ip_input.c revision 1.253
1/* $NetBSD: ip_input.c,v 1.253 2007/09/11 14:18:09 degroote 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 1.253 2007/09/11 14:18:09 degroote 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#include <sys/kauth.h> 125 126#include <net/if.h> 127#include <net/if_dl.h> 128#include <net/route.h> 129#include <net/pfil.h> 130 131#include <netinet/in.h> 132#include <netinet/in_systm.h> 133#include <netinet/ip.h> 134#include <netinet/in_pcb.h> 135#include <netinet/in_proto.h> 136#include <netinet/in_var.h> 137#include <netinet/ip_var.h> 138#include <netinet/ip_icmp.h> 139/* just for gif_ttl */ 140#include <netinet/in_gif.h> 141#include "gif.h" 142#include <net/if_gre.h> 143#include "gre.h" 144 145#ifdef MROUTING 146#include <netinet/ip_mroute.h> 147#endif 148 149#ifdef IPSEC 150#include <netinet6/ipsec.h> 151#include <netkey/key.h> 152#endif 153#ifdef FAST_IPSEC 154#include <netipsec/ipsec.h> 155#include <netipsec/key.h> 156#endif /* FAST_IPSEC*/ 157 158#ifndef IPFORWARDING 159#ifdef GATEWAY 160#define IPFORWARDING 1 /* forward IP packets not for us */ 161#else /* GATEWAY */ 162#define IPFORWARDING 0 /* don't forward IP packets not for us */ 163#endif /* GATEWAY */ 164#endif /* IPFORWARDING */ 165#ifndef IPSENDREDIRECTS 166#define IPSENDREDIRECTS 1 167#endif 168#ifndef IPFORWSRCRT 169#define IPFORWSRCRT 1 /* forward source-routed packets */ 170#endif 171#ifndef IPALLOWSRCRT 172#define IPALLOWSRCRT 1 /* allow source-routed packets */ 173#endif 174#ifndef IPMTUDISC 175#define IPMTUDISC 1 176#endif 177#ifndef IPMTUDISCTIMEOUT 178#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 179#endif 180 181/* 182 * Note: DIRECTED_BROADCAST is handled this way so that previous 183 * configuration using this option will Just Work. 184 */ 185#ifndef IPDIRECTEDBCAST 186#ifdef DIRECTED_BROADCAST 187#define IPDIRECTEDBCAST 1 188#else 189#define IPDIRECTEDBCAST 0 190#endif /* DIRECTED_BROADCAST */ 191#endif /* IPDIRECTEDBCAST */ 192int ipforwarding = IPFORWARDING; 193int ipsendredirects = IPSENDREDIRECTS; 194int ip_defttl = IPDEFTTL; 195int ip_forwsrcrt = IPFORWSRCRT; 196int ip_directedbcast = IPDIRECTEDBCAST; 197int ip_allowsrcrt = IPALLOWSRCRT; 198int ip_mtudisc = IPMTUDISC; 199int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 200#ifdef DIAGNOSTIC 201int ipprintfs = 0; 202#endif 203 204int ip_do_randomid = 0; 205 206/* 207 * XXX - Setting ip_checkinterface mostly implements the receive side of 208 * the Strong ES model described in RFC 1122, but since the routing table 209 * and transmit implementation do not implement the Strong ES model, 210 * setting this to 1 results in an odd hybrid. 211 * 212 * XXX - ip_checkinterface currently must be disabled if you use ipnat 213 * to translate the destination address to another local interface. 214 * 215 * XXX - ip_checkinterface must be disabled if you add IP aliases 216 * to the loopback interface instead of the interface where the 217 * packets for those addresses are received. 218 */ 219int ip_checkinterface = 0; 220 221 222struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 223 224int ipqmaxlen = IFQ_MAXLEN; 225u_long in_ifaddrhash; /* size of hash table - 1 */ 226int in_ifaddrentries; /* total number of addrs */ 227struct in_ifaddrhead in_ifaddrhead; 228struct in_ifaddrhashhead *in_ifaddrhashtbl; 229u_long in_multihash; /* size of hash table - 1 */ 230int in_multientries; /* total number of addrs */ 231struct in_multihashhead *in_multihashtbl; 232struct ifqueue ipintrq; 233struct ipstat ipstat; 234uint16_t ip_id; 235 236#ifdef PFIL_HOOKS 237struct pfil_head inet_pfil_hook; 238#endif 239 240/* 241 * Cached copy of nmbclusters. If nbclusters is different, 242 * recalculate IP parameters derived from nmbclusters. 243 */ 244static int ip_nmbclusters; /* copy of nmbclusters */ 245static void ip_nmbclusters_changed(void); /* recalc limits */ 246 247#define CHECK_NMBCLUSTER_PARAMS() \ 248do { \ 249 if (__predict_false(ip_nmbclusters != nmbclusters)) \ 250 ip_nmbclusters_changed(); \ 251} while (/*CONSTCOND*/0) 252 253/* IP datagram reassembly queues (hashed) */ 254#define IPREASS_NHASH_LOG2 6 255#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 256#define IPREASS_HMASK (IPREASS_NHASH - 1) 257#define IPREASS_HASH(x,y) \ 258 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 259struct ipqhead ipq[IPREASS_NHASH]; 260int ipq_locked; 261static int ip_nfragpackets; /* packets in reass queue */ 262static int ip_nfrags; /* total fragments in reass queues */ 263 264int ip_maxfragpackets = 200; /* limit on packets. XXX sysctl */ 265int ip_maxfrags; /* limit on fragments. XXX sysctl */ 266 267 268/* 269 * Additive-Increase/Multiplicative-Decrease (AIMD) strategy for 270 * IP reassembly queue buffer managment. 271 * 272 * We keep a count of total IP fragments (NB: not fragmented packets!) 273 * awaiting reassembly (ip_nfrags) and a limit (ip_maxfrags) on fragments. 274 * If ip_nfrags exceeds ip_maxfrags the limit, we drop half the 275 * total fragments in reassembly queues.This AIMD policy avoids 276 * repeatedly deleting single packets under heavy fragmentation load 277 * (e.g., from lossy NFS peers). 278 */ 279static u_int ip_reass_ttl_decr(u_int ticks); 280static void ip_reass_drophalf(void); 281 282 283static inline int ipq_lock_try(void); 284static inline void ipq_unlock(void); 285 286static inline int 287ipq_lock_try(void) 288{ 289 int s; 290 291 /* 292 * Use splvm() -- we're blocking things that would cause 293 * mbuf allocation. 294 */ 295 s = splvm(); 296 if (ipq_locked) { 297 splx(s); 298 return (0); 299 } 300 ipq_locked = 1; 301 splx(s); 302 return (1); 303} 304 305static inline void 306ipq_unlock(void) 307{ 308 int s; 309 310 s = splvm(); 311 ipq_locked = 0; 312 splx(s); 313} 314 315#ifdef DIAGNOSTIC 316#define IPQ_LOCK() \ 317do { \ 318 if (ipq_lock_try() == 0) { \ 319 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 320 panic("ipq_lock"); \ 321 } \ 322} while (/*CONSTCOND*/ 0) 323#define IPQ_LOCK_CHECK() \ 324do { \ 325 if (ipq_locked == 0) { \ 326 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 327 panic("ipq lock check"); \ 328 } \ 329} while (/*CONSTCOND*/ 0) 330#else 331#define IPQ_LOCK() (void) ipq_lock_try() 332#define IPQ_LOCK_CHECK() /* nothing */ 333#endif 334 335#define IPQ_UNLOCK() ipq_unlock() 336 337POOL_INIT(inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl", NULL, 338 IPL_SOFTNET); 339POOL_INIT(ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", NULL, 340 IPL_VM); 341 342#ifdef INET_CSUM_COUNTERS 343#include <sys/device.h> 344 345struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 346 NULL, "inet", "hwcsum bad"); 347struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 348 NULL, "inet", "hwcsum ok"); 349struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 350 NULL, "inet", "swcsum"); 351 352#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 353 354EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 355EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 356EVCNT_ATTACH_STATIC(ip_swcsum); 357 358#else 359 360#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 361 362#endif /* INET_CSUM_COUNTERS */ 363 364/* 365 * We need to save the IP options in case a protocol wants to respond 366 * to an incoming packet over the same route if the packet got here 367 * using IP source routing. This allows connection establishment and 368 * maintenance when the remote end is on a network that is not known 369 * to us. 370 */ 371int ip_nhops = 0; 372static struct ip_srcrt { 373 struct in_addr dst; /* final destination */ 374 char nop; /* one NOP to align */ 375 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 376 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 377} ip_srcrt; 378 379static void save_rte(u_char *, struct in_addr); 380 381#ifdef MBUFTRACE 382struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx"); 383struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx"); 384#endif 385 386/* 387 * Compute IP limits derived from the value of nmbclusters. 388 */ 389static void 390ip_nmbclusters_changed(void) 391{ 392 ip_maxfrags = nmbclusters / 4; 393 ip_nmbclusters = nmbclusters; 394} 395 396/* 397 * IP initialization: fill in IP protocol switch table. 398 * All protocols not implemented in kernel go to raw IP protocol handler. 399 */ 400void 401ip_init(void) 402{ 403 const struct protosw *pr; 404 int i; 405 406 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 407 if (pr == 0) 408 panic("ip_init"); 409 for (i = 0; i < IPPROTO_MAX; i++) 410 ip_protox[i] = pr - inetsw; 411 for (pr = inetdomain.dom_protosw; 412 pr < inetdomain.dom_protoswNPROTOSW; pr++) 413 if (pr->pr_domain->dom_family == PF_INET && 414 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 415 ip_protox[pr->pr_protocol] = pr - inetsw; 416 417 for (i = 0; i < IPREASS_NHASH; i++) 418 LIST_INIT(&ipq[i]); 419 420 ip_id = time_second & 0xfffff; 421 422 ipintrq.ifq_maxlen = ipqmaxlen; 423 ip_nmbclusters_changed(); 424 425 TAILQ_INIT(&in_ifaddrhead); 426 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 427 M_WAITOK, &in_ifaddrhash); 428 in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IPMADDR, 429 M_WAITOK, &in_multihash); 430 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 431#ifdef GATEWAY 432 ipflow_init(ip_hashsize); 433#endif 434 435#ifdef PFIL_HOOKS 436 /* Register our Packet Filter hook. */ 437 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 438 inet_pfil_hook.ph_af = AF_INET; 439 i = pfil_head_register(&inet_pfil_hook); 440 if (i != 0) 441 printf("ip_init: WARNING: unable to register pfil hook, " 442 "error %d\n", i); 443#endif /* PFIL_HOOKS */ 444 445#ifdef MBUFTRACE 446 MOWNER_ATTACH(&ip_tx_mowner); 447 MOWNER_ATTACH(&ip_rx_mowner); 448#endif /* MBUFTRACE */ 449} 450 451struct sockaddr_in ipaddr = { 452 .sin_len = sizeof(ipaddr), 453 .sin_family = AF_INET, 454}; 455struct route ipforward_rt; 456 457/* 458 * IP software interrupt routine 459 */ 460void 461ipintr(void) 462{ 463 int s; 464 struct mbuf *m; 465 466 while (!IF_IS_EMPTY(&ipintrq)) { 467 s = splnet(); 468 IF_DEQUEUE(&ipintrq, m); 469 splx(s); 470 if (m == 0) 471 return; 472 MCLAIM(m, &ip_rx_mowner); 473 ip_input(m); 474 } 475} 476 477/* 478 * Ip input routine. Checksum and byte swap header. If fragmented 479 * try to reassemble. Process options. Pass to next level. 480 */ 481void 482ip_input(struct mbuf *m) 483{ 484 struct ip *ip = NULL; 485 struct ipq *fp; 486 struct in_ifaddr *ia; 487 struct ifaddr *ifa; 488 struct ipqent *ipqe; 489 int hlen = 0, mff, len; 490 int downmatch; 491 int checkif; 492 int srcrt = 0; 493 int s; 494 u_int hash; 495#ifdef FAST_IPSEC 496 struct m_tag *mtag; 497 struct tdb_ident *tdbi; 498 struct secpolicy *sp; 499 int error; 500#endif /* FAST_IPSEC */ 501 502 MCLAIM(m, &ip_rx_mowner); 503#ifdef DIAGNOSTIC 504 if ((m->m_flags & M_PKTHDR) == 0) 505 panic("ipintr no HDR"); 506#endif 507 508 /* 509 * If no IP addresses have been set yet but the interfaces 510 * are receiving, can't do anything with incoming packets yet. 511 */ 512 if (TAILQ_FIRST(&in_ifaddrhead) == 0) 513 goto bad; 514 ipstat.ips_total++; 515 /* 516 * If the IP header is not aligned, slurp it up into a new 517 * mbuf with space for link headers, in the event we forward 518 * it. Otherwise, if it is aligned, make sure the entire 519 * base IP header is in the first mbuf of the chain. 520 */ 521 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 522 if ((m = m_copyup(m, sizeof(struct ip), 523 (max_linkhdr + 3) & ~3)) == NULL) { 524 /* XXXJRT new stat, please */ 525 ipstat.ips_toosmall++; 526 return; 527 } 528 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 529 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 530 ipstat.ips_toosmall++; 531 return; 532 } 533 } 534 ip = mtod(m, struct ip *); 535 if (ip->ip_v != IPVERSION) { 536 ipstat.ips_badvers++; 537 goto bad; 538 } 539 hlen = ip->ip_hl << 2; 540 if (hlen < sizeof(struct ip)) { /* minimum header length */ 541 ipstat.ips_badhlen++; 542 goto bad; 543 } 544 if (hlen > m->m_len) { 545 if ((m = m_pullup(m, hlen)) == 0) { 546 ipstat.ips_badhlen++; 547 return; 548 } 549 ip = mtod(m, struct ip *); 550 } 551 552 /* 553 * RFC1122: packets with a multicast source address are 554 * not allowed. 555 */ 556 if (IN_MULTICAST(ip->ip_src.s_addr)) { 557 ipstat.ips_badaddr++; 558 goto bad; 559 } 560 561 /* 127/8 must not appear on wire - RFC1122 */ 562 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 563 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 564 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 565 ipstat.ips_badaddr++; 566 goto bad; 567 } 568 } 569 570 switch (m->m_pkthdr.csum_flags & 571 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) | 572 M_CSUM_IPv4_BAD)) { 573 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 574 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 575 goto badcsum; 576 577 case M_CSUM_IPv4: 578 /* Checksum was okay. */ 579 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 580 break; 581 582 default: 583 /* 584 * Must compute it ourselves. Maybe skip checksum on 585 * loopback interfaces. 586 */ 587 if (__predict_true(!(m->m_pkthdr.rcvif->if_flags & 588 IFF_LOOPBACK) || ip_do_loopback_cksum)) { 589 INET_CSUM_COUNTER_INCR(&ip_swcsum); 590 if (in_cksum(m, hlen) != 0) 591 goto badcsum; 592 } 593 break; 594 } 595 596 /* Retrieve the packet length. */ 597 len = ntohs(ip->ip_len); 598 599 /* 600 * Check for additional length bogosity 601 */ 602 if (len < hlen) { 603 ipstat.ips_badlen++; 604 goto bad; 605 } 606 607 /* 608 * Check that the amount of data in the buffers 609 * is as at least much as the IP header would have us expect. 610 * Trim mbufs if longer than we expect. 611 * Drop packet if shorter than we expect. 612 */ 613 if (m->m_pkthdr.len < len) { 614 ipstat.ips_tooshort++; 615 goto bad; 616 } 617 if (m->m_pkthdr.len > len) { 618 if (m->m_len == m->m_pkthdr.len) { 619 m->m_len = len; 620 m->m_pkthdr.len = len; 621 } else 622 m_adj(m, len - m->m_pkthdr.len); 623 } 624 625#if defined(IPSEC) 626 /* ipflow (IP fast forwarding) is not compatible with IPsec. */ 627 m->m_flags &= ~M_CANFASTFWD; 628#else 629 /* 630 * Assume that we can create a fast-forward IP flow entry 631 * based on this packet. 632 */ 633 m->m_flags |= M_CANFASTFWD; 634#endif 635 636#ifdef PFIL_HOOKS 637 /* 638 * Run through list of hooks for input packets. If there are any 639 * filters which require that additional packets in the flow are 640 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 641 * Note that filters must _never_ set this flag, as another filter 642 * in the list may have previously cleared it. 643 */ 644 /* 645 * let ipfilter look at packet on the wire, 646 * not the decapsulated packet. 647 */ 648#ifdef IPSEC 649 if (!ipsec_getnhist(m)) 650#elif defined(FAST_IPSEC) 651 if (!ipsec_indone(m)) 652#else 653 if (1) 654#endif 655 { 656 struct in_addr odst; 657 658 odst = ip->ip_dst; 659 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 660 PFIL_IN) != 0) 661 return; 662 if (m == NULL) 663 return; 664 ip = mtod(m, struct ip *); 665 hlen = ip->ip_hl << 2; 666 /* 667 * XXX The setting of "srcrt" here is to prevent ip_forward() 668 * from generating ICMP redirects for packets that have 669 * been redirected by a hook back out on to the same LAN that 670 * they came from and is not an indication that the packet 671 * is being inffluenced by source routing options. This 672 * allows things like 673 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 674 * where tlp0 is both on the 1.1.1.0/24 network and is the 675 * default route for hosts on 1.1.1.0/24. Of course this 676 * also requires a "map tlp0 ..." to complete the story. 677 * One might argue whether or not this kind of network config. 678 * should be supported in this manner... 679 */ 680 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 681 } 682#endif /* PFIL_HOOKS */ 683 684#ifdef ALTQ 685 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 686 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 687 /* packet dropped by traffic conditioner */ 688 return; 689 } 690#endif 691 692 /* 693 * Process options and, if not destined for us, 694 * ship it on. ip_dooptions returns 1 when an 695 * error was detected (causing an icmp message 696 * to be sent and the original packet to be freed). 697 */ 698 ip_nhops = 0; /* for source routed packets */ 699 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 700 return; 701 702 /* 703 * Enable a consistency check between the destination address 704 * and the arrival interface for a unicast packet (the RFC 1122 705 * strong ES model) if IP forwarding is disabled and the packet 706 * is not locally generated. 707 * 708 * XXX - Checking also should be disabled if the destination 709 * address is ipnat'ed to a different interface. 710 * 711 * XXX - Checking is incompatible with IP aliases added 712 * to the loopback interface instead of the interface where 713 * the packets are received. 714 * 715 * XXX - We need to add a per ifaddr flag for this so that 716 * we get finer grain control. 717 */ 718 checkif = ip_checkinterface && (ipforwarding == 0) && 719 (m->m_pkthdr.rcvif != NULL) && 720 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0); 721 722 /* 723 * Check our list of addresses, to see if the packet is for us. 724 * 725 * Traditional 4.4BSD did not consult IFF_UP at all. 726 * The behavior here is to treat addresses on !IFF_UP interface 727 * as not mine. 728 */ 729 downmatch = 0; 730 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 731 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 732 if (checkif && ia->ia_ifp != m->m_pkthdr.rcvif) 733 continue; 734 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 735 break; 736 else 737 downmatch++; 738 } 739 } 740 if (ia != NULL) 741 goto ours; 742 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 743 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) { 744 if (ifa->ifa_addr->sa_family != AF_INET) 745 continue; 746 ia = ifatoia(ifa); 747 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 748 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 749 /* 750 * Look for all-0's host part (old broadcast addr), 751 * either for subnet or net. 752 */ 753 ip->ip_dst.s_addr == ia->ia_subnet || 754 ip->ip_dst.s_addr == ia->ia_net) 755 goto ours; 756 /* 757 * An interface with IP address zero accepts 758 * all packets that arrive on that interface. 759 */ 760 if (in_nullhost(ia->ia_addr.sin_addr)) 761 goto ours; 762 } 763 } 764 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 765 struct in_multi *inm; 766#ifdef MROUTING 767 extern struct socket *ip_mrouter; 768 769 if (ip_mrouter) { 770 /* 771 * If we are acting as a multicast router, all 772 * incoming multicast packets are passed to the 773 * kernel-level multicast forwarding function. 774 * The packet is returned (relatively) intact; if 775 * ip_mforward() returns a non-zero value, the packet 776 * must be discarded, else it may be accepted below. 777 * 778 * (The IP ident field is put in the same byte order 779 * as expected when ip_mforward() is called from 780 * ip_output().) 781 */ 782 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 783 ipstat.ips_cantforward++; 784 m_freem(m); 785 return; 786 } 787 788 /* 789 * The process-level routing demon needs to receive 790 * all multicast IGMP packets, whether or not this 791 * host belongs to their destination groups. 792 */ 793 if (ip->ip_p == IPPROTO_IGMP) 794 goto ours; 795 ipstat.ips_forward++; 796 } 797#endif 798 /* 799 * See if we belong to the destination multicast group on the 800 * arrival interface. 801 */ 802 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 803 if (inm == NULL) { 804 ipstat.ips_cantforward++; 805 m_freem(m); 806 return; 807 } 808 goto ours; 809 } 810 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 811 in_nullhost(ip->ip_dst)) 812 goto ours; 813 814 /* 815 * Not for us; forward if possible and desirable. 816 */ 817 if (ipforwarding == 0) { 818 ipstat.ips_cantforward++; 819 m_freem(m); 820 } else { 821 /* 822 * If ip_dst matched any of my address on !IFF_UP interface, 823 * and there's no IFF_UP interface that matches ip_dst, 824 * send icmp unreach. Forwarding it will result in in-kernel 825 * forwarding loop till TTL goes to 0. 826 */ 827 if (downmatch) { 828 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 829 ipstat.ips_cantforward++; 830 return; 831 } 832#ifdef IPSEC 833 if (ipsec4_in_reject(m, NULL)) { 834 ipsecstat.in_polvio++; 835 goto bad; 836 } 837#endif 838#ifdef FAST_IPSEC 839 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 840 s = splsoftnet(); 841 if (mtag != NULL) { 842 tdbi = (struct tdb_ident *)(mtag + 1); 843 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 844 } else { 845 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 846 IP_FORWARDING, &error); 847 } 848 if (sp == NULL) { /* NB: can happen if error */ 849 splx(s); 850 /*XXX error stat???*/ 851 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 852 goto bad; 853 } 854 855 /* 856 * Check security policy against packet attributes. 857 */ 858 error = ipsec_in_reject(sp, m); 859 KEY_FREESP(&sp); 860 splx(s); 861 if (error) { 862 ipstat.ips_cantforward++; 863 goto bad; 864 } 865 866 /* 867 * Peek at the outbound SP for this packet to determine if 868 * it's a Fast Forward candidate. 869 */ 870 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 871 if (mtag != NULL) 872 m->m_flags &= ~M_CANFASTFWD; 873 else { 874 s = splsoftnet(); 875 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, 876 (IP_FORWARDING | 877 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 878 &error, NULL); 879 if (sp != NULL) { 880 m->m_flags &= ~M_CANFASTFWD; 881 KEY_FREESP(&sp); 882 } 883 splx(s); 884 } 885#endif /* FAST_IPSEC */ 886 887 ip_forward(m, srcrt); 888 } 889 return; 890 891ours: 892 /* 893 * If offset or IP_MF are set, must reassemble. 894 * Otherwise, nothing need be done. 895 * (We could look in the reassembly queue to see 896 * if the packet was previously fragmented, 897 * but it's not worth the time; just let them time out.) 898 */ 899 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 900 901 /* 902 * Look for queue of fragments 903 * of this datagram. 904 */ 905 IPQ_LOCK(); 906 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 907 LIST_FOREACH(fp, &ipq[hash], ipq_q) { 908 if (ip->ip_id == fp->ipq_id && 909 in_hosteq(ip->ip_src, fp->ipq_src) && 910 in_hosteq(ip->ip_dst, fp->ipq_dst) && 911 ip->ip_p == fp->ipq_p) 912 goto found; 913 914 } 915 fp = 0; 916found: 917 918 /* 919 * Adjust ip_len to not reflect header, 920 * set ipqe_mff if more fragments are expected, 921 * convert offset of this to bytes. 922 */ 923 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 924 mff = (ip->ip_off & htons(IP_MF)) != 0; 925 if (mff) { 926 /* 927 * Make sure that fragments have a data length 928 * that's a non-zero multiple of 8 bytes. 929 */ 930 if (ntohs(ip->ip_len) == 0 || 931 (ntohs(ip->ip_len) & 0x7) != 0) { 932 ipstat.ips_badfrags++; 933 IPQ_UNLOCK(); 934 goto bad; 935 } 936 } 937 ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3); 938 939 /* 940 * If datagram marked as having more fragments 941 * or if this is not the first fragment, 942 * attempt reassembly; if it succeeds, proceed. 943 */ 944 if (mff || ip->ip_off != htons(0)) { 945 ipstat.ips_fragments++; 946 s = splvm(); 947 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 948 splx(s); 949 if (ipqe == NULL) { 950 ipstat.ips_rcvmemdrop++; 951 IPQ_UNLOCK(); 952 goto bad; 953 } 954 ipqe->ipqe_mff = mff; 955 ipqe->ipqe_m = m; 956 ipqe->ipqe_ip = ip; 957 m = ip_reass(ipqe, fp, &ipq[hash]); 958 if (m == 0) { 959 IPQ_UNLOCK(); 960 return; 961 } 962 ipstat.ips_reassembled++; 963 ip = mtod(m, struct ip *); 964 hlen = ip->ip_hl << 2; 965 ip->ip_len = htons(ntohs(ip->ip_len) + hlen); 966 } else 967 if (fp) 968 ip_freef(fp); 969 IPQ_UNLOCK(); 970 } 971 972#if defined(IPSEC) 973 /* 974 * enforce IPsec policy checking if we are seeing last header. 975 * note that we do not visit this with protocols with pcb layer 976 * code - like udp/tcp/raw ip. 977 */ 978 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 979 ipsec4_in_reject(m, NULL)) { 980 ipsecstat.in_polvio++; 981 goto bad; 982 } 983#endif 984#ifdef FAST_IPSEC 985 /* 986 * enforce IPsec policy checking if we are seeing last header. 987 * note that we do not visit this with protocols with pcb layer 988 * code - like udp/tcp/raw ip. 989 */ 990 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 991 /* 992 * Check if the packet has already had IPsec processing 993 * done. If so, then just pass it along. This tag gets 994 * set during AH, ESP, etc. input handling, before the 995 * packet is returned to the ip input queue for delivery. 996 */ 997 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 998 s = splsoftnet(); 999 if (mtag != NULL) { 1000 tdbi = (struct tdb_ident *)(mtag + 1); 1001 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 1002 } else { 1003 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 1004 IP_FORWARDING, &error); 1005 } 1006 if (sp != NULL) { 1007 /* 1008 * Check security policy against packet attributes. 1009 */ 1010 error = ipsec_in_reject(sp, m); 1011 KEY_FREESP(&sp); 1012 } else { 1013 /* XXX error stat??? */ 1014 error = EINVAL; 1015DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 1016 } 1017 splx(s); 1018 if (error) 1019 goto bad; 1020 } 1021#endif /* FAST_IPSEC */ 1022 1023 /* 1024 * Switch out to protocol's input routine. 1025 */ 1026#if IFA_STATS 1027 if (ia && ip) 1028 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 1029#endif 1030 ipstat.ips_delivered++; 1031 { 1032 int off = hlen, nh = ip->ip_p; 1033 1034 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 1035 return; 1036 } 1037bad: 1038 m_freem(m); 1039 return; 1040 1041badcsum: 1042 ipstat.ips_badsum++; 1043 m_freem(m); 1044} 1045 1046/* 1047 * Take incoming datagram fragment and try to 1048 * reassemble it into whole datagram. If a chain for 1049 * reassembly of this datagram already exists, then it 1050 * is given as fp; otherwise have to make a chain. 1051 */ 1052struct mbuf * 1053ip_reass(struct ipqent *ipqe, struct ipq *fp, struct ipqhead *ipqhead) 1054{ 1055 struct mbuf *m = ipqe->ipqe_m; 1056 struct ipqent *nq, *p, *q; 1057 struct ip *ip; 1058 struct mbuf *t; 1059 int hlen = ipqe->ipqe_ip->ip_hl << 2; 1060 int i, next, s; 1061 1062 IPQ_LOCK_CHECK(); 1063 1064 /* 1065 * Presence of header sizes in mbufs 1066 * would confuse code below. 1067 */ 1068 m->m_data += hlen; 1069 m->m_len -= hlen; 1070 1071#ifdef notyet 1072 /* make sure fragment limit is up-to-date */ 1073 CHECK_NMBCLUSTER_PARAMS(); 1074 1075 /* If we have too many fragments, drop the older half. */ 1076 if (ip_nfrags >= ip_maxfrags) 1077 ip_reass_drophalf(void); 1078#endif 1079 1080 /* 1081 * We are about to add a fragment; increment frag count. 1082 */ 1083 ip_nfrags++; 1084 1085 /* 1086 * If first fragment to arrive, create a reassembly queue. 1087 */ 1088 if (fp == 0) { 1089 /* 1090 * Enforce upper bound on number of fragmented packets 1091 * for which we attempt reassembly; 1092 * If maxfrag is 0, never accept fragments. 1093 * If maxfrag is -1, accept all fragments without limitation. 1094 */ 1095 if (ip_maxfragpackets < 0) 1096 ; 1097 else if (ip_nfragpackets >= ip_maxfragpackets) 1098 goto dropfrag; 1099 ip_nfragpackets++; 1100 MALLOC(fp, struct ipq *, sizeof (struct ipq), 1101 M_FTABLE, M_NOWAIT); 1102 if (fp == NULL) 1103 goto dropfrag; 1104 LIST_INSERT_HEAD(ipqhead, fp, ipq_q); 1105 fp->ipq_nfrags = 1; 1106 fp->ipq_ttl = IPFRAGTTL; 1107 fp->ipq_p = ipqe->ipqe_ip->ip_p; 1108 fp->ipq_id = ipqe->ipqe_ip->ip_id; 1109 TAILQ_INIT(&fp->ipq_fragq); 1110 fp->ipq_src = ipqe->ipqe_ip->ip_src; 1111 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 1112 p = NULL; 1113 goto insert; 1114 } else { 1115 fp->ipq_nfrags++; 1116 } 1117 1118 /* 1119 * Find a segment which begins after this one does. 1120 */ 1121 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1122 p = q, q = TAILQ_NEXT(q, ipqe_q)) 1123 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off)) 1124 break; 1125 1126 /* 1127 * If there is a preceding segment, it may provide some of 1128 * our data already. If so, drop the data from the incoming 1129 * segment. If it provides all of our data, drop us. 1130 */ 1131 if (p != NULL) { 1132 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) - 1133 ntohs(ipqe->ipqe_ip->ip_off); 1134 if (i > 0) { 1135 if (i >= ntohs(ipqe->ipqe_ip->ip_len)) 1136 goto dropfrag; 1137 m_adj(ipqe->ipqe_m, i); 1138 ipqe->ipqe_ip->ip_off = 1139 htons(ntohs(ipqe->ipqe_ip->ip_off) + i); 1140 ipqe->ipqe_ip->ip_len = 1141 htons(ntohs(ipqe->ipqe_ip->ip_len) - i); 1142 } 1143 } 1144 1145 /* 1146 * While we overlap succeeding segments trim them or, 1147 * if they are completely covered, dequeue them. 1148 */ 1149 for (; q != NULL && 1150 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) > 1151 ntohs(q->ipqe_ip->ip_off); q = nq) { 1152 i = (ntohs(ipqe->ipqe_ip->ip_off) + 1153 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off); 1154 if (i < ntohs(q->ipqe_ip->ip_len)) { 1155 q->ipqe_ip->ip_len = 1156 htons(ntohs(q->ipqe_ip->ip_len) - i); 1157 q->ipqe_ip->ip_off = 1158 htons(ntohs(q->ipqe_ip->ip_off) + i); 1159 m_adj(q->ipqe_m, i); 1160 break; 1161 } 1162 nq = TAILQ_NEXT(q, ipqe_q); 1163 m_freem(q->ipqe_m); 1164 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1165 s = splvm(); 1166 pool_put(&ipqent_pool, q); 1167 splx(s); 1168 fp->ipq_nfrags--; 1169 ip_nfrags--; 1170 } 1171 1172insert: 1173 /* 1174 * Stick new segment in its place; 1175 * check for complete reassembly. 1176 */ 1177 if (p == NULL) { 1178 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 1179 } else { 1180 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q); 1181 } 1182 next = 0; 1183 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1184 p = q, q = TAILQ_NEXT(q, ipqe_q)) { 1185 if (ntohs(q->ipqe_ip->ip_off) != next) 1186 return (0); 1187 next += ntohs(q->ipqe_ip->ip_len); 1188 } 1189 if (p->ipqe_mff) 1190 return (0); 1191 1192 /* 1193 * Reassembly is complete. Check for a bogus message size and 1194 * concatenate fragments. 1195 */ 1196 q = TAILQ_FIRST(&fp->ipq_fragq); 1197 ip = q->ipqe_ip; 1198 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 1199 ipstat.ips_toolong++; 1200 ip_freef(fp); 1201 return (0); 1202 } 1203 m = q->ipqe_m; 1204 t = m->m_next; 1205 m->m_next = 0; 1206 m_cat(m, t); 1207 nq = TAILQ_NEXT(q, ipqe_q); 1208 s = splvm(); 1209 pool_put(&ipqent_pool, q); 1210 splx(s); 1211 for (q = nq; q != NULL; q = nq) { 1212 t = q->ipqe_m; 1213 nq = TAILQ_NEXT(q, ipqe_q); 1214 s = splvm(); 1215 pool_put(&ipqent_pool, q); 1216 splx(s); 1217 m_cat(m, t); 1218 } 1219 ip_nfrags -= fp->ipq_nfrags; 1220 1221 /* 1222 * Create header for new ip packet by 1223 * modifying header of first packet; 1224 * dequeue and discard fragment reassembly header. 1225 * Make header visible. 1226 */ 1227 ip->ip_len = htons(next); 1228 ip->ip_src = fp->ipq_src; 1229 ip->ip_dst = fp->ipq_dst; 1230 LIST_REMOVE(fp, ipq_q); 1231 FREE(fp, M_FTABLE); 1232 ip_nfragpackets--; 1233 m->m_len += (ip->ip_hl << 2); 1234 m->m_data -= (ip->ip_hl << 2); 1235 /* some debugging cruft by sklower, below, will go away soon */ 1236 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1237 int plen = 0; 1238 for (t = m; t; t = t->m_next) 1239 plen += t->m_len; 1240 m->m_pkthdr.len = plen; 1241 m->m_pkthdr.csum_flags = 0; 1242 } 1243 return (m); 1244 1245dropfrag: 1246 if (fp != 0) 1247 fp->ipq_nfrags--; 1248 ip_nfrags--; 1249 ipstat.ips_fragdropped++; 1250 m_freem(m); 1251 s = splvm(); 1252 pool_put(&ipqent_pool, ipqe); 1253 splx(s); 1254 return (0); 1255} 1256 1257/* 1258 * Free a fragment reassembly header and all 1259 * associated datagrams. 1260 */ 1261void 1262ip_freef(struct ipq *fp) 1263{ 1264 struct ipqent *q, *p; 1265 u_int nfrags = 0; 1266 int s; 1267 1268 IPQ_LOCK_CHECK(); 1269 1270 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) { 1271 p = TAILQ_NEXT(q, ipqe_q); 1272 m_freem(q->ipqe_m); 1273 nfrags++; 1274 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1275 s = splvm(); 1276 pool_put(&ipqent_pool, q); 1277 splx(s); 1278 } 1279 1280 if (nfrags != fp->ipq_nfrags) 1281 printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags); 1282 ip_nfrags -= nfrags; 1283 LIST_REMOVE(fp, ipq_q); 1284 FREE(fp, M_FTABLE); 1285 ip_nfragpackets--; 1286} 1287 1288/* 1289 * IP reassembly TTL machinery for multiplicative drop. 1290 */ 1291static u_int fragttl_histo[(IPFRAGTTL+1)]; 1292 1293 1294/* 1295 * Decrement TTL of all reasembly queue entries by `ticks'. 1296 * Count number of distinct fragments (as opposed to partial, fragmented 1297 * datagrams) in the reassembly queue. While we traverse the entire 1298 * reassembly queue, compute and return the median TTL over all fragments. 1299 */ 1300static u_int 1301ip_reass_ttl_decr(u_int ticks) 1302{ 1303 u_int nfrags, median, dropfraction, keepfraction; 1304 struct ipq *fp, *nfp; 1305 int i; 1306 1307 nfrags = 0; 1308 memset(fragttl_histo, 0, sizeof fragttl_histo); 1309 1310 for (i = 0; i < IPREASS_NHASH; i++) { 1311 for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) { 1312 fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ? 1313 0 : fp->ipq_ttl - ticks); 1314 nfp = LIST_NEXT(fp, ipq_q); 1315 if (fp->ipq_ttl == 0) { 1316 ipstat.ips_fragtimeout++; 1317 ip_freef(fp); 1318 } else { 1319 nfrags += fp->ipq_nfrags; 1320 fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags; 1321 } 1322 } 1323 } 1324 1325 KASSERT(ip_nfrags == nfrags); 1326 1327 /* Find median (or other drop fraction) in histogram. */ 1328 dropfraction = (ip_nfrags / 2); 1329 keepfraction = ip_nfrags - dropfraction; 1330 for (i = IPFRAGTTL, median = 0; i >= 0; i--) { 1331 median += fragttl_histo[i]; 1332 if (median >= keepfraction) 1333 break; 1334 } 1335 1336 /* Return TTL of median (or other fraction). */ 1337 return (u_int)i; 1338} 1339 1340void 1341ip_reass_drophalf(void) 1342{ 1343 1344 u_int median_ticks; 1345 /* 1346 * Compute median TTL of all fragments, and count frags 1347 * with that TTL or lower (roughly half of all fragments). 1348 */ 1349 median_ticks = ip_reass_ttl_decr(0); 1350 1351 /* Drop half. */ 1352 median_ticks = ip_reass_ttl_decr(median_ticks); 1353 1354} 1355 1356/* 1357 * IP timer processing; 1358 * if a timer expires on a reassembly 1359 * queue, discard it. 1360 */ 1361void 1362ip_slowtimo(void) 1363{ 1364 static u_int dropscanidx = 0; 1365 u_int i; 1366 u_int median_ttl; 1367 int s = splsoftnet(); 1368 1369 IPQ_LOCK(); 1370 1371 /* Age TTL of all fragments by 1 tick .*/ 1372 median_ttl = ip_reass_ttl_decr(1); 1373 1374 /* make sure fragment limit is up-to-date */ 1375 CHECK_NMBCLUSTER_PARAMS(); 1376 1377 /* If we have too many fragments, drop the older half. */ 1378 if (ip_nfrags > ip_maxfrags) 1379 ip_reass_ttl_decr(median_ttl); 1380 1381 /* 1382 * If we are over the maximum number of fragmented packets 1383 * (due to the limit being lowered), drain off 1384 * enough to get down to the new limit. Start draining 1385 * from the reassembly hashqueue most recently drained. 1386 */ 1387 if (ip_maxfragpackets < 0) 1388 ; 1389 else { 1390 int wrapped = 0; 1391 1392 i = dropscanidx; 1393 while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) { 1394 while (LIST_FIRST(&ipq[i]) != NULL) 1395 ip_freef(LIST_FIRST(&ipq[i])); 1396 if (++i >= IPREASS_NHASH) { 1397 i = 0; 1398 } 1399 /* 1400 * Dont scan forever even if fragment counters are 1401 * wrong: stop after scanning entire reassembly queue. 1402 */ 1403 if (i == dropscanidx) 1404 wrapped = 1; 1405 } 1406 dropscanidx = i; 1407 } 1408 IPQ_UNLOCK(); 1409 splx(s); 1410} 1411 1412/* 1413 * Drain off all datagram fragments. 1414 */ 1415void 1416ip_drain(void) 1417{ 1418 1419 /* 1420 * We may be called from a device's interrupt context. If 1421 * the ipq is already busy, just bail out now. 1422 */ 1423 if (ipq_lock_try() == 0) 1424 return; 1425 1426 /* 1427 * Drop half the total fragments now. If more mbufs are needed, 1428 * we will be called again soon. 1429 */ 1430 ip_reass_drophalf(); 1431 1432 IPQ_UNLOCK(); 1433} 1434 1435/* 1436 * Do option processing on a datagram, 1437 * possibly discarding it if bad options are encountered, 1438 * or forwarding it if source-routed. 1439 * Returns 1 if packet has been forwarded/freed, 1440 * 0 if the packet should be processed further. 1441 */ 1442int 1443ip_dooptions(struct mbuf *m) 1444{ 1445 struct ip *ip = mtod(m, struct ip *); 1446 u_char *cp, *cp0; 1447 struct ip_timestamp *ipt; 1448 struct in_ifaddr *ia; 1449 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1450 struct in_addr dst; 1451 n_time ntime; 1452 1453 dst = ip->ip_dst; 1454 cp = (u_char *)(ip + 1); 1455 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1456 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1457 opt = cp[IPOPT_OPTVAL]; 1458 if (opt == IPOPT_EOL) 1459 break; 1460 if (opt == IPOPT_NOP) 1461 optlen = 1; 1462 else { 1463 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1464 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1465 goto bad; 1466 } 1467 optlen = cp[IPOPT_OLEN]; 1468 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1469 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1470 goto bad; 1471 } 1472 } 1473 switch (opt) { 1474 1475 default: 1476 break; 1477 1478 /* 1479 * Source routing with record. 1480 * Find interface with current destination address. 1481 * If none on this machine then drop if strictly routed, 1482 * or do nothing if loosely routed. 1483 * Record interface address and bring up next address 1484 * component. If strictly routed make sure next 1485 * address is on directly accessible net. 1486 */ 1487 case IPOPT_LSRR: 1488 case IPOPT_SSRR: 1489 if (ip_allowsrcrt == 0) { 1490 type = ICMP_UNREACH; 1491 code = ICMP_UNREACH_NET_PROHIB; 1492 goto bad; 1493 } 1494 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1495 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1496 goto bad; 1497 } 1498 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1499 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1500 goto bad; 1501 } 1502 ipaddr.sin_addr = ip->ip_dst; 1503 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1504 if (ia == 0) { 1505 if (opt == IPOPT_SSRR) { 1506 type = ICMP_UNREACH; 1507 code = ICMP_UNREACH_SRCFAIL; 1508 goto bad; 1509 } 1510 /* 1511 * Loose routing, and not at next destination 1512 * yet; nothing to do except forward. 1513 */ 1514 break; 1515 } 1516 off--; /* 0 origin */ 1517 if ((off + sizeof(struct in_addr)) > optlen) { 1518 /* 1519 * End of source route. Should be for us. 1520 */ 1521 save_rte(cp, ip->ip_src); 1522 break; 1523 } 1524 /* 1525 * locate outgoing interface 1526 */ 1527 bcopy((void *)(cp + off), (void *)&ipaddr.sin_addr, 1528 sizeof(ipaddr.sin_addr)); 1529 if (opt == IPOPT_SSRR) 1530 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); 1531 else 1532 ia = ip_rtaddr(ipaddr.sin_addr); 1533 if (ia == 0) { 1534 type = ICMP_UNREACH; 1535 code = ICMP_UNREACH_SRCFAIL; 1536 goto bad; 1537 } 1538 ip->ip_dst = ipaddr.sin_addr; 1539 bcopy((void *)&ia->ia_addr.sin_addr, 1540 (void *)(cp + off), sizeof(struct in_addr)); 1541 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1542 /* 1543 * Let ip_intr's mcast routing check handle mcast pkts 1544 */ 1545 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1546 break; 1547 1548 case IPOPT_RR: 1549 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1550 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1551 goto bad; 1552 } 1553 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1554 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1555 goto bad; 1556 } 1557 /* 1558 * If no space remains, ignore. 1559 */ 1560 off--; /* 0 origin */ 1561 if ((off + sizeof(struct in_addr)) > optlen) 1562 break; 1563 bcopy((void *)(&ip->ip_dst), (void *)&ipaddr.sin_addr, 1564 sizeof(ipaddr.sin_addr)); 1565 /* 1566 * locate outgoing interface; if we're the destination, 1567 * use the incoming interface (should be same). 1568 */ 1569 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1570 == NULL && 1571 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1572 type = ICMP_UNREACH; 1573 code = ICMP_UNREACH_HOST; 1574 goto bad; 1575 } 1576 bcopy((void *)&ia->ia_addr.sin_addr, 1577 (void *)(cp + off), sizeof(struct in_addr)); 1578 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1579 break; 1580 1581 case IPOPT_TS: 1582 code = cp - (u_char *)ip; 1583 ipt = (struct ip_timestamp *)cp; 1584 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1585 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1586 goto bad; 1587 } 1588 if (ipt->ipt_ptr < 5) { 1589 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1590 goto bad; 1591 } 1592 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1593 if (++ipt->ipt_oflw == 0) { 1594 code = (u_char *)&ipt->ipt_ptr - 1595 (u_char *)ip; 1596 goto bad; 1597 } 1598 break; 1599 } 1600 cp0 = (cp + ipt->ipt_ptr - 1); 1601 switch (ipt->ipt_flg) { 1602 1603 case IPOPT_TS_TSONLY: 1604 break; 1605 1606 case IPOPT_TS_TSANDADDR: 1607 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1608 sizeof(struct in_addr) > ipt->ipt_len) { 1609 code = (u_char *)&ipt->ipt_ptr - 1610 (u_char *)ip; 1611 goto bad; 1612 } 1613 ipaddr.sin_addr = dst; 1614 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1615 m->m_pkthdr.rcvif)); 1616 if (ia == 0) 1617 continue; 1618 bcopy(&ia->ia_addr.sin_addr, 1619 cp0, sizeof(struct in_addr)); 1620 ipt->ipt_ptr += sizeof(struct in_addr); 1621 break; 1622 1623 case IPOPT_TS_PRESPEC: 1624 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1625 sizeof(struct in_addr) > ipt->ipt_len) { 1626 code = (u_char *)&ipt->ipt_ptr - 1627 (u_char *)ip; 1628 goto bad; 1629 } 1630 bcopy(cp0, &ipaddr.sin_addr, 1631 sizeof(struct in_addr)); 1632 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1633 == NULL) 1634 continue; 1635 ipt->ipt_ptr += sizeof(struct in_addr); 1636 break; 1637 1638 default: 1639 /* XXX can't take &ipt->ipt_flg */ 1640 code = (u_char *)&ipt->ipt_ptr - 1641 (u_char *)ip + 1; 1642 goto bad; 1643 } 1644 ntime = iptime(); 1645 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1646 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1647 sizeof(n_time)); 1648 ipt->ipt_ptr += sizeof(n_time); 1649 } 1650 } 1651 if (forward) { 1652 if (ip_forwsrcrt == 0) { 1653 type = ICMP_UNREACH; 1654 code = ICMP_UNREACH_SRCFAIL; 1655 goto bad; 1656 } 1657 ip_forward(m, 1); 1658 return (1); 1659 } 1660 return (0); 1661bad: 1662 icmp_error(m, type, code, 0, 0); 1663 ipstat.ips_badoptions++; 1664 return (1); 1665} 1666 1667/* 1668 * Given address of next destination (final or next hop), 1669 * return internet address info of interface to be used to get there. 1670 */ 1671struct in_ifaddr * 1672ip_rtaddr(struct in_addr dst) 1673{ 1674 struct rtentry *rt; 1675 union { 1676 struct sockaddr dst; 1677 struct sockaddr_in dst4; 1678 } u; 1679 1680 sockaddr_in_init(&u.dst4, &dst, 0); 1681 1682 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) 1683 return NULL; 1684 1685 return ifatoia(rt->rt_ifa); 1686} 1687 1688/* 1689 * Save incoming source route for use in replies, 1690 * to be picked up later by ip_srcroute if the receiver is interested. 1691 */ 1692void 1693save_rte(u_char *option, struct in_addr dst) 1694{ 1695 unsigned olen; 1696 1697 olen = option[IPOPT_OLEN]; 1698#ifdef DIAGNOSTIC 1699 if (ipprintfs) 1700 printf("save_rte: olen %d\n", olen); 1701#endif /* 0 */ 1702 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1703 return; 1704 bcopy((void *)option, (void *)ip_srcrt.srcopt, olen); 1705 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1706 ip_srcrt.dst = dst; 1707} 1708 1709/* 1710 * Retrieve incoming source route for use in replies, 1711 * in the same form used by setsockopt. 1712 * The first hop is placed before the options, will be removed later. 1713 */ 1714struct mbuf * 1715ip_srcroute(void) 1716{ 1717 struct in_addr *p, *q; 1718 struct mbuf *m; 1719 1720 if (ip_nhops == 0) 1721 return NULL; 1722 m = m_get(M_DONTWAIT, MT_SOOPTS); 1723 if (m == 0) 1724 return NULL; 1725 1726 MCLAIM(m, &inetdomain.dom_mowner); 1727#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1728 1729 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1730 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1731 OPTSIZ; 1732#ifdef DIAGNOSTIC 1733 if (ipprintfs) 1734 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1735#endif 1736 1737 /* 1738 * First save first hop for return route 1739 */ 1740 p = &ip_srcrt.route[ip_nhops - 1]; 1741 *(mtod(m, struct in_addr *)) = *p--; 1742#ifdef DIAGNOSTIC 1743 if (ipprintfs) 1744 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1745#endif 1746 1747 /* 1748 * Copy option fields and padding (nop) to mbuf. 1749 */ 1750 ip_srcrt.nop = IPOPT_NOP; 1751 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1752 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop, 1753 OPTSIZ); 1754 q = (struct in_addr *)(mtod(m, char *) + 1755 sizeof(struct in_addr) + OPTSIZ); 1756#undef OPTSIZ 1757 /* 1758 * Record return path as an IP source route, 1759 * reversing the path (pointers are now aligned). 1760 */ 1761 while (p >= ip_srcrt.route) { 1762#ifdef DIAGNOSTIC 1763 if (ipprintfs) 1764 printf(" %x", ntohl(q->s_addr)); 1765#endif 1766 *q++ = *p--; 1767 } 1768 /* 1769 * Last hop goes to final destination. 1770 */ 1771 *q = ip_srcrt.dst; 1772#ifdef DIAGNOSTIC 1773 if (ipprintfs) 1774 printf(" %x\n", ntohl(q->s_addr)); 1775#endif 1776 return (m); 1777} 1778 1779/* 1780 * Strip out IP options, at higher 1781 * level protocol in the kernel. 1782 * Second argument is buffer to which options 1783 * will be moved, and return value is their length. 1784 * XXX should be deleted; last arg currently ignored. 1785 */ 1786void 1787ip_stripoptions(struct mbuf *m, struct mbuf *mopt) 1788{ 1789 int i; 1790 struct ip *ip = mtod(m, struct ip *); 1791 void *opts; 1792 int olen; 1793 1794 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1795 opts = (void *)(ip + 1); 1796 i = m->m_len - (sizeof (struct ip) + olen); 1797 memmove(opts, (char *)opts + olen, (unsigned)i); 1798 m->m_len -= olen; 1799 if (m->m_flags & M_PKTHDR) 1800 m->m_pkthdr.len -= olen; 1801 ip->ip_len = htons(ntohs(ip->ip_len) - olen); 1802 ip->ip_hl = sizeof (struct ip) >> 2; 1803} 1804 1805const int inetctlerrmap[PRC_NCMDS] = { 1806 0, 0, 0, 0, 1807 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1808 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1809 EMSGSIZE, EHOSTUNREACH, 0, 0, 1810 0, 0, 0, 0, 1811 ENOPROTOOPT 1812}; 1813 1814/* 1815 * Forward a packet. If some error occurs return the sender 1816 * an icmp packet. Note we can't always generate a meaningful 1817 * icmp message because icmp doesn't have a large enough repertoire 1818 * of codes and types. 1819 * 1820 * If not forwarding, just drop the packet. This could be confusing 1821 * if ipforwarding was zero but some routing protocol was advancing 1822 * us as a gateway to somewhere. However, we must let the routing 1823 * protocol deal with that. 1824 * 1825 * The srcrt parameter indicates whether the packet is being forwarded 1826 * via a source route. 1827 */ 1828void 1829ip_forward(struct mbuf *m, int srcrt) 1830{ 1831 struct ip *ip = mtod(m, struct ip *); 1832 struct rtentry *rt; 1833 int error, type = 0, code = 0, destmtu = 0; 1834 struct mbuf *mcopy; 1835 n_long dest; 1836 union { 1837 struct sockaddr dst; 1838 struct sockaddr_in dst4; 1839 } u; 1840 1841 /* 1842 * We are now in the output path. 1843 */ 1844 MCLAIM(m, &ip_tx_mowner); 1845 1846 /* 1847 * Clear any in-bound checksum flags for this packet. 1848 */ 1849 m->m_pkthdr.csum_flags = 0; 1850 1851 dest = 0; 1852#ifdef DIAGNOSTIC 1853 if (ipprintfs) { 1854 printf("forward: src %s ", inet_ntoa(ip->ip_src)); 1855 printf("dst %s ttl %x\n", inet_ntoa(ip->ip_dst), ip->ip_ttl); 1856 } 1857#endif 1858 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1859 ipstat.ips_cantforward++; 1860 m_freem(m); 1861 return; 1862 } 1863 if (ip->ip_ttl <= IPTTLDEC) { 1864 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1865 return; 1866 } 1867 1868 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1869 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) { 1870 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1871 return; 1872 } 1873 1874 /* 1875 * Save at most 68 bytes of the packet in case 1876 * we need to generate an ICMP message to the src. 1877 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1878 */ 1879 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1880 if (mcopy) 1881 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1882 1883 ip->ip_ttl -= IPTTLDEC; 1884 1885 /* 1886 * If forwarding packet using same interface that it came in on, 1887 * perhaps should send a redirect to sender to shortcut a hop. 1888 * Only send redirect if source is sending directly to us, 1889 * and if packet was not source routed (or has any options). 1890 * Also, don't send redirect if forwarding using a default route 1891 * or a route modified by a redirect. 1892 */ 1893 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1894 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1895 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1896 ipsendredirects && !srcrt) { 1897 if (rt->rt_ifa && 1898 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1899 ifatoia(rt->rt_ifa)->ia_subnet) { 1900 if (rt->rt_flags & RTF_GATEWAY) 1901 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1902 else 1903 dest = ip->ip_dst.s_addr; 1904 /* 1905 * Router requirements says to only send host 1906 * redirects. 1907 */ 1908 type = ICMP_REDIRECT; 1909 code = ICMP_REDIRECT_HOST; 1910#ifdef DIAGNOSTIC 1911 if (ipprintfs) 1912 printf("redirect (%d) to %x\n", code, 1913 (u_int32_t)dest); 1914#endif 1915 } 1916 } 1917 1918 error = ip_output(m, NULL, &ipforward_rt, 1919 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1920 (struct ip_moptions *)NULL, (struct socket *)NULL); 1921 1922 if (error) 1923 ipstat.ips_cantforward++; 1924 else { 1925 ipstat.ips_forward++; 1926 if (type) 1927 ipstat.ips_redirectsent++; 1928 else { 1929 if (mcopy) { 1930#ifdef GATEWAY 1931 if (mcopy->m_flags & M_CANFASTFWD) 1932 ipflow_create(&ipforward_rt, mcopy); 1933#endif 1934 m_freem(mcopy); 1935 } 1936 return; 1937 } 1938 } 1939 if (mcopy == NULL) 1940 return; 1941 1942 switch (error) { 1943 1944 case 0: /* forwarded, but need redirect */ 1945 /* type, code set above */ 1946 break; 1947 1948 case ENETUNREACH: /* shouldn't happen, checked above */ 1949 case EHOSTUNREACH: 1950 case ENETDOWN: 1951 case EHOSTDOWN: 1952 default: 1953 type = ICMP_UNREACH; 1954 code = ICMP_UNREACH_HOST; 1955 break; 1956 1957 case EMSGSIZE: 1958 type = ICMP_UNREACH; 1959 code = ICMP_UNREACH_NEEDFRAG; 1960#if !defined(IPSEC) && !defined(FAST_IPSEC) 1961 if (ipforward_rt.ro_rt != NULL) 1962 destmtu = ipforward_rt.ro_rt->rt_ifp->if_mtu; 1963#else 1964 /* 1965 * If the packet is routed over IPsec tunnel, tell the 1966 * originator the tunnel MTU. 1967 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1968 * XXX quickhack!!! 1969 */ 1970 if (ipforward_rt.ro_rt != NULL) { 1971 struct secpolicy *sp; 1972 int ipsecerror; 1973 size_t ipsechdr; 1974 struct route *ro; 1975 1976 sp = ipsec4_getpolicybyaddr(mcopy, 1977 IPSEC_DIR_OUTBOUND, IP_FORWARDING, 1978 &ipsecerror); 1979 1980 if (sp == NULL) 1981 destmtu = ipforward_rt.ro_rt->rt_ifp->if_mtu; 1982 else { 1983 /* count IPsec header size */ 1984 ipsechdr = ipsec4_hdrsiz(mcopy, 1985 IPSEC_DIR_OUTBOUND, NULL); 1986 1987 /* 1988 * find the correct route for outer IPv4 1989 * header, compute tunnel MTU. 1990 */ 1991 1992 if (sp->req != NULL 1993 && sp->req->sav != NULL 1994 && sp->req->sav->sah != NULL) { 1995 ro = &sp->req->sav->sah->sa_route; 1996 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1997 destmtu = 1998 ro->ro_rt->rt_rmx.rmx_mtu ? 1999 ro->ro_rt->rt_rmx.rmx_mtu : 2000 ro->ro_rt->rt_ifp->if_mtu; 2001 destmtu -= ipsechdr; 2002 } 2003 } 2004 2005#ifdef IPSEC 2006 key_freesp(sp); 2007#else 2008 KEY_FREESP(&sp); 2009#endif 2010 } 2011 } 2012#endif /*IPSEC*/ 2013 ipstat.ips_cantfrag++; 2014 break; 2015 2016 case ENOBUFS: 2017#if 1 2018 /* 2019 * a router should not generate ICMP_SOURCEQUENCH as 2020 * required in RFC1812 Requirements for IP Version 4 Routers. 2021 * source quench could be a big problem under DoS attacks, 2022 * or if the underlying interface is rate-limited. 2023 */ 2024 if (mcopy) 2025 m_freem(mcopy); 2026 return; 2027#else 2028 type = ICMP_SOURCEQUENCH; 2029 code = 0; 2030 break; 2031#endif 2032 } 2033 icmp_error(mcopy, type, code, dest, destmtu); 2034} 2035 2036void 2037ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2038 struct mbuf *m) 2039{ 2040 2041 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2042 struct timeval tv; 2043 2044 microtime(&tv); 2045 *mp = sbcreatecontrol((void *) &tv, sizeof(tv), 2046 SCM_TIMESTAMP, SOL_SOCKET); 2047 if (*mp) 2048 mp = &(*mp)->m_next; 2049 } 2050 if (inp->inp_flags & INP_RECVDSTADDR) { 2051 *mp = sbcreatecontrol((void *) &ip->ip_dst, 2052 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2053 if (*mp) 2054 mp = &(*mp)->m_next; 2055 } 2056#ifdef notyet 2057 /* 2058 * XXX 2059 * Moving these out of udp_input() made them even more broken 2060 * than they already were. 2061 * - fenner@parc.xerox.com 2062 */ 2063 /* options were tossed already */ 2064 if (inp->inp_flags & INP_RECVOPTS) { 2065 *mp = sbcreatecontrol((void *) opts_deleted_above, 2066 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2067 if (*mp) 2068 mp = &(*mp)->m_next; 2069 } 2070 /* ip_srcroute doesn't do what we want here, need to fix */ 2071 if (inp->inp_flags & INP_RECVRETOPTS) { 2072 *mp = sbcreatecontrol((void *) ip_srcroute(), 2073 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2074 if (*mp) 2075 mp = &(*mp)->m_next; 2076 } 2077#endif 2078 if (inp->inp_flags & INP_RECVIF) { 2079 struct sockaddr_dl sdl; 2080 2081 sockaddr_dl_init(&sdl, sizeof(sdl), 2082 (m->m_pkthdr.rcvif != NULL) 2083 ? m->m_pkthdr.rcvif->if_index 2084 : 0, 2085 0, NULL, 0, NULL, 0); 2086 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 2087 if (*mp) 2088 mp = &(*mp)->m_next; 2089 } 2090} 2091 2092/* 2093 * sysctl helper routine for net.inet.ip.forwsrcrt. 2094 */ 2095static int 2096sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 2097{ 2098 int error, tmp; 2099 struct sysctlnode node; 2100 2101 node = *rnode; 2102 tmp = ip_forwsrcrt; 2103 node.sysctl_data = &tmp; 2104 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2105 if (error || newp == NULL) 2106 return (error); 2107 2108 if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 2109 0, NULL, NULL, NULL)) 2110 return (EPERM); 2111 2112 ip_forwsrcrt = tmp; 2113 2114 return (0); 2115} 2116 2117/* 2118 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 2119 * range of the new value and tweaks timers if it changes. 2120 */ 2121static int 2122sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 2123{ 2124 int error, tmp; 2125 struct sysctlnode node; 2126 2127 node = *rnode; 2128 tmp = ip_mtudisc_timeout; 2129 node.sysctl_data = &tmp; 2130 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2131 if (error || newp == NULL) 2132 return (error); 2133 if (tmp < 0) 2134 return (EINVAL); 2135 2136 ip_mtudisc_timeout = tmp; 2137 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 2138 2139 return (0); 2140} 2141 2142#ifdef GATEWAY 2143/* 2144 * sysctl helper routine for net.inet.ip.maxflows. 2145 */ 2146static int 2147sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS) 2148{ 2149 int s; 2150 2151 s = sysctl_lookup(SYSCTLFN_CALL(rnode)); 2152 if (s || newp == NULL) 2153 return (s); 2154 2155 s = splsoftnet(); 2156 ipflow_reap(0); 2157 splx(s); 2158 2159 return (0); 2160} 2161 2162static int 2163sysctl_net_inet_ip_hashsize(SYSCTLFN_ARGS) 2164{ 2165 int error, tmp; 2166 struct sysctlnode node; 2167 2168 node = *rnode; 2169 tmp = ip_hashsize; 2170 node.sysctl_data = &tmp; 2171 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2172 if (error || newp == NULL) 2173 return (error); 2174 2175 if ((tmp & (tmp - 1)) == 0 && tmp != 0) { 2176 /* 2177 * Can only fail due to malloc() 2178 */ 2179 if (ipflow_invalidate_all(tmp)) 2180 return ENOMEM; 2181 } else { 2182 /* 2183 * EINVAL if not a power of 2 2184 */ 2185 return EINVAL; 2186 } 2187 2188 return (0); 2189} 2190#endif /* GATEWAY */ 2191 2192 2193SYSCTL_SETUP(sysctl_net_inet_ip_setup, "sysctl net.inet.ip subtree setup") 2194{ 2195 extern int subnetsarelocal, hostzeroisbroadcast; 2196 2197 sysctl_createv(clog, 0, NULL, NULL, 2198 CTLFLAG_PERMANENT, 2199 CTLTYPE_NODE, "net", NULL, 2200 NULL, 0, NULL, 0, 2201 CTL_NET, CTL_EOL); 2202 sysctl_createv(clog, 0, NULL, NULL, 2203 CTLFLAG_PERMANENT, 2204 CTLTYPE_NODE, "inet", 2205 SYSCTL_DESCR("PF_INET related settings"), 2206 NULL, 0, NULL, 0, 2207 CTL_NET, PF_INET, CTL_EOL); 2208 sysctl_createv(clog, 0, NULL, NULL, 2209 CTLFLAG_PERMANENT, 2210 CTLTYPE_NODE, "ip", 2211 SYSCTL_DESCR("IPv4 related settings"), 2212 NULL, 0, NULL, 0, 2213 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 2214 2215 sysctl_createv(clog, 0, NULL, NULL, 2216 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2217 CTLTYPE_INT, "forwarding", 2218 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 2219 NULL, 0, &ipforwarding, 0, 2220 CTL_NET, PF_INET, IPPROTO_IP, 2221 IPCTL_FORWARDING, CTL_EOL); 2222 sysctl_createv(clog, 0, NULL, NULL, 2223 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2224 CTLTYPE_INT, "redirect", 2225 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 2226 NULL, 0, &ipsendredirects, 0, 2227 CTL_NET, PF_INET, IPPROTO_IP, 2228 IPCTL_SENDREDIRECTS, CTL_EOL); 2229 sysctl_createv(clog, 0, NULL, NULL, 2230 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2231 CTLTYPE_INT, "ttl", 2232 SYSCTL_DESCR("Default TTL for an INET datagram"), 2233 NULL, 0, &ip_defttl, 0, 2234 CTL_NET, PF_INET, IPPROTO_IP, 2235 IPCTL_DEFTTL, CTL_EOL); 2236#ifdef IPCTL_DEFMTU 2237 sysctl_createv(clog, 0, NULL, NULL, 2238 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 2239 CTLTYPE_INT, "mtu", 2240 SYSCTL_DESCR("Default MTA for an INET route"), 2241 NULL, 0, &ip_mtu, 0, 2242 CTL_NET, PF_INET, IPPROTO_IP, 2243 IPCTL_DEFMTU, CTL_EOL); 2244#endif /* IPCTL_DEFMTU */ 2245 sysctl_createv(clog, 0, NULL, NULL, 2246 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2247 CTLTYPE_INT, "forwsrcrt", 2248 SYSCTL_DESCR("Enable forwarding of source-routed " 2249 "datagrams"), 2250 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 2251 CTL_NET, PF_INET, IPPROTO_IP, 2252 IPCTL_FORWSRCRT, CTL_EOL); 2253 sysctl_createv(clog, 0, NULL, NULL, 2254 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2255 CTLTYPE_INT, "directed-broadcast", 2256 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 2257 NULL, 0, &ip_directedbcast, 0, 2258 CTL_NET, PF_INET, IPPROTO_IP, 2259 IPCTL_DIRECTEDBCAST, CTL_EOL); 2260 sysctl_createv(clog, 0, NULL, NULL, 2261 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2262 CTLTYPE_INT, "allowsrcrt", 2263 SYSCTL_DESCR("Accept source-routed datagrams"), 2264 NULL, 0, &ip_allowsrcrt, 0, 2265 CTL_NET, PF_INET, IPPROTO_IP, 2266 IPCTL_ALLOWSRCRT, CTL_EOL); 2267 sysctl_createv(clog, 0, NULL, NULL, 2268 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2269 CTLTYPE_INT, "subnetsarelocal", 2270 SYSCTL_DESCR("Whether logical subnets are considered " 2271 "local"), 2272 NULL, 0, &subnetsarelocal, 0, 2273 CTL_NET, PF_INET, IPPROTO_IP, 2274 IPCTL_SUBNETSARELOCAL, CTL_EOL); 2275 sysctl_createv(clog, 0, NULL, NULL, 2276 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2277 CTLTYPE_INT, "mtudisc", 2278 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 2279 NULL, 0, &ip_mtudisc, 0, 2280 CTL_NET, PF_INET, IPPROTO_IP, 2281 IPCTL_MTUDISC, CTL_EOL); 2282 sysctl_createv(clog, 0, NULL, NULL, 2283 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2284 CTLTYPE_INT, "anonportmin", 2285 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 2286 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 2287 CTL_NET, PF_INET, IPPROTO_IP, 2288 IPCTL_ANONPORTMIN, CTL_EOL); 2289 sysctl_createv(clog, 0, NULL, NULL, 2290 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2291 CTLTYPE_INT, "anonportmax", 2292 SYSCTL_DESCR("Highest ephemeral port number to assign"), 2293 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 2294 CTL_NET, PF_INET, IPPROTO_IP, 2295 IPCTL_ANONPORTMAX, CTL_EOL); 2296 sysctl_createv(clog, 0, NULL, NULL, 2297 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2298 CTLTYPE_INT, "mtudisctimeout", 2299 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 2300 sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0, 2301 CTL_NET, PF_INET, IPPROTO_IP, 2302 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 2303#ifdef GATEWAY 2304 sysctl_createv(clog, 0, NULL, NULL, 2305 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2306 CTLTYPE_INT, "maxflows", 2307 SYSCTL_DESCR("Number of flows for fast forwarding"), 2308 sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0, 2309 CTL_NET, PF_INET, IPPROTO_IP, 2310 IPCTL_MAXFLOWS, CTL_EOL); 2311 sysctl_createv(clog, 0, NULL, NULL, 2312 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2313 CTLTYPE_INT, "hashsize", 2314 SYSCTL_DESCR("Size of hash table for fast forwarding (IPv4)"), 2315 sysctl_net_inet_ip_hashsize, 0, &ip_hashsize, 0, 2316 CTL_NET, PF_INET, IPPROTO_IP, 2317 CTL_CREATE, CTL_EOL); 2318#endif /* GATEWAY */ 2319 sysctl_createv(clog, 0, NULL, NULL, 2320 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2321 CTLTYPE_INT, "hostzerobroadcast", 2322 SYSCTL_DESCR("All zeroes address is broadcast address"), 2323 NULL, 0, &hostzeroisbroadcast, 0, 2324 CTL_NET, PF_INET, IPPROTO_IP, 2325 IPCTL_HOSTZEROBROADCAST, CTL_EOL); 2326#if NGIF > 0 2327 sysctl_createv(clog, 0, NULL, NULL, 2328 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2329 CTLTYPE_INT, "gifttl", 2330 SYSCTL_DESCR("Default TTL for a gif tunnel datagram"), 2331 NULL, 0, &ip_gif_ttl, 0, 2332 CTL_NET, PF_INET, IPPROTO_IP, 2333 IPCTL_GIF_TTL, CTL_EOL); 2334#endif /* NGIF */ 2335#ifndef IPNOPRIVPORTS 2336 sysctl_createv(clog, 0, NULL, NULL, 2337 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2338 CTLTYPE_INT, "lowportmin", 2339 SYSCTL_DESCR("Lowest privileged ephemeral port number " 2340 "to assign"), 2341 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 2342 CTL_NET, PF_INET, IPPROTO_IP, 2343 IPCTL_LOWPORTMIN, CTL_EOL); 2344 sysctl_createv(clog, 0, NULL, NULL, 2345 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2346 CTLTYPE_INT, "lowportmax", 2347 SYSCTL_DESCR("Highest privileged ephemeral port number " 2348 "to assign"), 2349 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 2350 CTL_NET, PF_INET, IPPROTO_IP, 2351 IPCTL_LOWPORTMAX, CTL_EOL); 2352#endif /* IPNOPRIVPORTS */ 2353 sysctl_createv(clog, 0, NULL, NULL, 2354 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2355 CTLTYPE_INT, "maxfragpackets", 2356 SYSCTL_DESCR("Maximum number of fragments to retain for " 2357 "possible reassembly"), 2358 NULL, 0, &ip_maxfragpackets, 0, 2359 CTL_NET, PF_INET, IPPROTO_IP, 2360 IPCTL_MAXFRAGPACKETS, CTL_EOL); 2361#if NGRE > 0 2362 sysctl_createv(clog, 0, NULL, NULL, 2363 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2364 CTLTYPE_INT, "grettl", 2365 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 2366 NULL, 0, &ip_gre_ttl, 0, 2367 CTL_NET, PF_INET, IPPROTO_IP, 2368 IPCTL_GRE_TTL, CTL_EOL); 2369#endif /* NGRE */ 2370 sysctl_createv(clog, 0, NULL, NULL, 2371 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2372 CTLTYPE_INT, "checkinterface", 2373 SYSCTL_DESCR("Enable receive side of Strong ES model " 2374 "from RFC1122"), 2375 NULL, 0, &ip_checkinterface, 0, 2376 CTL_NET, PF_INET, IPPROTO_IP, 2377 IPCTL_CHECKINTERFACE, CTL_EOL); 2378 sysctl_createv(clog, 0, NULL, NULL, 2379 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2380 CTLTYPE_INT, "random_id", 2381 SYSCTL_DESCR("Assign random ip_id values"), 2382 NULL, 0, &ip_do_randomid, 0, 2383 CTL_NET, PF_INET, IPPROTO_IP, 2384 IPCTL_RANDOMID, CTL_EOL); 2385 sysctl_createv(clog, 0, NULL, NULL, 2386 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2387 CTLTYPE_INT, "do_loopback_cksum", 2388 SYSCTL_DESCR("Perform IP checksum on loopback"), 2389 NULL, 0, &ip_do_loopback_cksum, 0, 2390 CTL_NET, PF_INET, IPPROTO_IP, 2391 IPCTL_LOOPBACKCKSUM, CTL_EOL); 2392 sysctl_createv(clog, 0, NULL, NULL, 2393 CTLFLAG_PERMANENT, 2394 CTLTYPE_STRUCT, "stats", 2395 SYSCTL_DESCR("IP statistics"), 2396 NULL, 0, &ipstat, sizeof(ipstat), 2397 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 2398 CTL_EOL); 2399} 2400