ip_input.c revision 1.205
1/* $NetBSD: ip_input.c,v 1.205 2004/10/06 01:34:11 darrenr Exp $ */ 2 3/* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32/*- 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 3. All advertising materials mentioning features or use of this software 49 * must display the following acknowledgement: 50 * This product includes software developed by the NetBSD 51 * Foundation, Inc. and its contributors. 52 * 4. Neither the name of The NetBSD Foundation nor the names of its 53 * contributors may be used to endorse or promote products derived 54 * from this software without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 57 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 58 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 59 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 60 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 61 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 62 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 63 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 64 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 65 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 66 * POSSIBILITY OF SUCH DAMAGE. 67 */ 68 69/* 70 * Copyright (c) 1982, 1986, 1988, 1993 71 * The Regents of the University of California. All rights reserved. 72 * 73 * Redistribution and use in source and binary forms, with or without 74 * modification, are permitted provided that the following conditions 75 * are met: 76 * 1. Redistributions of source code must retain the above copyright 77 * notice, this list of conditions and the following disclaimer. 78 * 2. Redistributions in binary form must reproduce the above copyright 79 * notice, this list of conditions and the following disclaimer in the 80 * documentation and/or other materials provided with the distribution. 81 * 3. Neither the name of the University nor the names of its contributors 82 * may be used to endorse or promote products derived from this software 83 * without specific prior written permission. 84 * 85 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 86 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 87 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 88 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 89 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 90 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 91 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 92 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 93 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 94 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 95 * SUCH DAMAGE. 96 * 97 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 98 */ 99 100#include <sys/cdefs.h> 101__KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.205 2004/10/06 01:34:11 darrenr Exp $"); 102 103#include "opt_inet.h" 104#include "opt_gateway.h" 105#include "opt_pfil_hooks.h" 106#include "opt_ipsec.h" 107#include "opt_mrouting.h" 108#include "opt_mbuftrace.h" 109#include "opt_inet_csum.h" 110 111#include <sys/param.h> 112#include <sys/systm.h> 113#include <sys/malloc.h> 114#include <sys/mbuf.h> 115#include <sys/domain.h> 116#include <sys/protosw.h> 117#include <sys/socket.h> 118#include <sys/socketvar.h> 119#include <sys/errno.h> 120#include <sys/time.h> 121#include <sys/kernel.h> 122#include <sys/pool.h> 123#include <sys/sysctl.h> 124 125#include <net/if.h> 126#include <net/if_dl.h> 127#include <net/route.h> 128#include <net/pfil.h> 129 130#include <netinet/in.h> 131#include <netinet/in_systm.h> 132#include <netinet/ip.h> 133#include <netinet/in_pcb.h> 134#include <netinet/in_var.h> 135#include <netinet/ip_var.h> 136#include <netinet/ip_icmp.h> 137/* just for gif_ttl */ 138#include <netinet/in_gif.h> 139#include "gif.h" 140#include <net/if_gre.h> 141#include "gre.h" 142 143#ifdef MROUTING 144#include <netinet/ip_mroute.h> 145#endif 146 147#ifdef IPSEC 148#include <netinet6/ipsec.h> 149#include <netkey/key.h> 150#endif 151#ifdef FAST_IPSEC 152#include <netipsec/ipsec.h> 153#include <netipsec/key.h> 154#endif /* FAST_IPSEC*/ 155 156#ifndef IPFORWARDING 157#ifdef GATEWAY 158#define IPFORWARDING 1 /* forward IP packets not for us */ 159#else /* GATEWAY */ 160#define IPFORWARDING 0 /* don't forward IP packets not for us */ 161#endif /* GATEWAY */ 162#endif /* IPFORWARDING */ 163#ifndef IPSENDREDIRECTS 164#define IPSENDREDIRECTS 1 165#endif 166#ifndef IPFORWSRCRT 167#define IPFORWSRCRT 1 /* forward source-routed packets */ 168#endif 169#ifndef IPALLOWSRCRT 170#define IPALLOWSRCRT 1 /* allow source-routed packets */ 171#endif 172#ifndef IPMTUDISC 173#define IPMTUDISC 1 174#endif 175#ifndef IPMTUDISCTIMEOUT 176#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 177#endif 178 179/* 180 * Note: DIRECTED_BROADCAST is handled this way so that previous 181 * configuration using this option will Just Work. 182 */ 183#ifndef IPDIRECTEDBCAST 184#ifdef DIRECTED_BROADCAST 185#define IPDIRECTEDBCAST 1 186#else 187#define IPDIRECTEDBCAST 0 188#endif /* DIRECTED_BROADCAST */ 189#endif /* IPDIRECTEDBCAST */ 190int ipforwarding = IPFORWARDING; 191int ipsendredirects = IPSENDREDIRECTS; 192int ip_defttl = IPDEFTTL; 193int ip_forwsrcrt = IPFORWSRCRT; 194int ip_directedbcast = IPDIRECTEDBCAST; 195int ip_allowsrcrt = IPALLOWSRCRT; 196int ip_mtudisc = IPMTUDISC; 197int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 198#ifdef DIAGNOSTIC 199int ipprintfs = 0; 200#endif 201 202int ip_do_randomid = 0; 203 204/* 205 * XXX - Setting ip_checkinterface mostly implements the receive side of 206 * the Strong ES model described in RFC 1122, but since the routing table 207 * and transmit implementation do not implement the Strong ES model, 208 * setting this to 1 results in an odd hybrid. 209 * 210 * XXX - ip_checkinterface currently must be disabled if you use ipnat 211 * to translate the destination address to another local interface. 212 * 213 * XXX - ip_checkinterface must be disabled if you add IP aliases 214 * to the loopback interface instead of the interface where the 215 * packets for those addresses are received. 216 */ 217int ip_checkinterface = 0; 218 219 220struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 221 222int ipqmaxlen = IFQ_MAXLEN; 223u_long in_ifaddrhash; /* size of hash table - 1 */ 224int in_ifaddrentries; /* total number of addrs */ 225struct in_ifaddrhead in_ifaddrhead; 226struct in_ifaddrhashhead *in_ifaddrhashtbl; 227u_long in_multihash; /* size of hash table - 1 */ 228int in_multientries; /* total number of addrs */ 229struct in_multihashhead *in_multihashtbl; 230struct ifqueue ipintrq; 231struct ipstat ipstat; 232uint16_t ip_id; 233 234#ifdef PFIL_HOOKS 235struct pfil_head inet_pfil_hook; 236#endif 237 238/* 239 * Cached copy of nmbclusters. If nbclusters is different, 240 * recalculate IP parameters derived from nmbclusters. 241 */ 242static int ip_nmbclusters; /* copy of nmbclusters */ 243static void ip_nmbclusters_changed __P((void)); /* recalc limits */ 244 245#define CHECK_NMBCLUSTER_PARAMS() \ 246do { \ 247 if (__predict_false(ip_nmbclusters != nmbclusters)) \ 248 ip_nmbclusters_changed(); \ 249} while (/*CONSTCOND*/0) 250 251/* IP datagram reassembly queues (hashed) */ 252#define IPREASS_NHASH_LOG2 6 253#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 254#define IPREASS_HMASK (IPREASS_NHASH - 1) 255#define IPREASS_HASH(x,y) \ 256 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 257struct ipqhead ipq[IPREASS_NHASH]; 258int ipq_locked; 259static int ip_nfragpackets; /* packets in reass queue */ 260static int ip_nfrags; /* total fragments in reass queues */ 261 262int ip_maxfragpackets = 200; /* limit on packets. XXX sysctl */ 263int ip_maxfrags; /* limit on fragments. XXX sysctl */ 264 265 266/* 267 * Additive-Increase/Multiplicative-Decrease (AIMD) strategy for 268 * IP reassembly queue buffer managment. 269 * 270 * We keep a count of total IP fragments (NB: not fragmented packets!) 271 * awaiting reassembly (ip_nfrags) and a limit (ip_maxfrags) on fragments. 272 * If ip_nfrags exceeds ip_maxfrags the limit, we drop half the 273 * total fragments in reassembly queues.This AIMD policy avoids 274 * repeatedly deleting single packets under heavy fragmentation load 275 * (e.g., from lossy NFS peers). 276 */ 277static u_int ip_reass_ttl_decr __P((u_int ticks)); 278static void ip_reass_drophalf __P((void)); 279 280 281static __inline int ipq_lock_try __P((void)); 282static __inline void ipq_unlock __P((void)); 283 284static __inline int 285ipq_lock_try() 286{ 287 int s; 288 289 /* 290 * Use splvm() -- we're blocking things that would cause 291 * mbuf allocation. 292 */ 293 s = splvm(); 294 if (ipq_locked) { 295 splx(s); 296 return (0); 297 } 298 ipq_locked = 1; 299 splx(s); 300 return (1); 301} 302 303static __inline void 304ipq_unlock() 305{ 306 int s; 307 308 s = splvm(); 309 ipq_locked = 0; 310 splx(s); 311} 312 313#ifdef DIAGNOSTIC 314#define IPQ_LOCK() \ 315do { \ 316 if (ipq_lock_try() == 0) { \ 317 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 318 panic("ipq_lock"); \ 319 } \ 320} while (/*CONSTCOND*/ 0) 321#define IPQ_LOCK_CHECK() \ 322do { \ 323 if (ipq_locked == 0) { \ 324 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 325 panic("ipq lock check"); \ 326 } \ 327} while (/*CONSTCOND*/ 0) 328#else 329#define IPQ_LOCK() (void) ipq_lock_try() 330#define IPQ_LOCK_CHECK() /* nothing */ 331#endif 332 333#define IPQ_UNLOCK() ipq_unlock() 334 335POOL_INIT(inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl", NULL); 336POOL_INIT(ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", NULL); 337 338#ifdef INET_CSUM_COUNTERS 339#include <sys/device.h> 340 341struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 342 NULL, "inet", "hwcsum bad"); 343struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 344 NULL, "inet", "hwcsum ok"); 345struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 346 NULL, "inet", "swcsum"); 347 348#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 349 350EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 351EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 352EVCNT_ATTACH_STATIC(ip_swcsum); 353 354#else 355 356#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 357 358#endif /* INET_CSUM_COUNTERS */ 359 360/* 361 * We need to save the IP options in case a protocol wants to respond 362 * to an incoming packet over the same route if the packet got here 363 * using IP source routing. This allows connection establishment and 364 * maintenance when the remote end is on a network that is not known 365 * to us. 366 */ 367int ip_nhops = 0; 368static struct ip_srcrt { 369 struct in_addr dst; /* final destination */ 370 char nop; /* one NOP to align */ 371 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 372 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 373} ip_srcrt; 374 375static void save_rte __P((u_char *, struct in_addr)); 376 377#ifdef MBUFTRACE 378struct mowner ip_rx_mowner = { "internet", "rx" }; 379struct mowner ip_tx_mowner = { "internet", "tx" }; 380#endif 381 382/* 383 * Compute IP limits derived from the value of nmbclusters. 384 */ 385static void 386ip_nmbclusters_changed(void) 387{ 388 ip_maxfrags = nmbclusters / 4; 389 ip_nmbclusters = nmbclusters; 390} 391 392/* 393 * IP initialization: fill in IP protocol switch table. 394 * All protocols not implemented in kernel go to raw IP protocol handler. 395 */ 396void 397ip_init() 398{ 399 const struct protosw *pr; 400 int i; 401 402 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 403 if (pr == 0) 404 panic("ip_init"); 405 for (i = 0; i < IPPROTO_MAX; i++) 406 ip_protox[i] = pr - inetsw; 407 for (pr = inetdomain.dom_protosw; 408 pr < inetdomain.dom_protoswNPROTOSW; pr++) 409 if (pr->pr_domain->dom_family == PF_INET && 410 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 411 ip_protox[pr->pr_protocol] = pr - inetsw; 412 413 for (i = 0; i < IPREASS_NHASH; i++) 414 LIST_INIT(&ipq[i]); 415 416 ip_id = time.tv_sec & 0xfffff; 417 418 ipintrq.ifq_maxlen = ipqmaxlen; 419 ip_nmbclusters_changed(); 420 421 TAILQ_INIT(&in_ifaddrhead); 422 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 423 M_WAITOK, &in_ifaddrhash); 424 in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IPMADDR, 425 M_WAITOK, &in_multihash); 426 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 427#ifdef GATEWAY 428 ipflow_init(); 429#endif 430 431#ifdef PFIL_HOOKS 432 /* Register our Packet Filter hook. */ 433 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 434 inet_pfil_hook.ph_af = AF_INET; 435 i = pfil_head_register(&inet_pfil_hook); 436 if (i != 0) 437 printf("ip_init: WARNING: unable to register pfil hook, " 438 "error %d\n", i); 439#endif /* PFIL_HOOKS */ 440 441#ifdef MBUFTRACE 442 MOWNER_ATTACH(&ip_tx_mowner); 443 MOWNER_ATTACH(&ip_rx_mowner); 444#endif /* MBUFTRACE */ 445} 446 447struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 448struct route ipforward_rt; 449 450/* 451 * IP software interrupt routine 452 */ 453void 454ipintr() 455{ 456 int s; 457 struct mbuf *m; 458 459 while (1) { 460 s = splnet(); 461 IF_DEQUEUE(&ipintrq, m); 462 splx(s); 463 if (m == 0) 464 return; 465 MCLAIM(m, &ip_rx_mowner); 466 ip_input(m); 467 } 468} 469 470/* 471 * Ip input routine. Checksum and byte swap header. If fragmented 472 * try to reassemble. Process options. Pass to next level. 473 */ 474void 475ip_input(struct mbuf *m) 476{ 477 struct ip *ip = NULL; 478 struct ipq *fp; 479 struct in_ifaddr *ia; 480 struct ifaddr *ifa; 481 struct ipqent *ipqe; 482 int hlen = 0, mff, len; 483 int downmatch; 484 int checkif; 485 int srcrt = 0; 486 u_int hash; 487#ifdef FAST_IPSEC 488 struct m_tag *mtag; 489 struct tdb_ident *tdbi; 490 struct secpolicy *sp; 491 int s, error; 492#endif /* FAST_IPSEC */ 493 494 MCLAIM(m, &ip_rx_mowner); 495#ifdef DIAGNOSTIC 496 if ((m->m_flags & M_PKTHDR) == 0) 497 panic("ipintr no HDR"); 498#endif 499 500 /* 501 * If no IP addresses have been set yet but the interfaces 502 * are receiving, can't do anything with incoming packets yet. 503 */ 504 if (TAILQ_FIRST(&in_ifaddrhead) == 0) 505 goto bad; 506 ipstat.ips_total++; 507 /* 508 * If the IP header is not aligned, slurp it up into a new 509 * mbuf with space for link headers, in the event we forward 510 * it. Otherwise, if it is aligned, make sure the entire 511 * base IP header is in the first mbuf of the chain. 512 */ 513 if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) { 514 if ((m = m_copyup(m, sizeof(struct ip), 515 (max_linkhdr + 3) & ~3)) == NULL) { 516 /* XXXJRT new stat, please */ 517 ipstat.ips_toosmall++; 518 return; 519 } 520 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 521 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 522 ipstat.ips_toosmall++; 523 return; 524 } 525 } 526 ip = mtod(m, struct ip *); 527 if (ip->ip_v != IPVERSION) { 528 ipstat.ips_badvers++; 529 goto bad; 530 } 531 hlen = ip->ip_hl << 2; 532 if (hlen < sizeof(struct ip)) { /* minimum header length */ 533 ipstat.ips_badhlen++; 534 goto bad; 535 } 536 if (hlen > m->m_len) { 537 if ((m = m_pullup(m, hlen)) == 0) { 538 ipstat.ips_badhlen++; 539 return; 540 } 541 ip = mtod(m, struct ip *); 542 } 543 544 /* 545 * RFC1122: packets with a multicast source address are 546 * not allowed. 547 */ 548 if (IN_MULTICAST(ip->ip_src.s_addr)) { 549 ipstat.ips_badaddr++; 550 goto bad; 551 } 552 553 /* 127/8 must not appear on wire - RFC1122 */ 554 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 555 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 556 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 557 ipstat.ips_badaddr++; 558 goto bad; 559 } 560 } 561 562 switch (m->m_pkthdr.csum_flags & 563 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) | 564 M_CSUM_IPv4_BAD)) { 565 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 566 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 567 goto badcsum; 568 569 case M_CSUM_IPv4: 570 /* Checksum was okay. */ 571 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 572 break; 573 574 default: 575 /* Must compute it ourselves. */ 576 INET_CSUM_COUNTER_INCR(&ip_swcsum); 577 if (in_cksum(m, hlen) != 0) 578 goto badcsum; 579 break; 580 } 581 582 /* Retrieve the packet length. */ 583 len = ntohs(ip->ip_len); 584 585 /* 586 * Check for additional length bogosity 587 */ 588 if (len < hlen) { 589 ipstat.ips_badlen++; 590 goto bad; 591 } 592 593 /* 594 * Check that the amount of data in the buffers 595 * is as at least much as the IP header would have us expect. 596 * Trim mbufs if longer than we expect. 597 * Drop packet if shorter than we expect. 598 */ 599 if (m->m_pkthdr.len < len) { 600 ipstat.ips_tooshort++; 601 goto bad; 602 } 603 if (m->m_pkthdr.len > len) { 604 if (m->m_len == m->m_pkthdr.len) { 605 m->m_len = len; 606 m->m_pkthdr.len = len; 607 } else 608 m_adj(m, len - m->m_pkthdr.len); 609 } 610 611#if defined(IPSEC) 612 /* ipflow (IP fast forwarding) is not compatible with IPsec. */ 613 m->m_flags &= ~M_CANFASTFWD; 614#else 615 /* 616 * Assume that we can create a fast-forward IP flow entry 617 * based on this packet. 618 */ 619 m->m_flags |= M_CANFASTFWD; 620#endif 621 622#ifdef PFIL_HOOKS 623 /* 624 * Run through list of hooks for input packets. If there are any 625 * filters which require that additional packets in the flow are 626 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 627 * Note that filters must _never_ set this flag, as another filter 628 * in the list may have previously cleared it. 629 */ 630 /* 631 * let ipfilter look at packet on the wire, 632 * not the decapsulated packet. 633 */ 634#ifdef IPSEC 635 if (!ipsec_getnhist(m)) 636#elif defined(FAST_IPSEC) 637 if (!ipsec_indone(m)) 638#else 639 if (1) 640#endif 641 { 642 struct in_addr odst; 643 644 odst = ip->ip_dst; 645 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 646 PFIL_IN) != 0) 647 return; 648 if (m == NULL) 649 return; 650 ip = mtod(m, struct ip *); 651 hlen = ip->ip_hl << 2; 652 /* 653 * XXX The setting of "srcrt" here is to prevent ip_forward() 654 * from generating ICMP redirects for packets that have 655 * been redirected by a hook back out on to the same LAN that 656 * they came from and is not an indication that the packet 657 * is being inffluenced by source routing options. This 658 * allows things like 659 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 660 * where tlp0 is both on the 1.1.1.0/24 network and is the 661 * default route for hosts on 1.1.1.0/24. Of course this 662 * also requires a "map tlp0 ..." to complete the story. 663 * One might argue whether or not this kind of network config. 664 * should be supported in this manner... 665 */ 666 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 667 } 668#endif /* PFIL_HOOKS */ 669 670#ifdef ALTQ 671 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 672 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 673 /* packet dropped by traffic conditioner */ 674 return; 675 } 676#endif 677 678 /* 679 * Process options and, if not destined for us, 680 * ship it on. ip_dooptions returns 1 when an 681 * error was detected (causing an icmp message 682 * to be sent and the original packet to be freed). 683 */ 684 ip_nhops = 0; /* for source routed packets */ 685 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 686 return; 687 688 /* 689 * Enable a consistency check between the destination address 690 * and the arrival interface for a unicast packet (the RFC 1122 691 * strong ES model) if IP forwarding is disabled and the packet 692 * is not locally generated. 693 * 694 * XXX - Checking also should be disabled if the destination 695 * address is ipnat'ed to a different interface. 696 * 697 * XXX - Checking is incompatible with IP aliases added 698 * to the loopback interface instead of the interface where 699 * the packets are received. 700 * 701 * XXX - We need to add a per ifaddr flag for this so that 702 * we get finer grain control. 703 */ 704 checkif = ip_checkinterface && (ipforwarding == 0) && 705 (m->m_pkthdr.rcvif != NULL) && 706 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0); 707 708 /* 709 * Check our list of addresses, to see if the packet is for us. 710 * 711 * Traditional 4.4BSD did not consult IFF_UP at all. 712 * The behavior here is to treat addresses on !IFF_UP interface 713 * as not mine. 714 */ 715 downmatch = 0; 716 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 717 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 718 if (checkif && ia->ia_ifp != m->m_pkthdr.rcvif) 719 continue; 720 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 721 break; 722 else 723 downmatch++; 724 } 725 } 726 if (ia != NULL) 727 goto ours; 728 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 729 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrlist, ifa_list) { 730 if (ifa->ifa_addr->sa_family != AF_INET) 731 continue; 732 ia = ifatoia(ifa); 733 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 734 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 735 /* 736 * Look for all-0's host part (old broadcast addr), 737 * either for subnet or net. 738 */ 739 ip->ip_dst.s_addr == ia->ia_subnet || 740 ip->ip_dst.s_addr == ia->ia_net) 741 goto ours; 742 /* 743 * An interface with IP address zero accepts 744 * all packets that arrive on that interface. 745 */ 746 if (in_nullhost(ia->ia_addr.sin_addr)) 747 goto ours; 748 } 749 } 750 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 751 struct in_multi *inm; 752#ifdef MROUTING 753 extern struct socket *ip_mrouter; 754 755 if (ip_mrouter) { 756 /* 757 * If we are acting as a multicast router, all 758 * incoming multicast packets are passed to the 759 * kernel-level multicast forwarding function. 760 * The packet is returned (relatively) intact; if 761 * ip_mforward() returns a non-zero value, the packet 762 * must be discarded, else it may be accepted below. 763 * 764 * (The IP ident field is put in the same byte order 765 * as expected when ip_mforward() is called from 766 * ip_output().) 767 */ 768 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 769 ipstat.ips_cantforward++; 770 m_freem(m); 771 return; 772 } 773 774 /* 775 * The process-level routing demon needs to receive 776 * all multicast IGMP packets, whether or not this 777 * host belongs to their destination groups. 778 */ 779 if (ip->ip_p == IPPROTO_IGMP) 780 goto ours; 781 ipstat.ips_forward++; 782 } 783#endif 784 /* 785 * See if we belong to the destination multicast group on the 786 * arrival interface. 787 */ 788 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 789 if (inm == NULL) { 790 ipstat.ips_cantforward++; 791 m_freem(m); 792 return; 793 } 794 goto ours; 795 } 796 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 797 in_nullhost(ip->ip_dst)) 798 goto ours; 799 800 /* 801 * Not for us; forward if possible and desirable. 802 */ 803 if (ipforwarding == 0) { 804 ipstat.ips_cantforward++; 805 m_freem(m); 806 } else { 807 /* 808 * If ip_dst matched any of my address on !IFF_UP interface, 809 * and there's no IFF_UP interface that matches ip_dst, 810 * send icmp unreach. Forwarding it will result in in-kernel 811 * forwarding loop till TTL goes to 0. 812 */ 813 if (downmatch) { 814 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 815 ipstat.ips_cantforward++; 816 return; 817 } 818#ifdef IPSEC 819 if (ipsec4_in_reject(m, NULL)) { 820 ipsecstat.in_polvio++; 821 goto bad; 822 } 823#endif 824#ifdef FAST_IPSEC 825 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 826 s = splsoftnet(); 827 if (mtag != NULL) { 828 tdbi = (struct tdb_ident *)(mtag + 1); 829 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 830 } else { 831 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 832 IP_FORWARDING, &error); 833 } 834 if (sp == NULL) { /* NB: can happen if error */ 835 splx(s); 836 /*XXX error stat???*/ 837 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 838 goto bad; 839 } 840 841 /* 842 * Check security policy against packet attributes. 843 */ 844 error = ipsec_in_reject(sp, m); 845 KEY_FREESP(&sp); 846 splx(s); 847 if (error) { 848 ipstat.ips_cantforward++; 849 goto bad; 850 } 851 852 /* 853 * Peek at the outbound SP for this packet to determine if 854 * it's a Fast Forward candidate. 855 */ 856 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 857 if (mtag != NULL) 858 m->m_flags &= ~M_CANFASTFWD; 859 else { 860 s = splsoftnet(); 861 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, 862 (IP_FORWARDING | 863 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 864 &error, NULL); 865 if (sp != NULL) { 866 m->m_flags &= ~M_CANFASTFWD; 867 KEY_FREESP(&sp); 868 } 869 splx(s); 870 } 871#endif /* FAST_IPSEC */ 872 873 ip_forward(m, srcrt); 874 } 875 return; 876 877ours: 878 /* 879 * If offset or IP_MF are set, must reassemble. 880 * Otherwise, nothing need be done. 881 * (We could look in the reassembly queue to see 882 * if the packet was previously fragmented, 883 * but it's not worth the time; just let them time out.) 884 */ 885 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 886 887 /* 888 * Look for queue of fragments 889 * of this datagram. 890 */ 891 IPQ_LOCK(); 892 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 893 /* XXX LIST_FOREACH(fp, &ipq[hash], ipq_q) */ 894 for (fp = LIST_FIRST(&ipq[hash]); fp != NULL; 895 fp = LIST_NEXT(fp, ipq_q)) { 896 if (ip->ip_id == fp->ipq_id && 897 in_hosteq(ip->ip_src, fp->ipq_src) && 898 in_hosteq(ip->ip_dst, fp->ipq_dst) && 899 ip->ip_p == fp->ipq_p) 900 goto found; 901 902 } 903 fp = 0; 904found: 905 906 /* 907 * Adjust ip_len to not reflect header, 908 * set ipqe_mff if more fragments are expected, 909 * convert offset of this to bytes. 910 */ 911 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 912 mff = (ip->ip_off & htons(IP_MF)) != 0; 913 if (mff) { 914 /* 915 * Make sure that fragments have a data length 916 * that's a non-zero multiple of 8 bytes. 917 */ 918 if (ntohs(ip->ip_len) == 0 || 919 (ntohs(ip->ip_len) & 0x7) != 0) { 920 ipstat.ips_badfrags++; 921 IPQ_UNLOCK(); 922 goto bad; 923 } 924 } 925 ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3); 926 927 /* 928 * If datagram marked as having more fragments 929 * or if this is not the first fragment, 930 * attempt reassembly; if it succeeds, proceed. 931 */ 932 if (mff || ip->ip_off != htons(0)) { 933 ipstat.ips_fragments++; 934 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 935 if (ipqe == NULL) { 936 ipstat.ips_rcvmemdrop++; 937 IPQ_UNLOCK(); 938 goto bad; 939 } 940 ipqe->ipqe_mff = mff; 941 ipqe->ipqe_m = m; 942 ipqe->ipqe_ip = ip; 943 m = ip_reass(ipqe, fp, &ipq[hash]); 944 if (m == 0) { 945 IPQ_UNLOCK(); 946 return; 947 } 948 ipstat.ips_reassembled++; 949 ip = mtod(m, struct ip *); 950 hlen = ip->ip_hl << 2; 951 ip->ip_len = htons(ntohs(ip->ip_len) + hlen); 952 } else 953 if (fp) 954 ip_freef(fp); 955 IPQ_UNLOCK(); 956 } 957 958#if defined(IPSEC) 959 /* 960 * enforce IPsec policy checking if we are seeing last header. 961 * note that we do not visit this with protocols with pcb layer 962 * code - like udp/tcp/raw ip. 963 */ 964 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 965 ipsec4_in_reject(m, NULL)) { 966 ipsecstat.in_polvio++; 967 goto bad; 968 } 969#endif 970#if FAST_IPSEC 971 /* 972 * enforce IPsec policy checking if we are seeing last header. 973 * note that we do not visit this with protocols with pcb layer 974 * code - like udp/tcp/raw ip. 975 */ 976 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 977 /* 978 * Check if the packet has already had IPsec processing 979 * done. If so, then just pass it along. This tag gets 980 * set during AH, ESP, etc. input handling, before the 981 * packet is returned to the ip input queue for delivery. 982 */ 983 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 984 s = splsoftnet(); 985 if (mtag != NULL) { 986 tdbi = (struct tdb_ident *)(mtag + 1); 987 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 988 } else { 989 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 990 IP_FORWARDING, &error); 991 } 992 if (sp != NULL) { 993 /* 994 * Check security policy against packet attributes. 995 */ 996 error = ipsec_in_reject(sp, m); 997 KEY_FREESP(&sp); 998 } else { 999 /* XXX error stat??? */ 1000 error = EINVAL; 1001DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 1002 goto bad; 1003 } 1004 splx(s); 1005 if (error) 1006 goto bad; 1007 } 1008#endif /* FAST_IPSEC */ 1009 1010 /* 1011 * Switch out to protocol's input routine. 1012 */ 1013#if IFA_STATS 1014 if (ia && ip) 1015 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 1016#endif 1017 ipstat.ips_delivered++; 1018 { 1019 int off = hlen, nh = ip->ip_p; 1020 1021 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 1022 return; 1023 } 1024bad: 1025 m_freem(m); 1026 return; 1027 1028badcsum: 1029 ipstat.ips_badsum++; 1030 m_freem(m); 1031} 1032 1033/* 1034 * Take incoming datagram fragment and try to 1035 * reassemble it into whole datagram. If a chain for 1036 * reassembly of this datagram already exists, then it 1037 * is given as fp; otherwise have to make a chain. 1038 */ 1039struct mbuf * 1040ip_reass(ipqe, fp, ipqhead) 1041 struct ipqent *ipqe; 1042 struct ipq *fp; 1043 struct ipqhead *ipqhead; 1044{ 1045 struct mbuf *m = ipqe->ipqe_m; 1046 struct ipqent *nq, *p, *q; 1047 struct ip *ip; 1048 struct mbuf *t; 1049 int hlen = ipqe->ipqe_ip->ip_hl << 2; 1050 int i, next; 1051 1052 IPQ_LOCK_CHECK(); 1053 1054 /* 1055 * Presence of header sizes in mbufs 1056 * would confuse code below. 1057 */ 1058 m->m_data += hlen; 1059 m->m_len -= hlen; 1060 1061#ifdef notyet 1062 /* make sure fragment limit is up-to-date */ 1063 CHECK_NMBCLUSTER_PARAMS(); 1064 1065 /* If we have too many fragments, drop the older half. */ 1066 if (ip_nfrags >= ip_maxfrags) 1067 ip_reass_drophalf(void); 1068#endif 1069 1070 /* 1071 * We are about to add a fragment; increment frag count. 1072 */ 1073 ip_nfrags++; 1074 1075 /* 1076 * If first fragment to arrive, create a reassembly queue. 1077 */ 1078 if (fp == 0) { 1079 /* 1080 * Enforce upper bound on number of fragmented packets 1081 * for which we attempt reassembly; 1082 * If maxfrag is 0, never accept fragments. 1083 * If maxfrag is -1, accept all fragments without limitation. 1084 */ 1085 if (ip_maxfragpackets < 0) 1086 ; 1087 else if (ip_nfragpackets >= ip_maxfragpackets) 1088 goto dropfrag; 1089 ip_nfragpackets++; 1090 MALLOC(fp, struct ipq *, sizeof (struct ipq), 1091 M_FTABLE, M_NOWAIT); 1092 if (fp == NULL) 1093 goto dropfrag; 1094 LIST_INSERT_HEAD(ipqhead, fp, ipq_q); 1095 fp->ipq_nfrags = 1; 1096 fp->ipq_ttl = IPFRAGTTL; 1097 fp->ipq_p = ipqe->ipqe_ip->ip_p; 1098 fp->ipq_id = ipqe->ipqe_ip->ip_id; 1099 TAILQ_INIT(&fp->ipq_fragq); 1100 fp->ipq_src = ipqe->ipqe_ip->ip_src; 1101 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 1102 p = NULL; 1103 goto insert; 1104 } else { 1105 fp->ipq_nfrags++; 1106 } 1107 1108 /* 1109 * Find a segment which begins after this one does. 1110 */ 1111 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1112 p = q, q = TAILQ_NEXT(q, ipqe_q)) 1113 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off)) 1114 break; 1115 1116 /* 1117 * If there is a preceding segment, it may provide some of 1118 * our data already. If so, drop the data from the incoming 1119 * segment. If it provides all of our data, drop us. 1120 */ 1121 if (p != NULL) { 1122 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) - 1123 ntohs(ipqe->ipqe_ip->ip_off); 1124 if (i > 0) { 1125 if (i >= ntohs(ipqe->ipqe_ip->ip_len)) 1126 goto dropfrag; 1127 m_adj(ipqe->ipqe_m, i); 1128 ipqe->ipqe_ip->ip_off = 1129 htons(ntohs(ipqe->ipqe_ip->ip_off) + i); 1130 ipqe->ipqe_ip->ip_len = 1131 htons(ntohs(ipqe->ipqe_ip->ip_len) - i); 1132 } 1133 } 1134 1135 /* 1136 * While we overlap succeeding segments trim them or, 1137 * if they are completely covered, dequeue them. 1138 */ 1139 for (; q != NULL && 1140 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) > 1141 ntohs(q->ipqe_ip->ip_off); q = nq) { 1142 i = (ntohs(ipqe->ipqe_ip->ip_off) + 1143 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off); 1144 if (i < ntohs(q->ipqe_ip->ip_len)) { 1145 q->ipqe_ip->ip_len = 1146 htons(ntohs(q->ipqe_ip->ip_len) - i); 1147 q->ipqe_ip->ip_off = 1148 htons(ntohs(q->ipqe_ip->ip_off) + i); 1149 m_adj(q->ipqe_m, i); 1150 break; 1151 } 1152 nq = TAILQ_NEXT(q, ipqe_q); 1153 m_freem(q->ipqe_m); 1154 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1155 pool_put(&ipqent_pool, q); 1156 fp->ipq_nfrags--; 1157 ip_nfrags--; 1158 } 1159 1160insert: 1161 /* 1162 * Stick new segment in its place; 1163 * check for complete reassembly. 1164 */ 1165 if (p == NULL) { 1166 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 1167 } else { 1168 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q); 1169 } 1170 next = 0; 1171 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1172 p = q, q = TAILQ_NEXT(q, ipqe_q)) { 1173 if (ntohs(q->ipqe_ip->ip_off) != next) 1174 return (0); 1175 next += ntohs(q->ipqe_ip->ip_len); 1176 } 1177 if (p->ipqe_mff) 1178 return (0); 1179 1180 /* 1181 * Reassembly is complete. Check for a bogus message size and 1182 * concatenate fragments. 1183 */ 1184 q = TAILQ_FIRST(&fp->ipq_fragq); 1185 ip = q->ipqe_ip; 1186 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 1187 ipstat.ips_toolong++; 1188 ip_freef(fp); 1189 return (0); 1190 } 1191 m = q->ipqe_m; 1192 t = m->m_next; 1193 m->m_next = 0; 1194 m_cat(m, t); 1195 nq = TAILQ_NEXT(q, ipqe_q); 1196 pool_put(&ipqent_pool, q); 1197 for (q = nq; q != NULL; q = nq) { 1198 t = q->ipqe_m; 1199 nq = TAILQ_NEXT(q, ipqe_q); 1200 pool_put(&ipqent_pool, q); 1201 m_cat(m, t); 1202 } 1203 ip_nfrags -= fp->ipq_nfrags; 1204 1205 /* 1206 * Create header for new ip packet by 1207 * modifying header of first packet; 1208 * dequeue and discard fragment reassembly header. 1209 * Make header visible. 1210 */ 1211 ip->ip_len = htons(next); 1212 ip->ip_src = fp->ipq_src; 1213 ip->ip_dst = fp->ipq_dst; 1214 LIST_REMOVE(fp, ipq_q); 1215 FREE(fp, M_FTABLE); 1216 ip_nfragpackets--; 1217 m->m_len += (ip->ip_hl << 2); 1218 m->m_data -= (ip->ip_hl << 2); 1219 /* some debugging cruft by sklower, below, will go away soon */ 1220 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1221 int plen = 0; 1222 for (t = m; t; t = t->m_next) 1223 plen += t->m_len; 1224 m->m_pkthdr.len = plen; 1225 } 1226 return (m); 1227 1228dropfrag: 1229 if (fp != 0) 1230 fp->ipq_nfrags--; 1231 ip_nfrags--; 1232 ipstat.ips_fragdropped++; 1233 m_freem(m); 1234 pool_put(&ipqent_pool, ipqe); 1235 return (0); 1236} 1237 1238/* 1239 * Free a fragment reassembly header and all 1240 * associated datagrams. 1241 */ 1242void 1243ip_freef(fp) 1244 struct ipq *fp; 1245{ 1246 struct ipqent *q, *p; 1247 u_int nfrags = 0; 1248 1249 IPQ_LOCK_CHECK(); 1250 1251 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) { 1252 p = TAILQ_NEXT(q, ipqe_q); 1253 m_freem(q->ipqe_m); 1254 nfrags++; 1255 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1256 pool_put(&ipqent_pool, q); 1257 } 1258 1259 if (nfrags != fp->ipq_nfrags) 1260 printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags); 1261 ip_nfrags -= nfrags; 1262 LIST_REMOVE(fp, ipq_q); 1263 FREE(fp, M_FTABLE); 1264 ip_nfragpackets--; 1265} 1266 1267/* 1268 * IP reassembly TTL machinery for multiplicative drop. 1269 */ 1270static u_int fragttl_histo[(IPFRAGTTL+1)]; 1271 1272 1273/* 1274 * Decrement TTL of all reasembly queue entries by `ticks'. 1275 * Count number of distinct fragments (as opposed to partial, fragmented 1276 * datagrams) in the reassembly queue. While we traverse the entire 1277 * reassembly queue, compute and return the median TTL over all fragments. 1278 */ 1279static u_int 1280ip_reass_ttl_decr(u_int ticks) 1281{ 1282 u_int nfrags, median, dropfraction, keepfraction; 1283 struct ipq *fp, *nfp; 1284 int i; 1285 1286 nfrags = 0; 1287 memset(fragttl_histo, 0, sizeof fragttl_histo); 1288 1289 for (i = 0; i < IPREASS_NHASH; i++) { 1290 for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) { 1291 fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ? 1292 0 : fp->ipq_ttl - ticks); 1293 nfp = LIST_NEXT(fp, ipq_q); 1294 if (fp->ipq_ttl == 0) { 1295 ipstat.ips_fragtimeout++; 1296 ip_freef(fp); 1297 } else { 1298 nfrags += fp->ipq_nfrags; 1299 fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags; 1300 } 1301 } 1302 } 1303 1304 KASSERT(ip_nfrags == nfrags); 1305 1306 /* Find median (or other drop fraction) in histogram. */ 1307 dropfraction = (ip_nfrags / 2); 1308 keepfraction = ip_nfrags - dropfraction; 1309 for (i = IPFRAGTTL, median = 0; i >= 0; i--) { 1310 median += fragttl_histo[i]; 1311 if (median >= keepfraction) 1312 break; 1313 } 1314 1315 /* Return TTL of median (or other fraction). */ 1316 return (u_int)i; 1317} 1318 1319void 1320ip_reass_drophalf(void) 1321{ 1322 1323 u_int median_ticks; 1324 /* 1325 * Compute median TTL of all fragments, and count frags 1326 * with that TTL or lower (roughly half of all fragments). 1327 */ 1328 median_ticks = ip_reass_ttl_decr(0); 1329 1330 /* Drop half. */ 1331 median_ticks = ip_reass_ttl_decr(median_ticks); 1332 1333} 1334 1335/* 1336 * IP timer processing; 1337 * if a timer expires on a reassembly 1338 * queue, discard it. 1339 */ 1340void 1341ip_slowtimo() 1342{ 1343 static u_int dropscanidx = 0; 1344 u_int i; 1345 u_int median_ttl; 1346 int s = splsoftnet(); 1347 1348 IPQ_LOCK(); 1349 1350 /* Age TTL of all fragments by 1 tick .*/ 1351 median_ttl = ip_reass_ttl_decr(1); 1352 1353 /* make sure fragment limit is up-to-date */ 1354 CHECK_NMBCLUSTER_PARAMS(); 1355 1356 /* If we have too many fragments, drop the older half. */ 1357 if (ip_nfrags > ip_maxfrags) 1358 ip_reass_ttl_decr(median_ttl); 1359 1360 /* 1361 * If we are over the maximum number of fragmented packets 1362 * (due to the limit being lowered), drain off 1363 * enough to get down to the new limit. Start draining 1364 * from the reassembly hashqueue most recently drained. 1365 */ 1366 if (ip_maxfragpackets < 0) 1367 ; 1368 else { 1369 int wrapped = 0; 1370 1371 i = dropscanidx; 1372 while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) { 1373 while (LIST_FIRST(&ipq[i]) != NULL) 1374 ip_freef(LIST_FIRST(&ipq[i])); 1375 if (++i >= IPREASS_NHASH) { 1376 i = 0; 1377 } 1378 /* 1379 * Dont scan forever even if fragment counters are 1380 * wrong: stop after scanning entire reassembly queue. 1381 */ 1382 if (i == dropscanidx) 1383 wrapped = 1; 1384 } 1385 dropscanidx = i; 1386 } 1387 IPQ_UNLOCK(); 1388#ifdef GATEWAY 1389 ipflow_slowtimo(); 1390#endif 1391 splx(s); 1392} 1393 1394/* 1395 * Drain off all datagram fragments. 1396 */ 1397void 1398ip_drain() 1399{ 1400 1401 /* 1402 * We may be called from a device's interrupt context. If 1403 * the ipq is already busy, just bail out now. 1404 */ 1405 if (ipq_lock_try() == 0) 1406 return; 1407 1408 /* 1409 * Drop half the total fragments now. If more mbufs are needed, 1410 * we will be called again soon. 1411 */ 1412 ip_reass_drophalf(); 1413 1414 IPQ_UNLOCK(); 1415} 1416 1417/* 1418 * Do option processing on a datagram, 1419 * possibly discarding it if bad options are encountered, 1420 * or forwarding it if source-routed. 1421 * Returns 1 if packet has been forwarded/freed, 1422 * 0 if the packet should be processed further. 1423 */ 1424int 1425ip_dooptions(m) 1426 struct mbuf *m; 1427{ 1428 struct ip *ip = mtod(m, struct ip *); 1429 u_char *cp, *cp0; 1430 struct ip_timestamp *ipt; 1431 struct in_ifaddr *ia; 1432 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1433 struct in_addr dst; 1434 n_time ntime; 1435 1436 dst = ip->ip_dst; 1437 cp = (u_char *)(ip + 1); 1438 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1439 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1440 opt = cp[IPOPT_OPTVAL]; 1441 if (opt == IPOPT_EOL) 1442 break; 1443 if (opt == IPOPT_NOP) 1444 optlen = 1; 1445 else { 1446 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1447 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1448 goto bad; 1449 } 1450 optlen = cp[IPOPT_OLEN]; 1451 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1452 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1453 goto bad; 1454 } 1455 } 1456 switch (opt) { 1457 1458 default: 1459 break; 1460 1461 /* 1462 * Source routing with record. 1463 * Find interface with current destination address. 1464 * If none on this machine then drop if strictly routed, 1465 * or do nothing if loosely routed. 1466 * Record interface address and bring up next address 1467 * component. If strictly routed make sure next 1468 * address is on directly accessible net. 1469 */ 1470 case IPOPT_LSRR: 1471 case IPOPT_SSRR: 1472 if (ip_allowsrcrt == 0) { 1473 type = ICMP_UNREACH; 1474 code = ICMP_UNREACH_NET_PROHIB; 1475 goto bad; 1476 } 1477 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1478 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1479 goto bad; 1480 } 1481 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1482 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1483 goto bad; 1484 } 1485 ipaddr.sin_addr = ip->ip_dst; 1486 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1487 if (ia == 0) { 1488 if (opt == IPOPT_SSRR) { 1489 type = ICMP_UNREACH; 1490 code = ICMP_UNREACH_SRCFAIL; 1491 goto bad; 1492 } 1493 /* 1494 * Loose routing, and not at next destination 1495 * yet; nothing to do except forward. 1496 */ 1497 break; 1498 } 1499 off--; /* 0 origin */ 1500 if ((off + sizeof(struct in_addr)) > optlen) { 1501 /* 1502 * End of source route. Should be for us. 1503 */ 1504 save_rte(cp, ip->ip_src); 1505 break; 1506 } 1507 /* 1508 * locate outgoing interface 1509 */ 1510 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1511 sizeof(ipaddr.sin_addr)); 1512 if (opt == IPOPT_SSRR) 1513 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); 1514 else 1515 ia = ip_rtaddr(ipaddr.sin_addr); 1516 if (ia == 0) { 1517 type = ICMP_UNREACH; 1518 code = ICMP_UNREACH_SRCFAIL; 1519 goto bad; 1520 } 1521 ip->ip_dst = ipaddr.sin_addr; 1522 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1523 (caddr_t)(cp + off), sizeof(struct in_addr)); 1524 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1525 /* 1526 * Let ip_intr's mcast routing check handle mcast pkts 1527 */ 1528 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1529 break; 1530 1531 case IPOPT_RR: 1532 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1533 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1534 goto bad; 1535 } 1536 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1537 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1538 goto bad; 1539 } 1540 /* 1541 * If no space remains, ignore. 1542 */ 1543 off--; /* 0 origin */ 1544 if ((off + sizeof(struct in_addr)) > optlen) 1545 break; 1546 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1547 sizeof(ipaddr.sin_addr)); 1548 /* 1549 * locate outgoing interface; if we're the destination, 1550 * use the incoming interface (should be same). 1551 */ 1552 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1553 == NULL && 1554 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1555 type = ICMP_UNREACH; 1556 code = ICMP_UNREACH_HOST; 1557 goto bad; 1558 } 1559 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1560 (caddr_t)(cp + off), sizeof(struct in_addr)); 1561 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1562 break; 1563 1564 case IPOPT_TS: 1565 code = cp - (u_char *)ip; 1566 ipt = (struct ip_timestamp *)cp; 1567 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1568 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1569 goto bad; 1570 } 1571 if (ipt->ipt_ptr < 5) { 1572 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1573 goto bad; 1574 } 1575 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1576 if (++ipt->ipt_oflw == 0) { 1577 code = (u_char *)&ipt->ipt_ptr - 1578 (u_char *)ip; 1579 goto bad; 1580 } 1581 break; 1582 } 1583 cp0 = (cp + ipt->ipt_ptr - 1); 1584 switch (ipt->ipt_flg) { 1585 1586 case IPOPT_TS_TSONLY: 1587 break; 1588 1589 case IPOPT_TS_TSANDADDR: 1590 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1591 sizeof(struct in_addr) > ipt->ipt_len) { 1592 code = (u_char *)&ipt->ipt_ptr - 1593 (u_char *)ip; 1594 goto bad; 1595 } 1596 ipaddr.sin_addr = dst; 1597 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1598 m->m_pkthdr.rcvif)); 1599 if (ia == 0) 1600 continue; 1601 bcopy(&ia->ia_addr.sin_addr, 1602 cp0, sizeof(struct in_addr)); 1603 ipt->ipt_ptr += sizeof(struct in_addr); 1604 break; 1605 1606 case IPOPT_TS_PRESPEC: 1607 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1608 sizeof(struct in_addr) > ipt->ipt_len) { 1609 code = (u_char *)&ipt->ipt_ptr - 1610 (u_char *)ip; 1611 goto bad; 1612 } 1613 bcopy(cp0, &ipaddr.sin_addr, 1614 sizeof(struct in_addr)); 1615 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1616 == NULL) 1617 continue; 1618 ipt->ipt_ptr += sizeof(struct in_addr); 1619 break; 1620 1621 default: 1622 /* XXX can't take &ipt->ipt_flg */ 1623 code = (u_char *)&ipt->ipt_ptr - 1624 (u_char *)ip + 1; 1625 goto bad; 1626 } 1627 ntime = iptime(); 1628 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1629 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1630 sizeof(n_time)); 1631 ipt->ipt_ptr += sizeof(n_time); 1632 } 1633 } 1634 if (forward) { 1635 if (ip_forwsrcrt == 0) { 1636 type = ICMP_UNREACH; 1637 code = ICMP_UNREACH_SRCFAIL; 1638 goto bad; 1639 } 1640 ip_forward(m, 1); 1641 return (1); 1642 } 1643 return (0); 1644bad: 1645 icmp_error(m, type, code, 0, 0); 1646 ipstat.ips_badoptions++; 1647 return (1); 1648} 1649 1650/* 1651 * Given address of next destination (final or next hop), 1652 * return internet address info of interface to be used to get there. 1653 */ 1654struct in_ifaddr * 1655ip_rtaddr(dst) 1656 struct in_addr dst; 1657{ 1658 struct sockaddr_in *sin; 1659 1660 sin = satosin(&ipforward_rt.ro_dst); 1661 1662 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1663 if (ipforward_rt.ro_rt) { 1664 RTFREE(ipforward_rt.ro_rt); 1665 ipforward_rt.ro_rt = 0; 1666 } 1667 sin->sin_family = AF_INET; 1668 sin->sin_len = sizeof(*sin); 1669 sin->sin_addr = dst; 1670 1671 rtalloc(&ipforward_rt); 1672 } 1673 if (ipforward_rt.ro_rt == 0) 1674 return ((struct in_ifaddr *)0); 1675 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1676} 1677 1678/* 1679 * Save incoming source route for use in replies, 1680 * to be picked up later by ip_srcroute if the receiver is interested. 1681 */ 1682void 1683save_rte(option, dst) 1684 u_char *option; 1685 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() 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(m, mopt) 1780 struct mbuf *m; 1781 struct mbuf *mopt; 1782{ 1783 int i; 1784 struct ip *ip = mtod(m, struct ip *); 1785 caddr_t opts; 1786 int olen; 1787 1788 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1789 opts = (caddr_t)(ip + 1); 1790 i = m->m_len - (sizeof (struct ip) + olen); 1791 bcopy(opts + olen, opts, (unsigned)i); 1792 m->m_len -= olen; 1793 if (m->m_flags & M_PKTHDR) 1794 m->m_pkthdr.len -= olen; 1795 ip->ip_len = htons(ntohs(ip->ip_len) - olen); 1796 ip->ip_hl = sizeof (struct ip) >> 2; 1797} 1798 1799const int inetctlerrmap[PRC_NCMDS] = { 1800 0, 0, 0, 0, 1801 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1802 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1803 EMSGSIZE, EHOSTUNREACH, 0, 0, 1804 0, 0, 0, 0, 1805 ENOPROTOOPT 1806}; 1807 1808/* 1809 * Forward a packet. If some error occurs return the sender 1810 * an icmp packet. Note we can't always generate a meaningful 1811 * icmp message because icmp doesn't have a large enough repertoire 1812 * of codes and types. 1813 * 1814 * If not forwarding, just drop the packet. This could be confusing 1815 * if ipforwarding was zero but some routing protocol was advancing 1816 * us as a gateway to somewhere. However, we must let the routing 1817 * protocol deal with that. 1818 * 1819 * The srcrt parameter indicates whether the packet is being forwarded 1820 * via a source route. 1821 */ 1822void 1823ip_forward(m, srcrt) 1824 struct mbuf *m; 1825 int srcrt; 1826{ 1827 struct ip *ip = mtod(m, struct ip *); 1828 struct sockaddr_in *sin; 1829 struct rtentry *rt; 1830 int error, type = 0, code = 0; 1831 struct mbuf *mcopy; 1832 n_long dest; 1833 struct ifnet *destifp; 1834#if defined(IPSEC) || defined(FAST_IPSEC) 1835 struct ifnet dummyifp; 1836#endif 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 %2.2x dst %2.2x ttl %x\n", 1852 ntohl(ip->ip_src.s_addr), 1853 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 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 ip->ip_ttl -= IPTTLDEC; 1865 1866 sin = satosin(&ipforward_rt.ro_dst); 1867 if ((rt = ipforward_rt.ro_rt) == 0 || 1868 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1869 if (ipforward_rt.ro_rt) { 1870 RTFREE(ipforward_rt.ro_rt); 1871 ipforward_rt.ro_rt = 0; 1872 } 1873 sin->sin_family = AF_INET; 1874 sin->sin_len = sizeof(struct sockaddr_in); 1875 sin->sin_addr = ip->ip_dst; 1876 1877 rtalloc(&ipforward_rt); 1878 if (ipforward_rt.ro_rt == 0) { 1879 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1880 return; 1881 } 1882 rt = ipforward_rt.ro_rt; 1883 } 1884 1885 /* 1886 * Save at most 68 bytes of the packet in case 1887 * we need to generate an ICMP message to the src. 1888 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1889 */ 1890 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1891 if (mcopy) 1892 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1893 1894 /* 1895 * If forwarding packet using same interface that it came in on, 1896 * perhaps should send a redirect to sender to shortcut a hop. 1897 * Only send redirect if source is sending directly to us, 1898 * and if packet was not source routed (or has any options). 1899 * Also, don't send redirect if forwarding using a default route 1900 * or a route modified by a redirect. 1901 */ 1902 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1903 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1904 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1905 ipsendredirects && !srcrt) { 1906 if (rt->rt_ifa && 1907 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1908 ifatoia(rt->rt_ifa)->ia_subnet) { 1909 if (rt->rt_flags & RTF_GATEWAY) 1910 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1911 else 1912 dest = ip->ip_dst.s_addr; 1913 /* 1914 * Router requirements says to only send host 1915 * redirects. 1916 */ 1917 type = ICMP_REDIRECT; 1918 code = ICMP_REDIRECT_HOST; 1919#ifdef DIAGNOSTIC 1920 if (ipprintfs) 1921 printf("redirect (%d) to %x\n", code, 1922 (u_int32_t)dest); 1923#endif 1924 } 1925 } 1926 1927 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1928 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1929 (struct ip_moptions *)NULL, (struct socket *)NULL); 1930 1931 if (error) 1932 ipstat.ips_cantforward++; 1933 else { 1934 ipstat.ips_forward++; 1935 if (type) 1936 ipstat.ips_redirectsent++; 1937 else { 1938 if (mcopy) { 1939#ifdef GATEWAY 1940 if (mcopy->m_flags & M_CANFASTFWD) 1941 ipflow_create(&ipforward_rt, mcopy); 1942#endif 1943 m_freem(mcopy); 1944 } 1945 return; 1946 } 1947 } 1948 if (mcopy == NULL) 1949 return; 1950 destifp = NULL; 1951 1952 switch (error) { 1953 1954 case 0: /* forwarded, but need redirect */ 1955 /* type, code set above */ 1956 break; 1957 1958 case ENETUNREACH: /* shouldn't happen, checked above */ 1959 case EHOSTUNREACH: 1960 case ENETDOWN: 1961 case EHOSTDOWN: 1962 default: 1963 type = ICMP_UNREACH; 1964 code = ICMP_UNREACH_HOST; 1965 break; 1966 1967 case EMSGSIZE: 1968 type = ICMP_UNREACH; 1969 code = ICMP_UNREACH_NEEDFRAG; 1970#if !defined(IPSEC) && !defined(FAST_IPSEC) 1971 if (ipforward_rt.ro_rt) 1972 destifp = ipforward_rt.ro_rt->rt_ifp; 1973#else 1974 /* 1975 * If the packet is routed over IPsec tunnel, tell the 1976 * originator the tunnel MTU. 1977 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1978 * XXX quickhack!!! 1979 */ 1980 if (ipforward_rt.ro_rt) { 1981 struct secpolicy *sp; 1982 int ipsecerror; 1983 size_t ipsechdr; 1984 struct route *ro; 1985 1986 sp = ipsec4_getpolicybyaddr(mcopy, 1987 IPSEC_DIR_OUTBOUND, IP_FORWARDING, 1988 &ipsecerror); 1989 1990 if (sp == NULL) 1991 destifp = ipforward_rt.ro_rt->rt_ifp; 1992 else { 1993 /* count IPsec header size */ 1994 ipsechdr = ipsec4_hdrsiz(mcopy, 1995 IPSEC_DIR_OUTBOUND, NULL); 1996 1997 /* 1998 * find the correct route for outer IPv4 1999 * header, compute tunnel MTU. 2000 * 2001 * XXX BUG ALERT 2002 * The "dummyifp" code relies upon the fact 2003 * that icmp_error() touches only ifp->if_mtu. 2004 */ 2005 /*XXX*/ 2006 destifp = NULL; 2007 if (sp->req != NULL 2008 && sp->req->sav != NULL 2009 && sp->req->sav->sah != NULL) { 2010 ro = &sp->req->sav->sah->sa_route; 2011 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 2012 dummyifp.if_mtu = 2013 ro->ro_rt->rt_rmx.rmx_mtu ? 2014 ro->ro_rt->rt_rmx.rmx_mtu : 2015 ro->ro_rt->rt_ifp->if_mtu; 2016 dummyifp.if_mtu -= ipsechdr; 2017 destifp = &dummyifp; 2018 } 2019 } 2020 2021#ifdef IPSEC 2022 key_freesp(sp); 2023#else 2024 KEY_FREESP(&sp); 2025#endif 2026 } 2027 } 2028#endif /*IPSEC*/ 2029 ipstat.ips_cantfrag++; 2030 break; 2031 2032 case ENOBUFS: 2033#if 1 2034 /* 2035 * a router should not generate ICMP_SOURCEQUENCH as 2036 * required in RFC1812 Requirements for IP Version 4 Routers. 2037 * source quench could be a big problem under DoS attacks, 2038 * or if the underlying interface is rate-limited. 2039 */ 2040 if (mcopy) 2041 m_freem(mcopy); 2042 return; 2043#else 2044 type = ICMP_SOURCEQUENCH; 2045 code = 0; 2046 break; 2047#endif 2048 } 2049 icmp_error(mcopy, type, code, dest, destifp); 2050} 2051 2052void 2053ip_savecontrol(inp, mp, ip, m) 2054 struct inpcb *inp; 2055 struct mbuf **mp; 2056 struct ip *ip; 2057 struct mbuf *m; 2058{ 2059 2060 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2061 struct timeval tv; 2062 2063 microtime(&tv); 2064 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2065 SCM_TIMESTAMP, SOL_SOCKET); 2066 if (*mp) 2067 mp = &(*mp)->m_next; 2068 } 2069 if (inp->inp_flags & INP_RECVDSTADDR) { 2070 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2071 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2072 if (*mp) 2073 mp = &(*mp)->m_next; 2074 } 2075#ifdef notyet 2076 /* 2077 * XXX 2078 * Moving these out of udp_input() made them even more broken 2079 * than they already were. 2080 * - fenner@parc.xerox.com 2081 */ 2082 /* options were tossed already */ 2083 if (inp->inp_flags & INP_RECVOPTS) { 2084 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2085 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2086 if (*mp) 2087 mp = &(*mp)->m_next; 2088 } 2089 /* ip_srcroute doesn't do what we want here, need to fix */ 2090 if (inp->inp_flags & INP_RECVRETOPTS) { 2091 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 2092 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2093 if (*mp) 2094 mp = &(*mp)->m_next; 2095 } 2096#endif 2097 if (inp->inp_flags & INP_RECVIF) { 2098 struct sockaddr_dl sdl; 2099 2100 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 2101 sdl.sdl_family = AF_LINK; 2102 sdl.sdl_index = m->m_pkthdr.rcvif ? 2103 m->m_pkthdr.rcvif->if_index : 0; 2104 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 2105 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 2106 IP_RECVIF, IPPROTO_IP); 2107 if (*mp) 2108 mp = &(*mp)->m_next; 2109 } 2110} 2111 2112/* 2113 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 2114 * range of the new value and tweaks timers if it changes. 2115 */ 2116static int 2117sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 2118{ 2119 int error, tmp; 2120 struct sysctlnode node; 2121 2122 node = *rnode; 2123 tmp = ip_mtudisc_timeout; 2124 node.sysctl_data = &tmp; 2125 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2126 if (error || newp == NULL) 2127 return (error); 2128 if (tmp < 0) 2129 return (EINVAL); 2130 2131 ip_mtudisc_timeout = tmp; 2132 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 2133 2134 return (0); 2135} 2136 2137#ifdef GATEWAY 2138/* 2139 * sysctl helper routine for net.inet.ip.maxflows. apparently if 2140 * maxflows is even looked up, we "reap flows". 2141 */ 2142static int 2143sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS) 2144{ 2145 int s; 2146 2147 s = sysctl_lookup(SYSCTLFN_CALL(rnode)); 2148 if (s) 2149 return (s); 2150 2151 s = splsoftnet(); 2152 ipflow_reap(0); 2153 splx(s); 2154 2155 return (0); 2156} 2157#endif /* GATEWAY */ 2158 2159 2160SYSCTL_SETUP(sysctl_net_inet_ip_setup, "sysctl net.inet.ip subtree setup") 2161{ 2162 extern int subnetsarelocal, hostzeroisbroadcast; 2163 2164 sysctl_createv(clog, 0, NULL, NULL, 2165 CTLFLAG_PERMANENT, 2166 CTLTYPE_NODE, "net", NULL, 2167 NULL, 0, NULL, 0, 2168 CTL_NET, CTL_EOL); 2169 sysctl_createv(clog, 0, NULL, NULL, 2170 CTLFLAG_PERMANENT, 2171 CTLTYPE_NODE, "inet", 2172 SYSCTL_DESCR("PF_INET related settings"), 2173 NULL, 0, NULL, 0, 2174 CTL_NET, PF_INET, CTL_EOL); 2175 sysctl_createv(clog, 0, NULL, NULL, 2176 CTLFLAG_PERMANENT, 2177 CTLTYPE_NODE, "ip", 2178 SYSCTL_DESCR("IPv4 related settings"), 2179 NULL, 0, NULL, 0, 2180 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 2181 2182 sysctl_createv(clog, 0, NULL, NULL, 2183 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2184 CTLTYPE_INT, "forwarding", 2185 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 2186 NULL, 0, &ipforwarding, 0, 2187 CTL_NET, PF_INET, IPPROTO_IP, 2188 IPCTL_FORWARDING, CTL_EOL); 2189 sysctl_createv(clog, 0, NULL, NULL, 2190 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2191 CTLTYPE_INT, "redirect", 2192 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 2193 NULL, 0, &ipsendredirects, 0, 2194 CTL_NET, PF_INET, IPPROTO_IP, 2195 IPCTL_SENDREDIRECTS, CTL_EOL); 2196 sysctl_createv(clog, 0, NULL, NULL, 2197 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2198 CTLTYPE_INT, "ttl", 2199 SYSCTL_DESCR("Default TTL for an INET datagram"), 2200 NULL, 0, &ip_defttl, 0, 2201 CTL_NET, PF_INET, IPPROTO_IP, 2202 IPCTL_DEFTTL, CTL_EOL); 2203#ifdef IPCTL_DEFMTU 2204 sysctl_createv(clog, 0, NULL, NULL, 2205 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 2206 CTLTYPE_INT, "mtu", 2207 SYSCTL_DESCR("Default MTA for an INET route"), 2208 NULL, 0, &ip_mtu, 0, 2209 CTL_NET, PF_INET, IPPROTO_IP, 2210 IPCTL_DEFMTU, CTL_EOL); 2211#endif /* IPCTL_DEFMTU */ 2212 sysctl_createv(clog, 0, NULL, NULL, 2213 CTLFLAG_PERMANENT|CTLFLAG_READONLY1, 2214 CTLTYPE_INT, "forwsrcrt", 2215 SYSCTL_DESCR("Enable forwarding of source-routed " 2216 "datagrams"), 2217 NULL, 0, &ip_forwsrcrt, 0, 2218 CTL_NET, PF_INET, IPPROTO_IP, 2219 IPCTL_FORWSRCRT, CTL_EOL); 2220 sysctl_createv(clog, 0, NULL, NULL, 2221 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2222 CTLTYPE_INT, "directed-broadcast", 2223 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 2224 NULL, 0, &ip_directedbcast, 0, 2225 CTL_NET, PF_INET, IPPROTO_IP, 2226 IPCTL_DIRECTEDBCAST, CTL_EOL); 2227 sysctl_createv(clog, 0, NULL, NULL, 2228 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2229 CTLTYPE_INT, "allowsrcrt", 2230 SYSCTL_DESCR("Accept source-routed datagrams"), 2231 NULL, 0, &ip_allowsrcrt, 0, 2232 CTL_NET, PF_INET, IPPROTO_IP, 2233 IPCTL_ALLOWSRCRT, CTL_EOL); 2234 sysctl_createv(clog, 0, NULL, NULL, 2235 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2236 CTLTYPE_INT, "subnetsarelocal", 2237 SYSCTL_DESCR("Whether logical subnets are considered " 2238 "local"), 2239 NULL, 0, &subnetsarelocal, 0, 2240 CTL_NET, PF_INET, IPPROTO_IP, 2241 IPCTL_SUBNETSARELOCAL, CTL_EOL); 2242 sysctl_createv(clog, 0, NULL, NULL, 2243 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2244 CTLTYPE_INT, "mtudisc", 2245 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 2246 NULL, 0, &ip_mtudisc, 0, 2247 CTL_NET, PF_INET, IPPROTO_IP, 2248 IPCTL_MTUDISC, CTL_EOL); 2249 sysctl_createv(clog, 0, NULL, NULL, 2250 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2251 CTLTYPE_INT, "anonportmin", 2252 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 2253 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 2254 CTL_NET, PF_INET, IPPROTO_IP, 2255 IPCTL_ANONPORTMIN, CTL_EOL); 2256 sysctl_createv(clog, 0, NULL, NULL, 2257 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2258 CTLTYPE_INT, "anonportmax", 2259 SYSCTL_DESCR("Highest ephemeral port number to assign"), 2260 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 2261 CTL_NET, PF_INET, IPPROTO_IP, 2262 IPCTL_ANONPORTMAX, CTL_EOL); 2263 sysctl_createv(clog, 0, NULL, NULL, 2264 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2265 CTLTYPE_INT, "mtudisctimeout", 2266 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 2267 sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0, 2268 CTL_NET, PF_INET, IPPROTO_IP, 2269 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 2270#ifdef GATEWAY 2271 sysctl_createv(clog, 0, NULL, NULL, 2272 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2273 CTLTYPE_INT, "maxflows", 2274 SYSCTL_DESCR("Number of flows for fast forwarding"), 2275 sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0, 2276 CTL_NET, PF_INET, IPPROTO_IP, 2277 IPCTL_MAXFLOWS, CTL_EOL); 2278#endif /* GATEWAY */ 2279 sysctl_createv(clog, 0, NULL, NULL, 2280 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2281 CTLTYPE_INT, "hostzerobroadcast", 2282 SYSCTL_DESCR("All zeroes address is broadcast address"), 2283 NULL, 0, &hostzeroisbroadcast, 0, 2284 CTL_NET, PF_INET, IPPROTO_IP, 2285 IPCTL_HOSTZEROBROADCAST, CTL_EOL); 2286#if NGIF > 0 2287 sysctl_createv(clog, 0, NULL, NULL, 2288 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2289 CTLTYPE_INT, "gifttl", 2290 SYSCTL_DESCR("Default TTL for a gif tunnel datagram"), 2291 NULL, 0, &ip_gif_ttl, 0, 2292 CTL_NET, PF_INET, IPPROTO_IP, 2293 IPCTL_GIF_TTL, CTL_EOL); 2294#endif /* NGIF */ 2295#ifndef IPNOPRIVPORTS 2296 sysctl_createv(clog, 0, NULL, NULL, 2297 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2298 CTLTYPE_INT, "lowportmin", 2299 SYSCTL_DESCR("Lowest privileged ephemeral port number " 2300 "to assign"), 2301 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 2302 CTL_NET, PF_INET, IPPROTO_IP, 2303 IPCTL_LOWPORTMIN, CTL_EOL); 2304 sysctl_createv(clog, 0, NULL, NULL, 2305 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2306 CTLTYPE_INT, "lowportmax", 2307 SYSCTL_DESCR("Highest privileged ephemeral port number " 2308 "to assign"), 2309 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 2310 CTL_NET, PF_INET, IPPROTO_IP, 2311 IPCTL_LOWPORTMAX, CTL_EOL); 2312#endif /* IPNOPRIVPORTS */ 2313 sysctl_createv(clog, 0, NULL, NULL, 2314 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2315 CTLTYPE_INT, "maxfragpackets", 2316 SYSCTL_DESCR("Maximum number of fragments to retain for " 2317 "possible reassembly"), 2318 NULL, 0, &ip_maxfragpackets, 0, 2319 CTL_NET, PF_INET, IPPROTO_IP, 2320 IPCTL_MAXFRAGPACKETS, CTL_EOL); 2321#if NGRE > 0 2322 sysctl_createv(clog, 0, NULL, NULL, 2323 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2324 CTLTYPE_INT, "grettl", 2325 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 2326 NULL, 0, &ip_gre_ttl, 0, 2327 CTL_NET, PF_INET, IPPROTO_IP, 2328 IPCTL_GRE_TTL, CTL_EOL); 2329#endif /* NGRE */ 2330 sysctl_createv(clog, 0, NULL, NULL, 2331 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2332 CTLTYPE_INT, "checkinterface", 2333 SYSCTL_DESCR("Enable receive side of Strong ES model " 2334 "from RFC1122"), 2335 NULL, 0, &ip_checkinterface, 0, 2336 CTL_NET, PF_INET, IPPROTO_IP, 2337 IPCTL_CHECKINTERFACE, CTL_EOL); 2338 sysctl_createv(clog, 0, NULL, NULL, 2339 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2340 CTLTYPE_INT, "random_id", 2341 SYSCTL_DESCR("Assign random ip_id values"), 2342 NULL, 0, &ip_do_randomid, 0, 2343 CTL_NET, PF_INET, IPPROTO_IP, 2344 IPCTL_RANDOMID, CTL_EOL); 2345} 2346