ip_input.c revision 1.227
1/* $NetBSD: ip_input.c,v 1.227 2006/06/07 22:34:01 kardel 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 1.227 2006/06/07 22:34:01 kardel Exp $"); 102 103#include "opt_inet.h" 104#include "opt_gateway.h" 105#include "opt_pfil_hooks.h" 106#include "opt_ipsec.h" 107#include "opt_mrouting.h" 108#include "opt_mbuftrace.h" 109#include "opt_inet_csum.h" 110 111#include <sys/param.h> 112#include <sys/systm.h> 113#include <sys/malloc.h> 114#include <sys/mbuf.h> 115#include <sys/domain.h> 116#include <sys/protosw.h> 117#include <sys/socket.h> 118#include <sys/socketvar.h> 119#include <sys/errno.h> 120#include <sys/time.h> 121#include <sys/kernel.h> 122#include <sys/pool.h> 123#include <sys/sysctl.h> 124 125#include <net/if.h> 126#include <net/if_dl.h> 127#include <net/route.h> 128#include <net/pfil.h> 129 130#include <netinet/in.h> 131#include <netinet/in_systm.h> 132#include <netinet/ip.h> 133#include <netinet/in_pcb.h> 134#include <netinet/in_proto.h> 135#include <netinet/in_var.h> 136#include <netinet/ip_var.h> 137#include <netinet/ip_icmp.h> 138/* just for gif_ttl */ 139#include <netinet/in_gif.h> 140#include "gif.h" 141#include <net/if_gre.h> 142#include "gre.h" 143 144#ifdef MROUTING 145#include <netinet/ip_mroute.h> 146#endif 147 148#ifdef IPSEC 149#include <netinet6/ipsec.h> 150#include <netkey/key.h> 151#endif 152#ifdef FAST_IPSEC 153#include <netipsec/ipsec.h> 154#include <netipsec/key.h> 155#endif /* FAST_IPSEC*/ 156 157#ifndef IPFORWARDING 158#ifdef GATEWAY 159#define IPFORWARDING 1 /* forward IP packets not for us */ 160#else /* GATEWAY */ 161#define IPFORWARDING 0 /* don't forward IP packets not for us */ 162#endif /* GATEWAY */ 163#endif /* IPFORWARDING */ 164#ifndef IPSENDREDIRECTS 165#define IPSENDREDIRECTS 1 166#endif 167#ifndef IPFORWSRCRT 168#define IPFORWSRCRT 1 /* forward source-routed packets */ 169#endif 170#ifndef IPALLOWSRCRT 171#define IPALLOWSRCRT 1 /* allow source-routed packets */ 172#endif 173#ifndef IPMTUDISC 174#define IPMTUDISC 1 175#endif 176#ifndef IPMTUDISCTIMEOUT 177#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 178#endif 179 180/* 181 * Note: DIRECTED_BROADCAST is handled this way so that previous 182 * configuration using this option will Just Work. 183 */ 184#ifndef IPDIRECTEDBCAST 185#ifdef DIRECTED_BROADCAST 186#define IPDIRECTEDBCAST 1 187#else 188#define IPDIRECTEDBCAST 0 189#endif /* DIRECTED_BROADCAST */ 190#endif /* IPDIRECTEDBCAST */ 191int ipforwarding = IPFORWARDING; 192int ipsendredirects = IPSENDREDIRECTS; 193int ip_defttl = IPDEFTTL; 194int ip_forwsrcrt = IPFORWSRCRT; 195int ip_directedbcast = IPDIRECTEDBCAST; 196int ip_allowsrcrt = IPALLOWSRCRT; 197int ip_mtudisc = IPMTUDISC; 198int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 199#ifdef DIAGNOSTIC 200int ipprintfs = 0; 201#endif 202 203int ip_do_randomid = 0; 204 205/* 206 * XXX - Setting ip_checkinterface mostly implements the receive side of 207 * the Strong ES model described in RFC 1122, but since the routing table 208 * and transmit implementation do not implement the Strong ES model, 209 * setting this to 1 results in an odd hybrid. 210 * 211 * XXX - ip_checkinterface currently must be disabled if you use ipnat 212 * to translate the destination address to another local interface. 213 * 214 * XXX - ip_checkinterface must be disabled if you add IP aliases 215 * to the loopback interface instead of the interface where the 216 * packets for those addresses are received. 217 */ 218int ip_checkinterface = 0; 219 220 221struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 222 223int ipqmaxlen = IFQ_MAXLEN; 224u_long in_ifaddrhash; /* size of hash table - 1 */ 225int in_ifaddrentries; /* total number of addrs */ 226struct in_ifaddrhead in_ifaddrhead; 227struct in_ifaddrhashhead *in_ifaddrhashtbl; 228u_long in_multihash; /* size of hash table - 1 */ 229int in_multientries; /* total number of addrs */ 230struct in_multihashhead *in_multihashtbl; 231struct ifqueue ipintrq; 232struct ipstat ipstat; 233uint16_t ip_id; 234 235#ifdef PFIL_HOOKS 236struct pfil_head inet_pfil_hook; 237#endif 238 239/* 240 * Cached copy of nmbclusters. If nbclusters is different, 241 * recalculate IP parameters derived from nmbclusters. 242 */ 243static int ip_nmbclusters; /* copy of nmbclusters */ 244static void ip_nmbclusters_changed(void); /* recalc limits */ 245 246#define CHECK_NMBCLUSTER_PARAMS() \ 247do { \ 248 if (__predict_false(ip_nmbclusters != nmbclusters)) \ 249 ip_nmbclusters_changed(); \ 250} while (/*CONSTCOND*/0) 251 252/* IP datagram reassembly queues (hashed) */ 253#define IPREASS_NHASH_LOG2 6 254#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 255#define IPREASS_HMASK (IPREASS_NHASH - 1) 256#define IPREASS_HASH(x,y) \ 257 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 258struct ipqhead ipq[IPREASS_NHASH]; 259int ipq_locked; 260static int ip_nfragpackets; /* packets in reass queue */ 261static int ip_nfrags; /* total fragments in reass queues */ 262 263int ip_maxfragpackets = 200; /* limit on packets. XXX sysctl */ 264int ip_maxfrags; /* limit on fragments. XXX sysctl */ 265 266 267/* 268 * Additive-Increase/Multiplicative-Decrease (AIMD) strategy for 269 * IP reassembly queue buffer managment. 270 * 271 * We keep a count of total IP fragments (NB: not fragmented packets!) 272 * awaiting reassembly (ip_nfrags) and a limit (ip_maxfrags) on fragments. 273 * If ip_nfrags exceeds ip_maxfrags the limit, we drop half the 274 * total fragments in reassembly queues.This AIMD policy avoids 275 * repeatedly deleting single packets under heavy fragmentation load 276 * (e.g., from lossy NFS peers). 277 */ 278static u_int ip_reass_ttl_decr(u_int ticks); 279static void ip_reass_drophalf(void); 280 281 282static inline int ipq_lock_try(void); 283static inline void ipq_unlock(void); 284 285static inline int 286ipq_lock_try(void) 287{ 288 int s; 289 290 /* 291 * Use splvm() -- we're blocking things that would cause 292 * mbuf allocation. 293 */ 294 s = splvm(); 295 if (ipq_locked) { 296 splx(s); 297 return (0); 298 } 299 ipq_locked = 1; 300 splx(s); 301 return (1); 302} 303 304static inline void 305ipq_unlock(void) 306{ 307 int s; 308 309 s = splvm(); 310 ipq_locked = 0; 311 splx(s); 312} 313 314#ifdef DIAGNOSTIC 315#define IPQ_LOCK() \ 316do { \ 317 if (ipq_lock_try() == 0) { \ 318 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 319 panic("ipq_lock"); \ 320 } \ 321} while (/*CONSTCOND*/ 0) 322#define IPQ_LOCK_CHECK() \ 323do { \ 324 if (ipq_locked == 0) { \ 325 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 326 panic("ipq lock check"); \ 327 } \ 328} while (/*CONSTCOND*/ 0) 329#else 330#define IPQ_LOCK() (void) ipq_lock_try() 331#define IPQ_LOCK_CHECK() /* nothing */ 332#endif 333 334#define IPQ_UNLOCK() ipq_unlock() 335 336POOL_INIT(inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl", NULL); 337POOL_INIT(ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", NULL); 338 339#ifdef INET_CSUM_COUNTERS 340#include <sys/device.h> 341 342struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 343 NULL, "inet", "hwcsum bad"); 344struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 345 NULL, "inet", "hwcsum ok"); 346struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 347 NULL, "inet", "swcsum"); 348 349#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 350 351EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 352EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 353EVCNT_ATTACH_STATIC(ip_swcsum); 354 355#else 356 357#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 358 359#endif /* INET_CSUM_COUNTERS */ 360 361/* 362 * We need to save the IP options in case a protocol wants to respond 363 * to an incoming packet over the same route if the packet got here 364 * using IP source routing. This allows connection establishment and 365 * maintenance when the remote end is on a network that is not known 366 * to us. 367 */ 368int ip_nhops = 0; 369static struct ip_srcrt { 370 struct in_addr dst; /* final destination */ 371 char nop; /* one NOP to align */ 372 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 373 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 374} ip_srcrt; 375 376static void save_rte(u_char *, struct in_addr); 377 378#ifdef MBUFTRACE 379struct mowner ip_rx_mowner = { "internet", "rx" }; 380struct mowner ip_tx_mowner = { "internet", "tx" }; 381#endif 382 383/* 384 * Compute IP limits derived from the value of nmbclusters. 385 */ 386static void 387ip_nmbclusters_changed(void) 388{ 389 ip_maxfrags = nmbclusters / 4; 390 ip_nmbclusters = nmbclusters; 391} 392 393/* 394 * IP initialization: fill in IP protocol switch table. 395 * All protocols not implemented in kernel go to raw IP protocol handler. 396 */ 397void 398ip_init(void) 399{ 400 const struct protosw *pr; 401 int i; 402 403 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 404 if (pr == 0) 405 panic("ip_init"); 406 for (i = 0; i < IPPROTO_MAX; i++) 407 ip_protox[i] = pr - inetsw; 408 for (pr = inetdomain.dom_protosw; 409 pr < inetdomain.dom_protoswNPROTOSW; pr++) 410 if (pr->pr_domain->dom_family == PF_INET && 411 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 412 ip_protox[pr->pr_protocol] = pr - inetsw; 413 414 for (i = 0; i < IPREASS_NHASH; i++) 415 LIST_INIT(&ipq[i]); 416 417 ip_id = time_second & 0xfffff; 418 419 ipintrq.ifq_maxlen = ipqmaxlen; 420 ip_nmbclusters_changed(); 421 422 TAILQ_INIT(&in_ifaddrhead); 423 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 424 M_WAITOK, &in_ifaddrhash); 425 in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IPMADDR, 426 M_WAITOK, &in_multihash); 427 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 428#ifdef GATEWAY 429 ipflow_init(); 430#endif 431 432#ifdef PFIL_HOOKS 433 /* Register our Packet Filter hook. */ 434 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 435 inet_pfil_hook.ph_af = AF_INET; 436 i = pfil_head_register(&inet_pfil_hook); 437 if (i != 0) 438 printf("ip_init: WARNING: unable to register pfil hook, " 439 "error %d\n", i); 440#endif /* PFIL_HOOKS */ 441 442#ifdef MBUFTRACE 443 MOWNER_ATTACH(&ip_tx_mowner); 444 MOWNER_ATTACH(&ip_rx_mowner); 445#endif /* MBUFTRACE */ 446} 447 448struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 449struct route ipforward_rt; 450 451/* 452 * IP software interrupt routine 453 */ 454void 455ipintr(void) 456{ 457 int s; 458 struct mbuf *m; 459 460 while (1) { 461 s = splnet(); 462 IF_DEQUEUE(&ipintrq, m); 463 splx(s); 464 if (m == 0) 465 return; 466 MCLAIM(m, &ip_rx_mowner); 467 ip_input(m); 468 } 469} 470 471/* 472 * Ip input routine. Checksum and byte swap header. If fragmented 473 * try to reassemble. Process options. Pass to next level. 474 */ 475void 476ip_input(struct mbuf *m) 477{ 478 struct ip *ip = NULL; 479 struct ipq *fp; 480 struct in_ifaddr *ia; 481 struct ifaddr *ifa; 482 struct ipqent *ipqe; 483 int hlen = 0, mff, len; 484 int downmatch; 485 int checkif; 486 int srcrt = 0; 487 u_int hash; 488#ifdef FAST_IPSEC 489 struct m_tag *mtag; 490 struct tdb_ident *tdbi; 491 struct secpolicy *sp; 492 int s, error; 493#endif /* FAST_IPSEC */ 494 495 MCLAIM(m, &ip_rx_mowner); 496#ifdef DIAGNOSTIC 497 if ((m->m_flags & M_PKTHDR) == 0) 498 panic("ipintr no HDR"); 499#endif 500 501 /* 502 * If no IP addresses have been set yet but the interfaces 503 * are receiving, can't do anything with incoming packets yet. 504 */ 505 if (TAILQ_FIRST(&in_ifaddrhead) == 0) 506 goto bad; 507 ipstat.ips_total++; 508 /* 509 * If the IP header is not aligned, slurp it up into a new 510 * mbuf with space for link headers, in the event we forward 511 * it. Otherwise, if it is aligned, make sure the entire 512 * base IP header is in the first mbuf of the chain. 513 */ 514 if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) { 515 if ((m = m_copyup(m, sizeof(struct ip), 516 (max_linkhdr + 3) & ~3)) == NULL) { 517 /* XXXJRT new stat, please */ 518 ipstat.ips_toosmall++; 519 return; 520 } 521 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 522 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 523 ipstat.ips_toosmall++; 524 return; 525 } 526 } 527 ip = mtod(m, struct ip *); 528 if (ip->ip_v != IPVERSION) { 529 ipstat.ips_badvers++; 530 goto bad; 531 } 532 hlen = ip->ip_hl << 2; 533 if (hlen < sizeof(struct ip)) { /* minimum header length */ 534 ipstat.ips_badhlen++; 535 goto bad; 536 } 537 if (hlen > m->m_len) { 538 if ((m = m_pullup(m, hlen)) == 0) { 539 ipstat.ips_badhlen++; 540 return; 541 } 542 ip = mtod(m, struct ip *); 543 } 544 545 /* 546 * RFC1122: packets with a multicast source address are 547 * not allowed. 548 */ 549 if (IN_MULTICAST(ip->ip_src.s_addr)) { 550 ipstat.ips_badaddr++; 551 goto bad; 552 } 553 554 /* 127/8 must not appear on wire - RFC1122 */ 555 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 556 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 557 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 558 ipstat.ips_badaddr++; 559 goto bad; 560 } 561 } 562 563 switch (m->m_pkthdr.csum_flags & 564 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) | 565 M_CSUM_IPv4_BAD)) { 566 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 567 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 568 goto badcsum; 569 570 case M_CSUM_IPv4: 571 /* Checksum was okay. */ 572 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 573 break; 574 575 default: 576 /* 577 * Must compute it ourselves. Maybe skip checksum on 578 * loopback interfaces. 579 */ 580 if (__predict_true(!(m->m_pkthdr.rcvif->if_flags & 581 IFF_LOOPBACK) || ip_do_loopback_cksum)) { 582 INET_CSUM_COUNTER_INCR(&ip_swcsum); 583 if (in_cksum(m, hlen) != 0) 584 goto badcsum; 585 } 586 break; 587 } 588 589 /* Retrieve the packet length. */ 590 len = ntohs(ip->ip_len); 591 592 /* 593 * Check for additional length bogosity 594 */ 595 if (len < hlen) { 596 ipstat.ips_badlen++; 597 goto bad; 598 } 599 600 /* 601 * Check that the amount of data in the buffers 602 * is as at least much as the IP header would have us expect. 603 * Trim mbufs if longer than we expect. 604 * Drop packet if shorter than we expect. 605 */ 606 if (m->m_pkthdr.len < len) { 607 ipstat.ips_tooshort++; 608 goto bad; 609 } 610 if (m->m_pkthdr.len > len) { 611 if (m->m_len == m->m_pkthdr.len) { 612 m->m_len = len; 613 m->m_pkthdr.len = len; 614 } else 615 m_adj(m, len - m->m_pkthdr.len); 616 } 617 618#if defined(IPSEC) 619 /* ipflow (IP fast forwarding) is not compatible with IPsec. */ 620 m->m_flags &= ~M_CANFASTFWD; 621#else 622 /* 623 * Assume that we can create a fast-forward IP flow entry 624 * based on this packet. 625 */ 626 m->m_flags |= M_CANFASTFWD; 627#endif 628 629#ifdef PFIL_HOOKS 630 /* 631 * Run through list of hooks for input packets. If there are any 632 * filters which require that additional packets in the flow are 633 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 634 * Note that filters must _never_ set this flag, as another filter 635 * in the list may have previously cleared it. 636 */ 637 /* 638 * let ipfilter look at packet on the wire, 639 * not the decapsulated packet. 640 */ 641#ifdef IPSEC 642 if (!ipsec_getnhist(m)) 643#elif defined(FAST_IPSEC) 644 if (!ipsec_indone(m)) 645#else 646 if (1) 647#endif 648 { 649 struct in_addr odst; 650 651 odst = ip->ip_dst; 652 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 653 PFIL_IN) != 0) 654 return; 655 if (m == NULL) 656 return; 657 ip = mtod(m, struct ip *); 658 hlen = ip->ip_hl << 2; 659 /* 660 * XXX The setting of "srcrt" here is to prevent ip_forward() 661 * from generating ICMP redirects for packets that have 662 * been redirected by a hook back out on to the same LAN that 663 * they came from and is not an indication that the packet 664 * is being inffluenced by source routing options. This 665 * allows things like 666 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 667 * where tlp0 is both on the 1.1.1.0/24 network and is the 668 * default route for hosts on 1.1.1.0/24. Of course this 669 * also requires a "map tlp0 ..." to complete the story. 670 * One might argue whether or not this kind of network config. 671 * should be supported in this manner... 672 */ 673 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 674 } 675#endif /* PFIL_HOOKS */ 676 677#ifdef ALTQ 678 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 679 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 680 /* packet dropped by traffic conditioner */ 681 return; 682 } 683#endif 684 685 /* 686 * Process options and, if not destined for us, 687 * ship it on. ip_dooptions returns 1 when an 688 * error was detected (causing an icmp message 689 * to be sent and the original packet to be freed). 690 */ 691 ip_nhops = 0; /* for source routed packets */ 692 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 693 return; 694 695 /* 696 * Enable a consistency check between the destination address 697 * and the arrival interface for a unicast packet (the RFC 1122 698 * strong ES model) if IP forwarding is disabled and the packet 699 * is not locally generated. 700 * 701 * XXX - Checking also should be disabled if the destination 702 * address is ipnat'ed to a different interface. 703 * 704 * XXX - Checking is incompatible with IP aliases added 705 * to the loopback interface instead of the interface where 706 * the packets are received. 707 * 708 * XXX - We need to add a per ifaddr flag for this so that 709 * we get finer grain control. 710 */ 711 checkif = ip_checkinterface && (ipforwarding == 0) && 712 (m->m_pkthdr.rcvif != NULL) && 713 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0); 714 715 /* 716 * Check our list of addresses, to see if the packet is for us. 717 * 718 * Traditional 4.4BSD did not consult IFF_UP at all. 719 * The behavior here is to treat addresses on !IFF_UP interface 720 * as not mine. 721 */ 722 downmatch = 0; 723 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 724 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 725 if (checkif && ia->ia_ifp != m->m_pkthdr.rcvif) 726 continue; 727 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 728 break; 729 else 730 downmatch++; 731 } 732 } 733 if (ia != NULL) 734 goto ours; 735 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 736 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) { 737 if (ifa->ifa_addr->sa_family != AF_INET) 738 continue; 739 ia = ifatoia(ifa); 740 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 741 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 742 /* 743 * Look for all-0's host part (old broadcast addr), 744 * either for subnet or net. 745 */ 746 ip->ip_dst.s_addr == ia->ia_subnet || 747 ip->ip_dst.s_addr == ia->ia_net) 748 goto ours; 749 /* 750 * An interface with IP address zero accepts 751 * all packets that arrive on that interface. 752 */ 753 if (in_nullhost(ia->ia_addr.sin_addr)) 754 goto ours; 755 } 756 } 757 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 758 struct in_multi *inm; 759#ifdef MROUTING 760 extern struct socket *ip_mrouter; 761 762 if (ip_mrouter) { 763 /* 764 * If we are acting as a multicast router, all 765 * incoming multicast packets are passed to the 766 * kernel-level multicast forwarding function. 767 * The packet is returned (relatively) intact; if 768 * ip_mforward() returns a non-zero value, the packet 769 * must be discarded, else it may be accepted below. 770 * 771 * (The IP ident field is put in the same byte order 772 * as expected when ip_mforward() is called from 773 * ip_output().) 774 */ 775 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 776 ipstat.ips_cantforward++; 777 m_freem(m); 778 return; 779 } 780 781 /* 782 * The process-level routing demon needs to receive 783 * all multicast IGMP packets, whether or not this 784 * host belongs to their destination groups. 785 */ 786 if (ip->ip_p == IPPROTO_IGMP) 787 goto ours; 788 ipstat.ips_forward++; 789 } 790#endif 791 /* 792 * See if we belong to the destination multicast group on the 793 * arrival interface. 794 */ 795 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 796 if (inm == NULL) { 797 ipstat.ips_cantforward++; 798 m_freem(m); 799 return; 800 } 801 goto ours; 802 } 803 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 804 in_nullhost(ip->ip_dst)) 805 goto ours; 806 807 /* 808 * Not for us; forward if possible and desirable. 809 */ 810 if (ipforwarding == 0) { 811 ipstat.ips_cantforward++; 812 m_freem(m); 813 } else { 814 /* 815 * If ip_dst matched any of my address on !IFF_UP interface, 816 * and there's no IFF_UP interface that matches ip_dst, 817 * send icmp unreach. Forwarding it will result in in-kernel 818 * forwarding loop till TTL goes to 0. 819 */ 820 if (downmatch) { 821 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 822 ipstat.ips_cantforward++; 823 return; 824 } 825#ifdef IPSEC 826 if (ipsec4_in_reject(m, NULL)) { 827 ipsecstat.in_polvio++; 828 goto bad; 829 } 830#endif 831#ifdef FAST_IPSEC 832 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 833 s = splsoftnet(); 834 if (mtag != NULL) { 835 tdbi = (struct tdb_ident *)(mtag + 1); 836 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 837 } else { 838 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 839 IP_FORWARDING, &error); 840 } 841 if (sp == NULL) { /* NB: can happen if error */ 842 splx(s); 843 /*XXX error stat???*/ 844 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 845 goto bad; 846 } 847 848 /* 849 * Check security policy against packet attributes. 850 */ 851 error = ipsec_in_reject(sp, m); 852 KEY_FREESP(&sp); 853 splx(s); 854 if (error) { 855 ipstat.ips_cantforward++; 856 goto bad; 857 } 858 859 /* 860 * Peek at the outbound SP for this packet to determine if 861 * it's a Fast Forward candidate. 862 */ 863 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 864 if (mtag != NULL) 865 m->m_flags &= ~M_CANFASTFWD; 866 else { 867 s = splsoftnet(); 868 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, 869 (IP_FORWARDING | 870 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 871 &error, NULL); 872 if (sp != NULL) { 873 m->m_flags &= ~M_CANFASTFWD; 874 KEY_FREESP(&sp); 875 } 876 splx(s); 877 } 878#endif /* FAST_IPSEC */ 879 880 ip_forward(m, srcrt); 881 } 882 return; 883 884ours: 885 /* 886 * If offset or IP_MF are set, must reassemble. 887 * Otherwise, nothing need be done. 888 * (We could look in the reassembly queue to see 889 * if the packet was previously fragmented, 890 * but it's not worth the time; just let them time out.) 891 */ 892 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 893 894 /* 895 * Look for queue of fragments 896 * of this datagram. 897 */ 898 IPQ_LOCK(); 899 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 900 /* XXX LIST_FOREACH(fp, &ipq[hash], ipq_q) */ 901 for (fp = LIST_FIRST(&ipq[hash]); fp != NULL; 902 fp = LIST_NEXT(fp, ipq_q)) { 903 if (ip->ip_id == fp->ipq_id && 904 in_hosteq(ip->ip_src, fp->ipq_src) && 905 in_hosteq(ip->ip_dst, fp->ipq_dst) && 906 ip->ip_p == fp->ipq_p) 907 goto found; 908 909 } 910 fp = 0; 911found: 912 913 /* 914 * Adjust ip_len to not reflect header, 915 * set ipqe_mff if more fragments are expected, 916 * convert offset of this to bytes. 917 */ 918 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 919 mff = (ip->ip_off & htons(IP_MF)) != 0; 920 if (mff) { 921 /* 922 * Make sure that fragments have a data length 923 * that's a non-zero multiple of 8 bytes. 924 */ 925 if (ntohs(ip->ip_len) == 0 || 926 (ntohs(ip->ip_len) & 0x7) != 0) { 927 ipstat.ips_badfrags++; 928 IPQ_UNLOCK(); 929 goto bad; 930 } 931 } 932 ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3); 933 934 /* 935 * If datagram marked as having more fragments 936 * or if this is not the first fragment, 937 * attempt reassembly; if it succeeds, proceed. 938 */ 939 if (mff || ip->ip_off != htons(0)) { 940 ipstat.ips_fragments++; 941 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 942 if (ipqe == NULL) { 943 ipstat.ips_rcvmemdrop++; 944 IPQ_UNLOCK(); 945 goto bad; 946 } 947 ipqe->ipqe_mff = mff; 948 ipqe->ipqe_m = m; 949 ipqe->ipqe_ip = ip; 950 m = ip_reass(ipqe, fp, &ipq[hash]); 951 if (m == 0) { 952 IPQ_UNLOCK(); 953 return; 954 } 955 ipstat.ips_reassembled++; 956 ip = mtod(m, struct ip *); 957 hlen = ip->ip_hl << 2; 958 ip->ip_len = htons(ntohs(ip->ip_len) + hlen); 959 } else 960 if (fp) 961 ip_freef(fp); 962 IPQ_UNLOCK(); 963 } 964 965#if defined(IPSEC) 966 /* 967 * enforce IPsec policy checking if we are seeing last header. 968 * note that we do not visit this with protocols with pcb layer 969 * code - like udp/tcp/raw ip. 970 */ 971 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 972 ipsec4_in_reject(m, NULL)) { 973 ipsecstat.in_polvio++; 974 goto bad; 975 } 976#endif 977#ifdef FAST_IPSEC 978 /* 979 * enforce IPsec policy checking if we are seeing last header. 980 * note that we do not visit this with protocols with pcb layer 981 * code - like udp/tcp/raw ip. 982 */ 983 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 984 /* 985 * Check if the packet has already had IPsec processing 986 * done. If so, then just pass it along. This tag gets 987 * set during AH, ESP, etc. input handling, before the 988 * packet is returned to the ip input queue for delivery. 989 */ 990 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 991 s = splsoftnet(); 992 if (mtag != NULL) { 993 tdbi = (struct tdb_ident *)(mtag + 1); 994 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 995 } else { 996 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 997 IP_FORWARDING, &error); 998 } 999 if (sp != NULL) { 1000 /* 1001 * Check security policy against packet attributes. 1002 */ 1003 error = ipsec_in_reject(sp, m); 1004 KEY_FREESP(&sp); 1005 } else { 1006 /* XXX error stat??? */ 1007 error = EINVAL; 1008DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 1009 goto bad; 1010 } 1011 splx(s); 1012 if (error) 1013 goto bad; 1014 } 1015#endif /* FAST_IPSEC */ 1016 1017 /* 1018 * Switch out to protocol's input routine. 1019 */ 1020#if IFA_STATS 1021 if (ia && ip) 1022 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 1023#endif 1024 ipstat.ips_delivered++; 1025 { 1026 int off = hlen, nh = ip->ip_p; 1027 1028 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 1029 return; 1030 } 1031bad: 1032 m_freem(m); 1033 return; 1034 1035badcsum: 1036 ipstat.ips_badsum++; 1037 m_freem(m); 1038} 1039 1040/* 1041 * Take incoming datagram fragment and try to 1042 * reassemble it into whole datagram. If a chain for 1043 * reassembly of this datagram already exists, then it 1044 * is given as fp; otherwise have to make a chain. 1045 */ 1046struct mbuf * 1047ip_reass(struct ipqent *ipqe, struct ipq *fp, struct ipqhead *ipqhead) 1048{ 1049 struct mbuf *m = ipqe->ipqe_m; 1050 struct ipqent *nq, *p, *q; 1051 struct ip *ip; 1052 struct mbuf *t; 1053 int hlen = ipqe->ipqe_ip->ip_hl << 2; 1054 int i, next; 1055 1056 IPQ_LOCK_CHECK(); 1057 1058 /* 1059 * Presence of header sizes in mbufs 1060 * would confuse code below. 1061 */ 1062 m->m_data += hlen; 1063 m->m_len -= hlen; 1064 1065#ifdef notyet 1066 /* make sure fragment limit is up-to-date */ 1067 CHECK_NMBCLUSTER_PARAMS(); 1068 1069 /* If we have too many fragments, drop the older half. */ 1070 if (ip_nfrags >= ip_maxfrags) 1071 ip_reass_drophalf(void); 1072#endif 1073 1074 /* 1075 * We are about to add a fragment; increment frag count. 1076 */ 1077 ip_nfrags++; 1078 1079 /* 1080 * If first fragment to arrive, create a reassembly queue. 1081 */ 1082 if (fp == 0) { 1083 /* 1084 * Enforce upper bound on number of fragmented packets 1085 * for which we attempt reassembly; 1086 * If maxfrag is 0, never accept fragments. 1087 * If maxfrag is -1, accept all fragments without limitation. 1088 */ 1089 if (ip_maxfragpackets < 0) 1090 ; 1091 else if (ip_nfragpackets >= ip_maxfragpackets) 1092 goto dropfrag; 1093 ip_nfragpackets++; 1094 MALLOC(fp, struct ipq *, sizeof (struct ipq), 1095 M_FTABLE, M_NOWAIT); 1096 if (fp == NULL) 1097 goto dropfrag; 1098 LIST_INSERT_HEAD(ipqhead, fp, ipq_q); 1099 fp->ipq_nfrags = 1; 1100 fp->ipq_ttl = IPFRAGTTL; 1101 fp->ipq_p = ipqe->ipqe_ip->ip_p; 1102 fp->ipq_id = ipqe->ipqe_ip->ip_id; 1103 TAILQ_INIT(&fp->ipq_fragq); 1104 fp->ipq_src = ipqe->ipqe_ip->ip_src; 1105 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 1106 p = NULL; 1107 goto insert; 1108 } else { 1109 fp->ipq_nfrags++; 1110 } 1111 1112 /* 1113 * Find a segment which begins after this one does. 1114 */ 1115 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1116 p = q, q = TAILQ_NEXT(q, ipqe_q)) 1117 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off)) 1118 break; 1119 1120 /* 1121 * If there is a preceding segment, it may provide some of 1122 * our data already. If so, drop the data from the incoming 1123 * segment. If it provides all of our data, drop us. 1124 */ 1125 if (p != NULL) { 1126 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) - 1127 ntohs(ipqe->ipqe_ip->ip_off); 1128 if (i > 0) { 1129 if (i >= ntohs(ipqe->ipqe_ip->ip_len)) 1130 goto dropfrag; 1131 m_adj(ipqe->ipqe_m, i); 1132 ipqe->ipqe_ip->ip_off = 1133 htons(ntohs(ipqe->ipqe_ip->ip_off) + i); 1134 ipqe->ipqe_ip->ip_len = 1135 htons(ntohs(ipqe->ipqe_ip->ip_len) - i); 1136 } 1137 } 1138 1139 /* 1140 * While we overlap succeeding segments trim them or, 1141 * if they are completely covered, dequeue them. 1142 */ 1143 for (; q != NULL && 1144 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) > 1145 ntohs(q->ipqe_ip->ip_off); q = nq) { 1146 i = (ntohs(ipqe->ipqe_ip->ip_off) + 1147 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off); 1148 if (i < ntohs(q->ipqe_ip->ip_len)) { 1149 q->ipqe_ip->ip_len = 1150 htons(ntohs(q->ipqe_ip->ip_len) - i); 1151 q->ipqe_ip->ip_off = 1152 htons(ntohs(q->ipqe_ip->ip_off) + i); 1153 m_adj(q->ipqe_m, i); 1154 break; 1155 } 1156 nq = TAILQ_NEXT(q, ipqe_q); 1157 m_freem(q->ipqe_m); 1158 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1159 pool_put(&ipqent_pool, q); 1160 fp->ipq_nfrags--; 1161 ip_nfrags--; 1162 } 1163 1164insert: 1165 /* 1166 * Stick new segment in its place; 1167 * check for complete reassembly. 1168 */ 1169 if (p == NULL) { 1170 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 1171 } else { 1172 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q); 1173 } 1174 next = 0; 1175 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1176 p = q, q = TAILQ_NEXT(q, ipqe_q)) { 1177 if (ntohs(q->ipqe_ip->ip_off) != next) 1178 return (0); 1179 next += ntohs(q->ipqe_ip->ip_len); 1180 } 1181 if (p->ipqe_mff) 1182 return (0); 1183 1184 /* 1185 * Reassembly is complete. Check for a bogus message size and 1186 * concatenate fragments. 1187 */ 1188 q = TAILQ_FIRST(&fp->ipq_fragq); 1189 ip = q->ipqe_ip; 1190 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 1191 ipstat.ips_toolong++; 1192 ip_freef(fp); 1193 return (0); 1194 } 1195 m = q->ipqe_m; 1196 t = m->m_next; 1197 m->m_next = 0; 1198 m_cat(m, t); 1199 nq = TAILQ_NEXT(q, ipqe_q); 1200 pool_put(&ipqent_pool, q); 1201 for (q = nq; q != NULL; q = nq) { 1202 t = q->ipqe_m; 1203 nq = TAILQ_NEXT(q, ipqe_q); 1204 pool_put(&ipqent_pool, q); 1205 m_cat(m, t); 1206 } 1207 ip_nfrags -= fp->ipq_nfrags; 1208 1209 /* 1210 * Create header for new ip packet by 1211 * modifying header of first packet; 1212 * dequeue and discard fragment reassembly header. 1213 * Make header visible. 1214 */ 1215 ip->ip_len = htons(next); 1216 ip->ip_src = fp->ipq_src; 1217 ip->ip_dst = fp->ipq_dst; 1218 LIST_REMOVE(fp, ipq_q); 1219 FREE(fp, M_FTABLE); 1220 ip_nfragpackets--; 1221 m->m_len += (ip->ip_hl << 2); 1222 m->m_data -= (ip->ip_hl << 2); 1223 /* some debugging cruft by sklower, below, will go away soon */ 1224 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1225 int plen = 0; 1226 for (t = m; t; t = t->m_next) 1227 plen += t->m_len; 1228 m->m_pkthdr.len = plen; 1229 m->m_pkthdr.csum_flags = 0; 1230 } 1231 return (m); 1232 1233dropfrag: 1234 if (fp != 0) 1235 fp->ipq_nfrags--; 1236 ip_nfrags--; 1237 ipstat.ips_fragdropped++; 1238 m_freem(m); 1239 pool_put(&ipqent_pool, ipqe); 1240 return (0); 1241} 1242 1243/* 1244 * Free a fragment reassembly header and all 1245 * associated datagrams. 1246 */ 1247void 1248ip_freef(struct ipq *fp) 1249{ 1250 struct ipqent *q, *p; 1251 u_int nfrags = 0; 1252 1253 IPQ_LOCK_CHECK(); 1254 1255 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) { 1256 p = TAILQ_NEXT(q, ipqe_q); 1257 m_freem(q->ipqe_m); 1258 nfrags++; 1259 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1260 pool_put(&ipqent_pool, q); 1261 } 1262 1263 if (nfrags != fp->ipq_nfrags) 1264 printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags); 1265 ip_nfrags -= nfrags; 1266 LIST_REMOVE(fp, ipq_q); 1267 FREE(fp, M_FTABLE); 1268 ip_nfragpackets--; 1269} 1270 1271/* 1272 * IP reassembly TTL machinery for multiplicative drop. 1273 */ 1274static u_int fragttl_histo[(IPFRAGTTL+1)]; 1275 1276 1277/* 1278 * Decrement TTL of all reasembly queue entries by `ticks'. 1279 * Count number of distinct fragments (as opposed to partial, fragmented 1280 * datagrams) in the reassembly queue. While we traverse the entire 1281 * reassembly queue, compute and return the median TTL over all fragments. 1282 */ 1283static u_int 1284ip_reass_ttl_decr(u_int ticks) 1285{ 1286 u_int nfrags, median, dropfraction, keepfraction; 1287 struct ipq *fp, *nfp; 1288 int i; 1289 1290 nfrags = 0; 1291 memset(fragttl_histo, 0, sizeof fragttl_histo); 1292 1293 for (i = 0; i < IPREASS_NHASH; i++) { 1294 for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) { 1295 fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ? 1296 0 : fp->ipq_ttl - ticks); 1297 nfp = LIST_NEXT(fp, ipq_q); 1298 if (fp->ipq_ttl == 0) { 1299 ipstat.ips_fragtimeout++; 1300 ip_freef(fp); 1301 } else { 1302 nfrags += fp->ipq_nfrags; 1303 fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags; 1304 } 1305 } 1306 } 1307 1308 KASSERT(ip_nfrags == nfrags); 1309 1310 /* Find median (or other drop fraction) in histogram. */ 1311 dropfraction = (ip_nfrags / 2); 1312 keepfraction = ip_nfrags - dropfraction; 1313 for (i = IPFRAGTTL, median = 0; i >= 0; i--) { 1314 median += fragttl_histo[i]; 1315 if (median >= keepfraction) 1316 break; 1317 } 1318 1319 /* Return TTL of median (or other fraction). */ 1320 return (u_int)i; 1321} 1322 1323void 1324ip_reass_drophalf(void) 1325{ 1326 1327 u_int median_ticks; 1328 /* 1329 * Compute median TTL of all fragments, and count frags 1330 * with that TTL or lower (roughly half of all fragments). 1331 */ 1332 median_ticks = ip_reass_ttl_decr(0); 1333 1334 /* Drop half. */ 1335 median_ticks = ip_reass_ttl_decr(median_ticks); 1336 1337} 1338 1339/* 1340 * IP timer processing; 1341 * if a timer expires on a reassembly 1342 * queue, discard it. 1343 */ 1344void 1345ip_slowtimo(void) 1346{ 1347 static u_int dropscanidx = 0; 1348 u_int i; 1349 u_int median_ttl; 1350 int s = splsoftnet(); 1351 1352 IPQ_LOCK(); 1353 1354 /* Age TTL of all fragments by 1 tick .*/ 1355 median_ttl = ip_reass_ttl_decr(1); 1356 1357 /* make sure fragment limit is up-to-date */ 1358 CHECK_NMBCLUSTER_PARAMS(); 1359 1360 /* If we have too many fragments, drop the older half. */ 1361 if (ip_nfrags > ip_maxfrags) 1362 ip_reass_ttl_decr(median_ttl); 1363 1364 /* 1365 * If we are over the maximum number of fragmented packets 1366 * (due to the limit being lowered), drain off 1367 * enough to get down to the new limit. Start draining 1368 * from the reassembly hashqueue most recently drained. 1369 */ 1370 if (ip_maxfragpackets < 0) 1371 ; 1372 else { 1373 int wrapped = 0; 1374 1375 i = dropscanidx; 1376 while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) { 1377 while (LIST_FIRST(&ipq[i]) != NULL) 1378 ip_freef(LIST_FIRST(&ipq[i])); 1379 if (++i >= IPREASS_NHASH) { 1380 i = 0; 1381 } 1382 /* 1383 * Dont scan forever even if fragment counters are 1384 * wrong: stop after scanning entire reassembly queue. 1385 */ 1386 if (i == dropscanidx) 1387 wrapped = 1; 1388 } 1389 dropscanidx = i; 1390 } 1391 IPQ_UNLOCK(); 1392#ifdef GATEWAY 1393 ipflow_slowtimo(); 1394#endif 1395 splx(s); 1396} 1397 1398/* 1399 * Drain off all datagram fragments. 1400 */ 1401void 1402ip_drain(void) 1403{ 1404 1405 /* 1406 * We may be called from a device's interrupt context. If 1407 * the ipq is already busy, just bail out now. 1408 */ 1409 if (ipq_lock_try() == 0) 1410 return; 1411 1412 /* 1413 * Drop half the total fragments now. If more mbufs are needed, 1414 * we will be called again soon. 1415 */ 1416 ip_reass_drophalf(); 1417 1418 IPQ_UNLOCK(); 1419} 1420 1421/* 1422 * Do option processing on a datagram, 1423 * possibly discarding it if bad options are encountered, 1424 * or forwarding it if source-routed. 1425 * Returns 1 if packet has been forwarded/freed, 1426 * 0 if the packet should be processed further. 1427 */ 1428int 1429ip_dooptions(struct mbuf *m) 1430{ 1431 struct ip *ip = mtod(m, struct ip *); 1432 u_char *cp, *cp0; 1433 struct ip_timestamp *ipt; 1434 struct in_ifaddr *ia; 1435 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1436 struct in_addr dst; 1437 n_time ntime; 1438 1439 dst = ip->ip_dst; 1440 cp = (u_char *)(ip + 1); 1441 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1442 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1443 opt = cp[IPOPT_OPTVAL]; 1444 if (opt == IPOPT_EOL) 1445 break; 1446 if (opt == IPOPT_NOP) 1447 optlen = 1; 1448 else { 1449 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1450 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1451 goto bad; 1452 } 1453 optlen = cp[IPOPT_OLEN]; 1454 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1455 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1456 goto bad; 1457 } 1458 } 1459 switch (opt) { 1460 1461 default: 1462 break; 1463 1464 /* 1465 * Source routing with record. 1466 * Find interface with current destination address. 1467 * If none on this machine then drop if strictly routed, 1468 * or do nothing if loosely routed. 1469 * Record interface address and bring up next address 1470 * component. If strictly routed make sure next 1471 * address is on directly accessible net. 1472 */ 1473 case IPOPT_LSRR: 1474 case IPOPT_SSRR: 1475 if (ip_allowsrcrt == 0) { 1476 type = ICMP_UNREACH; 1477 code = ICMP_UNREACH_NET_PROHIB; 1478 goto bad; 1479 } 1480 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1481 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1482 goto bad; 1483 } 1484 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1485 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1486 goto bad; 1487 } 1488 ipaddr.sin_addr = ip->ip_dst; 1489 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1490 if (ia == 0) { 1491 if (opt == IPOPT_SSRR) { 1492 type = ICMP_UNREACH; 1493 code = ICMP_UNREACH_SRCFAIL; 1494 goto bad; 1495 } 1496 /* 1497 * Loose routing, and not at next destination 1498 * yet; nothing to do except forward. 1499 */ 1500 break; 1501 } 1502 off--; /* 0 origin */ 1503 if ((off + sizeof(struct in_addr)) > optlen) { 1504 /* 1505 * End of source route. Should be for us. 1506 */ 1507 save_rte(cp, ip->ip_src); 1508 break; 1509 } 1510 /* 1511 * locate outgoing interface 1512 */ 1513 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1514 sizeof(ipaddr.sin_addr)); 1515 if (opt == IPOPT_SSRR) 1516 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); 1517 else 1518 ia = ip_rtaddr(ipaddr.sin_addr); 1519 if (ia == 0) { 1520 type = ICMP_UNREACH; 1521 code = ICMP_UNREACH_SRCFAIL; 1522 goto bad; 1523 } 1524 ip->ip_dst = ipaddr.sin_addr; 1525 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1526 (caddr_t)(cp + off), sizeof(struct in_addr)); 1527 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1528 /* 1529 * Let ip_intr's mcast routing check handle mcast pkts 1530 */ 1531 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1532 break; 1533 1534 case IPOPT_RR: 1535 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1536 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1537 goto bad; 1538 } 1539 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1540 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1541 goto bad; 1542 } 1543 /* 1544 * If no space remains, ignore. 1545 */ 1546 off--; /* 0 origin */ 1547 if ((off + sizeof(struct in_addr)) > optlen) 1548 break; 1549 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1550 sizeof(ipaddr.sin_addr)); 1551 /* 1552 * locate outgoing interface; if we're the destination, 1553 * use the incoming interface (should be same). 1554 */ 1555 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1556 == NULL && 1557 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1558 type = ICMP_UNREACH; 1559 code = ICMP_UNREACH_HOST; 1560 goto bad; 1561 } 1562 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1563 (caddr_t)(cp + off), sizeof(struct in_addr)); 1564 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1565 break; 1566 1567 case IPOPT_TS: 1568 code = cp - (u_char *)ip; 1569 ipt = (struct ip_timestamp *)cp; 1570 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1571 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1572 goto bad; 1573 } 1574 if (ipt->ipt_ptr < 5) { 1575 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1576 goto bad; 1577 } 1578 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1579 if (++ipt->ipt_oflw == 0) { 1580 code = (u_char *)&ipt->ipt_ptr - 1581 (u_char *)ip; 1582 goto bad; 1583 } 1584 break; 1585 } 1586 cp0 = (cp + ipt->ipt_ptr - 1); 1587 switch (ipt->ipt_flg) { 1588 1589 case IPOPT_TS_TSONLY: 1590 break; 1591 1592 case IPOPT_TS_TSANDADDR: 1593 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1594 sizeof(struct in_addr) > ipt->ipt_len) { 1595 code = (u_char *)&ipt->ipt_ptr - 1596 (u_char *)ip; 1597 goto bad; 1598 } 1599 ipaddr.sin_addr = dst; 1600 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1601 m->m_pkthdr.rcvif)); 1602 if (ia == 0) 1603 continue; 1604 bcopy(&ia->ia_addr.sin_addr, 1605 cp0, sizeof(struct in_addr)); 1606 ipt->ipt_ptr += sizeof(struct in_addr); 1607 break; 1608 1609 case IPOPT_TS_PRESPEC: 1610 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1611 sizeof(struct in_addr) > ipt->ipt_len) { 1612 code = (u_char *)&ipt->ipt_ptr - 1613 (u_char *)ip; 1614 goto bad; 1615 } 1616 bcopy(cp0, &ipaddr.sin_addr, 1617 sizeof(struct in_addr)); 1618 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1619 == NULL) 1620 continue; 1621 ipt->ipt_ptr += sizeof(struct in_addr); 1622 break; 1623 1624 default: 1625 /* XXX can't take &ipt->ipt_flg */ 1626 code = (u_char *)&ipt->ipt_ptr - 1627 (u_char *)ip + 1; 1628 goto bad; 1629 } 1630 ntime = iptime(); 1631 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1632 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1633 sizeof(n_time)); 1634 ipt->ipt_ptr += sizeof(n_time); 1635 } 1636 } 1637 if (forward) { 1638 if (ip_forwsrcrt == 0) { 1639 type = ICMP_UNREACH; 1640 code = ICMP_UNREACH_SRCFAIL; 1641 goto bad; 1642 } 1643 ip_forward(m, 1); 1644 return (1); 1645 } 1646 return (0); 1647bad: 1648 icmp_error(m, type, code, 0, 0); 1649 ipstat.ips_badoptions++; 1650 return (1); 1651} 1652 1653/* 1654 * Given address of next destination (final or next hop), 1655 * return internet address info of interface to be used to get there. 1656 */ 1657struct in_ifaddr * 1658ip_rtaddr(struct in_addr dst) 1659{ 1660 struct sockaddr_in *sin; 1661 1662 sin = satosin(&ipforward_rt.ro_dst); 1663 1664 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1665 if (ipforward_rt.ro_rt) { 1666 RTFREE(ipforward_rt.ro_rt); 1667 ipforward_rt.ro_rt = 0; 1668 } 1669 sin->sin_family = AF_INET; 1670 sin->sin_len = sizeof(*sin); 1671 sin->sin_addr = dst; 1672 1673 rtalloc(&ipforward_rt); 1674 } 1675 if (ipforward_rt.ro_rt == 0) 1676 return ((struct in_ifaddr *)0); 1677 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1678} 1679 1680/* 1681 * Save incoming source route for use in replies, 1682 * to be picked up later by ip_srcroute if the receiver is interested. 1683 */ 1684void 1685save_rte(u_char *option, struct in_addr dst) 1686{ 1687 unsigned olen; 1688 1689 olen = option[IPOPT_OLEN]; 1690#ifdef DIAGNOSTIC 1691 if (ipprintfs) 1692 printf("save_rte: olen %d\n", olen); 1693#endif /* 0 */ 1694 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1695 return; 1696 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1697 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1698 ip_srcrt.dst = dst; 1699} 1700 1701/* 1702 * Retrieve incoming source route for use in replies, 1703 * in the same form used by setsockopt. 1704 * The first hop is placed before the options, will be removed later. 1705 */ 1706struct mbuf * 1707ip_srcroute(void) 1708{ 1709 struct in_addr *p, *q; 1710 struct mbuf *m; 1711 1712 if (ip_nhops == 0) 1713 return ((struct mbuf *)0); 1714 m = m_get(M_DONTWAIT, MT_SOOPTS); 1715 if (m == 0) 1716 return ((struct mbuf *)0); 1717 1718 MCLAIM(m, &inetdomain.dom_mowner); 1719#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1720 1721 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1722 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1723 OPTSIZ; 1724#ifdef DIAGNOSTIC 1725 if (ipprintfs) 1726 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1727#endif 1728 1729 /* 1730 * First save first hop for return route 1731 */ 1732 p = &ip_srcrt.route[ip_nhops - 1]; 1733 *(mtod(m, struct in_addr *)) = *p--; 1734#ifdef DIAGNOSTIC 1735 if (ipprintfs) 1736 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1737#endif 1738 1739 /* 1740 * Copy option fields and padding (nop) to mbuf. 1741 */ 1742 ip_srcrt.nop = IPOPT_NOP; 1743 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1744 bcopy((caddr_t)&ip_srcrt.nop, 1745 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1746 q = (struct in_addr *)(mtod(m, caddr_t) + 1747 sizeof(struct in_addr) + OPTSIZ); 1748#undef OPTSIZ 1749 /* 1750 * Record return path as an IP source route, 1751 * reversing the path (pointers are now aligned). 1752 */ 1753 while (p >= ip_srcrt.route) { 1754#ifdef DIAGNOSTIC 1755 if (ipprintfs) 1756 printf(" %x", ntohl(q->s_addr)); 1757#endif 1758 *q++ = *p--; 1759 } 1760 /* 1761 * Last hop goes to final destination. 1762 */ 1763 *q = ip_srcrt.dst; 1764#ifdef DIAGNOSTIC 1765 if (ipprintfs) 1766 printf(" %x\n", ntohl(q->s_addr)); 1767#endif 1768 return (m); 1769} 1770 1771/* 1772 * Strip out IP options, at higher 1773 * level protocol in the kernel. 1774 * Second argument is buffer to which options 1775 * will be moved, and return value is their length. 1776 * XXX should be deleted; last arg currently ignored. 1777 */ 1778void 1779ip_stripoptions(struct mbuf *m, struct mbuf *mopt) 1780{ 1781 int i; 1782 struct ip *ip = mtod(m, struct ip *); 1783 caddr_t opts; 1784 int olen; 1785 1786 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1787 opts = (caddr_t)(ip + 1); 1788 i = m->m_len - (sizeof (struct ip) + olen); 1789 bcopy(opts + olen, opts, (unsigned)i); 1790 m->m_len -= olen; 1791 if (m->m_flags & M_PKTHDR) 1792 m->m_pkthdr.len -= olen; 1793 ip->ip_len = htons(ntohs(ip->ip_len) - olen); 1794 ip->ip_hl = sizeof (struct ip) >> 2; 1795} 1796 1797const int inetctlerrmap[PRC_NCMDS] = { 1798 0, 0, 0, 0, 1799 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1800 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1801 EMSGSIZE, EHOSTUNREACH, 0, 0, 1802 0, 0, 0, 0, 1803 ENOPROTOOPT 1804}; 1805 1806/* 1807 * Forward a packet. If some error occurs return the sender 1808 * an icmp packet. Note we can't always generate a meaningful 1809 * icmp message because icmp doesn't have a large enough repertoire 1810 * of codes and types. 1811 * 1812 * If not forwarding, just drop the packet. This could be confusing 1813 * if ipforwarding was zero but some routing protocol was advancing 1814 * us as a gateway to somewhere. However, we must let the routing 1815 * protocol deal with that. 1816 * 1817 * The srcrt parameter indicates whether the packet is being forwarded 1818 * via a source route. 1819 */ 1820void 1821ip_forward(struct mbuf *m, int srcrt) 1822{ 1823 struct ip *ip = mtod(m, struct ip *); 1824 struct sockaddr_in *sin; 1825 struct rtentry *rt; 1826 int error, type = 0, code = 0, destmtu = 0; 1827 struct mbuf *mcopy; 1828 n_long dest; 1829 1830 /* 1831 * We are now in the output path. 1832 */ 1833 MCLAIM(m, &ip_tx_mowner); 1834 1835 /* 1836 * Clear any in-bound checksum flags for this packet. 1837 */ 1838 m->m_pkthdr.csum_flags = 0; 1839 1840 dest = 0; 1841#ifdef DIAGNOSTIC 1842 if (ipprintfs) { 1843 printf("forward: src %s ", inet_ntoa(ip->ip_src)); 1844 printf("dst %s ttl %x\n", inet_ntoa(ip->ip_dst), ip->ip_ttl); 1845 } 1846#endif 1847 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1848 ipstat.ips_cantforward++; 1849 m_freem(m); 1850 return; 1851 } 1852 if (ip->ip_ttl <= IPTTLDEC) { 1853 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1854 return; 1855 } 1856 1857 sin = satosin(&ipforward_rt.ro_dst); 1858 if ((rt = ipforward_rt.ro_rt) == 0 || 1859 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1860 if (ipforward_rt.ro_rt) { 1861 RTFREE(ipforward_rt.ro_rt); 1862 ipforward_rt.ro_rt = 0; 1863 } 1864 sin->sin_family = AF_INET; 1865 sin->sin_len = sizeof(struct sockaddr_in); 1866 sin->sin_addr = ip->ip_dst; 1867 1868 rtalloc(&ipforward_rt); 1869 if (ipforward_rt.ro_rt == 0) { 1870 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1871 return; 1872 } 1873 rt = ipforward_rt.ro_rt; 1874 } 1875 1876 /* 1877 * Save at most 68 bytes of the packet in case 1878 * we need to generate an ICMP message to the src. 1879 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1880 */ 1881 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1882 if (mcopy) 1883 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1884 1885 ip->ip_ttl -= IPTTLDEC; 1886 1887 /* 1888 * If forwarding packet using same interface that it came in on, 1889 * perhaps should send a redirect to sender to shortcut a hop. 1890 * Only send redirect if source is sending directly to us, 1891 * and if packet was not source routed (or has any options). 1892 * Also, don't send redirect if forwarding using a default route 1893 * or a route modified by a redirect. 1894 */ 1895 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1896 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1897 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1898 ipsendredirects && !srcrt) { 1899 if (rt->rt_ifa && 1900 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1901 ifatoia(rt->rt_ifa)->ia_subnet) { 1902 if (rt->rt_flags & RTF_GATEWAY) 1903 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1904 else 1905 dest = ip->ip_dst.s_addr; 1906 /* 1907 * Router requirements says to only send host 1908 * redirects. 1909 */ 1910 type = ICMP_REDIRECT; 1911 code = ICMP_REDIRECT_HOST; 1912#ifdef DIAGNOSTIC 1913 if (ipprintfs) 1914 printf("redirect (%d) to %x\n", code, 1915 (u_int32_t)dest); 1916#endif 1917 } 1918 } 1919 1920 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1921 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1922 (struct ip_moptions *)NULL, (struct socket *)NULL); 1923 1924 if (error) 1925 ipstat.ips_cantforward++; 1926 else { 1927 ipstat.ips_forward++; 1928 if (type) 1929 ipstat.ips_redirectsent++; 1930 else { 1931 if (mcopy) { 1932#ifdef GATEWAY 1933 if (mcopy->m_flags & M_CANFASTFWD) 1934 ipflow_create(&ipforward_rt, mcopy); 1935#endif 1936 m_freem(mcopy); 1937 } 1938 return; 1939 } 1940 } 1941 if (mcopy == NULL) 1942 return; 1943 1944 switch (error) { 1945 1946 case 0: /* forwarded, but need redirect */ 1947 /* type, code set above */ 1948 break; 1949 1950 case ENETUNREACH: /* shouldn't happen, checked above */ 1951 case EHOSTUNREACH: 1952 case ENETDOWN: 1953 case EHOSTDOWN: 1954 default: 1955 type = ICMP_UNREACH; 1956 code = ICMP_UNREACH_HOST; 1957 break; 1958 1959 case EMSGSIZE: 1960 type = ICMP_UNREACH; 1961 code = ICMP_UNREACH_NEEDFRAG; 1962#if !defined(IPSEC) && !defined(FAST_IPSEC) 1963 if (ipforward_rt.ro_rt) 1964 destmtu = ipforward_rt.ro_rt->rt_ifp->if_mtu; 1965#else 1966 /* 1967 * If the packet is routed over IPsec tunnel, tell the 1968 * originator the tunnel MTU. 1969 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1970 * XXX quickhack!!! 1971 */ 1972 if (ipforward_rt.ro_rt) { 1973 struct secpolicy *sp; 1974 int ipsecerror; 1975 size_t ipsechdr; 1976 struct route *ro; 1977 1978 sp = ipsec4_getpolicybyaddr(mcopy, 1979 IPSEC_DIR_OUTBOUND, IP_FORWARDING, 1980 &ipsecerror); 1981 1982 if (sp == NULL) 1983 destmtu = ipforward_rt.ro_rt->rt_ifp->if_mtu; 1984 else { 1985 /* count IPsec header size */ 1986 ipsechdr = ipsec4_hdrsiz(mcopy, 1987 IPSEC_DIR_OUTBOUND, NULL); 1988 1989 /* 1990 * find the correct route for outer IPv4 1991 * header, compute tunnel MTU. 1992 */ 1993 1994 if (sp->req != NULL 1995 && sp->req->sav != NULL 1996 && sp->req->sav->sah != NULL) { 1997 ro = &sp->req->sav->sah->sa_route; 1998 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1999 destmtu = 2000 ro->ro_rt->rt_rmx.rmx_mtu ? 2001 ro->ro_rt->rt_rmx.rmx_mtu : 2002 ro->ro_rt->rt_ifp->if_mtu; 2003 destmtu -= ipsechdr; 2004 } 2005 } 2006 2007#ifdef IPSEC 2008 key_freesp(sp); 2009#else 2010 KEY_FREESP(&sp); 2011#endif 2012 } 2013 } 2014#endif /*IPSEC*/ 2015 ipstat.ips_cantfrag++; 2016 break; 2017 2018 case ENOBUFS: 2019#if 1 2020 /* 2021 * a router should not generate ICMP_SOURCEQUENCH as 2022 * required in RFC1812 Requirements for IP Version 4 Routers. 2023 * source quench could be a big problem under DoS attacks, 2024 * or if the underlying interface is rate-limited. 2025 */ 2026 if (mcopy) 2027 m_freem(mcopy); 2028 return; 2029#else 2030 type = ICMP_SOURCEQUENCH; 2031 code = 0; 2032 break; 2033#endif 2034 } 2035 icmp_error(mcopy, type, code, dest, destmtu); 2036} 2037 2038void 2039ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2040 struct mbuf *m) 2041{ 2042 2043 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2044 struct timeval tv; 2045 2046 microtime(&tv); 2047 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2048 SCM_TIMESTAMP, SOL_SOCKET); 2049 if (*mp) 2050 mp = &(*mp)->m_next; 2051 } 2052 if (inp->inp_flags & INP_RECVDSTADDR) { 2053 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2054 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2055 if (*mp) 2056 mp = &(*mp)->m_next; 2057 } 2058#ifdef notyet 2059 /* 2060 * XXX 2061 * Moving these out of udp_input() made them even more broken 2062 * than they already were. 2063 * - fenner@parc.xerox.com 2064 */ 2065 /* options were tossed already */ 2066 if (inp->inp_flags & INP_RECVOPTS) { 2067 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2068 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2069 if (*mp) 2070 mp = &(*mp)->m_next; 2071 } 2072 /* ip_srcroute doesn't do what we want here, need to fix */ 2073 if (inp->inp_flags & INP_RECVRETOPTS) { 2074 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 2075 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2076 if (*mp) 2077 mp = &(*mp)->m_next; 2078 } 2079#endif 2080 if (inp->inp_flags & INP_RECVIF) { 2081 struct sockaddr_dl sdl; 2082 2083 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 2084 sdl.sdl_family = AF_LINK; 2085 sdl.sdl_index = m->m_pkthdr.rcvif ? 2086 m->m_pkthdr.rcvif->if_index : 0; 2087 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 2088 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 2089 IP_RECVIF, IPPROTO_IP); 2090 if (*mp) 2091 mp = &(*mp)->m_next; 2092 } 2093} 2094 2095/* 2096 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 2097 * range of the new value and tweaks timers if it changes. 2098 */ 2099static int 2100sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 2101{ 2102 int error, tmp; 2103 struct sysctlnode node; 2104 2105 node = *rnode; 2106 tmp = ip_mtudisc_timeout; 2107 node.sysctl_data = &tmp; 2108 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2109 if (error || newp == NULL) 2110 return (error); 2111 if (tmp < 0) 2112 return (EINVAL); 2113 2114 ip_mtudisc_timeout = tmp; 2115 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 2116 2117 return (0); 2118} 2119 2120#ifdef GATEWAY 2121/* 2122 * sysctl helper routine for net.inet.ip.maxflows. apparently if 2123 * maxflows is even looked up, we "reap flows". 2124 */ 2125static int 2126sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS) 2127{ 2128 int s; 2129 2130 s = sysctl_lookup(SYSCTLFN_CALL(rnode)); 2131 if (s) 2132 return (s); 2133 2134 s = splsoftnet(); 2135 ipflow_reap(0); 2136 splx(s); 2137 2138 return (0); 2139} 2140#endif /* GATEWAY */ 2141 2142 2143SYSCTL_SETUP(sysctl_net_inet_ip_setup, "sysctl net.inet.ip subtree setup") 2144{ 2145 extern int subnetsarelocal, hostzeroisbroadcast; 2146 2147 sysctl_createv(clog, 0, NULL, NULL, 2148 CTLFLAG_PERMANENT, 2149 CTLTYPE_NODE, "net", NULL, 2150 NULL, 0, NULL, 0, 2151 CTL_NET, CTL_EOL); 2152 sysctl_createv(clog, 0, NULL, NULL, 2153 CTLFLAG_PERMANENT, 2154 CTLTYPE_NODE, "inet", 2155 SYSCTL_DESCR("PF_INET related settings"), 2156 NULL, 0, NULL, 0, 2157 CTL_NET, PF_INET, CTL_EOL); 2158 sysctl_createv(clog, 0, NULL, NULL, 2159 CTLFLAG_PERMANENT, 2160 CTLTYPE_NODE, "ip", 2161 SYSCTL_DESCR("IPv4 related settings"), 2162 NULL, 0, NULL, 0, 2163 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 2164 2165 sysctl_createv(clog, 0, NULL, NULL, 2166 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2167 CTLTYPE_INT, "forwarding", 2168 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 2169 NULL, 0, &ipforwarding, 0, 2170 CTL_NET, PF_INET, IPPROTO_IP, 2171 IPCTL_FORWARDING, CTL_EOL); 2172 sysctl_createv(clog, 0, NULL, NULL, 2173 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2174 CTLTYPE_INT, "redirect", 2175 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 2176 NULL, 0, &ipsendredirects, 0, 2177 CTL_NET, PF_INET, IPPROTO_IP, 2178 IPCTL_SENDREDIRECTS, CTL_EOL); 2179 sysctl_createv(clog, 0, NULL, NULL, 2180 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2181 CTLTYPE_INT, "ttl", 2182 SYSCTL_DESCR("Default TTL for an INET datagram"), 2183 NULL, 0, &ip_defttl, 0, 2184 CTL_NET, PF_INET, IPPROTO_IP, 2185 IPCTL_DEFTTL, CTL_EOL); 2186#ifdef IPCTL_DEFMTU 2187 sysctl_createv(clog, 0, NULL, NULL, 2188 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 2189 CTLTYPE_INT, "mtu", 2190 SYSCTL_DESCR("Default MTA for an INET route"), 2191 NULL, 0, &ip_mtu, 0, 2192 CTL_NET, PF_INET, IPPROTO_IP, 2193 IPCTL_DEFMTU, CTL_EOL); 2194#endif /* IPCTL_DEFMTU */ 2195 sysctl_createv(clog, 0, NULL, NULL, 2196 CTLFLAG_PERMANENT|CTLFLAG_READONLY1, 2197 CTLTYPE_INT, "forwsrcrt", 2198 SYSCTL_DESCR("Enable forwarding of source-routed " 2199 "datagrams"), 2200 NULL, 0, &ip_forwsrcrt, 0, 2201 CTL_NET, PF_INET, IPPROTO_IP, 2202 IPCTL_FORWSRCRT, CTL_EOL); 2203 sysctl_createv(clog, 0, NULL, NULL, 2204 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2205 CTLTYPE_INT, "directed-broadcast", 2206 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 2207 NULL, 0, &ip_directedbcast, 0, 2208 CTL_NET, PF_INET, IPPROTO_IP, 2209 IPCTL_DIRECTEDBCAST, CTL_EOL); 2210 sysctl_createv(clog, 0, NULL, NULL, 2211 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2212 CTLTYPE_INT, "allowsrcrt", 2213 SYSCTL_DESCR("Accept source-routed datagrams"), 2214 NULL, 0, &ip_allowsrcrt, 0, 2215 CTL_NET, PF_INET, IPPROTO_IP, 2216 IPCTL_ALLOWSRCRT, CTL_EOL); 2217 sysctl_createv(clog, 0, NULL, NULL, 2218 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2219 CTLTYPE_INT, "subnetsarelocal", 2220 SYSCTL_DESCR("Whether logical subnets are considered " 2221 "local"), 2222 NULL, 0, &subnetsarelocal, 0, 2223 CTL_NET, PF_INET, IPPROTO_IP, 2224 IPCTL_SUBNETSARELOCAL, CTL_EOL); 2225 sysctl_createv(clog, 0, NULL, NULL, 2226 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2227 CTLTYPE_INT, "mtudisc", 2228 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 2229 NULL, 0, &ip_mtudisc, 0, 2230 CTL_NET, PF_INET, IPPROTO_IP, 2231 IPCTL_MTUDISC, CTL_EOL); 2232 sysctl_createv(clog, 0, NULL, NULL, 2233 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2234 CTLTYPE_INT, "anonportmin", 2235 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 2236 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 2237 CTL_NET, PF_INET, IPPROTO_IP, 2238 IPCTL_ANONPORTMIN, CTL_EOL); 2239 sysctl_createv(clog, 0, NULL, NULL, 2240 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2241 CTLTYPE_INT, "anonportmax", 2242 SYSCTL_DESCR("Highest ephemeral port number to assign"), 2243 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 2244 CTL_NET, PF_INET, IPPROTO_IP, 2245 IPCTL_ANONPORTMAX, CTL_EOL); 2246 sysctl_createv(clog, 0, NULL, NULL, 2247 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2248 CTLTYPE_INT, "mtudisctimeout", 2249 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 2250 sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0, 2251 CTL_NET, PF_INET, IPPROTO_IP, 2252 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 2253#ifdef GATEWAY 2254 sysctl_createv(clog, 0, NULL, NULL, 2255 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2256 CTLTYPE_INT, "maxflows", 2257 SYSCTL_DESCR("Number of flows for fast forwarding"), 2258 sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0, 2259 CTL_NET, PF_INET, IPPROTO_IP, 2260 IPCTL_MAXFLOWS, CTL_EOL); 2261#endif /* GATEWAY */ 2262 sysctl_createv(clog, 0, NULL, NULL, 2263 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2264 CTLTYPE_INT, "hostzerobroadcast", 2265 SYSCTL_DESCR("All zeroes address is broadcast address"), 2266 NULL, 0, &hostzeroisbroadcast, 0, 2267 CTL_NET, PF_INET, IPPROTO_IP, 2268 IPCTL_HOSTZEROBROADCAST, CTL_EOL); 2269#if NGIF > 0 2270 sysctl_createv(clog, 0, NULL, NULL, 2271 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2272 CTLTYPE_INT, "gifttl", 2273 SYSCTL_DESCR("Default TTL for a gif tunnel datagram"), 2274 NULL, 0, &ip_gif_ttl, 0, 2275 CTL_NET, PF_INET, IPPROTO_IP, 2276 IPCTL_GIF_TTL, CTL_EOL); 2277#endif /* NGIF */ 2278#ifndef IPNOPRIVPORTS 2279 sysctl_createv(clog, 0, NULL, NULL, 2280 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2281 CTLTYPE_INT, "lowportmin", 2282 SYSCTL_DESCR("Lowest privileged ephemeral port number " 2283 "to assign"), 2284 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 2285 CTL_NET, PF_INET, IPPROTO_IP, 2286 IPCTL_LOWPORTMIN, CTL_EOL); 2287 sysctl_createv(clog, 0, NULL, NULL, 2288 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2289 CTLTYPE_INT, "lowportmax", 2290 SYSCTL_DESCR("Highest privileged ephemeral port number " 2291 "to assign"), 2292 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 2293 CTL_NET, PF_INET, IPPROTO_IP, 2294 IPCTL_LOWPORTMAX, CTL_EOL); 2295#endif /* IPNOPRIVPORTS */ 2296 sysctl_createv(clog, 0, NULL, NULL, 2297 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2298 CTLTYPE_INT, "maxfragpackets", 2299 SYSCTL_DESCR("Maximum number of fragments to retain for " 2300 "possible reassembly"), 2301 NULL, 0, &ip_maxfragpackets, 0, 2302 CTL_NET, PF_INET, IPPROTO_IP, 2303 IPCTL_MAXFRAGPACKETS, CTL_EOL); 2304#if NGRE > 0 2305 sysctl_createv(clog, 0, NULL, NULL, 2306 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2307 CTLTYPE_INT, "grettl", 2308 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 2309 NULL, 0, &ip_gre_ttl, 0, 2310 CTL_NET, PF_INET, IPPROTO_IP, 2311 IPCTL_GRE_TTL, CTL_EOL); 2312#endif /* NGRE */ 2313 sysctl_createv(clog, 0, NULL, NULL, 2314 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2315 CTLTYPE_INT, "checkinterface", 2316 SYSCTL_DESCR("Enable receive side of Strong ES model " 2317 "from RFC1122"), 2318 NULL, 0, &ip_checkinterface, 0, 2319 CTL_NET, PF_INET, IPPROTO_IP, 2320 IPCTL_CHECKINTERFACE, CTL_EOL); 2321 sysctl_createv(clog, 0, NULL, NULL, 2322 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2323 CTLTYPE_INT, "random_id", 2324 SYSCTL_DESCR("Assign random ip_id values"), 2325 NULL, 0, &ip_do_randomid, 0, 2326 CTL_NET, PF_INET, IPPROTO_IP, 2327 IPCTL_RANDOMID, CTL_EOL); 2328 sysctl_createv(clog, 0, NULL, NULL, 2329 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2330 CTLTYPE_INT, "do_loopback_cksum", 2331 SYSCTL_DESCR("Perform IP checksum on loopback"), 2332 NULL, 0, &ip_do_loopback_cksum, 0, 2333 CTL_NET, PF_INET, IPPROTO_IP, 2334 IPCTL_LOOPBACKCKSUM, CTL_EOL); 2335 sysctl_createv(clog, 0, NULL, NULL, 2336 CTLFLAG_PERMANENT, 2337 CTLTYPE_STRUCT, "stats", 2338 SYSCTL_DESCR("IP statistics"), 2339 NULL, 0, &ipstat, sizeof(ipstat), 2340 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 2341 CTL_EOL); 2342} 2343