drain off all datagram fragments. 1399 */ 1400void 1401ip_drain() 1402{ 1403 1404 /* 1405 * We may be called from a device's interrupt context. If 1406 * the ipq is already busy, just bail out now. 1407 */ 1408 if (ipq_lock_try() == 0) 1409 return; 1410 1411 /* 1412 * Drop half the total fragments now. If more mbufs are needed, 1413 * we will be called again soon. 1414 */ 1415 ip_reass_drophalf(); 1416 1417 IPQ_UNLOCK(); 1418} 1419 1420/* 1421 * Do option processing on a datagram, 1422 * possibly discarding it if bad options are encountered, 1423 * or forwarding it if source-routed. 1424 * Returns 1 if packet has been forwarded/freed, 1425 * 0 if the packet should be processed further. 1426 */ 1427int 1428ip_dooptions(m) 1429 struct mbuf *m; 1430{ 1431 struct ip *ip = mtod(m, struct ip *); 1432 u_char *cp, *cp0; 1433 struct ip_timestamp *ipt; 1434 struct in_ifaddr *ia; 1435 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1436 struct in_addr dst; 1437 n_time ntime; 1438 1439 dst = ip->ip_dst; 1440 cp = (u_char *)(ip + 1); 1441 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1442 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1443 opt = cp[IPOPT_OPTVAL]; 1444 if (opt == IPOPT_EOL) 1445 break; 1446 if (opt == IPOPT_NOP) 1447 optlen = 1; 1448 else { 1449 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1450 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1451 goto bad; 1452 } 1453 optlen = cp[IPOPT_OLEN]; 1454 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1455 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1456 goto bad; 1457 } 1458 } 1459 switch (opt) { 1460 1461 default: 1462 break; 1463 1464 /* 1465 * Source routing with record. 1466 * Find interface with current destination address. 1467 * If none on this machine then drop if strictly routed, 1468 * or do nothing if loosely routed. 1469 * Record interface address and bring up next address 1470 * component. If strictly routed make sure next 1471 * address is on directly accessible net. 1472 */ 1473 case IPOPT_LSRR: 1474 case IPOPT_SSRR: 1475 if (ip_allowsrcrt == 0) { 1476 type = ICMP_UNREACH; 1477 code = ICMP_UNREACH_NET_PROHIB; 1478 goto bad; 1479 } 1480 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1481 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1482 goto bad; 1483 } 1484 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1485 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1486 goto bad; 1487 } 1488 ipaddr.sin_addr = ip->ip_dst; 1489 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1490 if (ia == 0) { 1491 if (opt == IPOPT_SSRR) { 1492 type = ICMP_UNREACH; 1493 code = ICMP_UNREACH_SRCFAIL; 1494 goto bad; 1495 } 1496 /* 1497 * Loose routing, and not at next destination 1498 * yet; nothing to do except forward. 1499 */ 1500 break; 1501 } 1502 off--; /* 0 origin */ 1503 if ((off + sizeof(struct in_addr)) > optlen) { 1504 /* 1505 * End of source route. Should be for us. 1506 */ 1507 save_rte(cp, ip->ip_src); 1508 break; 1509 } 1510 /* 1511 * locate outgoing interface 1512 */ 1513 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1514 sizeof(ipaddr.sin_addr)); 1515 if (opt == IPOPT_SSRR) 1516 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); 1517 else 1518 ia = ip_rtaddr(ipaddr.sin_addr); 1519 if (ia == 0) { 1520 type = ICMP_UNREACH; 1521 code = ICMP_UNREACH_SRCFAIL; 1522 goto bad; 1523 } 1524 ip->ip_dst = ipaddr.sin_addr; 1525 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1526 (caddr_t)(cp + off), sizeof(struct in_addr)); 1527 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1528 /* 1529 * Let ip_intr's mcast routing check handle mcast pkts 1530 */ 1531 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1532 break; 1533 1534 case IPOPT_RR: 1535 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1536 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1537 goto bad; 1538 } 1539 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1540 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1541 goto bad; 1542 } 1543 /* 1544 * If no space remains, ignore. 1545 */ 1546 off--; /* 0 origin */ 1547 if ((off + sizeof(struct in_addr)) > optlen) 1548 break; 1549 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1550 sizeof(ipaddr.sin_addr)); 1551 /* 1552 * locate outgoing interface; if we're the destination, 1553 * use the incoming interface (should be same). 1554 */ 1555 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1556 == NULL && 1557 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1558 type = ICMP_UNREACH; 1559 code = ICMP_UNREACH_HOST; 1560 goto bad; 1561 } 1562 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1563 (caddr_t)(cp + off), sizeof(struct in_addr)); 1564 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1565 break; 1566 1567 case IPOPT_TS: 1568 code = cp - (u_char *)ip; 1569 ipt = (struct ip_timestamp *)cp; 1570 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1571 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1572 goto bad; 1573 } 1574 if (ipt->ipt_ptr < 5) { 1575 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1576 goto bad; 1577 } 1578 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1579 if (++ipt->ipt_oflw == 0) { 1580 code = (u_char *)&ipt->ipt_ptr - 1581 (u_char *)ip; 1582 goto bad; 1583 } 1584 break; 1585 } 1586 cp0 = (cp + ipt->ipt_ptr - 1); 1587 switch (ipt->ipt_flg) { 1588 1589 case IPOPT_TS_TSONLY: 1590 break; 1591 1592 case IPOPT_TS_TSANDADDR: 1593 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1594 sizeof(struct in_addr) > ipt->ipt_len) { 1595 code = (u_char *)&ipt->ipt_ptr - 1596 (u_char *)ip; 1597 goto bad; 1598 } 1599 ipaddr.sin_addr = dst; 1600 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1601 m->m_pkthdr.rcvif)); 1602 if (ia == 0) 1603 continue; 1604 bcopy(&ia->ia_addr.sin_addr, 1605 cp0, sizeof(struct in_addr)); 1606 ipt->ipt_ptr += sizeof(struct in_addr); 1607 break; 1608 1609 case IPOPT_TS_PRESPEC: 1610 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1611 sizeof(struct in_addr) > ipt->ipt_len) { 1612 code = (u_char *)&ipt->ipt_ptr - 1613 (u_char *)ip; 1614 goto bad; 1615 } 1616 bcopy(cp0, &ipaddr.sin_addr, 1617 sizeof(struct in_addr)); 1618 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1619 == NULL) 1620 continue; 1621 ipt->ipt_ptr += sizeof(struct in_addr); 1622 break; 1623 1624 default: 1625 /* XXX can't take &ipt->ipt_flg */ 1626 code = (u_char *)&ipt->ipt_ptr - 1627 (u_char *)ip + 1; 1628 goto bad; 1629 } 1630 ntime = iptime(); 1631 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1632 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1633 sizeof(n_time)); 1634 ipt->ipt_ptr += sizeof(n_time); 1635 } 1636 } 1637 if (forward) { 1638 if (ip_forwsrcrt == 0) { 1639 type = ICMP_UNREACH; 1640 code = ICMP_UNREACH_SRCFAIL; 1641 goto bad; 1642 } 1643 ip_forward(m, 1); 1644 return (1); 1645 } 1646 return (0); 1647bad: 1648 icmp_error(m, type, code, 0, 0); 1649 ipstat.ips_badoptions++; 1650 return (1); 1651} 1652 1653/* 1654 * Given address of next destination (final or next hop), 1655 * return internet address info of interface to be used to get there. 1656 */ 1657struct in_ifaddr * 1658ip_rtaddr(dst) 1659 struct in_addr dst; 1660{ 1661 struct sockaddr_in *sin; 1662 1663 sin = satosin(&ipforward_rt.ro_dst); 1664 1665 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1666 if (ipforward_rt.ro_rt) { 1667 RTFREE(ipforward_rt.ro_rt); 1668 ipforward_rt.ro_rt = 0; 1669 } 1670 sin->sin_family = AF_INET; 1671 sin->sin_len = sizeof(*sin); 1672 sin->sin_addr = dst; 1673 1674 rtalloc(&ipforward_rt); 1675 } 1676 if (ipforward_rt.ro_rt == 0) 1677 return ((struct in_ifaddr *)0); 1678 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1679} 1680 1681/* 1682 * Save incoming source route for use in replies, 1683 * to be picked up later by ip_srcroute if the receiver is interested. 1684 */ 1685void 1686save_rte(option, dst) 1687 u_char *option; 1688 struct in_addr dst; 1689{ 1690 unsigned olen; 1691 1692 olen = option[IPOPT_OLEN]; 1693#ifdef DIAGNOSTIC 1694 if (ipprintfs) 1695 printf("save_rte: olen %d\n", olen); 1696#endif /* 0 */ 1697 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1698 return; 1699 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1700 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1701 ip_srcrt.dst = dst; 1702} 1703 1704/* 1705 * Retrieve incoming source route for use in replies, 1706 * in the same form used by setsockopt. 1707 * The first hop is placed before the options, will be removed later. 1708 */ 1709struct mbuf * 1710ip_srcroute() 1711{ 1712 struct in_addr *p, *q; 1713 struct mbuf *m; 1714 1715 if (ip_nhops == 0) 1716 return ((struct mbuf *)0); 1717 m = m_get(M_DONTWAIT, MT_SOOPTS); 1718 if (m == 0) 1719 return ((struct mbuf *)0); 1720 1721 MCLAIM(m, &inetdomain.dom_mowner); 1722#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1723 1724 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1725 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1726 OPTSIZ; 1727#ifdef DIAGNOSTIC 1728 if (ipprintfs) 1729 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1730#endif 1731 1732 /* 1733 * First save first hop for return route 1734 */ 1735 p = &ip_srcrt.route[ip_nhops - 1]; 1736 *(mtod(m, struct in_addr *)) = *p--; 1737#ifdef DIAGNOSTIC 1738 if (ipprintfs) 1739 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1740#endif 1741 1742 /* 1743 * Copy option fields and padding (nop) to mbuf. 1744 */ 1745 ip_srcrt.nop = IPOPT_NOP; 1746 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1747 bcopy((caddr_t)&ip_srcrt.nop, 1748 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1749 q = (struct in_addr *)(mtod(m, caddr_t) + 1750 sizeof(struct in_addr) + OPTSIZ); 1751#undef OPTSIZ 1752 /* 1753 * Record return path as an IP source route, 1754 * reversing the path (pointers are now aligned). 1755 */ 1756 while (p >= ip_srcrt.route) { 1757#ifdef DIAGNOSTIC 1758 if (ipprintfs) 1759 printf(" %x", ntohl(q->s_addr)); 1760#endif 1761 *q++ = *p--; 1762 } 1763 /* 1764 * Last hop goes to final destination. 1765 */ 1766 *q = ip_srcrt.dst; 1767#ifdef DIAGNOSTIC 1768 if (ipprintfs) 1769 printf(" %x\n", ntohl(q->s_addr)); 1770#endif 1771 return (m); 1772} 1773 1774/* 1775 * Strip out IP options, at higher 1776 * level protocol in the kernel. 1777 * Second argument is buffer to which options 1778 * will be moved, and return value is their length. 1779 * XXX should be deleted; last arg currently ignored. 1780 */ 1781void 1782ip_stripoptions(m, mopt) 1783 struct mbuf *m; 1784 struct mbuf *mopt; 1785{ 1786 int i; 1787 struct ip *ip = mtod(m, struct ip *); 1788 caddr_t opts; 1789 int olen; 1790 1791 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1792 opts = (caddr_t)(ip + 1); 1793 i = m->m_len - (sizeof (struct ip) + olen); 1794 bcopy(opts + olen, opts, (unsigned)i); 1795 m->m_len -= olen; 1796 if (m->m_flags & M_PKTHDR) 1797 m->m_pkthdr.len -= olen; 1798 ip->ip_len = htons(ntohs(ip->ip_len) - olen); 1799 ip->ip_hl = sizeof (struct ip) >> 2; 1800} 1801 1802const int inetctlerrmap[PRC_NCMDS] = { 1803 0, 0, 0, 0, 1804 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1805 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1806 EMSGSIZE, EHOSTUNREACH, 0, 0, 1807 0, 0, 0, 0, 1808 ENOPROTOOPT 1809}; 1810 1811/* 1812 * Forward a packet. If some error occurs return the sender 1813 * an icmp packet. Note we can't always generate a meaningful 1814 * icmp message because icmp doesn't have a large enough repertoire 1815 * of codes and types. 1816 * 1817 * If not forwarding, just drop the packet. This could be confusing 1818 * if ipforwarding was zero but some routing protocol was advancing 1819 * us as a gateway to somewhere. However, we must let the routing 1820 * protocol deal with that. 1821 * 1822 * The srcrt parameter indicates whether the packet is being forwarded 1823 * via a source route. 1824 */ 1825void 1826ip_forward(m, srcrt) 1827 struct mbuf *m; 1828 int srcrt; 1829{ 1830 struct ip *ip = mtod(m, struct ip *); 1831 struct sockaddr_in *sin; 1832 struct rtentry *rt; 1833 int error, type = 0, code = 0; 1834 struct mbuf *mcopy; 1835 n_long dest; 1836 struct ifnet *destifp; 1837#if defined(IPSEC) || defined(FAST_IPSEC) 1838 struct ifnet dummyifp; 1839#endif 1840 1841 /* 1842 * We are now in the output path. 1843 */ 1844 MCLAIM(m, &ip_tx_mowner); 1845 1846 /* 1847 * Clear any in-bound checksum flags for this packet. 1848 */ 1849 m->m_pkthdr.csum_flags = 0; 1850 1851 dest = 0; 1852#ifdef DIAGNOSTIC 1853 if (ipprintfs) 1854 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1855 ntohl(ip->ip_src.s_addr), 1856 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1857#endif 1858 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1859 ipstat.ips_cantforward++; 1860 m_freem(m); 1861 return; 1862 } 1863 if (ip->ip_ttl <= IPTTLDEC) { 1864 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1865 return; 1866 } 1867 ip->ip_ttl -= IPTTLDEC; 1868 1869 sin = satosin(&ipforward_rt.ro_dst); 1870 if ((rt = ipforward_rt.ro_rt) == 0 || 1871 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1872 if (ipforward_rt.ro_rt) { 1873 RTFREE(ipforward_rt.ro_rt); 1874 ipforward_rt.ro_rt = 0; 1875 } 1876 sin->sin_family = AF_INET; 1877 sin->sin_len = sizeof(struct sockaddr_in); 1878 sin->sin_addr = ip->ip_dst; 1879 1880 rtalloc(&ipforward_rt); 1881 if (ipforward_rt.ro_rt == 0) { 1882 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1883 return; 1884 } 1885 rt = ipforward_rt.ro_rt; 1886 } 1887 1888 /* 1889 * Save at most 68 bytes of the packet in case 1890 * we need to generate an ICMP message to the src. 1891 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1892 */ 1893 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1894 if (mcopy) 1895 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1896 1897 /* 1898 * If forwarding packet using same interface that it came in on, 1899 * perhaps should send a redirect to sender to shortcut a hop. 1900 * Only send redirect if source is sending directly to us, 1901 * and if packet was not source routed (or has any options). 1902 * Also, don't send redirect if forwarding using a default route 1903 * or a route modified by a redirect. 1904 */ 1905 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1906 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1907 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1908 ipsendredirects && !srcrt) { 1909 if (rt->rt_ifa && 1910 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1911 ifatoia(rt->rt_ifa)->ia_subnet) { 1912 if (rt->rt_flags & RTF_GATEWAY) 1913 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1914 else 1915 dest = ip->ip_dst.s_addr; 1916 /* 1917 * Router requirements says to only send host 1918 * redirects. 1919 */ 1920 type = ICMP_REDIRECT; 1921 code = ICMP_REDIRECT_HOST; 1922#ifdef DIAGNOSTIC 1923 if (ipprintfs) 1924 printf("redirect (%d) to %x\n", code, 1925 (u_int32_t)dest); 1926#endif 1927 } 1928 } 1929 1930 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1931 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1932 (struct ip_moptions *)NULL, (struct socket *)NULL); 1933 1934 if (error) 1935 ipstat.ips_cantforward++; 1936 else { 1937 ipstat.ips_forward++; 1938 if (type) 1939 ipstat.ips_redirectsent++; 1940 else { 1941 if (mcopy) { 1942#ifdef GATEWAY 1943 if (mcopy->m_flags & M_CANFASTFWD) 1944 ipflow_create(&ipforward_rt, mcopy); 1945#endif 1946 m_freem(mcopy); 1947 } 1948 return; 1949 } 1950 } 1951 if (mcopy == NULL) 1952 return; 1953 destifp = NULL; 1954 1955 switch (error) { 1956 1957 case 0: /* forwarded, but need redirect */ 1958 /* type, code set above */ 1959 break; 1960 1961 case ENETUNREACH: /* shouldn't happen, checked above */ 1962 case EHOSTUNREACH: 1963 case ENETDOWN: 1964 case EHOSTDOWN: 1965 default: 1966 type = ICMP_UNREACH; 1967 code = ICMP_UNREACH_HOST; 1968 break; 1969 1970 case EMSGSIZE: 1971 type = ICMP_UNREACH; 1972 code = ICMP_UNREACH_NEEDFRAG; 1973#if !defined(IPSEC) && !defined(FAST_IPSEC) 1974 if (ipforward_rt.ro_rt) 1975 destifp = ipforward_rt.ro_rt->rt_ifp; 1976#else 1977 /* 1978 * If the packet is routed over IPsec tunnel, tell the 1979 * originator the tunnel MTU. 1980 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1981 * XXX quickhack!!! 1982 */ 1983 if (ipforward_rt.ro_rt) { 1984 struct secpolicy *sp; 1985 int ipsecerror; 1986 size_t ipsechdr; 1987 struct route *ro; 1988 1989 sp = ipsec4_getpolicybyaddr(mcopy, 1990 IPSEC_DIR_OUTBOUND, IP_FORWARDING, 1991 &ipsecerror); 1992 1993 if (sp == NULL) 1994 destifp = ipforward_rt.ro_rt->rt_ifp; 1995 else { 1996 /* count IPsec header size */ 1997 ipsechdr = ipsec4_hdrsiz(mcopy, 1998 IPSEC_DIR_OUTBOUND, NULL); 1999 2000 /* 2001 * find the correct route for outer IPv4 2002 * header, compute tunnel MTU. 2003 * 2004 * XXX BUG ALERT 2005 * The "dummyifp" code relies upon the fact 2006 * that icmp_error() touches only ifp->if_mtu. 2007 */ 2008 /*XXX*/ 2009 destifp = NULL; 2010 if (sp->req != NULL 2011 && sp->req->sav != NULL 2012 && sp->req->sav->sah != NULL) { 2013 ro = &sp->req->sav->sah->sa_route; 2014 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 2015 dummyifp.if_mtu = 2016 ro->ro_rt->rt_rmx.rmx_mtu ? 2017 ro->ro_rt->rt_rmx.rmx_mtu : 2018 ro->ro_rt->rt_ifp->if_mtu; 2019 dummyifp.if_mtu -= ipsechdr; 2020 destifp = &dummyifp; 2021 } 2022 } 2023 2024#ifdef IPSEC 2025 key_freesp(sp); 2026#else 2027 KEY_FREESP(&sp); 2028#endif 2029 } 2030 } 2031#endif /*IPSEC*/ 2032 ipstat.ips_cantfrag++; 2033 break; 2034 2035 case ENOBUFS: 2036#if 1 2037 /* 2038 * a router should not generate ICMP_SOURCEQUENCH as 2039 * required in RFC1812 Requirements for IP Version 4 Routers. 2040 * source quench could be a big problem under DoS attacks, 2041 * or if the underlying interface is rate-limited. 2042 */ 2043 if (mcopy) 2044 m_freem(mcopy); 2045 return; 2046#else 2047 type = ICMP_SOURCEQUENCH; 2048 code = 0; 2049 break; 2050#endif 2051 } 2052 icmp_error(mcopy, type, code, dest, destifp); 2053} 2054 2055void 2056ip_savecontrol(inp, mp, ip, m) 2057 struct inpcb *inp; 2058 struct mbuf **mp; 2059 struct ip *ip; 2060 struct mbuf *m; 2061{ 2062 2063 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2064 struct timeval tv; 2065 2066 microtime(&tv); 2067 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2068 SCM_TIMESTAMP, SOL_SOCKET); 2069 if (*mp) 2070 mp = &(*mp)->m_next; 2071 } 2072 if (inp->inp_flags & INP_RECVDSTADDR) { 2073 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2074 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2075 if (*mp) 2076 mp = &(*mp)->m_next; 2077 } 2078#ifdef notyet 2079 /* 2080 * XXX 2081 * Moving these out of udp_input() made them even more broken 2082 * than they already were. 2083 * - fenner@parc.xerox.com 2084 */ 2085 /* options were tossed already */ 2086 if (inp->inp_flags & INP_RECVOPTS) { 2087 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2088 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2089 if (*mp) 2090 mp = &(*mp)->m_next; 2091 } 2092 /* ip_srcroute doesn't do what we want here, need to fix */ 2093 if (inp->inp_flags & INP_RECVRETOPTS) { 2094 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 2095 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2096 if (*mp) 2097 mp = &(*mp)->m_next; 2098 } 2099#endif 2100 if (inp->inp_flags & INP_RECVIF) { 2101 struct sockaddr_dl sdl; 2102 2103 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 2104 sdl.sdl_family = AF_LINK; 2105 sdl.sdl_index = m->m_pkthdr.rcvif ? 2106 m->m_pkthdr.rcvif->if_index : 0; 2107 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 2108 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 2109 IP_RECVIF, IPPROTO_IP); 2110 if (*mp) 2111 mp = &(*mp)->m_next; 2112 } 2113} 2114 2115/* 2116 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 2117 * range of the new value and tweaks timers if it changes. 2118 */ 2119static int 2120sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 2121{ 2122 int error, tmp; 2123 struct sysctlnode node; 2124 2125 node = *rnode; 2126 tmp = ip_mtudisc_timeout; 2127 node.sysctl_data = &tmp; 2128 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2129 if (error || newp == NULL) 2130 return (error); 2131 if (tmp < 0) 2132 return (EINVAL); 2133 2134 ip_mtudisc_timeout = tmp; 2135 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 2136 2137 return (0); 2138} 2139 2140#ifdef GATEWAY 2141/* 2142 * sysctl helper routine for net.inet.ip.maxflows. apparently if 2143 * maxflows is even looked up, we "reap flows". 2144 */ 2145static int 2146sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS) 2147{ 2148 int s; 2149 2150 s = sysctl_lookup(SYSCTLFN_CALL(rnode)); 2151 if (s) 2152 return (s); 2153 2154 s = splsoftnet(); 2155 ipflow_reap(0); 2156 splx(s); 2157 2158 return (0); 2159} 2160#endif /* GATEWAY */ 2161 2162 2163SYSCTL_SETUP(sysctl_net_inet_ip_setup, "sysctl net.inet.ip subtree setup") 2164{ 2165 extern int subnetsarelocal, hostzeroisbroadcast; 2166 2167 sysctl_createv(clog, 0, NULL, NULL, 2168 CTLFLAG_PERMANENT, 2169 CTLTYPE_NODE, "net", NULL, 2170 NULL, 0, NULL, 0, 2171 CTL_NET, CTL_EOL); 2172 sysctl_createv(clog, 0, NULL, NULL, 2173 CTLFLAG_PERMANENT, 2174 CTLTYPE_NODE, "inet", NULL, 2175 NULL, 0, NULL, 0, 2176 CTL_NET, PF_INET, CTL_EOL); 2177 sysctl_createv(clog, 0, NULL, NULL, 2178 CTLFLAG_PERMANENT, 2179 CTLTYPE_NODE, "ip", NULL, 2180 NULL, 0, NULL, 0, 2181 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 2182 2183 sysctl_createv(clog, 0, NULL, NULL, 2184 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2185 CTLTYPE_INT, "forwarding", NULL, 2186 NULL, 0, &ipforwarding, 0, 2187 CTL_NET, PF_INET, IPPROTO_IP, 2188 IPCTL_FORWARDING, CTL_EOL); 2189 sysctl_createv(clog, 0, NULL, NULL, 2190 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2191 CTLTYPE_INT, "redirect", NULL, 2192 NULL, 0, &ipsendredirects, 0, 2193 CTL_NET, PF_INET, IPPROTO_IP, 2194 IPCTL_SENDREDIRECTS, CTL_EOL); 2195 sysctl_createv(clog, 0, NULL, NULL, 2196 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2197 CTLTYPE_INT, "ttl", NULL, 2198 NULL, 0, &ip_defttl, 0, 2199 CTL_NET, PF_INET, IPPROTO_IP, 2200 IPCTL_DEFTTL, CTL_EOL); 2201#ifdef IPCTL_DEFMTU 2202 sysctl_createv(clog, 0, NULL, NULL, 2203 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 2204 CTLTYPE_INT, "mtu", NULL, 2205 NULL, 0, &ip_mtu, 0, 2206 CTL_NET, PF_INET, IPPROTO_IP, 2207 IPCTL_DEFMTU, CTL_EOL); 2208#endif /* IPCTL_DEFMTU */ 2209 sysctl_createv(clog, 0, NULL, NULL, 2210 CTLFLAG_PERMANENT|CTLFLAG_READONLY1, 2211 CTLTYPE_INT, "forwsrcrt", NULL, 2212 NULL, 0, &ip_forwsrcrt, 0, 2213 CTL_NET, PF_INET, IPPROTO_IP, 2214 IPCTL_FORWSRCRT, CTL_EOL); 2215 sysctl_createv(clog, 0, NULL, NULL, 2216 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2217 CTLTYPE_INT, "directed-broadcast", NULL, 2218 NULL, 0, &ip_directedbcast, 0, 2219 CTL_NET, PF_INET, IPPROTO_IP, 2220 IPCTL_DIRECTEDBCAST, CTL_EOL); 2221 sysctl_createv(clog, 0, NULL, NULL, 2222 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2223 CTLTYPE_INT, "allowsrcrt", NULL, 2224 NULL, 0, &ip_allowsrcrt, 0, 2225 CTL_NET, PF_INET, IPPROTO_IP, 2226 IPCTL_ALLOWSRCRT, CTL_EOL); 2227 sysctl_createv(clog, 0, NULL, NULL, 2228 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2229 CTLTYPE_INT, "subnetsarelocal", NULL, 2230 NULL, 0, &subnetsarelocal, 0, 2231 CTL_NET, PF_INET, IPPROTO_IP, 2232 IPCTL_SUBNETSARELOCAL, CTL_EOL); 2233 sysctl_createv(clog, 0, NULL, NULL, 2234 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2235 CTLTYPE_INT, "mtudisc", NULL, 2236 NULL, 0, &ip_mtudisc, 0, 2237 CTL_NET, PF_INET, IPPROTO_IP, 2238 IPCTL_MTUDISC, CTL_EOL); 2239 sysctl_createv(clog, 0, NULL, NULL, 2240 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2241 CTLTYPE_INT, "anonportmin", NULL, 2242 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 2243 CTL_NET, PF_INET, IPPROTO_IP, 2244 IPCTL_ANONPORTMIN, CTL_EOL); 2245 sysctl_createv(clog, 0, NULL, NULL, 2246 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2247 CTLTYPE_INT, "anonportmax", NULL, 2248 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 2249 CTL_NET, PF_INET, IPPROTO_IP, 2250 IPCTL_ANONPORTMAX, CTL_EOL); 2251 sysctl_createv(clog, 0, NULL, NULL, 2252 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2253 CTLTYPE_INT, "mtudisctimeout", NULL, 2254 sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0, 2255 CTL_NET, PF_INET, IPPROTO_IP, 2256 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 2257#ifdef GATEWAY 2258 sysctl_createv(clog, 0, NULL, NULL, 2259 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2260 CTLTYPE_INT, "maxflows", NULL, 2261 sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0, 2262 CTL_NET, PF_INET, IPPROTO_IP, 2263 IPCTL_MAXFLOWS, CTL_EOL); 2264#endif /* GATEWAY */ 2265 sysctl_createv(clog, 0, NULL, NULL, 2266 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2267 CTLTYPE_INT, "hostzerobroadcast", NULL, 2268 NULL, 0, &hostzeroisbroadcast, 0, 2269 CTL_NET, PF_INET, IPPROTO_IP, 2270 IPCTL_HOSTZEROBROADCAST, CTL_EOL); 2271#if NGIF > 0 2272 sysctl_createv(clog, 0, NULL, NULL, 2273 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2274 CTLTYPE_INT, "gifttl", NULL, 2275 NULL, 0, &ip_gif_ttl, 0, 2276 CTL_NET, PF_INET, IPPROTO_IP, 2277 IPCTL_GIF_TTL, CTL_EOL); 2278#endif /* NGIF */ 2279#ifndef IPNOPRIVPORTS 2280 sysctl_createv(clog, 0, NULL, NULL, 2281 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2282 CTLTYPE_INT, "lowportmin", NULL, 2283 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 2284 CTL_NET, PF_INET, IPPROTO_IP, 2285 IPCTL_LOWPORTMIN, CTL_EOL); 2286 sysctl_createv(clog, 0, NULL, NULL, 2287 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2288 CTLTYPE_INT, "lowportmax", NULL, 2289 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 2290 CTL_NET, PF_INET, IPPROTO_IP, 2291 IPCTL_LOWPORTMAX, CTL_EOL); 2292#endif /* IPNOPRIVPORTS */ 2293 sysctl_createv(clog, 0, NULL, NULL, 2294 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2295 CTLTYPE_INT, "maxfragpackets", NULL, 2296 NULL, 0, &ip_maxfragpackets, 0, 2297 CTL_NET, PF_INET, IPPROTO_IP, 2298 IPCTL_MAXFRAGPACKETS, CTL_EOL); 2299#if NGRE > 0 2300 sysctl_createv(clog, 0, NULL, NULL, 2301 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2302 CTLTYPE_INT, "grettl", NULL, 2303 NULL, 0, &ip_gre_ttl, 0, 2304 CTL_NET, PF_INET, IPPROTO_IP, 2305 IPCTL_GRE_TTL, CTL_EOL); 2306#endif /* NGRE */ 2307 sysctl_createv(clog, 0, NULL, NULL, 2308 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2309 CTLTYPE_INT, "checkinterface", NULL, 2310 NULL, 0, &ip_checkinterface, 0, 2311 CTL_NET, PF_INET, IPPROTO_IP, 2312 IPCTL_CHECKINTERFACE, CTL_EOL); 2313 sysctl_createv(clog, 0, NULL, NULL, 2314 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2315 CTLTYPE_INT, "random_id", NULL, 2316 NULL, 0, &ip_do_randomid, 0, 2317 CTL_NET, PF_INET, IPPROTO_IP, 2318 IPCTL_RANDOMID, CTL_EOL); 2319} 2320