ip_input.c revision 1.202
1/* $NetBSD: ip_input.c,v 1.202 2004/05/02 05:02:53 darrenr 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 1.202 2004/05/02 05:02:53 darrenr 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 350EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 351EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 352EVCNT_ATTACH_STATIC(ip_swcsum); 353 354#else 355 356#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 357 358#endif /* INET_CSUM_COUNTERS */ 359 360/* 361 * We need to save the IP options in case a protocol wants to respond 362 * to an incoming packet over the same route if the packet got here 363 * using IP source routing. This allows connection establishment and 364 * maintenance when the remote end is on a network that is not known 365 * to us. 366 */ 367int ip_nhops = 0; 368static struct ip_srcrt { 369 struct in_addr dst; /* final destination */ 370 char nop; /* one NOP to align */ 371 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 372 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 373} ip_srcrt; 374 375static void save_rte __P((u_char *, struct in_addr)); 376 377#ifdef MBUFTRACE 378struct mowner ip_rx_mowner = { "internet", "rx" }; 379struct mowner ip_tx_mowner = { "internet", "tx" }; 380#endif 381 382/* 383 * Compute IP limits derived from the value of nmbclusters. 384 */ 385static void 386ip_nmbclusters_changed(void) 387{ 388 ip_maxfrags = nmbclusters / 4; 389 ip_nmbclusters = nmbclusters; 390} 391 392/* 393 * IP initialization: fill in IP protocol switch table. 394 * All protocols not implemented in kernel go to raw IP protocol handler. 395 */ 396void 397ip_init() 398{ 399 const struct protosw *pr; 400 int i; 401 402 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 403 if (pr == 0) 404 panic("ip_init"); 405 for (i = 0; i < IPPROTO_MAX; i++) 406 ip_protox[i] = pr - inetsw; 407 for (pr = inetdomain.dom_protosw; 408 pr < inetdomain.dom_protoswNPROTOSW; pr++) 409 if (pr->pr_domain->dom_family == PF_INET && 410 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 411 ip_protox[pr->pr_protocol] = pr - inetsw; 412 413 for (i = 0; i < IPREASS_NHASH; i++) 414 LIST_INIT(&ipq[i]); 415 416 ip_id = time.tv_sec & 0xfffff; 417 418 ipintrq.ifq_maxlen = ipqmaxlen; 419 ip_nmbclusters_changed(); 420 421 TAILQ_INIT(&in_ifaddrhead); 422 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 423 M_WAITOK, &in_ifaddrhash); 424 in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IPMADDR, 425 M_WAITOK, &in_multihash); 426 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 427#ifdef GATEWAY 428 ipflow_init(); 429#endif 430 431#ifdef PFIL_HOOKS 432 /* Register our Packet Filter hook. */ 433 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 434 inet_pfil_hook.ph_af = AF_INET; 435 i = pfil_head_register(&inet_pfil_hook); 436 if (i != 0) 437 printf("ip_init: WARNING: unable to register pfil hook, " 438 "error %d\n", i); 439#endif /* PFIL_HOOKS */ 440 441#ifdef MBUFTRACE 442 MOWNER_ATTACH(&ip_tx_mowner); 443 MOWNER_ATTACH(&ip_rx_mowner); 444#endif /* MBUFTRACE */ 445} 446 447struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 448struct route ipforward_rt; 449 450/* 451 * IP software interrupt routine 452 */ 453void 454ipintr() 455{ 456 int s; 457 struct mbuf *m; 458 459 while (1) { 460 s = splnet(); 461 IF_DEQUEUE(&ipintrq, m); 462 splx(s); 463 if (m == 0) 464 return; 465 MCLAIM(m, &ip_rx_mowner); 466 ip_input(m); 467 } 468} 469 470/* 471 * Ip input routine. Checksum and byte swap header. If fragmented 472 * try to reassemble. Process options. Pass to next level. 473 */ 474void 475ip_input(struct mbuf *m) 476{ 477 struct ip *ip = NULL; 478 struct ipq *fp; 479 struct in_ifaddr *ia; 480 struct ifaddr *ifa; 481 struct ipqent *ipqe; 482 int hlen = 0, mff, len; 483 int downmatch; 484 int checkif; 485 int srcrt = 0; 486 u_int hash; 487#ifdef FAST_IPSEC 488 struct m_tag *mtag; 489 struct tdb_ident *tdbi; 490 struct secpolicy *sp; 491 int s, error; 492#endif /* FAST_IPSEC */ 493 494 MCLAIM(m, &ip_rx_mowner); 495#ifdef DIAGNOSTIC 496 if ((m->m_flags & M_PKTHDR) == 0) 497 panic("ipintr no HDR"); 498#endif 499 500 /* 501 * If no IP addresses have been set yet but the interfaces 502 * are receiving, can't do anything with incoming packets yet. 503 */ 504 if (TAILQ_FIRST(&in_ifaddrhead) == 0) 505 goto bad; 506 ipstat.ips_total++; 507 /* 508 * If the IP header is not aligned, slurp it up into a new 509 * mbuf with space for link headers, in the event we forward 510 * it. Otherwise, if it is aligned, make sure the entire 511 * base IP header is in the first mbuf of the chain. 512 */ 513 if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) { 514 if ((m = m_copyup(m, sizeof(struct ip), 515 (max_linkhdr + 3) & ~3)) == NULL) { 516 /* XXXJRT new stat, please */ 517 ipstat.ips_toosmall++; 518 return; 519 } 520 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 521 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 522 ipstat.ips_toosmall++; 523 return; 524 } 525 } 526 ip = mtod(m, struct ip *); 527 if (ip->ip_v != IPVERSION) { 528 ipstat.ips_badvers++; 529 goto bad; 530 } 531 hlen = ip->ip_hl << 2; 532 if (hlen < sizeof(struct ip)) { /* minimum header length */ 533 ipstat.ips_badhlen++; 534 goto bad; 535 } 536 if (hlen > m->m_len) { 537 if ((m = m_pullup(m, hlen)) == 0) { 538 ipstat.ips_badhlen++; 539 return; 540 } 541 ip = mtod(m, struct ip *); 542 } 543 544 /* 545 * RFC1122: packets with a multicast source address are 546 * not allowed. 547 */ 548 if (IN_MULTICAST(ip->ip_src.s_addr)) { 549 ipstat.ips_badaddr++; 550 goto bad; 551 } 552 553 /* 127/8 must not appear on wire - RFC1122 */ 554 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 555 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 556 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 557 ipstat.ips_badaddr++; 558 goto bad; 559 } 560 } 561 562 switch (m->m_pkthdr.csum_flags & 563 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) | 564 M_CSUM_IPv4_BAD)) { 565 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 566 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 567 goto badcsum; 568 569 case M_CSUM_IPv4: 570 /* Checksum was okay. */ 571 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 572 break; 573 574 default: 575 /* Must compute it ourselves. */ 576 INET_CSUM_COUNTER_INCR(&ip_swcsum); 577 if (in_cksum(m, hlen) != 0) 578 goto bad; 579 break; 580 } 581 582 /* Retrieve the packet length. */ 583 len = ntohs(ip->ip_len); 584 585 /* 586 * Check for additional length bogosity 587 */ 588 if (len < hlen) { 589 ipstat.ips_badlen++; 590 goto bad; 591 } 592 593 /* 594 * Check that the amount of data in the buffers 595 * is as at least much as the IP header would have us expect. 596 * Trim mbufs if longer than we expect. 597 * Drop packet if shorter than we expect. 598 */ 599 if (m->m_pkthdr.len < len) { 600 ipstat.ips_tooshort++; 601 goto bad; 602 } 603 if (m->m_pkthdr.len > len) { 604 if (m->m_len == m->m_pkthdr.len) { 605 m->m_len = len; 606 m->m_pkthdr.len = len; 607 } else 608 m_adj(m, len - m->m_pkthdr.len); 609 } 610 611#if defined(IPSEC) 612 /* ipflow (IP fast forwarding) is not compatible with IPsec. */ 613 m->m_flags &= ~M_CANFASTFWD; 614#else 615 /* 616 * Assume that we can create a fast-forward IP flow entry 617 * based on this packet. 618 */ 619 m->m_flags |= M_CANFASTFWD; 620#endif 621 622#ifdef PFIL_HOOKS 623 /* 624 * Run through list of hooks for input packets. If there are any 625 * filters which require that additional packets in the flow are 626 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 627 * Note that filters must _never_ set this flag, as another filter 628 * in the list may have previously cleared it. 629 */ 630 /* 631 * let ipfilter look at packet on the wire, 632 * not the decapsulated packet. 633 */ 634#ifdef IPSEC 635 if (!ipsec_getnhist(m)) 636#elif defined(FAST_IPSEC) 637 if (!ipsec_indone(m)) 638#else 639 if (1) 640#endif 641 { 642 struct in_addr odst; 643 644 odst = ip->ip_dst; 645 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 646 PFIL_IN) != 0) 647 return; 648 if (m == NULL) 649 return; 650 ip = mtod(m, struct ip *); 651 hlen = ip->ip_hl << 2; 652 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 653 } 654#endif /* PFIL_HOOKS */ 655 656#ifdef ALTQ 657 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 658 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 659 /* packet dropped by traffic conditioner */ 660 return; 661 } 662#endif 663 664 /* 665 * Process options and, if not destined for us, 666 * ship it on. ip_dooptions returns 1 when an 667 * error was detected (causing an icmp message 668 * to be sent and the original packet to be freed). 669 */ 670 ip_nhops = 0; /* for source routed packets */ 671 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 672 return; 673 674 /* 675 * Enable a consistency check between the destination address 676 * and the arrival interface for a unicast packet (the RFC 1122 677 * strong ES model) if IP forwarding is disabled and the packet 678 * is not locally generated. 679 * 680 * XXX - Checking also should be disabled if the destination 681 * address is ipnat'ed to a different interface. 682 * 683 * XXX - Checking is incompatible with IP aliases added 684 * to the loopback interface instead of the interface where 685 * the packets are received. 686 * 687 * XXX - We need to add a per ifaddr flag for this so that 688 * we get finer grain control. 689 */ 690 checkif = ip_checkinterface && (ipforwarding == 0) && 691 (m->m_pkthdr.rcvif != NULL) && 692 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0); 693 694 /* 695 * Check our list of addresses, to see if the packet is for us. 696 * 697 * Traditional 4.4BSD did not consult IFF_UP at all. 698 * The behavior here is to treat addresses on !IFF_UP interface 699 * as not mine. 700 */ 701 downmatch = 0; 702 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 703 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 704 if (checkif && ia->ia_ifp != m->m_pkthdr.rcvif) 705 continue; 706 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 707 break; 708 else 709 downmatch++; 710 } 711 } 712 if (ia != NULL) 713 goto ours; 714 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 715 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrlist, ifa_list) { 716 if (ifa->ifa_addr->sa_family != AF_INET) 717 continue; 718 ia = ifatoia(ifa); 719 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 720 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 721 /* 722 * Look for all-0's host part (old broadcast addr), 723 * either for subnet or net. 724 */ 725 ip->ip_dst.s_addr == ia->ia_subnet || 726 ip->ip_dst.s_addr == ia->ia_net) 727 goto ours; 728 /* 729 * An interface with IP address zero accepts 730 * all packets that arrive on that interface. 731 */ 732 if (in_nullhost(ia->ia_addr.sin_addr)) 733 goto ours; 734 } 735 } 736 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 737 struct in_multi *inm; 738#ifdef MROUTING 739 extern struct socket *ip_mrouter; 740 741 if (ip_mrouter) { 742 /* 743 * If we are acting as a multicast router, all 744 * incoming multicast packets are passed to the 745 * kernel-level multicast forwarding function. 746 * The packet is returned (relatively) intact; if 747 * ip_mforward() returns a non-zero value, the packet 748 * must be discarded, else it may be accepted below. 749 * 750 * (The IP ident field is put in the same byte order 751 * as expected when ip_mforward() is called from 752 * ip_output().) 753 */ 754 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 755 ipstat.ips_cantforward++; 756 m_freem(m); 757 return; 758 } 759 760 /* 761 * The process-level routing demon needs to receive 762 * all multicast IGMP packets, whether or not this 763 * host belongs to their destination groups. 764 */ 765 if (ip->ip_p == IPPROTO_IGMP) 766 goto ours; 767 ipstat.ips_forward++; 768 } 769#endif 770 /* 771 * See if we belong to the destination multicast group on the 772 * arrival interface. 773 */ 774 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 775 if (inm == NULL) { 776 ipstat.ips_cantforward++; 777 m_freem(m); 778 return; 779 } 780 goto ours; 781 } 782 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 783 in_nullhost(ip->ip_dst)) 784 goto ours; 785 786 /* 787 * Not for us; forward if possible and desirable. 788 */ 789 if (ipforwarding == 0) { 790 ipstat.ips_cantforward++; 791 m_freem(m); 792 } else { 793 /* 794 * If ip_dst matched any of my address on !IFF_UP interface, 795 * and there's no IFF_UP interface that matches ip_dst, 796 * send icmp unreach. Forwarding it will result in in-kernel 797 * forwarding loop till TTL goes to 0. 798 */ 799 if (downmatch) { 800 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 801 ipstat.ips_cantforward++; 802 return; 803 } 804#ifdef IPSEC 805 if (ipsec4_in_reject(m, NULL)) { 806 ipsecstat.in_polvio++; 807 goto bad; 808 } 809#endif 810#ifdef FAST_IPSEC 811 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 812 s = splsoftnet(); 813 if (mtag != NULL) { 814 tdbi = (struct tdb_ident *)(mtag + 1); 815 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 816 } else { 817 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 818 IP_FORWARDING, &error); 819 } 820 if (sp == NULL) { /* NB: can happen if error */ 821 splx(s); 822 /*XXX error stat???*/ 823 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 824 goto bad; 825 } 826 827 /* 828 * Check security policy against packet attributes. 829 */ 830 error = ipsec_in_reject(sp, m); 831 KEY_FREESP(&sp); 832 splx(s); 833 if (error) { 834 ipstat.ips_cantforward++; 835 goto bad; 836 } 837 838 /* 839 * Peek at the outbound SP for this packet to determine if 840 * it's a Fast Forward candidate. 841 */ 842 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 843 if (mtag != NULL) 844 m->m_flags &= ~M_CANFASTFWD; 845 else { 846 s = splsoftnet(); 847 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, 848 (IP_FORWARDING | 849 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 850 &error, NULL); 851 if (sp != NULL) { 852 m->m_flags &= ~M_CANFASTFWD; 853 KEY_FREESP(&sp); 854 } 855 splx(s); 856 } 857#endif /* FAST_IPSEC */ 858 859 ip_forward(m, srcrt); 860 } 861 return; 862 863ours: 864 /* 865 * If offset or IP_MF are set, must reassemble. 866 * Otherwise, nothing need be done. 867 * (We could look in the reassembly queue to see 868 * if the packet was previously fragmented, 869 * but it's not worth the time; just let them time out.) 870 */ 871 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 872 873 /* 874 * Look for queue of fragments 875 * of this datagram. 876 */ 877 IPQ_LOCK(); 878 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 879 /* XXX LIST_FOREACH(fp, &ipq[hash], ipq_q) */ 880 for (fp = LIST_FIRST(&ipq[hash]); fp != NULL; 881 fp = LIST_NEXT(fp, ipq_q)) { 882 if (ip->ip_id == fp->ipq_id && 883 in_hosteq(ip->ip_src, fp->ipq_src) && 884 in_hosteq(ip->ip_dst, fp->ipq_dst) && 885 ip->ip_p == fp->ipq_p) 886 goto found; 887 888 } 889 fp = 0; 890found: 891 892 /* 893 * Adjust ip_len to not reflect header, 894 * set ipqe_mff if more fragments are expected, 895 * convert offset of this to bytes. 896 */ 897 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 898 mff = (ip->ip_off & htons(IP_MF)) != 0; 899 if (mff) { 900 /* 901 * Make sure that fragments have a data length 902 * that's a non-zero multiple of 8 bytes. 903 */ 904 if (ntohs(ip->ip_len) == 0 || 905 (ntohs(ip->ip_len) & 0x7) != 0) { 906 ipstat.ips_badfrags++; 907 IPQ_UNLOCK(); 908 goto bad; 909 } 910 } 911 ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3); 912 913 /* 914 * If datagram marked as having more fragments 915 * or if this is not the first fragment, 916 * attempt reassembly; if it succeeds, proceed. 917 */ 918 if (mff || ip->ip_off != htons(0)) { 919 ipstat.ips_fragments++; 920 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 921 if (ipqe == NULL) { 922 ipstat.ips_rcvmemdrop++; 923 IPQ_UNLOCK(); 924 goto bad; 925 } 926 ipqe->ipqe_mff = mff; 927 ipqe->ipqe_m = m; 928 ipqe->ipqe_ip = ip; 929 m = ip_reass(ipqe, fp, &ipq[hash]); 930 if (m == 0) { 931 IPQ_UNLOCK(); 932 return; 933 } 934 ipstat.ips_reassembled++; 935 ip = mtod(m, struct ip *); 936 hlen = ip->ip_hl << 2; 937 ip->ip_len = htons(ntohs(ip->ip_len) + hlen); 938 } else 939 if (fp) 940 ip_freef(fp); 941 IPQ_UNLOCK(); 942 } 943 944#if defined(IPSEC) 945 /* 946 * enforce IPsec policy checking if we are seeing last header. 947 * note that we do not visit this with protocols with pcb layer 948 * code - like udp/tcp/raw ip. 949 */ 950 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 951 ipsec4_in_reject(m, NULL)) { 952 ipsecstat.in_polvio++; 953 goto bad; 954 } 955#endif 956#if FAST_IPSEC 957 /* 958 * enforce IPsec policy checking if we are seeing last header. 959 * note that we do not visit this with protocols with pcb layer 960 * code - like udp/tcp/raw ip. 961 */ 962 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 963 /* 964 * Check if the packet has already had IPsec processing 965 * done. If so, then just pass it along. This tag gets 966 * set during AH, ESP, etc. input handling, before the 967 * packet is returned to the ip input queue for delivery. 968 */ 969 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 970 s = splsoftnet(); 971 if (mtag != NULL) { 972 tdbi = (struct tdb_ident *)(mtag + 1); 973 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 974 } else { 975 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 976 IP_FORWARDING, &error); 977 } 978 if (sp != NULL) { 979 /* 980 * Check security policy against packet attributes. 981 */ 982 error = ipsec_in_reject(sp, m); 983 KEY_FREESP(&sp); 984 } else { 985 /* XXX error stat??? */ 986 error = EINVAL; 987DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 988 goto bad; 989 } 990 splx(s); 991 if (error) 992 goto bad; 993 } 994#endif /* FAST_IPSEC */ 995 996 /* 997 * Switch out to protocol's input routine. 998 */ 999#if IFA_STATS 1000 if (ia && ip) 1001 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 1002#endif 1003 ipstat.ips_delivered++; 1004 { 1005 int off = hlen, nh = ip->ip_p; 1006 1007 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 1008 return; 1009 } 1010bad: 1011 m_freem(m); 1012 return; 1013 1014badcsum: 1015 ipstat.ips_badsum++; 1016 m_freem(m); 1017} 1018 1019/* 1020 * Take incoming datagram fragment and try to 1021 * reassemble it into whole datagram. If a chain for 1022 * reassembly of this datagram already exists, then it 1023 * is given as fp; otherwise have to make a chain. 1024 */ 1025struct mbuf * 1026ip_reass(ipqe, fp, ipqhead) 1027 struct ipqent *ipqe; 1028 struct ipq *fp; 1029 struct ipqhead *ipqhead; 1030{ 1031 struct mbuf *m = ipqe->ipqe_m; 1032 struct ipqent *nq, *p, *q; 1033 struct ip *ip; 1034 struct mbuf *t; 1035 int hlen = ipqe->ipqe_ip->ip_hl << 2; 1036 int i, next; 1037 1038 IPQ_LOCK_CHECK(); 1039 1040 /* 1041 * Presence of header sizes in mbufs 1042 * would confuse code below. 1043 */ 1044 m->m_data += hlen; 1045 m->m_len -= hlen; 1046 1047#ifdef notyet 1048 /* make sure fragment limit is up-to-date */ 1049 CHECK_NMBCLUSTER_PARAMS(); 1050 1051 /* If we have too many fragments, drop the older half. */ 1052 if (ip_nfrags >= ip_maxfrags) 1053 ip_reass_drophalf(void); 1054#endif 1055 1056 /* 1057 * We are about to add a fragment; increment frag count. 1058 */ 1059 ip_nfrags++; 1060 1061 /* 1062 * If first fragment to arrive, create a reassembly queue. 1063 */ 1064 if (fp == 0) { 1065 /* 1066 * Enforce upper bound on number of fragmented packets 1067 * for which we attempt reassembly; 1068 * If maxfrag is 0, never accept fragments. 1069 * If maxfrag is -1, accept all fragments without limitation. 1070 */ 1071 if (ip_maxfragpackets < 0) 1072 ; 1073 else if (ip_nfragpackets >= ip_maxfragpackets) 1074 goto dropfrag; 1075 ip_nfragpackets++; 1076 MALLOC(fp, struct ipq *, sizeof (struct ipq), 1077 M_FTABLE, M_NOWAIT); 1078 if (fp == NULL) 1079 goto dropfrag; 1080 LIST_INSERT_HEAD(ipqhead, fp, ipq_q); 1081 fp->ipq_nfrags = 1; 1082 fp->ipq_ttl = IPFRAGTTL; 1083 fp->ipq_p = ipqe->ipqe_ip->ip_p; 1084 fp->ipq_id = ipqe->ipqe_ip->ip_id; 1085 TAILQ_INIT(&fp->ipq_fragq); 1086 fp->ipq_src = ipqe->ipqe_ip->ip_src; 1087 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 1088 p = NULL; 1089 goto insert; 1090 } else { 1091 fp->ipq_nfrags++; 1092 } 1093 1094 /* 1095 * Find a segment which begins after this one does. 1096 */ 1097 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1098 p = q, q = TAILQ_NEXT(q, ipqe_q)) 1099 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off)) 1100 break; 1101 1102 /* 1103 * If there is a preceding segment, it may provide some of 1104 * our data already. If so, drop the data from the incoming 1105 * segment. If it provides all of our data, drop us. 1106 */ 1107 if (p != NULL) { 1108 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) - 1109 ntohs(ipqe->ipqe_ip->ip_off); 1110 if (i > 0) { 1111 if (i >= ntohs(ipqe->ipqe_ip->ip_len)) 1112 goto dropfrag; 1113 m_adj(ipqe->ipqe_m, i); 1114 ipqe->ipqe_ip->ip_off = 1115 htons(ntohs(ipqe->ipqe_ip->ip_off) + i); 1116 ipqe->ipqe_ip->ip_len = 1117 htons(ntohs(ipqe->ipqe_ip->ip_len) - i); 1118 } 1119 } 1120 1121 /* 1122 * While we overlap succeeding segments trim them or, 1123 * if they are completely covered, dequeue them. 1124 */ 1125 for (; q != NULL && 1126 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) > 1127 ntohs(q->ipqe_ip->ip_off); q = nq) { 1128 i = (ntohs(ipqe->ipqe_ip->ip_off) + 1129 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off); 1130 if (i < ntohs(q->ipqe_ip->ip_len)) { 1131 q->ipqe_ip->ip_len = 1132 htons(ntohs(q->ipqe_ip->ip_len) - i); 1133 q->ipqe_ip->ip_off = 1134 htons(ntohs(q->ipqe_ip->ip_off) + i); 1135 m_adj(q->ipqe_m, i); 1136 break; 1137 } 1138 nq = TAILQ_NEXT(q, ipqe_q); 1139 m_freem(q->ipqe_m); 1140 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1141 pool_put(&ipqent_pool, q); 1142 fp->ipq_nfrags--; 1143 ip_nfrags--; 1144 } 1145 1146insert: 1147 /* 1148 * Stick new segment in its place; 1149 * check for complete reassembly. 1150 */ 1151 if (p == NULL) { 1152 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 1153 } else { 1154 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q); 1155 } 1156 next = 0; 1157 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1158 p = q, q = TAILQ_NEXT(q, ipqe_q)) { 1159 if (ntohs(q->ipqe_ip->ip_off) != next) 1160 return (0); 1161 next += ntohs(q->ipqe_ip->ip_len); 1162 } 1163 if (p->ipqe_mff) 1164 return (0); 1165 1166 /* 1167 * Reassembly is complete. Check for a bogus message size and 1168 * concatenate fragments. 1169 */ 1170 q = TAILQ_FIRST(&fp->ipq_fragq); 1171 ip = q->ipqe_ip; 1172 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 1173 ipstat.ips_toolong++; 1174 ip_freef(fp); 1175 return (0); 1176 } 1177 m = q->ipqe_m; 1178 t = m->m_next; 1179 m->m_next = 0; 1180 m_cat(m, t); 1181 nq = TAILQ_NEXT(q, ipqe_q); 1182 pool_put(&ipqent_pool, q); 1183 for (q = nq; q != NULL; q = nq) { 1184 t = q->ipqe_m; 1185 nq = TAILQ_NEXT(q, ipqe_q); 1186 pool_put(&ipqent_pool, q); 1187 m_cat(m, t); 1188 } 1189 ip_nfrags -= fp->ipq_nfrags; 1190 1191 /* 1192 * Create header for new ip packet by 1193 * modifying header of first packet; 1194 * dequeue and discard fragment reassembly header. 1195 * Make header visible. 1196 */ 1197 ip->ip_len = htons(next); 1198 ip->ip_src = fp->ipq_src; 1199 ip->ip_dst = fp->ipq_dst; 1200 LIST_REMOVE(fp, ipq_q); 1201 FREE(fp, M_FTABLE); 1202 ip_nfragpackets--; 1203 m->m_len += (ip->ip_hl << 2); 1204 m->m_data -= (ip->ip_hl << 2); 1205 /* some debugging cruft by sklower, below, will go away soon */ 1206 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1207 int plen = 0; 1208 for (t = m; t; t = t->m_next) 1209 plen += t->m_len; 1210 m->m_pkthdr.len = plen; 1211 } 1212 return (m); 1213 1214dropfrag: 1215 if (fp != 0) 1216 fp->ipq_nfrags--; 1217 ip_nfrags--; 1218 ipstat.ips_fragdropped++; 1219 m_freem(m); 1220 pool_put(&ipqent_pool, ipqe); 1221 return (0); 1222} 1223 1224/* 1225 * Free a fragment reassembly header and all 1226 * associated datagrams. 1227 */ 1228void 1229ip_freef(fp) 1230 struct ipq *fp; 1231{ 1232 struct ipqent *q, *p; 1233 u_int nfrags = 0; 1234 1235 IPQ_LOCK_CHECK(); 1236 1237 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) { 1238 p = TAILQ_NEXT(q, ipqe_q); 1239 m_freem(q->ipqe_m); 1240 nfrags++; 1241 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1242 pool_put(&ipqent_pool, q); 1243 } 1244 1245 if (nfrags != fp->ipq_nfrags) 1246 printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags); 1247 ip_nfrags -= nfrags; 1248 LIST_REMOVE(fp, ipq_q); 1249 FREE(fp, M_FTABLE); 1250 ip_nfragpackets--; 1251} 1252 1253/* 1254 * IP reassembly TTL machinery for multiplicative drop. 1255 */ 1256static u_int fragttl_histo[(IPFRAGTTL+1)]; 1257 1258 1259/* 1260 * Decrement TTL of all reasembly queue entries by `ticks'. 1261 * Count number of distinct fragments (as opposed to partial, fragmented 1262 * datagrams) in the reassembly queue. While we traverse the entire 1263 * reassembly queue, compute and return the median TTL over all fragments. 1264 */ 1265static u_int 1266ip_reass_ttl_decr(u_int ticks) 1267{ 1268 u_int nfrags, median, dropfraction, keepfraction; 1269 struct ipq *fp, *nfp; 1270 int i; 1271 1272 nfrags = 0; 1273 memset(fragttl_histo, 0, sizeof fragttl_histo); 1274 1275 for (i = 0; i < IPREASS_NHASH; i++) { 1276 for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) { 1277 fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ? 1278 0 : fp->ipq_ttl - ticks); 1279 nfp = LIST_NEXT(fp, ipq_q); 1280 if (fp->ipq_ttl == 0) { 1281 ipstat.ips_fragtimeout++; 1282 ip_freef(fp); 1283 } else { 1284 nfrags += fp->ipq_nfrags; 1285 fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags; 1286 } 1287 } 1288 } 1289 1290 KASSERT(ip_nfrags == nfrags); 1291 1292 /* Find median (or other drop fraction) in histogram. */ 1293 dropfraction = (ip_nfrags / 2); 1294 keepfraction = ip_nfrags - dropfraction; 1295 for (i = IPFRAGTTL, median = 0; i >= 0; i--) { 1296 median += fragttl_histo[i]; 1297 if (median >= keepfraction) 1298 break; 1299 } 1300 1301 /* Return TTL of median (or other fraction). */ 1302 return (u_int)i; 1303} 1304 1305void 1306ip_reass_drophalf(void) 1307{ 1308 1309 u_int median_ticks; 1310 /* 1311 * Compute median TTL of all fragments, and count frags 1312 * with that TTL or lower (roughly half of all fragments). 1313 */ 1314 median_ticks = ip_reass_ttl_decr(0); 1315 1316 /* Drop half. */ 1317 median_ticks = ip_reass_ttl_decr(median_ticks); 1318 1319} 1320 1321/* 1322 * IP timer processing; 1323 * if a timer expires on a reassembly 1324 * queue, discard it. 1325 */ 1326void 1327ip_slowtimo() 1328{ 1329 static u_int dropscanidx = 0; 1330 u_int i; 1331 u_int median_ttl; 1332 int s = splsoftnet(); 1333 1334 IPQ_LOCK(); 1335 1336 /* Age TTL of all fragments by 1 tick .*/ 1337 median_ttl = ip_reass_ttl_decr(1); 1338 1339 /* make sure fragment limit is up-to-date */ 1340 CHECK_NMBCLUSTER_PARAMS(); 1341 1342 /* If we have too many fragments, drop the older half. */ 1343 if (ip_nfrags > ip_maxfrags) 1344 ip_reass_ttl_decr(median_ttl); 1345 1346 /* 1347 * If we are over the maximum number of fragmented packets 1348 * (due to the limit being lowered), drain off 1349 * enough to get down to the new limit. Start draining 1350 * from the reassembly hashqueue most recently drained. 1351 */ 1352 if (ip_maxfragpackets < 0) 1353 ; 1354 else { 1355 int wrapped = 0; 1356 1357 i = dropscanidx; 1358 while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) { 1359 while (LIST_FIRST(&ipq[i]) != NULL) 1360 ip_freef(LIST_FIRST(&ipq[i])); 1361 if (++i >= IPREASS_NHASH) { 1362 i = 0; 1363 } 1364 /* 1365 * Dont scan forever even if fragment counters are 1366 * wrong: stop after scanning entire reassembly queue. 1367 */ 1368 if (i == dropscanidx) 1369 wrapped = 1; 1370 } 1371 dropscanidx = i; 1372 } 1373 IPQ_UNLOCK(); 1374#ifdef GATEWAY 1375 ipflow_slowtimo(); 1376#endif 1377 splx(s); 1378} 1379 1380/* 1381 * Drain off all datagram fragments. 1382 */ 1383void 1384ip_drain() 1385{ 1386 1387 /* 1388 * We may be called from a device's interrupt context. If 1389 * the ipq is already busy, just bail out now. 1390 */ 1391 if (ipq_lock_try() == 0) 1392 return; 1393 1394 /* 1395 * Drop half the total fragments now. If more mbufs are needed, 1396 * we will be called again soon. 1397 */ 1398 ip_reass_drophalf(); 1399 1400 IPQ_UNLOCK(); 1401} 1402 1403/* 1404 * Do option processing on a datagram, 1405 * possibly discarding it if bad options are encountered, 1406 * or forwarding it if source-routed. 1407 * Returns 1 if packet has been forwarded/freed, 1408 * 0 if the packet should be processed further. 1409 */ 1410int 1411ip_dooptions(m) 1412 struct mbuf *m; 1413{ 1414 struct ip *ip = mtod(m, struct ip *); 1415 u_char *cp, *cp0; 1416 struct ip_timestamp *ipt; 1417 struct in_ifaddr *ia; 1418 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1419 struct in_addr dst; 1420 n_time ntime; 1421 1422 dst = ip->ip_dst; 1423 cp = (u_char *)(ip + 1); 1424 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1425 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1426 opt = cp[IPOPT_OPTVAL]; 1427 if (opt == IPOPT_EOL) 1428 break; 1429 if (opt == IPOPT_NOP) 1430 optlen = 1; 1431 else { 1432 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1433 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1434 goto bad; 1435 } 1436 optlen = cp[IPOPT_OLEN]; 1437 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1438 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1439 goto bad; 1440 } 1441 } 1442 switch (opt) { 1443 1444 default: 1445 break; 1446 1447 /* 1448 * Source routing with record. 1449 * Find interface with current destination address. 1450 * If none on this machine then drop if strictly routed, 1451 * or do nothing if loosely routed. 1452 * Record interface address and bring up next address 1453 * component. If strictly routed make sure next 1454 * address is on directly accessible net. 1455 */ 1456 case IPOPT_LSRR: 1457 case IPOPT_SSRR: 1458 if (ip_allowsrcrt == 0) { 1459 type = ICMP_UNREACH; 1460 code = ICMP_UNREACH_NET_PROHIB; 1461 goto bad; 1462 } 1463 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1464 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1465 goto bad; 1466 } 1467 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1468 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1469 goto bad; 1470 } 1471 ipaddr.sin_addr = ip->ip_dst; 1472 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1473 if (ia == 0) { 1474 if (opt == IPOPT_SSRR) { 1475 type = ICMP_UNREACH; 1476 code = ICMP_UNREACH_SRCFAIL; 1477 goto bad; 1478 } 1479 /* 1480 * Loose routing, and not at next destination 1481 * yet; nothing to do except forward. 1482 */ 1483 break; 1484 } 1485 off--; /* 0 origin */ 1486 if ((off + sizeof(struct in_addr)) > optlen) { 1487 /* 1488 * End of source route. Should be for us. 1489 */ 1490 save_rte(cp, ip->ip_src); 1491 break; 1492 } 1493 /* 1494 * locate outgoing interface 1495 */ 1496 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1497 sizeof(ipaddr.sin_addr)); 1498 if (opt == IPOPT_SSRR) 1499 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); 1500 else 1501 ia = ip_rtaddr(ipaddr.sin_addr); 1502 if (ia == 0) { 1503 type = ICMP_UNREACH; 1504 code = ICMP_UNREACH_SRCFAIL; 1505 goto bad; 1506 } 1507 ip->ip_dst = ipaddr.sin_addr; 1508 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1509 (caddr_t)(cp + off), sizeof(struct in_addr)); 1510 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1511 /* 1512 * Let ip_intr's mcast routing check handle mcast pkts 1513 */ 1514 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1515 break; 1516 1517 case IPOPT_RR: 1518 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1519 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1520 goto bad; 1521 } 1522 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1523 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1524 goto bad; 1525 } 1526 /* 1527 * If no space remains, ignore. 1528 */ 1529 off--; /* 0 origin */ 1530 if ((off + sizeof(struct in_addr)) > optlen) 1531 break; 1532 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1533 sizeof(ipaddr.sin_addr)); 1534 /* 1535 * locate outgoing interface; if we're the destination, 1536 * use the incoming interface (should be same). 1537 */ 1538 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1539 == NULL && 1540 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1541 type = ICMP_UNREACH; 1542 code = ICMP_UNREACH_HOST; 1543 goto bad; 1544 } 1545 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1546 (caddr_t)(cp + off), sizeof(struct in_addr)); 1547 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1548 break; 1549 1550 case IPOPT_TS: 1551 code = cp - (u_char *)ip; 1552 ipt = (struct ip_timestamp *)cp; 1553 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1554 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1555 goto bad; 1556 } 1557 if (ipt->ipt_ptr < 5) { 1558 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1559 goto bad; 1560 } 1561 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1562 if (++ipt->ipt_oflw == 0) { 1563 code = (u_char *)&ipt->ipt_ptr - 1564 (u_char *)ip; 1565 goto bad; 1566 } 1567 break; 1568 } 1569 cp0 = (cp + ipt->ipt_ptr - 1); 1570 switch (ipt->ipt_flg) { 1571 1572 case IPOPT_TS_TSONLY: 1573 break; 1574 1575 case IPOPT_TS_TSANDADDR: 1576 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1577 sizeof(struct in_addr) > ipt->ipt_len) { 1578 code = (u_char *)&ipt->ipt_ptr - 1579 (u_char *)ip; 1580 goto bad; 1581 } 1582 ipaddr.sin_addr = dst; 1583 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1584 m->m_pkthdr.rcvif)); 1585 if (ia == 0) 1586 continue; 1587 bcopy(&ia->ia_addr.sin_addr, 1588 cp0, sizeof(struct in_addr)); 1589 ipt->ipt_ptr += sizeof(struct in_addr); 1590 break; 1591 1592 case IPOPT_TS_PRESPEC: 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 bcopy(cp0, &ipaddr.sin_addr, 1600 sizeof(struct in_addr)); 1601 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1602 == NULL) 1603 continue; 1604 ipt->ipt_ptr += sizeof(struct in_addr); 1605 break; 1606 1607 default: 1608 /* XXX can't take &ipt->ipt_flg */ 1609 code = (u_char *)&ipt->ipt_ptr - 1610 (u_char *)ip + 1; 1611 goto bad; 1612 } 1613 ntime = iptime(); 1614 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1615 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1616 sizeof(n_time)); 1617 ipt->ipt_ptr += sizeof(n_time); 1618 } 1619 } 1620 if (forward) { 1621 if (ip_forwsrcrt == 0) { 1622 type = ICMP_UNREACH; 1623 code = ICMP_UNREACH_SRCFAIL; 1624 goto bad; 1625 } 1626 ip_forward(m, 1); 1627 return (1); 1628 } 1629 return (0); 1630bad: 1631 icmp_error(m, type, code, 0, 0); 1632 ipstat.ips_badoptions++; 1633 return (1); 1634} 1635 1636/* 1637 * Given address of next destination (final or next hop), 1638 * return internet address info of interface to be used to get there. 1639 */ 1640struct in_ifaddr * 1641ip_rtaddr(dst) 1642 struct in_addr dst; 1643{ 1644 struct sockaddr_in *sin; 1645 1646 sin = satosin(&ipforward_rt.ro_dst); 1647 1648 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1649 if (ipforward_rt.ro_rt) { 1650 RTFREE(ipforward_rt.ro_rt); 1651 ipforward_rt.ro_rt = 0; 1652 } 1653 sin->sin_family = AF_INET; 1654 sin->sin_len = sizeof(*sin); 1655 sin->sin_addr = dst; 1656 1657 rtalloc(&ipforward_rt); 1658 } 1659 if (ipforward_rt.ro_rt == 0) 1660 return ((struct in_ifaddr *)0); 1661 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1662} 1663 1664/* 1665 * Save incoming source route for use in replies, 1666 * to be picked up later by ip_srcroute if the receiver is interested. 1667 */ 1668void 1669save_rte(option, dst) 1670 u_char *option; 1671 struct in_addr dst; 1672{ 1673 unsigned olen; 1674 1675 olen = option[IPOPT_OLEN]; 1676#ifdef DIAGNOSTIC 1677 if (ipprintfs) 1678 printf("save_rte: olen %d\n", olen); 1679#endif /* 0 */ 1680 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1681 return; 1682 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1683 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1684 ip_srcrt.dst = dst; 1685} 1686 1687/* 1688 * Retrieve incoming source route for use in replies, 1689 * in the same form used by setsockopt. 1690 * The first hop is placed before the options, will be removed later. 1691 */ 1692struct mbuf * 1693ip_srcroute() 1694{ 1695 struct in_addr *p, *q; 1696 struct mbuf *m; 1697 1698 if (ip_nhops == 0) 1699 return ((struct mbuf *)0); 1700 m = m_get(M_DONTWAIT, MT_SOOPTS); 1701 if (m == 0) 1702 return ((struct mbuf *)0); 1703 1704 MCLAIM(m, &inetdomain.dom_mowner); 1705#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1706 1707 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1708 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1709 OPTSIZ; 1710#ifdef DIAGNOSTIC 1711 if (ipprintfs) 1712 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1713#endif 1714 1715 /* 1716 * First save first hop for return route 1717 */ 1718 p = &ip_srcrt.route[ip_nhops - 1]; 1719 *(mtod(m, struct in_addr *)) = *p--; 1720#ifdef DIAGNOSTIC 1721 if (ipprintfs) 1722 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1723#endif 1724 1725 /* 1726 * Copy option fields and padding (nop) to mbuf. 1727 */ 1728 ip_srcrt.nop = IPOPT_NOP; 1729 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1730 bcopy((caddr_t)&ip_srcrt.nop, 1731 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1732 q = (struct in_addr *)(mtod(m, caddr_t) + 1733 sizeof(struct in_addr) + OPTSIZ); 1734#undef OPTSIZ 1735 /* 1736 * Record return path as an IP source route, 1737 * reversing the path (pointers are now aligned). 1738 */ 1739 while (p >= ip_srcrt.route) { 1740#ifdef DIAGNOSTIC 1741 if (ipprintfs) 1742 printf(" %x", ntohl(q->s_addr)); 1743#endif 1744 *q++ = *p--; 1745 } 1746 /* 1747 * Last hop goes to final destination. 1748 */ 1749 *q = ip_srcrt.dst; 1750#ifdef DIAGNOSTIC 1751 if (ipprintfs) 1752 printf(" %x\n", ntohl(q->s_addr)); 1753#endif 1754 return (m); 1755} 1756 1757/* 1758 * Strip out IP options, at higher 1759 * level protocol in the kernel. 1760 * Second argument is buffer to which options 1761 * will be moved, and return value is their length. 1762 * XXX should be deleted; last arg currently ignored. 1763 */ 1764void 1765ip_stripoptions(m, mopt) 1766 struct mbuf *m; 1767 struct mbuf *mopt; 1768{ 1769 int i; 1770 struct ip *ip = mtod(m, struct ip *); 1771 caddr_t opts; 1772 int olen; 1773 1774 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1775 opts = (caddr_t)(ip + 1); 1776 i = m->m_len - (sizeof (struct ip) + olen); 1777 bcopy(opts + olen, opts, (unsigned)i); 1778 m->m_len -= olen; 1779 if (m->m_flags & M_PKTHDR) 1780 m->m_pkthdr.len -= olen; 1781 ip->ip_len = htons(ntohs(ip->ip_len) - olen); 1782 ip->ip_hl = sizeof (struct ip) >> 2; 1783} 1784 1785const int inetctlerrmap[PRC_NCMDS] = { 1786 0, 0, 0, 0, 1787 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1788 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1789 EMSGSIZE, EHOSTUNREACH, 0, 0, 1790 0, 0, 0, 0, 1791 ENOPROTOOPT 1792}; 1793 1794/* 1795 * Forward a packet. If some error occurs return the sender 1796 * an icmp packet. Note we can't always generate a meaningful 1797 * icmp message because icmp doesn't have a large enough repertoire 1798 * of codes and types. 1799 * 1800 * If not forwarding, just drop the packet. This could be confusing 1801 * if ipforwarding was zero but some routing protocol was advancing 1802 * us as a gateway to somewhere. However, we must let the routing 1803 * protocol deal with that. 1804 * 1805 * The srcrt parameter indicates whether the packet is being forwarded 1806 * via a source route. 1807 */ 1808void 1809ip_forward(m, srcrt) 1810 struct mbuf *m; 1811 int srcrt; 1812{ 1813 struct ip *ip = mtod(m, struct ip *); 1814 struct sockaddr_in *sin; 1815 struct rtentry *rt; 1816 int error, type = 0, code = 0; 1817 struct mbuf *mcopy; 1818 n_long dest; 1819 struct ifnet *destifp; 1820#if defined(IPSEC) || defined(FAST_IPSEC) 1821 struct ifnet dummyifp; 1822#endif 1823 1824 /* 1825 * We are now in the output path. 1826 */ 1827 MCLAIM(m, &ip_tx_mowner); 1828 1829 /* 1830 * Clear any in-bound checksum flags for this packet. 1831 */ 1832 m->m_pkthdr.csum_flags = 0; 1833 1834 dest = 0; 1835#ifdef DIAGNOSTIC 1836 if (ipprintfs) 1837 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1838 ntohl(ip->ip_src.s_addr), 1839 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1840#endif 1841 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1842 ipstat.ips_cantforward++; 1843 m_freem(m); 1844 return; 1845 } 1846 if (ip->ip_ttl <= IPTTLDEC) { 1847 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1848 return; 1849 } 1850 ip->ip_ttl -= IPTTLDEC; 1851 1852 sin = satosin(&ipforward_rt.ro_dst); 1853 if ((rt = ipforward_rt.ro_rt) == 0 || 1854 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1855 if (ipforward_rt.ro_rt) { 1856 RTFREE(ipforward_rt.ro_rt); 1857 ipforward_rt.ro_rt = 0; 1858 } 1859 sin->sin_family = AF_INET; 1860 sin->sin_len = sizeof(struct sockaddr_in); 1861 sin->sin_addr = ip->ip_dst; 1862 1863 rtalloc(&ipforward_rt); 1864 if (ipforward_rt.ro_rt == 0) { 1865 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1866 return; 1867 } 1868 rt = ipforward_rt.ro_rt; 1869 } 1870 1871 /* 1872 * Save at most 68 bytes of the packet in case 1873 * we need to generate an ICMP message to the src. 1874 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1875 */ 1876 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1877 if (mcopy) 1878 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1879 1880 /* 1881 * If forwarding packet using same interface that it came in on, 1882 * perhaps should send a redirect to sender to shortcut a hop. 1883 * Only send redirect if source is sending directly to us, 1884 * and if packet was not source routed (or has any options). 1885 * Also, don't send redirect if forwarding using a default route 1886 * or a route modified by a redirect. 1887 */ 1888 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1889 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1890 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1891 ipsendredirects && !srcrt) { 1892 if (rt->rt_ifa && 1893 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1894 ifatoia(rt->rt_ifa)->ia_subnet) { 1895 if (rt->rt_flags & RTF_GATEWAY) 1896 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1897 else 1898 dest = ip->ip_dst.s_addr; 1899 /* 1900 * Router requirements says to only send host 1901 * redirects. 1902 */ 1903 type = ICMP_REDIRECT; 1904 code = ICMP_REDIRECT_HOST; 1905#ifdef DIAGNOSTIC 1906 if (ipprintfs) 1907 printf("redirect (%d) to %x\n", code, 1908 (u_int32_t)dest); 1909#endif 1910 } 1911 } 1912 1913 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1914 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1915 (struct ip_moptions *)NULL, (struct socket *)NULL); 1916 1917 if (error) 1918 ipstat.ips_cantforward++; 1919 else { 1920 ipstat.ips_forward++; 1921 if (type) 1922 ipstat.ips_redirectsent++; 1923 else { 1924 if (mcopy) { 1925#ifdef GATEWAY 1926 if (mcopy->m_flags & M_CANFASTFWD) 1927 ipflow_create(&ipforward_rt, mcopy); 1928#endif 1929 m_freem(mcopy); 1930 } 1931 return; 1932 } 1933 } 1934 if (mcopy == NULL) 1935 return; 1936 destifp = NULL; 1937 1938 switch (error) { 1939 1940 case 0: /* forwarded, but need redirect */ 1941 /* type, code set above */ 1942 break; 1943 1944 case ENETUNREACH: /* shouldn't happen, checked above */ 1945 case EHOSTUNREACH: 1946 case ENETDOWN: 1947 case EHOSTDOWN: 1948 default: 1949 type = ICMP_UNREACH; 1950 code = ICMP_UNREACH_HOST; 1951 break; 1952 1953 case EMSGSIZE: 1954 type = ICMP_UNREACH; 1955 code = ICMP_UNREACH_NEEDFRAG; 1956#if !defined(IPSEC) && !defined(FAST_IPSEC) 1957 if (ipforward_rt.ro_rt) 1958 destifp = ipforward_rt.ro_rt->rt_ifp; 1959#else 1960 /* 1961 * If the packet is routed over IPsec tunnel, tell the 1962 * originator the tunnel MTU. 1963 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1964 * XXX quickhack!!! 1965 */ 1966 if (ipforward_rt.ro_rt) { 1967 struct secpolicy *sp; 1968 int ipsecerror; 1969 size_t ipsechdr; 1970 struct route *ro; 1971 1972 sp = ipsec4_getpolicybyaddr(mcopy, 1973 IPSEC_DIR_OUTBOUND, IP_FORWARDING, 1974 &ipsecerror); 1975 1976 if (sp == NULL) 1977 destifp = ipforward_rt.ro_rt->rt_ifp; 1978 else { 1979 /* count IPsec header size */ 1980 ipsechdr = ipsec4_hdrsiz(mcopy, 1981 IPSEC_DIR_OUTBOUND, NULL); 1982 1983 /* 1984 * find the correct route for outer IPv4 1985 * header, compute tunnel MTU. 1986 * 1987 * XXX BUG ALERT 1988 * The "dummyifp" code relies upon the fact 1989 * that icmp_error() touches only ifp->if_mtu. 1990 */ 1991 /*XXX*/ 1992 destifp = NULL; 1993 if (sp->req != NULL 1994 && sp->req->sav != NULL 1995 && sp->req->sav->sah != NULL) { 1996 ro = &sp->req->sav->sah->sa_route; 1997 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1998 dummyifp.if_mtu = 1999 ro->ro_rt->rt_rmx.rmx_mtu ? 2000 ro->ro_rt->rt_rmx.rmx_mtu : 2001 ro->ro_rt->rt_ifp->if_mtu; 2002 dummyifp.if_mtu -= ipsechdr; 2003 destifp = &dummyifp; 2004 } 2005 } 2006 2007#ifdef IPSEC 2008 key_freesp(sp); 2009#else 2010 KEY_FREESP(&sp); 2011#endif 2012 } 2013 } 2014#endif /*IPSEC*/ 2015 ipstat.ips_cantfrag++; 2016 break; 2017 2018 case ENOBUFS: 2019#if 1 2020 /* 2021 * a router should not generate ICMP_SOURCEQUENCH as 2022 * required in RFC1812 Requirements for IP Version 4 Routers. 2023 * source quench could be a big problem under DoS attacks, 2024 * or if the underlying interface is rate-limited. 2025 */ 2026 if (mcopy) 2027 m_freem(mcopy); 2028 return; 2029#else 2030 type = ICMP_SOURCEQUENCH; 2031 code = 0; 2032 break; 2033#endif 2034 } 2035 icmp_error(mcopy, type, code, dest, destifp); 2036} 2037 2038void 2039ip_savecontrol(inp, mp, ip, m) 2040 struct inpcb *inp; 2041 struct mbuf **mp; 2042 struct ip *ip; 2043 struct mbuf *m; 2044{ 2045 2046 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2047 struct timeval tv; 2048 2049 microtime(&tv); 2050 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2051 SCM_TIMESTAMP, SOL_SOCKET); 2052 if (*mp) 2053 mp = &(*mp)->m_next; 2054 } 2055 if (inp->inp_flags & INP_RECVDSTADDR) { 2056 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2057 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2058 if (*mp) 2059 mp = &(*mp)->m_next; 2060 } 2061#ifdef notyet 2062 /* 2063 * XXX 2064 * Moving these out of udp_input() made them even more broken 2065 * than they already were. 2066 * - fenner@parc.xerox.com 2067 */ 2068 /* options were tossed already */ 2069 if (inp->inp_flags & INP_RECVOPTS) { 2070 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2071 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2072 if (*mp) 2073 mp = &(*mp)->m_next; 2074 } 2075 /* ip_srcroute doesn't do what we want here, need to fix */ 2076 if (inp->inp_flags & INP_RECVRETOPTS) { 2077 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 2078 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2079 if (*mp) 2080 mp = &(*mp)->m_next; 2081 } 2082#endif 2083 if (inp->inp_flags & INP_RECVIF) { 2084 struct sockaddr_dl sdl; 2085 2086 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 2087 sdl.sdl_family = AF_LINK; 2088 sdl.sdl_index = m->m_pkthdr.rcvif ? 2089 m->m_pkthdr.rcvif->if_index : 0; 2090 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 2091 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 2092 IP_RECVIF, IPPROTO_IP); 2093 if (*mp) 2094 mp = &(*mp)->m_next; 2095 } 2096} 2097 2098/* 2099 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 2100 * range of the new value and tweaks timers if it changes. 2101 */ 2102static int 2103sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 2104{ 2105 int error, tmp; 2106 struct sysctlnode node; 2107 2108 node = *rnode; 2109 tmp = ip_mtudisc_timeout; 2110 node.sysctl_data = &tmp; 2111 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2112 if (error || newp == NULL) 2113 return (error); 2114 if (tmp < 0) 2115 return (EINVAL); 2116 2117 ip_mtudisc_timeout = tmp; 2118 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 2119 2120 return (0); 2121} 2122 2123#ifdef GATEWAY 2124/* 2125 * sysctl helper routine for net.inet.ip.maxflows. apparently if 2126 * maxflows is even looked up, we "reap flows". 2127 */ 2128static int 2129sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS) 2130{ 2131 int s; 2132 2133 s = sysctl_lookup(SYSCTLFN_CALL(rnode)); 2134 if (s) 2135 return (s); 2136 2137 s = splsoftnet(); 2138 ipflow_reap(0); 2139 splx(s); 2140 2141 return (0); 2142} 2143#endif /* GATEWAY */ 2144 2145 2146SYSCTL_SETUP(sysctl_net_inet_ip_setup, "sysctl net.inet.ip subtree setup") 2147{ 2148 extern int subnetsarelocal, hostzeroisbroadcast; 2149 2150 sysctl_createv(clog, 0, NULL, NULL, 2151 CTLFLAG_PERMANENT, 2152 CTLTYPE_NODE, "net", NULL, 2153 NULL, 0, NULL, 0, 2154 CTL_NET, CTL_EOL); 2155 sysctl_createv(clog, 0, NULL, NULL, 2156 CTLFLAG_PERMANENT, 2157 CTLTYPE_NODE, "inet", NULL, 2158 NULL, 0, NULL, 0, 2159 CTL_NET, PF_INET, CTL_EOL); 2160 sysctl_createv(clog, 0, NULL, NULL, 2161 CTLFLAG_PERMANENT, 2162 CTLTYPE_NODE, "ip", NULL, 2163 NULL, 0, NULL, 0, 2164 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 2165 2166 sysctl_createv(clog, 0, NULL, NULL, 2167 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2168 CTLTYPE_INT, "forwarding", NULL, 2169 NULL, 0, &ipforwarding, 0, 2170 CTL_NET, PF_INET, IPPROTO_IP, 2171 IPCTL_FORWARDING, CTL_EOL); 2172 sysctl_createv(clog, 0, NULL, NULL, 2173 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2174 CTLTYPE_INT, "redirect", NULL, 2175 NULL, 0, &ipsendredirects, 0, 2176 CTL_NET, PF_INET, IPPROTO_IP, 2177 IPCTL_SENDREDIRECTS, CTL_EOL); 2178 sysctl_createv(clog, 0, NULL, NULL, 2179 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2180 CTLTYPE_INT, "ttl", NULL, 2181 NULL, 0, &ip_defttl, 0, 2182 CTL_NET, PF_INET, IPPROTO_IP, 2183 IPCTL_DEFTTL, CTL_EOL); 2184#ifdef IPCTL_DEFMTU 2185 sysctl_createv(clog, 0, NULL, NULL, 2186 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 2187 CTLTYPE_INT, "mtu", NULL, 2188 NULL, 0, &ip_mtu, 0, 2189 CTL_NET, PF_INET, IPPROTO_IP, 2190 IPCTL_DEFMTU, CTL_EOL); 2191#endif /* IPCTL_DEFMTU */ 2192 sysctl_createv(clog, 0, NULL, NULL, 2193 CTLFLAG_PERMANENT|CTLFLAG_READONLY1, 2194 CTLTYPE_INT, "forwsrcrt", NULL, 2195 NULL, 0, &ip_forwsrcrt, 0, 2196 CTL_NET, PF_INET, IPPROTO_IP, 2197 IPCTL_FORWSRCRT, CTL_EOL); 2198 sysctl_createv(clog, 0, NULL, NULL, 2199 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2200 CTLTYPE_INT, "directed-broadcast", NULL, 2201 NULL, 0, &ip_directedbcast, 0, 2202 CTL_NET, PF_INET, IPPROTO_IP, 2203 IPCTL_DIRECTEDBCAST, CTL_EOL); 2204 sysctl_createv(clog, 0, NULL, NULL, 2205 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2206 CTLTYPE_INT, "allowsrcrt", NULL, 2207 NULL, 0, &ip_allowsrcrt, 0, 2208 CTL_NET, PF_INET, IPPROTO_IP, 2209 IPCTL_ALLOWSRCRT, CTL_EOL); 2210 sysctl_createv(clog, 0, NULL, NULL, 2211 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2212 CTLTYPE_INT, "subnetsarelocal", NULL, 2213 NULL, 0, &subnetsarelocal, 0, 2214 CTL_NET, PF_INET, IPPROTO_IP, 2215 IPCTL_SUBNETSARELOCAL, CTL_EOL); 2216 sysctl_createv(clog, 0, NULL, NULL, 2217 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2218 CTLTYPE_INT, "mtudisc", NULL, 2219 NULL, 0, &ip_mtudisc, 0, 2220 CTL_NET, PF_INET, IPPROTO_IP, 2221 IPCTL_MTUDISC, CTL_EOL); 2222 sysctl_createv(clog, 0, NULL, NULL, 2223 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2224 CTLTYPE_INT, "anonportmin", NULL, 2225 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 2226 CTL_NET, PF_INET, IPPROTO_IP, 2227 IPCTL_ANONPORTMIN, CTL_EOL); 2228 sysctl_createv(clog, 0, NULL, NULL, 2229 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2230 CTLTYPE_INT, "anonportmax", NULL, 2231 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 2232 CTL_NET, PF_INET, IPPROTO_IP, 2233 IPCTL_ANONPORTMAX, CTL_EOL); 2234 sysctl_createv(clog, 0, NULL, NULL, 2235 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2236 CTLTYPE_INT, "mtudisctimeout", NULL, 2237 sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0, 2238 CTL_NET, PF_INET, IPPROTO_IP, 2239 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 2240#ifdef GATEWAY 2241 sysctl_createv(clog, 0, NULL, NULL, 2242 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2243 CTLTYPE_INT, "maxflows", NULL, 2244 sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0, 2245 CTL_NET, PF_INET, IPPROTO_IP, 2246 IPCTL_MAXFLOWS, CTL_EOL); 2247#endif /* GATEWAY */ 2248 sysctl_createv(clog, 0, NULL, NULL, 2249 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2250 CTLTYPE_INT, "hostzerobroadcast", NULL, 2251 NULL, 0, &hostzeroisbroadcast, 0, 2252 CTL_NET, PF_INET, IPPROTO_IP, 2253 IPCTL_HOSTZEROBROADCAST, CTL_EOL); 2254#if NGIF > 0 2255 sysctl_createv(clog, 0, NULL, NULL, 2256 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2257 CTLTYPE_INT, "gifttl", NULL, 2258 NULL, 0, &ip_gif_ttl, 0, 2259 CTL_NET, PF_INET, IPPROTO_IP, 2260 IPCTL_GIF_TTL, CTL_EOL); 2261#endif /* NGIF */ 2262#ifndef IPNOPRIVPORTS 2263 sysctl_createv(clog, 0, NULL, NULL, 2264 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2265 CTLTYPE_INT, "lowportmin", NULL, 2266 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 2267 CTL_NET, PF_INET, IPPROTO_IP, 2268 IPCTL_LOWPORTMIN, CTL_EOL); 2269 sysctl_createv(clog, 0, NULL, NULL, 2270 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2271 CTLTYPE_INT, "lowportmax", NULL, 2272 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 2273 CTL_NET, PF_INET, IPPROTO_IP, 2274 IPCTL_LOWPORTMAX, CTL_EOL); 2275#endif /* IPNOPRIVPORTS */ 2276 sysctl_createv(clog, 0, NULL, NULL, 2277 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2278 CTLTYPE_INT, "maxfragpackets", NULL, 2279 NULL, 0, &ip_maxfragpackets, 0, 2280 CTL_NET, PF_INET, IPPROTO_IP, 2281 IPCTL_MAXFRAGPACKETS, CTL_EOL); 2282#if NGRE > 0 2283 sysctl_createv(clog, 0, NULL, NULL, 2284 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2285 CTLTYPE_INT, "grettl", NULL, 2286 NULL, 0, &ip_gre_ttl, 0, 2287 CTL_NET, PF_INET, IPPROTO_IP, 2288 IPCTL_GRE_TTL, CTL_EOL); 2289#endif /* NGRE */ 2290 sysctl_createv(clog, 0, NULL, NULL, 2291 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2292 CTLTYPE_INT, "checkinterface", NULL, 2293 NULL, 0, &ip_checkinterface, 0, 2294 CTL_NET, PF_INET, IPPROTO_IP, 2295 IPCTL_CHECKINTERFACE, CTL_EOL); 2296 sysctl_createv(clog, 0, NULL, NULL, 2297 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2298 CTLTYPE_INT, "random_id", NULL, 2299 NULL, 0, &ip_do_randomid, 0, 2300 CTL_NET, PF_INET, IPPROTO_IP, 2301 IPCTL_RANDOMID, CTL_EOL); 2302} 2303