ip_input.c revision 1.285
1/* $NetBSD: ip_input.c,v 1.285 2010/03/31 07:31:15 tls 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 1.285 2010/03/31 07:31:15 tls 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; 234uint16_t ip_id; 235 236percpu_t *ipstat_percpu; 237 238#ifdef PFIL_HOOKS 239struct pfil_head inet_pfil_hook; 240#endif 241 242/* 243 * Cached copy of nmbclusters. If nbclusters is different, 244 * recalculate IP parameters derived from nmbclusters. 245 */ 246static int ip_nmbclusters; /* copy of nmbclusters */ 247static void ip_nmbclusters_changed(void); /* recalc limits */ 248 249#define CHECK_NMBCLUSTER_PARAMS() \ 250do { \ 251 if (__predict_false(ip_nmbclusters != nmbclusters)) \ 252 ip_nmbclusters_changed(); \ 253} while (/*CONSTCOND*/0) 254 255/* IP datagram reassembly queues (hashed) */ 256#define IPREASS_NHASH_LOG2 6 257#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 258#define IPREASS_HMASK (IPREASS_NHASH - 1) 259#define IPREASS_HASH(x,y) \ 260 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 261struct ipqhead ipq[IPREASS_NHASH]; 262int ipq_locked; 263static int ip_nfragpackets; /* packets in reass queue */ 264static int ip_nfrags; /* total fragments in reass queues */ 265 266int ip_maxfragpackets = 200; /* limit on packets. XXX sysctl */ 267int ip_maxfrags; /* limit on fragments. XXX sysctl */ 268 269 270/* 271 * Additive-Increase/Multiplicative-Decrease (AIMD) strategy for 272 * IP reassembly queue buffer managment. 273 * 274 * We keep a count of total IP fragments (NB: not fragmented packets!) 275 * awaiting reassembly (ip_nfrags) and a limit (ip_maxfrags) on fragments. 276 * If ip_nfrags exceeds ip_maxfrags the limit, we drop half the 277 * total fragments in reassembly queues.This AIMD policy avoids 278 * repeatedly deleting single packets under heavy fragmentation load 279 * (e.g., from lossy NFS peers). 280 */ 281static u_int ip_reass_ttl_decr(u_int ticks); 282static void ip_reass_drophalf(void); 283 284 285static inline int ipq_lock_try(void); 286static inline void ipq_unlock(void); 287 288static inline int 289ipq_lock_try(void) 290{ 291 int s; 292 293 /* 294 * Use splvm() -- we're blocking things that would cause 295 * mbuf allocation. 296 */ 297 s = splvm(); 298 if (ipq_locked) { 299 splx(s); 300 return (0); 301 } 302 ipq_locked = 1; 303 splx(s); 304 return (1); 305} 306 307static inline void 308ipq_unlock(void) 309{ 310 int s; 311 312 s = splvm(); 313 ipq_locked = 0; 314 splx(s); 315} 316 317#ifdef DIAGNOSTIC 318#define IPQ_LOCK() \ 319do { \ 320 if (ipq_lock_try() == 0) { \ 321 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 322 panic("ipq_lock"); \ 323 } \ 324} while (/*CONSTCOND*/ 0) 325#define IPQ_LOCK_CHECK() \ 326do { \ 327 if (ipq_locked == 0) { \ 328 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 329 panic("ipq lock check"); \ 330 } \ 331} while (/*CONSTCOND*/ 0) 332#else 333#define IPQ_LOCK() (void) ipq_lock_try() 334#define IPQ_LOCK_CHECK() /* nothing */ 335#endif 336 337#define IPQ_UNLOCK() ipq_unlock() 338 339struct pool inmulti_pool; 340struct pool ipqent_pool; 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 386static void sysctl_net_inet_ip_setup(struct sysctllog **); 387 388/* 389 * Compute IP limits derived from the value of nmbclusters. 390 */ 391static void 392ip_nmbclusters_changed(void) 393{ 394 ip_maxfrags = nmbclusters / 4; 395 ip_nmbclusters = nmbclusters; 396} 397 398/* 399 * IP initialization: fill in IP protocol switch table. 400 * All protocols not implemented in kernel go to raw IP protocol handler. 401 */ 402void 403ip_init(void) 404{ 405 const struct protosw *pr; 406 int i; 407 408 sysctl_net_inet_ip_setup(NULL); 409 410 pool_init(&inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl", 411 NULL, IPL_SOFTNET); 412 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 413 NULL, IPL_VM); 414 415 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 416 if (pr == 0) 417 panic("ip_init"); 418 for (i = 0; i < IPPROTO_MAX; i++) 419 ip_protox[i] = pr - inetsw; 420 for (pr = inetdomain.dom_protosw; 421 pr < inetdomain.dom_protoswNPROTOSW; pr++) 422 if (pr->pr_domain->dom_family == PF_INET && 423 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 424 ip_protox[pr->pr_protocol] = pr - inetsw; 425 426 for (i = 0; i < IPREASS_NHASH; i++) 427 LIST_INIT(&ipq[i]); 428 429 ip_initid(); 430 ip_id = time_second & 0xfffff; 431 432 ipintrq.ifq_maxlen = ipqmaxlen; 433 ip_nmbclusters_changed(); 434 435 TAILQ_INIT(&in_ifaddrhead); 436 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true, 437 &in_ifaddrhash); 438 in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true, 439 &in_multihash); 440 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 441#ifdef GATEWAY 442 ipflow_init(ip_hashsize); 443#endif 444 445#ifdef PFIL_HOOKS 446 /* Register our Packet Filter hook. */ 447 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 448 inet_pfil_hook.ph_af = AF_INET; 449 i = pfil_head_register(&inet_pfil_hook); 450 if (i != 0) 451 printf("ip_init: WARNING: unable to register pfil hook, " 452 "error %d\n", i); 453#endif /* PFIL_HOOKS */ 454 455#ifdef MBUFTRACE 456 MOWNER_ATTACH(&ip_tx_mowner); 457 MOWNER_ATTACH(&ip_rx_mowner); 458#endif /* MBUFTRACE */ 459 460 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS); 461} 462 463struct sockaddr_in ipaddr = { 464 .sin_len = sizeof(ipaddr), 465 .sin_family = AF_INET, 466}; 467struct route ipforward_rt; 468 469/* 470 * IP software interrupt routine 471 */ 472void 473ipintr(void) 474{ 475 int s; 476 struct mbuf *m; 477 478 mutex_enter(softnet_lock); 479 KERNEL_LOCK(1, NULL); 480 while (!IF_IS_EMPTY(&ipintrq)) { 481 s = splnet(); 482 IF_DEQUEUE(&ipintrq, m); 483 splx(s); 484 if (m == NULL) 485 break; 486 KERNEL_UNLOCK_ONE(NULL); 487 ip_input(m); 488 KERNEL_LOCK(1, NULL); 489 } 490 KERNEL_UNLOCK_ONE(NULL); 491 mutex_exit(softnet_lock); 492} 493 494/* 495 * Ip input routine. Checksum and byte swap header. If fragmented 496 * try to reassemble. Process options. Pass to next level. 497 */ 498void 499ip_input(struct mbuf *m) 500{ 501 struct ip *ip = NULL; 502 struct ipq *fp; 503 struct in_ifaddr *ia; 504 struct ifaddr *ifa; 505 struct ipqent *ipqe; 506 int hlen = 0, mff, len; 507 int downmatch; 508 int checkif; 509 int srcrt = 0; 510 int s; 511 u_int hash; 512#ifdef FAST_IPSEC 513 struct m_tag *mtag; 514 struct tdb_ident *tdbi; 515 struct secpolicy *sp; 516 int error; 517#endif /* FAST_IPSEC */ 518 519 MCLAIM(m, &ip_rx_mowner); 520#ifdef DIAGNOSTIC 521 if ((m->m_flags & M_PKTHDR) == 0) 522 panic("ipintr no HDR"); 523#endif 524 525 /* 526 * If no IP addresses have been set yet but the interfaces 527 * are receiving, can't do anything with incoming packets yet. 528 */ 529 if (TAILQ_FIRST(&in_ifaddrhead) == 0) 530 goto bad; 531 IP_STATINC(IP_STAT_TOTAL); 532 /* 533 * If the IP header is not aligned, slurp it up into a new 534 * mbuf with space for link headers, in the event we forward 535 * it. Otherwise, if it is aligned, make sure the entire 536 * base IP header is in the first mbuf of the chain. 537 */ 538 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 539 if ((m = m_copyup(m, sizeof(struct ip), 540 (max_linkhdr + 3) & ~3)) == NULL) { 541 /* XXXJRT new stat, please */ 542 IP_STATINC(IP_STAT_TOOSMALL); 543 return; 544 } 545 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 546 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 547 IP_STATINC(IP_STAT_TOOSMALL); 548 return; 549 } 550 } 551 ip = mtod(m, struct ip *); 552 if (ip->ip_v != IPVERSION) { 553 IP_STATINC(IP_STAT_BADVERS); 554 goto bad; 555 } 556 hlen = ip->ip_hl << 2; 557 if (hlen < sizeof(struct ip)) { /* minimum header length */ 558 IP_STATINC(IP_STAT_BADHLEN); 559 goto bad; 560 } 561 if (hlen > m->m_len) { 562 if ((m = m_pullup(m, hlen)) == 0) { 563 IP_STATINC(IP_STAT_BADHLEN); 564 return; 565 } 566 ip = mtod(m, struct ip *); 567 } 568 569 /* 570 * RFC1122: packets with a multicast source address are 571 * not allowed. 572 */ 573 if (IN_MULTICAST(ip->ip_src.s_addr)) { 574 IP_STATINC(IP_STAT_BADADDR); 575 goto bad; 576 } 577 578 /* 127/8 must not appear on wire - RFC1122 */ 579 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 580 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 581 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 582 IP_STATINC(IP_STAT_BADADDR); 583 goto bad; 584 } 585 } 586 587 switch (m->m_pkthdr.csum_flags & 588 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) | 589 M_CSUM_IPv4_BAD)) { 590 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 591 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 592 goto badcsum; 593 594 case M_CSUM_IPv4: 595 /* Checksum was okay. */ 596 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 597 break; 598 599 default: 600 /* 601 * Must compute it ourselves. Maybe skip checksum on 602 * loopback interfaces. 603 */ 604 if (__predict_true(!(m->m_pkthdr.rcvif->if_flags & 605 IFF_LOOPBACK) || ip_do_loopback_cksum)) { 606 INET_CSUM_COUNTER_INCR(&ip_swcsum); 607 if (in_cksum(m, hlen) != 0) 608 goto badcsum; 609 } 610 break; 611 } 612 613 /* Retrieve the packet length. */ 614 len = ntohs(ip->ip_len); 615 616 /* 617 * Check for additional length bogosity 618 */ 619 if (len < hlen) { 620 IP_STATINC(IP_STAT_BADLEN); 621 goto bad; 622 } 623 624 /* 625 * Check that the amount of data in the buffers 626 * is as at least much as the IP header would have us expect. 627 * Trim mbufs if longer than we expect. 628 * Drop packet if shorter than we expect. 629 */ 630 if (m->m_pkthdr.len < len) { 631 IP_STATINC(IP_STAT_TOOSHORT); 632 goto bad; 633 } 634 if (m->m_pkthdr.len > len) { 635 if (m->m_len == m->m_pkthdr.len) { 636 m->m_len = len; 637 m->m_pkthdr.len = len; 638 } else 639 m_adj(m, len - m->m_pkthdr.len); 640 } 641 642#if defined(IPSEC) 643 /* ipflow (IP fast forwarding) is not compatible with IPsec. */ 644 m->m_flags &= ~M_CANFASTFWD; 645#else 646 /* 647 * Assume that we can create a fast-forward IP flow entry 648 * based on this packet. 649 */ 650 m->m_flags |= M_CANFASTFWD; 651#endif 652 653#ifdef PFIL_HOOKS 654 /* 655 * Run through list of hooks for input packets. If there are any 656 * filters which require that additional packets in the flow are 657 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 658 * Note that filters must _never_ set this flag, as another filter 659 * in the list may have previously cleared it. 660 */ 661 /* 662 * let ipfilter look at packet on the wire, 663 * not the decapsulated packet. 664 */ 665#ifdef IPSEC 666 if (!ipsec_getnhist(m)) 667#elif defined(FAST_IPSEC) 668 if (!ipsec_indone(m)) 669#else 670 if (1) 671#endif 672 { 673 struct in_addr odst; 674 675 odst = ip->ip_dst; 676 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 677 PFIL_IN) != 0) 678 return; 679 if (m == NULL) 680 return; 681 ip = mtod(m, struct ip *); 682 hlen = ip->ip_hl << 2; 683 /* 684 * XXX The setting of "srcrt" here is to prevent ip_forward() 685 * from generating ICMP redirects for packets that have 686 * been redirected by a hook back out on to the same LAN that 687 * they came from and is not an indication that the packet 688 * is being inffluenced by source routing options. This 689 * allows things like 690 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 691 * where tlp0 is both on the 1.1.1.0/24 network and is the 692 * default route for hosts on 1.1.1.0/24. Of course this 693 * also requires a "map tlp0 ..." to complete the story. 694 * One might argue whether or not this kind of network config. 695 * should be supported in this manner... 696 */ 697 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 698 } 699#endif /* PFIL_HOOKS */ 700 701#ifdef ALTQ 702 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 703 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 704 /* packet dropped by traffic conditioner */ 705 return; 706 } 707#endif 708 709 /* 710 * Process options and, if not destined for us, 711 * ship it on. ip_dooptions returns 1 when an 712 * error was detected (causing an icmp message 713 * to be sent and the original packet to be freed). 714 */ 715 ip_nhops = 0; /* for source routed packets */ 716 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 717 return; 718 719 /* 720 * Enable a consistency check between the destination address 721 * and the arrival interface for a unicast packet (the RFC 1122 722 * strong ES model) if IP forwarding is disabled and the packet 723 * is not locally generated. 724 * 725 * XXX - Checking also should be disabled if the destination 726 * address is ipnat'ed to a different interface. 727 * 728 * XXX - Checking is incompatible with IP aliases added 729 * to the loopback interface instead of the interface where 730 * the packets are received. 731 * 732 * XXX - We need to add a per ifaddr flag for this so that 733 * we get finer grain control. 734 */ 735 checkif = ip_checkinterface && (ipforwarding == 0) && 736 (m->m_pkthdr.rcvif != NULL) && 737 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0); 738 739 /* 740 * Check our list of addresses, to see if the packet is for us. 741 * 742 * Traditional 4.4BSD did not consult IFF_UP at all. 743 * The behavior here is to treat addresses on !IFF_UP interface 744 * as not mine. 745 */ 746 downmatch = 0; 747 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 748 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 749 if (checkif && ia->ia_ifp != m->m_pkthdr.rcvif) 750 continue; 751 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 752 break; 753 else 754 downmatch++; 755 } 756 } 757 if (ia != NULL) 758 goto ours; 759 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 760 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) { 761 if (ifa->ifa_addr->sa_family != AF_INET) 762 continue; 763 ia = ifatoia(ifa); 764 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 765 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 766 /* 767 * Look for all-0's host part (old broadcast addr), 768 * either for subnet or net. 769 */ 770 ip->ip_dst.s_addr == ia->ia_subnet || 771 ip->ip_dst.s_addr == ia->ia_net) 772 goto ours; 773 /* 774 * An interface with IP address zero accepts 775 * all packets that arrive on that interface. 776 */ 777 if (in_nullhost(ia->ia_addr.sin_addr)) 778 goto ours; 779 } 780 } 781 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 782 struct in_multi *inm; 783#ifdef MROUTING 784 extern struct socket *ip_mrouter; 785 786 if (ip_mrouter) { 787 /* 788 * If we are acting as a multicast router, all 789 * incoming multicast packets are passed to the 790 * kernel-level multicast forwarding function. 791 * The packet is returned (relatively) intact; if 792 * ip_mforward() returns a non-zero value, the packet 793 * must be discarded, else it may be accepted below. 794 * 795 * (The IP ident field is put in the same byte order 796 * as expected when ip_mforward() is called from 797 * ip_output().) 798 */ 799 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 800 IP_STATINC(IP_STAT_CANTFORWARD); 801 m_freem(m); 802 return; 803 } 804 805 /* 806 * The process-level routing demon needs to receive 807 * all multicast IGMP packets, whether or not this 808 * host belongs to their destination groups. 809 */ 810 if (ip->ip_p == IPPROTO_IGMP) 811 goto ours; 812 IP_STATINC(IP_STAT_CANTFORWARD); 813 } 814#endif 815 /* 816 * See if we belong to the destination multicast group on the 817 * arrival interface. 818 */ 819 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 820 if (inm == NULL) { 821 IP_STATINC(IP_STAT_CANTFORWARD); 822 m_freem(m); 823 return; 824 } 825 goto ours; 826 } 827 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 828 in_nullhost(ip->ip_dst)) 829 goto ours; 830 831 /* 832 * Not for us; forward if possible and desirable. 833 */ 834 if (ipforwarding == 0) { 835 IP_STATINC(IP_STAT_CANTFORWARD); 836 m_freem(m); 837 } else { 838 /* 839 * If ip_dst matched any of my address on !IFF_UP interface, 840 * and there's no IFF_UP interface that matches ip_dst, 841 * send icmp unreach. Forwarding it will result in in-kernel 842 * forwarding loop till TTL goes to 0. 843 */ 844 if (downmatch) { 845 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 846 IP_STATINC(IP_STAT_CANTFORWARD); 847 return; 848 } 849#ifdef IPSEC 850 if (ipsec4_in_reject(m, NULL)) { 851 IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); 852 goto bad; 853 } 854#endif 855#ifdef FAST_IPSEC 856 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 857 s = splsoftnet(); 858 if (mtag != NULL) { 859 tdbi = (struct tdb_ident *)(mtag + 1); 860 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 861 } else { 862 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 863 IP_FORWARDING, &error); 864 } 865 if (sp == NULL) { /* NB: can happen if error */ 866 splx(s); 867 /*XXX error stat???*/ 868 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 869 goto bad; 870 } 871 872 /* 873 * Check security policy against packet attributes. 874 */ 875 error = ipsec_in_reject(sp, m); 876 KEY_FREESP(&sp); 877 splx(s); 878 if (error) { 879 IP_STATINC(IP_STAT_CANTFORWARD); 880 goto bad; 881 } 882 883 /* 884 * Peek at the outbound SP for this packet to determine if 885 * it's a Fast Forward candidate. 886 */ 887 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 888 if (mtag != NULL) 889 m->m_flags &= ~M_CANFASTFWD; 890 else { 891 s = splsoftnet(); 892 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, 893 (IP_FORWARDING | 894 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 895 &error, NULL); 896 if (sp != NULL) { 897 m->m_flags &= ~M_CANFASTFWD; 898 KEY_FREESP(&sp); 899 } 900 splx(s); 901 } 902#endif /* FAST_IPSEC */ 903 904 ip_forward(m, srcrt); 905 } 906 return; 907 908ours: 909 /* 910 * If offset or IP_MF are set, must reassemble. 911 * Otherwise, nothing need be done. 912 * (We could look in the reassembly queue to see 913 * if the packet was previously fragmented, 914 * but it's not worth the time; just let them time out.) 915 */ 916 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 917 uint16_t off; 918 /* 919 * Prevent TCP blind data attacks by not allowing non-initial 920 * fragments to start at less than 68 bytes (minimal fragment 921 * size) and making sure the first fragment is at least 68 922 * bytes. 923 */ 924 off = (ntohs(ip->ip_off) & IP_OFFMASK) << 3; 925 if ((off > 0 ? off + hlen : len) < IP_MINFRAGSIZE - 1) { 926 IP_STATINC(IP_STAT_BADFRAGS); 927 goto bad; 928 } 929 /* 930 * Look for queue of fragments 931 * of this datagram. 932 */ 933 IPQ_LOCK(); 934 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 935 LIST_FOREACH(fp, &ipq[hash], ipq_q) { 936 if (ip->ip_id == fp->ipq_id && 937 in_hosteq(ip->ip_src, fp->ipq_src) && 938 in_hosteq(ip->ip_dst, fp->ipq_dst) && 939 ip->ip_p == fp->ipq_p) { 940 /* 941 * Make sure the TOS is matches previous 942 * fragments. 943 */ 944 if (ip->ip_tos != fp->ipq_tos) { 945 IP_STATINC(IP_STAT_BADFRAGS); 946 IPQ_UNLOCK(); 947 goto bad; 948 } 949 goto found; 950 } 951 } 952 fp = 0; 953found: 954 955 /* 956 * Adjust ip_len to not reflect header, 957 * set ipqe_mff if more fragments are expected, 958 * convert offset of this to bytes. 959 */ 960 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 961 mff = (ip->ip_off & htons(IP_MF)) != 0; 962 if (mff) { 963 /* 964 * Make sure that fragments have a data length 965 * that's a non-zero multiple of 8 bytes. 966 */ 967 if (ntohs(ip->ip_len) == 0 || 968 (ntohs(ip->ip_len) & 0x7) != 0) { 969 IP_STATINC(IP_STAT_BADFRAGS); 970 IPQ_UNLOCK(); 971 goto bad; 972 } 973 } 974 ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3); 975 976 /* 977 * If datagram marked as having more fragments 978 * or if this is not the first fragment, 979 * attempt reassembly; if it succeeds, proceed. 980 */ 981 if (mff || ip->ip_off != htons(0)) { 982 IP_STATINC(IP_STAT_FRAGMENTS); 983 s = splvm(); 984 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 985 splx(s); 986 if (ipqe == NULL) { 987 IP_STATINC(IP_STAT_RCVMEMDROP); 988 IPQ_UNLOCK(); 989 goto bad; 990 } 991 ipqe->ipqe_mff = mff; 992 ipqe->ipqe_m = m; 993 ipqe->ipqe_ip = ip; 994 m = ip_reass(ipqe, fp, &ipq[hash]); 995 if (m == 0) { 996 IPQ_UNLOCK(); 997 return; 998 } 999 IP_STATINC(IP_STAT_REASSEMBLED); 1000 ip = mtod(m, struct ip *); 1001 hlen = ip->ip_hl << 2; 1002 ip->ip_len = htons(ntohs(ip->ip_len) + hlen); 1003 } else 1004 if (fp) 1005 ip_freef(fp); 1006 IPQ_UNLOCK(); 1007 } 1008 1009#if defined(IPSEC) 1010 /* 1011 * enforce IPsec policy checking if we are seeing last header. 1012 * note that we do not visit this with protocols with pcb layer 1013 * code - like udp/tcp/raw ip. 1014 */ 1015 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 1016 ipsec4_in_reject(m, NULL)) { 1017 IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); 1018 goto bad; 1019 } 1020#endif 1021#ifdef FAST_IPSEC 1022 /* 1023 * enforce IPsec policy checking if we are seeing last header. 1024 * note that we do not visit this with protocols with pcb layer 1025 * code - like udp/tcp/raw ip. 1026 */ 1027 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 1028 /* 1029 * Check if the packet has already had IPsec processing 1030 * done. If so, then just pass it along. This tag gets 1031 * set during AH, ESP, etc. input handling, before the 1032 * packet is returned to the ip input queue for delivery. 1033 */ 1034 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 1035 s = splsoftnet(); 1036 if (mtag != NULL) { 1037 tdbi = (struct tdb_ident *)(mtag + 1); 1038 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 1039 } else { 1040 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 1041 IP_FORWARDING, &error); 1042 } 1043 if (sp != NULL) { 1044 /* 1045 * Check security policy against packet attributes. 1046 */ 1047 error = ipsec_in_reject(sp, m); 1048 KEY_FREESP(&sp); 1049 } else { 1050 /* XXX error stat??? */ 1051 error = EINVAL; 1052DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 1053 } 1054 splx(s); 1055 if (error) 1056 goto bad; 1057 } 1058#endif /* FAST_IPSEC */ 1059 1060 /* 1061 * Switch out to protocol's input routine. 1062 */ 1063#if IFA_STATS 1064 if (ia && ip) 1065 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 1066#endif 1067 IP_STATINC(IP_STAT_DELIVERED); 1068 { 1069 int off = hlen, nh = ip->ip_p; 1070 1071 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 1072 return; 1073 } 1074bad: 1075 m_freem(m); 1076 return; 1077 1078badcsum: 1079 IP_STATINC(IP_STAT_BADSUM); 1080 m_freem(m); 1081} 1082 1083/* 1084 * Take incoming datagram fragment and try to 1085 * reassemble it into whole datagram. If a chain for 1086 * reassembly of this datagram already exists, then it 1087 * is given as fp; otherwise have to make a chain. 1088 */ 1089struct mbuf * 1090ip_reass(struct ipqent *ipqe, struct ipq *fp, struct ipqhead *ipqhead) 1091{ 1092 struct mbuf *m = ipqe->ipqe_m; 1093 struct ipqent *nq, *p, *q; 1094 struct ip *ip; 1095 struct mbuf *t; 1096 int hlen = ipqe->ipqe_ip->ip_hl << 2; 1097 int i, next, s; 1098 1099 IPQ_LOCK_CHECK(); 1100 1101 /* 1102 * Presence of header sizes in mbufs 1103 * would confuse code below. 1104 */ 1105 m->m_data += hlen; 1106 m->m_len -= hlen; 1107 1108#ifdef notyet 1109 /* make sure fragment limit is up-to-date */ 1110 CHECK_NMBCLUSTER_PARAMS(); 1111 1112 /* If we have too many fragments, drop the older half. */ 1113 if (ip_nfrags >= ip_maxfrags) 1114 ip_reass_drophalf(void); 1115#endif 1116 1117 /* 1118 * We are about to add a fragment; increment frag count. 1119 */ 1120 ip_nfrags++; 1121 1122 /* 1123 * If first fragment to arrive, create a reassembly queue. 1124 */ 1125 if (fp == 0) { 1126 /* 1127 * Enforce upper bound on number of fragmented packets 1128 * for which we attempt reassembly; 1129 * If maxfrag is 0, never accept fragments. 1130 * If maxfrag is -1, accept all fragments without limitation. 1131 */ 1132 if (ip_maxfragpackets < 0) 1133 ; 1134 else if (ip_nfragpackets >= ip_maxfragpackets) 1135 goto dropfrag; 1136 ip_nfragpackets++; 1137 fp = malloc(sizeof (struct ipq), M_FTABLE, M_NOWAIT); 1138 if (fp == NULL) 1139 goto dropfrag; 1140 LIST_INSERT_HEAD(ipqhead, fp, ipq_q); 1141 fp->ipq_nfrags = 1; 1142 fp->ipq_ttl = IPFRAGTTL; 1143 fp->ipq_p = ipqe->ipqe_ip->ip_p; 1144 fp->ipq_id = ipqe->ipqe_ip->ip_id; 1145 fp->ipq_tos = ipqe->ipqe_ip->ip_tos; 1146 TAILQ_INIT(&fp->ipq_fragq); 1147 fp->ipq_src = ipqe->ipqe_ip->ip_src; 1148 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 1149 p = NULL; 1150 goto insert; 1151 } else { 1152 fp->ipq_nfrags++; 1153 } 1154 1155 /* 1156 * Find a segment which begins after this one does. 1157 */ 1158 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1159 p = q, q = TAILQ_NEXT(q, ipqe_q)) 1160 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off)) 1161 break; 1162 1163 /* 1164 * If there is a preceding segment, it may provide some of 1165 * our data already. If so, drop the data from the incoming 1166 * segment. If it provides all of our data, drop us. 1167 */ 1168 if (p != NULL) { 1169 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) - 1170 ntohs(ipqe->ipqe_ip->ip_off); 1171 if (i > 0) { 1172 if (i >= ntohs(ipqe->ipqe_ip->ip_len)) 1173 goto dropfrag; 1174 m_adj(ipqe->ipqe_m, i); 1175 ipqe->ipqe_ip->ip_off = 1176 htons(ntohs(ipqe->ipqe_ip->ip_off) + i); 1177 ipqe->ipqe_ip->ip_len = 1178 htons(ntohs(ipqe->ipqe_ip->ip_len) - i); 1179 } 1180 } 1181 1182 /* 1183 * While we overlap succeeding segments trim them or, 1184 * if they are completely covered, dequeue them. 1185 */ 1186 for (; q != NULL && 1187 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) > 1188 ntohs(q->ipqe_ip->ip_off); q = nq) { 1189 i = (ntohs(ipqe->ipqe_ip->ip_off) + 1190 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off); 1191 if (i < ntohs(q->ipqe_ip->ip_len)) { 1192 q->ipqe_ip->ip_len = 1193 htons(ntohs(q->ipqe_ip->ip_len) - i); 1194 q->ipqe_ip->ip_off = 1195 htons(ntohs(q->ipqe_ip->ip_off) + i); 1196 m_adj(q->ipqe_m, i); 1197 break; 1198 } 1199 nq = TAILQ_NEXT(q, ipqe_q); 1200 m_freem(q->ipqe_m); 1201 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1202 s = splvm(); 1203 pool_put(&ipqent_pool, q); 1204 splx(s); 1205 fp->ipq_nfrags--; 1206 ip_nfrags--; 1207 } 1208 1209insert: 1210 /* 1211 * Stick new segment in its place; 1212 * check for complete reassembly. 1213 */ 1214 if (p == NULL) { 1215 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 1216 } else { 1217 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q); 1218 } 1219 next = 0; 1220 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1221 p = q, q = TAILQ_NEXT(q, ipqe_q)) { 1222 if (ntohs(q->ipqe_ip->ip_off) != next) 1223 return (0); 1224 next += ntohs(q->ipqe_ip->ip_len); 1225 } 1226 if (p->ipqe_mff) 1227 return (0); 1228 1229 /* 1230 * Reassembly is complete. Check for a bogus message size and 1231 * concatenate fragments. 1232 */ 1233 q = TAILQ_FIRST(&fp->ipq_fragq); 1234 ip = q->ipqe_ip; 1235 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 1236 IP_STATINC(IP_STAT_TOOLONG); 1237 ip_freef(fp); 1238 return (0); 1239 } 1240 m = q->ipqe_m; 1241 t = m->m_next; 1242 m->m_next = 0; 1243 m_cat(m, t); 1244 nq = TAILQ_NEXT(q, ipqe_q); 1245 s = splvm(); 1246 pool_put(&ipqent_pool, q); 1247 splx(s); 1248 for (q = nq; q != NULL; q = nq) { 1249 t = q->ipqe_m; 1250 nq = TAILQ_NEXT(q, ipqe_q); 1251 s = splvm(); 1252 pool_put(&ipqent_pool, q); 1253 splx(s); 1254 m_cat(m, t); 1255 } 1256 ip_nfrags -= fp->ipq_nfrags; 1257 1258 /* 1259 * Create header for new ip packet by 1260 * modifying header of first packet; 1261 * dequeue and discard fragment reassembly header. 1262 * Make header visible. 1263 */ 1264 ip->ip_len = htons(next); 1265 ip->ip_src = fp->ipq_src; 1266 ip->ip_dst = fp->ipq_dst; 1267 LIST_REMOVE(fp, ipq_q); 1268 free(fp, M_FTABLE); 1269 ip_nfragpackets--; 1270 m->m_len += (ip->ip_hl << 2); 1271 m->m_data -= (ip->ip_hl << 2); 1272 /* some debugging cruft by sklower, below, will go away soon */ 1273 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1274 int plen = 0; 1275 for (t = m; t; t = t->m_next) 1276 plen += t->m_len; 1277 m->m_pkthdr.len = plen; 1278 m->m_pkthdr.csum_flags = 0; 1279 } 1280 return (m); 1281 1282dropfrag: 1283 if (fp != 0) 1284 fp->ipq_nfrags--; 1285 ip_nfrags--; 1286 IP_STATINC(IP_STAT_FRAGDROPPED); 1287 m_freem(m); 1288 s = splvm(); 1289 pool_put(&ipqent_pool, ipqe); 1290 splx(s); 1291 return (0); 1292} 1293 1294/* 1295 * Free a fragment reassembly header and all 1296 * associated datagrams. 1297 */ 1298void 1299ip_freef(struct ipq *fp) 1300{ 1301 struct ipqent *q, *p; 1302 u_int nfrags = 0; 1303 int s; 1304 1305 IPQ_LOCK_CHECK(); 1306 1307 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) { 1308 p = TAILQ_NEXT(q, ipqe_q); 1309 m_freem(q->ipqe_m); 1310 nfrags++; 1311 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1312 s = splvm(); 1313 pool_put(&ipqent_pool, q); 1314 splx(s); 1315 } 1316 1317 if (nfrags != fp->ipq_nfrags) 1318 printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags); 1319 ip_nfrags -= nfrags; 1320 LIST_REMOVE(fp, ipq_q); 1321 free(fp, M_FTABLE); 1322 ip_nfragpackets--; 1323} 1324 1325/* 1326 * IP reassembly TTL machinery for multiplicative drop. 1327 */ 1328static u_int fragttl_histo[(IPFRAGTTL+1)]; 1329 1330 1331/* 1332 * Decrement TTL of all reasembly queue entries by `ticks'. 1333 * Count number of distinct fragments (as opposed to partial, fragmented 1334 * datagrams) in the reassembly queue. While we traverse the entire 1335 * reassembly queue, compute and return the median TTL over all fragments. 1336 */ 1337static u_int 1338ip_reass_ttl_decr(u_int ticks) 1339{ 1340 u_int nfrags, median, dropfraction, keepfraction; 1341 struct ipq *fp, *nfp; 1342 int i; 1343 1344 nfrags = 0; 1345 memset(fragttl_histo, 0, sizeof fragttl_histo); 1346 1347 for (i = 0; i < IPREASS_NHASH; i++) { 1348 for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) { 1349 fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ? 1350 0 : fp->ipq_ttl - ticks); 1351 nfp = LIST_NEXT(fp, ipq_q); 1352 if (fp->ipq_ttl == 0) { 1353 IP_STATINC(IP_STAT_FRAGTIMEOUT); 1354 ip_freef(fp); 1355 } else { 1356 nfrags += fp->ipq_nfrags; 1357 fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags; 1358 } 1359 } 1360 } 1361 1362 KASSERT(ip_nfrags == nfrags); 1363 1364 /* Find median (or other drop fraction) in histogram. */ 1365 dropfraction = (ip_nfrags / 2); 1366 keepfraction = ip_nfrags - dropfraction; 1367 for (i = IPFRAGTTL, median = 0; i >= 0; i--) { 1368 median += fragttl_histo[i]; 1369 if (median >= keepfraction) 1370 break; 1371 } 1372 1373 /* Return TTL of median (or other fraction). */ 1374 return (u_int)i; 1375} 1376 1377void 1378ip_reass_drophalf(void) 1379{ 1380 1381 u_int median_ticks; 1382 /* 1383 * Compute median TTL of all fragments, and count frags 1384 * with that TTL or lower (roughly half of all fragments). 1385 */ 1386 median_ticks = ip_reass_ttl_decr(0); 1387 1388 /* Drop half. */ 1389 median_ticks = ip_reass_ttl_decr(median_ticks); 1390 1391} 1392 1393/* 1394 * IP timer processing; 1395 * if a timer expires on a reassembly 1396 * queue, discard it. 1397 */ 1398void 1399ip_slowtimo(void) 1400{ 1401 static u_int dropscanidx = 0; 1402 u_int i; 1403 u_int median_ttl; 1404 1405 mutex_enter(softnet_lock); 1406 KERNEL_LOCK(1, NULL); 1407 1408 IPQ_LOCK(); 1409 1410 /* Age TTL of all fragments by 1 tick .*/ 1411 median_ttl = ip_reass_ttl_decr(1); 1412 1413 /* make sure fragment limit is up-to-date */ 1414 CHECK_NMBCLUSTER_PARAMS(); 1415 1416 /* If we have too many fragments, drop the older half. */ 1417 if (ip_nfrags > ip_maxfrags) 1418 ip_reass_ttl_decr(median_ttl); 1419 1420 /* 1421 * If we are over the maximum number of fragmented packets 1422 * (due to the limit being lowered), drain off 1423 * enough to get down to the new limit. Start draining 1424 * from the reassembly hashqueue most recently drained. 1425 */ 1426 if (ip_maxfragpackets < 0) 1427 ; 1428 else { 1429 int wrapped = 0; 1430 1431 i = dropscanidx; 1432 while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) { 1433 while (LIST_FIRST(&ipq[i]) != NULL) 1434 ip_freef(LIST_FIRST(&ipq[i])); 1435 if (++i >= IPREASS_NHASH) { 1436 i = 0; 1437 } 1438 /* 1439 * Dont scan forever even if fragment counters are 1440 * wrong: stop after scanning entire reassembly queue. 1441 */ 1442 if (i == dropscanidx) 1443 wrapped = 1; 1444 } 1445 dropscanidx = i; 1446 } 1447 IPQ_UNLOCK(); 1448 1449 KERNEL_UNLOCK_ONE(NULL); 1450 mutex_exit(softnet_lock); 1451} 1452 1453/* 1454 * Drain off all datagram fragments. Don't acquire softnet_lock as 1455 * can be called from hardware interrupt context. 1456 */ 1457void 1458ip_drain(void) 1459{ 1460 1461 KERNEL_LOCK(1, NULL); 1462 1463 /* 1464 * We may be called from a device's interrupt context. If 1465 * the ipq is already busy, just bail out now. 1466 */ 1467 if (ipq_lock_try() != 0) { 1468 /* 1469 * Drop half the total fragments now. If more mbufs are 1470 * needed, we will be called again soon. 1471 */ 1472 ip_reass_drophalf(); 1473 IPQ_UNLOCK(); 1474 } 1475 1476 KERNEL_UNLOCK_ONE(NULL); 1477} 1478 1479/* 1480 * Do option processing on a datagram, 1481 * possibly discarding it if bad options are encountered, 1482 * or forwarding it if source-routed. 1483 * Returns 1 if packet has been forwarded/freed, 1484 * 0 if the packet should be processed further. 1485 */ 1486int 1487ip_dooptions(struct mbuf *m) 1488{ 1489 struct ip *ip = mtod(m, struct ip *); 1490 u_char *cp, *cp0; 1491 struct ip_timestamp *ipt; 1492 struct in_ifaddr *ia; 1493 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1494 struct in_addr dst; 1495 n_time ntime; 1496 1497 dst = ip->ip_dst; 1498 cp = (u_char *)(ip + 1); 1499 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1500 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1501 opt = cp[IPOPT_OPTVAL]; 1502 if (opt == IPOPT_EOL) 1503 break; 1504 if (opt == IPOPT_NOP) 1505 optlen = 1; 1506 else { 1507 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1508 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1509 goto bad; 1510 } 1511 optlen = cp[IPOPT_OLEN]; 1512 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1513 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1514 goto bad; 1515 } 1516 } 1517 switch (opt) { 1518 1519 default: 1520 break; 1521 1522 /* 1523 * Source routing with record. 1524 * Find interface with current destination address. 1525 * If none on this machine then drop if strictly routed, 1526 * or do nothing if loosely routed. 1527 * Record interface address and bring up next address 1528 * component. If strictly routed make sure next 1529 * address is on directly accessible net. 1530 */ 1531 case IPOPT_LSRR: 1532 case IPOPT_SSRR: 1533 if (ip_allowsrcrt == 0) { 1534 type = ICMP_UNREACH; 1535 code = ICMP_UNREACH_NET_PROHIB; 1536 goto bad; 1537 } 1538 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1539 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1540 goto bad; 1541 } 1542 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1543 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1544 goto bad; 1545 } 1546 ipaddr.sin_addr = ip->ip_dst; 1547 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1548 if (ia == 0) { 1549 if (opt == IPOPT_SSRR) { 1550 type = ICMP_UNREACH; 1551 code = ICMP_UNREACH_SRCFAIL; 1552 goto bad; 1553 } 1554 /* 1555 * Loose routing, and not at next destination 1556 * yet; nothing to do except forward. 1557 */ 1558 break; 1559 } 1560 off--; /* 0 origin */ 1561 if ((off + sizeof(struct in_addr)) > optlen) { 1562 /* 1563 * End of source route. Should be for us. 1564 */ 1565 save_rte(cp, ip->ip_src); 1566 break; 1567 } 1568 /* 1569 * locate outgoing interface 1570 */ 1571 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 1572 sizeof(ipaddr.sin_addr)); 1573 if (opt == IPOPT_SSRR) 1574 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); 1575 else 1576 ia = ip_rtaddr(ipaddr.sin_addr); 1577 if (ia == 0) { 1578 type = ICMP_UNREACH; 1579 code = ICMP_UNREACH_SRCFAIL; 1580 goto bad; 1581 } 1582 ip->ip_dst = ipaddr.sin_addr; 1583 bcopy((void *)&ia->ia_addr.sin_addr, 1584 (void *)(cp + off), sizeof(struct in_addr)); 1585 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1586 /* 1587 * Let ip_intr's mcast routing check handle mcast pkts 1588 */ 1589 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1590 break; 1591 1592 case IPOPT_RR: 1593 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1594 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1595 goto bad; 1596 } 1597 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1598 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1599 goto bad; 1600 } 1601 /* 1602 * If no space remains, ignore. 1603 */ 1604 off--; /* 0 origin */ 1605 if ((off + sizeof(struct in_addr)) > optlen) 1606 break; 1607 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst), 1608 sizeof(ipaddr.sin_addr)); 1609 /* 1610 * locate outgoing interface; if we're the destination, 1611 * use the incoming interface (should be same). 1612 */ 1613 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1614 == NULL && 1615 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1616 type = ICMP_UNREACH; 1617 code = ICMP_UNREACH_HOST; 1618 goto bad; 1619 } 1620 bcopy((void *)&ia->ia_addr.sin_addr, 1621 (void *)(cp + off), sizeof(struct in_addr)); 1622 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1623 break; 1624 1625 case IPOPT_TS: 1626 code = cp - (u_char *)ip; 1627 ipt = (struct ip_timestamp *)cp; 1628 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1629 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1630 goto bad; 1631 } 1632 if (ipt->ipt_ptr < 5) { 1633 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1634 goto bad; 1635 } 1636 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1637 if (++ipt->ipt_oflw == 0) { 1638 code = (u_char *)&ipt->ipt_ptr - 1639 (u_char *)ip; 1640 goto bad; 1641 } 1642 break; 1643 } 1644 cp0 = (cp + ipt->ipt_ptr - 1); 1645 switch (ipt->ipt_flg) { 1646 1647 case IPOPT_TS_TSONLY: 1648 break; 1649 1650 case IPOPT_TS_TSANDADDR: 1651 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1652 sizeof(struct in_addr) > ipt->ipt_len) { 1653 code = (u_char *)&ipt->ipt_ptr - 1654 (u_char *)ip; 1655 goto bad; 1656 } 1657 ipaddr.sin_addr = dst; 1658 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1659 m->m_pkthdr.rcvif)); 1660 if (ia == 0) 1661 continue; 1662 bcopy(&ia->ia_addr.sin_addr, 1663 cp0, sizeof(struct in_addr)); 1664 ipt->ipt_ptr += sizeof(struct in_addr); 1665 break; 1666 1667 case IPOPT_TS_PRESPEC: 1668 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1669 sizeof(struct in_addr) > ipt->ipt_len) { 1670 code = (u_char *)&ipt->ipt_ptr - 1671 (u_char *)ip; 1672 goto bad; 1673 } 1674 memcpy(&ipaddr.sin_addr, cp0, 1675 sizeof(struct in_addr)); 1676 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1677 == NULL) 1678 continue; 1679 ipt->ipt_ptr += sizeof(struct in_addr); 1680 break; 1681 1682 default: 1683 /* XXX can't take &ipt->ipt_flg */ 1684 code = (u_char *)&ipt->ipt_ptr - 1685 (u_char *)ip + 1; 1686 goto bad; 1687 } 1688 ntime = iptime(); 1689 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1690 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1691 sizeof(n_time)); 1692 ipt->ipt_ptr += sizeof(n_time); 1693 } 1694 } 1695 if (forward) { 1696 if (ip_forwsrcrt == 0) { 1697 type = ICMP_UNREACH; 1698 code = ICMP_UNREACH_SRCFAIL; 1699 goto bad; 1700 } 1701 ip_forward(m, 1); 1702 return (1); 1703 } 1704 return (0); 1705bad: 1706 icmp_error(m, type, code, 0, 0); 1707 IP_STATINC(IP_STAT_BADOPTIONS); 1708 return (1); 1709} 1710 1711/* 1712 * Given address of next destination (final or next hop), 1713 * return internet address info of interface to be used to get there. 1714 */ 1715struct in_ifaddr * 1716ip_rtaddr(struct in_addr dst) 1717{ 1718 struct rtentry *rt; 1719 union { 1720 struct sockaddr dst; 1721 struct sockaddr_in dst4; 1722 } u; 1723 1724 sockaddr_in_init(&u.dst4, &dst, 0); 1725 1726 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) 1727 return NULL; 1728 1729 return ifatoia(rt->rt_ifa); 1730} 1731 1732/* 1733 * Save incoming source route for use in replies, 1734 * to be picked up later by ip_srcroute if the receiver is interested. 1735 */ 1736void 1737save_rte(u_char *option, struct in_addr dst) 1738{ 1739 unsigned olen; 1740 1741 olen = option[IPOPT_OLEN]; 1742#ifdef DIAGNOSTIC 1743 if (ipprintfs) 1744 printf("save_rte: olen %d\n", olen); 1745#endif /* 0 */ 1746 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1747 return; 1748 memcpy((void *)ip_srcrt.srcopt, (void *)option, olen); 1749 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1750 ip_srcrt.dst = dst; 1751} 1752 1753/* 1754 * Retrieve incoming source route for use in replies, 1755 * in the same form used by setsockopt. 1756 * The first hop is placed before the options, will be removed later. 1757 */ 1758struct mbuf * 1759ip_srcroute(void) 1760{ 1761 struct in_addr *p, *q; 1762 struct mbuf *m; 1763 1764 if (ip_nhops == 0) 1765 return NULL; 1766 m = m_get(M_DONTWAIT, MT_SOOPTS); 1767 if (m == 0) 1768 return NULL; 1769 1770 MCLAIM(m, &inetdomain.dom_mowner); 1771#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1772 1773 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1774 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1775 OPTSIZ; 1776#ifdef DIAGNOSTIC 1777 if (ipprintfs) 1778 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1779#endif 1780 1781 /* 1782 * First save first hop for return route 1783 */ 1784 p = &ip_srcrt.route[ip_nhops - 1]; 1785 *(mtod(m, struct in_addr *)) = *p--; 1786#ifdef DIAGNOSTIC 1787 if (ipprintfs) 1788 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1789#endif 1790 1791 /* 1792 * Copy option fields and padding (nop) to mbuf. 1793 */ 1794 ip_srcrt.nop = IPOPT_NOP; 1795 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1796 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop, 1797 OPTSIZ); 1798 q = (struct in_addr *)(mtod(m, char *) + 1799 sizeof(struct in_addr) + OPTSIZ); 1800#undef OPTSIZ 1801 /* 1802 * Record return path as an IP source route, 1803 * reversing the path (pointers are now aligned). 1804 */ 1805 while (p >= ip_srcrt.route) { 1806#ifdef DIAGNOSTIC 1807 if (ipprintfs) 1808 printf(" %x", ntohl(q->s_addr)); 1809#endif 1810 *q++ = *p--; 1811 } 1812 /* 1813 * Last hop goes to final destination. 1814 */ 1815 *q = ip_srcrt.dst; 1816#ifdef DIAGNOSTIC 1817 if (ipprintfs) 1818 printf(" %x\n", ntohl(q->s_addr)); 1819#endif 1820 return (m); 1821} 1822 1823const int inetctlerrmap[PRC_NCMDS] = { 1824 [PRC_MSGSIZE] = EMSGSIZE, 1825 [PRC_HOSTDEAD] = EHOSTDOWN, 1826 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1827 [PRC_UNREACH_NET] = EHOSTUNREACH, 1828 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1829 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1830 [PRC_UNREACH_PORT] = ECONNREFUSED, 1831 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1832 [PRC_PARAMPROB] = ENOPROTOOPT, 1833}; 1834 1835/* 1836 * Forward a packet. If some error occurs return the sender 1837 * an icmp packet. Note we can't always generate a meaningful 1838 * icmp message because icmp doesn't have a large enough repertoire 1839 * of codes and types. 1840 * 1841 * If not forwarding, just drop the packet. This could be confusing 1842 * if ipforwarding was zero but some routing protocol was advancing 1843 * us as a gateway to somewhere. However, we must let the routing 1844 * protocol deal with that. 1845 * 1846 * The srcrt parameter indicates whether the packet is being forwarded 1847 * via a source route. 1848 */ 1849void 1850ip_forward(struct mbuf *m, int srcrt) 1851{ 1852 struct ip *ip = mtod(m, struct ip *); 1853 struct rtentry *rt; 1854 int error, type = 0, code = 0, destmtu = 0; 1855 struct mbuf *mcopy; 1856 n_long dest; 1857 union { 1858 struct sockaddr dst; 1859 struct sockaddr_in dst4; 1860 } u; 1861 1862 /* 1863 * We are now in the output path. 1864 */ 1865 MCLAIM(m, &ip_tx_mowner); 1866 1867 /* 1868 * Clear any in-bound checksum flags for this packet. 1869 */ 1870 m->m_pkthdr.csum_flags = 0; 1871 1872 dest = 0; 1873#ifdef DIAGNOSTIC 1874 if (ipprintfs) { 1875 printf("forward: src %s ", inet_ntoa(ip->ip_src)); 1876 printf("dst %s ttl %x\n", inet_ntoa(ip->ip_dst), ip->ip_ttl); 1877 } 1878#endif 1879 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1880 IP_STATINC(IP_STAT_CANTFORWARD); 1881 m_freem(m); 1882 return; 1883 } 1884 if (ip->ip_ttl <= IPTTLDEC) { 1885 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1886 return; 1887 } 1888 1889 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1890 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) { 1891 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1892 return; 1893 } 1894 1895 /* 1896 * Save at most 68 bytes of the packet in case 1897 * we need to generate an ICMP message to the src. 1898 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1899 */ 1900 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1901 if (mcopy) 1902 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1903 1904 ip->ip_ttl -= IPTTLDEC; 1905 1906 /* 1907 * If forwarding packet using same interface that it came in on, 1908 * perhaps should send a redirect to sender to shortcut a hop. 1909 * Only send redirect if source is sending directly to us, 1910 * and if packet was not source routed (or has any options). 1911 * Also, don't send redirect if forwarding using a default route 1912 * or a route modified by a redirect. 1913 */ 1914 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1915 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1916 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1917 ipsendredirects && !srcrt) { 1918 if (rt->rt_ifa && 1919 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1920 ifatoia(rt->rt_ifa)->ia_subnet) { 1921 if (rt->rt_flags & RTF_GATEWAY) 1922 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1923 else 1924 dest = ip->ip_dst.s_addr; 1925 /* 1926 * Router requirements says to only send host 1927 * redirects. 1928 */ 1929 type = ICMP_REDIRECT; 1930 code = ICMP_REDIRECT_HOST; 1931#ifdef DIAGNOSTIC 1932 if (ipprintfs) 1933 printf("redirect (%d) to %x\n", code, 1934 (u_int32_t)dest); 1935#endif 1936 } 1937 } 1938 1939 error = ip_output(m, NULL, &ipforward_rt, 1940 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1941 (struct ip_moptions *)NULL, (struct socket *)NULL); 1942 1943 if (error) 1944 IP_STATINC(IP_STAT_CANTFORWARD); 1945 else { 1946 uint64_t *ips = IP_STAT_GETREF(); 1947 ips[IP_STAT_FORWARD]++; 1948 if (type) { 1949 ips[IP_STAT_REDIRECTSENT]++; 1950 IP_STAT_PUTREF(); 1951 } else { 1952 IP_STAT_PUTREF(); 1953 if (mcopy) { 1954#ifdef GATEWAY 1955 if (mcopy->m_flags & M_CANFASTFWD) 1956 ipflow_create(&ipforward_rt, mcopy); 1957#endif 1958 m_freem(mcopy); 1959 } 1960 return; 1961 } 1962 } 1963 if (mcopy == NULL) 1964 return; 1965 1966 switch (error) { 1967 1968 case 0: /* forwarded, but need redirect */ 1969 /* type, code set above */ 1970 break; 1971 1972 case ENETUNREACH: /* shouldn't happen, checked above */ 1973 case EHOSTUNREACH: 1974 case ENETDOWN: 1975 case EHOSTDOWN: 1976 default: 1977 type = ICMP_UNREACH; 1978 code = ICMP_UNREACH_HOST; 1979 break; 1980 1981 case EMSGSIZE: 1982 type = ICMP_UNREACH; 1983 code = ICMP_UNREACH_NEEDFRAG; 1984 1985 if ((rt = rtcache_validate(&ipforward_rt)) != NULL) 1986 destmtu = rt->rt_ifp->if_mtu; 1987 1988#if defined(IPSEC) || defined(FAST_IPSEC) 1989 { 1990 /* 1991 * If the packet is routed over IPsec tunnel, tell the 1992 * originator the tunnel MTU. 1993 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1994 * XXX quickhack!!! 1995 */ 1996 1997 struct secpolicy *sp; 1998 int ipsecerror; 1999 size_t ipsechdr; 2000 struct route *ro; 2001 2002 sp = ipsec4_getpolicybyaddr(mcopy, 2003 IPSEC_DIR_OUTBOUND, IP_FORWARDING, 2004 &ipsecerror); 2005 2006 if (sp != NULL) { 2007 /* count IPsec header size */ 2008 ipsechdr = ipsec4_hdrsiz(mcopy, 2009 IPSEC_DIR_OUTBOUND, NULL); 2010 2011 /* 2012 * find the correct route for outer IPv4 2013 * header, compute tunnel MTU. 2014 */ 2015 2016 if (sp->req != NULL 2017 && sp->req->sav != NULL 2018 && sp->req->sav->sah != NULL) { 2019 ro = &sp->req->sav->sah->sa_route; 2020 rt = rtcache_validate(ro); 2021 if (rt && rt->rt_ifp) { 2022 destmtu = 2023 rt->rt_rmx.rmx_mtu ? 2024 rt->rt_rmx.rmx_mtu : 2025 rt->rt_ifp->if_mtu; 2026 destmtu -= ipsechdr; 2027 } 2028 } 2029 2030#ifdef IPSEC 2031 key_freesp(sp); 2032#else 2033 KEY_FREESP(&sp); 2034#endif 2035 } 2036 } 2037#endif /*defined(IPSEC) || defined(FAST_IPSEC)*/ 2038 IP_STATINC(IP_STAT_CANTFRAG); 2039 break; 2040 2041 case ENOBUFS: 2042#if 1 2043 /* 2044 * a router should not generate ICMP_SOURCEQUENCH as 2045 * required in RFC1812 Requirements for IP Version 4 Routers. 2046 * source quench could be a big problem under DoS attacks, 2047 * or if the underlying interface is rate-limited. 2048 */ 2049 if (mcopy) 2050 m_freem(mcopy); 2051 return; 2052#else 2053 type = ICMP_SOURCEQUENCH; 2054 code = 0; 2055 break; 2056#endif 2057 } 2058 icmp_error(mcopy, type, code, dest, destmtu); 2059} 2060 2061void 2062ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2063 struct mbuf *m) 2064{ 2065 2066 if (inp->inp_socket->so_options & SO_TIMESTAMP 2067#ifdef SO_OTIMESTAMP 2068 || inp->inp_socket->so_options & SO_OTIMESTAMP 2069#endif 2070 ) { 2071 struct timeval tv; 2072 2073 microtime(&tv); 2074#ifdef SO_OTIMESTAMP 2075 if (inp->inp_socket->so_options & SO_OTIMESTAMP) { 2076 struct timeval50 tv50; 2077 timeval_to_timeval50(&tv, &tv50); 2078 *mp = sbcreatecontrol((void *) &tv50, sizeof(tv50), 2079 SCM_OTIMESTAMP, SOL_SOCKET); 2080 } else 2081#endif 2082 *mp = sbcreatecontrol((void *) &tv, sizeof(tv), 2083 SCM_TIMESTAMP, SOL_SOCKET); 2084 if (*mp) 2085 mp = &(*mp)->m_next; 2086 } 2087 if (inp->inp_flags & INP_RECVDSTADDR) { 2088 *mp = sbcreatecontrol((void *) &ip->ip_dst, 2089 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2090 if (*mp) 2091 mp = &(*mp)->m_next; 2092 } 2093#ifdef notyet 2094 /* 2095 * XXX 2096 * Moving these out of udp_input() made them even more broken 2097 * than they already were. 2098 * - fenner@parc.xerox.com 2099 */ 2100 /* options were tossed already */ 2101 if (inp->inp_flags & INP_RECVOPTS) { 2102 *mp = sbcreatecontrol((void *) opts_deleted_above, 2103 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2104 if (*mp) 2105 mp = &(*mp)->m_next; 2106 } 2107 /* ip_srcroute doesn't do what we want here, need to fix */ 2108 if (inp->inp_flags & INP_RECVRETOPTS) { 2109 *mp = sbcreatecontrol((void *) ip_srcroute(), 2110 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2111 if (*mp) 2112 mp = &(*mp)->m_next; 2113 } 2114#endif 2115 if (inp->inp_flags & INP_RECVIF) { 2116 struct sockaddr_dl sdl; 2117 2118 sockaddr_dl_init(&sdl, sizeof(sdl), 2119 (m->m_pkthdr.rcvif != NULL) 2120 ? m->m_pkthdr.rcvif->if_index 2121 : 0, 2122 0, NULL, 0, NULL, 0); 2123 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 2124 if (*mp) 2125 mp = &(*mp)->m_next; 2126 } 2127 if (inp->inp_flags & INP_RECVTTL) { 2128 *mp = sbcreatecontrol((void *) &ip->ip_ttl, 2129 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 2130 if (*mp) 2131 mp = &(*mp)->m_next; 2132 } 2133} 2134 2135/* 2136 * sysctl helper routine for net.inet.ip.forwsrcrt. 2137 */ 2138static int 2139sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 2140{ 2141 int error, tmp; 2142 struct sysctlnode node; 2143 2144 node = *rnode; 2145 tmp = ip_forwsrcrt; 2146 node.sysctl_data = &tmp; 2147 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2148 if (error || newp == NULL) 2149 return (error); 2150 2151 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 2152 0, NULL, NULL, NULL); 2153 if (error) 2154 return (error); 2155 2156 ip_forwsrcrt = tmp; 2157 2158 return (0); 2159} 2160 2161/* 2162 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 2163 * range of the new value and tweaks timers if it changes. 2164 */ 2165static int 2166sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 2167{ 2168 int error, tmp; 2169 struct sysctlnode node; 2170 2171 node = *rnode; 2172 tmp = ip_mtudisc_timeout; 2173 node.sysctl_data = &tmp; 2174 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2175 if (error || newp == NULL) 2176 return (error); 2177 if (tmp < 0) 2178 return (EINVAL); 2179 2180 mutex_enter(softnet_lock); 2181 2182 ip_mtudisc_timeout = tmp; 2183 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 2184 2185 mutex_exit(softnet_lock); 2186 2187 return (0); 2188} 2189 2190#ifdef GATEWAY 2191/* 2192 * sysctl helper routine for net.inet.ip.maxflows. 2193 */ 2194static int 2195sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS) 2196{ 2197 int error; 2198 2199 error = sysctl_lookup(SYSCTLFN_CALL(rnode)); 2200 if (error || newp == NULL) 2201 return (error); 2202 2203 mutex_enter(softnet_lock); 2204 KERNEL_LOCK(1, NULL); 2205 2206 ipflow_prune(); 2207 2208 KERNEL_UNLOCK_ONE(NULL); 2209 mutex_exit(softnet_lock); 2210 2211 return (0); 2212} 2213 2214static int 2215sysctl_net_inet_ip_hashsize(SYSCTLFN_ARGS) 2216{ 2217 int error, tmp; 2218 struct sysctlnode node; 2219 2220 node = *rnode; 2221 tmp = ip_hashsize; 2222 node.sysctl_data = &tmp; 2223 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2224 if (error || newp == NULL) 2225 return (error); 2226 2227 if ((tmp & (tmp - 1)) == 0 && tmp != 0) { 2228 /* 2229 * Can only fail due to malloc() 2230 */ 2231 mutex_enter(softnet_lock); 2232 KERNEL_LOCK(1, NULL); 2233 2234 error = ipflow_invalidate_all(tmp); 2235 2236 KERNEL_UNLOCK_ONE(NULL); 2237 mutex_exit(softnet_lock); 2238 2239 } else { 2240 /* 2241 * EINVAL if not a power of 2 2242 */ 2243 error = EINVAL; 2244 } 2245 2246 return error; 2247} 2248#endif /* GATEWAY */ 2249 2250static int 2251sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 2252{ 2253 2254 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 2255} 2256 2257static void 2258sysctl_net_inet_ip_setup(struct sysctllog **clog) 2259{ 2260 extern int subnetsarelocal, hostzeroisbroadcast; 2261 2262 sysctl_createv(clog, 0, NULL, NULL, 2263 CTLFLAG_PERMANENT, 2264 CTLTYPE_NODE, "net", NULL, 2265 NULL, 0, NULL, 0, 2266 CTL_NET, CTL_EOL); 2267 sysctl_createv(clog, 0, NULL, NULL, 2268 CTLFLAG_PERMANENT, 2269 CTLTYPE_NODE, "inet", 2270 SYSCTL_DESCR("PF_INET related settings"), 2271 NULL, 0, NULL, 0, 2272 CTL_NET, PF_INET, CTL_EOL); 2273 sysctl_createv(clog, 0, NULL, NULL, 2274 CTLFLAG_PERMANENT, 2275 CTLTYPE_NODE, "ip", 2276 SYSCTL_DESCR("IPv4 related settings"), 2277 NULL, 0, NULL, 0, 2278 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 2279 2280 sysctl_createv(clog, 0, NULL, NULL, 2281 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2282 CTLTYPE_INT, "forwarding", 2283 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 2284 NULL, 0, &ipforwarding, 0, 2285 CTL_NET, PF_INET, IPPROTO_IP, 2286 IPCTL_FORWARDING, CTL_EOL); 2287 sysctl_createv(clog, 0, NULL, NULL, 2288 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2289 CTLTYPE_INT, "redirect", 2290 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 2291 NULL, 0, &ipsendredirects, 0, 2292 CTL_NET, PF_INET, IPPROTO_IP, 2293 IPCTL_SENDREDIRECTS, CTL_EOL); 2294 sysctl_createv(clog, 0, NULL, NULL, 2295 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2296 CTLTYPE_INT, "ttl", 2297 SYSCTL_DESCR("Default TTL for an INET datagram"), 2298 NULL, 0, &ip_defttl, 0, 2299 CTL_NET, PF_INET, IPPROTO_IP, 2300 IPCTL_DEFTTL, CTL_EOL); 2301#ifdef IPCTL_DEFMTU 2302 sysctl_createv(clog, 0, NULL, NULL, 2303 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 2304 CTLTYPE_INT, "mtu", 2305 SYSCTL_DESCR("Default MTA for an INET route"), 2306 NULL, 0, &ip_mtu, 0, 2307 CTL_NET, PF_INET, IPPROTO_IP, 2308 IPCTL_DEFMTU, CTL_EOL); 2309#endif /* IPCTL_DEFMTU */ 2310 sysctl_createv(clog, 0, NULL, NULL, 2311 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2312 CTLTYPE_INT, "forwsrcrt", 2313 SYSCTL_DESCR("Enable forwarding of source-routed " 2314 "datagrams"), 2315 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 2316 CTL_NET, PF_INET, IPPROTO_IP, 2317 IPCTL_FORWSRCRT, CTL_EOL); 2318 sysctl_createv(clog, 0, NULL, NULL, 2319 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2320 CTLTYPE_INT, "directed-broadcast", 2321 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 2322 NULL, 0, &ip_directedbcast, 0, 2323 CTL_NET, PF_INET, IPPROTO_IP, 2324 IPCTL_DIRECTEDBCAST, CTL_EOL); 2325 sysctl_createv(clog, 0, NULL, NULL, 2326 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2327 CTLTYPE_INT, "allowsrcrt", 2328 SYSCTL_DESCR("Accept source-routed datagrams"), 2329 NULL, 0, &ip_allowsrcrt, 0, 2330 CTL_NET, PF_INET, IPPROTO_IP, 2331 IPCTL_ALLOWSRCRT, CTL_EOL); 2332 sysctl_createv(clog, 0, NULL, NULL, 2333 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2334 CTLTYPE_INT, "subnetsarelocal", 2335 SYSCTL_DESCR("Whether logical subnets are considered " 2336 "local"), 2337 NULL, 0, &subnetsarelocal, 0, 2338 CTL_NET, PF_INET, IPPROTO_IP, 2339 IPCTL_SUBNETSARELOCAL, CTL_EOL); 2340 sysctl_createv(clog, 0, NULL, NULL, 2341 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2342 CTLTYPE_INT, "mtudisc", 2343 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 2344 NULL, 0, &ip_mtudisc, 0, 2345 CTL_NET, PF_INET, IPPROTO_IP, 2346 IPCTL_MTUDISC, CTL_EOL); 2347 sysctl_createv(clog, 0, NULL, NULL, 2348 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2349 CTLTYPE_INT, "anonportmin", 2350 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 2351 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 2352 CTL_NET, PF_INET, IPPROTO_IP, 2353 IPCTL_ANONPORTMIN, CTL_EOL); 2354 sysctl_createv(clog, 0, NULL, NULL, 2355 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2356 CTLTYPE_INT, "anonportmax", 2357 SYSCTL_DESCR("Highest ephemeral port number to assign"), 2358 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 2359 CTL_NET, PF_INET, IPPROTO_IP, 2360 IPCTL_ANONPORTMAX, CTL_EOL); 2361 sysctl_createv(clog, 0, NULL, NULL, 2362 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2363 CTLTYPE_INT, "mtudisctimeout", 2364 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 2365 sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0, 2366 CTL_NET, PF_INET, IPPROTO_IP, 2367 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 2368#ifdef GATEWAY 2369 sysctl_createv(clog, 0, NULL, NULL, 2370 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2371 CTLTYPE_INT, "maxflows", 2372 SYSCTL_DESCR("Number of flows for fast forwarding"), 2373 sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0, 2374 CTL_NET, PF_INET, IPPROTO_IP, 2375 IPCTL_MAXFLOWS, CTL_EOL); 2376 sysctl_createv(clog, 0, NULL, NULL, 2377 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2378 CTLTYPE_INT, "hashsize", 2379 SYSCTL_DESCR("Size of hash table for fast forwarding (IPv4)"), 2380 sysctl_net_inet_ip_hashsize, 0, &ip_hashsize, 0, 2381 CTL_NET, PF_INET, IPPROTO_IP, 2382 CTL_CREATE, CTL_EOL); 2383#endif /* GATEWAY */ 2384 sysctl_createv(clog, 0, NULL, NULL, 2385 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2386 CTLTYPE_INT, "hostzerobroadcast", 2387 SYSCTL_DESCR("All zeroes address is broadcast address"), 2388 NULL, 0, &hostzeroisbroadcast, 0, 2389 CTL_NET, PF_INET, IPPROTO_IP, 2390 IPCTL_HOSTZEROBROADCAST, CTL_EOL); 2391#if NGIF > 0 2392 sysctl_createv(clog, 0, NULL, NULL, 2393 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2394 CTLTYPE_INT, "gifttl", 2395 SYSCTL_DESCR("Default TTL for a gif tunnel datagram"), 2396 NULL, 0, &ip_gif_ttl, 0, 2397 CTL_NET, PF_INET, IPPROTO_IP, 2398 IPCTL_GIF_TTL, CTL_EOL); 2399#endif /* NGIF */ 2400#ifndef IPNOPRIVPORTS 2401 sysctl_createv(clog, 0, NULL, NULL, 2402 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2403 CTLTYPE_INT, "lowportmin", 2404 SYSCTL_DESCR("Lowest privileged ephemeral port number " 2405 "to assign"), 2406 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 2407 CTL_NET, PF_INET, IPPROTO_IP, 2408 IPCTL_LOWPORTMIN, CTL_EOL); 2409 sysctl_createv(clog, 0, NULL, NULL, 2410 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2411 CTLTYPE_INT, "lowportmax", 2412 SYSCTL_DESCR("Highest privileged ephemeral port number " 2413 "to assign"), 2414 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 2415 CTL_NET, PF_INET, IPPROTO_IP, 2416 IPCTL_LOWPORTMAX, CTL_EOL); 2417#endif /* IPNOPRIVPORTS */ 2418 sysctl_createv(clog, 0, NULL, NULL, 2419 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2420 CTLTYPE_INT, "maxfragpackets", 2421 SYSCTL_DESCR("Maximum number of fragments to retain for " 2422 "possible reassembly"), 2423 NULL, 0, &ip_maxfragpackets, 0, 2424 CTL_NET, PF_INET, IPPROTO_IP, 2425 IPCTL_MAXFRAGPACKETS, CTL_EOL); 2426#if NGRE > 0 2427 sysctl_createv(clog, 0, NULL, NULL, 2428 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2429 CTLTYPE_INT, "grettl", 2430 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 2431 NULL, 0, &ip_gre_ttl, 0, 2432 CTL_NET, PF_INET, IPPROTO_IP, 2433 IPCTL_GRE_TTL, CTL_EOL); 2434#endif /* NGRE */ 2435 sysctl_createv(clog, 0, NULL, NULL, 2436 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2437 CTLTYPE_INT, "checkinterface", 2438 SYSCTL_DESCR("Enable receive side of Strong ES model " 2439 "from RFC1122"), 2440 NULL, 0, &ip_checkinterface, 0, 2441 CTL_NET, PF_INET, IPPROTO_IP, 2442 IPCTL_CHECKINTERFACE, CTL_EOL); 2443 sysctl_createv(clog, 0, NULL, NULL, 2444 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2445 CTLTYPE_INT, "random_id", 2446 SYSCTL_DESCR("Assign random ip_id values"), 2447 NULL, 0, &ip_do_randomid, 0, 2448 CTL_NET, PF_INET, IPPROTO_IP, 2449 IPCTL_RANDOMID, CTL_EOL); 2450 sysctl_createv(clog, 0, NULL, NULL, 2451 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2452 CTLTYPE_INT, "do_loopback_cksum", 2453 SYSCTL_DESCR("Perform IP checksum on loopback"), 2454 NULL, 0, &ip_do_loopback_cksum, 0, 2455 CTL_NET, PF_INET, IPPROTO_IP, 2456 IPCTL_LOOPBACKCKSUM, CTL_EOL); 2457 sysctl_createv(clog, 0, NULL, NULL, 2458 CTLFLAG_PERMANENT, 2459 CTLTYPE_STRUCT, "stats", 2460 SYSCTL_DESCR("IP statistics"), 2461 sysctl_net_inet_ip_stats, 0, NULL, 0, 2462 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 2463 CTL_EOL); 2464} 2465 2466void 2467ip_statinc(u_int stat) 2468{ 2469 2470 KASSERT(stat < IP_NSTATS); 2471 IP_STATINC(stat); 2472} 2473