ip_input.c revision 1.248
1/* $NetBSD: ip_input.c,v 1.248 2007/03/25 20:12:20 liamjfoy 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 1.248 2007/03/25 20:12:20 liamjfoy Exp $"); 102 103#include "opt_inet.h" 104#include "opt_gateway.h" 105#include "opt_pfil_hooks.h" 106#include "opt_ipsec.h" 107#include "opt_mrouting.h" 108#include "opt_mbuftrace.h" 109#include "opt_inet_csum.h" 110 111#include <sys/param.h> 112#include <sys/systm.h> 113#include <sys/malloc.h> 114#include <sys/mbuf.h> 115#include <sys/domain.h> 116#include <sys/protosw.h> 117#include <sys/socket.h> 118#include <sys/socketvar.h> 119#include <sys/errno.h> 120#include <sys/time.h> 121#include <sys/kernel.h> 122#include <sys/pool.h> 123#include <sys/sysctl.h> 124#include <sys/kauth.h> 125 126#include <net/if.h> 127#include <net/if_dl.h> 128#include <net/route.h> 129#include <net/pfil.h> 130 131#include <netinet/in.h> 132#include <netinet/in_systm.h> 133#include <netinet/ip.h> 134#include <netinet/in_pcb.h> 135#include <netinet/in_proto.h> 136#include <netinet/in_var.h> 137#include <netinet/ip_var.h> 138#include <netinet/ip_icmp.h> 139/* just for gif_ttl */ 140#include <netinet/in_gif.h> 141#include "gif.h" 142#include <net/if_gre.h> 143#include "gre.h" 144 145#ifdef MROUTING 146#include <netinet/ip_mroute.h> 147#endif 148 149#ifdef IPSEC 150#include <netinet6/ipsec.h> 151#include <netkey/key.h> 152#endif 153#ifdef FAST_IPSEC 154#include <netipsec/ipsec.h> 155#include <netipsec/key.h> 156#endif /* FAST_IPSEC*/ 157 158#ifndef IPFORWARDING 159#ifdef GATEWAY 160#define IPFORWARDING 1 /* forward IP packets not for us */ 161#else /* GATEWAY */ 162#define IPFORWARDING 0 /* don't forward IP packets not for us */ 163#endif /* GATEWAY */ 164#endif /* IPFORWARDING */ 165#ifndef IPSENDREDIRECTS 166#define IPSENDREDIRECTS 1 167#endif 168#ifndef IPFORWSRCRT 169#define IPFORWSRCRT 1 /* forward source-routed packets */ 170#endif 171#ifndef IPALLOWSRCRT 172#define IPALLOWSRCRT 1 /* allow source-routed packets */ 173#endif 174#ifndef IPMTUDISC 175#define IPMTUDISC 1 176#endif 177#ifndef IPMTUDISCTIMEOUT 178#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 179#endif 180 181/* 182 * Note: DIRECTED_BROADCAST is handled this way so that previous 183 * configuration using this option will Just Work. 184 */ 185#ifndef IPDIRECTEDBCAST 186#ifdef DIRECTED_BROADCAST 187#define IPDIRECTEDBCAST 1 188#else 189#define IPDIRECTEDBCAST 0 190#endif /* DIRECTED_BROADCAST */ 191#endif /* IPDIRECTEDBCAST */ 192int ipforwarding = IPFORWARDING; 193int ipsendredirects = IPSENDREDIRECTS; 194int ip_defttl = IPDEFTTL; 195int ip_forwsrcrt = IPFORWSRCRT; 196int ip_directedbcast = IPDIRECTEDBCAST; 197int ip_allowsrcrt = IPALLOWSRCRT; 198int ip_mtudisc = IPMTUDISC; 199int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 200#ifdef DIAGNOSTIC 201int ipprintfs = 0; 202#endif 203 204int ip_do_randomid = 0; 205 206/* 207 * XXX - Setting ip_checkinterface mostly implements the receive side of 208 * the Strong ES model described in RFC 1122, but since the routing table 209 * and transmit implementation do not implement the Strong ES model, 210 * setting this to 1 results in an odd hybrid. 211 * 212 * XXX - ip_checkinterface currently must be disabled if you use ipnat 213 * to translate the destination address to another local interface. 214 * 215 * XXX - ip_checkinterface must be disabled if you add IP aliases 216 * to the loopback interface instead of the interface where the 217 * packets for those addresses are received. 218 */ 219int ip_checkinterface = 0; 220 221 222struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 223 224int ipqmaxlen = IFQ_MAXLEN; 225u_long in_ifaddrhash; /* size of hash table - 1 */ 226int in_ifaddrentries; /* total number of addrs */ 227struct in_ifaddrhead in_ifaddrhead; 228struct in_ifaddrhashhead *in_ifaddrhashtbl; 229u_long in_multihash; /* size of hash table - 1 */ 230int in_multientries; /* total number of addrs */ 231struct in_multihashhead *in_multihashtbl; 232struct ifqueue ipintrq; 233struct ipstat ipstat; 234uint16_t ip_id; 235 236#ifdef PFIL_HOOKS 237struct pfil_head inet_pfil_hook; 238#endif 239 240/* 241 * Cached copy of nmbclusters. If nbclusters is different, 242 * recalculate IP parameters derived from nmbclusters. 243 */ 244static int ip_nmbclusters; /* copy of nmbclusters */ 245static void ip_nmbclusters_changed(void); /* recalc limits */ 246 247#define CHECK_NMBCLUSTER_PARAMS() \ 248do { \ 249 if (__predict_false(ip_nmbclusters != nmbclusters)) \ 250 ip_nmbclusters_changed(); \ 251} while (/*CONSTCOND*/0) 252 253/* IP datagram reassembly queues (hashed) */ 254#define IPREASS_NHASH_LOG2 6 255#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 256#define IPREASS_HMASK (IPREASS_NHASH - 1) 257#define IPREASS_HASH(x,y) \ 258 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 259struct ipqhead ipq[IPREASS_NHASH]; 260int ipq_locked; 261static int ip_nfragpackets; /* packets in reass queue */ 262static int ip_nfrags; /* total fragments in reass queues */ 263 264int ip_maxfragpackets = 200; /* limit on packets. XXX sysctl */ 265int ip_maxfrags; /* limit on fragments. XXX sysctl */ 266 267 268/* 269 * Additive-Increase/Multiplicative-Decrease (AIMD) strategy for 270 * IP reassembly queue buffer managment. 271 * 272 * We keep a count of total IP fragments (NB: not fragmented packets!) 273 * awaiting reassembly (ip_nfrags) and a limit (ip_maxfrags) on fragments. 274 * If ip_nfrags exceeds ip_maxfrags the limit, we drop half the 275 * total fragments in reassembly queues.This AIMD policy avoids 276 * repeatedly deleting single packets under heavy fragmentation load 277 * (e.g., from lossy NFS peers). 278 */ 279static u_int ip_reass_ttl_decr(u_int ticks); 280static void ip_reass_drophalf(void); 281 282 283static inline int ipq_lock_try(void); 284static inline void ipq_unlock(void); 285 286static inline int 287ipq_lock_try(void) 288{ 289 int s; 290 291 /* 292 * Use splvm() -- we're blocking things that would cause 293 * mbuf allocation. 294 */ 295 s = splvm(); 296 if (ipq_locked) { 297 splx(s); 298 return (0); 299 } 300 ipq_locked = 1; 301 splx(s); 302 return (1); 303} 304 305static inline void 306ipq_unlock(void) 307{ 308 int s; 309 310 s = splvm(); 311 ipq_locked = 0; 312 splx(s); 313} 314 315#ifdef DIAGNOSTIC 316#define IPQ_LOCK() \ 317do { \ 318 if (ipq_lock_try() == 0) { \ 319 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 320 panic("ipq_lock"); \ 321 } \ 322} while (/*CONSTCOND*/ 0) 323#define IPQ_LOCK_CHECK() \ 324do { \ 325 if (ipq_locked == 0) { \ 326 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 327 panic("ipq lock check"); \ 328 } \ 329} while (/*CONSTCOND*/ 0) 330#else 331#define IPQ_LOCK() (void) ipq_lock_try() 332#define IPQ_LOCK_CHECK() /* nothing */ 333#endif 334 335#define IPQ_UNLOCK() ipq_unlock() 336 337POOL_INIT(inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl", NULL, 338 IPL_SOFTNET); 339POOL_INIT(ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", NULL, 340 IPL_VM); 341 342#ifdef INET_CSUM_COUNTERS 343#include <sys/device.h> 344 345struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 346 NULL, "inet", "hwcsum bad"); 347struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 348 NULL, "inet", "hwcsum ok"); 349struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 350 NULL, "inet", "swcsum"); 351 352#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 353 354EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 355EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 356EVCNT_ATTACH_STATIC(ip_swcsum); 357 358#else 359 360#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 361 362#endif /* INET_CSUM_COUNTERS */ 363 364/* 365 * We need to save the IP options in case a protocol wants to respond 366 * to an incoming packet over the same route if the packet got here 367 * using IP source routing. This allows connection establishment and 368 * maintenance when the remote end is on a network that is not known 369 * to us. 370 */ 371int ip_nhops = 0; 372static struct ip_srcrt { 373 struct in_addr dst; /* final destination */ 374 char nop; /* one NOP to align */ 375 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 376 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 377} ip_srcrt; 378 379static void save_rte(u_char *, struct in_addr); 380 381#ifdef MBUFTRACE 382struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx"); 383struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx"); 384#endif 385 386/* 387 * Compute IP limits derived from the value of nmbclusters. 388 */ 389static void 390ip_nmbclusters_changed(void) 391{ 392 ip_maxfrags = nmbclusters / 4; 393 ip_nmbclusters = nmbclusters; 394} 395 396/* 397 * IP initialization: fill in IP protocol switch table. 398 * All protocols not implemented in kernel go to raw IP protocol handler. 399 */ 400void 401ip_init(void) 402{ 403 const struct protosw *pr; 404 int i; 405 406 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 407 if (pr == 0) 408 panic("ip_init"); 409 for (i = 0; i < IPPROTO_MAX; i++) 410 ip_protox[i] = pr - inetsw; 411 for (pr = inetdomain.dom_protosw; 412 pr < inetdomain.dom_protoswNPROTOSW; pr++) 413 if (pr->pr_domain->dom_family == PF_INET && 414 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 415 ip_protox[pr->pr_protocol] = pr - inetsw; 416 417 for (i = 0; i < IPREASS_NHASH; i++) 418 LIST_INIT(&ipq[i]); 419 420 ip_id = time_second & 0xfffff; 421 422 ipintrq.ifq_maxlen = ipqmaxlen; 423 ip_nmbclusters_changed(); 424 425 TAILQ_INIT(&in_ifaddrhead); 426 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 427 M_WAITOK, &in_ifaddrhash); 428 in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IPMADDR, 429 M_WAITOK, &in_multihash); 430 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 431#ifdef GATEWAY 432 ipflow_init(ip_hashsize); 433#endif 434 435#ifdef PFIL_HOOKS 436 /* Register our Packet Filter hook. */ 437 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 438 inet_pfil_hook.ph_af = AF_INET; 439 i = pfil_head_register(&inet_pfil_hook); 440 if (i != 0) 441 printf("ip_init: WARNING: unable to register pfil hook, " 442 "error %d\n", i); 443#endif /* PFIL_HOOKS */ 444 445#ifdef MBUFTRACE 446 MOWNER_ATTACH(&ip_tx_mowner); 447 MOWNER_ATTACH(&ip_rx_mowner); 448#endif /* MBUFTRACE */ 449} 450 451struct sockaddr_in ipaddr = { 452 .sin_len = sizeof(ipaddr), 453 .sin_family = AF_INET, 454}; 455struct route ipforward_rt; 456 457/* 458 * IP software interrupt routine 459 */ 460void 461ipintr(void) 462{ 463 int s; 464 struct mbuf *m; 465 466 while (!IF_IS_EMPTY(&ipintrq)) { 467 s = splnet(); 468 IF_DEQUEUE(&ipintrq, m); 469 splx(s); 470 if (m == 0) 471 return; 472 MCLAIM(m, &ip_rx_mowner); 473 ip_input(m); 474 } 475} 476 477/* 478 * Ip input routine. Checksum and byte swap header. If fragmented 479 * try to reassemble. Process options. Pass to next level. 480 */ 481void 482ip_input(struct mbuf *m) 483{ 484 struct ip *ip = NULL; 485 struct ipq *fp; 486 struct in_ifaddr *ia; 487 struct ifaddr *ifa; 488 struct ipqent *ipqe; 489 int hlen = 0, mff, len; 490 int downmatch; 491 int checkif; 492 int srcrt = 0; 493 int s; 494 u_int hash; 495#ifdef FAST_IPSEC 496 struct m_tag *mtag; 497 struct tdb_ident *tdbi; 498 struct secpolicy *sp; 499 int error; 500#endif /* FAST_IPSEC */ 501 502 MCLAIM(m, &ip_rx_mowner); 503#ifdef DIAGNOSTIC 504 if ((m->m_flags & M_PKTHDR) == 0) 505 panic("ipintr no HDR"); 506#endif 507 508 /* 509 * If no IP addresses have been set yet but the interfaces 510 * are receiving, can't do anything with incoming packets yet. 511 */ 512 if (TAILQ_FIRST(&in_ifaddrhead) == 0) 513 goto bad; 514 ipstat.ips_total++; 515 /* 516 * If the IP header is not aligned, slurp it up into a new 517 * mbuf with space for link headers, in the event we forward 518 * it. Otherwise, if it is aligned, make sure the entire 519 * base IP header is in the first mbuf of the chain. 520 */ 521 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 522 if ((m = m_copyup(m, sizeof(struct ip), 523 (max_linkhdr + 3) & ~3)) == NULL) { 524 /* XXXJRT new stat, please */ 525 ipstat.ips_toosmall++; 526 return; 527 } 528 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 529 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 530 ipstat.ips_toosmall++; 531 return; 532 } 533 } 534 ip = mtod(m, struct ip *); 535 if (ip->ip_v != IPVERSION) { 536 ipstat.ips_badvers++; 537 goto bad; 538 } 539 hlen = ip->ip_hl << 2; 540 if (hlen < sizeof(struct ip)) { /* minimum header length */ 541 ipstat.ips_badhlen++; 542 goto bad; 543 } 544 if (hlen > m->m_len) { 545 if ((m = m_pullup(m, hlen)) == 0) { 546 ipstat.ips_badhlen++; 547 return; 548 } 549 ip = mtod(m, struct ip *); 550 } 551 552 /* 553 * RFC1122: packets with a multicast source address are 554 * not allowed. 555 */ 556 if (IN_MULTICAST(ip->ip_src.s_addr)) { 557 ipstat.ips_badaddr++; 558 goto bad; 559 } 560 561 /* 127/8 must not appear on wire - RFC1122 */ 562 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 563 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 564 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 565 ipstat.ips_badaddr++; 566 goto bad; 567 } 568 } 569 570 switch (m->m_pkthdr.csum_flags & 571 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) | 572 M_CSUM_IPv4_BAD)) { 573 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 574 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 575 goto badcsum; 576 577 case M_CSUM_IPv4: 578 /* Checksum was okay. */ 579 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 580 break; 581 582 default: 583 /* 584 * Must compute it ourselves. Maybe skip checksum on 585 * loopback interfaces. 586 */ 587 if (__predict_true(!(m->m_pkthdr.rcvif->if_flags & 588 IFF_LOOPBACK) || ip_do_loopback_cksum)) { 589 INET_CSUM_COUNTER_INCR(&ip_swcsum); 590 if (in_cksum(m, hlen) != 0) 591 goto badcsum; 592 } 593 break; 594 } 595 596 /* Retrieve the packet length. */ 597 len = ntohs(ip->ip_len); 598 599 /* 600 * Check for additional length bogosity 601 */ 602 if (len < hlen) { 603 ipstat.ips_badlen++; 604 goto bad; 605 } 606 607 /* 608 * Check that the amount of data in the buffers 609 * is as at least much as the IP header would have us expect. 610 * Trim mbufs if longer than we expect. 611 * Drop packet if shorter than we expect. 612 */ 613 if (m->m_pkthdr.len < len) { 614 ipstat.ips_tooshort++; 615 goto bad; 616 } 617 if (m->m_pkthdr.len > len) { 618 if (m->m_len == m->m_pkthdr.len) { 619 m->m_len = len; 620 m->m_pkthdr.len = len; 621 } else 622 m_adj(m, len - m->m_pkthdr.len); 623 } 624 625#if defined(IPSEC) 626 /* ipflow (IP fast forwarding) is not compatible with IPsec. */ 627 m->m_flags &= ~M_CANFASTFWD; 628#else 629 /* 630 * Assume that we can create a fast-forward IP flow entry 631 * based on this packet. 632 */ 633 m->m_flags |= M_CANFASTFWD; 634#endif 635 636#ifdef PFIL_HOOKS 637 /* 638 * Run through list of hooks for input packets. If there are any 639 * filters which require that additional packets in the flow are 640 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 641 * Note that filters must _never_ set this flag, as another filter 642 * in the list may have previously cleared it. 643 */ 644 /* 645 * let ipfilter look at packet on the wire, 646 * not the decapsulated packet. 647 */ 648#ifdef IPSEC 649 if (!ipsec_getnhist(m)) 650#elif defined(FAST_IPSEC) 651 if (!ipsec_indone(m)) 652#else 653 if (1) 654#endif 655 { 656 struct in_addr odst; 657 658 odst = ip->ip_dst; 659 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 660 PFIL_IN) != 0) 661 return; 662 if (m == NULL) 663 return; 664 ip = mtod(m, struct ip *); 665 hlen = ip->ip_hl << 2; 666 /* 667 * XXX The setting of "srcrt" here is to prevent ip_forward() 668 * from generating ICMP redirects for packets that have 669 * been redirected by a hook back out on to the same LAN that 670 * they came from and is not an indication that the packet 671 * is being inffluenced by source routing options. This 672 * allows things like 673 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 674 * where tlp0 is both on the 1.1.1.0/24 network and is the 675 * default route for hosts on 1.1.1.0/24. Of course this 676 * also requires a "map tlp0 ..." to complete the story. 677 * One might argue whether or not this kind of network config. 678 * should be supported in this manner... 679 */ 680 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 681 } 682#endif /* PFIL_HOOKS */ 683 684#ifdef ALTQ 685 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 686 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 687 /* packet dropped by traffic conditioner */ 688 return; 689 } 690#endif 691 692 /* 693 * Process options and, if not destined for us, 694 * ship it on. ip_dooptions returns 1 when an 695 * error was detected (causing an icmp message 696 * to be sent and the original packet to be freed). 697 */ 698 ip_nhops = 0; /* for source routed packets */ 699 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 700 return; 701 702 /* 703 * Enable a consistency check between the destination address 704 * and the arrival interface for a unicast packet (the RFC 1122 705 * strong ES model) if IP forwarding is disabled and the packet 706 * is not locally generated. 707 * 708 * XXX - Checking also should be disabled if the destination 709 * address is ipnat'ed to a different interface. 710 * 711 * XXX - Checking is incompatible with IP aliases added 712 * to the loopback interface instead of the interface where 713 * the packets are received. 714 * 715 * XXX - We need to add a per ifaddr flag for this so that 716 * we get finer grain control. 717 */ 718 checkif = ip_checkinterface && (ipforwarding == 0) && 719 (m->m_pkthdr.rcvif != NULL) && 720 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0); 721 722 /* 723 * Check our list of addresses, to see if the packet is for us. 724 * 725 * Traditional 4.4BSD did not consult IFF_UP at all. 726 * The behavior here is to treat addresses on !IFF_UP interface 727 * as not mine. 728 */ 729 downmatch = 0; 730 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 731 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 732 if (checkif && ia->ia_ifp != m->m_pkthdr.rcvif) 733 continue; 734 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 735 break; 736 else 737 downmatch++; 738 } 739 } 740 if (ia != NULL) 741 goto ours; 742 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 743 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) { 744 if (ifa->ifa_addr->sa_family != AF_INET) 745 continue; 746 ia = ifatoia(ifa); 747 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 748 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 749 /* 750 * Look for all-0's host part (old broadcast addr), 751 * either for subnet or net. 752 */ 753 ip->ip_dst.s_addr == ia->ia_subnet || 754 ip->ip_dst.s_addr == ia->ia_net) 755 goto ours; 756 /* 757 * An interface with IP address zero accepts 758 * all packets that arrive on that interface. 759 */ 760 if (in_nullhost(ia->ia_addr.sin_addr)) 761 goto ours; 762 } 763 } 764 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 765 struct in_multi *inm; 766#ifdef MROUTING 767 extern struct socket *ip_mrouter; 768 769 if (ip_mrouter) { 770 /* 771 * If we are acting as a multicast router, all 772 * incoming multicast packets are passed to the 773 * kernel-level multicast forwarding function. 774 * The packet is returned (relatively) intact; if 775 * ip_mforward() returns a non-zero value, the packet 776 * must be discarded, else it may be accepted below. 777 * 778 * (The IP ident field is put in the same byte order 779 * as expected when ip_mforward() is called from 780 * ip_output().) 781 */ 782 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 783 ipstat.ips_cantforward++; 784 m_freem(m); 785 return; 786 } 787 788 /* 789 * The process-level routing demon needs to receive 790 * all multicast IGMP packets, whether or not this 791 * host belongs to their destination groups. 792 */ 793 if (ip->ip_p == IPPROTO_IGMP) 794 goto ours; 795 ipstat.ips_forward++; 796 } 797#endif 798 /* 799 * See if we belong to the destination multicast group on the 800 * arrival interface. 801 */ 802 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 803 if (inm == NULL) { 804 ipstat.ips_cantforward++; 805 m_freem(m); 806 return; 807 } 808 goto ours; 809 } 810 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 811 in_nullhost(ip->ip_dst)) 812 goto ours; 813 814 /* 815 * Not for us; forward if possible and desirable. 816 */ 817 if (ipforwarding == 0) { 818 ipstat.ips_cantforward++; 819 m_freem(m); 820 } else { 821 /* 822 * If ip_dst matched any of my address on !IFF_UP interface, 823 * and there's no IFF_UP interface that matches ip_dst, 824 * send icmp unreach. Forwarding it will result in in-kernel 825 * forwarding loop till TTL goes to 0. 826 */ 827 if (downmatch) { 828 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 829 ipstat.ips_cantforward++; 830 return; 831 } 832#ifdef IPSEC 833 if (ipsec4_in_reject(m, NULL)) { 834 ipsecstat.in_polvio++; 835 goto bad; 836 } 837#endif 838#ifdef FAST_IPSEC 839 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 840 s = splsoftnet(); 841 if (mtag != NULL) { 842 tdbi = (struct tdb_ident *)(mtag + 1); 843 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 844 } else { 845 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 846 IP_FORWARDING, &error); 847 } 848 if (sp == NULL) { /* NB: can happen if error */ 849 splx(s); 850 /*XXX error stat???*/ 851 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 852 goto bad; 853 } 854 855 /* 856 * Check security policy against packet attributes. 857 */ 858 error = ipsec_in_reject(sp, m); 859 KEY_FREESP(&sp); 860 splx(s); 861 if (error) { 862 ipstat.ips_cantforward++; 863 goto bad; 864 } 865 866 /* 867 * Peek at the outbound SP for this packet to determine if 868 * it's a Fast Forward candidate. 869 */ 870 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 871 if (mtag != NULL) 872 m->m_flags &= ~M_CANFASTFWD; 873 else { 874 s = splsoftnet(); 875 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, 876 (IP_FORWARDING | 877 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 878 &error, NULL); 879 if (sp != NULL) { 880 m->m_flags &= ~M_CANFASTFWD; 881 KEY_FREESP(&sp); 882 } 883 splx(s); 884 } 885#endif /* FAST_IPSEC */ 886 887 ip_forward(m, srcrt); 888 } 889 return; 890 891ours: 892 /* 893 * If offset or IP_MF are set, must reassemble. 894 * Otherwise, nothing need be done. 895 * (We could look in the reassembly queue to see 896 * if the packet was previously fragmented, 897 * but it's not worth the time; just let them time out.) 898 */ 899 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 900 901 /* 902 * Look for queue of fragments 903 * of this datagram. 904 */ 905 IPQ_LOCK(); 906 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 907 /* XXX LIST_FOREACH(fp, &ipq[hash], ipq_q) */ 908 for (fp = LIST_FIRST(&ipq[hash]); fp != NULL; 909 fp = LIST_NEXT(fp, ipq_q)) { 910 if (ip->ip_id == fp->ipq_id && 911 in_hosteq(ip->ip_src, fp->ipq_src) && 912 in_hosteq(ip->ip_dst, fp->ipq_dst) && 913 ip->ip_p == fp->ipq_p) 914 goto found; 915 916 } 917 fp = 0; 918found: 919 920 /* 921 * Adjust ip_len to not reflect header, 922 * set ipqe_mff if more fragments are expected, 923 * convert offset of this to bytes. 924 */ 925 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 926 mff = (ip->ip_off & htons(IP_MF)) != 0; 927 if (mff) { 928 /* 929 * Make sure that fragments have a data length 930 * that's a non-zero multiple of 8 bytes. 931 */ 932 if (ntohs(ip->ip_len) == 0 || 933 (ntohs(ip->ip_len) & 0x7) != 0) { 934 ipstat.ips_badfrags++; 935 IPQ_UNLOCK(); 936 goto bad; 937 } 938 } 939 ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3); 940 941 /* 942 * If datagram marked as having more fragments 943 * or if this is not the first fragment, 944 * attempt reassembly; if it succeeds, proceed. 945 */ 946 if (mff || ip->ip_off != htons(0)) { 947 ipstat.ips_fragments++; 948 s = splvm(); 949 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 950 splx(s); 951 if (ipqe == NULL) { 952 ipstat.ips_rcvmemdrop++; 953 IPQ_UNLOCK(); 954 goto bad; 955 } 956 ipqe->ipqe_mff = mff; 957 ipqe->ipqe_m = m; 958 ipqe->ipqe_ip = ip; 959 m = ip_reass(ipqe, fp, &ipq[hash]); 960 if (m == 0) { 961 IPQ_UNLOCK(); 962 return; 963 } 964 ipstat.ips_reassembled++; 965 ip = mtod(m, struct ip *); 966 hlen = ip->ip_hl << 2; 967 ip->ip_len = htons(ntohs(ip->ip_len) + hlen); 968 } else 969 if (fp) 970 ip_freef(fp); 971 IPQ_UNLOCK(); 972 } 973 974#if defined(IPSEC) 975 /* 976 * enforce IPsec policy checking if we are seeing last header. 977 * note that we do not visit this with protocols with pcb layer 978 * code - like udp/tcp/raw ip. 979 */ 980 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 981 ipsec4_in_reject(m, NULL)) { 982 ipsecstat.in_polvio++; 983 goto bad; 984 } 985#endif 986#ifdef FAST_IPSEC 987 /* 988 * enforce IPsec policy checking if we are seeing last header. 989 * note that we do not visit this with protocols with pcb layer 990 * code - like udp/tcp/raw ip. 991 */ 992 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 993 /* 994 * Check if the packet has already had IPsec processing 995 * done. If so, then just pass it along. This tag gets 996 * set during AH, ESP, etc. input handling, before the 997 * packet is returned to the ip input queue for delivery. 998 */ 999 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 1000 s = splsoftnet(); 1001 if (mtag != NULL) { 1002 tdbi = (struct tdb_ident *)(mtag + 1); 1003 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 1004 } else { 1005 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 1006 IP_FORWARDING, &error); 1007 } 1008 if (sp != NULL) { 1009 /* 1010 * Check security policy against packet attributes. 1011 */ 1012 error = ipsec_in_reject(sp, m); 1013 KEY_FREESP(&sp); 1014 } else { 1015 /* XXX error stat??? */ 1016 error = EINVAL; 1017DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 1018 goto bad; 1019 } 1020 splx(s); 1021 if (error) 1022 goto bad; 1023 } 1024#endif /* FAST_IPSEC */ 1025 1026 /* 1027 * Switch out to protocol's input routine. 1028 */ 1029#if IFA_STATS 1030 if (ia && ip) 1031 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 1032#endif 1033 ipstat.ips_delivered++; 1034 { 1035 int off = hlen, nh = ip->ip_p; 1036 1037 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 1038 return; 1039 } 1040bad: 1041 m_freem(m); 1042 return; 1043 1044badcsum: 1045 ipstat.ips_badsum++; 1046 m_freem(m); 1047} 1048 1049/* 1050 * Take incoming datagram fragment and try to 1051 * reassemble it into whole datagram. If a chain for 1052 * reassembly of this datagram already exists, then it 1053 * is given as fp; otherwise have to make a chain. 1054 */ 1055struct mbuf * 1056ip_reass(struct ipqent *ipqe, struct ipq *fp, struct ipqhead *ipqhead) 1057{ 1058 struct mbuf *m = ipqe->ipqe_m; 1059 struct ipqent *nq, *p, *q; 1060 struct ip *ip; 1061 struct mbuf *t; 1062 int hlen = ipqe->ipqe_ip->ip_hl << 2; 1063 int i, next, s; 1064 1065 IPQ_LOCK_CHECK(); 1066 1067 /* 1068 * Presence of header sizes in mbufs 1069 * would confuse code below. 1070 */ 1071 m->m_data += hlen; 1072 m->m_len -= hlen; 1073 1074#ifdef notyet 1075 /* make sure fragment limit is up-to-date */ 1076 CHECK_NMBCLUSTER_PARAMS(); 1077 1078 /* If we have too many fragments, drop the older half. */ 1079 if (ip_nfrags >= ip_maxfrags) 1080 ip_reass_drophalf(void); 1081#endif 1082 1083 /* 1084 * We are about to add a fragment; increment frag count. 1085 */ 1086 ip_nfrags++; 1087 1088 /* 1089 * If first fragment to arrive, create a reassembly queue. 1090 */ 1091 if (fp == 0) { 1092 /* 1093 * Enforce upper bound on number of fragmented packets 1094 * for which we attempt reassembly; 1095 * If maxfrag is 0, never accept fragments. 1096 * If maxfrag is -1, accept all fragments without limitation. 1097 */ 1098 if (ip_maxfragpackets < 0) 1099 ; 1100 else if (ip_nfragpackets >= ip_maxfragpackets) 1101 goto dropfrag; 1102 ip_nfragpackets++; 1103 MALLOC(fp, struct ipq *, sizeof (struct ipq), 1104 M_FTABLE, M_NOWAIT); 1105 if (fp == NULL) 1106 goto dropfrag; 1107 LIST_INSERT_HEAD(ipqhead, fp, ipq_q); 1108 fp->ipq_nfrags = 1; 1109 fp->ipq_ttl = IPFRAGTTL; 1110 fp->ipq_p = ipqe->ipqe_ip->ip_p; 1111 fp->ipq_id = ipqe->ipqe_ip->ip_id; 1112 TAILQ_INIT(&fp->ipq_fragq); 1113 fp->ipq_src = ipqe->ipqe_ip->ip_src; 1114 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 1115 p = NULL; 1116 goto insert; 1117 } else { 1118 fp->ipq_nfrags++; 1119 } 1120 1121 /* 1122 * Find a segment which begins after this one does. 1123 */ 1124 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1125 p = q, q = TAILQ_NEXT(q, ipqe_q)) 1126 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off)) 1127 break; 1128 1129 /* 1130 * If there is a preceding segment, it may provide some of 1131 * our data already. If so, drop the data from the incoming 1132 * segment. If it provides all of our data, drop us. 1133 */ 1134 if (p != NULL) { 1135 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) - 1136 ntohs(ipqe->ipqe_ip->ip_off); 1137 if (i > 0) { 1138 if (i >= ntohs(ipqe->ipqe_ip->ip_len)) 1139 goto dropfrag; 1140 m_adj(ipqe->ipqe_m, i); 1141 ipqe->ipqe_ip->ip_off = 1142 htons(ntohs(ipqe->ipqe_ip->ip_off) + i); 1143 ipqe->ipqe_ip->ip_len = 1144 htons(ntohs(ipqe->ipqe_ip->ip_len) - i); 1145 } 1146 } 1147 1148 /* 1149 * While we overlap succeeding segments trim them or, 1150 * if they are completely covered, dequeue them. 1151 */ 1152 for (; q != NULL && 1153 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) > 1154 ntohs(q->ipqe_ip->ip_off); q = nq) { 1155 i = (ntohs(ipqe->ipqe_ip->ip_off) + 1156 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off); 1157 if (i < ntohs(q->ipqe_ip->ip_len)) { 1158 q->ipqe_ip->ip_len = 1159 htons(ntohs(q->ipqe_ip->ip_len) - i); 1160 q->ipqe_ip->ip_off = 1161 htons(ntohs(q->ipqe_ip->ip_off) + i); 1162 m_adj(q->ipqe_m, i); 1163 break; 1164 } 1165 nq = TAILQ_NEXT(q, ipqe_q); 1166 m_freem(q->ipqe_m); 1167 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1168 s = splvm(); 1169 pool_put(&ipqent_pool, q); 1170 splx(s); 1171 fp->ipq_nfrags--; 1172 ip_nfrags--; 1173 } 1174 1175insert: 1176 /* 1177 * Stick new segment in its place; 1178 * check for complete reassembly. 1179 */ 1180 if (p == NULL) { 1181 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 1182 } else { 1183 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q); 1184 } 1185 next = 0; 1186 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1187 p = q, q = TAILQ_NEXT(q, ipqe_q)) { 1188 if (ntohs(q->ipqe_ip->ip_off) != next) 1189 return (0); 1190 next += ntohs(q->ipqe_ip->ip_len); 1191 } 1192 if (p->ipqe_mff) 1193 return (0); 1194 1195 /* 1196 * Reassembly is complete. Check for a bogus message size and 1197 * concatenate fragments. 1198 */ 1199 q = TAILQ_FIRST(&fp->ipq_fragq); 1200 ip = q->ipqe_ip; 1201 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 1202 ipstat.ips_toolong++; 1203 ip_freef(fp); 1204 return (0); 1205 } 1206 m = q->ipqe_m; 1207 t = m->m_next; 1208 m->m_next = 0; 1209 m_cat(m, t); 1210 nq = TAILQ_NEXT(q, ipqe_q); 1211 s = splvm(); 1212 pool_put(&ipqent_pool, q); 1213 splx(s); 1214 for (q = nq; q != NULL; q = nq) { 1215 t = q->ipqe_m; 1216 nq = TAILQ_NEXT(q, ipqe_q); 1217 s = splvm(); 1218 pool_put(&ipqent_pool, q); 1219 splx(s); 1220 m_cat(m, t); 1221 } 1222 ip_nfrags -= fp->ipq_nfrags; 1223 1224 /* 1225 * Create header for new ip packet by 1226 * modifying header of first packet; 1227 * dequeue and discard fragment reassembly header. 1228 * Make header visible. 1229 */ 1230 ip->ip_len = htons(next); 1231 ip->ip_src = fp->ipq_src; 1232 ip->ip_dst = fp->ipq_dst; 1233 LIST_REMOVE(fp, ipq_q); 1234 FREE(fp, M_FTABLE); 1235 ip_nfragpackets--; 1236 m->m_len += (ip->ip_hl << 2); 1237 m->m_data -= (ip->ip_hl << 2); 1238 /* some debugging cruft by sklower, below, will go away soon */ 1239 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1240 int plen = 0; 1241 for (t = m; t; t = t->m_next) 1242 plen += t->m_len; 1243 m->m_pkthdr.len = plen; 1244 m->m_pkthdr.csum_flags = 0; 1245 } 1246 return (m); 1247 1248dropfrag: 1249 if (fp != 0) 1250 fp->ipq_nfrags--; 1251 ip_nfrags--; 1252 ipstat.ips_fragdropped++; 1253 m_freem(m); 1254 s = splvm(); 1255 pool_put(&ipqent_pool, ipqe); 1256 splx(s); 1257 return (0); 1258} 1259 1260/* 1261 * Free a fragment reassembly header and all 1262 * associated datagrams. 1263 */ 1264void 1265ip_freef(struct ipq *fp) 1266{ 1267 struct ipqent *q, *p; 1268 u_int nfrags = 0; 1269 int s; 1270 1271 IPQ_LOCK_CHECK(); 1272 1273 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) { 1274 p = TAILQ_NEXT(q, ipqe_q); 1275 m_freem(q->ipqe_m); 1276 nfrags++; 1277 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1278 s = splvm(); 1279 pool_put(&ipqent_pool, q); 1280 splx(s); 1281 } 1282 1283 if (nfrags != fp->ipq_nfrags) 1284 printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags); 1285 ip_nfrags -= nfrags; 1286 LIST_REMOVE(fp, ipq_q); 1287 FREE(fp, M_FTABLE); 1288 ip_nfragpackets--; 1289} 1290 1291/* 1292 * IP reassembly TTL machinery for multiplicative drop. 1293 */ 1294static u_int fragttl_histo[(IPFRAGTTL+1)]; 1295 1296 1297/* 1298 * Decrement TTL of all reasembly queue entries by `ticks'. 1299 * Count number of distinct fragments (as opposed to partial, fragmented 1300 * datagrams) in the reassembly queue. While we traverse the entire 1301 * reassembly queue, compute and return the median TTL over all fragments. 1302 */ 1303static u_int 1304ip_reass_ttl_decr(u_int ticks) 1305{ 1306 u_int nfrags, median, dropfraction, keepfraction; 1307 struct ipq *fp, *nfp; 1308 int i; 1309 1310 nfrags = 0; 1311 memset(fragttl_histo, 0, sizeof fragttl_histo); 1312 1313 for (i = 0; i < IPREASS_NHASH; i++) { 1314 for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) { 1315 fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ? 1316 0 : fp->ipq_ttl - ticks); 1317 nfp = LIST_NEXT(fp, ipq_q); 1318 if (fp->ipq_ttl == 0) { 1319 ipstat.ips_fragtimeout++; 1320 ip_freef(fp); 1321 } else { 1322 nfrags += fp->ipq_nfrags; 1323 fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags; 1324 } 1325 } 1326 } 1327 1328 KASSERT(ip_nfrags == nfrags); 1329 1330 /* Find median (or other drop fraction) in histogram. */ 1331 dropfraction = (ip_nfrags / 2); 1332 keepfraction = ip_nfrags - dropfraction; 1333 for (i = IPFRAGTTL, median = 0; i >= 0; i--) { 1334 median += fragttl_histo[i]; 1335 if (median >= keepfraction) 1336 break; 1337 } 1338 1339 /* Return TTL of median (or other fraction). */ 1340 return (u_int)i; 1341} 1342 1343void 1344ip_reass_drophalf(void) 1345{ 1346 1347 u_int median_ticks; 1348 /* 1349 * Compute median TTL of all fragments, and count frags 1350 * with that TTL or lower (roughly half of all fragments). 1351 */ 1352 median_ticks = ip_reass_ttl_decr(0); 1353 1354 /* Drop half. */ 1355 median_ticks = ip_reass_ttl_decr(median_ticks); 1356 1357} 1358 1359/* 1360 * IP timer processing; 1361 * if a timer expires on a reassembly 1362 * queue, discard it. 1363 */ 1364void 1365ip_slowtimo(void) 1366{ 1367 static u_int dropscanidx = 0; 1368 u_int i; 1369 u_int median_ttl; 1370 int s = splsoftnet(); 1371 1372 IPQ_LOCK(); 1373 1374 /* Age TTL of all fragments by 1 tick .*/ 1375 median_ttl = ip_reass_ttl_decr(1); 1376 1377 /* make sure fragment limit is up-to-date */ 1378 CHECK_NMBCLUSTER_PARAMS(); 1379 1380 /* If we have too many fragments, drop the older half. */ 1381 if (ip_nfrags > ip_maxfrags) 1382 ip_reass_ttl_decr(median_ttl); 1383 1384 /* 1385 * If we are over the maximum number of fragmented packets 1386 * (due to the limit being lowered), drain off 1387 * enough to get down to the new limit. Start draining 1388 * from the reassembly hashqueue most recently drained. 1389 */ 1390 if (ip_maxfragpackets < 0) 1391 ; 1392 else { 1393 int wrapped = 0; 1394 1395 i = dropscanidx; 1396 while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) { 1397 while (LIST_FIRST(&ipq[i]) != NULL) 1398 ip_freef(LIST_FIRST(&ipq[i])); 1399 if (++i >= IPREASS_NHASH) { 1400 i = 0; 1401 } 1402 /* 1403 * Dont scan forever even if fragment counters are 1404 * wrong: stop after scanning entire reassembly queue. 1405 */ 1406 if (i == dropscanidx) 1407 wrapped = 1; 1408 } 1409 dropscanidx = i; 1410 } 1411 IPQ_UNLOCK(); 1412 splx(s); 1413} 1414 1415/* 1416 * Drain off all datagram fragments. 1417 */ 1418void 1419ip_drain(void) 1420{ 1421 1422 /* 1423 * We may be called from a device's interrupt context. If 1424 * the ipq is already busy, just bail out now. 1425 */ 1426 if (ipq_lock_try() == 0) 1427 return; 1428 1429 /* 1430 * Drop half the total fragments now. If more mbufs are needed, 1431 * we will be called again soon. 1432 */ 1433 ip_reass_drophalf(); 1434 1435 IPQ_UNLOCK(); 1436} 1437 1438/* 1439 * Do option processing on a datagram, 1440 * possibly discarding it if bad options are encountered, 1441 * or forwarding it if source-routed. 1442 * Returns 1 if packet has been forwarded/freed, 1443 * 0 if the packet should be processed further. 1444 */ 1445int 1446ip_dooptions(struct mbuf *m) 1447{ 1448 struct ip *ip = mtod(m, struct ip *); 1449 u_char *cp, *cp0; 1450 struct ip_timestamp *ipt; 1451 struct in_ifaddr *ia; 1452 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1453 struct in_addr dst; 1454 n_time ntime; 1455 1456 dst = ip->ip_dst; 1457 cp = (u_char *)(ip + 1); 1458 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1459 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1460 opt = cp[IPOPT_OPTVAL]; 1461 if (opt == IPOPT_EOL) 1462 break; 1463 if (opt == IPOPT_NOP) 1464 optlen = 1; 1465 else { 1466 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1467 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1468 goto bad; 1469 } 1470 optlen = cp[IPOPT_OLEN]; 1471 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1472 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1473 goto bad; 1474 } 1475 } 1476 switch (opt) { 1477 1478 default: 1479 break; 1480 1481 /* 1482 * Source routing with record. 1483 * Find interface with current destination address. 1484 * If none on this machine then drop if strictly routed, 1485 * or do nothing if loosely routed. 1486 * Record interface address and bring up next address 1487 * component. If strictly routed make sure next 1488 * address is on directly accessible net. 1489 */ 1490 case IPOPT_LSRR: 1491 case IPOPT_SSRR: 1492 if (ip_allowsrcrt == 0) { 1493 type = ICMP_UNREACH; 1494 code = ICMP_UNREACH_NET_PROHIB; 1495 goto bad; 1496 } 1497 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1498 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1499 goto bad; 1500 } 1501 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1502 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1503 goto bad; 1504 } 1505 ipaddr.sin_addr = ip->ip_dst; 1506 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1507 if (ia == 0) { 1508 if (opt == IPOPT_SSRR) { 1509 type = ICMP_UNREACH; 1510 code = ICMP_UNREACH_SRCFAIL; 1511 goto bad; 1512 } 1513 /* 1514 * Loose routing, and not at next destination 1515 * yet; nothing to do except forward. 1516 */ 1517 break; 1518 } 1519 off--; /* 0 origin */ 1520 if ((off + sizeof(struct in_addr)) > optlen) { 1521 /* 1522 * End of source route. Should be for us. 1523 */ 1524 save_rte(cp, ip->ip_src); 1525 break; 1526 } 1527 /* 1528 * locate outgoing interface 1529 */ 1530 bcopy((void *)(cp + off), (void *)&ipaddr.sin_addr, 1531 sizeof(ipaddr.sin_addr)); 1532 if (opt == IPOPT_SSRR) 1533 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); 1534 else 1535 ia = ip_rtaddr(ipaddr.sin_addr); 1536 if (ia == 0) { 1537 type = ICMP_UNREACH; 1538 code = ICMP_UNREACH_SRCFAIL; 1539 goto bad; 1540 } 1541 ip->ip_dst = ipaddr.sin_addr; 1542 bcopy((void *)&ia->ia_addr.sin_addr, 1543 (void *)(cp + off), sizeof(struct in_addr)); 1544 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1545 /* 1546 * Let ip_intr's mcast routing check handle mcast pkts 1547 */ 1548 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1549 break; 1550 1551 case IPOPT_RR: 1552 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1553 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1554 goto bad; 1555 } 1556 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1557 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1558 goto bad; 1559 } 1560 /* 1561 * If no space remains, ignore. 1562 */ 1563 off--; /* 0 origin */ 1564 if ((off + sizeof(struct in_addr)) > optlen) 1565 break; 1566 bcopy((void *)(&ip->ip_dst), (void *)&ipaddr.sin_addr, 1567 sizeof(ipaddr.sin_addr)); 1568 /* 1569 * locate outgoing interface; if we're the destination, 1570 * use the incoming interface (should be same). 1571 */ 1572 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1573 == NULL && 1574 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1575 type = ICMP_UNREACH; 1576 code = ICMP_UNREACH_HOST; 1577 goto bad; 1578 } 1579 bcopy((void *)&ia->ia_addr.sin_addr, 1580 (void *)(cp + off), sizeof(struct in_addr)); 1581 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1582 break; 1583 1584 case IPOPT_TS: 1585 code = cp - (u_char *)ip; 1586 ipt = (struct ip_timestamp *)cp; 1587 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1588 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1589 goto bad; 1590 } 1591 if (ipt->ipt_ptr < 5) { 1592 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1593 goto bad; 1594 } 1595 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1596 if (++ipt->ipt_oflw == 0) { 1597 code = (u_char *)&ipt->ipt_ptr - 1598 (u_char *)ip; 1599 goto bad; 1600 } 1601 break; 1602 } 1603 cp0 = (cp + ipt->ipt_ptr - 1); 1604 switch (ipt->ipt_flg) { 1605 1606 case IPOPT_TS_TSONLY: 1607 break; 1608 1609 case IPOPT_TS_TSANDADDR: 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 ipaddr.sin_addr = dst; 1617 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1618 m->m_pkthdr.rcvif)); 1619 if (ia == 0) 1620 continue; 1621 bcopy(&ia->ia_addr.sin_addr, 1622 cp0, sizeof(struct in_addr)); 1623 ipt->ipt_ptr += sizeof(struct in_addr); 1624 break; 1625 1626 case IPOPT_TS_PRESPEC: 1627 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1628 sizeof(struct in_addr) > ipt->ipt_len) { 1629 code = (u_char *)&ipt->ipt_ptr - 1630 (u_char *)ip; 1631 goto bad; 1632 } 1633 bcopy(cp0, &ipaddr.sin_addr, 1634 sizeof(struct in_addr)); 1635 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1636 == NULL) 1637 continue; 1638 ipt->ipt_ptr += sizeof(struct in_addr); 1639 break; 1640 1641 default: 1642 /* XXX can't take &ipt->ipt_flg */ 1643 code = (u_char *)&ipt->ipt_ptr - 1644 (u_char *)ip + 1; 1645 goto bad; 1646 } 1647 ntime = iptime(); 1648 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1649 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1650 sizeof(n_time)); 1651 ipt->ipt_ptr += sizeof(n_time); 1652 } 1653 } 1654 if (forward) { 1655 if (ip_forwsrcrt == 0) { 1656 type = ICMP_UNREACH; 1657 code = ICMP_UNREACH_SRCFAIL; 1658 goto bad; 1659 } 1660 ip_forward(m, 1); 1661 return (1); 1662 } 1663 return (0); 1664bad: 1665 icmp_error(m, type, code, 0, 0); 1666 ipstat.ips_badoptions++; 1667 return (1); 1668} 1669 1670/* 1671 * Given address of next destination (final or next hop), 1672 * return internet address info of interface to be used to get there. 1673 */ 1674struct in_ifaddr * 1675ip_rtaddr(struct in_addr dst) 1676{ 1677 if (!in_hosteq(dst, satocsin(rtcache_getdst(&ipforward_rt))->sin_addr)) 1678 rtcache_free(&ipforward_rt); 1679 else 1680 rtcache_check(&ipforward_rt); 1681 1682 if (ipforward_rt.ro_rt == NULL) { 1683 struct sockaddr_in *sin = satosin(&ipforward_rt.ro_dst); 1684 1685 sin->sin_family = AF_INET; 1686 sin->sin_len = sizeof(*sin); 1687 sin->sin_addr = dst; 1688 1689 rtcache_init(&ipforward_rt); 1690 if (ipforward_rt.ro_rt == NULL) 1691 return NULL; 1692 } 1693 return ifatoia(ipforward_rt.ro_rt->rt_ifa); 1694} 1695 1696/* 1697 * Save incoming source route for use in replies, 1698 * to be picked up later by ip_srcroute if the receiver is interested. 1699 */ 1700void 1701save_rte(u_char *option, struct in_addr dst) 1702{ 1703 unsigned olen; 1704 1705 olen = option[IPOPT_OLEN]; 1706#ifdef DIAGNOSTIC 1707 if (ipprintfs) 1708 printf("save_rte: olen %d\n", olen); 1709#endif /* 0 */ 1710 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1711 return; 1712 bcopy((void *)option, (void *)ip_srcrt.srcopt, olen); 1713 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1714 ip_srcrt.dst = dst; 1715} 1716 1717/* 1718 * Retrieve incoming source route for use in replies, 1719 * in the same form used by setsockopt. 1720 * The first hop is placed before the options, will be removed later. 1721 */ 1722struct mbuf * 1723ip_srcroute(void) 1724{ 1725 struct in_addr *p, *q; 1726 struct mbuf *m; 1727 1728 if (ip_nhops == 0) 1729 return NULL; 1730 m = m_get(M_DONTWAIT, MT_SOOPTS); 1731 if (m == 0) 1732 return NULL; 1733 1734 MCLAIM(m, &inetdomain.dom_mowner); 1735#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1736 1737 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1738 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1739 OPTSIZ; 1740#ifdef DIAGNOSTIC 1741 if (ipprintfs) 1742 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1743#endif 1744 1745 /* 1746 * First save first hop for return route 1747 */ 1748 p = &ip_srcrt.route[ip_nhops - 1]; 1749 *(mtod(m, struct in_addr *)) = *p--; 1750#ifdef DIAGNOSTIC 1751 if (ipprintfs) 1752 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1753#endif 1754 1755 /* 1756 * Copy option fields and padding (nop) to mbuf. 1757 */ 1758 ip_srcrt.nop = IPOPT_NOP; 1759 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1760 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop, 1761 OPTSIZ); 1762 q = (struct in_addr *)(mtod(m, char *) + 1763 sizeof(struct in_addr) + OPTSIZ); 1764#undef OPTSIZ 1765 /* 1766 * Record return path as an IP source route, 1767 * reversing the path (pointers are now aligned). 1768 */ 1769 while (p >= ip_srcrt.route) { 1770#ifdef DIAGNOSTIC 1771 if (ipprintfs) 1772 printf(" %x", ntohl(q->s_addr)); 1773#endif 1774 *q++ = *p--; 1775 } 1776 /* 1777 * Last hop goes to final destination. 1778 */ 1779 *q = ip_srcrt.dst; 1780#ifdef DIAGNOSTIC 1781 if (ipprintfs) 1782 printf(" %x\n", ntohl(q->s_addr)); 1783#endif 1784 return (m); 1785} 1786 1787/* 1788 * Strip out IP options, at higher 1789 * level protocol in the kernel. 1790 * Second argument is buffer to which options 1791 * will be moved, and return value is their length. 1792 * XXX should be deleted; last arg currently ignored. 1793 */ 1794void 1795ip_stripoptions(struct mbuf *m, struct mbuf *mopt) 1796{ 1797 int i; 1798 struct ip *ip = mtod(m, struct ip *); 1799 void *opts; 1800 int olen; 1801 1802 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1803 opts = (void *)(ip + 1); 1804 i = m->m_len - (sizeof (struct ip) + olen); 1805 memmove(opts, (char *)opts + olen, (unsigned)i); 1806 m->m_len -= olen; 1807 if (m->m_flags & M_PKTHDR) 1808 m->m_pkthdr.len -= olen; 1809 ip->ip_len = htons(ntohs(ip->ip_len) - olen); 1810 ip->ip_hl = sizeof (struct ip) >> 2; 1811} 1812 1813const int inetctlerrmap[PRC_NCMDS] = { 1814 0, 0, 0, 0, 1815 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1816 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1817 EMSGSIZE, EHOSTUNREACH, 0, 0, 1818 0, 0, 0, 0, 1819 ENOPROTOOPT 1820}; 1821 1822/* 1823 * Forward a packet. If some error occurs return the sender 1824 * an icmp packet. Note we can't always generate a meaningful 1825 * icmp message because icmp doesn't have a large enough repertoire 1826 * of codes and types. 1827 * 1828 * If not forwarding, just drop the packet. This could be confusing 1829 * if ipforwarding was zero but some routing protocol was advancing 1830 * us as a gateway to somewhere. However, we must let the routing 1831 * protocol deal with that. 1832 * 1833 * The srcrt parameter indicates whether the packet is being forwarded 1834 * via a source route. 1835 */ 1836void 1837ip_forward(struct mbuf *m, int srcrt) 1838{ 1839 struct ip *ip = mtod(m, struct ip *); 1840 struct rtentry *rt; 1841 int error, type = 0, code = 0, destmtu = 0; 1842 struct mbuf *mcopy; 1843 n_long dest; 1844 1845 /* 1846 * We are now in the output path. 1847 */ 1848 MCLAIM(m, &ip_tx_mowner); 1849 1850 /* 1851 * Clear any in-bound checksum flags for this packet. 1852 */ 1853 m->m_pkthdr.csum_flags = 0; 1854 1855 dest = 0; 1856#ifdef DIAGNOSTIC 1857 if (ipprintfs) { 1858 printf("forward: src %s ", inet_ntoa(ip->ip_src)); 1859 printf("dst %s ttl %x\n", inet_ntoa(ip->ip_dst), ip->ip_ttl); 1860 } 1861#endif 1862 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1863 ipstat.ips_cantforward++; 1864 m_freem(m); 1865 return; 1866 } 1867 if (ip->ip_ttl <= IPTTLDEC) { 1868 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1869 return; 1870 } 1871 1872 if (!in_hosteq(ip->ip_dst, 1873 satocsin(rtcache_getdst(&ipforward_rt))->sin_addr)) 1874 rtcache_free(&ipforward_rt); 1875 else 1876 rtcache_check(&ipforward_rt); 1877 if (ipforward_rt.ro_rt == NULL) { 1878 struct sockaddr_in *sin = satosin(&ipforward_rt.ro_dst); 1879 1880 sin->sin_family = AF_INET; 1881 sin->sin_len = sizeof(*sin); 1882 sin->sin_addr = ip->ip_dst; 1883 1884 rtcache_init(&ipforward_rt); 1885 if (ipforward_rt.ro_rt == NULL) { 1886 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1887 return; 1888 } 1889 } 1890 rt = ipforward_rt.ro_rt; 1891 1892 /* 1893 * Save at most 68 bytes of the packet in case 1894 * we need to generate an ICMP message to the src. 1895 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1896 */ 1897 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1898 if (mcopy) 1899 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1900 1901 ip->ip_ttl -= IPTTLDEC; 1902 1903 /* 1904 * If forwarding packet using same interface that it came in on, 1905 * perhaps should send a redirect to sender to shortcut a hop. 1906 * Only send redirect if source is sending directly to us, 1907 * and if packet was not source routed (or has any options). 1908 * Also, don't send redirect if forwarding using a default route 1909 * or a route modified by a redirect. 1910 */ 1911 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1912 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1913 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1914 ipsendredirects && !srcrt) { 1915 if (rt->rt_ifa && 1916 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1917 ifatoia(rt->rt_ifa)->ia_subnet) { 1918 if (rt->rt_flags & RTF_GATEWAY) 1919 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1920 else 1921 dest = ip->ip_dst.s_addr; 1922 /* 1923 * Router requirements says to only send host 1924 * redirects. 1925 */ 1926 type = ICMP_REDIRECT; 1927 code = ICMP_REDIRECT_HOST; 1928#ifdef DIAGNOSTIC 1929 if (ipprintfs) 1930 printf("redirect (%d) to %x\n", code, 1931 (u_int32_t)dest); 1932#endif 1933 } 1934 } 1935 1936 error = ip_output(m, NULL, &ipforward_rt, 1937 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1938 (struct ip_moptions *)NULL, (struct socket *)NULL); 1939 1940 if (error) 1941 ipstat.ips_cantforward++; 1942 else { 1943 ipstat.ips_forward++; 1944 if (type) 1945 ipstat.ips_redirectsent++; 1946 else { 1947 if (mcopy) { 1948#ifdef GATEWAY 1949 if (mcopy->m_flags & M_CANFASTFWD) 1950 ipflow_create(&ipforward_rt, mcopy); 1951#endif 1952 m_freem(mcopy); 1953 } 1954 return; 1955 } 1956 } 1957 if (mcopy == NULL) 1958 return; 1959 1960 switch (error) { 1961 1962 case 0: /* forwarded, but need redirect */ 1963 /* type, code set above */ 1964 break; 1965 1966 case ENETUNREACH: /* shouldn't happen, checked above */ 1967 case EHOSTUNREACH: 1968 case ENETDOWN: 1969 case EHOSTDOWN: 1970 default: 1971 type = ICMP_UNREACH; 1972 code = ICMP_UNREACH_HOST; 1973 break; 1974 1975 case EMSGSIZE: 1976 type = ICMP_UNREACH; 1977 code = ICMP_UNREACH_NEEDFRAG; 1978#if !defined(IPSEC) && !defined(FAST_IPSEC) 1979 if (ipforward_rt.ro_rt != NULL) 1980 destmtu = ipforward_rt.ro_rt->rt_ifp->if_mtu; 1981#else 1982 /* 1983 * If the packet is routed over IPsec tunnel, tell the 1984 * originator the tunnel MTU. 1985 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1986 * XXX quickhack!!! 1987 */ 1988 if (ipforward_rt.ro_rt != NULL) { 1989 struct secpolicy *sp; 1990 int ipsecerror; 1991 size_t ipsechdr; 1992 struct route *ro; 1993 1994 sp = ipsec4_getpolicybyaddr(mcopy, 1995 IPSEC_DIR_OUTBOUND, IP_FORWARDING, 1996 &ipsecerror); 1997 1998 if (sp == NULL) 1999 destmtu = ipforward_rt.ro_rt->rt_ifp->if_mtu; 2000 else { 2001 /* count IPsec header size */ 2002 ipsechdr = ipsec4_hdrsiz(mcopy, 2003 IPSEC_DIR_OUTBOUND, NULL); 2004 2005 /* 2006 * find the correct route for outer IPv4 2007 * header, compute tunnel MTU. 2008 */ 2009 2010 if (sp->req != NULL 2011 && sp->req->sav != NULL 2012 && sp->req->sav->sah != NULL) { 2013 ro = &sp->req->sav->sah->sa_route; 2014 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 2015 destmtu = 2016 ro->ro_rt->rt_rmx.rmx_mtu ? 2017 ro->ro_rt->rt_rmx.rmx_mtu : 2018 ro->ro_rt->rt_ifp->if_mtu; 2019 destmtu -= ipsechdr; 2020 } 2021 } 2022 2023#ifdef IPSEC 2024 key_freesp(sp); 2025#else 2026 KEY_FREESP(&sp); 2027#endif 2028 } 2029 } 2030#endif /*IPSEC*/ 2031 ipstat.ips_cantfrag++; 2032 break; 2033 2034 case ENOBUFS: 2035#if 1 2036 /* 2037 * a router should not generate ICMP_SOURCEQUENCH as 2038 * required in RFC1812 Requirements for IP Version 4 Routers. 2039 * source quench could be a big problem under DoS attacks, 2040 * or if the underlying interface is rate-limited. 2041 */ 2042 if (mcopy) 2043 m_freem(mcopy); 2044 return; 2045#else 2046 type = ICMP_SOURCEQUENCH; 2047 code = 0; 2048 break; 2049#endif 2050 } 2051 icmp_error(mcopy, type, code, dest, destmtu); 2052} 2053 2054void 2055ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2056 struct mbuf *m) 2057{ 2058 2059 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2060 struct timeval tv; 2061 2062 microtime(&tv); 2063 *mp = sbcreatecontrol((void *) &tv, sizeof(tv), 2064 SCM_TIMESTAMP, SOL_SOCKET); 2065 if (*mp) 2066 mp = &(*mp)->m_next; 2067 } 2068 if (inp->inp_flags & INP_RECVDSTADDR) { 2069 *mp = sbcreatecontrol((void *) &ip->ip_dst, 2070 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2071 if (*mp) 2072 mp = &(*mp)->m_next; 2073 } 2074#ifdef notyet 2075 /* 2076 * XXX 2077 * Moving these out of udp_input() made them even more broken 2078 * than they already were. 2079 * - fenner@parc.xerox.com 2080 */ 2081 /* options were tossed already */ 2082 if (inp->inp_flags & INP_RECVOPTS) { 2083 *mp = sbcreatecontrol((void *) opts_deleted_above, 2084 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2085 if (*mp) 2086 mp = &(*mp)->m_next; 2087 } 2088 /* ip_srcroute doesn't do what we want here, need to fix */ 2089 if (inp->inp_flags & INP_RECVRETOPTS) { 2090 *mp = sbcreatecontrol((void *) ip_srcroute(), 2091 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2092 if (*mp) 2093 mp = &(*mp)->m_next; 2094 } 2095#endif 2096 if (inp->inp_flags & INP_RECVIF) { 2097 struct sockaddr_dl sdl; 2098 2099 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 2100 sdl.sdl_family = AF_LINK; 2101 sdl.sdl_index = m->m_pkthdr.rcvif ? 2102 m->m_pkthdr.rcvif->if_index : 0; 2103 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 2104 *mp = sbcreatecontrol((void *) &sdl, sdl.sdl_len, 2105 IP_RECVIF, IPPROTO_IP); 2106 if (*mp) 2107 mp = &(*mp)->m_next; 2108 } 2109} 2110 2111/* 2112 * sysctl helper routine for net.inet.ip.forwsrcrt. 2113 */ 2114static int 2115sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 2116{ 2117 int error, tmp; 2118 struct sysctlnode node; 2119 2120 node = *rnode; 2121 tmp = ip_forwsrcrt; 2122 node.sysctl_data = &tmp; 2123 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2124 if (error || newp == NULL) 2125 return (error); 2126 2127 if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 2128 0, NULL, NULL, NULL)) 2129 return (EPERM); 2130 2131 ip_forwsrcrt = tmp; 2132 2133 return (0); 2134} 2135 2136/* 2137 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 2138 * range of the new value and tweaks timers if it changes. 2139 */ 2140static int 2141sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 2142{ 2143 int error, tmp; 2144 struct sysctlnode node; 2145 2146 node = *rnode; 2147 tmp = ip_mtudisc_timeout; 2148 node.sysctl_data = &tmp; 2149 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2150 if (error || newp == NULL) 2151 return (error); 2152 if (tmp < 0) 2153 return (EINVAL); 2154 2155 ip_mtudisc_timeout = tmp; 2156 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 2157 2158 return (0); 2159} 2160 2161#ifdef GATEWAY 2162/* 2163 * sysctl helper routine for net.inet.ip.maxflows. 2164 */ 2165static int 2166sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS) 2167{ 2168 int s; 2169 2170 s = sysctl_lookup(SYSCTLFN_CALL(rnode)); 2171 if (s || newp == NULL) 2172 return (s); 2173 2174 s = splsoftnet(); 2175 ipflow_reap(0); 2176 splx(s); 2177 2178 return (0); 2179} 2180 2181static int 2182sysctl_net_inet_ip_hashsize(SYSCTLFN_ARGS) 2183{ 2184 int error, tmp; 2185 struct sysctlnode node; 2186 2187 node = *rnode; 2188 tmp = ip_hashsize; 2189 node.sysctl_data = &tmp; 2190 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2191 if (error || newp == NULL) 2192 return (error); 2193 2194 if ((tmp & (tmp - 1)) == 0 && tmp != 0) { 2195 /* 2196 * Can only fail due to malloc() 2197 */ 2198 if (ipflow_invalidate_all(tmp)) 2199 return ENOMEM; 2200 } else { 2201 /* 2202 * EINVAL if not a power of 2 2203 */ 2204 return EINVAL; 2205 } 2206 2207 return (0); 2208} 2209#endif /* GATEWAY */ 2210 2211 2212SYSCTL_SETUP(sysctl_net_inet_ip_setup, "sysctl net.inet.ip subtree setup") 2213{ 2214 extern int subnetsarelocal, hostzeroisbroadcast; 2215 2216 sysctl_createv(clog, 0, NULL, NULL, 2217 CTLFLAG_PERMANENT, 2218 CTLTYPE_NODE, "net", NULL, 2219 NULL, 0, NULL, 0, 2220 CTL_NET, CTL_EOL); 2221 sysctl_createv(clog, 0, NULL, NULL, 2222 CTLFLAG_PERMANENT, 2223 CTLTYPE_NODE, "inet", 2224 SYSCTL_DESCR("PF_INET related settings"), 2225 NULL, 0, NULL, 0, 2226 CTL_NET, PF_INET, CTL_EOL); 2227 sysctl_createv(clog, 0, NULL, NULL, 2228 CTLFLAG_PERMANENT, 2229 CTLTYPE_NODE, "ip", 2230 SYSCTL_DESCR("IPv4 related settings"), 2231 NULL, 0, NULL, 0, 2232 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 2233 2234 sysctl_createv(clog, 0, NULL, NULL, 2235 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2236 CTLTYPE_INT, "forwarding", 2237 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 2238 NULL, 0, &ipforwarding, 0, 2239 CTL_NET, PF_INET, IPPROTO_IP, 2240 IPCTL_FORWARDING, CTL_EOL); 2241 sysctl_createv(clog, 0, NULL, NULL, 2242 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2243 CTLTYPE_INT, "redirect", 2244 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 2245 NULL, 0, &ipsendredirects, 0, 2246 CTL_NET, PF_INET, IPPROTO_IP, 2247 IPCTL_SENDREDIRECTS, CTL_EOL); 2248 sysctl_createv(clog, 0, NULL, NULL, 2249 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2250 CTLTYPE_INT, "ttl", 2251 SYSCTL_DESCR("Default TTL for an INET datagram"), 2252 NULL, 0, &ip_defttl, 0, 2253 CTL_NET, PF_INET, IPPROTO_IP, 2254 IPCTL_DEFTTL, CTL_EOL); 2255#ifdef IPCTL_DEFMTU 2256 sysctl_createv(clog, 0, NULL, NULL, 2257 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 2258 CTLTYPE_INT, "mtu", 2259 SYSCTL_DESCR("Default MTA for an INET route"), 2260 NULL, 0, &ip_mtu, 0, 2261 CTL_NET, PF_INET, IPPROTO_IP, 2262 IPCTL_DEFMTU, CTL_EOL); 2263#endif /* IPCTL_DEFMTU */ 2264 sysctl_createv(clog, 0, NULL, NULL, 2265 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2266 CTLTYPE_INT, "forwsrcrt", 2267 SYSCTL_DESCR("Enable forwarding of source-routed " 2268 "datagrams"), 2269 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 2270 CTL_NET, PF_INET, IPPROTO_IP, 2271 IPCTL_FORWSRCRT, CTL_EOL); 2272 sysctl_createv(clog, 0, NULL, NULL, 2273 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2274 CTLTYPE_INT, "directed-broadcast", 2275 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 2276 NULL, 0, &ip_directedbcast, 0, 2277 CTL_NET, PF_INET, IPPROTO_IP, 2278 IPCTL_DIRECTEDBCAST, CTL_EOL); 2279 sysctl_createv(clog, 0, NULL, NULL, 2280 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2281 CTLTYPE_INT, "allowsrcrt", 2282 SYSCTL_DESCR("Accept source-routed datagrams"), 2283 NULL, 0, &ip_allowsrcrt, 0, 2284 CTL_NET, PF_INET, IPPROTO_IP, 2285 IPCTL_ALLOWSRCRT, CTL_EOL); 2286 sysctl_createv(clog, 0, NULL, NULL, 2287 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2288 CTLTYPE_INT, "subnetsarelocal", 2289 SYSCTL_DESCR("Whether logical subnets are considered " 2290 "local"), 2291 NULL, 0, &subnetsarelocal, 0, 2292 CTL_NET, PF_INET, IPPROTO_IP, 2293 IPCTL_SUBNETSARELOCAL, CTL_EOL); 2294 sysctl_createv(clog, 0, NULL, NULL, 2295 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2296 CTLTYPE_INT, "mtudisc", 2297 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 2298 NULL, 0, &ip_mtudisc, 0, 2299 CTL_NET, PF_INET, IPPROTO_IP, 2300 IPCTL_MTUDISC, CTL_EOL); 2301 sysctl_createv(clog, 0, NULL, NULL, 2302 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2303 CTLTYPE_INT, "anonportmin", 2304 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 2305 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 2306 CTL_NET, PF_INET, IPPROTO_IP, 2307 IPCTL_ANONPORTMIN, CTL_EOL); 2308 sysctl_createv(clog, 0, NULL, NULL, 2309 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2310 CTLTYPE_INT, "anonportmax", 2311 SYSCTL_DESCR("Highest ephemeral port number to assign"), 2312 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 2313 CTL_NET, PF_INET, IPPROTO_IP, 2314 IPCTL_ANONPORTMAX, CTL_EOL); 2315 sysctl_createv(clog, 0, NULL, NULL, 2316 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2317 CTLTYPE_INT, "mtudisctimeout", 2318 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 2319 sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0, 2320 CTL_NET, PF_INET, IPPROTO_IP, 2321 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 2322#ifdef GATEWAY 2323 sysctl_createv(clog, 0, NULL, NULL, 2324 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2325 CTLTYPE_INT, "maxflows", 2326 SYSCTL_DESCR("Number of flows for fast forwarding"), 2327 sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0, 2328 CTL_NET, PF_INET, IPPROTO_IP, 2329 IPCTL_MAXFLOWS, CTL_EOL); 2330 sysctl_createv(clog, 0, NULL, NULL, 2331 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2332 CTLTYPE_INT, "hashsize", 2333 SYSCTL_DESCR("Size of hash table for fast forwarding (IPv4)"), 2334 sysctl_net_inet_ip_hashsize, 0, &ip_hashsize, 0, 2335 CTL_NET, PF_INET, IPPROTO_IP, 2336 CTL_CREATE, CTL_EOL); 2337#endif /* GATEWAY */ 2338 sysctl_createv(clog, 0, NULL, NULL, 2339 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2340 CTLTYPE_INT, "hostzerobroadcast", 2341 SYSCTL_DESCR("All zeroes address is broadcast address"), 2342 NULL, 0, &hostzeroisbroadcast, 0, 2343 CTL_NET, PF_INET, IPPROTO_IP, 2344 IPCTL_HOSTZEROBROADCAST, CTL_EOL); 2345#if NGIF > 0 2346 sysctl_createv(clog, 0, NULL, NULL, 2347 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2348 CTLTYPE_INT, "gifttl", 2349 SYSCTL_DESCR("Default TTL for a gif tunnel datagram"), 2350 NULL, 0, &ip_gif_ttl, 0, 2351 CTL_NET, PF_INET, IPPROTO_IP, 2352 IPCTL_GIF_TTL, CTL_EOL); 2353#endif /* NGIF */ 2354#ifndef IPNOPRIVPORTS 2355 sysctl_createv(clog, 0, NULL, NULL, 2356 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2357 CTLTYPE_INT, "lowportmin", 2358 SYSCTL_DESCR("Lowest privileged ephemeral port number " 2359 "to assign"), 2360 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 2361 CTL_NET, PF_INET, IPPROTO_IP, 2362 IPCTL_LOWPORTMIN, CTL_EOL); 2363 sysctl_createv(clog, 0, NULL, NULL, 2364 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2365 CTLTYPE_INT, "lowportmax", 2366 SYSCTL_DESCR("Highest privileged ephemeral port number " 2367 "to assign"), 2368 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 2369 CTL_NET, PF_INET, IPPROTO_IP, 2370 IPCTL_LOWPORTMAX, CTL_EOL); 2371#endif /* IPNOPRIVPORTS */ 2372 sysctl_createv(clog, 0, NULL, NULL, 2373 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2374 CTLTYPE_INT, "maxfragpackets", 2375 SYSCTL_DESCR("Maximum number of fragments to retain for " 2376 "possible reassembly"), 2377 NULL, 0, &ip_maxfragpackets, 0, 2378 CTL_NET, PF_INET, IPPROTO_IP, 2379 IPCTL_MAXFRAGPACKETS, CTL_EOL); 2380#if NGRE > 0 2381 sysctl_createv(clog, 0, NULL, NULL, 2382 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2383 CTLTYPE_INT, "grettl", 2384 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 2385 NULL, 0, &ip_gre_ttl, 0, 2386 CTL_NET, PF_INET, IPPROTO_IP, 2387 IPCTL_GRE_TTL, CTL_EOL); 2388#endif /* NGRE */ 2389 sysctl_createv(clog, 0, NULL, NULL, 2390 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2391 CTLTYPE_INT, "checkinterface", 2392 SYSCTL_DESCR("Enable receive side of Strong ES model " 2393 "from RFC1122"), 2394 NULL, 0, &ip_checkinterface, 0, 2395 CTL_NET, PF_INET, IPPROTO_IP, 2396 IPCTL_CHECKINTERFACE, CTL_EOL); 2397 sysctl_createv(clog, 0, NULL, NULL, 2398 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2399 CTLTYPE_INT, "random_id", 2400 SYSCTL_DESCR("Assign random ip_id values"), 2401 NULL, 0, &ip_do_randomid, 0, 2402 CTL_NET, PF_INET, IPPROTO_IP, 2403 IPCTL_RANDOMID, CTL_EOL); 2404 sysctl_createv(clog, 0, NULL, NULL, 2405 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2406 CTLTYPE_INT, "do_loopback_cksum", 2407 SYSCTL_DESCR("Perform IP checksum on loopback"), 2408 NULL, 0, &ip_do_loopback_cksum, 0, 2409 CTL_NET, PF_INET, IPPROTO_IP, 2410 IPCTL_LOOPBACKCKSUM, CTL_EOL); 2411 sysctl_createv(clog, 0, NULL, NULL, 2412 CTLFLAG_PERMANENT, 2413 CTLTYPE_STRUCT, "stats", 2414 SYSCTL_DESCR("IP statistics"), 2415 NULL, 0, &ipstat, sizeof(ipstat), 2416 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 2417 CTL_EOL); 2418} 2419