ip_input.c revision 1.193
1/* $NetBSD: ip_input.c,v 1.193 2003/12/12 21:17:59 scw Exp $ */ 2 3/* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32/*- 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 3. All advertising materials mentioning features or use of this software 49 * must display the following acknowledgement: 50 * This product includes software developed by the NetBSD 51 * Foundation, Inc. and its contributors. 52 * 4. Neither the name of The NetBSD Foundation nor the names of its 53 * contributors may be used to endorse or promote products derived 54 * from this software without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 57 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 58 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 59 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 60 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 61 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 62 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 63 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 64 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 65 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 66 * POSSIBILITY OF SUCH DAMAGE. 67 */ 68 69/* 70 * Copyright (c) 1982, 1986, 1988, 1993 71 * The Regents of the University of California. All rights reserved. 72 * 73 * Redistribution and use in source and binary forms, with or without 74 * modification, are permitted provided that the following conditions 75 * are met: 76 * 1. Redistributions of source code must retain the above copyright 77 * notice, this list of conditions and the following disclaimer. 78 * 2. Redistributions in binary form must reproduce the above copyright 79 * notice, this list of conditions and the following disclaimer in the 80 * documentation and/or other materials provided with the distribution. 81 * 3. Neither the name of the University nor the names of its contributors 82 * may be used to endorse or promote products derived from this software 83 * without specific prior written permission. 84 * 85 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 86 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 87 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 88 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 89 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 90 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 91 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 92 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 93 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 94 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 95 * SUCH DAMAGE. 96 * 97 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 98 */ 99 100#include <sys/cdefs.h> 101__KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.193 2003/12/12 21:17:59 scw Exp $"); 102 103#include "opt_inet.h" 104#include "opt_gateway.h" 105#include "opt_pfil_hooks.h" 106#include "opt_ipsec.h" 107#include "opt_mrouting.h" 108#include "opt_mbuftrace.h" 109#include "opt_inet_csum.h" 110 111#include <sys/param.h> 112#include <sys/systm.h> 113#include <sys/malloc.h> 114#include <sys/mbuf.h> 115#include <sys/domain.h> 116#include <sys/protosw.h> 117#include <sys/socket.h> 118#include <sys/socketvar.h> 119#include <sys/errno.h> 120#include <sys/time.h> 121#include <sys/kernel.h> 122#include <sys/pool.h> 123#include <sys/sysctl.h> 124 125#include <net/if.h> 126#include <net/if_dl.h> 127#include <net/route.h> 128#include <net/pfil.h> 129 130#include <netinet/in.h> 131#include <netinet/in_systm.h> 132#include <netinet/ip.h> 133#include <netinet/in_pcb.h> 134#include <netinet/in_var.h> 135#include <netinet/ip_var.h> 136#include <netinet/ip_icmp.h> 137/* just for gif_ttl */ 138#include <netinet/in_gif.h> 139#include "gif.h" 140#include <net/if_gre.h> 141#include "gre.h" 142 143#ifdef MROUTING 144#include <netinet/ip_mroute.h> 145#endif 146 147#ifdef IPSEC 148#include <netinet6/ipsec.h> 149#include <netkey/key.h> 150#endif 151#ifdef FAST_IPSEC 152#include <netipsec/ipsec.h> 153#include <netipsec/key.h> 154#endif /* FAST_IPSEC*/ 155 156#ifndef IPFORWARDING 157#ifdef GATEWAY 158#define IPFORWARDING 1 /* forward IP packets not for us */ 159#else /* GATEWAY */ 160#define IPFORWARDING 0 /* don't forward IP packets not for us */ 161#endif /* GATEWAY */ 162#endif /* IPFORWARDING */ 163#ifndef IPSENDREDIRECTS 164#define IPSENDREDIRECTS 1 165#endif 166#ifndef IPFORWSRCRT 167#define IPFORWSRCRT 1 /* forward source-routed packets */ 168#endif 169#ifndef IPALLOWSRCRT 170#define IPALLOWSRCRT 1 /* allow source-routed packets */ 171#endif 172#ifndef IPMTUDISC 173#define IPMTUDISC 1 174#endif 175#ifndef IPMTUDISCTIMEOUT 176#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 177#endif 178 179/* 180 * Note: DIRECTED_BROADCAST is handled this way so that previous 181 * configuration using this option will Just Work. 182 */ 183#ifndef IPDIRECTEDBCAST 184#ifdef DIRECTED_BROADCAST 185#define IPDIRECTEDBCAST 1 186#else 187#define IPDIRECTEDBCAST 0 188#endif /* DIRECTED_BROADCAST */ 189#endif /* IPDIRECTEDBCAST */ 190int ipforwarding = IPFORWARDING; 191int ipsendredirects = IPSENDREDIRECTS; 192int ip_defttl = IPDEFTTL; 193int ip_forwsrcrt = IPFORWSRCRT; 194int ip_directedbcast = IPDIRECTEDBCAST; 195int ip_allowsrcrt = IPALLOWSRCRT; 196int ip_mtudisc = IPMTUDISC; 197int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 198#ifdef DIAGNOSTIC 199int ipprintfs = 0; 200#endif 201 202int ip_do_randomid = 0; 203 204/* 205 * XXX - Setting ip_checkinterface mostly implements the receive side of 206 * the Strong ES model described in RFC 1122, but since the routing table 207 * and transmit implementation do not implement the Strong ES model, 208 * setting this to 1 results in an odd hybrid. 209 * 210 * XXX - ip_checkinterface currently must be disabled if you use ipnat 211 * to translate the destination address to another local interface. 212 * 213 * XXX - ip_checkinterface must be disabled if you add IP aliases 214 * to the loopback interface instead of the interface where the 215 * packets for those addresses are received. 216 */ 217int ip_checkinterface = 0; 218 219 220struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 221 222extern struct domain inetdomain; 223int ipqmaxlen = IFQ_MAXLEN; 224u_long in_ifaddrhash; /* size of hash table - 1 */ 225int in_ifaddrentries; /* total number of addrs */ 226struct in_ifaddrhead in_ifaddrhead; 227struct in_ifaddrhashhead *in_ifaddrhashtbl; 228u_long in_multihash; /* size of hash table - 1 */ 229int in_multientries; /* total number of addrs */ 230struct in_multihashhead *in_multihashtbl; 231struct ifqueue ipintrq; 232struct ipstat ipstat; 233uint16_t ip_id; 234 235#ifdef PFIL_HOOKS 236struct pfil_head inet_pfil_hook; 237#endif 238 239/* IP datagram reassembly queues (hashed) */ 240#define IPREASS_NHASH_LOG2 6 241#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 242#define IPREASS_HMASK (IPREASS_NHASH - 1) 243#define IPREASS_HASH(x,y) \ 244 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 245struct ipqhead ipq[IPREASS_NHASH]; 246int ipq_locked; 247int ip_nfragpackets = 0; 248int ip_maxfragpackets = 200; 249int ip_nfrags = 0; /* total fragments in reass queues */ 250 251static __inline int ipq_lock_try __P((void)); 252static __inline void ipq_unlock __P((void)); 253 254static __inline int 255ipq_lock_try() 256{ 257 int s; 258 259 /* 260 * Use splvm() -- we're blocking things that would cause 261 * mbuf allocation. 262 */ 263 s = splvm(); 264 if (ipq_locked) { 265 splx(s); 266 return (0); 267 } 268 ipq_locked = 1; 269 splx(s); 270 return (1); 271} 272 273static __inline void 274ipq_unlock() 275{ 276 int s; 277 278 s = splvm(); 279 ipq_locked = 0; 280 splx(s); 281} 282 283#ifdef DIAGNOSTIC 284#define IPQ_LOCK() \ 285do { \ 286 if (ipq_lock_try() == 0) { \ 287 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 288 panic("ipq_lock"); \ 289 } \ 290} while (/*CONSTCOND*/ 0) 291#define IPQ_LOCK_CHECK() \ 292do { \ 293 if (ipq_locked == 0) { \ 294 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 295 panic("ipq lock check"); \ 296 } \ 297} while (/*CONSTCOND*/ 0) 298#else 299#define IPQ_LOCK() (void) ipq_lock_try() 300#define IPQ_LOCK_CHECK() /* nothing */ 301#endif 302 303#define IPQ_UNLOCK() ipq_unlock() 304 305struct pool inmulti_pool; 306struct pool ipqent_pool; 307 308#ifdef INET_CSUM_COUNTERS 309#include <sys/device.h> 310 311struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 312 NULL, "inet", "hwcsum bad"); 313struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 314 NULL, "inet", "hwcsum ok"); 315struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 316 NULL, "inet", "swcsum"); 317 318#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 319 320#else 321 322#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 323 324#endif /* INET_CSUM_COUNTERS */ 325 326/* 327 * We need to save the IP options in case a protocol wants to respond 328 * to an incoming packet over the same route if the packet got here 329 * using IP source routing. This allows connection establishment and 330 * maintenance when the remote end is on a network that is not known 331 * to us. 332 */ 333int ip_nhops = 0; 334static struct ip_srcrt { 335 struct in_addr dst; /* final destination */ 336 char nop; /* one NOP to align */ 337 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 338 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 339} ip_srcrt; 340 341static void save_rte __P((u_char *, struct in_addr)); 342 343#ifdef MBUFTRACE 344struct mowner ip_rx_mowner = { "internet", "rx" }; 345struct mowner ip_tx_mowner = { "internet", "tx" }; 346#endif 347 348/* 349 * IP initialization: fill in IP protocol switch table. 350 * All protocols not implemented in kernel go to raw IP protocol handler. 351 */ 352void 353ip_init() 354{ 355 struct protosw *pr; 356 int i; 357 358 pool_init(&inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl", 359 NULL); 360 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 361 NULL); 362 363 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 364 if (pr == 0) 365 panic("ip_init"); 366 for (i = 0; i < IPPROTO_MAX; i++) 367 ip_protox[i] = pr - inetsw; 368 for (pr = inetdomain.dom_protosw; 369 pr < inetdomain.dom_protoswNPROTOSW; pr++) 370 if (pr->pr_domain->dom_family == PF_INET && 371 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 372 ip_protox[pr->pr_protocol] = pr - inetsw; 373 374 for (i = 0; i < IPREASS_NHASH; i++) 375 LIST_INIT(&ipq[i]); 376 377 ip_id = time.tv_sec & 0xfffff; 378 ipintrq.ifq_maxlen = ipqmaxlen; 379 TAILQ_INIT(&in_ifaddrhead); 380 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 381 M_WAITOK, &in_ifaddrhash); 382 in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IPMADDR, 383 M_WAITOK, &in_multihash); 384 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 385#ifdef GATEWAY 386 ipflow_init(); 387#endif 388 389#ifdef PFIL_HOOKS 390 /* Register our Packet Filter hook. */ 391 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 392 inet_pfil_hook.ph_af = AF_INET; 393 i = pfil_head_register(&inet_pfil_hook); 394 if (i != 0) 395 printf("ip_init: WARNING: unable to register pfil hook, " 396 "error %d\n", i); 397#endif /* PFIL_HOOKS */ 398 399#ifdef INET_CSUM_COUNTERS 400 evcnt_attach_static(&ip_hwcsum_bad); 401 evcnt_attach_static(&ip_hwcsum_ok); 402 evcnt_attach_static(&ip_swcsum); 403#endif /* INET_CSUM_COUNTERS */ 404 405#ifdef MBUFTRACE 406 MOWNER_ATTACH(&ip_tx_mowner); 407 MOWNER_ATTACH(&ip_rx_mowner); 408#endif /* MBUFTRACE */ 409} 410 411struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 412struct route ipforward_rt; 413 414/* 415 * IP software interrupt routine 416 */ 417void 418ipintr() 419{ 420 int s; 421 struct mbuf *m; 422 423 while (1) { 424 s = splnet(); 425 IF_DEQUEUE(&ipintrq, m); 426 splx(s); 427 if (m == 0) 428 return; 429 MCLAIM(m, &ip_rx_mowner); 430 ip_input(m); 431 } 432} 433 434/* 435 * Ip input routine. Checksum and byte swap header. If fragmented 436 * try to reassemble. Process options. Pass to next level. 437 */ 438void 439ip_input(struct mbuf *m) 440{ 441 struct ip *ip = NULL; 442 struct ipq *fp; 443 struct in_ifaddr *ia; 444 struct ifaddr *ifa; 445 struct ipqent *ipqe; 446 int hlen = 0, mff, len; 447 int downmatch; 448 int checkif; 449 int srcrt = 0; 450 u_int hash; 451#ifdef FAST_IPSEC 452 struct m_tag *mtag; 453 struct tdb_ident *tdbi; 454 struct secpolicy *sp; 455 int s, error; 456#endif /* FAST_IPSEC */ 457 458 MCLAIM(m, &ip_rx_mowner); 459#ifdef DIAGNOSTIC 460 if ((m->m_flags & M_PKTHDR) == 0) 461 panic("ipintr no HDR"); 462#endif 463 464 /* 465 * If no IP addresses have been set yet but the interfaces 466 * are receiving, can't do anything with incoming packets yet. 467 */ 468 if (TAILQ_FIRST(&in_ifaddrhead) == 0) 469 goto bad; 470 ipstat.ips_total++; 471 /* 472 * If the IP header is not aligned, slurp it up into a new 473 * mbuf with space for link headers, in the event we forward 474 * it. Otherwise, if it is aligned, make sure the entire 475 * base IP header is in the first mbuf of the chain. 476 */ 477 if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) { 478 if ((m = m_copyup(m, sizeof(struct ip), 479 (max_linkhdr + 3) & ~3)) == NULL) { 480 /* XXXJRT new stat, please */ 481 ipstat.ips_toosmall++; 482 return; 483 } 484 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 485 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 486 ipstat.ips_toosmall++; 487 return; 488 } 489 } 490 ip = mtod(m, struct ip *); 491 if (ip->ip_v != IPVERSION) { 492 ipstat.ips_badvers++; 493 goto bad; 494 } 495 hlen = ip->ip_hl << 2; 496 if (hlen < sizeof(struct ip)) { /* minimum header length */ 497 ipstat.ips_badhlen++; 498 goto bad; 499 } 500 if (hlen > m->m_len) { 501 if ((m = m_pullup(m, hlen)) == 0) { 502 ipstat.ips_badhlen++; 503 return; 504 } 505 ip = mtod(m, struct ip *); 506 } 507 508 /* 509 * RFC1122: packets with a multicast source address are 510 * not allowed. 511 */ 512 if (IN_MULTICAST(ip->ip_src.s_addr)) { 513 ipstat.ips_badaddr++; 514 goto bad; 515 } 516 517 /* 127/8 must not appear on wire - RFC1122 */ 518 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 519 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 520 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 521 ipstat.ips_badaddr++; 522 goto bad; 523 } 524 } 525 526 switch (m->m_pkthdr.csum_flags & 527 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) | 528 M_CSUM_IPv4_BAD)) { 529 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 530 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 531 goto badcsum; 532 533 case M_CSUM_IPv4: 534 /* Checksum was okay. */ 535 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 536 break; 537 538 default: 539 /* Must compute it ourselves. */ 540 INET_CSUM_COUNTER_INCR(&ip_swcsum); 541 if (in_cksum(m, hlen) != 0) 542 goto bad; 543 break; 544 } 545 546 /* Retrieve the packet length. */ 547 len = ntohs(ip->ip_len); 548 549 /* 550 * Check for additional length bogosity 551 */ 552 if (len < hlen) { 553 ipstat.ips_badlen++; 554 goto bad; 555 } 556 557 /* 558 * Check that the amount of data in the buffers 559 * is as at least much as the IP header would have us expect. 560 * Trim mbufs if longer than we expect. 561 * Drop packet if shorter than we expect. 562 */ 563 if (m->m_pkthdr.len < len) { 564 ipstat.ips_tooshort++; 565 goto bad; 566 } 567 if (m->m_pkthdr.len > len) { 568 if (m->m_len == m->m_pkthdr.len) { 569 m->m_len = len; 570 m->m_pkthdr.len = len; 571 } else 572 m_adj(m, len - m->m_pkthdr.len); 573 } 574 575#if defined(IPSEC) 576 /* ipflow (IP fast forwarding) is not compatible with IPsec. */ 577 m->m_flags &= ~M_CANFASTFWD; 578#else 579 /* 580 * Assume that we can create a fast-forward IP flow entry 581 * based on this packet. 582 */ 583 m->m_flags |= M_CANFASTFWD; 584#endif 585 586#ifdef PFIL_HOOKS 587 /* 588 * Run through list of hooks for input packets. If there are any 589 * filters which require that additional packets in the flow are 590 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 591 * Note that filters must _never_ set this flag, as another filter 592 * in the list may have previously cleared it. 593 */ 594 /* 595 * let ipfilter look at packet on the wire, 596 * not the decapsulated packet. 597 */ 598#ifdef IPSEC 599 if (!ipsec_getnhist(m)) 600#elif defined(FAST_IPSEC) 601 if (!ipsec_indone(m)) 602#else 603 if (1) 604#endif 605 { 606 struct in_addr odst; 607 608 odst = ip->ip_dst; 609 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 610 PFIL_IN) != 0) 611 return; 612 if (m == NULL) 613 return; 614 ip = mtod(m, struct ip *); 615 hlen = ip->ip_hl << 2; 616 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 617 } 618#endif /* PFIL_HOOKS */ 619 620#ifdef ALTQ 621 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 622 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 623 /* packet dropped by traffic conditioner */ 624 return; 625 } 626#endif 627 628 /* 629 * Process options and, if not destined for us, 630 * ship it on. ip_dooptions returns 1 when an 631 * error was detected (causing an icmp message 632 * to be sent and the original packet to be freed). 633 */ 634 ip_nhops = 0; /* for source routed packets */ 635 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 636 return; 637 638 /* 639 * Enable a consistency check between the destination address 640 * and the arrival interface for a unicast packet (the RFC 1122 641 * strong ES model) if IP forwarding is disabled and the packet 642 * is not locally generated. 643 * 644 * XXX - Checking also should be disabled if the destination 645 * address is ipnat'ed to a different interface. 646 * 647 * XXX - Checking is incompatible with IP aliases added 648 * to the loopback interface instead of the interface where 649 * the packets are received. 650 * 651 * XXX - We need to add a per ifaddr flag for this so that 652 * we get finer grain control. 653 */ 654 checkif = ip_checkinterface && (ipforwarding == 0) && 655 (m->m_pkthdr.rcvif != NULL) && 656 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0); 657 658 /* 659 * Check our list of addresses, to see if the packet is for us. 660 * 661 * Traditional 4.4BSD did not consult IFF_UP at all. 662 * The behavior here is to treat addresses on !IFF_UP interface 663 * as not mine. 664 */ 665 downmatch = 0; 666 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 667 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 668 if (checkif && ia->ia_ifp != m->m_pkthdr.rcvif) 669 continue; 670 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 671 break; 672 else 673 downmatch++; 674 } 675 } 676 if (ia != NULL) 677 goto ours; 678 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 679 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrlist, ifa_list) { 680 if (ifa->ifa_addr->sa_family != AF_INET) 681 continue; 682 ia = ifatoia(ifa); 683 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 684 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 685 /* 686 * Look for all-0's host part (old broadcast addr), 687 * either for subnet or net. 688 */ 689 ip->ip_dst.s_addr == ia->ia_subnet || 690 ip->ip_dst.s_addr == ia->ia_net) 691 goto ours; 692 /* 693 * An interface with IP address zero accepts 694 * all packets that arrive on that interface. 695 */ 696 if (in_nullhost(ia->ia_addr.sin_addr)) 697 goto ours; 698 } 699 } 700 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 701 struct in_multi *inm; 702#ifdef MROUTING 703 extern struct socket *ip_mrouter; 704 705 if (M_READONLY(m)) { 706 if ((m = m_pullup(m, hlen)) == 0) { 707 ipstat.ips_toosmall++; 708 return; 709 } 710 ip = mtod(m, struct ip *); 711 } 712 713 if (ip_mrouter) { 714 /* 715 * If we are acting as a multicast router, all 716 * incoming multicast packets are passed to the 717 * kernel-level multicast forwarding function. 718 * The packet is returned (relatively) intact; if 719 * ip_mforward() returns a non-zero value, the packet 720 * must be discarded, else it may be accepted below. 721 * 722 * (The IP ident field is put in the same byte order 723 * as expected when ip_mforward() is called from 724 * ip_output().) 725 */ 726 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 727 ipstat.ips_cantforward++; 728 m_freem(m); 729 return; 730 } 731 732 /* 733 * The process-level routing demon needs to receive 734 * all multicast IGMP packets, whether or not this 735 * host belongs to their destination groups. 736 */ 737 if (ip->ip_p == IPPROTO_IGMP) 738 goto ours; 739 ipstat.ips_forward++; 740 } 741#endif 742 /* 743 * See if we belong to the destination multicast group on the 744 * arrival interface. 745 */ 746 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 747 if (inm == NULL) { 748 ipstat.ips_cantforward++; 749 m_freem(m); 750 return; 751 } 752 goto ours; 753 } 754 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 755 in_nullhost(ip->ip_dst)) 756 goto ours; 757 758 /* 759 * Not for us; forward if possible and desirable. 760 */ 761 if (ipforwarding == 0) { 762 ipstat.ips_cantforward++; 763 m_freem(m); 764 } else { 765 /* 766 * If ip_dst matched any of my address on !IFF_UP interface, 767 * and there's no IFF_UP interface that matches ip_dst, 768 * send icmp unreach. Forwarding it will result in in-kernel 769 * forwarding loop till TTL goes to 0. 770 */ 771 if (downmatch) { 772 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 773 ipstat.ips_cantforward++; 774 return; 775 } 776#ifdef IPSEC 777 if (ipsec4_in_reject(m, NULL)) { 778 ipsecstat.in_polvio++; 779 goto bad; 780 } 781#endif 782#ifdef FAST_IPSEC 783 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 784 s = splsoftnet(); 785 if (mtag != NULL) { 786 tdbi = (struct tdb_ident *)(mtag + 1); 787 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 788 } else { 789 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 790 IP_FORWARDING, &error); 791 } 792 if (sp == NULL) { /* NB: can happen if error */ 793 splx(s); 794 /*XXX error stat???*/ 795 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 796 goto bad; 797 } 798 799 /* 800 * Check security policy against packet attributes. 801 */ 802 error = ipsec_in_reject(sp, m); 803 KEY_FREESP(&sp); 804 splx(s); 805 if (error) { 806 ipstat.ips_cantforward++; 807 goto bad; 808 } 809 810 /* 811 * Peek at the outbound SP for this packet to determine if 812 * it's a Fast Forward candidate. 813 */ 814 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 815 if (mtag != NULL) 816 m->m_flags &= ~M_CANFASTFWD; 817 else { 818 s = splsoftnet(); 819 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, 820 (IP_FORWARDING | 821 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 822 &error, NULL); 823 if (sp != NULL) { 824 m->m_flags &= ~M_CANFASTFWD; 825 KEY_FREESP(&sp); 826 } 827 splx(s); 828 } 829#endif /* FAST_IPSEC */ 830 831 ip_forward(m, srcrt); 832 } 833 return; 834 835ours: 836 /* 837 * If offset or IP_MF are set, must reassemble. 838 * Otherwise, nothing need be done. 839 * (We could look in the reassembly queue to see 840 * if the packet was previously fragmented, 841 * but it's not worth the time; just let them time out.) 842 */ 843 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 844 if (M_READONLY(m)) { 845 if ((m = m_pullup(m, hlen)) == NULL) { 846 ipstat.ips_toosmall++; 847 goto bad; 848 } 849 ip = mtod(m, struct ip *); 850 } 851 852 /* 853 * Look for queue of fragments 854 * of this datagram. 855 */ 856 IPQ_LOCK(); 857 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 858 /* XXX LIST_FOREACH(fp, &ipq[hash], ipq_q) */ 859 for (fp = LIST_FIRST(&ipq[hash]); fp != NULL; 860 fp = LIST_NEXT(fp, ipq_q)) { 861 if (ip->ip_id == fp->ipq_id && 862 in_hosteq(ip->ip_src, fp->ipq_src) && 863 in_hosteq(ip->ip_dst, fp->ipq_dst) && 864 ip->ip_p == fp->ipq_p) 865 goto found; 866 867 } 868 fp = 0; 869found: 870 871 /* 872 * Adjust ip_len to not reflect header, 873 * set ipqe_mff if more fragments are expected, 874 * convert offset of this to bytes. 875 */ 876 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 877 mff = (ip->ip_off & htons(IP_MF)) != 0; 878 if (mff) { 879 /* 880 * Make sure that fragments have a data length 881 * that's a non-zero multiple of 8 bytes. 882 */ 883 if (ntohs(ip->ip_len) == 0 || 884 (ntohs(ip->ip_len) & 0x7) != 0) { 885 ipstat.ips_badfrags++; 886 IPQ_UNLOCK(); 887 goto bad; 888 } 889 } 890 ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3); 891 892 /* 893 * If datagram marked as having more fragments 894 * or if this is not the first fragment, 895 * attempt reassembly; if it succeeds, proceed. 896 */ 897 if (mff || ip->ip_off != htons(0)) { 898 ipstat.ips_fragments++; 899 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 900 if (ipqe == NULL) { 901 ipstat.ips_rcvmemdrop++; 902 IPQ_UNLOCK(); 903 goto bad; 904 } 905 ipqe->ipqe_mff = mff; 906 ipqe->ipqe_m = m; 907 ipqe->ipqe_ip = ip; 908 m = ip_reass(ipqe, fp, &ipq[hash]); 909 if (m == 0) { 910 IPQ_UNLOCK(); 911 return; 912 } 913 ipstat.ips_reassembled++; 914 ip = mtod(m, struct ip *); 915 hlen = ip->ip_hl << 2; 916 ip->ip_len = htons(ntohs(ip->ip_len) + hlen); 917 } else 918 if (fp) 919 ip_freef(fp); 920 IPQ_UNLOCK(); 921 } 922 923#if defined(IPSEC) 924 /* 925 * enforce IPsec policy checking if we are seeing last header. 926 * note that we do not visit this with protocols with pcb layer 927 * code - like udp/tcp/raw ip. 928 */ 929 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 930 ipsec4_in_reject(m, NULL)) { 931 ipsecstat.in_polvio++; 932 goto bad; 933 } 934#endif 935#if FAST_IPSEC 936 /* 937 * enforce IPsec policy checking if we are seeing last header. 938 * note that we do not visit this with protocols with pcb layer 939 * code - like udp/tcp/raw ip. 940 */ 941 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 942 /* 943 * Check if the packet has already had IPsec processing 944 * done. If so, then just pass it along. This tag gets 945 * set during AH, ESP, etc. input handling, before the 946 * packet is returned to the ip input queue for delivery. 947 */ 948 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 949 s = splsoftnet(); 950 if (mtag != NULL) { 951 tdbi = (struct tdb_ident *)(mtag + 1); 952 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 953 } else { 954 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 955 IP_FORWARDING, &error); 956 } 957 if (sp != NULL) { 958 /* 959 * Check security policy against packet attributes. 960 */ 961 error = ipsec_in_reject(sp, m); 962 KEY_FREESP(&sp); 963 } else { 964 /* XXX error stat??? */ 965 error = EINVAL; 966DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 967 goto bad; 968 } 969 splx(s); 970 if (error) 971 goto bad; 972 } 973#endif /* FAST_IPSEC */ 974 975 /* 976 * Switch out to protocol's input routine. 977 */ 978#if IFA_STATS 979 if (ia && ip) 980 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 981#endif 982 ipstat.ips_delivered++; 983 { 984 int off = hlen, nh = ip->ip_p; 985 986 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 987 return; 988 } 989bad: 990 m_freem(m); 991 return; 992 993badcsum: 994 ipstat.ips_badsum++; 995 m_freem(m); 996} 997 998/* 999 * Take incoming datagram fragment and try to 1000 * reassemble it into whole datagram. If a chain for 1001 * reassembly of this datagram already exists, then it 1002 * is given as fp; otherwise have to make a chain. 1003 */ 1004struct mbuf * 1005ip_reass(ipqe, fp, ipqhead) 1006 struct ipqent *ipqe; 1007 struct ipq *fp; 1008 struct ipqhead *ipqhead; 1009{ 1010 struct mbuf *m = ipqe->ipqe_m; 1011 struct ipqent *nq, *p, *q; 1012 struct ip *ip; 1013 struct mbuf *t; 1014 int hlen = ipqe->ipqe_ip->ip_hl << 2; 1015 int i, next; 1016 1017 IPQ_LOCK_CHECK(); 1018 1019 /* 1020 * Presence of header sizes in mbufs 1021 * would confuse code below. 1022 */ 1023 m->m_data += hlen; 1024 m->m_len -= hlen; 1025 1026 /* 1027 * We are about to add a fragment; increment frag count. 1028 */ 1029 ip_nfrags++; 1030 1031 /* 1032 * If first fragment to arrive, create a reassembly queue. 1033 */ 1034 if (fp == 0) { 1035 /* 1036 * Enforce upper bound on number of fragmented packets 1037 * for which we attempt reassembly; 1038 * If maxfrag is 0, never accept fragments. 1039 * If maxfrag is -1, accept all fragments without limitation. 1040 */ 1041 if (ip_maxfragpackets < 0) 1042 ; 1043 else if (ip_nfragpackets >= ip_maxfragpackets) 1044 goto dropfrag; 1045 ip_nfragpackets++; 1046 MALLOC(fp, struct ipq *, sizeof (struct ipq), 1047 M_FTABLE, M_NOWAIT); 1048 if (fp == NULL) 1049 goto dropfrag; 1050 LIST_INSERT_HEAD(ipqhead, fp, ipq_q); 1051 fp->ipq_nfrags = 1; 1052 fp->ipq_ttl = IPFRAGTTL; 1053 fp->ipq_p = ipqe->ipqe_ip->ip_p; 1054 fp->ipq_id = ipqe->ipqe_ip->ip_id; 1055 TAILQ_INIT(&fp->ipq_fragq); 1056 fp->ipq_src = ipqe->ipqe_ip->ip_src; 1057 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 1058 p = NULL; 1059 goto insert; 1060 } else { 1061 fp->ipq_nfrags++; 1062 } 1063 1064 /* 1065 * Find a segment which begins after this one does. 1066 */ 1067 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1068 p = q, q = TAILQ_NEXT(q, ipqe_q)) 1069 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off)) 1070 break; 1071 1072 /* 1073 * If there is a preceding segment, it may provide some of 1074 * our data already. If so, drop the data from the incoming 1075 * segment. If it provides all of our data, drop us. 1076 */ 1077 if (p != NULL) { 1078 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) - 1079 ntohs(ipqe->ipqe_ip->ip_off); 1080 if (i > 0) { 1081 if (i >= ntohs(ipqe->ipqe_ip->ip_len)) 1082 goto dropfrag; 1083 m_adj(ipqe->ipqe_m, i); 1084 ipqe->ipqe_ip->ip_off = 1085 htons(ntohs(ipqe->ipqe_ip->ip_off) + i); 1086 ipqe->ipqe_ip->ip_len = 1087 htons(ntohs(ipqe->ipqe_ip->ip_len) - i); 1088 } 1089 } 1090 1091 /* 1092 * While we overlap succeeding segments trim them or, 1093 * if they are completely covered, dequeue them. 1094 */ 1095 for (; q != NULL && 1096 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) > 1097 ntohs(q->ipqe_ip->ip_off); q = nq) { 1098 i = (ntohs(ipqe->ipqe_ip->ip_off) + 1099 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off); 1100 if (i < ntohs(q->ipqe_ip->ip_len)) { 1101 q->ipqe_ip->ip_len = 1102 htons(ntohs(q->ipqe_ip->ip_len) - i); 1103 q->ipqe_ip->ip_off = 1104 htons(ntohs(q->ipqe_ip->ip_off) + i); 1105 m_adj(q->ipqe_m, i); 1106 break; 1107 } 1108 nq = TAILQ_NEXT(q, ipqe_q); 1109 m_freem(q->ipqe_m); 1110 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1111 pool_put(&ipqent_pool, q); 1112 fp->ipq_nfrags--; 1113 ip_nfrags--; 1114 } 1115 1116insert: 1117 /* 1118 * Stick new segment in its place; 1119 * check for complete reassembly. 1120 */ 1121 if (p == NULL) { 1122 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 1123 } else { 1124 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q); 1125 } 1126 next = 0; 1127 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1128 p = q, q = TAILQ_NEXT(q, ipqe_q)) { 1129 if (ntohs(q->ipqe_ip->ip_off) != next) 1130 return (0); 1131 next += ntohs(q->ipqe_ip->ip_len); 1132 } 1133 if (p->ipqe_mff) 1134 return (0); 1135 1136 /* 1137 * Reassembly is complete. Check for a bogus message size and 1138 * concatenate fragments. 1139 */ 1140 q = TAILQ_FIRST(&fp->ipq_fragq); 1141 ip = q->ipqe_ip; 1142 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 1143 ipstat.ips_toolong++; 1144 ip_freef(fp); 1145 return (0); 1146 } 1147 m = q->ipqe_m; 1148 t = m->m_next; 1149 m->m_next = 0; 1150 m_cat(m, t); 1151 nq = TAILQ_NEXT(q, ipqe_q); 1152 pool_put(&ipqent_pool, q); 1153 for (q = nq; q != NULL; q = nq) { 1154 t = q->ipqe_m; 1155 nq = TAILQ_NEXT(q, ipqe_q); 1156 pool_put(&ipqent_pool, q); 1157 m_cat(m, t); 1158 } 1159 ip_nfrags -= fp->ipq_nfrags; 1160 1161 /* 1162 * Create header for new ip packet by 1163 * modifying header of first packet; 1164 * dequeue and discard fragment reassembly header. 1165 * Make header visible. 1166 */ 1167 ip->ip_len = htons(next); 1168 ip->ip_src = fp->ipq_src; 1169 ip->ip_dst = fp->ipq_dst; 1170 LIST_REMOVE(fp, ipq_q); 1171 FREE(fp, M_FTABLE); 1172 ip_nfragpackets--; 1173 m->m_len += (ip->ip_hl << 2); 1174 m->m_data -= (ip->ip_hl << 2); 1175 /* some debugging cruft by sklower, below, will go away soon */ 1176 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1177 int plen = 0; 1178 for (t = m; t; t = t->m_next) 1179 plen += t->m_len; 1180 m->m_pkthdr.len = plen; 1181 } 1182 return (m); 1183 1184dropfrag: 1185 if (fp != 0) 1186 fp->ipq_nfrags--; 1187 ip_nfrags--; 1188 ipstat.ips_fragdropped++; 1189 m_freem(m); 1190 pool_put(&ipqent_pool, ipqe); 1191 return (0); 1192} 1193 1194/* 1195 * Free a fragment reassembly header and all 1196 * associated datagrams. 1197 */ 1198void 1199ip_freef(fp) 1200 struct ipq *fp; 1201{ 1202 struct ipqent *q, *p; 1203 u_int nfrags = 0; 1204 1205 IPQ_LOCK_CHECK(); 1206 1207 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) { 1208 p = TAILQ_NEXT(q, ipqe_q); 1209 m_freem(q->ipqe_m); 1210 nfrags++; 1211 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1212 pool_put(&ipqent_pool, q); 1213 } 1214 1215 if (nfrags != fp->ipq_nfrags) 1216 printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags); 1217 ip_nfrags -= nfrags; 1218 LIST_REMOVE(fp, ipq_q); 1219 FREE(fp, M_FTABLE); 1220 ip_nfragpackets--; 1221} 1222 1223/* 1224 * IP timer processing; 1225 * if a timer expires on a reassembly 1226 * queue, discard it. 1227 */ 1228void 1229ip_slowtimo() 1230{ 1231 static u_int dropscanidx = 0; 1232 u_int i; 1233 struct ipq *fp, *nfp; 1234 int s = splsoftnet(); 1235 1236 IPQ_LOCK(); 1237 for (i = 0; i < IPREASS_NHASH; i++) { 1238 for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) { 1239 nfp = LIST_NEXT(fp, ipq_q); 1240 if (--fp->ipq_ttl == 0) { 1241 ipstat.ips_fragtimeout++; 1242 ip_freef(fp); 1243 } 1244 } 1245 } 1246 /* 1247 * If we are over the maximum number of fragments 1248 * (due to the limit being lowered), drain off 1249 * enough to get down to the new limit. Start draining 1250 * from the reassembly hashqueue most recently drained. 1251 */ 1252 if (ip_maxfragpackets < 0) 1253 ; 1254 else { 1255 int wrapped = 0; 1256 1257 i = dropscanidx; 1258 while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) { 1259 while (LIST_FIRST(&ipq[i]) != NULL) 1260 ip_freef(LIST_FIRST(&ipq[i])); 1261 if (++i >= IPREASS_NHASH) { 1262 i = 0; 1263 } 1264 /* 1265 * Dont scan forever even if fragment counters are 1266 * wrong: stop after scanning entire reassembly queue. 1267 */ 1268 if (i == dropscanidx) 1269 wrapped = 1; 1270 } 1271 dropscanidx = i; 1272 } 1273 IPQ_UNLOCK(); 1274#ifdef GATEWAY 1275 ipflow_slowtimo(); 1276#endif 1277 splx(s); 1278} 1279 1280/* 1281 * Drain off all datagram fragments. 1282 */ 1283void 1284ip_drain() 1285{ 1286 int i; 1287 1288 /* 1289 * We may be called from a device's interrupt context. If 1290 * the ipq is already busy, just bail out now. 1291 */ 1292 if (ipq_lock_try() == 0) 1293 return; 1294 1295 for (i = 0; i < IPREASS_NHASH; i++) { 1296 struct ipqhead *ipqh = &ipq[i]; 1297 struct ipq *fp, *nfp; 1298 for (fp = LIST_FIRST(ipqh); fp != NULL; fp = nfp) { 1299 nfp = LIST_NEXT(fp, ipq_q); 1300 ip_freef(fp); 1301 ipstat.ips_fragdropped++; 1302 } 1303 } 1304 1305 IPQ_UNLOCK(); 1306} 1307 1308/* 1309 * Do option processing on a datagram, 1310 * possibly discarding it if bad options are encountered, 1311 * or forwarding it if source-routed. 1312 * Returns 1 if packet has been forwarded/freed, 1313 * 0 if the packet should be processed further. 1314 */ 1315int 1316ip_dooptions(m) 1317 struct mbuf *m; 1318{ 1319 struct ip *ip = mtod(m, struct ip *); 1320 u_char *cp, *cp0; 1321 struct ip_timestamp *ipt; 1322 struct in_ifaddr *ia; 1323 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1324 struct in_addr dst; 1325 n_time ntime; 1326 1327 dst = ip->ip_dst; 1328 cp = (u_char *)(ip + 1); 1329 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1330 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1331 opt = cp[IPOPT_OPTVAL]; 1332 if (opt == IPOPT_EOL) 1333 break; 1334 if (opt == IPOPT_NOP) 1335 optlen = 1; 1336 else { 1337 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1338 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1339 goto bad; 1340 } 1341 optlen = cp[IPOPT_OLEN]; 1342 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1343 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1344 goto bad; 1345 } 1346 } 1347 switch (opt) { 1348 1349 default: 1350 break; 1351 1352 /* 1353 * Source routing with record. 1354 * Find interface with current destination address. 1355 * If none on this machine then drop if strictly routed, 1356 * or do nothing if loosely routed. 1357 * Record interface address and bring up next address 1358 * component. If strictly routed make sure next 1359 * address is on directly accessible net. 1360 */ 1361 case IPOPT_LSRR: 1362 case IPOPT_SSRR: 1363 if (ip_allowsrcrt == 0) { 1364 type = ICMP_UNREACH; 1365 code = ICMP_UNREACH_NET_PROHIB; 1366 goto bad; 1367 } 1368 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1369 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1370 goto bad; 1371 } 1372 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1373 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1374 goto bad; 1375 } 1376 ipaddr.sin_addr = ip->ip_dst; 1377 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1378 if (ia == 0) { 1379 if (opt == IPOPT_SSRR) { 1380 type = ICMP_UNREACH; 1381 code = ICMP_UNREACH_SRCFAIL; 1382 goto bad; 1383 } 1384 /* 1385 * Loose routing, and not at next destination 1386 * yet; nothing to do except forward. 1387 */ 1388 break; 1389 } 1390 off--; /* 0 origin */ 1391 if ((off + sizeof(struct in_addr)) > optlen) { 1392 /* 1393 * End of source route. Should be for us. 1394 */ 1395 save_rte(cp, ip->ip_src); 1396 break; 1397 } 1398 /* 1399 * locate outgoing interface 1400 */ 1401 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1402 sizeof(ipaddr.sin_addr)); 1403 if (opt == IPOPT_SSRR) 1404 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1405 else 1406 ia = ip_rtaddr(ipaddr.sin_addr); 1407 if (ia == 0) { 1408 type = ICMP_UNREACH; 1409 code = ICMP_UNREACH_SRCFAIL; 1410 goto bad; 1411 } 1412 ip->ip_dst = ipaddr.sin_addr; 1413 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1414 (caddr_t)(cp + off), sizeof(struct in_addr)); 1415 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1416 /* 1417 * Let ip_intr's mcast routing check handle mcast pkts 1418 */ 1419 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1420 break; 1421 1422 case IPOPT_RR: 1423 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1424 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1425 goto bad; 1426 } 1427 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1428 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1429 goto bad; 1430 } 1431 /* 1432 * If no space remains, ignore. 1433 */ 1434 off--; /* 0 origin */ 1435 if ((off + sizeof(struct in_addr)) > optlen) 1436 break; 1437 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1438 sizeof(ipaddr.sin_addr)); 1439 /* 1440 * locate outgoing interface; if we're the destination, 1441 * use the incoming interface (should be same). 1442 */ 1443 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1444 == NULL && 1445 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1446 type = ICMP_UNREACH; 1447 code = ICMP_UNREACH_HOST; 1448 goto bad; 1449 } 1450 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1451 (caddr_t)(cp + off), sizeof(struct in_addr)); 1452 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1453 break; 1454 1455 case IPOPT_TS: 1456 code = cp - (u_char *)ip; 1457 ipt = (struct ip_timestamp *)cp; 1458 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1459 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1460 goto bad; 1461 } 1462 if (ipt->ipt_ptr < 5) { 1463 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1464 goto bad; 1465 } 1466 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1467 if (++ipt->ipt_oflw == 0) { 1468 code = (u_char *)&ipt->ipt_ptr - 1469 (u_char *)ip; 1470 goto bad; 1471 } 1472 break; 1473 } 1474 cp0 = (cp + ipt->ipt_ptr - 1); 1475 switch (ipt->ipt_flg) { 1476 1477 case IPOPT_TS_TSONLY: 1478 break; 1479 1480 case IPOPT_TS_TSANDADDR: 1481 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1482 sizeof(struct in_addr) > ipt->ipt_len) { 1483 code = (u_char *)&ipt->ipt_ptr - 1484 (u_char *)ip; 1485 goto bad; 1486 } 1487 ipaddr.sin_addr = dst; 1488 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1489 m->m_pkthdr.rcvif)); 1490 if (ia == 0) 1491 continue; 1492 bcopy(&ia->ia_addr.sin_addr, 1493 cp0, sizeof(struct in_addr)); 1494 ipt->ipt_ptr += sizeof(struct in_addr); 1495 break; 1496 1497 case IPOPT_TS_PRESPEC: 1498 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1499 sizeof(struct in_addr) > ipt->ipt_len) { 1500 code = (u_char *)&ipt->ipt_ptr - 1501 (u_char *)ip; 1502 goto bad; 1503 } 1504 bcopy(cp0, &ipaddr.sin_addr, 1505 sizeof(struct in_addr)); 1506 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1507 == NULL) 1508 continue; 1509 ipt->ipt_ptr += sizeof(struct in_addr); 1510 break; 1511 1512 default: 1513 /* XXX can't take &ipt->ipt_flg */ 1514 code = (u_char *)&ipt->ipt_ptr - 1515 (u_char *)ip + 1; 1516 goto bad; 1517 } 1518 ntime = iptime(); 1519 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1520 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1521 sizeof(n_time)); 1522 ipt->ipt_ptr += sizeof(n_time); 1523 } 1524 } 1525 if (forward) { 1526 if (ip_forwsrcrt == 0) { 1527 type = ICMP_UNREACH; 1528 code = ICMP_UNREACH_SRCFAIL; 1529 goto bad; 1530 } 1531 ip_forward(m, 1); 1532 return (1); 1533 } 1534 return (0); 1535bad: 1536 icmp_error(m, type, code, 0, 0); 1537 ipstat.ips_badoptions++; 1538 return (1); 1539} 1540 1541/* 1542 * Given address of next destination (final or next hop), 1543 * return internet address info of interface to be used to get there. 1544 */ 1545struct in_ifaddr * 1546ip_rtaddr(dst) 1547 struct in_addr dst; 1548{ 1549 struct sockaddr_in *sin; 1550 1551 sin = satosin(&ipforward_rt.ro_dst); 1552 1553 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1554 if (ipforward_rt.ro_rt) { 1555 RTFREE(ipforward_rt.ro_rt); 1556 ipforward_rt.ro_rt = 0; 1557 } 1558 sin->sin_family = AF_INET; 1559 sin->sin_len = sizeof(*sin); 1560 sin->sin_addr = dst; 1561 1562 rtalloc(&ipforward_rt); 1563 } 1564 if (ipforward_rt.ro_rt == 0) 1565 return ((struct in_ifaddr *)0); 1566 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1567} 1568 1569/* 1570 * Save incoming source route for use in replies, 1571 * to be picked up later by ip_srcroute if the receiver is interested. 1572 */ 1573void 1574save_rte(option, dst) 1575 u_char *option; 1576 struct in_addr dst; 1577{ 1578 unsigned olen; 1579 1580 olen = option[IPOPT_OLEN]; 1581#ifdef DIAGNOSTIC 1582 if (ipprintfs) 1583 printf("save_rte: olen %d\n", olen); 1584#endif /* 0 */ 1585 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1586 return; 1587 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1588 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1589 ip_srcrt.dst = dst; 1590} 1591 1592/* 1593 * Retrieve incoming source route for use in replies, 1594 * in the same form used by setsockopt. 1595 * The first hop is placed before the options, will be removed later. 1596 */ 1597struct mbuf * 1598ip_srcroute() 1599{ 1600 struct in_addr *p, *q; 1601 struct mbuf *m; 1602 1603 if (ip_nhops == 0) 1604 return ((struct mbuf *)0); 1605 m = m_get(M_DONTWAIT, MT_SOOPTS); 1606 if (m == 0) 1607 return ((struct mbuf *)0); 1608 1609 MCLAIM(m, &inetdomain.dom_mowner); 1610#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1611 1612 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1613 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1614 OPTSIZ; 1615#ifdef DIAGNOSTIC 1616 if (ipprintfs) 1617 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1618#endif 1619 1620 /* 1621 * First save first hop for return route 1622 */ 1623 p = &ip_srcrt.route[ip_nhops - 1]; 1624 *(mtod(m, struct in_addr *)) = *p--; 1625#ifdef DIAGNOSTIC 1626 if (ipprintfs) 1627 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1628#endif 1629 1630 /* 1631 * Copy option fields and padding (nop) to mbuf. 1632 */ 1633 ip_srcrt.nop = IPOPT_NOP; 1634 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1635 bcopy((caddr_t)&ip_srcrt.nop, 1636 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1637 q = (struct in_addr *)(mtod(m, caddr_t) + 1638 sizeof(struct in_addr) + OPTSIZ); 1639#undef OPTSIZ 1640 /* 1641 * Record return path as an IP source route, 1642 * reversing the path (pointers are now aligned). 1643 */ 1644 while (p >= ip_srcrt.route) { 1645#ifdef DIAGNOSTIC 1646 if (ipprintfs) 1647 printf(" %x", ntohl(q->s_addr)); 1648#endif 1649 *q++ = *p--; 1650 } 1651 /* 1652 * Last hop goes to final destination. 1653 */ 1654 *q = ip_srcrt.dst; 1655#ifdef DIAGNOSTIC 1656 if (ipprintfs) 1657 printf(" %x\n", ntohl(q->s_addr)); 1658#endif 1659 return (m); 1660} 1661 1662/* 1663 * Strip out IP options, at higher 1664 * level protocol in the kernel. 1665 * Second argument is buffer to which options 1666 * will be moved, and return value is their length. 1667 * XXX should be deleted; last arg currently ignored. 1668 */ 1669void 1670ip_stripoptions(m, mopt) 1671 struct mbuf *m; 1672 struct mbuf *mopt; 1673{ 1674 int i; 1675 struct ip *ip = mtod(m, struct ip *); 1676 caddr_t opts; 1677 int olen; 1678 1679 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1680 opts = (caddr_t)(ip + 1); 1681 i = m->m_len - (sizeof (struct ip) + olen); 1682 bcopy(opts + olen, opts, (unsigned)i); 1683 m->m_len -= olen; 1684 if (m->m_flags & M_PKTHDR) 1685 m->m_pkthdr.len -= olen; 1686 ip->ip_len = htons(ntohs(ip->ip_len) - olen); 1687 ip->ip_hl = sizeof (struct ip) >> 2; 1688} 1689 1690const int inetctlerrmap[PRC_NCMDS] = { 1691 0, 0, 0, 0, 1692 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1693 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1694 EMSGSIZE, EHOSTUNREACH, 0, 0, 1695 0, 0, 0, 0, 1696 ENOPROTOOPT 1697}; 1698 1699/* 1700 * Forward a packet. If some error occurs return the sender 1701 * an icmp packet. Note we can't always generate a meaningful 1702 * icmp message because icmp doesn't have a large enough repertoire 1703 * of codes and types. 1704 * 1705 * If not forwarding, just drop the packet. This could be confusing 1706 * if ipforwarding was zero but some routing protocol was advancing 1707 * us as a gateway to somewhere. However, we must let the routing 1708 * protocol deal with that. 1709 * 1710 * The srcrt parameter indicates whether the packet is being forwarded 1711 * via a source route. 1712 */ 1713void 1714ip_forward(m, srcrt) 1715 struct mbuf *m; 1716 int srcrt; 1717{ 1718 struct ip *ip = mtod(m, struct ip *); 1719 struct sockaddr_in *sin; 1720 struct rtentry *rt; 1721 int error, type = 0, code = 0; 1722 struct mbuf *mcopy; 1723 n_long dest; 1724 struct ifnet *destifp; 1725#if defined(IPSEC) || defined(FAST_IPSEC) 1726 struct ifnet dummyifp; 1727#endif 1728 1729 /* 1730 * We are now in the output path. 1731 */ 1732 MCLAIM(m, &ip_tx_mowner); 1733 1734 /* 1735 * Clear any in-bound checksum flags for this packet. 1736 */ 1737 m->m_pkthdr.csum_flags = 0; 1738 1739 dest = 0; 1740#ifdef DIAGNOSTIC 1741 if (ipprintfs) 1742 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1743 ntohl(ip->ip_src.s_addr), 1744 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1745#endif 1746 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1747 ipstat.ips_cantforward++; 1748 m_freem(m); 1749 return; 1750 } 1751 if (ip->ip_ttl <= IPTTLDEC) { 1752 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1753 return; 1754 } 1755 ip->ip_ttl -= IPTTLDEC; 1756 1757 sin = satosin(&ipforward_rt.ro_dst); 1758 if ((rt = ipforward_rt.ro_rt) == 0 || 1759 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1760 if (ipforward_rt.ro_rt) { 1761 RTFREE(ipforward_rt.ro_rt); 1762 ipforward_rt.ro_rt = 0; 1763 } 1764 sin->sin_family = AF_INET; 1765 sin->sin_len = sizeof(struct sockaddr_in); 1766 sin->sin_addr = ip->ip_dst; 1767 1768 rtalloc(&ipforward_rt); 1769 if (ipforward_rt.ro_rt == 0) { 1770 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1771 return; 1772 } 1773 rt = ipforward_rt.ro_rt; 1774 } 1775 1776 /* 1777 * Save at most 68 bytes of the packet in case 1778 * we need to generate an ICMP message to the src. 1779 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1780 */ 1781 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1782 if (mcopy) 1783 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1784 1785 /* 1786 * If forwarding packet using same interface that it came in on, 1787 * perhaps should send a redirect to sender to shortcut a hop. 1788 * Only send redirect if source is sending directly to us, 1789 * and if packet was not source routed (or has any options). 1790 * Also, don't send redirect if forwarding using a default route 1791 * or a route modified by a redirect. 1792 */ 1793 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1794 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1795 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1796 ipsendredirects && !srcrt) { 1797 if (rt->rt_ifa && 1798 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1799 ifatoia(rt->rt_ifa)->ia_subnet) { 1800 if (rt->rt_flags & RTF_GATEWAY) 1801 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1802 else 1803 dest = ip->ip_dst.s_addr; 1804 /* 1805 * Router requirements says to only send host 1806 * redirects. 1807 */ 1808 type = ICMP_REDIRECT; 1809 code = ICMP_REDIRECT_HOST; 1810#ifdef DIAGNOSTIC 1811 if (ipprintfs) 1812 printf("redirect (%d) to %x\n", code, 1813 (u_int32_t)dest); 1814#endif 1815 } 1816 } 1817 1818 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1819 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1820 (struct ip_moptions *)NULL, (struct socket *)NULL); 1821 1822 if (error) 1823 ipstat.ips_cantforward++; 1824 else { 1825 ipstat.ips_forward++; 1826 if (type) 1827 ipstat.ips_redirectsent++; 1828 else { 1829 if (mcopy) { 1830#ifdef GATEWAY 1831 if (mcopy->m_flags & M_CANFASTFWD) 1832 ipflow_create(&ipforward_rt, mcopy); 1833#endif 1834 m_freem(mcopy); 1835 } 1836 return; 1837 } 1838 } 1839 if (mcopy == NULL) 1840 return; 1841 destifp = NULL; 1842 1843 switch (error) { 1844 1845 case 0: /* forwarded, but need redirect */ 1846 /* type, code set above */ 1847 break; 1848 1849 case ENETUNREACH: /* shouldn't happen, checked above */ 1850 case EHOSTUNREACH: 1851 case ENETDOWN: 1852 case EHOSTDOWN: 1853 default: 1854 type = ICMP_UNREACH; 1855 code = ICMP_UNREACH_HOST; 1856 break; 1857 1858 case EMSGSIZE: 1859 type = ICMP_UNREACH; 1860 code = ICMP_UNREACH_NEEDFRAG; 1861#if !defined(IPSEC) && !defined(FAST_IPSEC) 1862 if (ipforward_rt.ro_rt) 1863 destifp = ipforward_rt.ro_rt->rt_ifp; 1864#else 1865 /* 1866 * If the packet is routed over IPsec tunnel, tell the 1867 * originator the tunnel MTU. 1868 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1869 * XXX quickhack!!! 1870 */ 1871 if (ipforward_rt.ro_rt) { 1872 struct secpolicy *sp; 1873 int ipsecerror; 1874 size_t ipsechdr; 1875 struct route *ro; 1876 1877 sp = ipsec4_getpolicybyaddr(mcopy, 1878 IPSEC_DIR_OUTBOUND, IP_FORWARDING, 1879 &ipsecerror); 1880 1881 if (sp == NULL) 1882 destifp = ipforward_rt.ro_rt->rt_ifp; 1883 else { 1884 /* count IPsec header size */ 1885 ipsechdr = ipsec4_hdrsiz(mcopy, 1886 IPSEC_DIR_OUTBOUND, NULL); 1887 1888 /* 1889 * find the correct route for outer IPv4 1890 * header, compute tunnel MTU. 1891 * 1892 * XXX BUG ALERT 1893 * The "dummyifp" code relies upon the fact 1894 * that icmp_error() touches only ifp->if_mtu. 1895 */ 1896 /*XXX*/ 1897 destifp = NULL; 1898 if (sp->req != NULL 1899 && sp->req->sav != NULL 1900 && sp->req->sav->sah != NULL) { 1901 ro = &sp->req->sav->sah->sa_route; 1902 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1903 dummyifp.if_mtu = 1904 ro->ro_rt->rt_rmx.rmx_mtu ? 1905 ro->ro_rt->rt_rmx.rmx_mtu : 1906 ro->ro_rt->rt_ifp->if_mtu; 1907 dummyifp.if_mtu -= ipsechdr; 1908 destifp = &dummyifp; 1909 } 1910 } 1911 1912#ifdef IPSEC 1913 key_freesp(sp); 1914#else 1915 KEY_FREESP(&sp); 1916#endif 1917 } 1918 } 1919#endif /*IPSEC*/ 1920 ipstat.ips_cantfrag++; 1921 break; 1922 1923 case ENOBUFS: 1924#if 1 1925 /* 1926 * a router should not generate ICMP_SOURCEQUENCH as 1927 * required in RFC1812 Requirements for IP Version 4 Routers. 1928 * source quench could be a big problem under DoS attacks, 1929 * or if the underlying interface is rate-limited. 1930 */ 1931 if (mcopy) 1932 m_freem(mcopy); 1933 return; 1934#else 1935 type = ICMP_SOURCEQUENCH; 1936 code = 0; 1937 break; 1938#endif 1939 } 1940 icmp_error(mcopy, type, code, dest, destifp); 1941} 1942 1943void 1944ip_savecontrol(inp, mp, ip, m) 1945 struct inpcb *inp; 1946 struct mbuf **mp; 1947 struct ip *ip; 1948 struct mbuf *m; 1949{ 1950 1951 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1952 struct timeval tv; 1953 1954 microtime(&tv); 1955 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1956 SCM_TIMESTAMP, SOL_SOCKET); 1957 if (*mp) 1958 mp = &(*mp)->m_next; 1959 } 1960 if (inp->inp_flags & INP_RECVDSTADDR) { 1961 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1962 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1963 if (*mp) 1964 mp = &(*mp)->m_next; 1965 } 1966#ifdef notyet 1967 /* 1968 * XXX 1969 * Moving these out of udp_input() made them even more broken 1970 * than they already were. 1971 * - fenner@parc.xerox.com 1972 */ 1973 /* options were tossed already */ 1974 if (inp->inp_flags & INP_RECVOPTS) { 1975 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1976 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1977 if (*mp) 1978 mp = &(*mp)->m_next; 1979 } 1980 /* ip_srcroute doesn't do what we want here, need to fix */ 1981 if (inp->inp_flags & INP_RECVRETOPTS) { 1982 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1983 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1984 if (*mp) 1985 mp = &(*mp)->m_next; 1986 } 1987#endif 1988 if (inp->inp_flags & INP_RECVIF) { 1989 struct sockaddr_dl sdl; 1990 1991 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1992 sdl.sdl_family = AF_LINK; 1993 sdl.sdl_index = m->m_pkthdr.rcvif ? 1994 m->m_pkthdr.rcvif->if_index : 0; 1995 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1996 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1997 IP_RECVIF, IPPROTO_IP); 1998 if (*mp) 1999 mp = &(*mp)->m_next; 2000 } 2001} 2002 2003/* 2004 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 2005 * range of the new value and tweaks timers if it changes. 2006 */ 2007static int 2008sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 2009{ 2010 int error, tmp; 2011 struct sysctlnode node; 2012 2013 node = *rnode; 2014 tmp = ip_mtudisc_timeout; 2015 node.sysctl_data = &tmp; 2016 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2017 if (error || newp == NULL) 2018 return (error); 2019 if (tmp < 0) 2020 return (EINVAL); 2021 2022 ip_mtudisc_timeout = tmp; 2023 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 2024 2025 return (0); 2026} 2027 2028#ifdef GATEWAY 2029/* 2030 * sysctl helper routine for net.inet.ip.maxflows. apparently if 2031 * maxflows is even looked up, we "reap flows". 2032 */ 2033static int 2034sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS) 2035{ 2036 int s; 2037 2038 s = sysctl_lookup(SYSCTLFN_CALL(rnode)); 2039 if (s) 2040 return (s); 2041 2042 s = splsoftnet(); 2043 ipflow_reap(0); 2044 splx(s); 2045 2046 return (0); 2047} 2048#endif /* GATEWAY */ 2049 2050 2051SYSCTL_SETUP(sysctl_net_inet_ip_setup, "sysctl net.inet.ip subtree setup") 2052{ 2053 extern int subnetsarelocal, hostzeroisbroadcast; 2054 2055 sysctl_createv(SYSCTL_PERMANENT, 2056 CTLTYPE_NODE, "net", NULL, 2057 NULL, 0, NULL, 0, 2058 CTL_NET, CTL_EOL); 2059 sysctl_createv(SYSCTL_PERMANENT, 2060 CTLTYPE_NODE, "inet", NULL, 2061 NULL, 0, NULL, 0, 2062 CTL_NET, PF_INET, CTL_EOL); 2063 sysctl_createv(SYSCTL_PERMANENT, 2064 CTLTYPE_NODE, "ip", NULL, 2065 NULL, 0, NULL, 0, 2066 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 2067 2068 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2069 CTLTYPE_INT, "forwarding", NULL, 2070 NULL, 0, &ipforwarding, 0, 2071 CTL_NET, PF_INET, IPPROTO_IP, 2072 IPCTL_FORWARDING, CTL_EOL); 2073 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2074 CTLTYPE_INT, "redirect", NULL, 2075 NULL, 0, &ipsendredirects, 0, 2076 CTL_NET, PF_INET, IPPROTO_IP, 2077 IPCTL_SENDREDIRECTS, CTL_EOL); 2078 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2079 CTLTYPE_INT, "ttl", NULL, 2080 NULL, 0, &ip_defttl, 0, 2081 CTL_NET, PF_INET, IPPROTO_IP, 2082 IPCTL_DEFTTL, CTL_EOL); 2083#ifdef IPCTL_DEFMTU 2084 sysctl_createv(SYSCTL_PERMANENT /* |SYSCTL_READWRITE? */, 2085 CTLTYPE_INT, "mtu", NULL, 2086 NULL, 0, &ip_mtu, 0, 2087 CTL_NET, PF_INET, IPPROTO_IP, 2088 IPCTL_DEFMTU, CTL_EOL); 2089#endif /* IPCTL_DEFMTU */ 2090 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READONLY1, 2091 CTLTYPE_INT, "forwsrcrt", NULL, 2092 NULL, 0, &ip_forwsrcrt, 0, 2093 CTL_NET, PF_INET, IPPROTO_IP, 2094 IPCTL_FORWSRCRT, CTL_EOL); 2095 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2096 CTLTYPE_INT, "directed-broadcast", NULL, 2097 NULL, 0, &ip_directedbcast, 0, 2098 CTL_NET, PF_INET, IPPROTO_IP, 2099 IPCTL_DIRECTEDBCAST, CTL_EOL); 2100 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2101 CTLTYPE_INT, "allowsrcrt", NULL, 2102 NULL, 0, &ip_allowsrcrt, 0, 2103 CTL_NET, PF_INET, IPPROTO_IP, 2104 IPCTL_ALLOWSRCRT, CTL_EOL); 2105 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2106 CTLTYPE_INT, "subnetsarelocal", NULL, 2107 NULL, 0, &subnetsarelocal, 0, 2108 CTL_NET, PF_INET, IPPROTO_IP, 2109 IPCTL_SUBNETSARELOCAL, CTL_EOL); 2110 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2111 CTLTYPE_INT, "mtudisc", NULL, 2112 NULL, 0, &ip_mtudisc, 0, 2113 CTL_NET, PF_INET, IPPROTO_IP, 2114 IPCTL_MTUDISC, CTL_EOL); 2115 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2116 CTLTYPE_INT, "anonportmin", NULL, 2117 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 2118 CTL_NET, PF_INET, IPPROTO_IP, 2119 IPCTL_ANONPORTMIN, CTL_EOL); 2120 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2121 CTLTYPE_INT, "anonportmax", NULL, 2122 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 2123 CTL_NET, PF_INET, IPPROTO_IP, 2124 IPCTL_ANONPORTMAX, CTL_EOL); 2125 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2126 CTLTYPE_INT, "mtudisctimeout", NULL, 2127 sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0, 2128 CTL_NET, PF_INET, IPPROTO_IP, 2129 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 2130#ifdef GATEWAY 2131 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2132 CTLTYPE_INT, "maxflows", NULL, 2133 sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0, 2134 CTL_NET, PF_INET, IPPROTO_IP, 2135 IPCTL_MAXFLOWS, CTL_EOL); 2136#endif /* GATEWAY */ 2137 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2138 CTLTYPE_INT, "hostzerobroadcast", NULL, 2139 NULL, 0, &hostzeroisbroadcast, 0, 2140 CTL_NET, PF_INET, IPPROTO_IP, 2141 IPCTL_HOSTZEROBROADCAST, CTL_EOL); 2142#if NGIF > 0 2143 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2144 CTLTYPE_INT, "gifttl", NULL, 2145 NULL, 0, &ip_gif_ttl, 0, 2146 CTL_NET, PF_INET, IPPROTO_IP, 2147 IPCTL_GIF_TTL, CTL_EOL); 2148#endif /* NGIF */ 2149#ifndef IPNOPRIVPORTS 2150 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2151 CTLTYPE_INT, "lowportmin", NULL, 2152 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 2153 CTL_NET, PF_INET, IPPROTO_IP, 2154 IPCTL_LOWPORTMIN, CTL_EOL); 2155 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2156 CTLTYPE_INT, "lowportmax", NULL, 2157 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 2158 CTL_NET, PF_INET, IPPROTO_IP, 2159 IPCTL_LOWPORTMAX, CTL_EOL); 2160#endif /* IPNOPRIVPORTS */ 2161 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2162 CTLTYPE_INT, "maxfragpackets", NULL, 2163 NULL, 0, &ip_maxfragpackets, 0, 2164 CTL_NET, PF_INET, IPPROTO_IP, 2165 IPCTL_MAXFRAGPACKETS, CTL_EOL); 2166#if NGRE > 0 2167 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2168 CTLTYPE_INT, "grettl", NULL, 2169 NULL, 0, &ip_gre_ttl, 0, 2170 CTL_NET, PF_INET, IPPROTO_IP, 2171 IPCTL_GRE_TTL, CTL_EOL); 2172#endif /* NGRE */ 2173 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2174 CTLTYPE_INT, "checkinterface", NULL, 2175 NULL, 0, &ip_checkinterface, 0, 2176 CTL_NET, PF_INET, IPPROTO_IP, 2177 IPCTL_CHECKINTERFACE, CTL_EOL); 2178 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2179 CTLTYPE_INT, "random_id", NULL, 2180 NULL, 0, &ip_do_randomid, 0, 2181 CTL_NET, PF_INET, IPPROTO_IP, 2182 IPCTL_RANDOMID, CTL_EOL); 2183} 2184