ip_input.c revision 1.260
1/* $NetBSD: ip_input.c,v 1.260 2007/12/22 15:41:11 matt 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 1.260 2007/12/22 15:41:11 matt 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 ip_input(m); 473 } 474} 475 476/* 477 * Ip input routine. Checksum and byte swap header. If fragmented 478 * try to reassemble. Process options. Pass to next level. 479 */ 480void 481ip_input(struct mbuf *m) 482{ 483 struct ip *ip = NULL; 484 struct ipq *fp; 485 struct in_ifaddr *ia; 486 struct ifaddr *ifa; 487 struct ipqent *ipqe; 488 int hlen = 0, mff, len; 489 int downmatch; 490 int checkif; 491 int srcrt = 0; 492 int s; 493 u_int hash; 494#ifdef FAST_IPSEC 495 struct m_tag *mtag; 496 struct tdb_ident *tdbi; 497 struct secpolicy *sp; 498 int error; 499#endif /* FAST_IPSEC */ 500 501 MCLAIM(m, &ip_rx_mowner); 502#ifdef DIAGNOSTIC 503 if ((m->m_flags & M_PKTHDR) == 0) 504 panic("ipintr no HDR"); 505#endif 506 507 /* 508 * If no IP addresses have been set yet but the interfaces 509 * are receiving, can't do anything with incoming packets yet. 510 */ 511 if (TAILQ_FIRST(&in_ifaddrhead) == 0) 512 goto bad; 513 ipstat.ips_total++; 514 /* 515 * If the IP header is not aligned, slurp it up into a new 516 * mbuf with space for link headers, in the event we forward 517 * it. Otherwise, if it is aligned, make sure the entire 518 * base IP header is in the first mbuf of the chain. 519 */ 520 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 521 if ((m = m_copyup(m, sizeof(struct ip), 522 (max_linkhdr + 3) & ~3)) == NULL) { 523 /* XXXJRT new stat, please */ 524 ipstat.ips_toosmall++; 525 return; 526 } 527 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 528 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 529 ipstat.ips_toosmall++; 530 return; 531 } 532 } 533 ip = mtod(m, struct ip *); 534 if (ip->ip_v != IPVERSION) { 535 ipstat.ips_badvers++; 536 goto bad; 537 } 538 hlen = ip->ip_hl << 2; 539 if (hlen < sizeof(struct ip)) { /* minimum header length */ 540 ipstat.ips_badhlen++; 541 goto bad; 542 } 543 if (hlen > m->m_len) { 544 if ((m = m_pullup(m, hlen)) == 0) { 545 ipstat.ips_badhlen++; 546 return; 547 } 548 ip = mtod(m, struct ip *); 549 } 550 551 /* 552 * RFC1122: packets with a multicast source address are 553 * not allowed. 554 */ 555 if (IN_MULTICAST(ip->ip_src.s_addr)) { 556 ipstat.ips_badaddr++; 557 goto bad; 558 } 559 560 /* 127/8 must not appear on wire - RFC1122 */ 561 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 562 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 563 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 564 ipstat.ips_badaddr++; 565 goto bad; 566 } 567 } 568 569 switch (m->m_pkthdr.csum_flags & 570 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) | 571 M_CSUM_IPv4_BAD)) { 572 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 573 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 574 goto badcsum; 575 576 case M_CSUM_IPv4: 577 /* Checksum was okay. */ 578 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 579 break; 580 581 default: 582 /* 583 * Must compute it ourselves. Maybe skip checksum on 584 * loopback interfaces. 585 */ 586 if (__predict_true(!(m->m_pkthdr.rcvif->if_flags & 587 IFF_LOOPBACK) || ip_do_loopback_cksum)) { 588 INET_CSUM_COUNTER_INCR(&ip_swcsum); 589 if (in_cksum(m, hlen) != 0) 590 goto badcsum; 591 } 592 break; 593 } 594 595 /* Retrieve the packet length. */ 596 len = ntohs(ip->ip_len); 597 598 /* 599 * Check for additional length bogosity 600 */ 601 if (len < hlen) { 602 ipstat.ips_badlen++; 603 goto bad; 604 } 605 606 /* 607 * Check that the amount of data in the buffers 608 * is as at least much as the IP header would have us expect. 609 * Trim mbufs if longer than we expect. 610 * Drop packet if shorter than we expect. 611 */ 612 if (m->m_pkthdr.len < len) { 613 ipstat.ips_tooshort++; 614 goto bad; 615 } 616 if (m->m_pkthdr.len > len) { 617 if (m->m_len == m->m_pkthdr.len) { 618 m->m_len = len; 619 m->m_pkthdr.len = len; 620 } else 621 m_adj(m, len - m->m_pkthdr.len); 622 } 623 624#if defined(IPSEC) 625 /* ipflow (IP fast forwarding) is not compatible with IPsec. */ 626 m->m_flags &= ~M_CANFASTFWD; 627#else 628 /* 629 * Assume that we can create a fast-forward IP flow entry 630 * based on this packet. 631 */ 632 m->m_flags |= M_CANFASTFWD; 633#endif 634 635#ifdef PFIL_HOOKS 636 /* 637 * Run through list of hooks for input packets. If there are any 638 * filters which require that additional packets in the flow are 639 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 640 * Note that filters must _never_ set this flag, as another filter 641 * in the list may have previously cleared it. 642 */ 643 /* 644 * let ipfilter look at packet on the wire, 645 * not the decapsulated packet. 646 */ 647#ifdef IPSEC 648 if (!ipsec_getnhist(m)) 649#elif defined(FAST_IPSEC) 650 if (!ipsec_indone(m)) 651#else 652 if (1) 653#endif 654 { 655 struct in_addr odst; 656 657 odst = ip->ip_dst; 658 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 659 PFIL_IN) != 0) 660 return; 661 if (m == NULL) 662 return; 663 ip = mtod(m, struct ip *); 664 hlen = ip->ip_hl << 2; 665 /* 666 * XXX The setting of "srcrt" here is to prevent ip_forward() 667 * from generating ICMP redirects for packets that have 668 * been redirected by a hook back out on to the same LAN that 669 * they came from and is not an indication that the packet 670 * is being inffluenced by source routing options. This 671 * allows things like 672 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 673 * where tlp0 is both on the 1.1.1.0/24 network and is the 674 * default route for hosts on 1.1.1.0/24. Of course this 675 * also requires a "map tlp0 ..." to complete the story. 676 * One might argue whether or not this kind of network config. 677 * should be supported in this manner... 678 */ 679 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 680 } 681#endif /* PFIL_HOOKS */ 682 683#ifdef ALTQ 684 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 685 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 686 /* packet dropped by traffic conditioner */ 687 return; 688 } 689#endif 690 691 /* 692 * Process options and, if not destined for us, 693 * ship it on. ip_dooptions returns 1 when an 694 * error was detected (causing an icmp message 695 * to be sent and the original packet to be freed). 696 */ 697 ip_nhops = 0; /* for source routed packets */ 698 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 699 return; 700 701 /* 702 * Enable a consistency check between the destination address 703 * and the arrival interface for a unicast packet (the RFC 1122 704 * strong ES model) if IP forwarding is disabled and the packet 705 * is not locally generated. 706 * 707 * XXX - Checking also should be disabled if the destination 708 * address is ipnat'ed to a different interface. 709 * 710 * XXX - Checking is incompatible with IP aliases added 711 * to the loopback interface instead of the interface where 712 * the packets are received. 713 * 714 * XXX - We need to add a per ifaddr flag for this so that 715 * we get finer grain control. 716 */ 717 checkif = ip_checkinterface && (ipforwarding == 0) && 718 (m->m_pkthdr.rcvif != NULL) && 719 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0); 720 721 /* 722 * Check our list of addresses, to see if the packet is for us. 723 * 724 * Traditional 4.4BSD did not consult IFF_UP at all. 725 * The behavior here is to treat addresses on !IFF_UP interface 726 * as not mine. 727 */ 728 downmatch = 0; 729 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 730 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 731 if (checkif && ia->ia_ifp != m->m_pkthdr.rcvif) 732 continue; 733 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 734 break; 735 else 736 downmatch++; 737 } 738 } 739 if (ia != NULL) 740 goto ours; 741 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 742 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) { 743 if (ifa->ifa_addr->sa_family != AF_INET) 744 continue; 745 ia = ifatoia(ifa); 746 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 747 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 748 /* 749 * Look for all-0's host part (old broadcast addr), 750 * either for subnet or net. 751 */ 752 ip->ip_dst.s_addr == ia->ia_subnet || 753 ip->ip_dst.s_addr == ia->ia_net) 754 goto ours; 755 /* 756 * An interface with IP address zero accepts 757 * all packets that arrive on that interface. 758 */ 759 if (in_nullhost(ia->ia_addr.sin_addr)) 760 goto ours; 761 } 762 } 763 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 764 struct in_multi *inm; 765#ifdef MROUTING 766 extern struct socket *ip_mrouter; 767 768 if (ip_mrouter) { 769 /* 770 * If we are acting as a multicast router, all 771 * incoming multicast packets are passed to the 772 * kernel-level multicast forwarding function. 773 * The packet is returned (relatively) intact; if 774 * ip_mforward() returns a non-zero value, the packet 775 * must be discarded, else it may be accepted below. 776 * 777 * (The IP ident field is put in the same byte order 778 * as expected when ip_mforward() is called from 779 * ip_output().) 780 */ 781 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 782 ipstat.ips_cantforward++; 783 m_freem(m); 784 return; 785 } 786 787 /* 788 * The process-level routing demon needs to receive 789 * all multicast IGMP packets, whether or not this 790 * host belongs to their destination groups. 791 */ 792 if (ip->ip_p == IPPROTO_IGMP) 793 goto ours; 794 ipstat.ips_forward++; 795 } 796#endif 797 /* 798 * See if we belong to the destination multicast group on the 799 * arrival interface. 800 */ 801 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 802 if (inm == NULL) { 803 ipstat.ips_cantforward++; 804 m_freem(m); 805 return; 806 } 807 goto ours; 808 } 809 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 810 in_nullhost(ip->ip_dst)) 811 goto ours; 812 813 /* 814 * Not for us; forward if possible and desirable. 815 */ 816 if (ipforwarding == 0) { 817 ipstat.ips_cantforward++; 818 m_freem(m); 819 } else { 820 /* 821 * If ip_dst matched any of my address on !IFF_UP interface, 822 * and there's no IFF_UP interface that matches ip_dst, 823 * send icmp unreach. Forwarding it will result in in-kernel 824 * forwarding loop till TTL goes to 0. 825 */ 826 if (downmatch) { 827 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 828 ipstat.ips_cantforward++; 829 return; 830 } 831#ifdef IPSEC 832 if (ipsec4_in_reject(m, NULL)) { 833 ipsecstat.in_polvio++; 834 goto bad; 835 } 836#endif 837#ifdef FAST_IPSEC 838 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 839 s = splsoftnet(); 840 if (mtag != NULL) { 841 tdbi = (struct tdb_ident *)(mtag + 1); 842 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 843 } else { 844 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 845 IP_FORWARDING, &error); 846 } 847 if (sp == NULL) { /* NB: can happen if error */ 848 splx(s); 849 /*XXX error stat???*/ 850 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 851 goto bad; 852 } 853 854 /* 855 * Check security policy against packet attributes. 856 */ 857 error = ipsec_in_reject(sp, m); 858 KEY_FREESP(&sp); 859 splx(s); 860 if (error) { 861 ipstat.ips_cantforward++; 862 goto bad; 863 } 864 865 /* 866 * Peek at the outbound SP for this packet to determine if 867 * it's a Fast Forward candidate. 868 */ 869 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 870 if (mtag != NULL) 871 m->m_flags &= ~M_CANFASTFWD; 872 else { 873 s = splsoftnet(); 874 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, 875 (IP_FORWARDING | 876 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 877 &error, NULL); 878 if (sp != NULL) { 879 m->m_flags &= ~M_CANFASTFWD; 880 KEY_FREESP(&sp); 881 } 882 splx(s); 883 } 884#endif /* FAST_IPSEC */ 885 886 ip_forward(m, srcrt); 887 } 888 return; 889 890ours: 891 /* 892 * If offset or IP_MF are set, must reassemble. 893 * Otherwise, nothing need be done. 894 * (We could look in the reassembly queue to see 895 * if the packet was previously fragmented, 896 * but it's not worth the time; just let them time out.) 897 */ 898 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 899 uint16_t off; 900 /* 901 * Prevent TCP blind data attacks by not allowing non-initial 902 * fragments to start at less than 68 bytes (minimal fragment 903 * size) and making sure the first fragment is at least 68 904 * bytes. 905 */ 906 off = (ntohs(ip->ip_off) & IP_OFFMASK) << 3; 907 if ((off > 0 ? off + hlen : len) < IP_MINFRAGSIZE - 1) { 908 ipstat.ips_badfrags++; 909 goto bad; 910 } 911 /* 912 * Look for queue of fragments 913 * of this datagram. 914 */ 915 IPQ_LOCK(); 916 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 917 LIST_FOREACH(fp, &ipq[hash], ipq_q) { 918 if (ip->ip_id == fp->ipq_id && 919 in_hosteq(ip->ip_src, fp->ipq_src) && 920 in_hosteq(ip->ip_dst, fp->ipq_dst) && 921 ip->ip_p == fp->ipq_p) { 922 /* 923 * Make sure the TOS is matches previous 924 * fragments. 925 */ 926 if (ip->ip_tos != fp->ipq_tos) { 927 ipstat.ips_badfrags++; 928 goto bad; 929 } 930 goto found; 931 } 932 } 933 fp = 0; 934found: 935 936 /* 937 * Adjust ip_len to not reflect header, 938 * set ipqe_mff if more fragments are expected, 939 * convert offset of this to bytes. 940 */ 941 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 942 mff = (ip->ip_off & htons(IP_MF)) != 0; 943 if (mff) { 944 /* 945 * Make sure that fragments have a data length 946 * that's a non-zero multiple of 8 bytes. 947 */ 948 if (ntohs(ip->ip_len) == 0 || 949 (ntohs(ip->ip_len) & 0x7) != 0) { 950 ipstat.ips_badfrags++; 951 IPQ_UNLOCK(); 952 goto bad; 953 } 954 } 955 ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3); 956 957 /* 958 * If datagram marked as having more fragments 959 * or if this is not the first fragment, 960 * attempt reassembly; if it succeeds, proceed. 961 */ 962 if (mff || ip->ip_off != htons(0)) { 963 ipstat.ips_fragments++; 964 s = splvm(); 965 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 966 splx(s); 967 if (ipqe == NULL) { 968 ipstat.ips_rcvmemdrop++; 969 IPQ_UNLOCK(); 970 goto bad; 971 } 972 ipqe->ipqe_mff = mff; 973 ipqe->ipqe_m = m; 974 ipqe->ipqe_ip = ip; 975 m = ip_reass(ipqe, fp, &ipq[hash]); 976 if (m == 0) { 977 IPQ_UNLOCK(); 978 return; 979 } 980 ipstat.ips_reassembled++; 981 ip = mtod(m, struct ip *); 982 hlen = ip->ip_hl << 2; 983 ip->ip_len = htons(ntohs(ip->ip_len) + hlen); 984 } else 985 if (fp) 986 ip_freef(fp); 987 IPQ_UNLOCK(); 988 } 989 990#if defined(IPSEC) 991 /* 992 * enforce IPsec policy checking if we are seeing last header. 993 * note that we do not visit this with protocols with pcb layer 994 * code - like udp/tcp/raw ip. 995 */ 996 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 997 ipsec4_in_reject(m, NULL)) { 998 ipsecstat.in_polvio++; 999 goto bad; 1000 } 1001#endif 1002#ifdef FAST_IPSEC 1003 /* 1004 * enforce IPsec policy checking if we are seeing last header. 1005 * note that we do not visit this with protocols with pcb layer 1006 * code - like udp/tcp/raw ip. 1007 */ 1008 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 1009 /* 1010 * Check if the packet has already had IPsec processing 1011 * done. If so, then just pass it along. This tag gets 1012 * set during AH, ESP, etc. input handling, before the 1013 * packet is returned to the ip input queue for delivery. 1014 */ 1015 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 1016 s = splsoftnet(); 1017 if (mtag != NULL) { 1018 tdbi = (struct tdb_ident *)(mtag + 1); 1019 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 1020 } else { 1021 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 1022 IP_FORWARDING, &error); 1023 } 1024 if (sp != NULL) { 1025 /* 1026 * Check security policy against packet attributes. 1027 */ 1028 error = ipsec_in_reject(sp, m); 1029 KEY_FREESP(&sp); 1030 } else { 1031 /* XXX error stat??? */ 1032 error = EINVAL; 1033DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 1034 } 1035 splx(s); 1036 if (error) 1037 goto bad; 1038 } 1039#endif /* FAST_IPSEC */ 1040 1041 /* 1042 * Switch out to protocol's input routine. 1043 */ 1044#if IFA_STATS 1045 if (ia && ip) 1046 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 1047#endif 1048 ipstat.ips_delivered++; 1049 { 1050 int off = hlen, nh = ip->ip_p; 1051 1052 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 1053 return; 1054 } 1055bad: 1056 m_freem(m); 1057 return; 1058 1059badcsum: 1060 ipstat.ips_badsum++; 1061 m_freem(m); 1062} 1063 1064/* 1065 * Take incoming datagram fragment and try to 1066 * reassemble it into whole datagram. If a chain for 1067 * reassembly of this datagram already exists, then it 1068 * is given as fp; otherwise have to make a chain. 1069 */ 1070struct mbuf * 1071ip_reass(struct ipqent *ipqe, struct ipq *fp, struct ipqhead *ipqhead) 1072{ 1073 struct mbuf *m = ipqe->ipqe_m; 1074 struct ipqent *nq, *p, *q; 1075 struct ip *ip; 1076 struct mbuf *t; 1077 int hlen = ipqe->ipqe_ip->ip_hl << 2; 1078 int i, next, s; 1079 1080 IPQ_LOCK_CHECK(); 1081 1082 /* 1083 * Presence of header sizes in mbufs 1084 * would confuse code below. 1085 */ 1086 m->m_data += hlen; 1087 m->m_len -= hlen; 1088 1089#ifdef notyet 1090 /* make sure fragment limit is up-to-date */ 1091 CHECK_NMBCLUSTER_PARAMS(); 1092 1093 /* If we have too many fragments, drop the older half. */ 1094 if (ip_nfrags >= ip_maxfrags) 1095 ip_reass_drophalf(void); 1096#endif 1097 1098 /* 1099 * We are about to add a fragment; increment frag count. 1100 */ 1101 ip_nfrags++; 1102 1103 /* 1104 * If first fragment to arrive, create a reassembly queue. 1105 */ 1106 if (fp == 0) { 1107 /* 1108 * Enforce upper bound on number of fragmented packets 1109 * for which we attempt reassembly; 1110 * If maxfrag is 0, never accept fragments. 1111 * If maxfrag is -1, accept all fragments without limitation. 1112 */ 1113 if (ip_maxfragpackets < 0) 1114 ; 1115 else if (ip_nfragpackets >= ip_maxfragpackets) 1116 goto dropfrag; 1117 ip_nfragpackets++; 1118 MALLOC(fp, struct ipq *, sizeof (struct ipq), 1119 M_FTABLE, M_NOWAIT); 1120 if (fp == NULL) 1121 goto dropfrag; 1122 LIST_INSERT_HEAD(ipqhead, fp, ipq_q); 1123 fp->ipq_nfrags = 1; 1124 fp->ipq_ttl = IPFRAGTTL; 1125 fp->ipq_p = ipqe->ipqe_ip->ip_p; 1126 fp->ipq_id = ipqe->ipqe_ip->ip_id; 1127 fp->ipq_tos = ipqe->ipqe_ip->ip_tos; 1128 TAILQ_INIT(&fp->ipq_fragq); 1129 fp->ipq_src = ipqe->ipqe_ip->ip_src; 1130 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 1131 p = NULL; 1132 goto insert; 1133 } else { 1134 fp->ipq_nfrags++; 1135 } 1136 1137 /* 1138 * Find a segment which begins after this one does. 1139 */ 1140 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1141 p = q, q = TAILQ_NEXT(q, ipqe_q)) 1142 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off)) 1143 break; 1144 1145 /* 1146 * If there is a preceding segment, it may provide some of 1147 * our data already. If so, drop the data from the incoming 1148 * segment. If it provides all of our data, drop us. 1149 */ 1150 if (p != NULL) { 1151 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) - 1152 ntohs(ipqe->ipqe_ip->ip_off); 1153 if (i > 0) { 1154 if (i >= ntohs(ipqe->ipqe_ip->ip_len)) 1155 goto dropfrag; 1156 m_adj(ipqe->ipqe_m, i); 1157 ipqe->ipqe_ip->ip_off = 1158 htons(ntohs(ipqe->ipqe_ip->ip_off) + i); 1159 ipqe->ipqe_ip->ip_len = 1160 htons(ntohs(ipqe->ipqe_ip->ip_len) - i); 1161 } 1162 } 1163 1164 /* 1165 * While we overlap succeeding segments trim them or, 1166 * if they are completely covered, dequeue them. 1167 */ 1168 for (; q != NULL && 1169 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) > 1170 ntohs(q->ipqe_ip->ip_off); q = nq) { 1171 i = (ntohs(ipqe->ipqe_ip->ip_off) + 1172 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off); 1173 if (i < ntohs(q->ipqe_ip->ip_len)) { 1174 q->ipqe_ip->ip_len = 1175 htons(ntohs(q->ipqe_ip->ip_len) - i); 1176 q->ipqe_ip->ip_off = 1177 htons(ntohs(q->ipqe_ip->ip_off) + i); 1178 m_adj(q->ipqe_m, i); 1179 break; 1180 } 1181 nq = TAILQ_NEXT(q, ipqe_q); 1182 m_freem(q->ipqe_m); 1183 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1184 s = splvm(); 1185 pool_put(&ipqent_pool, q); 1186 splx(s); 1187 fp->ipq_nfrags--; 1188 ip_nfrags--; 1189 } 1190 1191insert: 1192 /* 1193 * Stick new segment in its place; 1194 * check for complete reassembly. 1195 */ 1196 if (p == NULL) { 1197 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 1198 } else { 1199 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q); 1200 } 1201 next = 0; 1202 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1203 p = q, q = TAILQ_NEXT(q, ipqe_q)) { 1204 if (ntohs(q->ipqe_ip->ip_off) != next) 1205 return (0); 1206 next += ntohs(q->ipqe_ip->ip_len); 1207 } 1208 if (p->ipqe_mff) 1209 return (0); 1210 1211 /* 1212 * Reassembly is complete. Check for a bogus message size and 1213 * concatenate fragments. 1214 */ 1215 q = TAILQ_FIRST(&fp->ipq_fragq); 1216 ip = q->ipqe_ip; 1217 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 1218 ipstat.ips_toolong++; 1219 ip_freef(fp); 1220 return (0); 1221 } 1222 m = q->ipqe_m; 1223 t = m->m_next; 1224 m->m_next = 0; 1225 m_cat(m, t); 1226 nq = TAILQ_NEXT(q, ipqe_q); 1227 s = splvm(); 1228 pool_put(&ipqent_pool, q); 1229 splx(s); 1230 for (q = nq; q != NULL; q = nq) { 1231 t = q->ipqe_m; 1232 nq = TAILQ_NEXT(q, ipqe_q); 1233 s = splvm(); 1234 pool_put(&ipqent_pool, q); 1235 splx(s); 1236 m_cat(m, t); 1237 } 1238 ip_nfrags -= fp->ipq_nfrags; 1239 1240 /* 1241 * Create header for new ip packet by 1242 * modifying header of first packet; 1243 * dequeue and discard fragment reassembly header. 1244 * Make header visible. 1245 */ 1246 ip->ip_len = htons(next); 1247 ip->ip_src = fp->ipq_src; 1248 ip->ip_dst = fp->ipq_dst; 1249 LIST_REMOVE(fp, ipq_q); 1250 FREE(fp, M_FTABLE); 1251 ip_nfragpackets--; 1252 m->m_len += (ip->ip_hl << 2); 1253 m->m_data -= (ip->ip_hl << 2); 1254 /* some debugging cruft by sklower, below, will go away soon */ 1255 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1256 int plen = 0; 1257 for (t = m; t; t = t->m_next) 1258 plen += t->m_len; 1259 m->m_pkthdr.len = plen; 1260 m->m_pkthdr.csum_flags = 0; 1261 } 1262 return (m); 1263 1264dropfrag: 1265 if (fp != 0) 1266 fp->ipq_nfrags--; 1267 ip_nfrags--; 1268 ipstat.ips_fragdropped++; 1269 m_freem(m); 1270 s = splvm(); 1271 pool_put(&ipqent_pool, ipqe); 1272 splx(s); 1273 return (0); 1274} 1275 1276/* 1277 * Free a fragment reassembly header and all 1278 * associated datagrams. 1279 */ 1280void 1281ip_freef(struct ipq *fp) 1282{ 1283 struct ipqent *q, *p; 1284 u_int nfrags = 0; 1285 int s; 1286 1287 IPQ_LOCK_CHECK(); 1288 1289 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) { 1290 p = TAILQ_NEXT(q, ipqe_q); 1291 m_freem(q->ipqe_m); 1292 nfrags++; 1293 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1294 s = splvm(); 1295 pool_put(&ipqent_pool, q); 1296 splx(s); 1297 } 1298 1299 if (nfrags != fp->ipq_nfrags) 1300 printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags); 1301 ip_nfrags -= nfrags; 1302 LIST_REMOVE(fp, ipq_q); 1303 FREE(fp, M_FTABLE); 1304 ip_nfragpackets--; 1305} 1306 1307/* 1308 * IP reassembly TTL machinery for multiplicative drop. 1309 */ 1310static u_int fragttl_histo[(IPFRAGTTL+1)]; 1311 1312 1313/* 1314 * Decrement TTL of all reasembly queue entries by `ticks'. 1315 * Count number of distinct fragments (as opposed to partial, fragmented 1316 * datagrams) in the reassembly queue. While we traverse the entire 1317 * reassembly queue, compute and return the median TTL over all fragments. 1318 */ 1319static u_int 1320ip_reass_ttl_decr(u_int ticks) 1321{ 1322 u_int nfrags, median, dropfraction, keepfraction; 1323 struct ipq *fp, *nfp; 1324 int i; 1325 1326 nfrags = 0; 1327 memset(fragttl_histo, 0, sizeof fragttl_histo); 1328 1329 for (i = 0; i < IPREASS_NHASH; i++) { 1330 for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) { 1331 fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ? 1332 0 : fp->ipq_ttl - ticks); 1333 nfp = LIST_NEXT(fp, ipq_q); 1334 if (fp->ipq_ttl == 0) { 1335 ipstat.ips_fragtimeout++; 1336 ip_freef(fp); 1337 } else { 1338 nfrags += fp->ipq_nfrags; 1339 fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags; 1340 } 1341 } 1342 } 1343 1344 KASSERT(ip_nfrags == nfrags); 1345 1346 /* Find median (or other drop fraction) in histogram. */ 1347 dropfraction = (ip_nfrags / 2); 1348 keepfraction = ip_nfrags - dropfraction; 1349 for (i = IPFRAGTTL, median = 0; i >= 0; i--) { 1350 median += fragttl_histo[i]; 1351 if (median >= keepfraction) 1352 break; 1353 } 1354 1355 /* Return TTL of median (or other fraction). */ 1356 return (u_int)i; 1357} 1358 1359void 1360ip_reass_drophalf(void) 1361{ 1362 1363 u_int median_ticks; 1364 /* 1365 * Compute median TTL of all fragments, and count frags 1366 * with that TTL or lower (roughly half of all fragments). 1367 */ 1368 median_ticks = ip_reass_ttl_decr(0); 1369 1370 /* Drop half. */ 1371 median_ticks = ip_reass_ttl_decr(median_ticks); 1372 1373} 1374 1375/* 1376 * IP timer processing; 1377 * if a timer expires on a reassembly 1378 * queue, discard it. 1379 */ 1380void 1381ip_slowtimo(void) 1382{ 1383 static u_int dropscanidx = 0; 1384 u_int i; 1385 u_int median_ttl; 1386 int s = splsoftnet(); 1387 1388 IPQ_LOCK(); 1389 1390 /* Age TTL of all fragments by 1 tick .*/ 1391 median_ttl = ip_reass_ttl_decr(1); 1392 1393 /* make sure fragment limit is up-to-date */ 1394 CHECK_NMBCLUSTER_PARAMS(); 1395 1396 /* If we have too many fragments, drop the older half. */ 1397 if (ip_nfrags > ip_maxfrags) 1398 ip_reass_ttl_decr(median_ttl); 1399 1400 /* 1401 * If we are over the maximum number of fragmented packets 1402 * (due to the limit being lowered), drain off 1403 * enough to get down to the new limit. Start draining 1404 * from the reassembly hashqueue most recently drained. 1405 */ 1406 if (ip_maxfragpackets < 0) 1407 ; 1408 else { 1409 int wrapped = 0; 1410 1411 i = dropscanidx; 1412 while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) { 1413 while (LIST_FIRST(&ipq[i]) != NULL) 1414 ip_freef(LIST_FIRST(&ipq[i])); 1415 if (++i >= IPREASS_NHASH) { 1416 i = 0; 1417 } 1418 /* 1419 * Dont scan forever even if fragment counters are 1420 * wrong: stop after scanning entire reassembly queue. 1421 */ 1422 if (i == dropscanidx) 1423 wrapped = 1; 1424 } 1425 dropscanidx = i; 1426 } 1427 IPQ_UNLOCK(); 1428 splx(s); 1429} 1430 1431/* 1432 * Drain off all datagram fragments. 1433 */ 1434void 1435ip_drain(void) 1436{ 1437 1438 /* 1439 * We may be called from a device's interrupt context. If 1440 * the ipq is already busy, just bail out now. 1441 */ 1442 if (ipq_lock_try() == 0) 1443 return; 1444 1445 /* 1446 * Drop half the total fragments now. If more mbufs are needed, 1447 * we will be called again soon. 1448 */ 1449 ip_reass_drophalf(); 1450 1451 IPQ_UNLOCK(); 1452} 1453 1454/* 1455 * Do option processing on a datagram, 1456 * possibly discarding it if bad options are encountered, 1457 * or forwarding it if source-routed. 1458 * Returns 1 if packet has been forwarded/freed, 1459 * 0 if the packet should be processed further. 1460 */ 1461int 1462ip_dooptions(struct mbuf *m) 1463{ 1464 struct ip *ip = mtod(m, struct ip *); 1465 u_char *cp, *cp0; 1466 struct ip_timestamp *ipt; 1467 struct in_ifaddr *ia; 1468 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1469 struct in_addr dst; 1470 n_time ntime; 1471 1472 dst = ip->ip_dst; 1473 cp = (u_char *)(ip + 1); 1474 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1475 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1476 opt = cp[IPOPT_OPTVAL]; 1477 if (opt == IPOPT_EOL) 1478 break; 1479 if (opt == IPOPT_NOP) 1480 optlen = 1; 1481 else { 1482 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1483 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1484 goto bad; 1485 } 1486 optlen = cp[IPOPT_OLEN]; 1487 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1488 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1489 goto bad; 1490 } 1491 } 1492 switch (opt) { 1493 1494 default: 1495 break; 1496 1497 /* 1498 * Source routing with record. 1499 * Find interface with current destination address. 1500 * If none on this machine then drop if strictly routed, 1501 * or do nothing if loosely routed. 1502 * Record interface address and bring up next address 1503 * component. If strictly routed make sure next 1504 * address is on directly accessible net. 1505 */ 1506 case IPOPT_LSRR: 1507 case IPOPT_SSRR: 1508 if (ip_allowsrcrt == 0) { 1509 type = ICMP_UNREACH; 1510 code = ICMP_UNREACH_NET_PROHIB; 1511 goto bad; 1512 } 1513 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1514 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1515 goto bad; 1516 } 1517 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1518 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1519 goto bad; 1520 } 1521 ipaddr.sin_addr = ip->ip_dst; 1522 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1523 if (ia == 0) { 1524 if (opt == IPOPT_SSRR) { 1525 type = ICMP_UNREACH; 1526 code = ICMP_UNREACH_SRCFAIL; 1527 goto bad; 1528 } 1529 /* 1530 * Loose routing, and not at next destination 1531 * yet; nothing to do except forward. 1532 */ 1533 break; 1534 } 1535 off--; /* 0 origin */ 1536 if ((off + sizeof(struct in_addr)) > optlen) { 1537 /* 1538 * End of source route. Should be for us. 1539 */ 1540 save_rte(cp, ip->ip_src); 1541 break; 1542 } 1543 /* 1544 * locate outgoing interface 1545 */ 1546 bcopy((void *)(cp + off), (void *)&ipaddr.sin_addr, 1547 sizeof(ipaddr.sin_addr)); 1548 if (opt == IPOPT_SSRR) 1549 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); 1550 else 1551 ia = ip_rtaddr(ipaddr.sin_addr); 1552 if (ia == 0) { 1553 type = ICMP_UNREACH; 1554 code = ICMP_UNREACH_SRCFAIL; 1555 goto bad; 1556 } 1557 ip->ip_dst = ipaddr.sin_addr; 1558 bcopy((void *)&ia->ia_addr.sin_addr, 1559 (void *)(cp + off), sizeof(struct in_addr)); 1560 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1561 /* 1562 * Let ip_intr's mcast routing check handle mcast pkts 1563 */ 1564 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1565 break; 1566 1567 case IPOPT_RR: 1568 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1569 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1570 goto bad; 1571 } 1572 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1573 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1574 goto bad; 1575 } 1576 /* 1577 * If no space remains, ignore. 1578 */ 1579 off--; /* 0 origin */ 1580 if ((off + sizeof(struct in_addr)) > optlen) 1581 break; 1582 bcopy((void *)(&ip->ip_dst), (void *)&ipaddr.sin_addr, 1583 sizeof(ipaddr.sin_addr)); 1584 /* 1585 * locate outgoing interface; if we're the destination, 1586 * use the incoming interface (should be same). 1587 */ 1588 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1589 == NULL && 1590 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1591 type = ICMP_UNREACH; 1592 code = ICMP_UNREACH_HOST; 1593 goto bad; 1594 } 1595 bcopy((void *)&ia->ia_addr.sin_addr, 1596 (void *)(cp + off), sizeof(struct in_addr)); 1597 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1598 break; 1599 1600 case IPOPT_TS: 1601 code = cp - (u_char *)ip; 1602 ipt = (struct ip_timestamp *)cp; 1603 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1604 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1605 goto bad; 1606 } 1607 if (ipt->ipt_ptr < 5) { 1608 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1609 goto bad; 1610 } 1611 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1612 if (++ipt->ipt_oflw == 0) { 1613 code = (u_char *)&ipt->ipt_ptr - 1614 (u_char *)ip; 1615 goto bad; 1616 } 1617 break; 1618 } 1619 cp0 = (cp + ipt->ipt_ptr - 1); 1620 switch (ipt->ipt_flg) { 1621 1622 case IPOPT_TS_TSONLY: 1623 break; 1624 1625 case IPOPT_TS_TSANDADDR: 1626 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1627 sizeof(struct in_addr) > ipt->ipt_len) { 1628 code = (u_char *)&ipt->ipt_ptr - 1629 (u_char *)ip; 1630 goto bad; 1631 } 1632 ipaddr.sin_addr = dst; 1633 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1634 m->m_pkthdr.rcvif)); 1635 if (ia == 0) 1636 continue; 1637 bcopy(&ia->ia_addr.sin_addr, 1638 cp0, sizeof(struct in_addr)); 1639 ipt->ipt_ptr += sizeof(struct in_addr); 1640 break; 1641 1642 case IPOPT_TS_PRESPEC: 1643 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1644 sizeof(struct in_addr) > ipt->ipt_len) { 1645 code = (u_char *)&ipt->ipt_ptr - 1646 (u_char *)ip; 1647 goto bad; 1648 } 1649 bcopy(cp0, &ipaddr.sin_addr, 1650 sizeof(struct in_addr)); 1651 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1652 == NULL) 1653 continue; 1654 ipt->ipt_ptr += sizeof(struct in_addr); 1655 break; 1656 1657 default: 1658 /* XXX can't take &ipt->ipt_flg */ 1659 code = (u_char *)&ipt->ipt_ptr - 1660 (u_char *)ip + 1; 1661 goto bad; 1662 } 1663 ntime = iptime(); 1664 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1665 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1666 sizeof(n_time)); 1667 ipt->ipt_ptr += sizeof(n_time); 1668 } 1669 } 1670 if (forward) { 1671 if (ip_forwsrcrt == 0) { 1672 type = ICMP_UNREACH; 1673 code = ICMP_UNREACH_SRCFAIL; 1674 goto bad; 1675 } 1676 ip_forward(m, 1); 1677 return (1); 1678 } 1679 return (0); 1680bad: 1681 icmp_error(m, type, code, 0, 0); 1682 ipstat.ips_badoptions++; 1683 return (1); 1684} 1685 1686/* 1687 * Given address of next destination (final or next hop), 1688 * return internet address info of interface to be used to get there. 1689 */ 1690struct in_ifaddr * 1691ip_rtaddr(struct in_addr dst) 1692{ 1693 struct rtentry *rt; 1694 union { 1695 struct sockaddr dst; 1696 struct sockaddr_in dst4; 1697 } u; 1698 1699 sockaddr_in_init(&u.dst4, &dst, 0); 1700 1701 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) 1702 return NULL; 1703 1704 return ifatoia(rt->rt_ifa); 1705} 1706 1707/* 1708 * Save incoming source route for use in replies, 1709 * to be picked up later by ip_srcroute if the receiver is interested. 1710 */ 1711void 1712save_rte(u_char *option, struct in_addr dst) 1713{ 1714 unsigned olen; 1715 1716 olen = option[IPOPT_OLEN]; 1717#ifdef DIAGNOSTIC 1718 if (ipprintfs) 1719 printf("save_rte: olen %d\n", olen); 1720#endif /* 0 */ 1721 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1722 return; 1723 bcopy((void *)option, (void *)ip_srcrt.srcopt, olen); 1724 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1725 ip_srcrt.dst = dst; 1726} 1727 1728/* 1729 * Retrieve incoming source route for use in replies, 1730 * in the same form used by setsockopt. 1731 * The first hop is placed before the options, will be removed later. 1732 */ 1733struct mbuf * 1734ip_srcroute(void) 1735{ 1736 struct in_addr *p, *q; 1737 struct mbuf *m; 1738 1739 if (ip_nhops == 0) 1740 return NULL; 1741 m = m_get(M_DONTWAIT, MT_SOOPTS); 1742 if (m == 0) 1743 return NULL; 1744 1745 MCLAIM(m, &inetdomain.dom_mowner); 1746#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1747 1748 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1749 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1750 OPTSIZ; 1751#ifdef DIAGNOSTIC 1752 if (ipprintfs) 1753 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1754#endif 1755 1756 /* 1757 * First save first hop for return route 1758 */ 1759 p = &ip_srcrt.route[ip_nhops - 1]; 1760 *(mtod(m, struct in_addr *)) = *p--; 1761#ifdef DIAGNOSTIC 1762 if (ipprintfs) 1763 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1764#endif 1765 1766 /* 1767 * Copy option fields and padding (nop) to mbuf. 1768 */ 1769 ip_srcrt.nop = IPOPT_NOP; 1770 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1771 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop, 1772 OPTSIZ); 1773 q = (struct in_addr *)(mtod(m, char *) + 1774 sizeof(struct in_addr) + OPTSIZ); 1775#undef OPTSIZ 1776 /* 1777 * Record return path as an IP source route, 1778 * reversing the path (pointers are now aligned). 1779 */ 1780 while (p >= ip_srcrt.route) { 1781#ifdef DIAGNOSTIC 1782 if (ipprintfs) 1783 printf(" %x", ntohl(q->s_addr)); 1784#endif 1785 *q++ = *p--; 1786 } 1787 /* 1788 * Last hop goes to final destination. 1789 */ 1790 *q = ip_srcrt.dst; 1791#ifdef DIAGNOSTIC 1792 if (ipprintfs) 1793 printf(" %x\n", ntohl(q->s_addr)); 1794#endif 1795 return (m); 1796} 1797 1798const int inetctlerrmap[PRC_NCMDS] = { 1799 [PRC_MSGSIZE] = EMSGSIZE, 1800 [PRC_HOSTDEAD] = EHOSTDOWN, 1801 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1802 [PRC_UNREACH_NET] = EHOSTUNREACH, 1803 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1804 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1805 [PRC_UNREACH_PORT] = ECONNREFUSED, 1806 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1807 [PRC_PARAMPROB] = ENOPROTOOPT, 1808}; 1809 1810/* 1811 * Forward a packet. If some error occurs return the sender 1812 * an icmp packet. Note we can't always generate a meaningful 1813 * icmp message because icmp doesn't have a large enough repertoire 1814 * of codes and types. 1815 * 1816 * If not forwarding, just drop the packet. This could be confusing 1817 * if ipforwarding was zero but some routing protocol was advancing 1818 * us as a gateway to somewhere. However, we must let the routing 1819 * protocol deal with that. 1820 * 1821 * The srcrt parameter indicates whether the packet is being forwarded 1822 * via a source route. 1823 */ 1824void 1825ip_forward(struct mbuf *m, int srcrt) 1826{ 1827 struct ip *ip = mtod(m, struct ip *); 1828 struct rtentry *rt; 1829 int error, type = 0, code = 0, destmtu = 0; 1830 struct mbuf *mcopy; 1831 n_long dest; 1832 union { 1833 struct sockaddr dst; 1834 struct sockaddr_in dst4; 1835 } u; 1836 1837 /* 1838 * We are now in the output path. 1839 */ 1840 MCLAIM(m, &ip_tx_mowner); 1841 1842 /* 1843 * Clear any in-bound checksum flags for this packet. 1844 */ 1845 m->m_pkthdr.csum_flags = 0; 1846 1847 dest = 0; 1848#ifdef DIAGNOSTIC 1849 if (ipprintfs) { 1850 printf("forward: src %s ", inet_ntoa(ip->ip_src)); 1851 printf("dst %s ttl %x\n", inet_ntoa(ip->ip_dst), ip->ip_ttl); 1852 } 1853#endif 1854 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1855 ipstat.ips_cantforward++; 1856 m_freem(m); 1857 return; 1858 } 1859 if (ip->ip_ttl <= IPTTLDEC) { 1860 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1861 return; 1862 } 1863 1864 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1865 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) { 1866 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1867 return; 1868 } 1869 1870 /* 1871 * Save at most 68 bytes of the packet in case 1872 * we need to generate an ICMP message to the src. 1873 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1874 */ 1875 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1876 if (mcopy) 1877 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1878 1879 ip->ip_ttl -= IPTTLDEC; 1880 1881 /* 1882 * If forwarding packet using same interface that it came in on, 1883 * perhaps should send a redirect to sender to shortcut a hop. 1884 * Only send redirect if source is sending directly to us, 1885 * and if packet was not source routed (or has any options). 1886 * Also, don't send redirect if forwarding using a default route 1887 * or a route modified by a redirect. 1888 */ 1889 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1890 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1891 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1892 ipsendredirects && !srcrt) { 1893 if (rt->rt_ifa && 1894 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1895 ifatoia(rt->rt_ifa)->ia_subnet) { 1896 if (rt->rt_flags & RTF_GATEWAY) 1897 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1898 else 1899 dest = ip->ip_dst.s_addr; 1900 /* 1901 * Router requirements says to only send host 1902 * redirects. 1903 */ 1904 type = ICMP_REDIRECT; 1905 code = ICMP_REDIRECT_HOST; 1906#ifdef DIAGNOSTIC 1907 if (ipprintfs) 1908 printf("redirect (%d) to %x\n", code, 1909 (u_int32_t)dest); 1910#endif 1911 } 1912 } 1913 1914 error = ip_output(m, NULL, &ipforward_rt, 1915 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1916 (struct ip_moptions *)NULL, (struct socket *)NULL); 1917 1918 if (error) 1919 ipstat.ips_cantforward++; 1920 else { 1921 ipstat.ips_forward++; 1922 if (type) 1923 ipstat.ips_redirectsent++; 1924 else { 1925 if (mcopy) { 1926#ifdef GATEWAY 1927 if (mcopy->m_flags & M_CANFASTFWD) 1928 ipflow_create(&ipforward_rt, mcopy); 1929#endif 1930 m_freem(mcopy); 1931 } 1932 return; 1933 } 1934 } 1935 if (mcopy == NULL) 1936 return; 1937 1938 switch (error) { 1939 1940 case 0: /* forwarded, but need redirect */ 1941 /* type, code set above */ 1942 break; 1943 1944 case ENETUNREACH: /* shouldn't happen, checked above */ 1945 case EHOSTUNREACH: 1946 case ENETDOWN: 1947 case EHOSTDOWN: 1948 default: 1949 type = ICMP_UNREACH; 1950 code = ICMP_UNREACH_HOST; 1951 break; 1952 1953 case EMSGSIZE: 1954 type = ICMP_UNREACH; 1955 code = ICMP_UNREACH_NEEDFRAG; 1956#if !defined(IPSEC) && !defined(FAST_IPSEC) 1957 if ((rt = rtcache_getrt(&ipforward_rt)) != NULL) 1958 destmtu = rt->rt_ifp->if_mtu; 1959#else 1960 /* 1961 * If the packet is routed over IPsec tunnel, tell the 1962 * originator the tunnel MTU. 1963 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1964 * XXX quickhack!!! 1965 */ 1966 if ((rt = rtcache_getrt(&ipforward_rt)) != NULL) { 1967 struct secpolicy *sp; 1968 int ipsecerror; 1969 size_t ipsechdr; 1970 struct route *ro; 1971 1972 sp = ipsec4_getpolicybyaddr(mcopy, 1973 IPSEC_DIR_OUTBOUND, IP_FORWARDING, 1974 &ipsecerror); 1975 1976 if (sp == NULL) 1977 destmtu = rt->rt_ifp->if_mtu; 1978 else { 1979 /* count IPsec header size */ 1980 ipsechdr = ipsec4_hdrsiz(mcopy, 1981 IPSEC_DIR_OUTBOUND, NULL); 1982 1983 /* 1984 * find the correct route for outer IPv4 1985 * header, compute tunnel MTU. 1986 */ 1987 1988 if (sp->req != NULL 1989 && sp->req->sav != NULL 1990 && sp->req->sav->sah != NULL) { 1991 ro = &sp->req->sav->sah->sa_route; 1992 if (rt && rt->rt_ifp) { 1993 destmtu = 1994 rt->rt_rmx.rmx_mtu ? 1995 rt->rt_rmx.rmx_mtu : 1996 rt->rt_ifp->if_mtu; 1997 destmtu -= ipsechdr; 1998 } 1999 } 2000 2001#ifdef IPSEC 2002 key_freesp(sp); 2003#else 2004 KEY_FREESP(&sp); 2005#endif 2006 } 2007 } 2008#endif /*IPSEC*/ 2009 ipstat.ips_cantfrag++; 2010 break; 2011 2012 case ENOBUFS: 2013#if 1 2014 /* 2015 * a router should not generate ICMP_SOURCEQUENCH as 2016 * required in RFC1812 Requirements for IP Version 4 Routers. 2017 * source quench could be a big problem under DoS attacks, 2018 * or if the underlying interface is rate-limited. 2019 */ 2020 if (mcopy) 2021 m_freem(mcopy); 2022 return; 2023#else 2024 type = ICMP_SOURCEQUENCH; 2025 code = 0; 2026 break; 2027#endif 2028 } 2029 icmp_error(mcopy, type, code, dest, destmtu); 2030} 2031 2032void 2033ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2034 struct mbuf *m) 2035{ 2036 2037 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2038 struct timeval tv; 2039 2040 microtime(&tv); 2041 *mp = sbcreatecontrol((void *) &tv, sizeof(tv), 2042 SCM_TIMESTAMP, SOL_SOCKET); 2043 if (*mp) 2044 mp = &(*mp)->m_next; 2045 } 2046 if (inp->inp_flags & INP_RECVDSTADDR) { 2047 *mp = sbcreatecontrol((void *) &ip->ip_dst, 2048 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2049 if (*mp) 2050 mp = &(*mp)->m_next; 2051 } 2052#ifdef notyet 2053 /* 2054 * XXX 2055 * Moving these out of udp_input() made them even more broken 2056 * than they already were. 2057 * - fenner@parc.xerox.com 2058 */ 2059 /* options were tossed already */ 2060 if (inp->inp_flags & INP_RECVOPTS) { 2061 *mp = sbcreatecontrol((void *) opts_deleted_above, 2062 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2063 if (*mp) 2064 mp = &(*mp)->m_next; 2065 } 2066 /* ip_srcroute doesn't do what we want here, need to fix */ 2067 if (inp->inp_flags & INP_RECVRETOPTS) { 2068 *mp = sbcreatecontrol((void *) ip_srcroute(), 2069 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2070 if (*mp) 2071 mp = &(*mp)->m_next; 2072 } 2073#endif 2074 if (inp->inp_flags & INP_RECVIF) { 2075 struct sockaddr_dl sdl; 2076 2077 sockaddr_dl_init(&sdl, sizeof(sdl), 2078 (m->m_pkthdr.rcvif != NULL) 2079 ? m->m_pkthdr.rcvif->if_index 2080 : 0, 2081 0, NULL, 0, NULL, 0); 2082 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 2083 if (*mp) 2084 mp = &(*mp)->m_next; 2085 } 2086} 2087 2088/* 2089 * sysctl helper routine for net.inet.ip.forwsrcrt. 2090 */ 2091static int 2092sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 2093{ 2094 int error, tmp; 2095 struct sysctlnode node; 2096 2097 node = *rnode; 2098 tmp = ip_forwsrcrt; 2099 node.sysctl_data = &tmp; 2100 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2101 if (error || newp == NULL) 2102 return (error); 2103 2104 if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 2105 0, NULL, NULL, NULL)) 2106 return (EPERM); 2107 2108 ip_forwsrcrt = tmp; 2109 2110 return (0); 2111} 2112 2113/* 2114 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 2115 * range of the new value and tweaks timers if it changes. 2116 */ 2117static int 2118sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 2119{ 2120 int error, tmp; 2121 struct sysctlnode node; 2122 2123 node = *rnode; 2124 tmp = ip_mtudisc_timeout; 2125 node.sysctl_data = &tmp; 2126 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2127 if (error || newp == NULL) 2128 return (error); 2129 if (tmp < 0) 2130 return (EINVAL); 2131 2132 ip_mtudisc_timeout = tmp; 2133 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 2134 2135 return (0); 2136} 2137 2138#ifdef GATEWAY 2139/* 2140 * sysctl helper routine for net.inet.ip.maxflows. 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 || newp == NULL) 2149 return (s); 2150 2151 s = splsoftnet(); 2152 ipflow_reap(0); 2153 splx(s); 2154 2155 return (0); 2156} 2157 2158static int 2159sysctl_net_inet_ip_hashsize(SYSCTLFN_ARGS) 2160{ 2161 int error, tmp; 2162 struct sysctlnode node; 2163 2164 node = *rnode; 2165 tmp = ip_hashsize; 2166 node.sysctl_data = &tmp; 2167 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2168 if (error || newp == NULL) 2169 return (error); 2170 2171 if ((tmp & (tmp - 1)) == 0 && tmp != 0) { 2172 /* 2173 * Can only fail due to malloc() 2174 */ 2175 if (ipflow_invalidate_all(tmp)) 2176 return ENOMEM; 2177 } else { 2178 /* 2179 * EINVAL if not a power of 2 2180 */ 2181 return EINVAL; 2182 } 2183 2184 return (0); 2185} 2186#endif /* GATEWAY */ 2187 2188 2189SYSCTL_SETUP(sysctl_net_inet_ip_setup, "sysctl net.inet.ip subtree setup") 2190{ 2191 extern int subnetsarelocal, hostzeroisbroadcast; 2192 2193 sysctl_createv(clog, 0, NULL, NULL, 2194 CTLFLAG_PERMANENT, 2195 CTLTYPE_NODE, "net", NULL, 2196 NULL, 0, NULL, 0, 2197 CTL_NET, CTL_EOL); 2198 sysctl_createv(clog, 0, NULL, NULL, 2199 CTLFLAG_PERMANENT, 2200 CTLTYPE_NODE, "inet", 2201 SYSCTL_DESCR("PF_INET related settings"), 2202 NULL, 0, NULL, 0, 2203 CTL_NET, PF_INET, CTL_EOL); 2204 sysctl_createv(clog, 0, NULL, NULL, 2205 CTLFLAG_PERMANENT, 2206 CTLTYPE_NODE, "ip", 2207 SYSCTL_DESCR("IPv4 related settings"), 2208 NULL, 0, NULL, 0, 2209 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 2210 2211 sysctl_createv(clog, 0, NULL, NULL, 2212 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2213 CTLTYPE_INT, "forwarding", 2214 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 2215 NULL, 0, &ipforwarding, 0, 2216 CTL_NET, PF_INET, IPPROTO_IP, 2217 IPCTL_FORWARDING, CTL_EOL); 2218 sysctl_createv(clog, 0, NULL, NULL, 2219 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2220 CTLTYPE_INT, "redirect", 2221 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 2222 NULL, 0, &ipsendredirects, 0, 2223 CTL_NET, PF_INET, IPPROTO_IP, 2224 IPCTL_SENDREDIRECTS, CTL_EOL); 2225 sysctl_createv(clog, 0, NULL, NULL, 2226 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2227 CTLTYPE_INT, "ttl", 2228 SYSCTL_DESCR("Default TTL for an INET datagram"), 2229 NULL, 0, &ip_defttl, 0, 2230 CTL_NET, PF_INET, IPPROTO_IP, 2231 IPCTL_DEFTTL, CTL_EOL); 2232#ifdef IPCTL_DEFMTU 2233 sysctl_createv(clog, 0, NULL, NULL, 2234 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 2235 CTLTYPE_INT, "mtu", 2236 SYSCTL_DESCR("Default MTA for an INET route"), 2237 NULL, 0, &ip_mtu, 0, 2238 CTL_NET, PF_INET, IPPROTO_IP, 2239 IPCTL_DEFMTU, CTL_EOL); 2240#endif /* IPCTL_DEFMTU */ 2241 sysctl_createv(clog, 0, NULL, NULL, 2242 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2243 CTLTYPE_INT, "forwsrcrt", 2244 SYSCTL_DESCR("Enable forwarding of source-routed " 2245 "datagrams"), 2246 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 2247 CTL_NET, PF_INET, IPPROTO_IP, 2248 IPCTL_FORWSRCRT, CTL_EOL); 2249 sysctl_createv(clog, 0, NULL, NULL, 2250 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2251 CTLTYPE_INT, "directed-broadcast", 2252 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 2253 NULL, 0, &ip_directedbcast, 0, 2254 CTL_NET, PF_INET, IPPROTO_IP, 2255 IPCTL_DIRECTEDBCAST, CTL_EOL); 2256 sysctl_createv(clog, 0, NULL, NULL, 2257 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2258 CTLTYPE_INT, "allowsrcrt", 2259 SYSCTL_DESCR("Accept source-routed datagrams"), 2260 NULL, 0, &ip_allowsrcrt, 0, 2261 CTL_NET, PF_INET, IPPROTO_IP, 2262 IPCTL_ALLOWSRCRT, CTL_EOL); 2263 sysctl_createv(clog, 0, NULL, NULL, 2264 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2265 CTLTYPE_INT, "subnetsarelocal", 2266 SYSCTL_DESCR("Whether logical subnets are considered " 2267 "local"), 2268 NULL, 0, &subnetsarelocal, 0, 2269 CTL_NET, PF_INET, IPPROTO_IP, 2270 IPCTL_SUBNETSARELOCAL, CTL_EOL); 2271 sysctl_createv(clog, 0, NULL, NULL, 2272 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2273 CTLTYPE_INT, "mtudisc", 2274 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 2275 NULL, 0, &ip_mtudisc, 0, 2276 CTL_NET, PF_INET, IPPROTO_IP, 2277 IPCTL_MTUDISC, CTL_EOL); 2278 sysctl_createv(clog, 0, NULL, NULL, 2279 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2280 CTLTYPE_INT, "anonportmin", 2281 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 2282 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 2283 CTL_NET, PF_INET, IPPROTO_IP, 2284 IPCTL_ANONPORTMIN, CTL_EOL); 2285 sysctl_createv(clog, 0, NULL, NULL, 2286 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2287 CTLTYPE_INT, "anonportmax", 2288 SYSCTL_DESCR("Highest ephemeral port number to assign"), 2289 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 2290 CTL_NET, PF_INET, IPPROTO_IP, 2291 IPCTL_ANONPORTMAX, CTL_EOL); 2292 sysctl_createv(clog, 0, NULL, NULL, 2293 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2294 CTLTYPE_INT, "mtudisctimeout", 2295 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 2296 sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0, 2297 CTL_NET, PF_INET, IPPROTO_IP, 2298 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 2299#ifdef GATEWAY 2300 sysctl_createv(clog, 0, NULL, NULL, 2301 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2302 CTLTYPE_INT, "maxflows", 2303 SYSCTL_DESCR("Number of flows for fast forwarding"), 2304 sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0, 2305 CTL_NET, PF_INET, IPPROTO_IP, 2306 IPCTL_MAXFLOWS, CTL_EOL); 2307 sysctl_createv(clog, 0, NULL, NULL, 2308 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2309 CTLTYPE_INT, "hashsize", 2310 SYSCTL_DESCR("Size of hash table for fast forwarding (IPv4)"), 2311 sysctl_net_inet_ip_hashsize, 0, &ip_hashsize, 0, 2312 CTL_NET, PF_INET, IPPROTO_IP, 2313 CTL_CREATE, CTL_EOL); 2314#endif /* GATEWAY */ 2315 sysctl_createv(clog, 0, NULL, NULL, 2316 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2317 CTLTYPE_INT, "hostzerobroadcast", 2318 SYSCTL_DESCR("All zeroes address is broadcast address"), 2319 NULL, 0, &hostzeroisbroadcast, 0, 2320 CTL_NET, PF_INET, IPPROTO_IP, 2321 IPCTL_HOSTZEROBROADCAST, CTL_EOL); 2322#if NGIF > 0 2323 sysctl_createv(clog, 0, NULL, NULL, 2324 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2325 CTLTYPE_INT, "gifttl", 2326 SYSCTL_DESCR("Default TTL for a gif tunnel datagram"), 2327 NULL, 0, &ip_gif_ttl, 0, 2328 CTL_NET, PF_INET, IPPROTO_IP, 2329 IPCTL_GIF_TTL, CTL_EOL); 2330#endif /* NGIF */ 2331#ifndef IPNOPRIVPORTS 2332 sysctl_createv(clog, 0, NULL, NULL, 2333 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2334 CTLTYPE_INT, "lowportmin", 2335 SYSCTL_DESCR("Lowest privileged ephemeral port number " 2336 "to assign"), 2337 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 2338 CTL_NET, PF_INET, IPPROTO_IP, 2339 IPCTL_LOWPORTMIN, CTL_EOL); 2340 sysctl_createv(clog, 0, NULL, NULL, 2341 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2342 CTLTYPE_INT, "lowportmax", 2343 SYSCTL_DESCR("Highest privileged ephemeral port number " 2344 "to assign"), 2345 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 2346 CTL_NET, PF_INET, IPPROTO_IP, 2347 IPCTL_LOWPORTMAX, CTL_EOL); 2348#endif /* IPNOPRIVPORTS */ 2349 sysctl_createv(clog, 0, NULL, NULL, 2350 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2351 CTLTYPE_INT, "maxfragpackets", 2352 SYSCTL_DESCR("Maximum number of fragments to retain for " 2353 "possible reassembly"), 2354 NULL, 0, &ip_maxfragpackets, 0, 2355 CTL_NET, PF_INET, IPPROTO_IP, 2356 IPCTL_MAXFRAGPACKETS, CTL_EOL); 2357#if NGRE > 0 2358 sysctl_createv(clog, 0, NULL, NULL, 2359 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2360 CTLTYPE_INT, "grettl", 2361 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 2362 NULL, 0, &ip_gre_ttl, 0, 2363 CTL_NET, PF_INET, IPPROTO_IP, 2364 IPCTL_GRE_TTL, CTL_EOL); 2365#endif /* NGRE */ 2366 sysctl_createv(clog, 0, NULL, NULL, 2367 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2368 CTLTYPE_INT, "checkinterface", 2369 SYSCTL_DESCR("Enable receive side of Strong ES model " 2370 "from RFC1122"), 2371 NULL, 0, &ip_checkinterface, 0, 2372 CTL_NET, PF_INET, IPPROTO_IP, 2373 IPCTL_CHECKINTERFACE, CTL_EOL); 2374 sysctl_createv(clog, 0, NULL, NULL, 2375 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2376 CTLTYPE_INT, "random_id", 2377 SYSCTL_DESCR("Assign random ip_id values"), 2378 NULL, 0, &ip_do_randomid, 0, 2379 CTL_NET, PF_INET, IPPROTO_IP, 2380 IPCTL_RANDOMID, CTL_EOL); 2381 sysctl_createv(clog, 0, NULL, NULL, 2382 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2383 CTLTYPE_INT, "do_loopback_cksum", 2384 SYSCTL_DESCR("Perform IP checksum on loopback"), 2385 NULL, 0, &ip_do_loopback_cksum, 0, 2386 CTL_NET, PF_INET, IPPROTO_IP, 2387 IPCTL_LOOPBACKCKSUM, CTL_EOL); 2388 sysctl_createv(clog, 0, NULL, NULL, 2389 CTLFLAG_PERMANENT, 2390 CTLTYPE_STRUCT, "stats", 2391 SYSCTL_DESCR("IP statistics"), 2392 NULL, 0, &ipstat, sizeof(ipstat), 2393 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 2394 CTL_EOL); 2395} 2396