ip_input.c revision 1.259
1/* $NetBSD: ip_input.c,v 1.259 2007/12/21 23:49:09 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 1.259 2007/12/21 23:49:09 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 = htons(ip->ip_off) & ~(IP_DF|IP_EF|IP_MF); 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 goto found; 923 924 } 925 fp = 0; 926found: 927 928 /* 929 * Adjust ip_len to not reflect header, 930 * set ipqe_mff if more fragments are expected, 931 * convert offset of this to bytes. 932 */ 933 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 934 mff = (ip->ip_off & htons(IP_MF)) != 0; 935 if (mff) { 936 /* 937 * Make sure that fragments have a data length 938 * that's a non-zero multiple of 8 bytes. 939 */ 940 if (ntohs(ip->ip_len) == 0 || 941 (ntohs(ip->ip_len) & 0x7) != 0) { 942 ipstat.ips_badfrags++; 943 IPQ_UNLOCK(); 944 goto bad; 945 } 946 } 947 ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3); 948 949 /* 950 * If datagram marked as having more fragments 951 * or if this is not the first fragment, 952 * attempt reassembly; if it succeeds, proceed. 953 */ 954 if (mff || ip->ip_off != htons(0)) { 955 ipstat.ips_fragments++; 956 s = splvm(); 957 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 958 splx(s); 959 if (ipqe == NULL) { 960 ipstat.ips_rcvmemdrop++; 961 IPQ_UNLOCK(); 962 goto bad; 963 } 964 ipqe->ipqe_mff = mff; 965 ipqe->ipqe_m = m; 966 ipqe->ipqe_ip = ip; 967 m = ip_reass(ipqe, fp, &ipq[hash]); 968 if (m == 0) { 969 IPQ_UNLOCK(); 970 return; 971 } 972 ipstat.ips_reassembled++; 973 ip = mtod(m, struct ip *); 974 hlen = ip->ip_hl << 2; 975 ip->ip_len = htons(ntohs(ip->ip_len) + hlen); 976 } else 977 if (fp) 978 ip_freef(fp); 979 IPQ_UNLOCK(); 980 } 981 982#if defined(IPSEC) 983 /* 984 * enforce IPsec policy checking if we are seeing last header. 985 * note that we do not visit this with protocols with pcb layer 986 * code - like udp/tcp/raw ip. 987 */ 988 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 989 ipsec4_in_reject(m, NULL)) { 990 ipsecstat.in_polvio++; 991 goto bad; 992 } 993#endif 994#ifdef FAST_IPSEC 995 /* 996 * enforce IPsec policy checking if we are seeing last header. 997 * note that we do not visit this with protocols with pcb layer 998 * code - like udp/tcp/raw ip. 999 */ 1000 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 1001 /* 1002 * Check if the packet has already had IPsec processing 1003 * done. If so, then just pass it along. This tag gets 1004 * set during AH, ESP, etc. input handling, before the 1005 * packet is returned to the ip input queue for delivery. 1006 */ 1007 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 1008 s = splsoftnet(); 1009 if (mtag != NULL) { 1010 tdbi = (struct tdb_ident *)(mtag + 1); 1011 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 1012 } else { 1013 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 1014 IP_FORWARDING, &error); 1015 } 1016 if (sp != NULL) { 1017 /* 1018 * Check security policy against packet attributes. 1019 */ 1020 error = ipsec_in_reject(sp, m); 1021 KEY_FREESP(&sp); 1022 } else { 1023 /* XXX error stat??? */ 1024 error = EINVAL; 1025DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 1026 } 1027 splx(s); 1028 if (error) 1029 goto bad; 1030 } 1031#endif /* FAST_IPSEC */ 1032 1033 /* 1034 * Switch out to protocol's input routine. 1035 */ 1036#if IFA_STATS 1037 if (ia && ip) 1038 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 1039#endif 1040 ipstat.ips_delivered++; 1041 { 1042 int off = hlen, nh = ip->ip_p; 1043 1044 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 1045 return; 1046 } 1047bad: 1048 m_freem(m); 1049 return; 1050 1051badcsum: 1052 ipstat.ips_badsum++; 1053 m_freem(m); 1054} 1055 1056/* 1057 * Take incoming datagram fragment and try to 1058 * reassemble it into whole datagram. If a chain for 1059 * reassembly of this datagram already exists, then it 1060 * is given as fp; otherwise have to make a chain. 1061 */ 1062struct mbuf * 1063ip_reass(struct ipqent *ipqe, struct ipq *fp, struct ipqhead *ipqhead) 1064{ 1065 struct mbuf *m = ipqe->ipqe_m; 1066 struct ipqent *nq, *p, *q; 1067 struct ip *ip; 1068 struct mbuf *t; 1069 int hlen = ipqe->ipqe_ip->ip_hl << 2; 1070 int i, next, s; 1071 1072 IPQ_LOCK_CHECK(); 1073 1074 /* 1075 * Presence of header sizes in mbufs 1076 * would confuse code below. 1077 */ 1078 m->m_data += hlen; 1079 m->m_len -= hlen; 1080 1081#ifdef notyet 1082 /* make sure fragment limit is up-to-date */ 1083 CHECK_NMBCLUSTER_PARAMS(); 1084 1085 /* If we have too many fragments, drop the older half. */ 1086 if (ip_nfrags >= ip_maxfrags) 1087 ip_reass_drophalf(void); 1088#endif 1089 1090 /* 1091 * We are about to add a fragment; increment frag count. 1092 */ 1093 ip_nfrags++; 1094 1095 /* 1096 * If first fragment to arrive, create a reassembly queue. 1097 */ 1098 if (fp == 0) { 1099 /* 1100 * Enforce upper bound on number of fragmented packets 1101 * for which we attempt reassembly; 1102 * If maxfrag is 0, never accept fragments. 1103 * If maxfrag is -1, accept all fragments without limitation. 1104 */ 1105 if (ip_maxfragpackets < 0) 1106 ; 1107 else if (ip_nfragpackets >= ip_maxfragpackets) 1108 goto dropfrag; 1109 ip_nfragpackets++; 1110 MALLOC(fp, struct ipq *, sizeof (struct ipq), 1111 M_FTABLE, M_NOWAIT); 1112 if (fp == NULL) 1113 goto dropfrag; 1114 LIST_INSERT_HEAD(ipqhead, fp, ipq_q); 1115 fp->ipq_nfrags = 1; 1116 fp->ipq_ttl = IPFRAGTTL; 1117 fp->ipq_p = ipqe->ipqe_ip->ip_p; 1118 fp->ipq_id = ipqe->ipqe_ip->ip_id; 1119 TAILQ_INIT(&fp->ipq_fragq); 1120 fp->ipq_src = ipqe->ipqe_ip->ip_src; 1121 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 1122 p = NULL; 1123 goto insert; 1124 } else { 1125 fp->ipq_nfrags++; 1126 } 1127 1128 /* 1129 * Find a segment which begins after this one does. 1130 */ 1131 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1132 p = q, q = TAILQ_NEXT(q, ipqe_q)) 1133 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off)) 1134 break; 1135 1136 /* 1137 * If there is a preceding segment, it may provide some of 1138 * our data already. If so, drop the data from the incoming 1139 * segment. If it provides all of our data, drop us. 1140 */ 1141 if (p != NULL) { 1142 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) - 1143 ntohs(ipqe->ipqe_ip->ip_off); 1144 if (i > 0) { 1145 if (i >= ntohs(ipqe->ipqe_ip->ip_len)) 1146 goto dropfrag; 1147 m_adj(ipqe->ipqe_m, i); 1148 ipqe->ipqe_ip->ip_off = 1149 htons(ntohs(ipqe->ipqe_ip->ip_off) + i); 1150 ipqe->ipqe_ip->ip_len = 1151 htons(ntohs(ipqe->ipqe_ip->ip_len) - i); 1152 } 1153 } 1154 1155 /* 1156 * While we overlap succeeding segments trim them or, 1157 * if they are completely covered, dequeue them. 1158 */ 1159 for (; q != NULL && 1160 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) > 1161 ntohs(q->ipqe_ip->ip_off); q = nq) { 1162 i = (ntohs(ipqe->ipqe_ip->ip_off) + 1163 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off); 1164 if (i < ntohs(q->ipqe_ip->ip_len)) { 1165 q->ipqe_ip->ip_len = 1166 htons(ntohs(q->ipqe_ip->ip_len) - i); 1167 q->ipqe_ip->ip_off = 1168 htons(ntohs(q->ipqe_ip->ip_off) + i); 1169 m_adj(q->ipqe_m, i); 1170 break; 1171 } 1172 nq = TAILQ_NEXT(q, ipqe_q); 1173 m_freem(q->ipqe_m); 1174 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1175 s = splvm(); 1176 pool_put(&ipqent_pool, q); 1177 splx(s); 1178 fp->ipq_nfrags--; 1179 ip_nfrags--; 1180 } 1181 1182insert: 1183 /* 1184 * Stick new segment in its place; 1185 * check for complete reassembly. 1186 */ 1187 if (p == NULL) { 1188 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 1189 } else { 1190 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q); 1191 } 1192 next = 0; 1193 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1194 p = q, q = TAILQ_NEXT(q, ipqe_q)) { 1195 if (ntohs(q->ipqe_ip->ip_off) != next) 1196 return (0); 1197 next += ntohs(q->ipqe_ip->ip_len); 1198 } 1199 if (p->ipqe_mff) 1200 return (0); 1201 1202 /* 1203 * Reassembly is complete. Check for a bogus message size and 1204 * concatenate fragments. 1205 */ 1206 q = TAILQ_FIRST(&fp->ipq_fragq); 1207 ip = q->ipqe_ip; 1208 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 1209 ipstat.ips_toolong++; 1210 ip_freef(fp); 1211 return (0); 1212 } 1213 m = q->ipqe_m; 1214 t = m->m_next; 1215 m->m_next = 0; 1216 m_cat(m, t); 1217 nq = TAILQ_NEXT(q, ipqe_q); 1218 s = splvm(); 1219 pool_put(&ipqent_pool, q); 1220 splx(s); 1221 for (q = nq; q != NULL; q = nq) { 1222 t = q->ipqe_m; 1223 nq = TAILQ_NEXT(q, ipqe_q); 1224 s = splvm(); 1225 pool_put(&ipqent_pool, q); 1226 splx(s); 1227 m_cat(m, t); 1228 } 1229 ip_nfrags -= fp->ipq_nfrags; 1230 1231 /* 1232 * Create header for new ip packet by 1233 * modifying header of first packet; 1234 * dequeue and discard fragment reassembly header. 1235 * Make header visible. 1236 */ 1237 ip->ip_len = htons(next); 1238 ip->ip_src = fp->ipq_src; 1239 ip->ip_dst = fp->ipq_dst; 1240 LIST_REMOVE(fp, ipq_q); 1241 FREE(fp, M_FTABLE); 1242 ip_nfragpackets--; 1243 m->m_len += (ip->ip_hl << 2); 1244 m->m_data -= (ip->ip_hl << 2); 1245 /* some debugging cruft by sklower, below, will go away soon */ 1246 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1247 int plen = 0; 1248 for (t = m; t; t = t->m_next) 1249 plen += t->m_len; 1250 m->m_pkthdr.len = plen; 1251 m->m_pkthdr.csum_flags = 0; 1252 } 1253 return (m); 1254 1255dropfrag: 1256 if (fp != 0) 1257 fp->ipq_nfrags--; 1258 ip_nfrags--; 1259 ipstat.ips_fragdropped++; 1260 m_freem(m); 1261 s = splvm(); 1262 pool_put(&ipqent_pool, ipqe); 1263 splx(s); 1264 return (0); 1265} 1266 1267/* 1268 * Free a fragment reassembly header and all 1269 * associated datagrams. 1270 */ 1271void 1272ip_freef(struct ipq *fp) 1273{ 1274 struct ipqent *q, *p; 1275 u_int nfrags = 0; 1276 int s; 1277 1278 IPQ_LOCK_CHECK(); 1279 1280 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) { 1281 p = TAILQ_NEXT(q, ipqe_q); 1282 m_freem(q->ipqe_m); 1283 nfrags++; 1284 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1285 s = splvm(); 1286 pool_put(&ipqent_pool, q); 1287 splx(s); 1288 } 1289 1290 if (nfrags != fp->ipq_nfrags) 1291 printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags); 1292 ip_nfrags -= nfrags; 1293 LIST_REMOVE(fp, ipq_q); 1294 FREE(fp, M_FTABLE); 1295 ip_nfragpackets--; 1296} 1297 1298/* 1299 * IP reassembly TTL machinery for multiplicative drop. 1300 */ 1301static u_int fragttl_histo[(IPFRAGTTL+1)]; 1302 1303 1304/* 1305 * Decrement TTL of all reasembly queue entries by `ticks'. 1306 * Count number of distinct fragments (as opposed to partial, fragmented 1307 * datagrams) in the reassembly queue. While we traverse the entire 1308 * reassembly queue, compute and return the median TTL over all fragments. 1309 */ 1310static u_int 1311ip_reass_ttl_decr(u_int ticks) 1312{ 1313 u_int nfrags, median, dropfraction, keepfraction; 1314 struct ipq *fp, *nfp; 1315 int i; 1316 1317 nfrags = 0; 1318 memset(fragttl_histo, 0, sizeof fragttl_histo); 1319 1320 for (i = 0; i < IPREASS_NHASH; i++) { 1321 for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) { 1322 fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ? 1323 0 : fp->ipq_ttl - ticks); 1324 nfp = LIST_NEXT(fp, ipq_q); 1325 if (fp->ipq_ttl == 0) { 1326 ipstat.ips_fragtimeout++; 1327 ip_freef(fp); 1328 } else { 1329 nfrags += fp->ipq_nfrags; 1330 fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags; 1331 } 1332 } 1333 } 1334 1335 KASSERT(ip_nfrags == nfrags); 1336 1337 /* Find median (or other drop fraction) in histogram. */ 1338 dropfraction = (ip_nfrags / 2); 1339 keepfraction = ip_nfrags - dropfraction; 1340 for (i = IPFRAGTTL, median = 0; i >= 0; i--) { 1341 median += fragttl_histo[i]; 1342 if (median >= keepfraction) 1343 break; 1344 } 1345 1346 /* Return TTL of median (or other fraction). */ 1347 return (u_int)i; 1348} 1349 1350void 1351ip_reass_drophalf(void) 1352{ 1353 1354 u_int median_ticks; 1355 /* 1356 * Compute median TTL of all fragments, and count frags 1357 * with that TTL or lower (roughly half of all fragments). 1358 */ 1359 median_ticks = ip_reass_ttl_decr(0); 1360 1361 /* Drop half. */ 1362 median_ticks = ip_reass_ttl_decr(median_ticks); 1363 1364} 1365 1366/* 1367 * IP timer processing; 1368 * if a timer expires on a reassembly 1369 * queue, discard it. 1370 */ 1371void 1372ip_slowtimo(void) 1373{ 1374 static u_int dropscanidx = 0; 1375 u_int i; 1376 u_int median_ttl; 1377 int s = splsoftnet(); 1378 1379 IPQ_LOCK(); 1380 1381 /* Age TTL of all fragments by 1 tick .*/ 1382 median_ttl = ip_reass_ttl_decr(1); 1383 1384 /* make sure fragment limit is up-to-date */ 1385 CHECK_NMBCLUSTER_PARAMS(); 1386 1387 /* If we have too many fragments, drop the older half. */ 1388 if (ip_nfrags > ip_maxfrags) 1389 ip_reass_ttl_decr(median_ttl); 1390 1391 /* 1392 * If we are over the maximum number of fragmented packets 1393 * (due to the limit being lowered), drain off 1394 * enough to get down to the new limit. Start draining 1395 * from the reassembly hashqueue most recently drained. 1396 */ 1397 if (ip_maxfragpackets < 0) 1398 ; 1399 else { 1400 int wrapped = 0; 1401 1402 i = dropscanidx; 1403 while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) { 1404 while (LIST_FIRST(&ipq[i]) != NULL) 1405 ip_freef(LIST_FIRST(&ipq[i])); 1406 if (++i >= IPREASS_NHASH) { 1407 i = 0; 1408 } 1409 /* 1410 * Dont scan forever even if fragment counters are 1411 * wrong: stop after scanning entire reassembly queue. 1412 */ 1413 if (i == dropscanidx) 1414 wrapped = 1; 1415 } 1416 dropscanidx = i; 1417 } 1418 IPQ_UNLOCK(); 1419 splx(s); 1420} 1421 1422/* 1423 * Drain off all datagram fragments. 1424 */ 1425void 1426ip_drain(void) 1427{ 1428 1429 /* 1430 * We may be called from a device's interrupt context. If 1431 * the ipq is already busy, just bail out now. 1432 */ 1433 if (ipq_lock_try() == 0) 1434 return; 1435 1436 /* 1437 * Drop half the total fragments now. If more mbufs are needed, 1438 * we will be called again soon. 1439 */ 1440 ip_reass_drophalf(); 1441 1442 IPQ_UNLOCK(); 1443} 1444 1445/* 1446 * Do option processing on a datagram, 1447 * possibly discarding it if bad options are encountered, 1448 * or forwarding it if source-routed. 1449 * Returns 1 if packet has been forwarded/freed, 1450 * 0 if the packet should be processed further. 1451 */ 1452int 1453ip_dooptions(struct mbuf *m) 1454{ 1455 struct ip *ip = mtod(m, struct ip *); 1456 u_char *cp, *cp0; 1457 struct ip_timestamp *ipt; 1458 struct in_ifaddr *ia; 1459 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1460 struct in_addr dst; 1461 n_time ntime; 1462 1463 dst = ip->ip_dst; 1464 cp = (u_char *)(ip + 1); 1465 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1466 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1467 opt = cp[IPOPT_OPTVAL]; 1468 if (opt == IPOPT_EOL) 1469 break; 1470 if (opt == IPOPT_NOP) 1471 optlen = 1; 1472 else { 1473 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1474 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1475 goto bad; 1476 } 1477 optlen = cp[IPOPT_OLEN]; 1478 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1479 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1480 goto bad; 1481 } 1482 } 1483 switch (opt) { 1484 1485 default: 1486 break; 1487 1488 /* 1489 * Source routing with record. 1490 * Find interface with current destination address. 1491 * If none on this machine then drop if strictly routed, 1492 * or do nothing if loosely routed. 1493 * Record interface address and bring up next address 1494 * component. If strictly routed make sure next 1495 * address is on directly accessible net. 1496 */ 1497 case IPOPT_LSRR: 1498 case IPOPT_SSRR: 1499 if (ip_allowsrcrt == 0) { 1500 type = ICMP_UNREACH; 1501 code = ICMP_UNREACH_NET_PROHIB; 1502 goto bad; 1503 } 1504 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1505 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1506 goto bad; 1507 } 1508 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1509 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1510 goto bad; 1511 } 1512 ipaddr.sin_addr = ip->ip_dst; 1513 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1514 if (ia == 0) { 1515 if (opt == IPOPT_SSRR) { 1516 type = ICMP_UNREACH; 1517 code = ICMP_UNREACH_SRCFAIL; 1518 goto bad; 1519 } 1520 /* 1521 * Loose routing, and not at next destination 1522 * yet; nothing to do except forward. 1523 */ 1524 break; 1525 } 1526 off--; /* 0 origin */ 1527 if ((off + sizeof(struct in_addr)) > optlen) { 1528 /* 1529 * End of source route. Should be for us. 1530 */ 1531 save_rte(cp, ip->ip_src); 1532 break; 1533 } 1534 /* 1535 * locate outgoing interface 1536 */ 1537 bcopy((void *)(cp + off), (void *)&ipaddr.sin_addr, 1538 sizeof(ipaddr.sin_addr)); 1539 if (opt == IPOPT_SSRR) 1540 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); 1541 else 1542 ia = ip_rtaddr(ipaddr.sin_addr); 1543 if (ia == 0) { 1544 type = ICMP_UNREACH; 1545 code = ICMP_UNREACH_SRCFAIL; 1546 goto bad; 1547 } 1548 ip->ip_dst = ipaddr.sin_addr; 1549 bcopy((void *)&ia->ia_addr.sin_addr, 1550 (void *)(cp + off), sizeof(struct in_addr)); 1551 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1552 /* 1553 * Let ip_intr's mcast routing check handle mcast pkts 1554 */ 1555 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1556 break; 1557 1558 case IPOPT_RR: 1559 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1560 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1561 goto bad; 1562 } 1563 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1564 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1565 goto bad; 1566 } 1567 /* 1568 * If no space remains, ignore. 1569 */ 1570 off--; /* 0 origin */ 1571 if ((off + sizeof(struct in_addr)) > optlen) 1572 break; 1573 bcopy((void *)(&ip->ip_dst), (void *)&ipaddr.sin_addr, 1574 sizeof(ipaddr.sin_addr)); 1575 /* 1576 * locate outgoing interface; if we're the destination, 1577 * use the incoming interface (should be same). 1578 */ 1579 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1580 == NULL && 1581 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1582 type = ICMP_UNREACH; 1583 code = ICMP_UNREACH_HOST; 1584 goto bad; 1585 } 1586 bcopy((void *)&ia->ia_addr.sin_addr, 1587 (void *)(cp + off), sizeof(struct in_addr)); 1588 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1589 break; 1590 1591 case IPOPT_TS: 1592 code = cp - (u_char *)ip; 1593 ipt = (struct ip_timestamp *)cp; 1594 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1595 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1596 goto bad; 1597 } 1598 if (ipt->ipt_ptr < 5) { 1599 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1600 goto bad; 1601 } 1602 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1603 if (++ipt->ipt_oflw == 0) { 1604 code = (u_char *)&ipt->ipt_ptr - 1605 (u_char *)ip; 1606 goto bad; 1607 } 1608 break; 1609 } 1610 cp0 = (cp + ipt->ipt_ptr - 1); 1611 switch (ipt->ipt_flg) { 1612 1613 case IPOPT_TS_TSONLY: 1614 break; 1615 1616 case IPOPT_TS_TSANDADDR: 1617 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1618 sizeof(struct in_addr) > ipt->ipt_len) { 1619 code = (u_char *)&ipt->ipt_ptr - 1620 (u_char *)ip; 1621 goto bad; 1622 } 1623 ipaddr.sin_addr = dst; 1624 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1625 m->m_pkthdr.rcvif)); 1626 if (ia == 0) 1627 continue; 1628 bcopy(&ia->ia_addr.sin_addr, 1629 cp0, sizeof(struct in_addr)); 1630 ipt->ipt_ptr += sizeof(struct in_addr); 1631 break; 1632 1633 case IPOPT_TS_PRESPEC: 1634 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1635 sizeof(struct in_addr) > ipt->ipt_len) { 1636 code = (u_char *)&ipt->ipt_ptr - 1637 (u_char *)ip; 1638 goto bad; 1639 } 1640 bcopy(cp0, &ipaddr.sin_addr, 1641 sizeof(struct in_addr)); 1642 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1643 == NULL) 1644 continue; 1645 ipt->ipt_ptr += sizeof(struct in_addr); 1646 break; 1647 1648 default: 1649 /* XXX can't take &ipt->ipt_flg */ 1650 code = (u_char *)&ipt->ipt_ptr - 1651 (u_char *)ip + 1; 1652 goto bad; 1653 } 1654 ntime = iptime(); 1655 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1656 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1657 sizeof(n_time)); 1658 ipt->ipt_ptr += sizeof(n_time); 1659 } 1660 } 1661 if (forward) { 1662 if (ip_forwsrcrt == 0) { 1663 type = ICMP_UNREACH; 1664 code = ICMP_UNREACH_SRCFAIL; 1665 goto bad; 1666 } 1667 ip_forward(m, 1); 1668 return (1); 1669 } 1670 return (0); 1671bad: 1672 icmp_error(m, type, code, 0, 0); 1673 ipstat.ips_badoptions++; 1674 return (1); 1675} 1676 1677/* 1678 * Given address of next destination (final or next hop), 1679 * return internet address info of interface to be used to get there. 1680 */ 1681struct in_ifaddr * 1682ip_rtaddr(struct in_addr dst) 1683{ 1684 struct rtentry *rt; 1685 union { 1686 struct sockaddr dst; 1687 struct sockaddr_in dst4; 1688 } u; 1689 1690 sockaddr_in_init(&u.dst4, &dst, 0); 1691 1692 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) 1693 return NULL; 1694 1695 return ifatoia(rt->rt_ifa); 1696} 1697 1698/* 1699 * Save incoming source route for use in replies, 1700 * to be picked up later by ip_srcroute if the receiver is interested. 1701 */ 1702void 1703save_rte(u_char *option, struct in_addr dst) 1704{ 1705 unsigned olen; 1706 1707 olen = option[IPOPT_OLEN]; 1708#ifdef DIAGNOSTIC 1709 if (ipprintfs) 1710 printf("save_rte: olen %d\n", olen); 1711#endif /* 0 */ 1712 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1713 return; 1714 bcopy((void *)option, (void *)ip_srcrt.srcopt, olen); 1715 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1716 ip_srcrt.dst = dst; 1717} 1718 1719/* 1720 * Retrieve incoming source route for use in replies, 1721 * in the same form used by setsockopt. 1722 * The first hop is placed before the options, will be removed later. 1723 */ 1724struct mbuf * 1725ip_srcroute(void) 1726{ 1727 struct in_addr *p, *q; 1728 struct mbuf *m; 1729 1730 if (ip_nhops == 0) 1731 return NULL; 1732 m = m_get(M_DONTWAIT, MT_SOOPTS); 1733 if (m == 0) 1734 return NULL; 1735 1736 MCLAIM(m, &inetdomain.dom_mowner); 1737#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1738 1739 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1740 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1741 OPTSIZ; 1742#ifdef DIAGNOSTIC 1743 if (ipprintfs) 1744 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1745#endif 1746 1747 /* 1748 * First save first hop for return route 1749 */ 1750 p = &ip_srcrt.route[ip_nhops - 1]; 1751 *(mtod(m, struct in_addr *)) = *p--; 1752#ifdef DIAGNOSTIC 1753 if (ipprintfs) 1754 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1755#endif 1756 1757 /* 1758 * Copy option fields and padding (nop) to mbuf. 1759 */ 1760 ip_srcrt.nop = IPOPT_NOP; 1761 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1762 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop, 1763 OPTSIZ); 1764 q = (struct in_addr *)(mtod(m, char *) + 1765 sizeof(struct in_addr) + OPTSIZ); 1766#undef OPTSIZ 1767 /* 1768 * Record return path as an IP source route, 1769 * reversing the path (pointers are now aligned). 1770 */ 1771 while (p >= ip_srcrt.route) { 1772#ifdef DIAGNOSTIC 1773 if (ipprintfs) 1774 printf(" %x", ntohl(q->s_addr)); 1775#endif 1776 *q++ = *p--; 1777 } 1778 /* 1779 * Last hop goes to final destination. 1780 */ 1781 *q = ip_srcrt.dst; 1782#ifdef DIAGNOSTIC 1783 if (ipprintfs) 1784 printf(" %x\n", ntohl(q->s_addr)); 1785#endif 1786 return (m); 1787} 1788 1789const int inetctlerrmap[PRC_NCMDS] = { 1790 [PRC_MSGSIZE] = EMSGSIZE, 1791 [PRC_HOSTDEAD] = EHOSTDOWN, 1792 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1793 [PRC_UNREACH_NET] = EHOSTUNREACH, 1794 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1795 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1796 [PRC_UNREACH_PORT] = ECONNREFUSED, 1797 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1798 [PRC_PARAMPROB] = ENOPROTOOPT, 1799}; 1800 1801/* 1802 * Forward a packet. If some error occurs return the sender 1803 * an icmp packet. Note we can't always generate a meaningful 1804 * icmp message because icmp doesn't have a large enough repertoire 1805 * of codes and types. 1806 * 1807 * If not forwarding, just drop the packet. This could be confusing 1808 * if ipforwarding was zero but some routing protocol was advancing 1809 * us as a gateway to somewhere. However, we must let the routing 1810 * protocol deal with that. 1811 * 1812 * The srcrt parameter indicates whether the packet is being forwarded 1813 * via a source route. 1814 */ 1815void 1816ip_forward(struct mbuf *m, int srcrt) 1817{ 1818 struct ip *ip = mtod(m, struct ip *); 1819 struct rtentry *rt; 1820 int error, type = 0, code = 0, destmtu = 0; 1821 struct mbuf *mcopy; 1822 n_long dest; 1823 union { 1824 struct sockaddr dst; 1825 struct sockaddr_in dst4; 1826 } u; 1827 1828 /* 1829 * We are now in the output path. 1830 */ 1831 MCLAIM(m, &ip_tx_mowner); 1832 1833 /* 1834 * Clear any in-bound checksum flags for this packet. 1835 */ 1836 m->m_pkthdr.csum_flags = 0; 1837 1838 dest = 0; 1839#ifdef DIAGNOSTIC 1840 if (ipprintfs) { 1841 printf("forward: src %s ", inet_ntoa(ip->ip_src)); 1842 printf("dst %s ttl %x\n", inet_ntoa(ip->ip_dst), ip->ip_ttl); 1843 } 1844#endif 1845 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1846 ipstat.ips_cantforward++; 1847 m_freem(m); 1848 return; 1849 } 1850 if (ip->ip_ttl <= IPTTLDEC) { 1851 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1852 return; 1853 } 1854 1855 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1856 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) { 1857 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1858 return; 1859 } 1860 1861 /* 1862 * Save at most 68 bytes of the packet in case 1863 * we need to generate an ICMP message to the src. 1864 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1865 */ 1866 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1867 if (mcopy) 1868 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1869 1870 ip->ip_ttl -= IPTTLDEC; 1871 1872 /* 1873 * If forwarding packet using same interface that it came in on, 1874 * perhaps should send a redirect to sender to shortcut a hop. 1875 * Only send redirect if source is sending directly to us, 1876 * and if packet was not source routed (or has any options). 1877 * Also, don't send redirect if forwarding using a default route 1878 * or a route modified by a redirect. 1879 */ 1880 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1881 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1882 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1883 ipsendredirects && !srcrt) { 1884 if (rt->rt_ifa && 1885 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1886 ifatoia(rt->rt_ifa)->ia_subnet) { 1887 if (rt->rt_flags & RTF_GATEWAY) 1888 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1889 else 1890 dest = ip->ip_dst.s_addr; 1891 /* 1892 * Router requirements says to only send host 1893 * redirects. 1894 */ 1895 type = ICMP_REDIRECT; 1896 code = ICMP_REDIRECT_HOST; 1897#ifdef DIAGNOSTIC 1898 if (ipprintfs) 1899 printf("redirect (%d) to %x\n", code, 1900 (u_int32_t)dest); 1901#endif 1902 } 1903 } 1904 1905 error = ip_output(m, NULL, &ipforward_rt, 1906 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1907 (struct ip_moptions *)NULL, (struct socket *)NULL); 1908 1909 if (error) 1910 ipstat.ips_cantforward++; 1911 else { 1912 ipstat.ips_forward++; 1913 if (type) 1914 ipstat.ips_redirectsent++; 1915 else { 1916 if (mcopy) { 1917#ifdef GATEWAY 1918 if (mcopy->m_flags & M_CANFASTFWD) 1919 ipflow_create(&ipforward_rt, mcopy); 1920#endif 1921 m_freem(mcopy); 1922 } 1923 return; 1924 } 1925 } 1926 if (mcopy == NULL) 1927 return; 1928 1929 switch (error) { 1930 1931 case 0: /* forwarded, but need redirect */ 1932 /* type, code set above */ 1933 break; 1934 1935 case ENETUNREACH: /* shouldn't happen, checked above */ 1936 case EHOSTUNREACH: 1937 case ENETDOWN: 1938 case EHOSTDOWN: 1939 default: 1940 type = ICMP_UNREACH; 1941 code = ICMP_UNREACH_HOST; 1942 break; 1943 1944 case EMSGSIZE: 1945 type = ICMP_UNREACH; 1946 code = ICMP_UNREACH_NEEDFRAG; 1947#if !defined(IPSEC) && !defined(FAST_IPSEC) 1948 if ((rt = rtcache_getrt(&ipforward_rt)) != NULL) 1949 destmtu = rt->rt_ifp->if_mtu; 1950#else 1951 /* 1952 * If the packet is routed over IPsec tunnel, tell the 1953 * originator the tunnel MTU. 1954 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1955 * XXX quickhack!!! 1956 */ 1957 if ((rt = rtcache_getrt(&ipforward_rt)) != NULL) { 1958 struct secpolicy *sp; 1959 int ipsecerror; 1960 size_t ipsechdr; 1961 struct route *ro; 1962 1963 sp = ipsec4_getpolicybyaddr(mcopy, 1964 IPSEC_DIR_OUTBOUND, IP_FORWARDING, 1965 &ipsecerror); 1966 1967 if (sp == NULL) 1968 destmtu = rt->rt_ifp->if_mtu; 1969 else { 1970 /* count IPsec header size */ 1971 ipsechdr = ipsec4_hdrsiz(mcopy, 1972 IPSEC_DIR_OUTBOUND, NULL); 1973 1974 /* 1975 * find the correct route for outer IPv4 1976 * header, compute tunnel MTU. 1977 */ 1978 1979 if (sp->req != NULL 1980 && sp->req->sav != NULL 1981 && sp->req->sav->sah != NULL) { 1982 ro = &sp->req->sav->sah->sa_route; 1983 if (rt && rt->rt_ifp) { 1984 destmtu = 1985 rt->rt_rmx.rmx_mtu ? 1986 rt->rt_rmx.rmx_mtu : 1987 rt->rt_ifp->if_mtu; 1988 destmtu -= ipsechdr; 1989 } 1990 } 1991 1992#ifdef IPSEC 1993 key_freesp(sp); 1994#else 1995 KEY_FREESP(&sp); 1996#endif 1997 } 1998 } 1999#endif /*IPSEC*/ 2000 ipstat.ips_cantfrag++; 2001 break; 2002 2003 case ENOBUFS: 2004#if 1 2005 /* 2006 * a router should not generate ICMP_SOURCEQUENCH as 2007 * required in RFC1812 Requirements for IP Version 4 Routers. 2008 * source quench could be a big problem under DoS attacks, 2009 * or if the underlying interface is rate-limited. 2010 */ 2011 if (mcopy) 2012 m_freem(mcopy); 2013 return; 2014#else 2015 type = ICMP_SOURCEQUENCH; 2016 code = 0; 2017 break; 2018#endif 2019 } 2020 icmp_error(mcopy, type, code, dest, destmtu); 2021} 2022 2023void 2024ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2025 struct mbuf *m) 2026{ 2027 2028 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2029 struct timeval tv; 2030 2031 microtime(&tv); 2032 *mp = sbcreatecontrol((void *) &tv, sizeof(tv), 2033 SCM_TIMESTAMP, SOL_SOCKET); 2034 if (*mp) 2035 mp = &(*mp)->m_next; 2036 } 2037 if (inp->inp_flags & INP_RECVDSTADDR) { 2038 *mp = sbcreatecontrol((void *) &ip->ip_dst, 2039 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2040 if (*mp) 2041 mp = &(*mp)->m_next; 2042 } 2043#ifdef notyet 2044 /* 2045 * XXX 2046 * Moving these out of udp_input() made them even more broken 2047 * than they already were. 2048 * - fenner@parc.xerox.com 2049 */ 2050 /* options were tossed already */ 2051 if (inp->inp_flags & INP_RECVOPTS) { 2052 *mp = sbcreatecontrol((void *) opts_deleted_above, 2053 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2054 if (*mp) 2055 mp = &(*mp)->m_next; 2056 } 2057 /* ip_srcroute doesn't do what we want here, need to fix */ 2058 if (inp->inp_flags & INP_RECVRETOPTS) { 2059 *mp = sbcreatecontrol((void *) ip_srcroute(), 2060 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2061 if (*mp) 2062 mp = &(*mp)->m_next; 2063 } 2064#endif 2065 if (inp->inp_flags & INP_RECVIF) { 2066 struct sockaddr_dl sdl; 2067 2068 sockaddr_dl_init(&sdl, sizeof(sdl), 2069 (m->m_pkthdr.rcvif != NULL) 2070 ? m->m_pkthdr.rcvif->if_index 2071 : 0, 2072 0, NULL, 0, NULL, 0); 2073 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 2074 if (*mp) 2075 mp = &(*mp)->m_next; 2076 } 2077} 2078 2079/* 2080 * sysctl helper routine for net.inet.ip.forwsrcrt. 2081 */ 2082static int 2083sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 2084{ 2085 int error, tmp; 2086 struct sysctlnode node; 2087 2088 node = *rnode; 2089 tmp = ip_forwsrcrt; 2090 node.sysctl_data = &tmp; 2091 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2092 if (error || newp == NULL) 2093 return (error); 2094 2095 if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 2096 0, NULL, NULL, NULL)) 2097 return (EPERM); 2098 2099 ip_forwsrcrt = tmp; 2100 2101 return (0); 2102} 2103 2104/* 2105 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 2106 * range of the new value and tweaks timers if it changes. 2107 */ 2108static int 2109sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 2110{ 2111 int error, tmp; 2112 struct sysctlnode node; 2113 2114 node = *rnode; 2115 tmp = ip_mtudisc_timeout; 2116 node.sysctl_data = &tmp; 2117 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2118 if (error || newp == NULL) 2119 return (error); 2120 if (tmp < 0) 2121 return (EINVAL); 2122 2123 ip_mtudisc_timeout = tmp; 2124 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 2125 2126 return (0); 2127} 2128 2129#ifdef GATEWAY 2130/* 2131 * sysctl helper routine for net.inet.ip.maxflows. 2132 */ 2133static int 2134sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS) 2135{ 2136 int s; 2137 2138 s = sysctl_lookup(SYSCTLFN_CALL(rnode)); 2139 if (s || newp == NULL) 2140 return (s); 2141 2142 s = splsoftnet(); 2143 ipflow_reap(0); 2144 splx(s); 2145 2146 return (0); 2147} 2148 2149static int 2150sysctl_net_inet_ip_hashsize(SYSCTLFN_ARGS) 2151{ 2152 int error, tmp; 2153 struct sysctlnode node; 2154 2155 node = *rnode; 2156 tmp = ip_hashsize; 2157 node.sysctl_data = &tmp; 2158 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2159 if (error || newp == NULL) 2160 return (error); 2161 2162 if ((tmp & (tmp - 1)) == 0 && tmp != 0) { 2163 /* 2164 * Can only fail due to malloc() 2165 */ 2166 if (ipflow_invalidate_all(tmp)) 2167 return ENOMEM; 2168 } else { 2169 /* 2170 * EINVAL if not a power of 2 2171 */ 2172 return EINVAL; 2173 } 2174 2175 return (0); 2176} 2177#endif /* GATEWAY */ 2178 2179 2180SYSCTL_SETUP(sysctl_net_inet_ip_setup, "sysctl net.inet.ip subtree setup") 2181{ 2182 extern int subnetsarelocal, hostzeroisbroadcast; 2183 2184 sysctl_createv(clog, 0, NULL, NULL, 2185 CTLFLAG_PERMANENT, 2186 CTLTYPE_NODE, "net", NULL, 2187 NULL, 0, NULL, 0, 2188 CTL_NET, CTL_EOL); 2189 sysctl_createv(clog, 0, NULL, NULL, 2190 CTLFLAG_PERMANENT, 2191 CTLTYPE_NODE, "inet", 2192 SYSCTL_DESCR("PF_INET related settings"), 2193 NULL, 0, NULL, 0, 2194 CTL_NET, PF_INET, CTL_EOL); 2195 sysctl_createv(clog, 0, NULL, NULL, 2196 CTLFLAG_PERMANENT, 2197 CTLTYPE_NODE, "ip", 2198 SYSCTL_DESCR("IPv4 related settings"), 2199 NULL, 0, NULL, 0, 2200 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 2201 2202 sysctl_createv(clog, 0, NULL, NULL, 2203 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2204 CTLTYPE_INT, "forwarding", 2205 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 2206 NULL, 0, &ipforwarding, 0, 2207 CTL_NET, PF_INET, IPPROTO_IP, 2208 IPCTL_FORWARDING, CTL_EOL); 2209 sysctl_createv(clog, 0, NULL, NULL, 2210 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2211 CTLTYPE_INT, "redirect", 2212 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 2213 NULL, 0, &ipsendredirects, 0, 2214 CTL_NET, PF_INET, IPPROTO_IP, 2215 IPCTL_SENDREDIRECTS, CTL_EOL); 2216 sysctl_createv(clog, 0, NULL, NULL, 2217 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2218 CTLTYPE_INT, "ttl", 2219 SYSCTL_DESCR("Default TTL for an INET datagram"), 2220 NULL, 0, &ip_defttl, 0, 2221 CTL_NET, PF_INET, IPPROTO_IP, 2222 IPCTL_DEFTTL, CTL_EOL); 2223#ifdef IPCTL_DEFMTU 2224 sysctl_createv(clog, 0, NULL, NULL, 2225 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 2226 CTLTYPE_INT, "mtu", 2227 SYSCTL_DESCR("Default MTA for an INET route"), 2228 NULL, 0, &ip_mtu, 0, 2229 CTL_NET, PF_INET, IPPROTO_IP, 2230 IPCTL_DEFMTU, CTL_EOL); 2231#endif /* IPCTL_DEFMTU */ 2232 sysctl_createv(clog, 0, NULL, NULL, 2233 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2234 CTLTYPE_INT, "forwsrcrt", 2235 SYSCTL_DESCR("Enable forwarding of source-routed " 2236 "datagrams"), 2237 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 2238 CTL_NET, PF_INET, IPPROTO_IP, 2239 IPCTL_FORWSRCRT, CTL_EOL); 2240 sysctl_createv(clog, 0, NULL, NULL, 2241 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2242 CTLTYPE_INT, "directed-broadcast", 2243 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 2244 NULL, 0, &ip_directedbcast, 0, 2245 CTL_NET, PF_INET, IPPROTO_IP, 2246 IPCTL_DIRECTEDBCAST, CTL_EOL); 2247 sysctl_createv(clog, 0, NULL, NULL, 2248 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2249 CTLTYPE_INT, "allowsrcrt", 2250 SYSCTL_DESCR("Accept source-routed datagrams"), 2251 NULL, 0, &ip_allowsrcrt, 0, 2252 CTL_NET, PF_INET, IPPROTO_IP, 2253 IPCTL_ALLOWSRCRT, CTL_EOL); 2254 sysctl_createv(clog, 0, NULL, NULL, 2255 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2256 CTLTYPE_INT, "subnetsarelocal", 2257 SYSCTL_DESCR("Whether logical subnets are considered " 2258 "local"), 2259 NULL, 0, &subnetsarelocal, 0, 2260 CTL_NET, PF_INET, IPPROTO_IP, 2261 IPCTL_SUBNETSARELOCAL, CTL_EOL); 2262 sysctl_createv(clog, 0, NULL, NULL, 2263 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2264 CTLTYPE_INT, "mtudisc", 2265 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 2266 NULL, 0, &ip_mtudisc, 0, 2267 CTL_NET, PF_INET, IPPROTO_IP, 2268 IPCTL_MTUDISC, CTL_EOL); 2269 sysctl_createv(clog, 0, NULL, NULL, 2270 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2271 CTLTYPE_INT, "anonportmin", 2272 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 2273 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 2274 CTL_NET, PF_INET, IPPROTO_IP, 2275 IPCTL_ANONPORTMIN, CTL_EOL); 2276 sysctl_createv(clog, 0, NULL, NULL, 2277 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2278 CTLTYPE_INT, "anonportmax", 2279 SYSCTL_DESCR("Highest ephemeral port number to assign"), 2280 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 2281 CTL_NET, PF_INET, IPPROTO_IP, 2282 IPCTL_ANONPORTMAX, CTL_EOL); 2283 sysctl_createv(clog, 0, NULL, NULL, 2284 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2285 CTLTYPE_INT, "mtudisctimeout", 2286 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 2287 sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0, 2288 CTL_NET, PF_INET, IPPROTO_IP, 2289 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 2290#ifdef GATEWAY 2291 sysctl_createv(clog, 0, NULL, NULL, 2292 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2293 CTLTYPE_INT, "maxflows", 2294 SYSCTL_DESCR("Number of flows for fast forwarding"), 2295 sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0, 2296 CTL_NET, PF_INET, IPPROTO_IP, 2297 IPCTL_MAXFLOWS, CTL_EOL); 2298 sysctl_createv(clog, 0, NULL, NULL, 2299 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2300 CTLTYPE_INT, "hashsize", 2301 SYSCTL_DESCR("Size of hash table for fast forwarding (IPv4)"), 2302 sysctl_net_inet_ip_hashsize, 0, &ip_hashsize, 0, 2303 CTL_NET, PF_INET, IPPROTO_IP, 2304 CTL_CREATE, CTL_EOL); 2305#endif /* GATEWAY */ 2306 sysctl_createv(clog, 0, NULL, NULL, 2307 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2308 CTLTYPE_INT, "hostzerobroadcast", 2309 SYSCTL_DESCR("All zeroes address is broadcast address"), 2310 NULL, 0, &hostzeroisbroadcast, 0, 2311 CTL_NET, PF_INET, IPPROTO_IP, 2312 IPCTL_HOSTZEROBROADCAST, CTL_EOL); 2313#if NGIF > 0 2314 sysctl_createv(clog, 0, NULL, NULL, 2315 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2316 CTLTYPE_INT, "gifttl", 2317 SYSCTL_DESCR("Default TTL for a gif tunnel datagram"), 2318 NULL, 0, &ip_gif_ttl, 0, 2319 CTL_NET, PF_INET, IPPROTO_IP, 2320 IPCTL_GIF_TTL, CTL_EOL); 2321#endif /* NGIF */ 2322#ifndef IPNOPRIVPORTS 2323 sysctl_createv(clog, 0, NULL, NULL, 2324 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2325 CTLTYPE_INT, "lowportmin", 2326 SYSCTL_DESCR("Lowest privileged ephemeral port number " 2327 "to assign"), 2328 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 2329 CTL_NET, PF_INET, IPPROTO_IP, 2330 IPCTL_LOWPORTMIN, CTL_EOL); 2331 sysctl_createv(clog, 0, NULL, NULL, 2332 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2333 CTLTYPE_INT, "lowportmax", 2334 SYSCTL_DESCR("Highest privileged ephemeral port number " 2335 "to assign"), 2336 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 2337 CTL_NET, PF_INET, IPPROTO_IP, 2338 IPCTL_LOWPORTMAX, CTL_EOL); 2339#endif /* IPNOPRIVPORTS */ 2340 sysctl_createv(clog, 0, NULL, NULL, 2341 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2342 CTLTYPE_INT, "maxfragpackets", 2343 SYSCTL_DESCR("Maximum number of fragments to retain for " 2344 "possible reassembly"), 2345 NULL, 0, &ip_maxfragpackets, 0, 2346 CTL_NET, PF_INET, IPPROTO_IP, 2347 IPCTL_MAXFRAGPACKETS, CTL_EOL); 2348#if NGRE > 0 2349 sysctl_createv(clog, 0, NULL, NULL, 2350 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2351 CTLTYPE_INT, "grettl", 2352 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 2353 NULL, 0, &ip_gre_ttl, 0, 2354 CTL_NET, PF_INET, IPPROTO_IP, 2355 IPCTL_GRE_TTL, CTL_EOL); 2356#endif /* NGRE */ 2357 sysctl_createv(clog, 0, NULL, NULL, 2358 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2359 CTLTYPE_INT, "checkinterface", 2360 SYSCTL_DESCR("Enable receive side of Strong ES model " 2361 "from RFC1122"), 2362 NULL, 0, &ip_checkinterface, 0, 2363 CTL_NET, PF_INET, IPPROTO_IP, 2364 IPCTL_CHECKINTERFACE, CTL_EOL); 2365 sysctl_createv(clog, 0, NULL, NULL, 2366 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2367 CTLTYPE_INT, "random_id", 2368 SYSCTL_DESCR("Assign random ip_id values"), 2369 NULL, 0, &ip_do_randomid, 0, 2370 CTL_NET, PF_INET, IPPROTO_IP, 2371 IPCTL_RANDOMID, CTL_EOL); 2372 sysctl_createv(clog, 0, NULL, NULL, 2373 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2374 CTLTYPE_INT, "do_loopback_cksum", 2375 SYSCTL_DESCR("Perform IP checksum on loopback"), 2376 NULL, 0, &ip_do_loopback_cksum, 0, 2377 CTL_NET, PF_INET, IPPROTO_IP, 2378 IPCTL_LOOPBACKCKSUM, CTL_EOL); 2379 sysctl_createv(clog, 0, NULL, NULL, 2380 CTLFLAG_PERMANENT, 2381 CTLTYPE_STRUCT, "stats", 2382 SYSCTL_DESCR("IP statistics"), 2383 NULL, 0, &ipstat, sizeof(ipstat), 2384 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 2385 CTL_EOL); 2386} 2387