ip_input.c revision 1.263
1/* $NetBSD: ip_input.c,v 1.263 2008/03/27 00:18:56 cube 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 1.263 2008/03/27 00:18:56 cube 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_initid(); 421 ip_id = time_second & 0xfffff; 422 423 ipintrq.ifq_maxlen = ipqmaxlen; 424 ip_nmbclusters_changed(); 425 426 TAILQ_INIT(&in_ifaddrhead); 427 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 428 M_WAITOK, &in_ifaddrhash); 429 in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IPMADDR, 430 M_WAITOK, &in_multihash); 431 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 432#ifdef GATEWAY 433 ipflow_init(ip_hashsize); 434#endif 435 436#ifdef PFIL_HOOKS 437 /* Register our Packet Filter hook. */ 438 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 439 inet_pfil_hook.ph_af = AF_INET; 440 i = pfil_head_register(&inet_pfil_hook); 441 if (i != 0) 442 printf("ip_init: WARNING: unable to register pfil hook, " 443 "error %d\n", i); 444#endif /* PFIL_HOOKS */ 445 446#ifdef MBUFTRACE 447 MOWNER_ATTACH(&ip_tx_mowner); 448 MOWNER_ATTACH(&ip_rx_mowner); 449#endif /* MBUFTRACE */ 450} 451 452struct sockaddr_in ipaddr = { 453 .sin_len = sizeof(ipaddr), 454 .sin_family = AF_INET, 455}; 456struct route ipforward_rt; 457 458/* 459 * IP software interrupt routine 460 */ 461void 462ipintr(void) 463{ 464 int s; 465 struct mbuf *m; 466 467 while (!IF_IS_EMPTY(&ipintrq)) { 468 s = splnet(); 469 IF_DEQUEUE(&ipintrq, m); 470 splx(s); 471 if (m == 0) 472 return; 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 uint16_t off; 901 /* 902 * Prevent TCP blind data attacks by not allowing non-initial 903 * fragments to start at less than 68 bytes (minimal fragment 904 * size) and making sure the first fragment is at least 68 905 * bytes. 906 */ 907 off = (ntohs(ip->ip_off) & IP_OFFMASK) << 3; 908 if ((off > 0 ? off + hlen : len) < IP_MINFRAGSIZE - 1) { 909 ipstat.ips_badfrags++; 910 goto bad; 911 } 912 /* 913 * Look for queue of fragments 914 * of this datagram. 915 */ 916 IPQ_LOCK(); 917 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 918 LIST_FOREACH(fp, &ipq[hash], ipq_q) { 919 if (ip->ip_id == fp->ipq_id && 920 in_hosteq(ip->ip_src, fp->ipq_src) && 921 in_hosteq(ip->ip_dst, fp->ipq_dst) && 922 ip->ip_p == fp->ipq_p) { 923 /* 924 * Make sure the TOS is matches previous 925 * fragments. 926 */ 927 if (ip->ip_tos != fp->ipq_tos) { 928 ipstat.ips_badfrags++; 929 goto bad; 930 } 931 goto found; 932 } 933 } 934 fp = 0; 935found: 936 937 /* 938 * Adjust ip_len to not reflect header, 939 * set ipqe_mff if more fragments are expected, 940 * convert offset of this to bytes. 941 */ 942 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 943 mff = (ip->ip_off & htons(IP_MF)) != 0; 944 if (mff) { 945 /* 946 * Make sure that fragments have a data length 947 * that's a non-zero multiple of 8 bytes. 948 */ 949 if (ntohs(ip->ip_len) == 0 || 950 (ntohs(ip->ip_len) & 0x7) != 0) { 951 ipstat.ips_badfrags++; 952 IPQ_UNLOCK(); 953 goto bad; 954 } 955 } 956 ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3); 957 958 /* 959 * If datagram marked as having more fragments 960 * or if this is not the first fragment, 961 * attempt reassembly; if it succeeds, proceed. 962 */ 963 if (mff || ip->ip_off != htons(0)) { 964 ipstat.ips_fragments++; 965 s = splvm(); 966 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 967 splx(s); 968 if (ipqe == NULL) { 969 ipstat.ips_rcvmemdrop++; 970 IPQ_UNLOCK(); 971 goto bad; 972 } 973 ipqe->ipqe_mff = mff; 974 ipqe->ipqe_m = m; 975 ipqe->ipqe_ip = ip; 976 m = ip_reass(ipqe, fp, &ipq[hash]); 977 if (m == 0) { 978 IPQ_UNLOCK(); 979 return; 980 } 981 ipstat.ips_reassembled++; 982 ip = mtod(m, struct ip *); 983 hlen = ip->ip_hl << 2; 984 ip->ip_len = htons(ntohs(ip->ip_len) + hlen); 985 } else 986 if (fp) 987 ip_freef(fp); 988 IPQ_UNLOCK(); 989 } 990 991#if defined(IPSEC) 992 /* 993 * enforce IPsec policy checking if we are seeing last header. 994 * note that we do not visit this with protocols with pcb layer 995 * code - like udp/tcp/raw ip. 996 */ 997 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 998 ipsec4_in_reject(m, NULL)) { 999 ipsecstat.in_polvio++; 1000 goto bad; 1001 } 1002#endif 1003#ifdef FAST_IPSEC 1004 /* 1005 * enforce IPsec policy checking if we are seeing last header. 1006 * note that we do not visit this with protocols with pcb layer 1007 * code - like udp/tcp/raw ip. 1008 */ 1009 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 1010 /* 1011 * Check if the packet has already had IPsec processing 1012 * done. If so, then just pass it along. This tag gets 1013 * set during AH, ESP, etc. input handling, before the 1014 * packet is returned to the ip input queue for delivery. 1015 */ 1016 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 1017 s = splsoftnet(); 1018 if (mtag != NULL) { 1019 tdbi = (struct tdb_ident *)(mtag + 1); 1020 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 1021 } else { 1022 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 1023 IP_FORWARDING, &error); 1024 } 1025 if (sp != NULL) { 1026 /* 1027 * Check security policy against packet attributes. 1028 */ 1029 error = ipsec_in_reject(sp, m); 1030 KEY_FREESP(&sp); 1031 } else { 1032 /* XXX error stat??? */ 1033 error = EINVAL; 1034DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 1035 } 1036 splx(s); 1037 if (error) 1038 goto bad; 1039 } 1040#endif /* FAST_IPSEC */ 1041 1042 /* 1043 * Switch out to protocol's input routine. 1044 */ 1045#if IFA_STATS 1046 if (ia && ip) 1047 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 1048#endif 1049 ipstat.ips_delivered++; 1050 { 1051 int off = hlen, nh = ip->ip_p; 1052 1053 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 1054 return; 1055 } 1056bad: 1057 m_freem(m); 1058 return; 1059 1060badcsum: 1061 ipstat.ips_badsum++; 1062 m_freem(m); 1063} 1064 1065/* 1066 * Take incoming datagram fragment and try to 1067 * reassemble it into whole datagram. If a chain for 1068 * reassembly of this datagram already exists, then it 1069 * is given as fp; otherwise have to make a chain. 1070 */ 1071struct mbuf * 1072ip_reass(struct ipqent *ipqe, struct ipq *fp, struct ipqhead *ipqhead) 1073{ 1074 struct mbuf *m = ipqe->ipqe_m; 1075 struct ipqent *nq, *p, *q; 1076 struct ip *ip; 1077 struct mbuf *t; 1078 int hlen = ipqe->ipqe_ip->ip_hl << 2; 1079 int i, next, s; 1080 1081 IPQ_LOCK_CHECK(); 1082 1083 /* 1084 * Presence of header sizes in mbufs 1085 * would confuse code below. 1086 */ 1087 m->m_data += hlen; 1088 m->m_len -= hlen; 1089 1090#ifdef notyet 1091 /* make sure fragment limit is up-to-date */ 1092 CHECK_NMBCLUSTER_PARAMS(); 1093 1094 /* If we have too many fragments, drop the older half. */ 1095 if (ip_nfrags >= ip_maxfrags) 1096 ip_reass_drophalf(void); 1097#endif 1098 1099 /* 1100 * We are about to add a fragment; increment frag count. 1101 */ 1102 ip_nfrags++; 1103 1104 /* 1105 * If first fragment to arrive, create a reassembly queue. 1106 */ 1107 if (fp == 0) { 1108 /* 1109 * Enforce upper bound on number of fragmented packets 1110 * for which we attempt reassembly; 1111 * If maxfrag is 0, never accept fragments. 1112 * If maxfrag is -1, accept all fragments without limitation. 1113 */ 1114 if (ip_maxfragpackets < 0) 1115 ; 1116 else if (ip_nfragpackets >= ip_maxfragpackets) 1117 goto dropfrag; 1118 ip_nfragpackets++; 1119 MALLOC(fp, struct ipq *, sizeof (struct ipq), 1120 M_FTABLE, M_NOWAIT); 1121 if (fp == NULL) 1122 goto dropfrag; 1123 LIST_INSERT_HEAD(ipqhead, fp, ipq_q); 1124 fp->ipq_nfrags = 1; 1125 fp->ipq_ttl = IPFRAGTTL; 1126 fp->ipq_p = ipqe->ipqe_ip->ip_p; 1127 fp->ipq_id = ipqe->ipqe_ip->ip_id; 1128 fp->ipq_tos = ipqe->ipqe_ip->ip_tos; 1129 TAILQ_INIT(&fp->ipq_fragq); 1130 fp->ipq_src = ipqe->ipqe_ip->ip_src; 1131 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 1132 p = NULL; 1133 goto insert; 1134 } else { 1135 fp->ipq_nfrags++; 1136 } 1137 1138 /* 1139 * Find a segment which begins after this one does. 1140 */ 1141 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1142 p = q, q = TAILQ_NEXT(q, ipqe_q)) 1143 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off)) 1144 break; 1145 1146 /* 1147 * If there is a preceding segment, it may provide some of 1148 * our data already. If so, drop the data from the incoming 1149 * segment. If it provides all of our data, drop us. 1150 */ 1151 if (p != NULL) { 1152 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) - 1153 ntohs(ipqe->ipqe_ip->ip_off); 1154 if (i > 0) { 1155 if (i >= ntohs(ipqe->ipqe_ip->ip_len)) 1156 goto dropfrag; 1157 m_adj(ipqe->ipqe_m, i); 1158 ipqe->ipqe_ip->ip_off = 1159 htons(ntohs(ipqe->ipqe_ip->ip_off) + i); 1160 ipqe->ipqe_ip->ip_len = 1161 htons(ntohs(ipqe->ipqe_ip->ip_len) - i); 1162 } 1163 } 1164 1165 /* 1166 * While we overlap succeeding segments trim them or, 1167 * if they are completely covered, dequeue them. 1168 */ 1169 for (; q != NULL && 1170 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) > 1171 ntohs(q->ipqe_ip->ip_off); q = nq) { 1172 i = (ntohs(ipqe->ipqe_ip->ip_off) + 1173 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off); 1174 if (i < ntohs(q->ipqe_ip->ip_len)) { 1175 q->ipqe_ip->ip_len = 1176 htons(ntohs(q->ipqe_ip->ip_len) - i); 1177 q->ipqe_ip->ip_off = 1178 htons(ntohs(q->ipqe_ip->ip_off) + i); 1179 m_adj(q->ipqe_m, i); 1180 break; 1181 } 1182 nq = TAILQ_NEXT(q, ipqe_q); 1183 m_freem(q->ipqe_m); 1184 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1185 s = splvm(); 1186 pool_put(&ipqent_pool, q); 1187 splx(s); 1188 fp->ipq_nfrags--; 1189 ip_nfrags--; 1190 } 1191 1192insert: 1193 /* 1194 * Stick new segment in its place; 1195 * check for complete reassembly. 1196 */ 1197 if (p == NULL) { 1198 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 1199 } else { 1200 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q); 1201 } 1202 next = 0; 1203 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1204 p = q, q = TAILQ_NEXT(q, ipqe_q)) { 1205 if (ntohs(q->ipqe_ip->ip_off) != next) 1206 return (0); 1207 next += ntohs(q->ipqe_ip->ip_len); 1208 } 1209 if (p->ipqe_mff) 1210 return (0); 1211 1212 /* 1213 * Reassembly is complete. Check for a bogus message size and 1214 * concatenate fragments. 1215 */ 1216 q = TAILQ_FIRST(&fp->ipq_fragq); 1217 ip = q->ipqe_ip; 1218 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 1219 ipstat.ips_toolong++; 1220 ip_freef(fp); 1221 return (0); 1222 } 1223 m = q->ipqe_m; 1224 t = m->m_next; 1225 m->m_next = 0; 1226 m_cat(m, t); 1227 nq = TAILQ_NEXT(q, ipqe_q); 1228 s = splvm(); 1229 pool_put(&ipqent_pool, q); 1230 splx(s); 1231 for (q = nq; q != NULL; q = nq) { 1232 t = q->ipqe_m; 1233 nq = TAILQ_NEXT(q, ipqe_q); 1234 s = splvm(); 1235 pool_put(&ipqent_pool, q); 1236 splx(s); 1237 m_cat(m, t); 1238 } 1239 ip_nfrags -= fp->ipq_nfrags; 1240 1241 /* 1242 * Create header for new ip packet by 1243 * modifying header of first packet; 1244 * dequeue and discard fragment reassembly header. 1245 * Make header visible. 1246 */ 1247 ip->ip_len = htons(next); 1248 ip->ip_src = fp->ipq_src; 1249 ip->ip_dst = fp->ipq_dst; 1250 LIST_REMOVE(fp, ipq_q); 1251 FREE(fp, M_FTABLE); 1252 ip_nfragpackets--; 1253 m->m_len += (ip->ip_hl << 2); 1254 m->m_data -= (ip->ip_hl << 2); 1255 /* some debugging cruft by sklower, below, will go away soon */ 1256 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1257 int plen = 0; 1258 for (t = m; t; t = t->m_next) 1259 plen += t->m_len; 1260 m->m_pkthdr.len = plen; 1261 m->m_pkthdr.csum_flags = 0; 1262 } 1263 return (m); 1264 1265dropfrag: 1266 if (fp != 0) 1267 fp->ipq_nfrags--; 1268 ip_nfrags--; 1269 ipstat.ips_fragdropped++; 1270 m_freem(m); 1271 s = splvm(); 1272 pool_put(&ipqent_pool, ipqe); 1273 splx(s); 1274 return (0); 1275} 1276 1277/* 1278 * Free a fragment reassembly header and all 1279 * associated datagrams. 1280 */ 1281void 1282ip_freef(struct ipq *fp) 1283{ 1284 struct ipqent *q, *p; 1285 u_int nfrags = 0; 1286 int s; 1287 1288 IPQ_LOCK_CHECK(); 1289 1290 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) { 1291 p = TAILQ_NEXT(q, ipqe_q); 1292 m_freem(q->ipqe_m); 1293 nfrags++; 1294 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1295 s = splvm(); 1296 pool_put(&ipqent_pool, q); 1297 splx(s); 1298 } 1299 1300 if (nfrags != fp->ipq_nfrags) 1301 printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags); 1302 ip_nfrags -= nfrags; 1303 LIST_REMOVE(fp, ipq_q); 1304 FREE(fp, M_FTABLE); 1305 ip_nfragpackets--; 1306} 1307 1308/* 1309 * IP reassembly TTL machinery for multiplicative drop. 1310 */ 1311static u_int fragttl_histo[(IPFRAGTTL+1)]; 1312 1313 1314/* 1315 * Decrement TTL of all reasembly queue entries by `ticks'. 1316 * Count number of distinct fragments (as opposed to partial, fragmented 1317 * datagrams) in the reassembly queue. While we traverse the entire 1318 * reassembly queue, compute and return the median TTL over all fragments. 1319 */ 1320static u_int 1321ip_reass_ttl_decr(u_int ticks) 1322{ 1323 u_int nfrags, median, dropfraction, keepfraction; 1324 struct ipq *fp, *nfp; 1325 int i; 1326 1327 nfrags = 0; 1328 memset(fragttl_histo, 0, sizeof fragttl_histo); 1329 1330 for (i = 0; i < IPREASS_NHASH; i++) { 1331 for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) { 1332 fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ? 1333 0 : fp->ipq_ttl - ticks); 1334 nfp = LIST_NEXT(fp, ipq_q); 1335 if (fp->ipq_ttl == 0) { 1336 ipstat.ips_fragtimeout++; 1337 ip_freef(fp); 1338 } else { 1339 nfrags += fp->ipq_nfrags; 1340 fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags; 1341 } 1342 } 1343 } 1344 1345 KASSERT(ip_nfrags == nfrags); 1346 1347 /* Find median (or other drop fraction) in histogram. */ 1348 dropfraction = (ip_nfrags / 2); 1349 keepfraction = ip_nfrags - dropfraction; 1350 for (i = IPFRAGTTL, median = 0; i >= 0; i--) { 1351 median += fragttl_histo[i]; 1352 if (median >= keepfraction) 1353 break; 1354 } 1355 1356 /* Return TTL of median (or other fraction). */ 1357 return (u_int)i; 1358} 1359 1360void 1361ip_reass_drophalf(void) 1362{ 1363 1364 u_int median_ticks; 1365 /* 1366 * Compute median TTL of all fragments, and count frags 1367 * with that TTL or lower (roughly half of all fragments). 1368 */ 1369 median_ticks = ip_reass_ttl_decr(0); 1370 1371 /* Drop half. */ 1372 median_ticks = ip_reass_ttl_decr(median_ticks); 1373 1374} 1375 1376/* 1377 * IP timer processing; 1378 * if a timer expires on a reassembly 1379 * queue, discard it. 1380 */ 1381void 1382ip_slowtimo(void) 1383{ 1384 static u_int dropscanidx = 0; 1385 u_int i; 1386 u_int median_ttl; 1387 int s = splsoftnet(); 1388 1389 IPQ_LOCK(); 1390 1391 /* Age TTL of all fragments by 1 tick .*/ 1392 median_ttl = ip_reass_ttl_decr(1); 1393 1394 /* make sure fragment limit is up-to-date */ 1395 CHECK_NMBCLUSTER_PARAMS(); 1396 1397 /* If we have too many fragments, drop the older half. */ 1398 if (ip_nfrags > ip_maxfrags) 1399 ip_reass_ttl_decr(median_ttl); 1400 1401 /* 1402 * If we are over the maximum number of fragmented packets 1403 * (due to the limit being lowered), drain off 1404 * enough to get down to the new limit. Start draining 1405 * from the reassembly hashqueue most recently drained. 1406 */ 1407 if (ip_maxfragpackets < 0) 1408 ; 1409 else { 1410 int wrapped = 0; 1411 1412 i = dropscanidx; 1413 while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) { 1414 while (LIST_FIRST(&ipq[i]) != NULL) 1415 ip_freef(LIST_FIRST(&ipq[i])); 1416 if (++i >= IPREASS_NHASH) { 1417 i = 0; 1418 } 1419 /* 1420 * Dont scan forever even if fragment counters are 1421 * wrong: stop after scanning entire reassembly queue. 1422 */ 1423 if (i == dropscanidx) 1424 wrapped = 1; 1425 } 1426 dropscanidx = i; 1427 } 1428 IPQ_UNLOCK(); 1429 splx(s); 1430} 1431 1432/* 1433 * Drain off all datagram fragments. 1434 */ 1435void 1436ip_drain(void) 1437{ 1438 1439 /* 1440 * We may be called from a device's interrupt context. If 1441 * the ipq is already busy, just bail out now. 1442 */ 1443 if (ipq_lock_try() == 0) 1444 return; 1445 1446 /* 1447 * Drop half the total fragments now. If more mbufs are needed, 1448 * we will be called again soon. 1449 */ 1450 ip_reass_drophalf(); 1451 1452 IPQ_UNLOCK(); 1453} 1454 1455/* 1456 * Do option processing on a datagram, 1457 * possibly discarding it if bad options are encountered, 1458 * or forwarding it if source-routed. 1459 * Returns 1 if packet has been forwarded/freed, 1460 * 0 if the packet should be processed further. 1461 */ 1462int 1463ip_dooptions(struct mbuf *m) 1464{ 1465 struct ip *ip = mtod(m, struct ip *); 1466 u_char *cp, *cp0; 1467 struct ip_timestamp *ipt; 1468 struct in_ifaddr *ia; 1469 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1470 struct in_addr dst; 1471 n_time ntime; 1472 1473 dst = ip->ip_dst; 1474 cp = (u_char *)(ip + 1); 1475 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1476 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1477 opt = cp[IPOPT_OPTVAL]; 1478 if (opt == IPOPT_EOL) 1479 break; 1480 if (opt == IPOPT_NOP) 1481 optlen = 1; 1482 else { 1483 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1484 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1485 goto bad; 1486 } 1487 optlen = cp[IPOPT_OLEN]; 1488 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1489 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1490 goto bad; 1491 } 1492 } 1493 switch (opt) { 1494 1495 default: 1496 break; 1497 1498 /* 1499 * Source routing with record. 1500 * Find interface with current destination address. 1501 * If none on this machine then drop if strictly routed, 1502 * or do nothing if loosely routed. 1503 * Record interface address and bring up next address 1504 * component. If strictly routed make sure next 1505 * address is on directly accessible net. 1506 */ 1507 case IPOPT_LSRR: 1508 case IPOPT_SSRR: 1509 if (ip_allowsrcrt == 0) { 1510 type = ICMP_UNREACH; 1511 code = ICMP_UNREACH_NET_PROHIB; 1512 goto bad; 1513 } 1514 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1515 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1516 goto bad; 1517 } 1518 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1519 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1520 goto bad; 1521 } 1522 ipaddr.sin_addr = ip->ip_dst; 1523 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1524 if (ia == 0) { 1525 if (opt == IPOPT_SSRR) { 1526 type = ICMP_UNREACH; 1527 code = ICMP_UNREACH_SRCFAIL; 1528 goto bad; 1529 } 1530 /* 1531 * Loose routing, and not at next destination 1532 * yet; nothing to do except forward. 1533 */ 1534 break; 1535 } 1536 off--; /* 0 origin */ 1537 if ((off + sizeof(struct in_addr)) > optlen) { 1538 /* 1539 * End of source route. Should be for us. 1540 */ 1541 save_rte(cp, ip->ip_src); 1542 break; 1543 } 1544 /* 1545 * locate outgoing interface 1546 */ 1547 bcopy((void *)(cp + off), (void *)&ipaddr.sin_addr, 1548 sizeof(ipaddr.sin_addr)); 1549 if (opt == IPOPT_SSRR) 1550 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); 1551 else 1552 ia = ip_rtaddr(ipaddr.sin_addr); 1553 if (ia == 0) { 1554 type = ICMP_UNREACH; 1555 code = ICMP_UNREACH_SRCFAIL; 1556 goto bad; 1557 } 1558 ip->ip_dst = ipaddr.sin_addr; 1559 bcopy((void *)&ia->ia_addr.sin_addr, 1560 (void *)(cp + off), sizeof(struct in_addr)); 1561 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1562 /* 1563 * Let ip_intr's mcast routing check handle mcast pkts 1564 */ 1565 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1566 break; 1567 1568 case IPOPT_RR: 1569 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1570 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1571 goto bad; 1572 } 1573 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1574 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1575 goto bad; 1576 } 1577 /* 1578 * If no space remains, ignore. 1579 */ 1580 off--; /* 0 origin */ 1581 if ((off + sizeof(struct in_addr)) > optlen) 1582 break; 1583 bcopy((void *)(&ip->ip_dst), (void *)&ipaddr.sin_addr, 1584 sizeof(ipaddr.sin_addr)); 1585 /* 1586 * locate outgoing interface; if we're the destination, 1587 * use the incoming interface (should be same). 1588 */ 1589 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1590 == NULL && 1591 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1592 type = ICMP_UNREACH; 1593 code = ICMP_UNREACH_HOST; 1594 goto bad; 1595 } 1596 bcopy((void *)&ia->ia_addr.sin_addr, 1597 (void *)(cp + off), sizeof(struct in_addr)); 1598 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1599 break; 1600 1601 case IPOPT_TS: 1602 code = cp - (u_char *)ip; 1603 ipt = (struct ip_timestamp *)cp; 1604 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1605 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1606 goto bad; 1607 } 1608 if (ipt->ipt_ptr < 5) { 1609 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1610 goto bad; 1611 } 1612 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1613 if (++ipt->ipt_oflw == 0) { 1614 code = (u_char *)&ipt->ipt_ptr - 1615 (u_char *)ip; 1616 goto bad; 1617 } 1618 break; 1619 } 1620 cp0 = (cp + ipt->ipt_ptr - 1); 1621 switch (ipt->ipt_flg) { 1622 1623 case IPOPT_TS_TSONLY: 1624 break; 1625 1626 case IPOPT_TS_TSANDADDR: 1627 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1628 sizeof(struct in_addr) > ipt->ipt_len) { 1629 code = (u_char *)&ipt->ipt_ptr - 1630 (u_char *)ip; 1631 goto bad; 1632 } 1633 ipaddr.sin_addr = dst; 1634 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1635 m->m_pkthdr.rcvif)); 1636 if (ia == 0) 1637 continue; 1638 bcopy(&ia->ia_addr.sin_addr, 1639 cp0, sizeof(struct in_addr)); 1640 ipt->ipt_ptr += sizeof(struct in_addr); 1641 break; 1642 1643 case IPOPT_TS_PRESPEC: 1644 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1645 sizeof(struct in_addr) > ipt->ipt_len) { 1646 code = (u_char *)&ipt->ipt_ptr - 1647 (u_char *)ip; 1648 goto bad; 1649 } 1650 bcopy(cp0, &ipaddr.sin_addr, 1651 sizeof(struct in_addr)); 1652 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1653 == NULL) 1654 continue; 1655 ipt->ipt_ptr += sizeof(struct in_addr); 1656 break; 1657 1658 default: 1659 /* XXX can't take &ipt->ipt_flg */ 1660 code = (u_char *)&ipt->ipt_ptr - 1661 (u_char *)ip + 1; 1662 goto bad; 1663 } 1664 ntime = iptime(); 1665 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1666 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1667 sizeof(n_time)); 1668 ipt->ipt_ptr += sizeof(n_time); 1669 } 1670 } 1671 if (forward) { 1672 if (ip_forwsrcrt == 0) { 1673 type = ICMP_UNREACH; 1674 code = ICMP_UNREACH_SRCFAIL; 1675 goto bad; 1676 } 1677 ip_forward(m, 1); 1678 return (1); 1679 } 1680 return (0); 1681bad: 1682 icmp_error(m, type, code, 0, 0); 1683 ipstat.ips_badoptions++; 1684 return (1); 1685} 1686 1687/* 1688 * Given address of next destination (final or next hop), 1689 * return internet address info of interface to be used to get there. 1690 */ 1691struct in_ifaddr * 1692ip_rtaddr(struct in_addr dst) 1693{ 1694 struct rtentry *rt; 1695 union { 1696 struct sockaddr dst; 1697 struct sockaddr_in dst4; 1698 } u; 1699 1700 sockaddr_in_init(&u.dst4, &dst, 0); 1701 1702 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) 1703 return NULL; 1704 1705 return ifatoia(rt->rt_ifa); 1706} 1707 1708/* 1709 * Save incoming source route for use in replies, 1710 * to be picked up later by ip_srcroute if the receiver is interested. 1711 */ 1712void 1713save_rte(u_char *option, struct in_addr dst) 1714{ 1715 unsigned olen; 1716 1717 olen = option[IPOPT_OLEN]; 1718#ifdef DIAGNOSTIC 1719 if (ipprintfs) 1720 printf("save_rte: olen %d\n", olen); 1721#endif /* 0 */ 1722 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1723 return; 1724 bcopy((void *)option, (void *)ip_srcrt.srcopt, olen); 1725 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1726 ip_srcrt.dst = dst; 1727} 1728 1729/* 1730 * Retrieve incoming source route for use in replies, 1731 * in the same form used by setsockopt. 1732 * The first hop is placed before the options, will be removed later. 1733 */ 1734struct mbuf * 1735ip_srcroute(void) 1736{ 1737 struct in_addr *p, *q; 1738 struct mbuf *m; 1739 1740 if (ip_nhops == 0) 1741 return NULL; 1742 m = m_get(M_DONTWAIT, MT_SOOPTS); 1743 if (m == 0) 1744 return NULL; 1745 1746 MCLAIM(m, &inetdomain.dom_mowner); 1747#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1748 1749 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1750 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1751 OPTSIZ; 1752#ifdef DIAGNOSTIC 1753 if (ipprintfs) 1754 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1755#endif 1756 1757 /* 1758 * First save first hop for return route 1759 */ 1760 p = &ip_srcrt.route[ip_nhops - 1]; 1761 *(mtod(m, struct in_addr *)) = *p--; 1762#ifdef DIAGNOSTIC 1763 if (ipprintfs) 1764 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1765#endif 1766 1767 /* 1768 * Copy option fields and padding (nop) to mbuf. 1769 */ 1770 ip_srcrt.nop = IPOPT_NOP; 1771 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1772 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop, 1773 OPTSIZ); 1774 q = (struct in_addr *)(mtod(m, char *) + 1775 sizeof(struct in_addr) + OPTSIZ); 1776#undef OPTSIZ 1777 /* 1778 * Record return path as an IP source route, 1779 * reversing the path (pointers are now aligned). 1780 */ 1781 while (p >= ip_srcrt.route) { 1782#ifdef DIAGNOSTIC 1783 if (ipprintfs) 1784 printf(" %x", ntohl(q->s_addr)); 1785#endif 1786 *q++ = *p--; 1787 } 1788 /* 1789 * Last hop goes to final destination. 1790 */ 1791 *q = ip_srcrt.dst; 1792#ifdef DIAGNOSTIC 1793 if (ipprintfs) 1794 printf(" %x\n", ntohl(q->s_addr)); 1795#endif 1796 return (m); 1797} 1798 1799const int inetctlerrmap[PRC_NCMDS] = { 1800 [PRC_MSGSIZE] = EMSGSIZE, 1801 [PRC_HOSTDEAD] = EHOSTDOWN, 1802 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1803 [PRC_UNREACH_NET] = EHOSTUNREACH, 1804 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1805 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1806 [PRC_UNREACH_PORT] = ECONNREFUSED, 1807 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1808 [PRC_PARAMPROB] = ENOPROTOOPT, 1809}; 1810 1811/* 1812 * Forward a packet. If some error occurs return the sender 1813 * an icmp packet. Note we can't always generate a meaningful 1814 * icmp message because icmp doesn't have a large enough repertoire 1815 * of codes and types. 1816 * 1817 * If not forwarding, just drop the packet. This could be confusing 1818 * if ipforwarding was zero but some routing protocol was advancing 1819 * us as a gateway to somewhere. However, we must let the routing 1820 * protocol deal with that. 1821 * 1822 * The srcrt parameter indicates whether the packet is being forwarded 1823 * via a source route. 1824 */ 1825void 1826ip_forward(struct mbuf *m, int srcrt) 1827{ 1828 struct ip *ip = mtod(m, struct ip *); 1829 struct rtentry *rt; 1830 int error, type = 0, code = 0, destmtu = 0; 1831 struct mbuf *mcopy; 1832 n_long dest; 1833 union { 1834 struct sockaddr dst; 1835 struct sockaddr_in dst4; 1836 } u; 1837 1838 /* 1839 * We are now in the output path. 1840 */ 1841 MCLAIM(m, &ip_tx_mowner); 1842 1843 /* 1844 * Clear any in-bound checksum flags for this packet. 1845 */ 1846 m->m_pkthdr.csum_flags = 0; 1847 1848 dest = 0; 1849#ifdef DIAGNOSTIC 1850 if (ipprintfs) { 1851 printf("forward: src %s ", inet_ntoa(ip->ip_src)); 1852 printf("dst %s ttl %x\n", inet_ntoa(ip->ip_dst), ip->ip_ttl); 1853 } 1854#endif 1855 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1856 ipstat.ips_cantforward++; 1857 m_freem(m); 1858 return; 1859 } 1860 if (ip->ip_ttl <= IPTTLDEC) { 1861 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1862 return; 1863 } 1864 1865 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1866 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) { 1867 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1868 return; 1869 } 1870 1871 /* 1872 * Save at most 68 bytes of the packet in case 1873 * we need to generate an ICMP message to the src. 1874 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1875 */ 1876 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1877 if (mcopy) 1878 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1879 1880 ip->ip_ttl -= IPTTLDEC; 1881 1882 /* 1883 * If forwarding packet using same interface that it came in on, 1884 * perhaps should send a redirect to sender to shortcut a hop. 1885 * Only send redirect if source is sending directly to us, 1886 * and if packet was not source routed (or has any options). 1887 * Also, don't send redirect if forwarding using a default route 1888 * or a route modified by a redirect. 1889 */ 1890 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1891 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1892 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1893 ipsendredirects && !srcrt) { 1894 if (rt->rt_ifa && 1895 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1896 ifatoia(rt->rt_ifa)->ia_subnet) { 1897 if (rt->rt_flags & RTF_GATEWAY) 1898 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1899 else 1900 dest = ip->ip_dst.s_addr; 1901 /* 1902 * Router requirements says to only send host 1903 * redirects. 1904 */ 1905 type = ICMP_REDIRECT; 1906 code = ICMP_REDIRECT_HOST; 1907#ifdef DIAGNOSTIC 1908 if (ipprintfs) 1909 printf("redirect (%d) to %x\n", code, 1910 (u_int32_t)dest); 1911#endif 1912 } 1913 } 1914 1915 error = ip_output(m, NULL, &ipforward_rt, 1916 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1917 (struct ip_moptions *)NULL, (struct socket *)NULL); 1918 1919 if (error) 1920 ipstat.ips_cantforward++; 1921 else { 1922 ipstat.ips_forward++; 1923 if (type) 1924 ipstat.ips_redirectsent++; 1925 else { 1926 if (mcopy) { 1927#ifdef GATEWAY 1928 if (mcopy->m_flags & M_CANFASTFWD) 1929 ipflow_create(&ipforward_rt, mcopy); 1930#endif 1931 m_freem(mcopy); 1932 } 1933 return; 1934 } 1935 } 1936 if (mcopy == NULL) 1937 return; 1938 1939 switch (error) { 1940 1941 case 0: /* forwarded, but need redirect */ 1942 /* type, code set above */ 1943 break; 1944 1945 case ENETUNREACH: /* shouldn't happen, checked above */ 1946 case EHOSTUNREACH: 1947 case ENETDOWN: 1948 case EHOSTDOWN: 1949 default: 1950 type = ICMP_UNREACH; 1951 code = ICMP_UNREACH_HOST; 1952 break; 1953 1954 case EMSGSIZE: 1955 type = ICMP_UNREACH; 1956 code = ICMP_UNREACH_NEEDFRAG; 1957 1958 if ((rt = rtcache_validate(&ipforward_rt)) != NULL) { 1959 1960#if defined(IPSEC) || defined(FAST_IPSEC) 1961 /* 1962 * If the packet is routed over IPsec tunnel, tell the 1963 * originator the tunnel MTU. 1964 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1965 * XXX quickhack!!! 1966 */ 1967 1968 struct secpolicy *sp; 1969 int ipsecerror; 1970 size_t ipsechdr; 1971 struct route *ro; 1972 1973 sp = ipsec4_getpolicybyaddr(mcopy, 1974 IPSEC_DIR_OUTBOUND, IP_FORWARDING, 1975 &ipsecerror); 1976#endif 1977 1978 destmtu = rt->rt_ifp->if_mtu; 1979#if defined(IPSEC) || defined(FAST_IPSEC) 1980 if (sp != NULL) { 1981 /* count IPsec header size */ 1982 ipsechdr = ipsec4_hdrsiz(mcopy, 1983 IPSEC_DIR_OUTBOUND, NULL); 1984 1985 /* 1986 * find the correct route for outer IPv4 1987 * header, compute tunnel MTU. 1988 */ 1989 1990 if (sp->req != NULL 1991 && sp->req->sav != NULL 1992 && sp->req->sav->sah != NULL) { 1993 ro = &sp->req->sav->sah->sa_route; 1994 if (rt && rt->rt_ifp) { 1995 destmtu = 1996 rt->rt_rmx.rmx_mtu ? 1997 rt->rt_rmx.rmx_mtu : 1998 rt->rt_ifp->if_mtu; 1999 destmtu -= ipsechdr; 2000 } 2001 } 2002 2003#ifdef IPSEC 2004 key_freesp(sp); 2005#else 2006 KEY_FREESP(&sp); 2007#endif 2008 } 2009#endif /*defined(IPSEC) || defined(FAST_IPSEC)*/ 2010 } 2011 ipstat.ips_cantfrag++; 2012 break; 2013 2014 case ENOBUFS: 2015#if 1 2016 /* 2017 * a router should not generate ICMP_SOURCEQUENCH as 2018 * required in RFC1812 Requirements for IP Version 4 Routers. 2019 * source quench could be a big problem under DoS attacks, 2020 * or if the underlying interface is rate-limited. 2021 */ 2022 if (mcopy) 2023 m_freem(mcopy); 2024 return; 2025#else 2026 type = ICMP_SOURCEQUENCH; 2027 code = 0; 2028 break; 2029#endif 2030 } 2031 icmp_error(mcopy, type, code, dest, destmtu); 2032} 2033 2034void 2035ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2036 struct mbuf *m) 2037{ 2038 2039 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2040 struct timeval tv; 2041 2042 microtime(&tv); 2043 *mp = sbcreatecontrol((void *) &tv, sizeof(tv), 2044 SCM_TIMESTAMP, SOL_SOCKET); 2045 if (*mp) 2046 mp = &(*mp)->m_next; 2047 } 2048 if (inp->inp_flags & INP_RECVDSTADDR) { 2049 *mp = sbcreatecontrol((void *) &ip->ip_dst, 2050 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2051 if (*mp) 2052 mp = &(*mp)->m_next; 2053 } 2054#ifdef notyet 2055 /* 2056 * XXX 2057 * Moving these out of udp_input() made them even more broken 2058 * than they already were. 2059 * - fenner@parc.xerox.com 2060 */ 2061 /* options were tossed already */ 2062 if (inp->inp_flags & INP_RECVOPTS) { 2063 *mp = sbcreatecontrol((void *) opts_deleted_above, 2064 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2065 if (*mp) 2066 mp = &(*mp)->m_next; 2067 } 2068 /* ip_srcroute doesn't do what we want here, need to fix */ 2069 if (inp->inp_flags & INP_RECVRETOPTS) { 2070 *mp = sbcreatecontrol((void *) ip_srcroute(), 2071 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2072 if (*mp) 2073 mp = &(*mp)->m_next; 2074 } 2075#endif 2076 if (inp->inp_flags & INP_RECVIF) { 2077 struct sockaddr_dl sdl; 2078 2079 sockaddr_dl_init(&sdl, sizeof(sdl), 2080 (m->m_pkthdr.rcvif != NULL) 2081 ? m->m_pkthdr.rcvif->if_index 2082 : 0, 2083 0, NULL, 0, NULL, 0); 2084 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 2085 if (*mp) 2086 mp = &(*mp)->m_next; 2087 } 2088} 2089 2090/* 2091 * sysctl helper routine for net.inet.ip.forwsrcrt. 2092 */ 2093static int 2094sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 2095{ 2096 int error, tmp; 2097 struct sysctlnode node; 2098 2099 node = *rnode; 2100 tmp = ip_forwsrcrt; 2101 node.sysctl_data = &tmp; 2102 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2103 if (error || newp == NULL) 2104 return (error); 2105 2106 if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 2107 0, NULL, NULL, NULL)) 2108 return (EPERM); 2109 2110 ip_forwsrcrt = tmp; 2111 2112 return (0); 2113} 2114 2115/* 2116 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 2117 * range of the new value and tweaks timers if it changes. 2118 */ 2119static int 2120sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 2121{ 2122 int error, tmp; 2123 struct sysctlnode node; 2124 2125 node = *rnode; 2126 tmp = ip_mtudisc_timeout; 2127 node.sysctl_data = &tmp; 2128 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2129 if (error || newp == NULL) 2130 return (error); 2131 if (tmp < 0) 2132 return (EINVAL); 2133 2134 ip_mtudisc_timeout = tmp; 2135 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 2136 2137 return (0); 2138} 2139 2140#ifdef GATEWAY 2141/* 2142 * sysctl helper routine for net.inet.ip.maxflows. 2143 */ 2144static int 2145sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS) 2146{ 2147 int s; 2148 2149 s = sysctl_lookup(SYSCTLFN_CALL(rnode)); 2150 if (s || newp == NULL) 2151 return (s); 2152 2153 s = splsoftnet(); 2154 ipflow_reap(0); 2155 splx(s); 2156 2157 return (0); 2158} 2159 2160static int 2161sysctl_net_inet_ip_hashsize(SYSCTLFN_ARGS) 2162{ 2163 int error, tmp; 2164 struct sysctlnode node; 2165 2166 node = *rnode; 2167 tmp = ip_hashsize; 2168 node.sysctl_data = &tmp; 2169 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2170 if (error || newp == NULL) 2171 return (error); 2172 2173 if ((tmp & (tmp - 1)) == 0 && tmp != 0) { 2174 /* 2175 * Can only fail due to malloc() 2176 */ 2177 if (ipflow_invalidate_all(tmp)) 2178 return ENOMEM; 2179 } else { 2180 /* 2181 * EINVAL if not a power of 2 2182 */ 2183 return EINVAL; 2184 } 2185 2186 return (0); 2187} 2188#endif /* GATEWAY */ 2189 2190 2191SYSCTL_SETUP(sysctl_net_inet_ip_setup, "sysctl net.inet.ip subtree setup") 2192{ 2193 extern int subnetsarelocal, hostzeroisbroadcast; 2194 2195 sysctl_createv(clog, 0, NULL, NULL, 2196 CTLFLAG_PERMANENT, 2197 CTLTYPE_NODE, "net", NULL, 2198 NULL, 0, NULL, 0, 2199 CTL_NET, CTL_EOL); 2200 sysctl_createv(clog, 0, NULL, NULL, 2201 CTLFLAG_PERMANENT, 2202 CTLTYPE_NODE, "inet", 2203 SYSCTL_DESCR("PF_INET related settings"), 2204 NULL, 0, NULL, 0, 2205 CTL_NET, PF_INET, CTL_EOL); 2206 sysctl_createv(clog, 0, NULL, NULL, 2207 CTLFLAG_PERMANENT, 2208 CTLTYPE_NODE, "ip", 2209 SYSCTL_DESCR("IPv4 related settings"), 2210 NULL, 0, NULL, 0, 2211 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 2212 2213 sysctl_createv(clog, 0, NULL, NULL, 2214 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2215 CTLTYPE_INT, "forwarding", 2216 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 2217 NULL, 0, &ipforwarding, 0, 2218 CTL_NET, PF_INET, IPPROTO_IP, 2219 IPCTL_FORWARDING, CTL_EOL); 2220 sysctl_createv(clog, 0, NULL, NULL, 2221 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2222 CTLTYPE_INT, "redirect", 2223 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 2224 NULL, 0, &ipsendredirects, 0, 2225 CTL_NET, PF_INET, IPPROTO_IP, 2226 IPCTL_SENDREDIRECTS, CTL_EOL); 2227 sysctl_createv(clog, 0, NULL, NULL, 2228 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2229 CTLTYPE_INT, "ttl", 2230 SYSCTL_DESCR("Default TTL for an INET datagram"), 2231 NULL, 0, &ip_defttl, 0, 2232 CTL_NET, PF_INET, IPPROTO_IP, 2233 IPCTL_DEFTTL, CTL_EOL); 2234#ifdef IPCTL_DEFMTU 2235 sysctl_createv(clog, 0, NULL, NULL, 2236 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 2237 CTLTYPE_INT, "mtu", 2238 SYSCTL_DESCR("Default MTA for an INET route"), 2239 NULL, 0, &ip_mtu, 0, 2240 CTL_NET, PF_INET, IPPROTO_IP, 2241 IPCTL_DEFMTU, CTL_EOL); 2242#endif /* IPCTL_DEFMTU */ 2243 sysctl_createv(clog, 0, NULL, NULL, 2244 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2245 CTLTYPE_INT, "forwsrcrt", 2246 SYSCTL_DESCR("Enable forwarding of source-routed " 2247 "datagrams"), 2248 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 2249 CTL_NET, PF_INET, IPPROTO_IP, 2250 IPCTL_FORWSRCRT, CTL_EOL); 2251 sysctl_createv(clog, 0, NULL, NULL, 2252 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2253 CTLTYPE_INT, "directed-broadcast", 2254 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 2255 NULL, 0, &ip_directedbcast, 0, 2256 CTL_NET, PF_INET, IPPROTO_IP, 2257 IPCTL_DIRECTEDBCAST, CTL_EOL); 2258 sysctl_createv(clog, 0, NULL, NULL, 2259 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2260 CTLTYPE_INT, "allowsrcrt", 2261 SYSCTL_DESCR("Accept source-routed datagrams"), 2262 NULL, 0, &ip_allowsrcrt, 0, 2263 CTL_NET, PF_INET, IPPROTO_IP, 2264 IPCTL_ALLOWSRCRT, CTL_EOL); 2265 sysctl_createv(clog, 0, NULL, NULL, 2266 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2267 CTLTYPE_INT, "subnetsarelocal", 2268 SYSCTL_DESCR("Whether logical subnets are considered " 2269 "local"), 2270 NULL, 0, &subnetsarelocal, 0, 2271 CTL_NET, PF_INET, IPPROTO_IP, 2272 IPCTL_SUBNETSARELOCAL, CTL_EOL); 2273 sysctl_createv(clog, 0, NULL, NULL, 2274 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2275 CTLTYPE_INT, "mtudisc", 2276 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 2277 NULL, 0, &ip_mtudisc, 0, 2278 CTL_NET, PF_INET, IPPROTO_IP, 2279 IPCTL_MTUDISC, CTL_EOL); 2280 sysctl_createv(clog, 0, NULL, NULL, 2281 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2282 CTLTYPE_INT, "anonportmin", 2283 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 2284 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 2285 CTL_NET, PF_INET, IPPROTO_IP, 2286 IPCTL_ANONPORTMIN, CTL_EOL); 2287 sysctl_createv(clog, 0, NULL, NULL, 2288 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2289 CTLTYPE_INT, "anonportmax", 2290 SYSCTL_DESCR("Highest ephemeral port number to assign"), 2291 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 2292 CTL_NET, PF_INET, IPPROTO_IP, 2293 IPCTL_ANONPORTMAX, CTL_EOL); 2294 sysctl_createv(clog, 0, NULL, NULL, 2295 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2296 CTLTYPE_INT, "mtudisctimeout", 2297 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 2298 sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0, 2299 CTL_NET, PF_INET, IPPROTO_IP, 2300 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 2301#ifdef GATEWAY 2302 sysctl_createv(clog, 0, NULL, NULL, 2303 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2304 CTLTYPE_INT, "maxflows", 2305 SYSCTL_DESCR("Number of flows for fast forwarding"), 2306 sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0, 2307 CTL_NET, PF_INET, IPPROTO_IP, 2308 IPCTL_MAXFLOWS, CTL_EOL); 2309 sysctl_createv(clog, 0, NULL, NULL, 2310 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2311 CTLTYPE_INT, "hashsize", 2312 SYSCTL_DESCR("Size of hash table for fast forwarding (IPv4)"), 2313 sysctl_net_inet_ip_hashsize, 0, &ip_hashsize, 0, 2314 CTL_NET, PF_INET, IPPROTO_IP, 2315 CTL_CREATE, CTL_EOL); 2316#endif /* GATEWAY */ 2317 sysctl_createv(clog, 0, NULL, NULL, 2318 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2319 CTLTYPE_INT, "hostzerobroadcast", 2320 SYSCTL_DESCR("All zeroes address is broadcast address"), 2321 NULL, 0, &hostzeroisbroadcast, 0, 2322 CTL_NET, PF_INET, IPPROTO_IP, 2323 IPCTL_HOSTZEROBROADCAST, CTL_EOL); 2324#if NGIF > 0 2325 sysctl_createv(clog, 0, NULL, NULL, 2326 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2327 CTLTYPE_INT, "gifttl", 2328 SYSCTL_DESCR("Default TTL for a gif tunnel datagram"), 2329 NULL, 0, &ip_gif_ttl, 0, 2330 CTL_NET, PF_INET, IPPROTO_IP, 2331 IPCTL_GIF_TTL, CTL_EOL); 2332#endif /* NGIF */ 2333#ifndef IPNOPRIVPORTS 2334 sysctl_createv(clog, 0, NULL, NULL, 2335 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2336 CTLTYPE_INT, "lowportmin", 2337 SYSCTL_DESCR("Lowest privileged ephemeral port number " 2338 "to assign"), 2339 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 2340 CTL_NET, PF_INET, IPPROTO_IP, 2341 IPCTL_LOWPORTMIN, CTL_EOL); 2342 sysctl_createv(clog, 0, NULL, NULL, 2343 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2344 CTLTYPE_INT, "lowportmax", 2345 SYSCTL_DESCR("Highest privileged ephemeral port number " 2346 "to assign"), 2347 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 2348 CTL_NET, PF_INET, IPPROTO_IP, 2349 IPCTL_LOWPORTMAX, CTL_EOL); 2350#endif /* IPNOPRIVPORTS */ 2351 sysctl_createv(clog, 0, NULL, NULL, 2352 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2353 CTLTYPE_INT, "maxfragpackets", 2354 SYSCTL_DESCR("Maximum number of fragments to retain for " 2355 "possible reassembly"), 2356 NULL, 0, &ip_maxfragpackets, 0, 2357 CTL_NET, PF_INET, IPPROTO_IP, 2358 IPCTL_MAXFRAGPACKETS, CTL_EOL); 2359#if NGRE > 0 2360 sysctl_createv(clog, 0, NULL, NULL, 2361 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2362 CTLTYPE_INT, "grettl", 2363 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 2364 NULL, 0, &ip_gre_ttl, 0, 2365 CTL_NET, PF_INET, IPPROTO_IP, 2366 IPCTL_GRE_TTL, CTL_EOL); 2367#endif /* NGRE */ 2368 sysctl_createv(clog, 0, NULL, NULL, 2369 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2370 CTLTYPE_INT, "checkinterface", 2371 SYSCTL_DESCR("Enable receive side of Strong ES model " 2372 "from RFC1122"), 2373 NULL, 0, &ip_checkinterface, 0, 2374 CTL_NET, PF_INET, IPPROTO_IP, 2375 IPCTL_CHECKINTERFACE, CTL_EOL); 2376 sysctl_createv(clog, 0, NULL, NULL, 2377 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2378 CTLTYPE_INT, "random_id", 2379 SYSCTL_DESCR("Assign random ip_id values"), 2380 NULL, 0, &ip_do_randomid, 0, 2381 CTL_NET, PF_INET, IPPROTO_IP, 2382 IPCTL_RANDOMID, CTL_EOL); 2383 sysctl_createv(clog, 0, NULL, NULL, 2384 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2385 CTLTYPE_INT, "do_loopback_cksum", 2386 SYSCTL_DESCR("Perform IP checksum on loopback"), 2387 NULL, 0, &ip_do_loopback_cksum, 0, 2388 CTL_NET, PF_INET, IPPROTO_IP, 2389 IPCTL_LOOPBACKCKSUM, CTL_EOL); 2390 sysctl_createv(clog, 0, NULL, NULL, 2391 CTLFLAG_PERMANENT, 2392 CTLTYPE_STRUCT, "stats", 2393 SYSCTL_DESCR("IP statistics"), 2394 NULL, 0, &ipstat, sizeof(ipstat), 2395 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 2396 CTL_EOL); 2397} 2398