ip_input.c revision 1.287
1/* $NetBSD: ip_input.c,v 1.287 2010/07/09 18:42:46 rmind 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 * 49 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 50 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 51 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 52 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 53 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 54 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 55 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 56 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 57 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 58 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 59 * POSSIBILITY OF SUCH DAMAGE. 60 */ 61 62/* 63 * Copyright (c) 1982, 1986, 1988, 1993 64 * The Regents of the University of California. All rights reserved. 65 * 66 * Redistribution and use in source and binary forms, with or without 67 * modification, are permitted provided that the following conditions 68 * are met: 69 * 1. Redistributions of source code must retain the above copyright 70 * notice, this list of conditions and the following disclaimer. 71 * 2. Redistributions in binary form must reproduce the above copyright 72 * notice, this list of conditions and the following disclaimer in the 73 * documentation and/or other materials provided with the distribution. 74 * 3. Neither the name of the University nor the names of its contributors 75 * may be used to endorse or promote products derived from this software 76 * without specific prior written permission. 77 * 78 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 79 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 80 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 81 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 82 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 83 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 84 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 85 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 86 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 87 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 88 * SUCH DAMAGE. 89 * 90 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 91 */ 92 93#include <sys/cdefs.h> 94__KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.287 2010/07/09 18:42:46 rmind Exp $"); 95 96#include "opt_inet.h" 97#include "opt_compat_netbsd.h" 98#include "opt_gateway.h" 99#include "opt_pfil_hooks.h" 100#include "opt_ipsec.h" 101#include "opt_mrouting.h" 102#include "opt_mbuftrace.h" 103#include "opt_inet_csum.h" 104 105#include <sys/param.h> 106#include <sys/systm.h> 107#include <sys/malloc.h> 108#include <sys/mbuf.h> 109#include <sys/domain.h> 110#include <sys/protosw.h> 111#include <sys/socket.h> 112#include <sys/socketvar.h> 113#include <sys/errno.h> 114#include <sys/time.h> 115#include <sys/kernel.h> 116#include <sys/pool.h> 117#include <sys/sysctl.h> 118#include <sys/kauth.h> 119 120#include <net/if.h> 121#include <net/if_dl.h> 122#include <net/route.h> 123#include <net/pfil.h> 124 125#include <netinet/in.h> 126#include <netinet/in_systm.h> 127#include <netinet/ip.h> 128#include <netinet/in_pcb.h> 129#include <netinet/in_proto.h> 130#include <netinet/in_var.h> 131#include <netinet/ip_var.h> 132#include <netinet/ip_private.h> 133#include <netinet/ip_icmp.h> 134/* just for gif_ttl */ 135#include <netinet/in_gif.h> 136#include "gif.h" 137#include <net/if_gre.h> 138#include "gre.h" 139 140#ifdef MROUTING 141#include <netinet/ip_mroute.h> 142#endif 143 144#ifdef IPSEC 145#include <netinet6/ipsec.h> 146#include <netinet6/ipsec_private.h> 147#include <netkey/key.h> 148#endif 149#ifdef FAST_IPSEC 150#include <netipsec/ipsec.h> 151#include <netipsec/key.h> 152#endif /* FAST_IPSEC*/ 153 154#ifndef IPFORWARDING 155#ifdef GATEWAY 156#define IPFORWARDING 1 /* forward IP packets not for us */ 157#else /* GATEWAY */ 158#define IPFORWARDING 0 /* don't forward IP packets not for us */ 159#endif /* GATEWAY */ 160#endif /* IPFORWARDING */ 161#ifndef IPSENDREDIRECTS 162#define IPSENDREDIRECTS 1 163#endif 164#ifndef IPFORWSRCRT 165#define IPFORWSRCRT 1 /* forward source-routed packets */ 166#endif 167#ifndef IPALLOWSRCRT 168#define IPALLOWSRCRT 1 /* allow source-routed packets */ 169#endif 170#ifndef IPMTUDISC 171#define IPMTUDISC 1 172#endif 173#ifndef IPMTUDISCTIMEOUT 174#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 175#endif 176 177#ifdef COMPAT_50 178#include <compat/sys/time.h> 179#include <compat/sys/socket.h> 180#endif 181 182/* 183 * Note: DIRECTED_BROADCAST is handled this way so that previous 184 * configuration using this option will Just Work. 185 */ 186#ifndef IPDIRECTEDBCAST 187#ifdef DIRECTED_BROADCAST 188#define IPDIRECTEDBCAST 1 189#else 190#define IPDIRECTEDBCAST 0 191#endif /* DIRECTED_BROADCAST */ 192#endif /* IPDIRECTEDBCAST */ 193int ipforwarding = IPFORWARDING; 194int ipsendredirects = IPSENDREDIRECTS; 195int ip_defttl = IPDEFTTL; 196int ip_forwsrcrt = IPFORWSRCRT; 197int ip_directedbcast = IPDIRECTEDBCAST; 198int ip_allowsrcrt = IPALLOWSRCRT; 199int ip_mtudisc = IPMTUDISC; 200int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 201#ifdef DIAGNOSTIC 202int ipprintfs = 0; 203#endif 204 205int ip_do_randomid = 0; 206 207/* 208 * XXX - Setting ip_checkinterface mostly implements the receive side of 209 * the Strong ES model described in RFC 1122, but since the routing table 210 * and transmit implementation do not implement the Strong ES model, 211 * setting this to 1 results in an odd hybrid. 212 * 213 * XXX - ip_checkinterface currently must be disabled if you use ipnat 214 * to translate the destination address to another local interface. 215 * 216 * XXX - ip_checkinterface must be disabled if you add IP aliases 217 * to the loopback interface instead of the interface where the 218 * packets for those addresses are received. 219 */ 220int ip_checkinterface = 0; 221 222 223struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 224 225int ipqmaxlen = IFQ_MAXLEN; 226u_long in_ifaddrhash; /* size of hash table - 1 */ 227int in_ifaddrentries; /* total number of addrs */ 228struct in_ifaddrhead in_ifaddrhead; 229struct in_ifaddrhashhead *in_ifaddrhashtbl; 230u_long in_multihash; /* size of hash table - 1 */ 231int in_multientries; /* total number of addrs */ 232struct in_multihashhead *in_multihashtbl; 233struct ifqueue ipintrq; 234 235uint16_t ip_id; 236 237percpu_t *ipstat_percpu; 238 239#ifdef PFIL_HOOKS 240struct pfil_head inet_pfil_hook; 241#endif 242 243/* 244 * Cached copy of nmbclusters. If nbclusters is different, 245 * recalculate IP parameters derived from nmbclusters. 246 */ 247static int ip_nmbclusters; /* copy of nmbclusters */ 248static void ip_nmbclusters_changed(void); /* recalc limits */ 249 250#define CHECK_NMBCLUSTER_PARAMS() \ 251do { \ 252 if (__predict_false(ip_nmbclusters != nmbclusters)) \ 253 ip_nmbclusters_changed(); \ 254} while (/*CONSTCOND*/0) 255 256/* IP datagram reassembly queues (hashed) */ 257#define IPREASS_NHASH_LOG2 6 258#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 259#define IPREASS_HMASK (IPREASS_NHASH - 1) 260#define IPREASS_HASH(x,y) \ 261 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 262struct ipqhead ipq[IPREASS_NHASH]; 263int ipq_locked; 264static int ip_nfragpackets; /* packets in reass queue */ 265static int ip_nfrags; /* total fragments in reass queues */ 266 267int ip_maxfragpackets = 200; /* limit on packets. XXX sysctl */ 268int ip_maxfrags; /* limit on fragments. XXX sysctl */ 269 270 271/* 272 * Additive-Increase/Multiplicative-Decrease (AIMD) strategy for 273 * IP reassembly queue buffer managment. 274 * 275 * We keep a count of total IP fragments (NB: not fragmented packets!) 276 * awaiting reassembly (ip_nfrags) and a limit (ip_maxfrags) on fragments. 277 * If ip_nfrags exceeds ip_maxfrags the limit, we drop half the 278 * total fragments in reassembly queues.This AIMD policy avoids 279 * repeatedly deleting single packets under heavy fragmentation load 280 * (e.g., from lossy NFS peers). 281 */ 282static u_int ip_reass_ttl_decr(u_int ticks); 283static void ip_reass_drophalf(void); 284 285 286static inline int ipq_lock_try(void); 287static inline void ipq_unlock(void); 288 289static inline int 290ipq_lock_try(void) 291{ 292 int s; 293 294 /* 295 * Use splvm() -- we're blocking things that would cause 296 * mbuf allocation. 297 */ 298 s = splvm(); 299 if (ipq_locked) { 300 splx(s); 301 return (0); 302 } 303 ipq_locked = 1; 304 splx(s); 305 return (1); 306} 307 308static inline void 309ipq_unlock(void) 310{ 311 int s; 312 313 s = splvm(); 314 ipq_locked = 0; 315 splx(s); 316} 317 318#ifdef DIAGNOSTIC 319#define IPQ_LOCK() \ 320do { \ 321 if (ipq_lock_try() == 0) { \ 322 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 323 panic("ipq_lock"); \ 324 } \ 325} while (/*CONSTCOND*/ 0) 326#define IPQ_LOCK_CHECK() \ 327do { \ 328 if (ipq_locked == 0) { \ 329 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 330 panic("ipq lock check"); \ 331 } \ 332} while (/*CONSTCOND*/ 0) 333#else 334#define IPQ_LOCK() (void) ipq_lock_try() 335#define IPQ_LOCK_CHECK() /* nothing */ 336#endif 337 338#define IPQ_UNLOCK() ipq_unlock() 339 340struct pool inmulti_pool; 341struct pool ipqent_pool; 342 343#ifdef INET_CSUM_COUNTERS 344#include <sys/device.h> 345 346struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 347 NULL, "inet", "hwcsum bad"); 348struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 349 NULL, "inet", "hwcsum ok"); 350struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 351 NULL, "inet", "swcsum"); 352 353#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 354 355EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 356EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 357EVCNT_ATTACH_STATIC(ip_swcsum); 358 359#else 360 361#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 362 363#endif /* INET_CSUM_COUNTERS */ 364 365/* 366 * We need to save the IP options in case a protocol wants to respond 367 * to an incoming packet over the same route if the packet got here 368 * using IP source routing. This allows connection establishment and 369 * maintenance when the remote end is on a network that is not known 370 * to us. 371 */ 372int ip_nhops = 0; 373static struct ip_srcrt { 374 struct in_addr dst; /* final destination */ 375 char nop; /* one NOP to align */ 376 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 377 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 378} ip_srcrt; 379 380static void save_rte(u_char *, struct in_addr); 381 382#ifdef MBUFTRACE 383struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx"); 384struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx"); 385#endif 386 387static void sysctl_net_inet_ip_setup(struct sysctllog **); 388 389/* 390 * Compute IP limits derived from the value of nmbclusters. 391 */ 392static void 393ip_nmbclusters_changed(void) 394{ 395 ip_maxfrags = nmbclusters / 4; 396 ip_nmbclusters = nmbclusters; 397} 398 399/* 400 * IP initialization: fill in IP protocol switch table. 401 * All protocols not implemented in kernel go to raw IP protocol handler. 402 */ 403void 404ip_init(void) 405{ 406 const struct protosw *pr; 407 int i; 408 409 sysctl_net_inet_ip_setup(NULL); 410 411 pool_init(&inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl", 412 NULL, IPL_SOFTNET); 413 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 414 NULL, IPL_VM); 415 416 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 417 if (pr == 0) 418 panic("ip_init"); 419 for (i = 0; i < IPPROTO_MAX; i++) 420 ip_protox[i] = pr - inetsw; 421 for (pr = inetdomain.dom_protosw; 422 pr < inetdomain.dom_protoswNPROTOSW; pr++) 423 if (pr->pr_domain->dom_family == PF_INET && 424 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 425 ip_protox[pr->pr_protocol] = pr - inetsw; 426 427 for (i = 0; i < IPREASS_NHASH; i++) 428 LIST_INIT(&ipq[i]); 429 430 ip_initid(); 431 ip_id = time_second & 0xfffff; 432 433 ipintrq.ifq_maxlen = ipqmaxlen; 434 ip_nmbclusters_changed(); 435 436 TAILQ_INIT(&in_ifaddrhead); 437 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true, 438 &in_ifaddrhash); 439 in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true, 440 &in_multihash); 441 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 442#ifdef GATEWAY 443 ipflow_init(ip_hashsize); 444#endif 445 446#ifdef PFIL_HOOKS 447 /* Register our Packet Filter hook. */ 448 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 449 inet_pfil_hook.ph_af = AF_INET; 450 i = pfil_head_register(&inet_pfil_hook); 451 if (i != 0) 452 printf("ip_init: WARNING: unable to register pfil hook, " 453 "error %d\n", i); 454#endif /* PFIL_HOOKS */ 455 456#ifdef MBUFTRACE 457 MOWNER_ATTACH(&ip_tx_mowner); 458 MOWNER_ATTACH(&ip_rx_mowner); 459#endif /* MBUFTRACE */ 460 461 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS); 462} 463 464struct sockaddr_in ipaddr = { 465 .sin_len = sizeof(ipaddr), 466 .sin_family = AF_INET, 467}; 468struct route ipforward_rt; 469 470/* 471 * IP software interrupt routine 472 */ 473void 474ipintr(void) 475{ 476 int s; 477 struct mbuf *m; 478 struct ifqueue lcl_intrq; 479 480 memset(&lcl_intrq, 0, sizeof(lcl_intrq)); 481 ipintrq.ifq_maxlen = ipqmaxlen; 482 483 mutex_enter(softnet_lock); 484 KERNEL_LOCK(1, NULL); 485 if (!IF_IS_EMPTY(&ipintrq)) { 486 s = splnet(); 487 488 /* Take existing queue onto stack */ 489 lcl_intrq = ipintrq; 490 491 /* Zero out global queue, preserving maxlen and drops */ 492 ipintrq.ifq_head = NULL; 493 ipintrq.ifq_tail = NULL; 494 ipintrq.ifq_len = 0; 495 ipintrq.ifq_maxlen = lcl_intrq.ifq_maxlen; 496 ipintrq.ifq_drops = lcl_intrq.ifq_drops; 497 498 splx(s); 499 } 500 KERNEL_UNLOCK_ONE(NULL); 501 while (!IF_IS_EMPTY(&lcl_intrq)) { 502 IF_DEQUEUE(&lcl_intrq, m); 503 if (m == NULL) 504 break; 505 ip_input(m); 506 } 507 mutex_exit(softnet_lock); 508} 509 510/* 511 * Ip input routine. Checksum and byte swap header. If fragmented 512 * try to reassemble. Process options. Pass to next level. 513 */ 514void 515ip_input(struct mbuf *m) 516{ 517 struct ip *ip = NULL; 518 struct ipq *fp; 519 struct in_ifaddr *ia; 520 struct ifaddr *ifa; 521 struct ipqent *ipqe; 522 int hlen = 0, mff, len; 523 int downmatch; 524 int checkif; 525 int srcrt = 0; 526 int s; 527 u_int hash; 528#ifdef FAST_IPSEC 529 struct m_tag *mtag; 530 struct tdb_ident *tdbi; 531 struct secpolicy *sp; 532 int error; 533#endif /* FAST_IPSEC */ 534 535 MCLAIM(m, &ip_rx_mowner); 536#ifdef DIAGNOSTIC 537 if ((m->m_flags & M_PKTHDR) == 0) 538 panic("ipintr no HDR"); 539#endif 540 541 /* 542 * If no IP addresses have been set yet but the interfaces 543 * are receiving, can't do anything with incoming packets yet. 544 */ 545 if (TAILQ_FIRST(&in_ifaddrhead) == 0) 546 goto bad; 547 IP_STATINC(IP_STAT_TOTAL); 548 /* 549 * If the IP header is not aligned, slurp it up into a new 550 * mbuf with space for link headers, in the event we forward 551 * it. Otherwise, if it is aligned, make sure the entire 552 * base IP header is in the first mbuf of the chain. 553 */ 554 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 555 if ((m = m_copyup(m, sizeof(struct ip), 556 (max_linkhdr + 3) & ~3)) == NULL) { 557 /* XXXJRT new stat, please */ 558 IP_STATINC(IP_STAT_TOOSMALL); 559 return; 560 } 561 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 562 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 563 IP_STATINC(IP_STAT_TOOSMALL); 564 return; 565 } 566 } 567 ip = mtod(m, struct ip *); 568 if (ip->ip_v != IPVERSION) { 569 IP_STATINC(IP_STAT_BADVERS); 570 goto bad; 571 } 572 hlen = ip->ip_hl << 2; 573 if (hlen < sizeof(struct ip)) { /* minimum header length */ 574 IP_STATINC(IP_STAT_BADHLEN); 575 goto bad; 576 } 577 if (hlen > m->m_len) { 578 if ((m = m_pullup(m, hlen)) == 0) { 579 IP_STATINC(IP_STAT_BADHLEN); 580 return; 581 } 582 ip = mtod(m, struct ip *); 583 } 584 585 /* 586 * RFC1122: packets with a multicast source address are 587 * not allowed. 588 */ 589 if (IN_MULTICAST(ip->ip_src.s_addr)) { 590 IP_STATINC(IP_STAT_BADADDR); 591 goto bad; 592 } 593 594 /* 127/8 must not appear on wire - RFC1122 */ 595 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 596 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 597 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 598 IP_STATINC(IP_STAT_BADADDR); 599 goto bad; 600 } 601 } 602 603 switch (m->m_pkthdr.csum_flags & 604 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) | 605 M_CSUM_IPv4_BAD)) { 606 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 607 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 608 goto badcsum; 609 610 case M_CSUM_IPv4: 611 /* Checksum was okay. */ 612 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 613 break; 614 615 default: 616 /* 617 * Must compute it ourselves. Maybe skip checksum on 618 * loopback interfaces. 619 */ 620 if (__predict_true(!(m->m_pkthdr.rcvif->if_flags & 621 IFF_LOOPBACK) || ip_do_loopback_cksum)) { 622 INET_CSUM_COUNTER_INCR(&ip_swcsum); 623 if (in_cksum(m, hlen) != 0) 624 goto badcsum; 625 } 626 break; 627 } 628 629 /* Retrieve the packet length. */ 630 len = ntohs(ip->ip_len); 631 632 /* 633 * Check for additional length bogosity 634 */ 635 if (len < hlen) { 636 IP_STATINC(IP_STAT_BADLEN); 637 goto bad; 638 } 639 640 /* 641 * Check that the amount of data in the buffers 642 * is as at least much as the IP header would have us expect. 643 * Trim mbufs if longer than we expect. 644 * Drop packet if shorter than we expect. 645 */ 646 if (m->m_pkthdr.len < len) { 647 IP_STATINC(IP_STAT_TOOSHORT); 648 goto bad; 649 } 650 if (m->m_pkthdr.len > len) { 651 if (m->m_len == m->m_pkthdr.len) { 652 m->m_len = len; 653 m->m_pkthdr.len = len; 654 } else 655 m_adj(m, len - m->m_pkthdr.len); 656 } 657 658#if defined(IPSEC) 659 /* ipflow (IP fast forwarding) is not compatible with IPsec. */ 660 m->m_flags &= ~M_CANFASTFWD; 661#else 662 /* 663 * Assume that we can create a fast-forward IP flow entry 664 * based on this packet. 665 */ 666 m->m_flags |= M_CANFASTFWD; 667#endif 668 669#ifdef PFIL_HOOKS 670 /* 671 * Run through list of hooks for input packets. If there are any 672 * filters which require that additional packets in the flow are 673 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 674 * Note that filters must _never_ set this flag, as another filter 675 * in the list may have previously cleared it. 676 */ 677 /* 678 * let ipfilter look at packet on the wire, 679 * not the decapsulated packet. 680 */ 681#ifdef IPSEC 682 if (!ipsec_getnhist(m)) 683#elif defined(FAST_IPSEC) 684 if (!ipsec_indone(m)) 685#else 686 if (1) 687#endif 688 { 689 struct in_addr odst; 690 691 odst = ip->ip_dst; 692 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 693 PFIL_IN) != 0) 694 return; 695 if (m == NULL) 696 return; 697 ip = mtod(m, struct ip *); 698 hlen = ip->ip_hl << 2; 699 /* 700 * XXX The setting of "srcrt" here is to prevent ip_forward() 701 * from generating ICMP redirects for packets that have 702 * been redirected by a hook back out on to the same LAN that 703 * they came from and is not an indication that the packet 704 * is being inffluenced by source routing options. This 705 * allows things like 706 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 707 * where tlp0 is both on the 1.1.1.0/24 network and is the 708 * default route for hosts on 1.1.1.0/24. Of course this 709 * also requires a "map tlp0 ..." to complete the story. 710 * One might argue whether or not this kind of network config. 711 * should be supported in this manner... 712 */ 713 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 714 } 715#endif /* PFIL_HOOKS */ 716 717#ifdef ALTQ 718 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 719 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 720 /* packet dropped by traffic conditioner */ 721 return; 722 } 723#endif 724 725 /* 726 * Process options and, if not destined for us, 727 * ship it on. ip_dooptions returns 1 when an 728 * error was detected (causing an icmp message 729 * to be sent and the original packet to be freed). 730 */ 731 ip_nhops = 0; /* for source routed packets */ 732 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 733 return; 734 735 /* 736 * Enable a consistency check between the destination address 737 * and the arrival interface for a unicast packet (the RFC 1122 738 * strong ES model) if IP forwarding is disabled and the packet 739 * is not locally generated. 740 * 741 * XXX - Checking also should be disabled if the destination 742 * address is ipnat'ed to a different interface. 743 * 744 * XXX - Checking is incompatible with IP aliases added 745 * to the loopback interface instead of the interface where 746 * the packets are received. 747 * 748 * XXX - We need to add a per ifaddr flag for this so that 749 * we get finer grain control. 750 */ 751 checkif = ip_checkinterface && (ipforwarding == 0) && 752 (m->m_pkthdr.rcvif != NULL) && 753 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0); 754 755 /* 756 * Check our list of addresses, to see if the packet is for us. 757 * 758 * Traditional 4.4BSD did not consult IFF_UP at all. 759 * The behavior here is to treat addresses on !IFF_UP interface 760 * as not mine. 761 */ 762 downmatch = 0; 763 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 764 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 765 if (checkif && ia->ia_ifp != m->m_pkthdr.rcvif) 766 continue; 767 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 768 break; 769 else 770 downmatch++; 771 } 772 } 773 if (ia != NULL) 774 goto ours; 775 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 776 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) { 777 if (ifa->ifa_addr->sa_family != AF_INET) 778 continue; 779 ia = ifatoia(ifa); 780 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 781 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 782 /* 783 * Look for all-0's host part (old broadcast addr), 784 * either for subnet or net. 785 */ 786 ip->ip_dst.s_addr == ia->ia_subnet || 787 ip->ip_dst.s_addr == ia->ia_net) 788 goto ours; 789 /* 790 * An interface with IP address zero accepts 791 * all packets that arrive on that interface. 792 */ 793 if (in_nullhost(ia->ia_addr.sin_addr)) 794 goto ours; 795 } 796 } 797 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 798 struct in_multi *inm; 799#ifdef MROUTING 800 extern struct socket *ip_mrouter; 801 802 if (ip_mrouter) { 803 /* 804 * If we are acting as a multicast router, all 805 * incoming multicast packets are passed to the 806 * kernel-level multicast forwarding function. 807 * The packet is returned (relatively) intact; if 808 * ip_mforward() returns a non-zero value, the packet 809 * must be discarded, else it may be accepted below. 810 * 811 * (The IP ident field is put in the same byte order 812 * as expected when ip_mforward() is called from 813 * ip_output().) 814 */ 815 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 816 IP_STATINC(IP_STAT_CANTFORWARD); 817 m_freem(m); 818 return; 819 } 820 821 /* 822 * The process-level routing demon needs to receive 823 * all multicast IGMP packets, whether or not this 824 * host belongs to their destination groups. 825 */ 826 if (ip->ip_p == IPPROTO_IGMP) 827 goto ours; 828 IP_STATINC(IP_STAT_CANTFORWARD); 829 } 830#endif 831 /* 832 * See if we belong to the destination multicast group on the 833 * arrival interface. 834 */ 835 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 836 if (inm == NULL) { 837 IP_STATINC(IP_STAT_CANTFORWARD); 838 m_freem(m); 839 return; 840 } 841 goto ours; 842 } 843 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 844 in_nullhost(ip->ip_dst)) 845 goto ours; 846 847 /* 848 * Not for us; forward if possible and desirable. 849 */ 850 if (ipforwarding == 0) { 851 IP_STATINC(IP_STAT_CANTFORWARD); 852 m_freem(m); 853 } else { 854 /* 855 * If ip_dst matched any of my address on !IFF_UP interface, 856 * and there's no IFF_UP interface that matches ip_dst, 857 * send icmp unreach. Forwarding it will result in in-kernel 858 * forwarding loop till TTL goes to 0. 859 */ 860 if (downmatch) { 861 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 862 IP_STATINC(IP_STAT_CANTFORWARD); 863 return; 864 } 865#ifdef IPSEC 866 if (ipsec4_in_reject(m, NULL)) { 867 IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); 868 goto bad; 869 } 870#endif 871#ifdef FAST_IPSEC 872 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 873 s = splsoftnet(); 874 if (mtag != NULL) { 875 tdbi = (struct tdb_ident *)(mtag + 1); 876 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 877 } else { 878 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 879 IP_FORWARDING, &error); 880 } 881 if (sp == NULL) { /* NB: can happen if error */ 882 splx(s); 883 /*XXX error stat???*/ 884 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 885 goto bad; 886 } 887 888 /* 889 * Check security policy against packet attributes. 890 */ 891 error = ipsec_in_reject(sp, m); 892 KEY_FREESP(&sp); 893 splx(s); 894 if (error) { 895 IP_STATINC(IP_STAT_CANTFORWARD); 896 goto bad; 897 } 898 899 /* 900 * Peek at the outbound SP for this packet to determine if 901 * it's a Fast Forward candidate. 902 */ 903 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 904 if (mtag != NULL) 905 m->m_flags &= ~M_CANFASTFWD; 906 else { 907 s = splsoftnet(); 908 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, 909 (IP_FORWARDING | 910 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 911 &error, NULL); 912 if (sp != NULL) { 913 m->m_flags &= ~M_CANFASTFWD; 914 KEY_FREESP(&sp); 915 } 916 splx(s); 917 } 918#endif /* FAST_IPSEC */ 919 920 ip_forward(m, srcrt); 921 } 922 return; 923 924ours: 925 /* 926 * If offset or IP_MF are set, must reassemble. 927 * Otherwise, nothing need be done. 928 * (We could look in the reassembly queue to see 929 * if the packet was previously fragmented, 930 * but it's not worth the time; just let them time out.) 931 */ 932 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 933 u_int off; 934 /* 935 * Prevent TCP blind data attacks by not allowing non-initial 936 * fragments to start at less than 68 bytes (minimal fragment 937 * size) and making sure the first fragment is at least 68 938 * bytes. 939 */ 940 off = (ntohs(ip->ip_off) & IP_OFFMASK) << 3; 941 if ((off > 0 ? off + hlen : len) < IP_MINFRAGSIZE - 1) { 942 IP_STATINC(IP_STAT_BADFRAGS); 943 goto bad; 944 } 945 946 /* 947 * Adjust ip_len to not reflect header, 948 * set ipqe_mff if more fragments are expected, 949 * convert offset of this to bytes. 950 */ 951 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 952 mff = (ip->ip_off & htons(IP_MF)) != 0; 953 if (mff) { 954 /* 955 * Make sure that fragments have a data length 956 * that's a non-zero multiple of 8 bytes. 957 */ 958 if (ntohs(ip->ip_len) == 0 || 959 (ntohs(ip->ip_len) & 0x7) != 0) { 960 IP_STATINC(IP_STAT_BADFRAGS); 961 goto bad; 962 } 963 } 964 ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3); 965 966 /* 967 * Look for queue of fragments of this datagram. 968 */ 969 IPQ_LOCK(); 970 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 971 LIST_FOREACH(fp, &ipq[hash], ipq_q) { 972 if (ip->ip_id != fp->ipq_id) 973 continue; 974 if (!in_hosteq(ip->ip_src, fp->ipq_src)) 975 continue; 976 if (!in_hosteq(ip->ip_dst, fp->ipq_dst)) 977 continue; 978 if (ip->ip_p != fp->ipq_p) 979 continue; 980 /* 981 * Make sure the TOS is matches previous fragments. 982 */ 983 if (ip->ip_tos != fp->ipq_tos) { 984 IP_STATINC(IP_STAT_BADFRAGS); 985 IPQ_UNLOCK(); 986 goto bad; 987 } 988 break; 989 } 990 991 /* 992 * If datagram marked as having more fragments 993 * or if this is not the first fragment, 994 * attempt reassembly; if it succeeds, proceed. 995 */ 996 if (mff || ip->ip_off != htons(0)) { 997 IP_STATINC(IP_STAT_FRAGMENTS); 998 s = splvm(); 999 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 1000 splx(s); 1001 if (ipqe == NULL) { 1002 IP_STATINC(IP_STAT_RCVMEMDROP); 1003 IPQ_UNLOCK(); 1004 goto bad; 1005 } 1006 ipqe->ipqe_mff = mff; 1007 ipqe->ipqe_m = m; 1008 ipqe->ipqe_ip = ip; 1009 m = ip_reass(ipqe, fp, &ipq[hash]); 1010 if (m == NULL) { 1011 IPQ_UNLOCK(); 1012 return; 1013 } 1014 IP_STATINC(IP_STAT_REASSEMBLED); 1015 ip = mtod(m, struct ip *); 1016 hlen = ip->ip_hl << 2; 1017 ip->ip_len = htons(ntohs(ip->ip_len) + hlen); 1018 } else if (fp) { 1019 ip_freef(fp); 1020 } 1021 IPQ_UNLOCK(); 1022 } 1023 1024#if defined(IPSEC) 1025 /* 1026 * enforce IPsec policy checking if we are seeing last header. 1027 * note that we do not visit this with protocols with pcb layer 1028 * code - like udp/tcp/raw ip. 1029 */ 1030 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 1031 ipsec4_in_reject(m, NULL)) { 1032 IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); 1033 goto bad; 1034 } 1035#endif 1036#ifdef FAST_IPSEC 1037 /* 1038 * enforce IPsec policy checking if we are seeing last header. 1039 * note that we do not visit this with protocols with pcb layer 1040 * code - like udp/tcp/raw ip. 1041 */ 1042 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 1043 /* 1044 * Check if the packet has already had IPsec processing 1045 * done. If so, then just pass it along. This tag gets 1046 * set during AH, ESP, etc. input handling, before the 1047 * packet is returned to the ip input queue for delivery. 1048 */ 1049 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 1050 s = splsoftnet(); 1051 if (mtag != NULL) { 1052 tdbi = (struct tdb_ident *)(mtag + 1); 1053 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 1054 } else { 1055 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 1056 IP_FORWARDING, &error); 1057 } 1058 if (sp != NULL) { 1059 /* 1060 * Check security policy against packet attributes. 1061 */ 1062 error = ipsec_in_reject(sp, m); 1063 KEY_FREESP(&sp); 1064 } else { 1065 /* XXX error stat??? */ 1066 error = EINVAL; 1067DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 1068 } 1069 splx(s); 1070 if (error) 1071 goto bad; 1072 } 1073#endif /* FAST_IPSEC */ 1074 1075 /* 1076 * Switch out to protocol's input routine. 1077 */ 1078#if IFA_STATS 1079 if (ia && ip) 1080 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 1081#endif 1082 IP_STATINC(IP_STAT_DELIVERED); 1083 { 1084 int off = hlen, nh = ip->ip_p; 1085 1086 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 1087 return; 1088 } 1089bad: 1090 m_freem(m); 1091 return; 1092 1093badcsum: 1094 IP_STATINC(IP_STAT_BADSUM); 1095 m_freem(m); 1096} 1097 1098/* 1099 * Take incoming datagram fragment and try to 1100 * reassemble it into whole datagram. If a chain for 1101 * reassembly of this datagram already exists, then it 1102 * is given as fp; otherwise have to make a chain. 1103 */ 1104struct mbuf * 1105ip_reass(struct ipqent *ipqe, struct ipq *fp, struct ipqhead *ipqhead) 1106{ 1107 struct mbuf *m = ipqe->ipqe_m; 1108 struct ipqent *nq, *p, *q; 1109 struct ip *ip; 1110 struct mbuf *t; 1111 int hlen = ipqe->ipqe_ip->ip_hl << 2; 1112 int i, next, s; 1113 1114 IPQ_LOCK_CHECK(); 1115 1116 /* 1117 * Presence of header sizes in mbufs 1118 * would confuse code below. 1119 */ 1120 m->m_data += hlen; 1121 m->m_len -= hlen; 1122 1123#ifdef notyet 1124 /* make sure fragment limit is up-to-date */ 1125 CHECK_NMBCLUSTER_PARAMS(); 1126 1127 /* If we have too many fragments, drop the older half. */ 1128 if (ip_nfrags >= ip_maxfrags) 1129 ip_reass_drophalf(void); 1130#endif 1131 1132 /* 1133 * We are about to add a fragment; increment frag count. 1134 */ 1135 ip_nfrags++; 1136 1137 /* 1138 * If first fragment to arrive, create a reassembly queue. 1139 */ 1140 if (fp == 0) { 1141 /* 1142 * Enforce upper bound on number of fragmented packets 1143 * for which we attempt reassembly; 1144 * If maxfrag is 0, never accept fragments. 1145 * If maxfrag is -1, accept all fragments without limitation. 1146 */ 1147 if (ip_maxfragpackets < 0) 1148 ; 1149 else if (ip_nfragpackets >= ip_maxfragpackets) 1150 goto dropfrag; 1151 ip_nfragpackets++; 1152 fp = malloc(sizeof (struct ipq), M_FTABLE, M_NOWAIT); 1153 if (fp == NULL) 1154 goto dropfrag; 1155 LIST_INSERT_HEAD(ipqhead, fp, ipq_q); 1156 fp->ipq_nfrags = 1; 1157 fp->ipq_ttl = IPFRAGTTL; 1158 fp->ipq_p = ipqe->ipqe_ip->ip_p; 1159 fp->ipq_id = ipqe->ipqe_ip->ip_id; 1160 fp->ipq_tos = ipqe->ipqe_ip->ip_tos; 1161 TAILQ_INIT(&fp->ipq_fragq); 1162 fp->ipq_src = ipqe->ipqe_ip->ip_src; 1163 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 1164 p = NULL; 1165 goto insert; 1166 } else { 1167 fp->ipq_nfrags++; 1168 } 1169 1170 /* 1171 * Find a segment which begins after this one does. 1172 */ 1173 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1174 p = q, q = TAILQ_NEXT(q, ipqe_q)) 1175 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off)) 1176 break; 1177 1178 /* 1179 * If there is a preceding segment, it may provide some of 1180 * our data already. If so, drop the data from the incoming 1181 * segment. If it provides all of our data, drop us. 1182 */ 1183 if (p != NULL) { 1184 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) - 1185 ntohs(ipqe->ipqe_ip->ip_off); 1186 if (i > 0) { 1187 if (i >= ntohs(ipqe->ipqe_ip->ip_len)) 1188 goto dropfrag; 1189 m_adj(ipqe->ipqe_m, i); 1190 ipqe->ipqe_ip->ip_off = 1191 htons(ntohs(ipqe->ipqe_ip->ip_off) + i); 1192 ipqe->ipqe_ip->ip_len = 1193 htons(ntohs(ipqe->ipqe_ip->ip_len) - i); 1194 } 1195 } 1196 1197 /* 1198 * While we overlap succeeding segments trim them or, 1199 * if they are completely covered, dequeue them. 1200 */ 1201 for (; q != NULL && 1202 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) > 1203 ntohs(q->ipqe_ip->ip_off); q = nq) { 1204 i = (ntohs(ipqe->ipqe_ip->ip_off) + 1205 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off); 1206 if (i < ntohs(q->ipqe_ip->ip_len)) { 1207 q->ipqe_ip->ip_len = 1208 htons(ntohs(q->ipqe_ip->ip_len) - i); 1209 q->ipqe_ip->ip_off = 1210 htons(ntohs(q->ipqe_ip->ip_off) + i); 1211 m_adj(q->ipqe_m, i); 1212 break; 1213 } 1214 nq = TAILQ_NEXT(q, ipqe_q); 1215 m_freem(q->ipqe_m); 1216 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1217 s = splvm(); 1218 pool_put(&ipqent_pool, q); 1219 splx(s); 1220 fp->ipq_nfrags--; 1221 ip_nfrags--; 1222 } 1223 1224insert: 1225 /* 1226 * Stick new segment in its place; 1227 * check for complete reassembly. 1228 */ 1229 if (p == NULL) { 1230 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 1231 } else { 1232 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q); 1233 } 1234 next = 0; 1235 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1236 p = q, q = TAILQ_NEXT(q, ipqe_q)) { 1237 if (ntohs(q->ipqe_ip->ip_off) != next) 1238 return (0); 1239 next += ntohs(q->ipqe_ip->ip_len); 1240 } 1241 if (p->ipqe_mff) 1242 return (0); 1243 1244 /* 1245 * Reassembly is complete. Check for a bogus message size and 1246 * concatenate fragments. 1247 */ 1248 q = TAILQ_FIRST(&fp->ipq_fragq); 1249 ip = q->ipqe_ip; 1250 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 1251 IP_STATINC(IP_STAT_TOOLONG); 1252 ip_freef(fp); 1253 return (0); 1254 } 1255 m = q->ipqe_m; 1256 t = m->m_next; 1257 m->m_next = 0; 1258 m_cat(m, t); 1259 nq = TAILQ_NEXT(q, ipqe_q); 1260 s = splvm(); 1261 pool_put(&ipqent_pool, q); 1262 splx(s); 1263 for (q = nq; q != NULL; q = nq) { 1264 t = q->ipqe_m; 1265 nq = TAILQ_NEXT(q, ipqe_q); 1266 s = splvm(); 1267 pool_put(&ipqent_pool, q); 1268 splx(s); 1269 m_cat(m, t); 1270 } 1271 ip_nfrags -= fp->ipq_nfrags; 1272 1273 /* 1274 * Create header for new ip packet by 1275 * modifying header of first packet; 1276 * dequeue and discard fragment reassembly header. 1277 * Make header visible. 1278 */ 1279 ip->ip_len = htons(next); 1280 ip->ip_src = fp->ipq_src; 1281 ip->ip_dst = fp->ipq_dst; 1282 LIST_REMOVE(fp, ipq_q); 1283 free(fp, M_FTABLE); 1284 ip_nfragpackets--; 1285 m->m_len += (ip->ip_hl << 2); 1286 m->m_data -= (ip->ip_hl << 2); 1287 /* some debugging cruft by sklower, below, will go away soon */ 1288 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1289 int plen = 0; 1290 for (t = m; t; t = t->m_next) 1291 plen += t->m_len; 1292 m->m_pkthdr.len = plen; 1293 m->m_pkthdr.csum_flags = 0; 1294 } 1295 return (m); 1296 1297dropfrag: 1298 if (fp != 0) 1299 fp->ipq_nfrags--; 1300 ip_nfrags--; 1301 IP_STATINC(IP_STAT_FRAGDROPPED); 1302 m_freem(m); 1303 s = splvm(); 1304 pool_put(&ipqent_pool, ipqe); 1305 splx(s); 1306 return (0); 1307} 1308 1309/* 1310 * Free a fragment reassembly header and all 1311 * associated datagrams. 1312 */ 1313void 1314ip_freef(struct ipq *fp) 1315{ 1316 struct ipqent *q, *p; 1317 u_int nfrags = 0; 1318 int s; 1319 1320 IPQ_LOCK_CHECK(); 1321 1322 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) { 1323 p = TAILQ_NEXT(q, ipqe_q); 1324 m_freem(q->ipqe_m); 1325 nfrags++; 1326 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1327 s = splvm(); 1328 pool_put(&ipqent_pool, q); 1329 splx(s); 1330 } 1331 1332 if (nfrags != fp->ipq_nfrags) 1333 printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags); 1334 ip_nfrags -= nfrags; 1335 LIST_REMOVE(fp, ipq_q); 1336 free(fp, M_FTABLE); 1337 ip_nfragpackets--; 1338} 1339 1340/* 1341 * IP reassembly TTL machinery for multiplicative drop. 1342 */ 1343static u_int fragttl_histo[(IPFRAGTTL+1)]; 1344 1345 1346/* 1347 * Decrement TTL of all reasembly queue entries by `ticks'. 1348 * Count number of distinct fragments (as opposed to partial, fragmented 1349 * datagrams) in the reassembly queue. While we traverse the entire 1350 * reassembly queue, compute and return the median TTL over all fragments. 1351 */ 1352static u_int 1353ip_reass_ttl_decr(u_int ticks) 1354{ 1355 u_int nfrags, median, dropfraction, keepfraction; 1356 struct ipq *fp, *nfp; 1357 int i; 1358 1359 nfrags = 0; 1360 memset(fragttl_histo, 0, sizeof fragttl_histo); 1361 1362 for (i = 0; i < IPREASS_NHASH; i++) { 1363 for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) { 1364 fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ? 1365 0 : fp->ipq_ttl - ticks); 1366 nfp = LIST_NEXT(fp, ipq_q); 1367 if (fp->ipq_ttl == 0) { 1368 IP_STATINC(IP_STAT_FRAGTIMEOUT); 1369 ip_freef(fp); 1370 } else { 1371 nfrags += fp->ipq_nfrags; 1372 fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags; 1373 } 1374 } 1375 } 1376 1377 KASSERT(ip_nfrags == nfrags); 1378 1379 /* Find median (or other drop fraction) in histogram. */ 1380 dropfraction = (ip_nfrags / 2); 1381 keepfraction = ip_nfrags - dropfraction; 1382 for (i = IPFRAGTTL, median = 0; i >= 0; i--) { 1383 median += fragttl_histo[i]; 1384 if (median >= keepfraction) 1385 break; 1386 } 1387 1388 /* Return TTL of median (or other fraction). */ 1389 return (u_int)i; 1390} 1391 1392void 1393ip_reass_drophalf(void) 1394{ 1395 1396 u_int median_ticks; 1397 /* 1398 * Compute median TTL of all fragments, and count frags 1399 * with that TTL or lower (roughly half of all fragments). 1400 */ 1401 median_ticks = ip_reass_ttl_decr(0); 1402 1403 /* Drop half. */ 1404 median_ticks = ip_reass_ttl_decr(median_ticks); 1405 1406} 1407 1408/* 1409 * IP timer processing; 1410 * if a timer expires on a reassembly 1411 * queue, discard it. 1412 */ 1413void 1414ip_slowtimo(void) 1415{ 1416 static u_int dropscanidx = 0; 1417 u_int i; 1418 u_int median_ttl; 1419 1420 mutex_enter(softnet_lock); 1421 KERNEL_LOCK(1, NULL); 1422 1423 IPQ_LOCK(); 1424 1425 /* Age TTL of all fragments by 1 tick .*/ 1426 median_ttl = ip_reass_ttl_decr(1); 1427 1428 /* make sure fragment limit is up-to-date */ 1429 CHECK_NMBCLUSTER_PARAMS(); 1430 1431 /* If we have too many fragments, drop the older half. */ 1432 if (ip_nfrags > ip_maxfrags) 1433 ip_reass_ttl_decr(median_ttl); 1434 1435 /* 1436 * If we are over the maximum number of fragmented packets 1437 * (due to the limit being lowered), drain off 1438 * enough to get down to the new limit. Start draining 1439 * from the reassembly hashqueue most recently drained. 1440 */ 1441 if (ip_maxfragpackets < 0) 1442 ; 1443 else { 1444 int wrapped = 0; 1445 1446 i = dropscanidx; 1447 while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) { 1448 while (LIST_FIRST(&ipq[i]) != NULL) 1449 ip_freef(LIST_FIRST(&ipq[i])); 1450 if (++i >= IPREASS_NHASH) { 1451 i = 0; 1452 } 1453 /* 1454 * Dont scan forever even if fragment counters are 1455 * wrong: stop after scanning entire reassembly queue. 1456 */ 1457 if (i == dropscanidx) 1458 wrapped = 1; 1459 } 1460 dropscanidx = i; 1461 } 1462 IPQ_UNLOCK(); 1463 1464 KERNEL_UNLOCK_ONE(NULL); 1465 mutex_exit(softnet_lock); 1466} 1467 1468/* 1469 * Drain off all datagram fragments. Don't acquire softnet_lock as 1470 * can be called from hardware interrupt context. 1471 */ 1472void 1473ip_drain(void) 1474{ 1475 1476 KERNEL_LOCK(1, NULL); 1477 1478 /* 1479 * We may be called from a device's interrupt context. If 1480 * the ipq is already busy, just bail out now. 1481 */ 1482 if (ipq_lock_try() != 0) { 1483 /* 1484 * Drop half the total fragments now. If more mbufs are 1485 * needed, we will be called again soon. 1486 */ 1487 ip_reass_drophalf(); 1488 IPQ_UNLOCK(); 1489 } 1490 1491 KERNEL_UNLOCK_ONE(NULL); 1492} 1493 1494/* 1495 * Do option processing on a datagram, 1496 * possibly discarding it if bad options are encountered, 1497 * or forwarding it if source-routed. 1498 * Returns 1 if packet has been forwarded/freed, 1499 * 0 if the packet should be processed further. 1500 */ 1501int 1502ip_dooptions(struct mbuf *m) 1503{ 1504 struct ip *ip = mtod(m, struct ip *); 1505 u_char *cp, *cp0; 1506 struct ip_timestamp *ipt; 1507 struct in_ifaddr *ia; 1508 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1509 struct in_addr dst; 1510 n_time ntime; 1511 1512 dst = ip->ip_dst; 1513 cp = (u_char *)(ip + 1); 1514 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1515 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1516 opt = cp[IPOPT_OPTVAL]; 1517 if (opt == IPOPT_EOL) 1518 break; 1519 if (opt == IPOPT_NOP) 1520 optlen = 1; 1521 else { 1522 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1523 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1524 goto bad; 1525 } 1526 optlen = cp[IPOPT_OLEN]; 1527 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1528 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1529 goto bad; 1530 } 1531 } 1532 switch (opt) { 1533 1534 default: 1535 break; 1536 1537 /* 1538 * Source routing with record. 1539 * Find interface with current destination address. 1540 * If none on this machine then drop if strictly routed, 1541 * or do nothing if loosely routed. 1542 * Record interface address and bring up next address 1543 * component. If strictly routed make sure next 1544 * address is on directly accessible net. 1545 */ 1546 case IPOPT_LSRR: 1547 case IPOPT_SSRR: 1548 if (ip_allowsrcrt == 0) { 1549 type = ICMP_UNREACH; 1550 code = ICMP_UNREACH_NET_PROHIB; 1551 goto bad; 1552 } 1553 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1554 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1555 goto bad; 1556 } 1557 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1558 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1559 goto bad; 1560 } 1561 ipaddr.sin_addr = ip->ip_dst; 1562 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1563 if (ia == 0) { 1564 if (opt == IPOPT_SSRR) { 1565 type = ICMP_UNREACH; 1566 code = ICMP_UNREACH_SRCFAIL; 1567 goto bad; 1568 } 1569 /* 1570 * Loose routing, and not at next destination 1571 * yet; nothing to do except forward. 1572 */ 1573 break; 1574 } 1575 off--; /* 0 origin */ 1576 if ((off + sizeof(struct in_addr)) > optlen) { 1577 /* 1578 * End of source route. Should be for us. 1579 */ 1580 save_rte(cp, ip->ip_src); 1581 break; 1582 } 1583 /* 1584 * locate outgoing interface 1585 */ 1586 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 1587 sizeof(ipaddr.sin_addr)); 1588 if (opt == IPOPT_SSRR) 1589 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); 1590 else 1591 ia = ip_rtaddr(ipaddr.sin_addr); 1592 if (ia == 0) { 1593 type = ICMP_UNREACH; 1594 code = ICMP_UNREACH_SRCFAIL; 1595 goto bad; 1596 } 1597 ip->ip_dst = ipaddr.sin_addr; 1598 bcopy((void *)&ia->ia_addr.sin_addr, 1599 (void *)(cp + off), sizeof(struct in_addr)); 1600 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1601 /* 1602 * Let ip_intr's mcast routing check handle mcast pkts 1603 */ 1604 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1605 break; 1606 1607 case IPOPT_RR: 1608 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1609 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1610 goto bad; 1611 } 1612 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1613 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1614 goto bad; 1615 } 1616 /* 1617 * If no space remains, ignore. 1618 */ 1619 off--; /* 0 origin */ 1620 if ((off + sizeof(struct in_addr)) > optlen) 1621 break; 1622 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst), 1623 sizeof(ipaddr.sin_addr)); 1624 /* 1625 * locate outgoing interface; if we're the destination, 1626 * use the incoming interface (should be same). 1627 */ 1628 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1629 == NULL && 1630 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1631 type = ICMP_UNREACH; 1632 code = ICMP_UNREACH_HOST; 1633 goto bad; 1634 } 1635 bcopy((void *)&ia->ia_addr.sin_addr, 1636 (void *)(cp + off), sizeof(struct in_addr)); 1637 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1638 break; 1639 1640 case IPOPT_TS: 1641 code = cp - (u_char *)ip; 1642 ipt = (struct ip_timestamp *)cp; 1643 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1644 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1645 goto bad; 1646 } 1647 if (ipt->ipt_ptr < 5) { 1648 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1649 goto bad; 1650 } 1651 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1652 if (++ipt->ipt_oflw == 0) { 1653 code = (u_char *)&ipt->ipt_ptr - 1654 (u_char *)ip; 1655 goto bad; 1656 } 1657 break; 1658 } 1659 cp0 = (cp + ipt->ipt_ptr - 1); 1660 switch (ipt->ipt_flg) { 1661 1662 case IPOPT_TS_TSONLY: 1663 break; 1664 1665 case IPOPT_TS_TSANDADDR: 1666 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1667 sizeof(struct in_addr) > ipt->ipt_len) { 1668 code = (u_char *)&ipt->ipt_ptr - 1669 (u_char *)ip; 1670 goto bad; 1671 } 1672 ipaddr.sin_addr = dst; 1673 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1674 m->m_pkthdr.rcvif)); 1675 if (ia == 0) 1676 continue; 1677 bcopy(&ia->ia_addr.sin_addr, 1678 cp0, sizeof(struct in_addr)); 1679 ipt->ipt_ptr += sizeof(struct in_addr); 1680 break; 1681 1682 case IPOPT_TS_PRESPEC: 1683 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1684 sizeof(struct in_addr) > ipt->ipt_len) { 1685 code = (u_char *)&ipt->ipt_ptr - 1686 (u_char *)ip; 1687 goto bad; 1688 } 1689 memcpy(&ipaddr.sin_addr, cp0, 1690 sizeof(struct in_addr)); 1691 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1692 == NULL) 1693 continue; 1694 ipt->ipt_ptr += sizeof(struct in_addr); 1695 break; 1696 1697 default: 1698 /* XXX can't take &ipt->ipt_flg */ 1699 code = (u_char *)&ipt->ipt_ptr - 1700 (u_char *)ip + 1; 1701 goto bad; 1702 } 1703 ntime = iptime(); 1704 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1705 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1706 sizeof(n_time)); 1707 ipt->ipt_ptr += sizeof(n_time); 1708 } 1709 } 1710 if (forward) { 1711 if (ip_forwsrcrt == 0) { 1712 type = ICMP_UNREACH; 1713 code = ICMP_UNREACH_SRCFAIL; 1714 goto bad; 1715 } 1716 ip_forward(m, 1); 1717 return (1); 1718 } 1719 return (0); 1720bad: 1721 icmp_error(m, type, code, 0, 0); 1722 IP_STATINC(IP_STAT_BADOPTIONS); 1723 return (1); 1724} 1725 1726/* 1727 * Given address of next destination (final or next hop), 1728 * return internet address info of interface to be used to get there. 1729 */ 1730struct in_ifaddr * 1731ip_rtaddr(struct in_addr dst) 1732{ 1733 struct rtentry *rt; 1734 union { 1735 struct sockaddr dst; 1736 struct sockaddr_in dst4; 1737 } u; 1738 1739 sockaddr_in_init(&u.dst4, &dst, 0); 1740 1741 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) 1742 return NULL; 1743 1744 return ifatoia(rt->rt_ifa); 1745} 1746 1747/* 1748 * Save incoming source route for use in replies, 1749 * to be picked up later by ip_srcroute if the receiver is interested. 1750 */ 1751void 1752save_rte(u_char *option, struct in_addr dst) 1753{ 1754 unsigned olen; 1755 1756 olen = option[IPOPT_OLEN]; 1757#ifdef DIAGNOSTIC 1758 if (ipprintfs) 1759 printf("save_rte: olen %d\n", olen); 1760#endif /* 0 */ 1761 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1762 return; 1763 memcpy((void *)ip_srcrt.srcopt, (void *)option, olen); 1764 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1765 ip_srcrt.dst = dst; 1766} 1767 1768/* 1769 * Retrieve incoming source route for use in replies, 1770 * in the same form used by setsockopt. 1771 * The first hop is placed before the options, will be removed later. 1772 */ 1773struct mbuf * 1774ip_srcroute(void) 1775{ 1776 struct in_addr *p, *q; 1777 struct mbuf *m; 1778 1779 if (ip_nhops == 0) 1780 return NULL; 1781 m = m_get(M_DONTWAIT, MT_SOOPTS); 1782 if (m == 0) 1783 return NULL; 1784 1785 MCLAIM(m, &inetdomain.dom_mowner); 1786#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1787 1788 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1789 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1790 OPTSIZ; 1791#ifdef DIAGNOSTIC 1792 if (ipprintfs) 1793 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1794#endif 1795 1796 /* 1797 * First save first hop for return route 1798 */ 1799 p = &ip_srcrt.route[ip_nhops - 1]; 1800 *(mtod(m, struct in_addr *)) = *p--; 1801#ifdef DIAGNOSTIC 1802 if (ipprintfs) 1803 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1804#endif 1805 1806 /* 1807 * Copy option fields and padding (nop) to mbuf. 1808 */ 1809 ip_srcrt.nop = IPOPT_NOP; 1810 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1811 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop, 1812 OPTSIZ); 1813 q = (struct in_addr *)(mtod(m, char *) + 1814 sizeof(struct in_addr) + OPTSIZ); 1815#undef OPTSIZ 1816 /* 1817 * Record return path as an IP source route, 1818 * reversing the path (pointers are now aligned). 1819 */ 1820 while (p >= ip_srcrt.route) { 1821#ifdef DIAGNOSTIC 1822 if (ipprintfs) 1823 printf(" %x", ntohl(q->s_addr)); 1824#endif 1825 *q++ = *p--; 1826 } 1827 /* 1828 * Last hop goes to final destination. 1829 */ 1830 *q = ip_srcrt.dst; 1831#ifdef DIAGNOSTIC 1832 if (ipprintfs) 1833 printf(" %x\n", ntohl(q->s_addr)); 1834#endif 1835 return (m); 1836} 1837 1838const int inetctlerrmap[PRC_NCMDS] = { 1839 [PRC_MSGSIZE] = EMSGSIZE, 1840 [PRC_HOSTDEAD] = EHOSTDOWN, 1841 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1842 [PRC_UNREACH_NET] = EHOSTUNREACH, 1843 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1844 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1845 [PRC_UNREACH_PORT] = ECONNREFUSED, 1846 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1847 [PRC_PARAMPROB] = ENOPROTOOPT, 1848}; 1849 1850/* 1851 * Forward a packet. If some error occurs return the sender 1852 * an icmp packet. Note we can't always generate a meaningful 1853 * icmp message because icmp doesn't have a large enough repertoire 1854 * of codes and types. 1855 * 1856 * If not forwarding, just drop the packet. This could be confusing 1857 * if ipforwarding was zero but some routing protocol was advancing 1858 * us as a gateway to somewhere. However, we must let the routing 1859 * protocol deal with that. 1860 * 1861 * The srcrt parameter indicates whether the packet is being forwarded 1862 * via a source route. 1863 */ 1864void 1865ip_forward(struct mbuf *m, int srcrt) 1866{ 1867 struct ip *ip = mtod(m, struct ip *); 1868 struct rtentry *rt; 1869 int error, type = 0, code = 0, destmtu = 0; 1870 struct mbuf *mcopy; 1871 n_long dest; 1872 union { 1873 struct sockaddr dst; 1874 struct sockaddr_in dst4; 1875 } u; 1876 1877 /* 1878 * We are now in the output path. 1879 */ 1880 MCLAIM(m, &ip_tx_mowner); 1881 1882 /* 1883 * Clear any in-bound checksum flags for this packet. 1884 */ 1885 m->m_pkthdr.csum_flags = 0; 1886 1887 dest = 0; 1888#ifdef DIAGNOSTIC 1889 if (ipprintfs) { 1890 printf("forward: src %s ", inet_ntoa(ip->ip_src)); 1891 printf("dst %s ttl %x\n", inet_ntoa(ip->ip_dst), ip->ip_ttl); 1892 } 1893#endif 1894 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1895 IP_STATINC(IP_STAT_CANTFORWARD); 1896 m_freem(m); 1897 return; 1898 } 1899 if (ip->ip_ttl <= IPTTLDEC) { 1900 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1901 return; 1902 } 1903 1904 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1905 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) { 1906 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1907 return; 1908 } 1909 1910 /* 1911 * Save at most 68 bytes of the packet in case 1912 * we need to generate an ICMP message to the src. 1913 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1914 */ 1915 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1916 if (mcopy) 1917 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1918 1919 ip->ip_ttl -= IPTTLDEC; 1920 1921 /* 1922 * If forwarding packet using same interface that it came in on, 1923 * perhaps should send a redirect to sender to shortcut a hop. 1924 * Only send redirect if source is sending directly to us, 1925 * and if packet was not source routed (or has any options). 1926 * Also, don't send redirect if forwarding using a default route 1927 * or a route modified by a redirect. 1928 */ 1929 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1930 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1931 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1932 ipsendredirects && !srcrt) { 1933 if (rt->rt_ifa && 1934 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1935 ifatoia(rt->rt_ifa)->ia_subnet) { 1936 if (rt->rt_flags & RTF_GATEWAY) 1937 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1938 else 1939 dest = ip->ip_dst.s_addr; 1940 /* 1941 * Router requirements says to only send host 1942 * redirects. 1943 */ 1944 type = ICMP_REDIRECT; 1945 code = ICMP_REDIRECT_HOST; 1946#ifdef DIAGNOSTIC 1947 if (ipprintfs) 1948 printf("redirect (%d) to %x\n", code, 1949 (u_int32_t)dest); 1950#endif 1951 } 1952 } 1953 1954 error = ip_output(m, NULL, &ipforward_rt, 1955 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1956 (struct ip_moptions *)NULL, (struct socket *)NULL); 1957 1958 if (error) 1959 IP_STATINC(IP_STAT_CANTFORWARD); 1960 else { 1961 uint64_t *ips = IP_STAT_GETREF(); 1962 ips[IP_STAT_FORWARD]++; 1963 if (type) { 1964 ips[IP_STAT_REDIRECTSENT]++; 1965 IP_STAT_PUTREF(); 1966 } else { 1967 IP_STAT_PUTREF(); 1968 if (mcopy) { 1969#ifdef GATEWAY 1970 if (mcopy->m_flags & M_CANFASTFWD) 1971 ipflow_create(&ipforward_rt, mcopy); 1972#endif 1973 m_freem(mcopy); 1974 } 1975 return; 1976 } 1977 } 1978 if (mcopy == NULL) 1979 return; 1980 1981 switch (error) { 1982 1983 case 0: /* forwarded, but need redirect */ 1984 /* type, code set above */ 1985 break; 1986 1987 case ENETUNREACH: /* shouldn't happen, checked above */ 1988 case EHOSTUNREACH: 1989 case ENETDOWN: 1990 case EHOSTDOWN: 1991 default: 1992 type = ICMP_UNREACH; 1993 code = ICMP_UNREACH_HOST; 1994 break; 1995 1996 case EMSGSIZE: 1997 type = ICMP_UNREACH; 1998 code = ICMP_UNREACH_NEEDFRAG; 1999 2000 if ((rt = rtcache_validate(&ipforward_rt)) != NULL) 2001 destmtu = rt->rt_ifp->if_mtu; 2002 2003#if defined(IPSEC) || defined(FAST_IPSEC) 2004 { 2005 /* 2006 * If the packet is routed over IPsec tunnel, tell the 2007 * originator the tunnel MTU. 2008 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 2009 * XXX quickhack!!! 2010 */ 2011 2012 struct secpolicy *sp; 2013 int ipsecerror; 2014 size_t ipsechdr; 2015 struct route *ro; 2016 2017 sp = ipsec4_getpolicybyaddr(mcopy, 2018 IPSEC_DIR_OUTBOUND, IP_FORWARDING, 2019 &ipsecerror); 2020 2021 if (sp != NULL) { 2022 /* count IPsec header size */ 2023 ipsechdr = ipsec4_hdrsiz(mcopy, 2024 IPSEC_DIR_OUTBOUND, NULL); 2025 2026 /* 2027 * find the correct route for outer IPv4 2028 * header, compute tunnel MTU. 2029 */ 2030 2031 if (sp->req != NULL 2032 && sp->req->sav != NULL 2033 && sp->req->sav->sah != NULL) { 2034 ro = &sp->req->sav->sah->sa_route; 2035 rt = rtcache_validate(ro); 2036 if (rt && rt->rt_ifp) { 2037 destmtu = 2038 rt->rt_rmx.rmx_mtu ? 2039 rt->rt_rmx.rmx_mtu : 2040 rt->rt_ifp->if_mtu; 2041 destmtu -= ipsechdr; 2042 } 2043 } 2044 2045#ifdef IPSEC 2046 key_freesp(sp); 2047#else 2048 KEY_FREESP(&sp); 2049#endif 2050 } 2051 } 2052#endif /*defined(IPSEC) || defined(FAST_IPSEC)*/ 2053 IP_STATINC(IP_STAT_CANTFRAG); 2054 break; 2055 2056 case ENOBUFS: 2057#if 1 2058 /* 2059 * a router should not generate ICMP_SOURCEQUENCH as 2060 * required in RFC1812 Requirements for IP Version 4 Routers. 2061 * source quench could be a big problem under DoS attacks, 2062 * or if the underlying interface is rate-limited. 2063 */ 2064 if (mcopy) 2065 m_freem(mcopy); 2066 return; 2067#else 2068 type = ICMP_SOURCEQUENCH; 2069 code = 0; 2070 break; 2071#endif 2072 } 2073 icmp_error(mcopy, type, code, dest, destmtu); 2074} 2075 2076void 2077ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2078 struct mbuf *m) 2079{ 2080 2081 if (inp->inp_socket->so_options & SO_TIMESTAMP 2082#ifdef SO_OTIMESTAMP 2083 || inp->inp_socket->so_options & SO_OTIMESTAMP 2084#endif 2085 ) { 2086 struct timeval tv; 2087 2088 microtime(&tv); 2089#ifdef SO_OTIMESTAMP 2090 if (inp->inp_socket->so_options & SO_OTIMESTAMP) { 2091 struct timeval50 tv50; 2092 timeval_to_timeval50(&tv, &tv50); 2093 *mp = sbcreatecontrol((void *) &tv50, sizeof(tv50), 2094 SCM_OTIMESTAMP, SOL_SOCKET); 2095 } else 2096#endif 2097 *mp = sbcreatecontrol((void *) &tv, sizeof(tv), 2098 SCM_TIMESTAMP, SOL_SOCKET); 2099 if (*mp) 2100 mp = &(*mp)->m_next; 2101 } 2102 if (inp->inp_flags & INP_RECVDSTADDR) { 2103 *mp = sbcreatecontrol((void *) &ip->ip_dst, 2104 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2105 if (*mp) 2106 mp = &(*mp)->m_next; 2107 } 2108#ifdef notyet 2109 /* 2110 * XXX 2111 * Moving these out of udp_input() made them even more broken 2112 * than they already were. 2113 * - fenner@parc.xerox.com 2114 */ 2115 /* options were tossed already */ 2116 if (inp->inp_flags & INP_RECVOPTS) { 2117 *mp = sbcreatecontrol((void *) opts_deleted_above, 2118 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2119 if (*mp) 2120 mp = &(*mp)->m_next; 2121 } 2122 /* ip_srcroute doesn't do what we want here, need to fix */ 2123 if (inp->inp_flags & INP_RECVRETOPTS) { 2124 *mp = sbcreatecontrol((void *) ip_srcroute(), 2125 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2126 if (*mp) 2127 mp = &(*mp)->m_next; 2128 } 2129#endif 2130 if (inp->inp_flags & INP_RECVIF) { 2131 struct sockaddr_dl sdl; 2132 2133 sockaddr_dl_init(&sdl, sizeof(sdl), 2134 (m->m_pkthdr.rcvif != NULL) 2135 ? m->m_pkthdr.rcvif->if_index 2136 : 0, 2137 0, NULL, 0, NULL, 0); 2138 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 2139 if (*mp) 2140 mp = &(*mp)->m_next; 2141 } 2142 if (inp->inp_flags & INP_RECVTTL) { 2143 *mp = sbcreatecontrol((void *) &ip->ip_ttl, 2144 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 2145 if (*mp) 2146 mp = &(*mp)->m_next; 2147 } 2148} 2149 2150/* 2151 * sysctl helper routine for net.inet.ip.forwsrcrt. 2152 */ 2153static int 2154sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 2155{ 2156 int error, tmp; 2157 struct sysctlnode node; 2158 2159 node = *rnode; 2160 tmp = ip_forwsrcrt; 2161 node.sysctl_data = &tmp; 2162 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2163 if (error || newp == NULL) 2164 return (error); 2165 2166 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 2167 0, NULL, NULL, NULL); 2168 if (error) 2169 return (error); 2170 2171 ip_forwsrcrt = tmp; 2172 2173 return (0); 2174} 2175 2176/* 2177 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 2178 * range of the new value and tweaks timers if it changes. 2179 */ 2180static int 2181sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 2182{ 2183 int error, tmp; 2184 struct sysctlnode node; 2185 2186 node = *rnode; 2187 tmp = ip_mtudisc_timeout; 2188 node.sysctl_data = &tmp; 2189 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2190 if (error || newp == NULL) 2191 return (error); 2192 if (tmp < 0) 2193 return (EINVAL); 2194 2195 mutex_enter(softnet_lock); 2196 2197 ip_mtudisc_timeout = tmp; 2198 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 2199 2200 mutex_exit(softnet_lock); 2201 2202 return (0); 2203} 2204 2205#ifdef GATEWAY 2206/* 2207 * sysctl helper routine for net.inet.ip.maxflows. 2208 */ 2209static int 2210sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS) 2211{ 2212 int error; 2213 2214 error = sysctl_lookup(SYSCTLFN_CALL(rnode)); 2215 if (error || newp == NULL) 2216 return (error); 2217 2218 mutex_enter(softnet_lock); 2219 KERNEL_LOCK(1, NULL); 2220 2221 ipflow_prune(); 2222 2223 KERNEL_UNLOCK_ONE(NULL); 2224 mutex_exit(softnet_lock); 2225 2226 return (0); 2227} 2228 2229static int 2230sysctl_net_inet_ip_hashsize(SYSCTLFN_ARGS) 2231{ 2232 int error, tmp; 2233 struct sysctlnode node; 2234 2235 node = *rnode; 2236 tmp = ip_hashsize; 2237 node.sysctl_data = &tmp; 2238 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2239 if (error || newp == NULL) 2240 return (error); 2241 2242 if ((tmp & (tmp - 1)) == 0 && tmp != 0) { 2243 /* 2244 * Can only fail due to malloc() 2245 */ 2246 mutex_enter(softnet_lock); 2247 KERNEL_LOCK(1, NULL); 2248 2249 error = ipflow_invalidate_all(tmp); 2250 2251 KERNEL_UNLOCK_ONE(NULL); 2252 mutex_exit(softnet_lock); 2253 2254 } else { 2255 /* 2256 * EINVAL if not a power of 2 2257 */ 2258 error = EINVAL; 2259 } 2260 2261 return error; 2262} 2263#endif /* GATEWAY */ 2264 2265static int 2266sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 2267{ 2268 2269 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 2270} 2271 2272static void 2273sysctl_net_inet_ip_setup(struct sysctllog **clog) 2274{ 2275 extern int subnetsarelocal, hostzeroisbroadcast; 2276 2277 sysctl_createv(clog, 0, NULL, NULL, 2278 CTLFLAG_PERMANENT, 2279 CTLTYPE_NODE, "net", NULL, 2280 NULL, 0, NULL, 0, 2281 CTL_NET, CTL_EOL); 2282 sysctl_createv(clog, 0, NULL, NULL, 2283 CTLFLAG_PERMANENT, 2284 CTLTYPE_NODE, "inet", 2285 SYSCTL_DESCR("PF_INET related settings"), 2286 NULL, 0, NULL, 0, 2287 CTL_NET, PF_INET, CTL_EOL); 2288 sysctl_createv(clog, 0, NULL, NULL, 2289 CTLFLAG_PERMANENT, 2290 CTLTYPE_NODE, "ip", 2291 SYSCTL_DESCR("IPv4 related settings"), 2292 NULL, 0, NULL, 0, 2293 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 2294 2295 sysctl_createv(clog, 0, NULL, NULL, 2296 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2297 CTLTYPE_INT, "forwarding", 2298 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 2299 NULL, 0, &ipforwarding, 0, 2300 CTL_NET, PF_INET, IPPROTO_IP, 2301 IPCTL_FORWARDING, CTL_EOL); 2302 sysctl_createv(clog, 0, NULL, NULL, 2303 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2304 CTLTYPE_INT, "redirect", 2305 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 2306 NULL, 0, &ipsendredirects, 0, 2307 CTL_NET, PF_INET, IPPROTO_IP, 2308 IPCTL_SENDREDIRECTS, CTL_EOL); 2309 sysctl_createv(clog, 0, NULL, NULL, 2310 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2311 CTLTYPE_INT, "ttl", 2312 SYSCTL_DESCR("Default TTL for an INET datagram"), 2313 NULL, 0, &ip_defttl, 0, 2314 CTL_NET, PF_INET, IPPROTO_IP, 2315 IPCTL_DEFTTL, CTL_EOL); 2316#ifdef IPCTL_DEFMTU 2317 sysctl_createv(clog, 0, NULL, NULL, 2318 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 2319 CTLTYPE_INT, "mtu", 2320 SYSCTL_DESCR("Default MTA for an INET route"), 2321 NULL, 0, &ip_mtu, 0, 2322 CTL_NET, PF_INET, IPPROTO_IP, 2323 IPCTL_DEFMTU, CTL_EOL); 2324#endif /* IPCTL_DEFMTU */ 2325 sysctl_createv(clog, 0, NULL, NULL, 2326 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2327 CTLTYPE_INT, "forwsrcrt", 2328 SYSCTL_DESCR("Enable forwarding of source-routed " 2329 "datagrams"), 2330 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 2331 CTL_NET, PF_INET, IPPROTO_IP, 2332 IPCTL_FORWSRCRT, CTL_EOL); 2333 sysctl_createv(clog, 0, NULL, NULL, 2334 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2335 CTLTYPE_INT, "directed-broadcast", 2336 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 2337 NULL, 0, &ip_directedbcast, 0, 2338 CTL_NET, PF_INET, IPPROTO_IP, 2339 IPCTL_DIRECTEDBCAST, CTL_EOL); 2340 sysctl_createv(clog, 0, NULL, NULL, 2341 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2342 CTLTYPE_INT, "allowsrcrt", 2343 SYSCTL_DESCR("Accept source-routed datagrams"), 2344 NULL, 0, &ip_allowsrcrt, 0, 2345 CTL_NET, PF_INET, IPPROTO_IP, 2346 IPCTL_ALLOWSRCRT, CTL_EOL); 2347 sysctl_createv(clog, 0, NULL, NULL, 2348 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2349 CTLTYPE_INT, "subnetsarelocal", 2350 SYSCTL_DESCR("Whether logical subnets are considered " 2351 "local"), 2352 NULL, 0, &subnetsarelocal, 0, 2353 CTL_NET, PF_INET, IPPROTO_IP, 2354 IPCTL_SUBNETSARELOCAL, CTL_EOL); 2355 sysctl_createv(clog, 0, NULL, NULL, 2356 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2357 CTLTYPE_INT, "mtudisc", 2358 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 2359 NULL, 0, &ip_mtudisc, 0, 2360 CTL_NET, PF_INET, IPPROTO_IP, 2361 IPCTL_MTUDISC, CTL_EOL); 2362 sysctl_createv(clog, 0, NULL, NULL, 2363 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2364 CTLTYPE_INT, "anonportmin", 2365 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 2366 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 2367 CTL_NET, PF_INET, IPPROTO_IP, 2368 IPCTL_ANONPORTMIN, CTL_EOL); 2369 sysctl_createv(clog, 0, NULL, NULL, 2370 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2371 CTLTYPE_INT, "anonportmax", 2372 SYSCTL_DESCR("Highest ephemeral port number to assign"), 2373 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 2374 CTL_NET, PF_INET, IPPROTO_IP, 2375 IPCTL_ANONPORTMAX, CTL_EOL); 2376 sysctl_createv(clog, 0, NULL, NULL, 2377 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2378 CTLTYPE_INT, "mtudisctimeout", 2379 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 2380 sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0, 2381 CTL_NET, PF_INET, IPPROTO_IP, 2382 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 2383#ifdef GATEWAY 2384 sysctl_createv(clog, 0, NULL, NULL, 2385 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2386 CTLTYPE_INT, "maxflows", 2387 SYSCTL_DESCR("Number of flows for fast forwarding"), 2388 sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0, 2389 CTL_NET, PF_INET, IPPROTO_IP, 2390 IPCTL_MAXFLOWS, CTL_EOL); 2391 sysctl_createv(clog, 0, NULL, NULL, 2392 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2393 CTLTYPE_INT, "hashsize", 2394 SYSCTL_DESCR("Size of hash table for fast forwarding (IPv4)"), 2395 sysctl_net_inet_ip_hashsize, 0, &ip_hashsize, 0, 2396 CTL_NET, PF_INET, IPPROTO_IP, 2397 CTL_CREATE, CTL_EOL); 2398#endif /* GATEWAY */ 2399 sysctl_createv(clog, 0, NULL, NULL, 2400 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2401 CTLTYPE_INT, "hostzerobroadcast", 2402 SYSCTL_DESCR("All zeroes address is broadcast address"), 2403 NULL, 0, &hostzeroisbroadcast, 0, 2404 CTL_NET, PF_INET, IPPROTO_IP, 2405 IPCTL_HOSTZEROBROADCAST, CTL_EOL); 2406#if NGIF > 0 2407 sysctl_createv(clog, 0, NULL, NULL, 2408 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2409 CTLTYPE_INT, "gifttl", 2410 SYSCTL_DESCR("Default TTL for a gif tunnel datagram"), 2411 NULL, 0, &ip_gif_ttl, 0, 2412 CTL_NET, PF_INET, IPPROTO_IP, 2413 IPCTL_GIF_TTL, CTL_EOL); 2414#endif /* NGIF */ 2415#ifndef IPNOPRIVPORTS 2416 sysctl_createv(clog, 0, NULL, NULL, 2417 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2418 CTLTYPE_INT, "lowportmin", 2419 SYSCTL_DESCR("Lowest privileged ephemeral port number " 2420 "to assign"), 2421 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 2422 CTL_NET, PF_INET, IPPROTO_IP, 2423 IPCTL_LOWPORTMIN, CTL_EOL); 2424 sysctl_createv(clog, 0, NULL, NULL, 2425 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2426 CTLTYPE_INT, "lowportmax", 2427 SYSCTL_DESCR("Highest privileged ephemeral port number " 2428 "to assign"), 2429 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 2430 CTL_NET, PF_INET, IPPROTO_IP, 2431 IPCTL_LOWPORTMAX, CTL_EOL); 2432#endif /* IPNOPRIVPORTS */ 2433 sysctl_createv(clog, 0, NULL, NULL, 2434 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2435 CTLTYPE_INT, "maxfragpackets", 2436 SYSCTL_DESCR("Maximum number of fragments to retain for " 2437 "possible reassembly"), 2438 NULL, 0, &ip_maxfragpackets, 0, 2439 CTL_NET, PF_INET, IPPROTO_IP, 2440 IPCTL_MAXFRAGPACKETS, CTL_EOL); 2441#if NGRE > 0 2442 sysctl_createv(clog, 0, NULL, NULL, 2443 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2444 CTLTYPE_INT, "grettl", 2445 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 2446 NULL, 0, &ip_gre_ttl, 0, 2447 CTL_NET, PF_INET, IPPROTO_IP, 2448 IPCTL_GRE_TTL, CTL_EOL); 2449#endif /* NGRE */ 2450 sysctl_createv(clog, 0, NULL, NULL, 2451 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2452 CTLTYPE_INT, "checkinterface", 2453 SYSCTL_DESCR("Enable receive side of Strong ES model " 2454 "from RFC1122"), 2455 NULL, 0, &ip_checkinterface, 0, 2456 CTL_NET, PF_INET, IPPROTO_IP, 2457 IPCTL_CHECKINTERFACE, CTL_EOL); 2458 sysctl_createv(clog, 0, NULL, NULL, 2459 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2460 CTLTYPE_INT, "random_id", 2461 SYSCTL_DESCR("Assign random ip_id values"), 2462 NULL, 0, &ip_do_randomid, 0, 2463 CTL_NET, PF_INET, IPPROTO_IP, 2464 IPCTL_RANDOMID, CTL_EOL); 2465 sysctl_createv(clog, 0, NULL, NULL, 2466 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2467 CTLTYPE_INT, "do_loopback_cksum", 2468 SYSCTL_DESCR("Perform IP checksum on loopback"), 2469 NULL, 0, &ip_do_loopback_cksum, 0, 2470 CTL_NET, PF_INET, IPPROTO_IP, 2471 IPCTL_LOOPBACKCKSUM, CTL_EOL); 2472 sysctl_createv(clog, 0, NULL, NULL, 2473 CTLFLAG_PERMANENT, 2474 CTLTYPE_STRUCT, "stats", 2475 SYSCTL_DESCR("IP statistics"), 2476 sysctl_net_inet_ip_stats, 0, NULL, 0, 2477 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 2478 CTL_EOL); 2479} 2480 2481void 2482ip_statinc(u_int stat) 2483{ 2484 2485 KASSERT(stat < IP_NSTATS); 2486 IP_STATINC(stat); 2487} 2488