ip_input.c revision 1.201
1/* $NetBSD: ip_input.c,v 1.201 2004/05/01 02:20:42 matt 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 1.201 2004/05/01 02:20:42 matt 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 (M_READONLY(m)) { 742 if ((m = m_pullup(m, hlen)) == 0) { 743 ipstat.ips_toosmall++; 744 return; 745 } 746 ip = mtod(m, struct ip *); 747 } 748 749 if (ip_mrouter) { 750 /* 751 * If we are acting as a multicast router, all 752 * incoming multicast packets are passed to the 753 * kernel-level multicast forwarding function. 754 * The packet is returned (relatively) intact; if 755 * ip_mforward() returns a non-zero value, the packet 756 * must be discarded, else it may be accepted below. 757 * 758 * (The IP ident field is put in the same byte order 759 * as expected when ip_mforward() is called from 760 * ip_output().) 761 */ 762 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 763 ipstat.ips_cantforward++; 764 m_freem(m); 765 return; 766 } 767 768 /* 769 * The process-level routing demon needs to receive 770 * all multicast IGMP packets, whether or not this 771 * host belongs to their destination groups. 772 */ 773 if (ip->ip_p == IPPROTO_IGMP) 774 goto ours; 775 ipstat.ips_forward++; 776 } 777#endif 778 /* 779 * See if we belong to the destination multicast group on the 780 * arrival interface. 781 */ 782 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 783 if (inm == NULL) { 784 ipstat.ips_cantforward++; 785 m_freem(m); 786 return; 787 } 788 goto ours; 789 } 790 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 791 in_nullhost(ip->ip_dst)) 792 goto ours; 793 794 /* 795 * Not for us; forward if possible and desirable. 796 */ 797 if (ipforwarding == 0) { 798 ipstat.ips_cantforward++; 799 m_freem(m); 800 } else { 801 /* 802 * If ip_dst matched any of my address on !IFF_UP interface, 803 * and there's no IFF_UP interface that matches ip_dst, 804 * send icmp unreach. Forwarding it will result in in-kernel 805 * forwarding loop till TTL goes to 0. 806 */ 807 if (downmatch) { 808 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 809 ipstat.ips_cantforward++; 810 return; 811 } 812#ifdef IPSEC 813 if (ipsec4_in_reject(m, NULL)) { 814 ipsecstat.in_polvio++; 815 goto bad; 816 } 817#endif 818#ifdef FAST_IPSEC 819 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 820 s = splsoftnet(); 821 if (mtag != NULL) { 822 tdbi = (struct tdb_ident *)(mtag + 1); 823 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 824 } else { 825 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 826 IP_FORWARDING, &error); 827 } 828 if (sp == NULL) { /* NB: can happen if error */ 829 splx(s); 830 /*XXX error stat???*/ 831 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 832 goto bad; 833 } 834 835 /* 836 * Check security policy against packet attributes. 837 */ 838 error = ipsec_in_reject(sp, m); 839 KEY_FREESP(&sp); 840 splx(s); 841 if (error) { 842 ipstat.ips_cantforward++; 843 goto bad; 844 } 845 846 /* 847 * Peek at the outbound SP for this packet to determine if 848 * it's a Fast Forward candidate. 849 */ 850 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 851 if (mtag != NULL) 852 m->m_flags &= ~M_CANFASTFWD; 853 else { 854 s = splsoftnet(); 855 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, 856 (IP_FORWARDING | 857 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 858 &error, NULL); 859 if (sp != NULL) { 860 m->m_flags &= ~M_CANFASTFWD; 861 KEY_FREESP(&sp); 862 } 863 splx(s); 864 } 865#endif /* FAST_IPSEC */ 866 867 ip_forward(m, srcrt); 868 } 869 return; 870 871ours: 872 /* 873 * If offset or IP_MF are set, must reassemble. 874 * Otherwise, nothing need be done. 875 * (We could look in the reassembly queue to see 876 * if the packet was previously fragmented, 877 * but it's not worth the time; just let them time out.) 878 */ 879 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 880 if (M_READONLY(m)) { 881 if ((m = m_pullup(m, hlen)) == NULL) { 882 ipstat.ips_toosmall++; 883 goto bad; 884 } 885 ip = mtod(m, struct ip *); 886 } 887 888 /* 889 * Look for queue of fragments 890 * of this datagram. 891 */ 892 IPQ_LOCK(); 893 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 894 /* XXX LIST_FOREACH(fp, &ipq[hash], ipq_q) */ 895 for (fp = LIST_FIRST(&ipq[hash]); fp != NULL; 896 fp = LIST_NEXT(fp, ipq_q)) { 897 if (ip->ip_id == fp->ipq_id && 898 in_hosteq(ip->ip_src, fp->ipq_src) && 899 in_hosteq(ip->ip_dst, fp->ipq_dst) && 900 ip->ip_p == fp->ipq_p) 901 goto found; 902 903 } 904 fp = 0; 905found: 906 907 /* 908 * Adjust ip_len to not reflect header, 909 * set ipqe_mff if more fragments are expected, 910 * convert offset of this to bytes. 911 */ 912 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 913 mff = (ip->ip_off & htons(IP_MF)) != 0; 914 if (mff) { 915 /* 916 * Make sure that fragments have a data length 917 * that's a non-zero multiple of 8 bytes. 918 */ 919 if (ntohs(ip->ip_len) == 0 || 920 (ntohs(ip->ip_len) & 0x7) != 0) { 921 ipstat.ips_badfrags++; 922 IPQ_UNLOCK(); 923 goto bad; 924 } 925 } 926 ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3); 927 928 /* 929 * If datagram marked as having more fragments 930 * or if this is not the first fragment, 931 * attempt reassembly; if it succeeds, proceed. 932 */ 933 if (mff || ip->ip_off != htons(0)) { 934 ipstat.ips_fragments++; 935 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 936 if (ipqe == NULL) { 937 ipstat.ips_rcvmemdrop++; 938 IPQ_UNLOCK(); 939 goto bad; 940 } 941 ipqe->ipqe_mff = mff; 942 ipqe->ipqe_m = m; 943 ipqe->ipqe_ip = ip; 944 m = ip_reass(ipqe, fp, &ipq[hash]); 945 if (m == 0) { 946 IPQ_UNLOCK(); 947 return; 948 } 949 ipstat.ips_reassembled++; 950 ip = mtod(m, struct ip *); 951 hlen = ip->ip_hl << 2; 952 ip->ip_len = htons(ntohs(ip->ip_len) + hlen); 953 } else 954 if (fp) 955 ip_freef(fp); 956 IPQ_UNLOCK(); 957 } 958 959#if defined(IPSEC) 960 /* 961 * enforce IPsec policy checking if we are seeing last header. 962 * note that we do not visit this with protocols with pcb layer 963 * code - like udp/tcp/raw ip. 964 */ 965 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 966 ipsec4_in_reject(m, NULL)) { 967 ipsecstat.in_polvio++; 968 goto bad; 969 } 970#endif 971#if FAST_IPSEC 972 /* 973 * enforce IPsec policy checking if we are seeing last header. 974 * note that we do not visit this with protocols with pcb layer 975 * code - like udp/tcp/raw ip. 976 */ 977 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 978 /* 979 * Check if the packet has already had IPsec processing 980 * done. If so, then just pass it along. This tag gets 981 * set during AH, ESP, etc. input handling, before the 982 * packet is returned to the ip input queue for delivery. 983 */ 984 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 985 s = splsoftnet(); 986 if (mtag != NULL) { 987 tdbi = (struct tdb_ident *)(mtag + 1); 988 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 989 } else { 990 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 991 IP_FORWARDING, &error); 992 } 993 if (sp != NULL) { 994 /* 995 * Check security policy against packet attributes. 996 */ 997 error = ipsec_in_reject(sp, m); 998 KEY_FREESP(&sp); 999 } else { 1000 /* XXX error stat??? */ 1001 error = EINVAL; 1002DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 1003 goto bad; 1004 } 1005 splx(s); 1006 if (error) 1007 goto bad; 1008 } 1009#endif /* FAST_IPSEC */ 1010 1011 /* 1012 * Switch out to protocol's input routine. 1013 */ 1014#if IFA_STATS 1015 if (ia && ip) 1016 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 1017#endif 1018 ipstat.ips_delivered++; 1019 { 1020 int off = hlen, nh = ip->ip_p; 1021 1022 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 1023 return; 1024 } 1025bad: 1026 m_freem(m); 1027 return; 1028 1029badcsum: 1030 ipstat.ips_badsum++; 1031 m_freem(m); 1032} 1033 1034/* 1035 * Take incoming datagram fragment and try to 1036 * reassemble it into whole datagram. If a chain for 1037 * reassembly of this datagram already exists, then it 1038 * is given as fp; otherwise have to make a chain. 1039 */ 1040struct mbuf * 1041ip_reass(ipqe, fp, ipqhead) 1042 struct ipqent *ipqe; 1043 struct ipq *fp; 1044 struct ipqhead *ipqhead; 1045{ 1046 struct mbuf *m = ipqe->ipqe_m; 1047 struct ipqent *nq, *p, *q; 1048 struct ip *ip; 1049 struct mbuf *t; 1050 int hlen = ipqe->ipqe_ip->ip_hl << 2; 1051 int i, next; 1052 1053 IPQ_LOCK_CHECK(); 1054 1055 /* 1056 * Presence of header sizes in mbufs 1057 * would confuse code below. 1058 */ 1059 m->m_data += hlen; 1060 m->m_len -= hlen; 1061 1062#ifdef notyet 1063 /* make sure fragment limit is up-to-date */ 1064 CHECK_NMBCLUSTER_PARAMS(); 1065 1066 /* If we have too many fragments, drop the older half. */ 1067 if (ip_nfrags >= ip_maxfrags) 1068 ip_reass_drophalf(void); 1069#endif 1070 1071 /* 1072 * We are about to add a fragment; increment frag count. 1073 */ 1074 ip_nfrags++; 1075 1076 /* 1077 * If first fragment to arrive, create a reassembly queue. 1078 */ 1079 if (fp == 0) { 1080 /* 1081 * Enforce upper bound on number of fragmented packets 1082 * for which we attempt reassembly; 1083 * If maxfrag is 0, never accept fragments. 1084 * If maxfrag is -1, accept all fragments without limitation. 1085 */ 1086 if (ip_maxfragpackets < 0) 1087 ; 1088 else if (ip_nfragpackets >= ip_maxfragpackets) 1089 goto dropfrag; 1090 ip_nfragpackets++; 1091 MALLOC(fp, struct ipq *, sizeof (struct ipq), 1092 M_FTABLE, M_NOWAIT); 1093 if (fp == NULL) 1094 goto dropfrag; 1095 LIST_INSERT_HEAD(ipqhead, fp, ipq_q); 1096 fp->ipq_nfrags = 1; 1097 fp->ipq_ttl = IPFRAGTTL; 1098 fp->ipq_p = ipqe->ipqe_ip->ip_p; 1099 fp->ipq_id = ipqe->ipqe_ip->ip_id; 1100 TAILQ_INIT(&fp->ipq_fragq); 1101 fp->ipq_src = ipqe->ipqe_ip->ip_src; 1102 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 1103 p = NULL; 1104 goto insert; 1105 } else { 1106 fp->ipq_nfrags++; 1107 } 1108 1109 /* 1110 * Find a segment which begins after this one does. 1111 */ 1112 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1113 p = q, q = TAILQ_NEXT(q, ipqe_q)) 1114 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off)) 1115 break; 1116 1117 /* 1118 * If there is a preceding segment, it may provide some of 1119 * our data already. If so, drop the data from the incoming 1120 * segment. If it provides all of our data, drop us. 1121 */ 1122 if (p != NULL) { 1123 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) - 1124 ntohs(ipqe->ipqe_ip->ip_off); 1125 if (i > 0) { 1126 if (i >= ntohs(ipqe->ipqe_ip->ip_len)) 1127 goto dropfrag; 1128 m_adj(ipqe->ipqe_m, i); 1129 ipqe->ipqe_ip->ip_off = 1130 htons(ntohs(ipqe->ipqe_ip->ip_off) + i); 1131 ipqe->ipqe_ip->ip_len = 1132 htons(ntohs(ipqe->ipqe_ip->ip_len) - i); 1133 } 1134 } 1135 1136 /* 1137 * While we overlap succeeding segments trim them or, 1138 * if they are completely covered, dequeue them. 1139 */ 1140 for (; q != NULL && 1141 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) > 1142 ntohs(q->ipqe_ip->ip_off); q = nq) { 1143 i = (ntohs(ipqe->ipqe_ip->ip_off) + 1144 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off); 1145 if (i < ntohs(q->ipqe_ip->ip_len)) { 1146 q->ipqe_ip->ip_len = 1147 htons(ntohs(q->ipqe_ip->ip_len) - i); 1148 q->ipqe_ip->ip_off = 1149 htons(ntohs(q->ipqe_ip->ip_off) + i); 1150 m_adj(q->ipqe_m, i); 1151 break; 1152 } 1153 nq = TAILQ_NEXT(q, ipqe_q); 1154 m_freem(q->ipqe_m); 1155 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1156 pool_put(&ipqent_pool, q); 1157 fp->ipq_nfrags--; 1158 ip_nfrags--; 1159 } 1160 1161insert: 1162 /* 1163 * Stick new segment in its place; 1164 * check for complete reassembly. 1165 */ 1166 if (p == NULL) { 1167 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 1168 } else { 1169 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q); 1170 } 1171 next = 0; 1172 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1173 p = q, q = TAILQ_NEXT(q, ipqe_q)) { 1174 if (ntohs(q->ipqe_ip->ip_off) != next) 1175 return (0); 1176 next += ntohs(q->ipqe_ip->ip_len); 1177 } 1178 if (p->ipqe_mff) 1179 return (0); 1180 1181 /* 1182 * Reassembly is complete. Check for a bogus message size and 1183 * concatenate fragments. 1184 */ 1185 q = TAILQ_FIRST(&fp->ipq_fragq); 1186 ip = q->ipqe_ip; 1187 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 1188 ipstat.ips_toolong++; 1189 ip_freef(fp); 1190 return (0); 1191 } 1192 m = q->ipqe_m; 1193 t = m->m_next; 1194 m->m_next = 0; 1195 m_cat(m, t); 1196 nq = TAILQ_NEXT(q, ipqe_q); 1197 pool_put(&ipqent_pool, q); 1198 for (q = nq; q != NULL; q = nq) { 1199 t = q->ipqe_m; 1200 nq = TAILQ_NEXT(q, ipqe_q); 1201 pool_put(&ipqent_pool, q); 1202 m_cat(m, t); 1203 } 1204 ip_nfrags -= fp->ipq_nfrags; 1205 1206 /* 1207 * Create header for new ip packet by 1208 * modifying header of first packet; 1209 * dequeue and discard fragment reassembly header. 1210 * Make header visible. 1211 */ 1212 ip->ip_len = htons(next); 1213 ip->ip_src = fp->ipq_src; 1214 ip->ip_dst = fp->ipq_dst; 1215 LIST_REMOVE(fp, ipq_q); 1216 FREE(fp, M_FTABLE); 1217 ip_nfragpackets--; 1218 m->m_len += (ip->ip_hl << 2); 1219 m->m_data -= (ip->ip_hl << 2); 1220 /* some debugging cruft by sklower, below, will go away soon */ 1221 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1222 int plen = 0; 1223 for (t = m; t; t = t->m_next) 1224 plen += t->m_len; 1225 m->m_pkthdr.len = plen; 1226 } 1227 return (m); 1228 1229dropfrag: 1230 if (fp != 0) 1231 fp->ipq_nfrags--; 1232 ip_nfrags--; 1233 ipstat.ips_fragdropped++; 1234 m_freem(m); 1235 pool_put(&ipqent_pool, ipqe); 1236 return (0); 1237} 1238 1239/* 1240 * Free a fragment reassembly header and all 1241 * associated datagrams. 1242 */ 1243void 1244ip_freef(fp) 1245 struct ipq *fp; 1246{ 1247 struct ipqent *q, *p; 1248 u_int nfrags = 0; 1249 1250 IPQ_LOCK_CHECK(); 1251 1252 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) { 1253 p = TAILQ_NEXT(q, ipqe_q); 1254 m_freem(q->ipqe_m); 1255 nfrags++; 1256 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1257 pool_put(&ipqent_pool, q); 1258 } 1259 1260 if (nfrags != fp->ipq_nfrags) 1261 printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags); 1262 ip_nfrags -= nfrags; 1263 LIST_REMOVE(fp, ipq_q); 1264 FREE(fp, M_FTABLE); 1265 ip_nfragpackets--; 1266} 1267 1268/* 1269 * IP reassembly TTL machinery for multiplicative drop. 1270 */ 1271static u_int fragttl_histo[(IPFRAGTTL+1)]; 1272 1273 1274/* 1275 * Decrement TTL of all reasembly queue entries by `ticks'. 1276 * Count number of distinct fragments (as opposed to partial, fragmented 1277 * datagrams) in the reassembly queue. While we traverse the entire 1278 * reassembly queue, compute and return the median TTL over all fragments. 1279 */ 1280static u_int 1281ip_reass_ttl_decr(u_int ticks) 1282{ 1283 u_int nfrags, median, dropfraction, keepfraction; 1284 struct ipq *fp, *nfp; 1285 int i; 1286 1287 nfrags = 0; 1288 memset(fragttl_histo, 0, sizeof fragttl_histo); 1289 1290 for (i = 0; i < IPREASS_NHASH; i++) { 1291 for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) { 1292 fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ? 1293 0 : fp->ipq_ttl - ticks); 1294 nfp = LIST_NEXT(fp, ipq_q); 1295 if (fp->ipq_ttl == 0) { 1296 ipstat.ips_fragtimeout++; 1297 ip_freef(fp); 1298 } else { 1299 nfrags += fp->ipq_nfrags; 1300 fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags; 1301 } 1302 } 1303 } 1304 1305 KASSERT(ip_nfrags == nfrags); 1306 1307 /* Find median (or other drop fraction) in histogram. */ 1308 dropfraction = (ip_nfrags / 2); 1309 keepfraction = ip_nfrags - dropfraction; 1310 for (i = IPFRAGTTL, median = 0; i >= 0; i--) { 1311 median += fragttl_histo[i]; 1312 if (median >= keepfraction) 1313 break; 1314 } 1315 1316 /* Return TTL of median (or other fraction). */ 1317 return (u_int)i; 1318} 1319 1320void 1321ip_reass_drophalf(void) 1322{ 1323 1324 u_int median_ticks; 1325 /* 1326 * Compute median TTL of all fragments, and count frags 1327 * with that TTL or lower (roughly half of all fragments). 1328 */ 1329 median_ticks = ip_reass_ttl_decr(0); 1330 1331 /* Drop half. */ 1332 median_ticks = ip_reass_ttl_decr(median_ticks); 1333 1334} 1335 1336/* 1337 * IP timer processing; 1338 * if a timer expires on a reassembly 1339 * queue, discard it. 1340 */ 1341void 1342ip_slowtimo() 1343{ 1344 static u_int dropscanidx = 0; 1345 u_int i; 1346 u_int median_ttl; 1347 int s = splsoftnet(); 1348 1349 IPQ_LOCK(); 1350 1351 /* Age TTL of all fragments by 1 tick .*/ 1352 median_ttl = ip_reass_ttl_decr(1); 1353 1354 /* make sure fragment limit is up-to-date */ 1355 CHECK_NMBCLUSTER_PARAMS(); 1356 1357 /* If we have too many fragments, drop the older half. */ 1358 if (ip_nfrags > ip_maxfrags) 1359 ip_reass_ttl_decr(median_ttl); 1360 1361 /* 1362 * If we are over the maximum number of fragmented packets 1363 * (due to the limit being lowered), drain off 1364 * enough to get down to the new limit. Start draining 1365 * from the reassembly hashqueue most recently drained. 1366 */ 1367 if (ip_maxfragpackets < 0) 1368 ; 1369 else { 1370 int wrapped = 0; 1371 1372 i = dropscanidx; 1373 while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) { 1374 while (LIST_FIRST(&ipq[i]) != NULL) 1375 ip_freef(LIST_FIRST(&ipq[i])); 1376 if (++i >= IPREASS_NHASH) { 1377 i = 0; 1378 } 1379 /* 1380 * Dont scan forever even if fragment counters are 1381 * wrong: stop after scanning entire reassembly queue. 1382 */ 1383 if (i == dropscanidx) 1384 wrapped = 1; 1385 } 1386 dropscanidx = i; 1387 } 1388 IPQ_UNLOCK(); 1389#ifdef GATEWAY 1390 ipflow_slowtimo(); 1391#endif 1392 splx(s); 1393} 1394 1395/* 1396 * Drain off all datagram fragments. 1397 */ 1398void 1399ip_drain() 1400{ 1401 1402 /* 1403 * We may be called from a device's interrupt context. If 1404 * the ipq is already busy, just bail out now. 1405 */ 1406 if (ipq_lock_try() == 0) 1407 return; 1408 1409 /* 1410 * Drop half the total fragments now. If more mbufs are needed, 1411 * we will be called again soon. 1412 */ 1413 ip_reass_drophalf(); 1414 1415 IPQ_UNLOCK(); 1416} 1417 1418/* 1419 * Do option processing on a datagram, 1420 * possibly discarding it if bad options are encountered, 1421 * or forwarding it if source-routed. 1422 * Returns 1 if packet has been forwarded/freed, 1423 * 0 if the packet should be processed further. 1424 */ 1425int 1426ip_dooptions(m) 1427 struct mbuf *m; 1428{ 1429 struct ip *ip = mtod(m, struct ip *); 1430 u_char *cp, *cp0; 1431 struct ip_timestamp *ipt; 1432 struct in_ifaddr *ia; 1433 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1434 struct in_addr dst; 1435 n_time ntime; 1436 1437 dst = ip->ip_dst; 1438 cp = (u_char *)(ip + 1); 1439 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1440 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1441 opt = cp[IPOPT_OPTVAL]; 1442 if (opt == IPOPT_EOL) 1443 break; 1444 if (opt == IPOPT_NOP) 1445 optlen = 1; 1446 else { 1447 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1448 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1449 goto bad; 1450 } 1451 optlen = cp[IPOPT_OLEN]; 1452 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1453 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1454 goto bad; 1455 } 1456 } 1457 switch (opt) { 1458 1459 default: 1460 break; 1461 1462 /* 1463 * Source routing with record. 1464 * Find interface with current destination address. 1465 * If none on this machine then drop if strictly routed, 1466 * or do nothing if loosely routed. 1467 * Record interface address and bring up next address 1468 * component. If strictly routed make sure next 1469 * address is on directly accessible net. 1470 */ 1471 case IPOPT_LSRR: 1472 case IPOPT_SSRR: 1473 if (ip_allowsrcrt == 0) { 1474 type = ICMP_UNREACH; 1475 code = ICMP_UNREACH_NET_PROHIB; 1476 goto bad; 1477 } 1478 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1479 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1480 goto bad; 1481 } 1482 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1483 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1484 goto bad; 1485 } 1486 ipaddr.sin_addr = ip->ip_dst; 1487 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1488 if (ia == 0) { 1489 if (opt == IPOPT_SSRR) { 1490 type = ICMP_UNREACH; 1491 code = ICMP_UNREACH_SRCFAIL; 1492 goto bad; 1493 } 1494 /* 1495 * Loose routing, and not at next destination 1496 * yet; nothing to do except forward. 1497 */ 1498 break; 1499 } 1500 off--; /* 0 origin */ 1501 if ((off + sizeof(struct in_addr)) > optlen) { 1502 /* 1503 * End of source route. Should be for us. 1504 */ 1505 save_rte(cp, ip->ip_src); 1506 break; 1507 } 1508 /* 1509 * locate outgoing interface 1510 */ 1511 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1512 sizeof(ipaddr.sin_addr)); 1513 if (opt == IPOPT_SSRR) 1514 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); 1515 else 1516 ia = ip_rtaddr(ipaddr.sin_addr); 1517 if (ia == 0) { 1518 type = ICMP_UNREACH; 1519 code = ICMP_UNREACH_SRCFAIL; 1520 goto bad; 1521 } 1522 ip->ip_dst = ipaddr.sin_addr; 1523 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1524 (caddr_t)(cp + off), sizeof(struct in_addr)); 1525 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1526 /* 1527 * Let ip_intr's mcast routing check handle mcast pkts 1528 */ 1529 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1530 break; 1531 1532 case IPOPT_RR: 1533 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1534 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1535 goto bad; 1536 } 1537 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1538 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1539 goto bad; 1540 } 1541 /* 1542 * If no space remains, ignore. 1543 */ 1544 off--; /* 0 origin */ 1545 if ((off + sizeof(struct in_addr)) > optlen) 1546 break; 1547 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1548 sizeof(ipaddr.sin_addr)); 1549 /* 1550 * locate outgoing interface; if we're the destination, 1551 * use the incoming interface (should be same). 1552 */ 1553 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1554 == NULL && 1555 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1556 type = ICMP_UNREACH; 1557 code = ICMP_UNREACH_HOST; 1558 goto bad; 1559 } 1560 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1561 (caddr_t)(cp + off), sizeof(struct in_addr)); 1562 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1563 break; 1564 1565 case IPOPT_TS: 1566 code = cp - (u_char *)ip; 1567 ipt = (struct ip_timestamp *)cp; 1568 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1569 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1570 goto bad; 1571 } 1572 if (ipt->ipt_ptr < 5) { 1573 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1574 goto bad; 1575 } 1576 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1577 if (++ipt->ipt_oflw == 0) { 1578 code = (u_char *)&ipt->ipt_ptr - 1579 (u_char *)ip; 1580 goto bad; 1581 } 1582 break; 1583 } 1584 cp0 = (cp + ipt->ipt_ptr - 1); 1585 switch (ipt->ipt_flg) { 1586 1587 case IPOPT_TS_TSONLY: 1588 break; 1589 1590 case IPOPT_TS_TSANDADDR: 1591 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1592 sizeof(struct in_addr) > ipt->ipt_len) { 1593 code = (u_char *)&ipt->ipt_ptr - 1594 (u_char *)ip; 1595 goto bad; 1596 } 1597 ipaddr.sin_addr = dst; 1598 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1599 m->m_pkthdr.rcvif)); 1600 if (ia == 0) 1601 continue; 1602 bcopy(&ia->ia_addr.sin_addr, 1603 cp0, sizeof(struct in_addr)); 1604 ipt->ipt_ptr += sizeof(struct in_addr); 1605 break; 1606 1607 case IPOPT_TS_PRESPEC: 1608 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1609 sizeof(struct in_addr) > ipt->ipt_len) { 1610 code = (u_char *)&ipt->ipt_ptr - 1611 (u_char *)ip; 1612 goto bad; 1613 } 1614 bcopy(cp0, &ipaddr.sin_addr, 1615 sizeof(struct in_addr)); 1616 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1617 == NULL) 1618 continue; 1619 ipt->ipt_ptr += sizeof(struct in_addr); 1620 break; 1621 1622 default: 1623 /* XXX can't take &ipt->ipt_flg */ 1624 code = (u_char *)&ipt->ipt_ptr - 1625 (u_char *)ip + 1; 1626 goto bad; 1627 } 1628 ntime = iptime(); 1629 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1630 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1631 sizeof(n_time)); 1632 ipt->ipt_ptr += sizeof(n_time); 1633 } 1634 } 1635 if (forward) { 1636 if (ip_forwsrcrt == 0) { 1637 type = ICMP_UNREACH; 1638 code = ICMP_UNREACH_SRCFAIL; 1639 goto bad; 1640 } 1641 ip_forward(m, 1); 1642 return (1); 1643 } 1644 return (0); 1645bad: 1646 icmp_error(m, type, code, 0, 0); 1647 ipstat.ips_badoptions++; 1648 return (1); 1649} 1650 1651/* 1652 * Given address of next destination (final or next hop), 1653 * return internet address info of interface to be used to get there. 1654 */ 1655struct in_ifaddr * 1656ip_rtaddr(dst) 1657 struct in_addr dst; 1658{ 1659 struct sockaddr_in *sin; 1660 1661 sin = satosin(&ipforward_rt.ro_dst); 1662 1663 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1664 if (ipforward_rt.ro_rt) { 1665 RTFREE(ipforward_rt.ro_rt); 1666 ipforward_rt.ro_rt = 0; 1667 } 1668 sin->sin_family = AF_INET; 1669 sin->sin_len = sizeof(*sin); 1670 sin->sin_addr = dst; 1671 1672 rtalloc(&ipforward_rt); 1673 } 1674 if (ipforward_rt.ro_rt == 0) 1675 return ((struct in_ifaddr *)0); 1676 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1677} 1678 1679/* 1680 * Save incoming source route for use in replies, 1681 * to be picked up later by ip_srcroute if the receiver is interested. 1682 */ 1683void 1684save_rte(option, dst) 1685 u_char *option; 1686 struct in_addr dst; 1687{ 1688 unsigned olen; 1689 1690 olen = option[IPOPT_OLEN]; 1691#ifdef DIAGNOSTIC 1692 if (ipprintfs) 1693 printf("save_rte: olen %d\n", olen); 1694#endif /* 0 */ 1695 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1696 return; 1697 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1698 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1699 ip_srcrt.dst = dst; 1700} 1701 1702/* 1703 * Retrieve incoming source route for use in replies, 1704 * in the same form used by setsockopt. 1705 * The first hop is placed before the options, will be removed later. 1706 */ 1707struct mbuf * 1708ip_srcroute() 1709{ 1710 struct in_addr *p, *q; 1711 struct mbuf *m; 1712 1713 if (ip_nhops == 0) 1714 return ((struct mbuf *)0); 1715 m = m_get(M_DONTWAIT, MT_SOOPTS); 1716 if (m == 0) 1717 return ((struct mbuf *)0); 1718 1719 MCLAIM(m, &inetdomain.dom_mowner); 1720#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1721 1722 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1723 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1724 OPTSIZ; 1725#ifdef DIAGNOSTIC 1726 if (ipprintfs) 1727 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1728#endif 1729 1730 /* 1731 * First save first hop for return route 1732 */ 1733 p = &ip_srcrt.route[ip_nhops - 1]; 1734 *(mtod(m, struct in_addr *)) = *p--; 1735#ifdef DIAGNOSTIC 1736 if (ipprintfs) 1737 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1738#endif 1739 1740 /* 1741 * Copy option fields and padding (nop) to mbuf. 1742 */ 1743 ip_srcrt.nop = IPOPT_NOP; 1744 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1745 bcopy((caddr_t)&ip_srcrt.nop, 1746 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1747 q = (struct in_addr *)(mtod(m, caddr_t) + 1748 sizeof(struct in_addr) + OPTSIZ); 1749#undef OPTSIZ 1750 /* 1751 * Record return path as an IP source route, 1752 * reversing the path (pointers are now aligned). 1753 */ 1754 while (p >= ip_srcrt.route) { 1755#ifdef DIAGNOSTIC 1756 if (ipprintfs) 1757 printf(" %x", ntohl(q->s_addr)); 1758#endif 1759 *q++ = *p--; 1760 } 1761 /* 1762 * Last hop goes to final destination. 1763 */ 1764 *q = ip_srcrt.dst; 1765#ifdef DIAGNOSTIC 1766 if (ipprintfs) 1767 printf(" %x\n", ntohl(q->s_addr)); 1768#endif 1769 return (m); 1770} 1771 1772/* 1773 * Strip out IP options, at higher 1774 * level protocol in the kernel. 1775 * Second argument is buffer to which options 1776 * will be moved, and return value is their length. 1777 * XXX should be deleted; last arg currently ignored. 1778 */ 1779void 1780ip_stripoptions(m, mopt) 1781 struct mbuf *m; 1782 struct mbuf *mopt; 1783{ 1784 int i; 1785 struct ip *ip = mtod(m, struct ip *); 1786 caddr_t opts; 1787 int olen; 1788 1789 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1790 opts = (caddr_t)(ip + 1); 1791 i = m->m_len - (sizeof (struct ip) + olen); 1792 bcopy(opts + olen, opts, (unsigned)i); 1793 m->m_len -= olen; 1794 if (m->m_flags & M_PKTHDR) 1795 m->m_pkthdr.len -= olen; 1796 ip->ip_len = htons(ntohs(ip->ip_len) - olen); 1797 ip->ip_hl = sizeof (struct ip) >> 2; 1798} 1799 1800const int inetctlerrmap[PRC_NCMDS] = { 1801 0, 0, 0, 0, 1802 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1803 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1804 EMSGSIZE, EHOSTUNREACH, 0, 0, 1805 0, 0, 0, 0, 1806 ENOPROTOOPT 1807}; 1808 1809/* 1810 * Forward a packet. If some error occurs return the sender 1811 * an icmp packet. Note we can't always generate a meaningful 1812 * icmp message because icmp doesn't have a large enough repertoire 1813 * of codes and types. 1814 * 1815 * If not forwarding, just drop the packet. This could be confusing 1816 * if ipforwarding was zero but some routing protocol was advancing 1817 * us as a gateway to somewhere. However, we must let the routing 1818 * protocol deal with that. 1819 * 1820 * The srcrt parameter indicates whether the packet is being forwarded 1821 * via a source route. 1822 */ 1823void 1824ip_forward(m, srcrt) 1825 struct mbuf *m; 1826 int srcrt; 1827{ 1828 struct ip *ip = mtod(m, struct ip *); 1829 struct sockaddr_in *sin; 1830 struct rtentry *rt; 1831 int error, type = 0, code = 0; 1832 struct mbuf *mcopy; 1833 n_long dest; 1834 struct ifnet *destifp; 1835#if defined(IPSEC) || defined(FAST_IPSEC) 1836 struct ifnet dummyifp; 1837#endif 1838 1839 /* 1840 * We are now in the output path. 1841 */ 1842 MCLAIM(m, &ip_tx_mowner); 1843 1844 /* 1845 * Clear any in-bound checksum flags for this packet. 1846 */ 1847 m->m_pkthdr.csum_flags = 0; 1848 1849 dest = 0; 1850#ifdef DIAGNOSTIC 1851 if (ipprintfs) 1852 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1853 ntohl(ip->ip_src.s_addr), 1854 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1855#endif 1856 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1857 ipstat.ips_cantforward++; 1858 m_freem(m); 1859 return; 1860 } 1861 if (ip->ip_ttl <= IPTTLDEC) { 1862 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1863 return; 1864 } 1865 ip->ip_ttl -= IPTTLDEC; 1866 1867 sin = satosin(&ipforward_rt.ro_dst); 1868 if ((rt = ipforward_rt.ro_rt) == 0 || 1869 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1870 if (ipforward_rt.ro_rt) { 1871 RTFREE(ipforward_rt.ro_rt); 1872 ipforward_rt.ro_rt = 0; 1873 } 1874 sin->sin_family = AF_INET; 1875 sin->sin_len = sizeof(struct sockaddr_in); 1876 sin->sin_addr = ip->ip_dst; 1877 1878 rtalloc(&ipforward_rt); 1879 if (ipforward_rt.ro_rt == 0) { 1880 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1881 return; 1882 } 1883 rt = ipforward_rt.ro_rt; 1884 } 1885 1886 /* 1887 * Save at most 68 bytes of the packet in case 1888 * we need to generate an ICMP message to the src. 1889 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1890 */ 1891 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1892 if (mcopy) 1893 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1894 1895 /* 1896 * If forwarding packet using same interface that it came in on, 1897 * perhaps should send a redirect to sender to shortcut a hop. 1898 * Only send redirect if source is sending directly to us, 1899 * and if packet was not source routed (or has any options). 1900 * Also, don't send redirect if forwarding using a default route 1901 * or a route modified by a redirect. 1902 */ 1903 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1904 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1905 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1906 ipsendredirects && !srcrt) { 1907 if (rt->rt_ifa && 1908 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1909 ifatoia(rt->rt_ifa)->ia_subnet) { 1910 if (rt->rt_flags & RTF_GATEWAY) 1911 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1912 else 1913 dest = ip->ip_dst.s_addr; 1914 /* 1915 * Router requirements says to only send host 1916 * redirects. 1917 */ 1918 type = ICMP_REDIRECT; 1919 code = ICMP_REDIRECT_HOST; 1920#ifdef DIAGNOSTIC 1921 if (ipprintfs) 1922 printf("redirect (%d) to %x\n", code, 1923 (u_int32_t)dest); 1924#endif 1925 } 1926 } 1927 1928 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1929 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1930 (struct ip_moptions *)NULL, (struct socket *)NULL); 1931 1932 if (error) 1933 ipstat.ips_cantforward++; 1934 else { 1935 ipstat.ips_forward++; 1936 if (type) 1937 ipstat.ips_redirectsent++; 1938 else { 1939 if (mcopy) { 1940#ifdef GATEWAY 1941 if (mcopy->m_flags & M_CANFASTFWD) 1942 ipflow_create(&ipforward_rt, mcopy); 1943#endif 1944 m_freem(mcopy); 1945 } 1946 return; 1947 } 1948 } 1949 if (mcopy == NULL) 1950 return; 1951 destifp = NULL; 1952 1953 switch (error) { 1954 1955 case 0: /* forwarded, but need redirect */ 1956 /* type, code set above */ 1957 break; 1958 1959 case ENETUNREACH: /* shouldn't happen, checked above */ 1960 case EHOSTUNREACH: 1961 case ENETDOWN: 1962 case EHOSTDOWN: 1963 default: 1964 type = ICMP_UNREACH; 1965 code = ICMP_UNREACH_HOST; 1966 break; 1967 1968 case EMSGSIZE: 1969 type = ICMP_UNREACH; 1970 code = ICMP_UNREACH_NEEDFRAG; 1971#if !defined(IPSEC) && !defined(FAST_IPSEC) 1972 if (ipforward_rt.ro_rt) 1973 destifp = ipforward_rt.ro_rt->rt_ifp; 1974#else 1975 /* 1976 * If the packet is routed over IPsec tunnel, tell the 1977 * originator the tunnel MTU. 1978 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1979 * XXX quickhack!!! 1980 */ 1981 if (ipforward_rt.ro_rt) { 1982 struct secpolicy *sp; 1983 int ipsecerror; 1984 size_t ipsechdr; 1985 struct route *ro; 1986 1987 sp = ipsec4_getpolicybyaddr(mcopy, 1988 IPSEC_DIR_OUTBOUND, IP_FORWARDING, 1989 &ipsecerror); 1990 1991 if (sp == NULL) 1992 destifp = ipforward_rt.ro_rt->rt_ifp; 1993 else { 1994 /* count IPsec header size */ 1995 ipsechdr = ipsec4_hdrsiz(mcopy, 1996 IPSEC_DIR_OUTBOUND, NULL); 1997 1998 /* 1999 * find the correct route for outer IPv4 2000 * header, compute tunnel MTU. 2001 * 2002 * XXX BUG ALERT 2003 * The "dummyifp" code relies upon the fact 2004 * that icmp_error() touches only ifp->if_mtu. 2005 */ 2006 /*XXX*/ 2007 destifp = NULL; 2008 if (sp->req != NULL 2009 && sp->req->sav != NULL 2010 && sp->req->sav->sah != NULL) { 2011 ro = &sp->req->sav->sah->sa_route; 2012 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 2013 dummyifp.if_mtu = 2014 ro->ro_rt->rt_rmx.rmx_mtu ? 2015 ro->ro_rt->rt_rmx.rmx_mtu : 2016 ro->ro_rt->rt_ifp->if_mtu; 2017 dummyifp.if_mtu -= ipsechdr; 2018 destifp = &dummyifp; 2019 } 2020 } 2021 2022#ifdef IPSEC 2023 key_freesp(sp); 2024#else 2025 KEY_FREESP(&sp); 2026#endif 2027 } 2028 } 2029#endif /*IPSEC*/ 2030 ipstat.ips_cantfrag++; 2031 break; 2032 2033 case ENOBUFS: 2034#if 1 2035 /* 2036 * a router should not generate ICMP_SOURCEQUENCH as 2037 * required in RFC1812 Requirements for IP Version 4 Routers. 2038 * source quench could be a big problem under DoS attacks, 2039 * or if the underlying interface is rate-limited. 2040 */ 2041 if (mcopy) 2042 m_freem(mcopy); 2043 return; 2044#else 2045 type = ICMP_SOURCEQUENCH; 2046 code = 0; 2047 break; 2048#endif 2049 } 2050 icmp_error(mcopy, type, code, dest, destifp); 2051} 2052 2053void 2054ip_savecontrol(inp, mp, ip, m) 2055 struct inpcb *inp; 2056 struct mbuf **mp; 2057 struct ip *ip; 2058 struct mbuf *m; 2059{ 2060 2061 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2062 struct timeval tv; 2063 2064 microtime(&tv); 2065 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2066 SCM_TIMESTAMP, SOL_SOCKET); 2067 if (*mp) 2068 mp = &(*mp)->m_next; 2069 } 2070 if (inp->inp_flags & INP_RECVDSTADDR) { 2071 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2072 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2073 if (*mp) 2074 mp = &(*mp)->m_next; 2075 } 2076#ifdef notyet 2077 /* 2078 * XXX 2079 * Moving these out of udp_input() made them even more broken 2080 * than they already were. 2081 * - fenner@parc.xerox.com 2082 */ 2083 /* options were tossed already */ 2084 if (inp->inp_flags & INP_RECVOPTS) { 2085 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2086 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2087 if (*mp) 2088 mp = &(*mp)->m_next; 2089 } 2090 /* ip_srcroute doesn't do what we want here, need to fix */ 2091 if (inp->inp_flags & INP_RECVRETOPTS) { 2092 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 2093 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2094 if (*mp) 2095 mp = &(*mp)->m_next; 2096 } 2097#endif 2098 if (inp->inp_flags & INP_RECVIF) { 2099 struct sockaddr_dl sdl; 2100 2101 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 2102 sdl.sdl_family = AF_LINK; 2103 sdl.sdl_index = m->m_pkthdr.rcvif ? 2104 m->m_pkthdr.rcvif->if_index : 0; 2105 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 2106 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 2107 IP_RECVIF, IPPROTO_IP); 2108 if (*mp) 2109 mp = &(*mp)->m_next; 2110 } 2111} 2112 2113/* 2114 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 2115 * range of the new value and tweaks timers if it changes. 2116 */ 2117static int 2118sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 2119{ 2120 int error, tmp; 2121 struct sysctlnode node; 2122 2123 node = *rnode; 2124 tmp = ip_mtudisc_timeout; 2125 node.sysctl_data = &tmp; 2126 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2127 if (error || newp == NULL) 2128 return (error); 2129 if (tmp < 0) 2130 return (EINVAL); 2131 2132 ip_mtudisc_timeout = tmp; 2133 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 2134 2135 return (0); 2136} 2137 2138#ifdef GATEWAY 2139/* 2140 * sysctl helper routine for net.inet.ip.maxflows. apparently if 2141 * maxflows is even looked up, we "reap flows". 2142 */ 2143static int 2144sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS) 2145{ 2146 int s; 2147 2148 s = sysctl_lookup(SYSCTLFN_CALL(rnode)); 2149 if (s) 2150 return (s); 2151 2152 s = splsoftnet(); 2153 ipflow_reap(0); 2154 splx(s); 2155 2156 return (0); 2157} 2158#endif /* GATEWAY */ 2159 2160 2161SYSCTL_SETUP(sysctl_net_inet_ip_setup, "sysctl net.inet.ip subtree setup") 2162{ 2163 extern int subnetsarelocal, hostzeroisbroadcast; 2164 2165 sysctl_createv(clog, 0, NULL, NULL, 2166 CTLFLAG_PERMANENT, 2167 CTLTYPE_NODE, "net", NULL, 2168 NULL, 0, NULL, 0, 2169 CTL_NET, CTL_EOL); 2170 sysctl_createv(clog, 0, NULL, NULL, 2171 CTLFLAG_PERMANENT, 2172 CTLTYPE_NODE, "inet", NULL, 2173 NULL, 0, NULL, 0, 2174 CTL_NET, PF_INET, CTL_EOL); 2175 sysctl_createv(clog, 0, NULL, NULL, 2176 CTLFLAG_PERMANENT, 2177 CTLTYPE_NODE, "ip", NULL, 2178 NULL, 0, NULL, 0, 2179 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 2180 2181 sysctl_createv(clog, 0, NULL, NULL, 2182 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2183 CTLTYPE_INT, "forwarding", NULL, 2184 NULL, 0, &ipforwarding, 0, 2185 CTL_NET, PF_INET, IPPROTO_IP, 2186 IPCTL_FORWARDING, CTL_EOL); 2187 sysctl_createv(clog, 0, NULL, NULL, 2188 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2189 CTLTYPE_INT, "redirect", NULL, 2190 NULL, 0, &ipsendredirects, 0, 2191 CTL_NET, PF_INET, IPPROTO_IP, 2192 IPCTL_SENDREDIRECTS, CTL_EOL); 2193 sysctl_createv(clog, 0, NULL, NULL, 2194 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2195 CTLTYPE_INT, "ttl", NULL, 2196 NULL, 0, &ip_defttl, 0, 2197 CTL_NET, PF_INET, IPPROTO_IP, 2198 IPCTL_DEFTTL, CTL_EOL); 2199#ifdef IPCTL_DEFMTU 2200 sysctl_createv(clog, 0, NULL, NULL, 2201 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 2202 CTLTYPE_INT, "mtu", NULL, 2203 NULL, 0, &ip_mtu, 0, 2204 CTL_NET, PF_INET, IPPROTO_IP, 2205 IPCTL_DEFMTU, CTL_EOL); 2206#endif /* IPCTL_DEFMTU */ 2207 sysctl_createv(clog, 0, NULL, NULL, 2208 CTLFLAG_PERMANENT|CTLFLAG_READONLY1, 2209 CTLTYPE_INT, "forwsrcrt", NULL, 2210 NULL, 0, &ip_forwsrcrt, 0, 2211 CTL_NET, PF_INET, IPPROTO_IP, 2212 IPCTL_FORWSRCRT, CTL_EOL); 2213 sysctl_createv(clog, 0, NULL, NULL, 2214 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2215 CTLTYPE_INT, "directed-broadcast", NULL, 2216 NULL, 0, &ip_directedbcast, 0, 2217 CTL_NET, PF_INET, IPPROTO_IP, 2218 IPCTL_DIRECTEDBCAST, CTL_EOL); 2219 sysctl_createv(clog, 0, NULL, NULL, 2220 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2221 CTLTYPE_INT, "allowsrcrt", NULL, 2222 NULL, 0, &ip_allowsrcrt, 0, 2223 CTL_NET, PF_INET, IPPROTO_IP, 2224 IPCTL_ALLOWSRCRT, CTL_EOL); 2225 sysctl_createv(clog, 0, NULL, NULL, 2226 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2227 CTLTYPE_INT, "subnetsarelocal", NULL, 2228 NULL, 0, &subnetsarelocal, 0, 2229 CTL_NET, PF_INET, IPPROTO_IP, 2230 IPCTL_SUBNETSARELOCAL, CTL_EOL); 2231 sysctl_createv(clog, 0, NULL, NULL, 2232 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2233 CTLTYPE_INT, "mtudisc", NULL, 2234 NULL, 0, &ip_mtudisc, 0, 2235 CTL_NET, PF_INET, IPPROTO_IP, 2236 IPCTL_MTUDISC, CTL_EOL); 2237 sysctl_createv(clog, 0, NULL, NULL, 2238 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2239 CTLTYPE_INT, "anonportmin", NULL, 2240 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 2241 CTL_NET, PF_INET, IPPROTO_IP, 2242 IPCTL_ANONPORTMIN, CTL_EOL); 2243 sysctl_createv(clog, 0, NULL, NULL, 2244 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2245 CTLTYPE_INT, "anonportmax", NULL, 2246 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 2247 CTL_NET, PF_INET, IPPROTO_IP, 2248 IPCTL_ANONPORTMAX, CTL_EOL); 2249 sysctl_createv(clog, 0, NULL, NULL, 2250 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2251 CTLTYPE_INT, "mtudisctimeout", NULL, 2252 sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0, 2253 CTL_NET, PF_INET, IPPROTO_IP, 2254 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 2255#ifdef GATEWAY 2256 sysctl_createv(clog, 0, NULL, NULL, 2257 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2258 CTLTYPE_INT, "maxflows", NULL, 2259 sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0, 2260 CTL_NET, PF_INET, IPPROTO_IP, 2261 IPCTL_MAXFLOWS, CTL_EOL); 2262#endif /* GATEWAY */ 2263 sysctl_createv(clog, 0, NULL, NULL, 2264 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2265 CTLTYPE_INT, "hostzerobroadcast", NULL, 2266 NULL, 0, &hostzeroisbroadcast, 0, 2267 CTL_NET, PF_INET, IPPROTO_IP, 2268 IPCTL_HOSTZEROBROADCAST, CTL_EOL); 2269#if NGIF > 0 2270 sysctl_createv(clog, 0, NULL, NULL, 2271 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2272 CTLTYPE_INT, "gifttl", NULL, 2273 NULL, 0, &ip_gif_ttl, 0, 2274 CTL_NET, PF_INET, IPPROTO_IP, 2275 IPCTL_GIF_TTL, CTL_EOL); 2276#endif /* NGIF */ 2277#ifndef IPNOPRIVPORTS 2278 sysctl_createv(clog, 0, NULL, NULL, 2279 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2280 CTLTYPE_INT, "lowportmin", NULL, 2281 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 2282 CTL_NET, PF_INET, IPPROTO_IP, 2283 IPCTL_LOWPORTMIN, CTL_EOL); 2284 sysctl_createv(clog, 0, NULL, NULL, 2285 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2286 CTLTYPE_INT, "lowportmax", NULL, 2287 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 2288 CTL_NET, PF_INET, IPPROTO_IP, 2289 IPCTL_LOWPORTMAX, CTL_EOL); 2290#endif /* IPNOPRIVPORTS */ 2291 sysctl_createv(clog, 0, NULL, NULL, 2292 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2293 CTLTYPE_INT, "maxfragpackets", NULL, 2294 NULL, 0, &ip_maxfragpackets, 0, 2295 CTL_NET, PF_INET, IPPROTO_IP, 2296 IPCTL_MAXFRAGPACKETS, CTL_EOL); 2297#if NGRE > 0 2298 sysctl_createv(clog, 0, NULL, NULL, 2299 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2300 CTLTYPE_INT, "grettl", NULL, 2301 NULL, 0, &ip_gre_ttl, 0, 2302 CTL_NET, PF_INET, IPPROTO_IP, 2303 IPCTL_GRE_TTL, CTL_EOL); 2304#endif /* NGRE */ 2305 sysctl_createv(clog, 0, NULL, NULL, 2306 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2307 CTLTYPE_INT, "checkinterface", NULL, 2308 NULL, 0, &ip_checkinterface, 0, 2309 CTL_NET, PF_INET, IPPROTO_IP, 2310 IPCTL_CHECKINTERFACE, CTL_EOL); 2311 sysctl_createv(clog, 0, NULL, NULL, 2312 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2313 CTLTYPE_INT, "random_id", NULL, 2314 NULL, 0, &ip_do_randomid, 0, 2315 CTL_NET, PF_INET, IPPROTO_IP, 2316 IPCTL_RANDOMID, CTL_EOL); 2317} 2318