ip_input.c revision 1.255
1/* $NetBSD: ip_input.c,v 1.255 2007/11/09 06:59:33 kefren 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 1.255 2007/11/09 06:59:33 kefren Exp $"); 102 103#include "opt_inet.h" 104#include "opt_gateway.h" 105#include "opt_pfil_hooks.h" 106#include "opt_ipsec.h" 107#include "opt_mrouting.h" 108#include "opt_mbuftrace.h" 109#include "opt_inet_csum.h" 110 111#include <sys/param.h> 112#include <sys/systm.h> 113#include <sys/malloc.h> 114#include <sys/mbuf.h> 115#include <sys/domain.h> 116#include <sys/protosw.h> 117#include <sys/socket.h> 118#include <sys/socketvar.h> 119#include <sys/errno.h> 120#include <sys/time.h> 121#include <sys/kernel.h> 122#include <sys/pool.h> 123#include <sys/sysctl.h> 124#include <sys/kauth.h> 125 126#include <net/if.h> 127#include <net/if_dl.h> 128#include <net/route.h> 129#include <net/pfil.h> 130 131#include <netinet/in.h> 132#include <netinet/in_systm.h> 133#include <netinet/ip.h> 134#include <netinet/in_pcb.h> 135#include <netinet/in_proto.h> 136#include <netinet/in_var.h> 137#include <netinet/ip_var.h> 138#include <netinet/ip_icmp.h> 139/* just for gif_ttl */ 140#include <netinet/in_gif.h> 141#include "gif.h" 142#include <net/if_gre.h> 143#include "gre.h" 144 145#ifdef MROUTING 146#include <netinet/ip_mroute.h> 147#endif 148 149#ifdef IPSEC 150#include <netinet6/ipsec.h> 151#include <netkey/key.h> 152#endif 153#ifdef FAST_IPSEC 154#include <netipsec/ipsec.h> 155#include <netipsec/key.h> 156#endif /* FAST_IPSEC*/ 157 158#ifndef IPFORWARDING 159#ifdef GATEWAY 160#define IPFORWARDING 1 /* forward IP packets not for us */ 161#else /* GATEWAY */ 162#define IPFORWARDING 0 /* don't forward IP packets not for us */ 163#endif /* GATEWAY */ 164#endif /* IPFORWARDING */ 165#ifndef IPSENDREDIRECTS 166#define IPSENDREDIRECTS 1 167#endif 168#ifndef IPFORWSRCRT 169#define IPFORWSRCRT 1 /* forward source-routed packets */ 170#endif 171#ifndef IPALLOWSRCRT 172#define IPALLOWSRCRT 1 /* allow source-routed packets */ 173#endif 174#ifndef IPMTUDISC 175#define IPMTUDISC 1 176#endif 177#ifndef IPMTUDISCTIMEOUT 178#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 179#endif 180 181/* 182 * Note: DIRECTED_BROADCAST is handled this way so that previous 183 * configuration using this option will Just Work. 184 */ 185#ifndef IPDIRECTEDBCAST 186#ifdef DIRECTED_BROADCAST 187#define IPDIRECTEDBCAST 1 188#else 189#define IPDIRECTEDBCAST 0 190#endif /* DIRECTED_BROADCAST */ 191#endif /* IPDIRECTEDBCAST */ 192int ipforwarding = IPFORWARDING; 193int ipsendredirects = IPSENDREDIRECTS; 194int ip_defttl = IPDEFTTL; 195int ip_forwsrcrt = IPFORWSRCRT; 196int ip_directedbcast = IPDIRECTEDBCAST; 197int ip_allowsrcrt = IPALLOWSRCRT; 198int ip_mtudisc = IPMTUDISC; 199int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 200#ifdef DIAGNOSTIC 201int ipprintfs = 0; 202#endif 203 204int ip_do_randomid = 0; 205 206/* 207 * XXX - Setting ip_checkinterface mostly implements the receive side of 208 * the Strong ES model described in RFC 1122, but since the routing table 209 * and transmit implementation do not implement the Strong ES model, 210 * setting this to 1 results in an odd hybrid. 211 * 212 * XXX - ip_checkinterface currently must be disabled if you use ipnat 213 * to translate the destination address to another local interface. 214 * 215 * XXX - ip_checkinterface must be disabled if you add IP aliases 216 * to the loopback interface instead of the interface where the 217 * packets for those addresses are received. 218 */ 219int ip_checkinterface = 0; 220 221 222struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 223 224int ipqmaxlen = IFQ_MAXLEN; 225u_long in_ifaddrhash; /* size of hash table - 1 */ 226int in_ifaddrentries; /* total number of addrs */ 227struct in_ifaddrhead in_ifaddrhead; 228struct in_ifaddrhashhead *in_ifaddrhashtbl; 229u_long in_multihash; /* size of hash table - 1 */ 230int in_multientries; /* total number of addrs */ 231struct in_multihashhead *in_multihashtbl; 232struct ifqueue ipintrq; 233struct ipstat ipstat; 234uint16_t ip_id; 235 236#ifdef PFIL_HOOKS 237struct pfil_head inet_pfil_hook; 238#endif 239 240/* 241 * Cached copy of nmbclusters. If nbclusters is different, 242 * recalculate IP parameters derived from nmbclusters. 243 */ 244static int ip_nmbclusters; /* copy of nmbclusters */ 245static void ip_nmbclusters_changed(void); /* recalc limits */ 246 247#define CHECK_NMBCLUSTER_PARAMS() \ 248do { \ 249 if (__predict_false(ip_nmbclusters != nmbclusters)) \ 250 ip_nmbclusters_changed(); \ 251} while (/*CONSTCOND*/0) 252 253/* IP datagram reassembly queues (hashed) */ 254#define IPREASS_NHASH_LOG2 6 255#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 256#define IPREASS_HMASK (IPREASS_NHASH - 1) 257#define IPREASS_HASH(x,y) \ 258 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 259struct ipqhead ipq[IPREASS_NHASH]; 260int ipq_locked; 261static int ip_nfragpackets; /* packets in reass queue */ 262static int ip_nfrags; /* total fragments in reass queues */ 263 264int ip_maxfragpackets = 200; /* limit on packets. XXX sysctl */ 265int ip_maxfrags; /* limit on fragments. XXX sysctl */ 266 267 268/* 269 * Additive-Increase/Multiplicative-Decrease (AIMD) strategy for 270 * IP reassembly queue buffer managment. 271 * 272 * We keep a count of total IP fragments (NB: not fragmented packets!) 273 * awaiting reassembly (ip_nfrags) and a limit (ip_maxfrags) on fragments. 274 * If ip_nfrags exceeds ip_maxfrags the limit, we drop half the 275 * total fragments in reassembly queues.This AIMD policy avoids 276 * repeatedly deleting single packets under heavy fragmentation load 277 * (e.g., from lossy NFS peers). 278 */ 279static u_int ip_reass_ttl_decr(u_int ticks); 280static void ip_reass_drophalf(void); 281 282 283static inline int ipq_lock_try(void); 284static inline void ipq_unlock(void); 285 286static inline int 287ipq_lock_try(void) 288{ 289 int s; 290 291 /* 292 * Use splvm() -- we're blocking things that would cause 293 * mbuf allocation. 294 */ 295 s = splvm(); 296 if (ipq_locked) { 297 splx(s); 298 return (0); 299 } 300 ipq_locked = 1; 301 splx(s); 302 return (1); 303} 304 305static inline void 306ipq_unlock(void) 307{ 308 int s; 309 310 s = splvm(); 311 ipq_locked = 0; 312 splx(s); 313} 314 315#ifdef DIAGNOSTIC 316#define IPQ_LOCK() \ 317do { \ 318 if (ipq_lock_try() == 0) { \ 319 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 320 panic("ipq_lock"); \ 321 } \ 322} while (/*CONSTCOND*/ 0) 323#define IPQ_LOCK_CHECK() \ 324do { \ 325 if (ipq_locked == 0) { \ 326 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 327 panic("ipq lock check"); \ 328 } \ 329} while (/*CONSTCOND*/ 0) 330#else 331#define IPQ_LOCK() (void) ipq_lock_try() 332#define IPQ_LOCK_CHECK() /* nothing */ 333#endif 334 335#define IPQ_UNLOCK() ipq_unlock() 336 337POOL_INIT(inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl", NULL, 338 IPL_SOFTNET); 339POOL_INIT(ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", NULL, 340 IPL_VM); 341 342#ifdef INET_CSUM_COUNTERS 343#include <sys/device.h> 344 345struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 346 NULL, "inet", "hwcsum bad"); 347struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 348 NULL, "inet", "hwcsum ok"); 349struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 350 NULL, "inet", "swcsum"); 351 352#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 353 354EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 355EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 356EVCNT_ATTACH_STATIC(ip_swcsum); 357 358#else 359 360#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 361 362#endif /* INET_CSUM_COUNTERS */ 363 364/* 365 * We need to save the IP options in case a protocol wants to respond 366 * to an incoming packet over the same route if the packet got here 367 * using IP source routing. This allows connection establishment and 368 * maintenance when the remote end is on a network that is not known 369 * to us. 370 */ 371int ip_nhops = 0; 372static struct ip_srcrt { 373 struct in_addr dst; /* final destination */ 374 char nop; /* one NOP to align */ 375 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 376 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 377} ip_srcrt; 378 379static void save_rte(u_char *, struct in_addr); 380 381#ifdef MBUFTRACE 382struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx"); 383struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx"); 384#endif 385 386/* 387 * Compute IP limits derived from the value of nmbclusters. 388 */ 389static void 390ip_nmbclusters_changed(void) 391{ 392 ip_maxfrags = nmbclusters / 4; 393 ip_nmbclusters = nmbclusters; 394} 395 396/* 397 * IP initialization: fill in IP protocol switch table. 398 * All protocols not implemented in kernel go to raw IP protocol handler. 399 */ 400void 401ip_init(void) 402{ 403 const struct protosw *pr; 404 int i; 405 406 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 407 if (pr == 0) 408 panic("ip_init"); 409 for (i = 0; i < IPPROTO_MAX; i++) 410 ip_protox[i] = pr - inetsw; 411 for (pr = inetdomain.dom_protosw; 412 pr < inetdomain.dom_protoswNPROTOSW; pr++) 413 if (pr->pr_domain->dom_family == PF_INET && 414 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 415 ip_protox[pr->pr_protocol] = pr - inetsw; 416 417 for (i = 0; i < IPREASS_NHASH; i++) 418 LIST_INIT(&ipq[i]); 419 420 ip_id = time_second & 0xfffff; 421 422 ipintrq.ifq_maxlen = ipqmaxlen; 423 ip_nmbclusters_changed(); 424 425 TAILQ_INIT(&in_ifaddrhead); 426 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 427 M_WAITOK, &in_ifaddrhash); 428 in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IPMADDR, 429 M_WAITOK, &in_multihash); 430 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 431#ifdef GATEWAY 432 ipflow_init(ip_hashsize); 433#endif 434 435#ifdef PFIL_HOOKS 436 /* Register our Packet Filter hook. */ 437 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 438 inet_pfil_hook.ph_af = AF_INET; 439 i = pfil_head_register(&inet_pfil_hook); 440 if (i != 0) 441 printf("ip_init: WARNING: unable to register pfil hook, " 442 "error %d\n", i); 443#endif /* PFIL_HOOKS */ 444 445#ifdef MBUFTRACE 446 MOWNER_ATTACH(&ip_tx_mowner); 447 MOWNER_ATTACH(&ip_rx_mowner); 448#endif /* MBUFTRACE */ 449} 450 451struct sockaddr_in ipaddr = { 452 .sin_len = sizeof(ipaddr), 453 .sin_family = AF_INET, 454}; 455struct route ipforward_rt; 456 457/* 458 * IP software interrupt routine 459 */ 460void 461ipintr(void) 462{ 463 int s; 464 struct mbuf *m; 465 466 while (!IF_IS_EMPTY(&ipintrq)) { 467 s = splnet(); 468 IF_DEQUEUE(&ipintrq, m); 469 splx(s); 470 if (m == 0) 471 return; 472 ip_input(m); 473 } 474} 475 476/* 477 * Ip input routine. Checksum and byte swap header. If fragmented 478 * try to reassemble. Process options. Pass to next level. 479 */ 480void 481ip_input(struct mbuf *m) 482{ 483 struct ip *ip = NULL; 484 struct ipq *fp; 485 struct in_ifaddr *ia; 486 struct ifaddr *ifa; 487 struct ipqent *ipqe; 488 int hlen = 0, mff, len; 489 int downmatch; 490 int checkif; 491 int srcrt = 0; 492 int s; 493 u_int hash; 494#ifdef FAST_IPSEC 495 struct m_tag *mtag; 496 struct tdb_ident *tdbi; 497 struct secpolicy *sp; 498 int error; 499#endif /* FAST_IPSEC */ 500 501 MCLAIM(m, &ip_rx_mowner); 502#ifdef DIAGNOSTIC 503 if ((m->m_flags & M_PKTHDR) == 0) 504 panic("ipintr no HDR"); 505#endif 506 507 /* 508 * If no IP addresses have been set yet but the interfaces 509 * are receiving, can't do anything with incoming packets yet. 510 */ 511 if (TAILQ_FIRST(&in_ifaddrhead) == 0) 512 goto bad; 513 ipstat.ips_total++; 514 /* 515 * If the IP header is not aligned, slurp it up into a new 516 * mbuf with space for link headers, in the event we forward 517 * it. Otherwise, if it is aligned, make sure the entire 518 * base IP header is in the first mbuf of the chain. 519 */ 520 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 521 if ((m = m_copyup(m, sizeof(struct ip), 522 (max_linkhdr + 3) & ~3)) == NULL) { 523 /* XXXJRT new stat, please */ 524 ipstat.ips_toosmall++; 525 return; 526 } 527 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 528 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 529 ipstat.ips_toosmall++; 530 return; 531 } 532 } 533 ip = mtod(m, struct ip *); 534 if (ip->ip_v != IPVERSION) { 535 ipstat.ips_badvers++; 536 goto bad; 537 } 538 hlen = ip->ip_hl << 2; 539 if (hlen < sizeof(struct ip)) { /* minimum header length */ 540 ipstat.ips_badhlen++; 541 goto bad; 542 } 543 if (hlen > m->m_len) { 544 if ((m = m_pullup(m, hlen)) == 0) { 545 ipstat.ips_badhlen++; 546 return; 547 } 548 ip = mtod(m, struct ip *); 549 } 550 551 /* 552 * RFC1122: packets with a multicast source address are 553 * not allowed. 554 */ 555 if (IN_MULTICAST(ip->ip_src.s_addr)) { 556 ipstat.ips_badaddr++; 557 goto bad; 558 } 559 560 /* 127/8 must not appear on wire - RFC1122 */ 561 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 562 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 563 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 564 ipstat.ips_badaddr++; 565 goto bad; 566 } 567 } 568 569 switch (m->m_pkthdr.csum_flags & 570 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) | 571 M_CSUM_IPv4_BAD)) { 572 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 573 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 574 goto badcsum; 575 576 case M_CSUM_IPv4: 577 /* Checksum was okay. */ 578 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 579 break; 580 581 default: 582 /* 583 * Must compute it ourselves. Maybe skip checksum on 584 * loopback interfaces. 585 */ 586 if (__predict_true(!(m->m_pkthdr.rcvif->if_flags & 587 IFF_LOOPBACK) || ip_do_loopback_cksum)) { 588 INET_CSUM_COUNTER_INCR(&ip_swcsum); 589 if (in_cksum(m, hlen) != 0) 590 goto badcsum; 591 } 592 break; 593 } 594 595 /* Retrieve the packet length. */ 596 len = ntohs(ip->ip_len); 597 598 /* 599 * Check for additional length bogosity 600 */ 601 if (len < hlen) { 602 ipstat.ips_badlen++; 603 goto bad; 604 } 605 606 /* 607 * Check that the amount of data in the buffers 608 * is as at least much as the IP header would have us expect. 609 * Trim mbufs if longer than we expect. 610 * Drop packet if shorter than we expect. 611 */ 612 if (m->m_pkthdr.len < len) { 613 ipstat.ips_tooshort++; 614 goto bad; 615 } 616 if (m->m_pkthdr.len > len) { 617 if (m->m_len == m->m_pkthdr.len) { 618 m->m_len = len; 619 m->m_pkthdr.len = len; 620 } else 621 m_adj(m, len - m->m_pkthdr.len); 622 } 623 624#if defined(IPSEC) 625 /* ipflow (IP fast forwarding) is not compatible with IPsec. */ 626 m->m_flags &= ~M_CANFASTFWD; 627#else 628 /* 629 * Assume that we can create a fast-forward IP flow entry 630 * based on this packet. 631 */ 632 m->m_flags |= M_CANFASTFWD; 633#endif 634 635#ifdef PFIL_HOOKS 636 /* 637 * Run through list of hooks for input packets. If there are any 638 * filters which require that additional packets in the flow are 639 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 640 * Note that filters must _never_ set this flag, as another filter 641 * in the list may have previously cleared it. 642 */ 643 /* 644 * let ipfilter look at packet on the wire, 645 * not the decapsulated packet. 646 */ 647#ifdef IPSEC 648 if (!ipsec_getnhist(m)) 649#elif defined(FAST_IPSEC) 650 if (!ipsec_indone(m)) 651#else 652 if (1) 653#endif 654 { 655 struct in_addr odst; 656 657 odst = ip->ip_dst; 658 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 659 PFIL_IN) != 0) 660 return; 661 if (m == NULL) 662 return; 663 ip = mtod(m, struct ip *); 664 hlen = ip->ip_hl << 2; 665 /* 666 * XXX The setting of "srcrt" here is to prevent ip_forward() 667 * from generating ICMP redirects for packets that have 668 * been redirected by a hook back out on to the same LAN that 669 * they came from and is not an indication that the packet 670 * is being inffluenced by source routing options. This 671 * allows things like 672 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 673 * where tlp0 is both on the 1.1.1.0/24 network and is the 674 * default route for hosts on 1.1.1.0/24. Of course this 675 * also requires a "map tlp0 ..." to complete the story. 676 * One might argue whether or not this kind of network config. 677 * should be supported in this manner... 678 */ 679 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 680 } 681#endif /* PFIL_HOOKS */ 682 683#ifdef ALTQ 684 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 685 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 686 /* packet dropped by traffic conditioner */ 687 return; 688 } 689#endif 690 691 /* 692 * Process options and, if not destined for us, 693 * ship it on. ip_dooptions returns 1 when an 694 * error was detected (causing an icmp message 695 * to be sent and the original packet to be freed). 696 */ 697 ip_nhops = 0; /* for source routed packets */ 698 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 699 return; 700 701 /* 702 * Enable a consistency check between the destination address 703 * and the arrival interface for a unicast packet (the RFC 1122 704 * strong ES model) if IP forwarding is disabled and the packet 705 * is not locally generated. 706 * 707 * XXX - Checking also should be disabled if the destination 708 * address is ipnat'ed to a different interface. 709 * 710 * XXX - Checking is incompatible with IP aliases added 711 * to the loopback interface instead of the interface where 712 * the packets are received. 713 * 714 * XXX - We need to add a per ifaddr flag for this so that 715 * we get finer grain control. 716 */ 717 checkif = ip_checkinterface && (ipforwarding == 0) && 718 (m->m_pkthdr.rcvif != NULL) && 719 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0); 720 721 /* 722 * Check our list of addresses, to see if the packet is for us. 723 * 724 * Traditional 4.4BSD did not consult IFF_UP at all. 725 * The behavior here is to treat addresses on !IFF_UP interface 726 * as not mine. 727 */ 728 downmatch = 0; 729 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 730 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 731 if (checkif && ia->ia_ifp != m->m_pkthdr.rcvif) 732 continue; 733 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 734 break; 735 else 736 downmatch++; 737 } 738 } 739 if (ia != NULL) 740 goto ours; 741 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 742 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) { 743 if (ifa->ifa_addr->sa_family != AF_INET) 744 continue; 745 ia = ifatoia(ifa); 746 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 747 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 748 /* 749 * Look for all-0's host part (old broadcast addr), 750 * either for subnet or net. 751 */ 752 ip->ip_dst.s_addr == ia->ia_subnet || 753 ip->ip_dst.s_addr == ia->ia_net) 754 goto ours; 755 /* 756 * An interface with IP address zero accepts 757 * all packets that arrive on that interface. 758 */ 759 if (in_nullhost(ia->ia_addr.sin_addr)) 760 goto ours; 761 } 762 } 763 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 764 struct in_multi *inm; 765#ifdef MROUTING 766 extern struct socket *ip_mrouter; 767 768 if (ip_mrouter) { 769 /* 770 * If we are acting as a multicast router, all 771 * incoming multicast packets are passed to the 772 * kernel-level multicast forwarding function. 773 * The packet is returned (relatively) intact; if 774 * ip_mforward() returns a non-zero value, the packet 775 * must be discarded, else it may be accepted below. 776 * 777 * (The IP ident field is put in the same byte order 778 * as expected when ip_mforward() is called from 779 * ip_output().) 780 */ 781 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 782 ipstat.ips_cantforward++; 783 m_freem(m); 784 return; 785 } 786 787 /* 788 * The process-level routing demon needs to receive 789 * all multicast IGMP packets, whether or not this 790 * host belongs to their destination groups. 791 */ 792 if (ip->ip_p == IPPROTO_IGMP) 793 goto ours; 794 ipstat.ips_forward++; 795 } 796#endif 797 /* 798 * See if we belong to the destination multicast group on the 799 * arrival interface. 800 */ 801 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 802 if (inm == NULL) { 803 ipstat.ips_cantforward++; 804 m_freem(m); 805 return; 806 } 807 goto ours; 808 } 809 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 810 in_nullhost(ip->ip_dst)) 811 goto ours; 812 813 /* 814 * Not for us; forward if possible and desirable. 815 */ 816 if (ipforwarding == 0) { 817 ipstat.ips_cantforward++; 818 m_freem(m); 819 } else { 820 /* 821 * If ip_dst matched any of my address on !IFF_UP interface, 822 * and there's no IFF_UP interface that matches ip_dst, 823 * send icmp unreach. Forwarding it will result in in-kernel 824 * forwarding loop till TTL goes to 0. 825 */ 826 if (downmatch) { 827 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 828 ipstat.ips_cantforward++; 829 return; 830 } 831#ifdef IPSEC 832 if (ipsec4_in_reject(m, NULL)) { 833 ipsecstat.in_polvio++; 834 goto bad; 835 } 836#endif 837#ifdef FAST_IPSEC 838 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 839 s = splsoftnet(); 840 if (mtag != NULL) { 841 tdbi = (struct tdb_ident *)(mtag + 1); 842 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 843 } else { 844 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 845 IP_FORWARDING, &error); 846 } 847 if (sp == NULL) { /* NB: can happen if error */ 848 splx(s); 849 /*XXX error stat???*/ 850 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 851 goto bad; 852 } 853 854 /* 855 * Check security policy against packet attributes. 856 */ 857 error = ipsec_in_reject(sp, m); 858 KEY_FREESP(&sp); 859 splx(s); 860 if (error) { 861 ipstat.ips_cantforward++; 862 goto bad; 863 } 864 865 /* 866 * Peek at the outbound SP for this packet to determine if 867 * it's a Fast Forward candidate. 868 */ 869 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 870 if (mtag != NULL) 871 m->m_flags &= ~M_CANFASTFWD; 872 else { 873 s = splsoftnet(); 874 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, 875 (IP_FORWARDING | 876 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 877 &error, NULL); 878 if (sp != NULL) { 879 m->m_flags &= ~M_CANFASTFWD; 880 KEY_FREESP(&sp); 881 } 882 splx(s); 883 } 884#endif /* FAST_IPSEC */ 885 886 ip_forward(m, srcrt); 887 } 888 return; 889 890ours: 891 /* 892 * If offset or IP_MF are set, must reassemble. 893 * Otherwise, nothing need be done. 894 * (We could look in the reassembly queue to see 895 * if the packet was previously fragmented, 896 * but it's not worth the time; just let them time out.) 897 */ 898 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 899 900 /* 901 * Look for queue of fragments 902 * of this datagram. 903 */ 904 IPQ_LOCK(); 905 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 906 LIST_FOREACH(fp, &ipq[hash], ipq_q) { 907 if (ip->ip_id == fp->ipq_id && 908 in_hosteq(ip->ip_src, fp->ipq_src) && 909 in_hosteq(ip->ip_dst, fp->ipq_dst) && 910 ip->ip_p == fp->ipq_p) 911 goto found; 912 913 } 914 fp = 0; 915found: 916 917 /* 918 * Adjust ip_len to not reflect header, 919 * set ipqe_mff if more fragments are expected, 920 * convert offset of this to bytes. 921 */ 922 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 923 mff = (ip->ip_off & htons(IP_MF)) != 0; 924 if (mff) { 925 /* 926 * Make sure that fragments have a data length 927 * that's a non-zero multiple of 8 bytes. 928 */ 929 if (ntohs(ip->ip_len) == 0 || 930 (ntohs(ip->ip_len) & 0x7) != 0) { 931 ipstat.ips_badfrags++; 932 IPQ_UNLOCK(); 933 goto bad; 934 } 935 } 936 ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3); 937 938 /* 939 * If datagram marked as having more fragments 940 * or if this is not the first fragment, 941 * attempt reassembly; if it succeeds, proceed. 942 */ 943 if (mff || ip->ip_off != htons(0)) { 944 ipstat.ips_fragments++; 945 s = splvm(); 946 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 947 splx(s); 948 if (ipqe == NULL) { 949 ipstat.ips_rcvmemdrop++; 950 IPQ_UNLOCK(); 951 goto bad; 952 } 953 ipqe->ipqe_mff = mff; 954 ipqe->ipqe_m = m; 955 ipqe->ipqe_ip = ip; 956 m = ip_reass(ipqe, fp, &ipq[hash]); 957 if (m == 0) { 958 IPQ_UNLOCK(); 959 return; 960 } 961 ipstat.ips_reassembled++; 962 ip = mtod(m, struct ip *); 963 hlen = ip->ip_hl << 2; 964 ip->ip_len = htons(ntohs(ip->ip_len) + hlen); 965 } else 966 if (fp) 967 ip_freef(fp); 968 IPQ_UNLOCK(); 969 } 970 971#if defined(IPSEC) 972 /* 973 * enforce IPsec policy checking if we are seeing last header. 974 * note that we do not visit this with protocols with pcb layer 975 * code - like udp/tcp/raw ip. 976 */ 977 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 978 ipsec4_in_reject(m, NULL)) { 979 ipsecstat.in_polvio++; 980 goto bad; 981 } 982#endif 983#ifdef FAST_IPSEC 984 /* 985 * enforce IPsec policy checking if we are seeing last header. 986 * note that we do not visit this with protocols with pcb layer 987 * code - like udp/tcp/raw ip. 988 */ 989 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 990 /* 991 * Check if the packet has already had IPsec processing 992 * done. If so, then just pass it along. This tag gets 993 * set during AH, ESP, etc. input handling, before the 994 * packet is returned to the ip input queue for delivery. 995 */ 996 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 997 s = splsoftnet(); 998 if (mtag != NULL) { 999 tdbi = (struct tdb_ident *)(mtag + 1); 1000 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 1001 } else { 1002 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 1003 IP_FORWARDING, &error); 1004 } 1005 if (sp != NULL) { 1006 /* 1007 * Check security policy against packet attributes. 1008 */ 1009 error = ipsec_in_reject(sp, m); 1010 KEY_FREESP(&sp); 1011 } else { 1012 /* XXX error stat??? */ 1013 error = EINVAL; 1014DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 1015 } 1016 splx(s); 1017 if (error) 1018 goto bad; 1019 } 1020#endif /* FAST_IPSEC */ 1021 1022 /* 1023 * Switch out to protocol's input routine. 1024 */ 1025#if IFA_STATS 1026 if (ia && ip) 1027 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 1028#endif 1029 ipstat.ips_delivered++; 1030 { 1031 int off = hlen, nh = ip->ip_p; 1032 1033 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 1034 return; 1035 } 1036bad: 1037 m_freem(m); 1038 return; 1039 1040badcsum: 1041 ipstat.ips_badsum++; 1042 m_freem(m); 1043} 1044 1045/* 1046 * Take incoming datagram fragment and try to 1047 * reassemble it into whole datagram. If a chain for 1048 * reassembly of this datagram already exists, then it 1049 * is given as fp; otherwise have to make a chain. 1050 */ 1051struct mbuf * 1052ip_reass(struct ipqent *ipqe, struct ipq *fp, struct ipqhead *ipqhead) 1053{ 1054 struct mbuf *m = ipqe->ipqe_m; 1055 struct ipqent *nq, *p, *q; 1056 struct ip *ip; 1057 struct mbuf *t; 1058 int hlen = ipqe->ipqe_ip->ip_hl << 2; 1059 int i, next, s; 1060 1061 IPQ_LOCK_CHECK(); 1062 1063 /* 1064 * Presence of header sizes in mbufs 1065 * would confuse code below. 1066 */ 1067 m->m_data += hlen; 1068 m->m_len -= hlen; 1069 1070#ifdef notyet 1071 /* make sure fragment limit is up-to-date */ 1072 CHECK_NMBCLUSTER_PARAMS(); 1073 1074 /* If we have too many fragments, drop the older half. */ 1075 if (ip_nfrags >= ip_maxfrags) 1076 ip_reass_drophalf(void); 1077#endif 1078 1079 /* 1080 * We are about to add a fragment; increment frag count. 1081 */ 1082 ip_nfrags++; 1083 1084 /* 1085 * If first fragment to arrive, create a reassembly queue. 1086 */ 1087 if (fp == 0) { 1088 /* 1089 * Enforce upper bound on number of fragmented packets 1090 * for which we attempt reassembly; 1091 * If maxfrag is 0, never accept fragments. 1092 * If maxfrag is -1, accept all fragments without limitation. 1093 */ 1094 if (ip_maxfragpackets < 0) 1095 ; 1096 else if (ip_nfragpackets >= ip_maxfragpackets) 1097 goto dropfrag; 1098 ip_nfragpackets++; 1099 MALLOC(fp, struct ipq *, sizeof (struct ipq), 1100 M_FTABLE, M_NOWAIT); 1101 if (fp == NULL) 1102 goto dropfrag; 1103 LIST_INSERT_HEAD(ipqhead, fp, ipq_q); 1104 fp->ipq_nfrags = 1; 1105 fp->ipq_ttl = IPFRAGTTL; 1106 fp->ipq_p = ipqe->ipqe_ip->ip_p; 1107 fp->ipq_id = ipqe->ipqe_ip->ip_id; 1108 TAILQ_INIT(&fp->ipq_fragq); 1109 fp->ipq_src = ipqe->ipqe_ip->ip_src; 1110 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 1111 p = NULL; 1112 goto insert; 1113 } else { 1114 fp->ipq_nfrags++; 1115 } 1116 1117 /* 1118 * Find a segment which begins after this one does. 1119 */ 1120 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1121 p = q, q = TAILQ_NEXT(q, ipqe_q)) 1122 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off)) 1123 break; 1124 1125 /* 1126 * If there is a preceding segment, it may provide some of 1127 * our data already. If so, drop the data from the incoming 1128 * segment. If it provides all of our data, drop us. 1129 */ 1130 if (p != NULL) { 1131 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) - 1132 ntohs(ipqe->ipqe_ip->ip_off); 1133 if (i > 0) { 1134 if (i >= ntohs(ipqe->ipqe_ip->ip_len)) 1135 goto dropfrag; 1136 m_adj(ipqe->ipqe_m, i); 1137 ipqe->ipqe_ip->ip_off = 1138 htons(ntohs(ipqe->ipqe_ip->ip_off) + i); 1139 ipqe->ipqe_ip->ip_len = 1140 htons(ntohs(ipqe->ipqe_ip->ip_len) - i); 1141 } 1142 } 1143 1144 /* 1145 * While we overlap succeeding segments trim them or, 1146 * if they are completely covered, dequeue them. 1147 */ 1148 for (; q != NULL && 1149 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) > 1150 ntohs(q->ipqe_ip->ip_off); q = nq) { 1151 i = (ntohs(ipqe->ipqe_ip->ip_off) + 1152 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off); 1153 if (i < ntohs(q->ipqe_ip->ip_len)) { 1154 q->ipqe_ip->ip_len = 1155 htons(ntohs(q->ipqe_ip->ip_len) - i); 1156 q->ipqe_ip->ip_off = 1157 htons(ntohs(q->ipqe_ip->ip_off) + i); 1158 m_adj(q->ipqe_m, i); 1159 break; 1160 } 1161 nq = TAILQ_NEXT(q, ipqe_q); 1162 m_freem(q->ipqe_m); 1163 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1164 s = splvm(); 1165 pool_put(&ipqent_pool, q); 1166 splx(s); 1167 fp->ipq_nfrags--; 1168 ip_nfrags--; 1169 } 1170 1171insert: 1172 /* 1173 * Stick new segment in its place; 1174 * check for complete reassembly. 1175 */ 1176 if (p == NULL) { 1177 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 1178 } else { 1179 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q); 1180 } 1181 next = 0; 1182 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 1183 p = q, q = TAILQ_NEXT(q, ipqe_q)) { 1184 if (ntohs(q->ipqe_ip->ip_off) != next) 1185 return (0); 1186 next += ntohs(q->ipqe_ip->ip_len); 1187 } 1188 if (p->ipqe_mff) 1189 return (0); 1190 1191 /* 1192 * Reassembly is complete. Check for a bogus message size and 1193 * concatenate fragments. 1194 */ 1195 q = TAILQ_FIRST(&fp->ipq_fragq); 1196 ip = q->ipqe_ip; 1197 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 1198 ipstat.ips_toolong++; 1199 ip_freef(fp); 1200 return (0); 1201 } 1202 m = q->ipqe_m; 1203 t = m->m_next; 1204 m->m_next = 0; 1205 m_cat(m, t); 1206 nq = TAILQ_NEXT(q, ipqe_q); 1207 s = splvm(); 1208 pool_put(&ipqent_pool, q); 1209 splx(s); 1210 for (q = nq; q != NULL; q = nq) { 1211 t = q->ipqe_m; 1212 nq = TAILQ_NEXT(q, ipqe_q); 1213 s = splvm(); 1214 pool_put(&ipqent_pool, q); 1215 splx(s); 1216 m_cat(m, t); 1217 } 1218 ip_nfrags -= fp->ipq_nfrags; 1219 1220 /* 1221 * Create header for new ip packet by 1222 * modifying header of first packet; 1223 * dequeue and discard fragment reassembly header. 1224 * Make header visible. 1225 */ 1226 ip->ip_len = htons(next); 1227 ip->ip_src = fp->ipq_src; 1228 ip->ip_dst = fp->ipq_dst; 1229 LIST_REMOVE(fp, ipq_q); 1230 FREE(fp, M_FTABLE); 1231 ip_nfragpackets--; 1232 m->m_len += (ip->ip_hl << 2); 1233 m->m_data -= (ip->ip_hl << 2); 1234 /* some debugging cruft by sklower, below, will go away soon */ 1235 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1236 int plen = 0; 1237 for (t = m; t; t = t->m_next) 1238 plen += t->m_len; 1239 m->m_pkthdr.len = plen; 1240 m->m_pkthdr.csum_flags = 0; 1241 } 1242 return (m); 1243 1244dropfrag: 1245 if (fp != 0) 1246 fp->ipq_nfrags--; 1247 ip_nfrags--; 1248 ipstat.ips_fragdropped++; 1249 m_freem(m); 1250 s = splvm(); 1251 pool_put(&ipqent_pool, ipqe); 1252 splx(s); 1253 return (0); 1254} 1255 1256/* 1257 * Free a fragment reassembly header and all 1258 * associated datagrams. 1259 */ 1260void 1261ip_freef(struct ipq *fp) 1262{ 1263 struct ipqent *q, *p; 1264 u_int nfrags = 0; 1265 int s; 1266 1267 IPQ_LOCK_CHECK(); 1268 1269 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) { 1270 p = TAILQ_NEXT(q, ipqe_q); 1271 m_freem(q->ipqe_m); 1272 nfrags++; 1273 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1274 s = splvm(); 1275 pool_put(&ipqent_pool, q); 1276 splx(s); 1277 } 1278 1279 if (nfrags != fp->ipq_nfrags) 1280 printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags); 1281 ip_nfrags -= nfrags; 1282 LIST_REMOVE(fp, ipq_q); 1283 FREE(fp, M_FTABLE); 1284 ip_nfragpackets--; 1285} 1286 1287/* 1288 * IP reassembly TTL machinery for multiplicative drop. 1289 */ 1290static u_int fragttl_histo[(IPFRAGTTL+1)]; 1291 1292 1293/* 1294 * Decrement TTL of all reasembly queue entries by `ticks'. 1295 * Count number of distinct fragments (as opposed to partial, fragmented 1296 * datagrams) in the reassembly queue. While we traverse the entire 1297 * reassembly queue, compute and return the median TTL over all fragments. 1298 */ 1299static u_int 1300ip_reass_ttl_decr(u_int ticks) 1301{ 1302 u_int nfrags, median, dropfraction, keepfraction; 1303 struct ipq *fp, *nfp; 1304 int i; 1305 1306 nfrags = 0; 1307 memset(fragttl_histo, 0, sizeof fragttl_histo); 1308 1309 for (i = 0; i < IPREASS_NHASH; i++) { 1310 for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) { 1311 fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ? 1312 0 : fp->ipq_ttl - ticks); 1313 nfp = LIST_NEXT(fp, ipq_q); 1314 if (fp->ipq_ttl == 0) { 1315 ipstat.ips_fragtimeout++; 1316 ip_freef(fp); 1317 } else { 1318 nfrags += fp->ipq_nfrags; 1319 fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags; 1320 } 1321 } 1322 } 1323 1324 KASSERT(ip_nfrags == nfrags); 1325 1326 /* Find median (or other drop fraction) in histogram. */ 1327 dropfraction = (ip_nfrags / 2); 1328 keepfraction = ip_nfrags - dropfraction; 1329 for (i = IPFRAGTTL, median = 0; i >= 0; i--) { 1330 median += fragttl_histo[i]; 1331 if (median >= keepfraction) 1332 break; 1333 } 1334 1335 /* Return TTL of median (or other fraction). */ 1336 return (u_int)i; 1337} 1338 1339void 1340ip_reass_drophalf(void) 1341{ 1342 1343 u_int median_ticks; 1344 /* 1345 * Compute median TTL of all fragments, and count frags 1346 * with that TTL or lower (roughly half of all fragments). 1347 */ 1348 median_ticks = ip_reass_ttl_decr(0); 1349 1350 /* Drop half. */ 1351 median_ticks = ip_reass_ttl_decr(median_ticks); 1352 1353} 1354 1355/* 1356 * IP timer processing; 1357 * if a timer expires on a reassembly 1358 * queue, discard it. 1359 */ 1360void 1361ip_slowtimo(void) 1362{ 1363 static u_int dropscanidx = 0; 1364 u_int i; 1365 u_int median_ttl; 1366 int s = splsoftnet(); 1367 1368 IPQ_LOCK(); 1369 1370 /* Age TTL of all fragments by 1 tick .*/ 1371 median_ttl = ip_reass_ttl_decr(1); 1372 1373 /* make sure fragment limit is up-to-date */ 1374 CHECK_NMBCLUSTER_PARAMS(); 1375 1376 /* If we have too many fragments, drop the older half. */ 1377 if (ip_nfrags > ip_maxfrags) 1378 ip_reass_ttl_decr(median_ttl); 1379 1380 /* 1381 * If we are over the maximum number of fragmented packets 1382 * (due to the limit being lowered), drain off 1383 * enough to get down to the new limit. Start draining 1384 * from the reassembly hashqueue most recently drained. 1385 */ 1386 if (ip_maxfragpackets < 0) 1387 ; 1388 else { 1389 int wrapped = 0; 1390 1391 i = dropscanidx; 1392 while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) { 1393 while (LIST_FIRST(&ipq[i]) != NULL) 1394 ip_freef(LIST_FIRST(&ipq[i])); 1395 if (++i >= IPREASS_NHASH) { 1396 i = 0; 1397 } 1398 /* 1399 * Dont scan forever even if fragment counters are 1400 * wrong: stop after scanning entire reassembly queue. 1401 */ 1402 if (i == dropscanidx) 1403 wrapped = 1; 1404 } 1405 dropscanidx = i; 1406 } 1407 IPQ_UNLOCK(); 1408 splx(s); 1409} 1410 1411/* 1412 * Drain off all datagram fragments. 1413 */ 1414void 1415ip_drain(void) 1416{ 1417 1418 /* 1419 * We may be called from a device's interrupt context. If 1420 * the ipq is already busy, just bail out now. 1421 */ 1422 if (ipq_lock_try() == 0) 1423 return; 1424 1425 /* 1426 * Drop half the total fragments now. If more mbufs are needed, 1427 * we will be called again soon. 1428 */ 1429 ip_reass_drophalf(); 1430 1431 IPQ_UNLOCK(); 1432} 1433 1434/* 1435 * Do option processing on a datagram, 1436 * possibly discarding it if bad options are encountered, 1437 * or forwarding it if source-routed. 1438 * Returns 1 if packet has been forwarded/freed, 1439 * 0 if the packet should be processed further. 1440 */ 1441int 1442ip_dooptions(struct mbuf *m) 1443{ 1444 struct ip *ip = mtod(m, struct ip *); 1445 u_char *cp, *cp0; 1446 struct ip_timestamp *ipt; 1447 struct in_ifaddr *ia; 1448 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1449 struct in_addr dst; 1450 n_time ntime; 1451 1452 dst = ip->ip_dst; 1453 cp = (u_char *)(ip + 1); 1454 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1455 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1456 opt = cp[IPOPT_OPTVAL]; 1457 if (opt == IPOPT_EOL) 1458 break; 1459 if (opt == IPOPT_NOP) 1460 optlen = 1; 1461 else { 1462 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1463 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1464 goto bad; 1465 } 1466 optlen = cp[IPOPT_OLEN]; 1467 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1468 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1469 goto bad; 1470 } 1471 } 1472 switch (opt) { 1473 1474 default: 1475 break; 1476 1477 /* 1478 * Source routing with record. 1479 * Find interface with current destination address. 1480 * If none on this machine then drop if strictly routed, 1481 * or do nothing if loosely routed. 1482 * Record interface address and bring up next address 1483 * component. If strictly routed make sure next 1484 * address is on directly accessible net. 1485 */ 1486 case IPOPT_LSRR: 1487 case IPOPT_SSRR: 1488 if (ip_allowsrcrt == 0) { 1489 type = ICMP_UNREACH; 1490 code = ICMP_UNREACH_NET_PROHIB; 1491 goto bad; 1492 } 1493 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1494 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1495 goto bad; 1496 } 1497 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1498 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1499 goto bad; 1500 } 1501 ipaddr.sin_addr = ip->ip_dst; 1502 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1503 if (ia == 0) { 1504 if (opt == IPOPT_SSRR) { 1505 type = ICMP_UNREACH; 1506 code = ICMP_UNREACH_SRCFAIL; 1507 goto bad; 1508 } 1509 /* 1510 * Loose routing, and not at next destination 1511 * yet; nothing to do except forward. 1512 */ 1513 break; 1514 } 1515 off--; /* 0 origin */ 1516 if ((off + sizeof(struct in_addr)) > optlen) { 1517 /* 1518 * End of source route. Should be for us. 1519 */ 1520 save_rte(cp, ip->ip_src); 1521 break; 1522 } 1523 /* 1524 * locate outgoing interface 1525 */ 1526 bcopy((void *)(cp + off), (void *)&ipaddr.sin_addr, 1527 sizeof(ipaddr.sin_addr)); 1528 if (opt == IPOPT_SSRR) 1529 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); 1530 else 1531 ia = ip_rtaddr(ipaddr.sin_addr); 1532 if (ia == 0) { 1533 type = ICMP_UNREACH; 1534 code = ICMP_UNREACH_SRCFAIL; 1535 goto bad; 1536 } 1537 ip->ip_dst = ipaddr.sin_addr; 1538 bcopy((void *)&ia->ia_addr.sin_addr, 1539 (void *)(cp + off), sizeof(struct in_addr)); 1540 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1541 /* 1542 * Let ip_intr's mcast routing check handle mcast pkts 1543 */ 1544 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1545 break; 1546 1547 case IPOPT_RR: 1548 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1549 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1550 goto bad; 1551 } 1552 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1553 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1554 goto bad; 1555 } 1556 /* 1557 * If no space remains, ignore. 1558 */ 1559 off--; /* 0 origin */ 1560 if ((off + sizeof(struct in_addr)) > optlen) 1561 break; 1562 bcopy((void *)(&ip->ip_dst), (void *)&ipaddr.sin_addr, 1563 sizeof(ipaddr.sin_addr)); 1564 /* 1565 * locate outgoing interface; if we're the destination, 1566 * use the incoming interface (should be same). 1567 */ 1568 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1569 == NULL && 1570 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1571 type = ICMP_UNREACH; 1572 code = ICMP_UNREACH_HOST; 1573 goto bad; 1574 } 1575 bcopy((void *)&ia->ia_addr.sin_addr, 1576 (void *)(cp + off), sizeof(struct in_addr)); 1577 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1578 break; 1579 1580 case IPOPT_TS: 1581 code = cp - (u_char *)ip; 1582 ipt = (struct ip_timestamp *)cp; 1583 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1584 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1585 goto bad; 1586 } 1587 if (ipt->ipt_ptr < 5) { 1588 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1589 goto bad; 1590 } 1591 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1592 if (++ipt->ipt_oflw == 0) { 1593 code = (u_char *)&ipt->ipt_ptr - 1594 (u_char *)ip; 1595 goto bad; 1596 } 1597 break; 1598 } 1599 cp0 = (cp + ipt->ipt_ptr - 1); 1600 switch (ipt->ipt_flg) { 1601 1602 case IPOPT_TS_TSONLY: 1603 break; 1604 1605 case IPOPT_TS_TSANDADDR: 1606 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1607 sizeof(struct in_addr) > ipt->ipt_len) { 1608 code = (u_char *)&ipt->ipt_ptr - 1609 (u_char *)ip; 1610 goto bad; 1611 } 1612 ipaddr.sin_addr = dst; 1613 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1614 m->m_pkthdr.rcvif)); 1615 if (ia == 0) 1616 continue; 1617 bcopy(&ia->ia_addr.sin_addr, 1618 cp0, sizeof(struct in_addr)); 1619 ipt->ipt_ptr += sizeof(struct in_addr); 1620 break; 1621 1622 case IPOPT_TS_PRESPEC: 1623 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1624 sizeof(struct in_addr) > ipt->ipt_len) { 1625 code = (u_char *)&ipt->ipt_ptr - 1626 (u_char *)ip; 1627 goto bad; 1628 } 1629 bcopy(cp0, &ipaddr.sin_addr, 1630 sizeof(struct in_addr)); 1631 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1632 == NULL) 1633 continue; 1634 ipt->ipt_ptr += sizeof(struct in_addr); 1635 break; 1636 1637 default: 1638 /* XXX can't take &ipt->ipt_flg */ 1639 code = (u_char *)&ipt->ipt_ptr - 1640 (u_char *)ip + 1; 1641 goto bad; 1642 } 1643 ntime = iptime(); 1644 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1645 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1646 sizeof(n_time)); 1647 ipt->ipt_ptr += sizeof(n_time); 1648 } 1649 } 1650 if (forward) { 1651 if (ip_forwsrcrt == 0) { 1652 type = ICMP_UNREACH; 1653 code = ICMP_UNREACH_SRCFAIL; 1654 goto bad; 1655 } 1656 ip_forward(m, 1); 1657 return (1); 1658 } 1659 return (0); 1660bad: 1661 icmp_error(m, type, code, 0, 0); 1662 ipstat.ips_badoptions++; 1663 return (1); 1664} 1665 1666/* 1667 * Given address of next destination (final or next hop), 1668 * return internet address info of interface to be used to get there. 1669 */ 1670struct in_ifaddr * 1671ip_rtaddr(struct in_addr dst) 1672{ 1673 struct rtentry *rt; 1674 union { 1675 struct sockaddr dst; 1676 struct sockaddr_in dst4; 1677 } u; 1678 1679 sockaddr_in_init(&u.dst4, &dst, 0); 1680 1681 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) 1682 return NULL; 1683 1684 return ifatoia(rt->rt_ifa); 1685} 1686 1687/* 1688 * Save incoming source route for use in replies, 1689 * to be picked up later by ip_srcroute if the receiver is interested. 1690 */ 1691void 1692save_rte(u_char *option, struct in_addr dst) 1693{ 1694 unsigned olen; 1695 1696 olen = option[IPOPT_OLEN]; 1697#ifdef DIAGNOSTIC 1698 if (ipprintfs) 1699 printf("save_rte: olen %d\n", olen); 1700#endif /* 0 */ 1701 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1702 return; 1703 bcopy((void *)option, (void *)ip_srcrt.srcopt, olen); 1704 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1705 ip_srcrt.dst = dst; 1706} 1707 1708/* 1709 * Retrieve incoming source route for use in replies, 1710 * in the same form used by setsockopt. 1711 * The first hop is placed before the options, will be removed later. 1712 */ 1713struct mbuf * 1714ip_srcroute(void) 1715{ 1716 struct in_addr *p, *q; 1717 struct mbuf *m; 1718 1719 if (ip_nhops == 0) 1720 return NULL; 1721 m = m_get(M_DONTWAIT, MT_SOOPTS); 1722 if (m == 0) 1723 return NULL; 1724 1725 MCLAIM(m, &inetdomain.dom_mowner); 1726#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1727 1728 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1729 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1730 OPTSIZ; 1731#ifdef DIAGNOSTIC 1732 if (ipprintfs) 1733 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1734#endif 1735 1736 /* 1737 * First save first hop for return route 1738 */ 1739 p = &ip_srcrt.route[ip_nhops - 1]; 1740 *(mtod(m, struct in_addr *)) = *p--; 1741#ifdef DIAGNOSTIC 1742 if (ipprintfs) 1743 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1744#endif 1745 1746 /* 1747 * Copy option fields and padding (nop) to mbuf. 1748 */ 1749 ip_srcrt.nop = IPOPT_NOP; 1750 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1751 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop, 1752 OPTSIZ); 1753 q = (struct in_addr *)(mtod(m, char *) + 1754 sizeof(struct in_addr) + OPTSIZ); 1755#undef OPTSIZ 1756 /* 1757 * Record return path as an IP source route, 1758 * reversing the path (pointers are now aligned). 1759 */ 1760 while (p >= ip_srcrt.route) { 1761#ifdef DIAGNOSTIC 1762 if (ipprintfs) 1763 printf(" %x", ntohl(q->s_addr)); 1764#endif 1765 *q++ = *p--; 1766 } 1767 /* 1768 * Last hop goes to final destination. 1769 */ 1770 *q = ip_srcrt.dst; 1771#ifdef DIAGNOSTIC 1772 if (ipprintfs) 1773 printf(" %x\n", ntohl(q->s_addr)); 1774#endif 1775 return (m); 1776} 1777 1778const int inetctlerrmap[PRC_NCMDS] = { 1779 0, 0, 0, 0, 1780 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1781 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1782 EMSGSIZE, EHOSTUNREACH, 0, 0, 1783 0, 0, 0, 0, 1784 ENOPROTOOPT 1785}; 1786 1787/* 1788 * Forward a packet. If some error occurs return the sender 1789 * an icmp packet. Note we can't always generate a meaningful 1790 * icmp message because icmp doesn't have a large enough repertoire 1791 * of codes and types. 1792 * 1793 * If not forwarding, just drop the packet. This could be confusing 1794 * if ipforwarding was zero but some routing protocol was advancing 1795 * us as a gateway to somewhere. However, we must let the routing 1796 * protocol deal with that. 1797 * 1798 * The srcrt parameter indicates whether the packet is being forwarded 1799 * via a source route. 1800 */ 1801void 1802ip_forward(struct mbuf *m, int srcrt) 1803{ 1804 struct ip *ip = mtod(m, struct ip *); 1805 struct rtentry *rt; 1806 int error, type = 0, code = 0, destmtu = 0; 1807 struct mbuf *mcopy; 1808 n_long dest; 1809 union { 1810 struct sockaddr dst; 1811 struct sockaddr_in dst4; 1812 } u; 1813 1814 /* 1815 * We are now in the output path. 1816 */ 1817 MCLAIM(m, &ip_tx_mowner); 1818 1819 /* 1820 * Clear any in-bound checksum flags for this packet. 1821 */ 1822 m->m_pkthdr.csum_flags = 0; 1823 1824 dest = 0; 1825#ifdef DIAGNOSTIC 1826 if (ipprintfs) { 1827 printf("forward: src %s ", inet_ntoa(ip->ip_src)); 1828 printf("dst %s ttl %x\n", inet_ntoa(ip->ip_dst), ip->ip_ttl); 1829 } 1830#endif 1831 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1832 ipstat.ips_cantforward++; 1833 m_freem(m); 1834 return; 1835 } 1836 if (ip->ip_ttl <= IPTTLDEC) { 1837 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1838 return; 1839 } 1840 1841 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1842 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) { 1843 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1844 return; 1845 } 1846 1847 /* 1848 * Save at most 68 bytes of the packet in case 1849 * we need to generate an ICMP message to the src. 1850 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1851 */ 1852 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1853 if (mcopy) 1854 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1855 1856 ip->ip_ttl -= IPTTLDEC; 1857 1858 /* 1859 * If forwarding packet using same interface that it came in on, 1860 * perhaps should send a redirect to sender to shortcut a hop. 1861 * Only send redirect if source is sending directly to us, 1862 * and if packet was not source routed (or has any options). 1863 * Also, don't send redirect if forwarding using a default route 1864 * or a route modified by a redirect. 1865 */ 1866 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1867 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1868 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1869 ipsendredirects && !srcrt) { 1870 if (rt->rt_ifa && 1871 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1872 ifatoia(rt->rt_ifa)->ia_subnet) { 1873 if (rt->rt_flags & RTF_GATEWAY) 1874 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1875 else 1876 dest = ip->ip_dst.s_addr; 1877 /* 1878 * Router requirements says to only send host 1879 * redirects. 1880 */ 1881 type = ICMP_REDIRECT; 1882 code = ICMP_REDIRECT_HOST; 1883#ifdef DIAGNOSTIC 1884 if (ipprintfs) 1885 printf("redirect (%d) to %x\n", code, 1886 (u_int32_t)dest); 1887#endif 1888 } 1889 } 1890 1891 error = ip_output(m, NULL, &ipforward_rt, 1892 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1893 (struct ip_moptions *)NULL, (struct socket *)NULL); 1894 1895 if (error) 1896 ipstat.ips_cantforward++; 1897 else { 1898 ipstat.ips_forward++; 1899 if (type) 1900 ipstat.ips_redirectsent++; 1901 else { 1902 if (mcopy) { 1903#ifdef GATEWAY 1904 if (mcopy->m_flags & M_CANFASTFWD) 1905 ipflow_create(&ipforward_rt, mcopy); 1906#endif 1907 m_freem(mcopy); 1908 } 1909 return; 1910 } 1911 } 1912 if (mcopy == NULL) 1913 return; 1914 1915 switch (error) { 1916 1917 case 0: /* forwarded, but need redirect */ 1918 /* type, code set above */ 1919 break; 1920 1921 case ENETUNREACH: /* shouldn't happen, checked above */ 1922 case EHOSTUNREACH: 1923 case ENETDOWN: 1924 case EHOSTDOWN: 1925 default: 1926 type = ICMP_UNREACH; 1927 code = ICMP_UNREACH_HOST; 1928 break; 1929 1930 case EMSGSIZE: 1931 type = ICMP_UNREACH; 1932 code = ICMP_UNREACH_NEEDFRAG; 1933#if !defined(IPSEC) && !defined(FAST_IPSEC) 1934 if (ipforward_rt.ro_rt != NULL) 1935 destmtu = ipforward_rt.ro_rt->rt_ifp->if_mtu; 1936#else 1937 /* 1938 * If the packet is routed over IPsec tunnel, tell the 1939 * originator the tunnel MTU. 1940 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1941 * XXX quickhack!!! 1942 */ 1943 if (ipforward_rt.ro_rt != NULL) { 1944 struct secpolicy *sp; 1945 int ipsecerror; 1946 size_t ipsechdr; 1947 struct route *ro; 1948 1949 sp = ipsec4_getpolicybyaddr(mcopy, 1950 IPSEC_DIR_OUTBOUND, IP_FORWARDING, 1951 &ipsecerror); 1952 1953 if (sp == NULL) 1954 destmtu = ipforward_rt.ro_rt->rt_ifp->if_mtu; 1955 else { 1956 /* count IPsec header size */ 1957 ipsechdr = ipsec4_hdrsiz(mcopy, 1958 IPSEC_DIR_OUTBOUND, NULL); 1959 1960 /* 1961 * find the correct route for outer IPv4 1962 * header, compute tunnel MTU. 1963 */ 1964 1965 if (sp->req != NULL 1966 && sp->req->sav != NULL 1967 && sp->req->sav->sah != NULL) { 1968 ro = &sp->req->sav->sah->sa_route; 1969 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1970 destmtu = 1971 ro->ro_rt->rt_rmx.rmx_mtu ? 1972 ro->ro_rt->rt_rmx.rmx_mtu : 1973 ro->ro_rt->rt_ifp->if_mtu; 1974 destmtu -= ipsechdr; 1975 } 1976 } 1977 1978#ifdef IPSEC 1979 key_freesp(sp); 1980#else 1981 KEY_FREESP(&sp); 1982#endif 1983 } 1984 } 1985#endif /*IPSEC*/ 1986 ipstat.ips_cantfrag++; 1987 break; 1988 1989 case ENOBUFS: 1990#if 1 1991 /* 1992 * a router should not generate ICMP_SOURCEQUENCH as 1993 * required in RFC1812 Requirements for IP Version 4 Routers. 1994 * source quench could be a big problem under DoS attacks, 1995 * or if the underlying interface is rate-limited. 1996 */ 1997 if (mcopy) 1998 m_freem(mcopy); 1999 return; 2000#else 2001 type = ICMP_SOURCEQUENCH; 2002 code = 0; 2003 break; 2004#endif 2005 } 2006 icmp_error(mcopy, type, code, dest, destmtu); 2007} 2008 2009void 2010ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2011 struct mbuf *m) 2012{ 2013 2014 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2015 struct timeval tv; 2016 2017 microtime(&tv); 2018 *mp = sbcreatecontrol((void *) &tv, sizeof(tv), 2019 SCM_TIMESTAMP, SOL_SOCKET); 2020 if (*mp) 2021 mp = &(*mp)->m_next; 2022 } 2023 if (inp->inp_flags & INP_RECVDSTADDR) { 2024 *mp = sbcreatecontrol((void *) &ip->ip_dst, 2025 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2026 if (*mp) 2027 mp = &(*mp)->m_next; 2028 } 2029#ifdef notyet 2030 /* 2031 * XXX 2032 * Moving these out of udp_input() made them even more broken 2033 * than they already were. 2034 * - fenner@parc.xerox.com 2035 */ 2036 /* options were tossed already */ 2037 if (inp->inp_flags & INP_RECVOPTS) { 2038 *mp = sbcreatecontrol((void *) opts_deleted_above, 2039 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2040 if (*mp) 2041 mp = &(*mp)->m_next; 2042 } 2043 /* ip_srcroute doesn't do what we want here, need to fix */ 2044 if (inp->inp_flags & INP_RECVRETOPTS) { 2045 *mp = sbcreatecontrol((void *) ip_srcroute(), 2046 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2047 if (*mp) 2048 mp = &(*mp)->m_next; 2049 } 2050#endif 2051 if (inp->inp_flags & INP_RECVIF) { 2052 struct sockaddr_dl sdl; 2053 2054 sockaddr_dl_init(&sdl, sizeof(sdl), 2055 (m->m_pkthdr.rcvif != NULL) 2056 ? m->m_pkthdr.rcvif->if_index 2057 : 0, 2058 0, NULL, 0, NULL, 0); 2059 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 2060 if (*mp) 2061 mp = &(*mp)->m_next; 2062 } 2063} 2064 2065/* 2066 * sysctl helper routine for net.inet.ip.forwsrcrt. 2067 */ 2068static int 2069sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 2070{ 2071 int error, tmp; 2072 struct sysctlnode node; 2073 2074 node = *rnode; 2075 tmp = ip_forwsrcrt; 2076 node.sysctl_data = &tmp; 2077 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2078 if (error || newp == NULL) 2079 return (error); 2080 2081 if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 2082 0, NULL, NULL, NULL)) 2083 return (EPERM); 2084 2085 ip_forwsrcrt = tmp; 2086 2087 return (0); 2088} 2089 2090/* 2091 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 2092 * range of the new value and tweaks timers if it changes. 2093 */ 2094static int 2095sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 2096{ 2097 int error, tmp; 2098 struct sysctlnode node; 2099 2100 node = *rnode; 2101 tmp = ip_mtudisc_timeout; 2102 node.sysctl_data = &tmp; 2103 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2104 if (error || newp == NULL) 2105 return (error); 2106 if (tmp < 0) 2107 return (EINVAL); 2108 2109 ip_mtudisc_timeout = tmp; 2110 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 2111 2112 return (0); 2113} 2114 2115#ifdef GATEWAY 2116/* 2117 * sysctl helper routine for net.inet.ip.maxflows. 2118 */ 2119static int 2120sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS) 2121{ 2122 int s; 2123 2124 s = sysctl_lookup(SYSCTLFN_CALL(rnode)); 2125 if (s || newp == NULL) 2126 return (s); 2127 2128 s = splsoftnet(); 2129 ipflow_reap(0); 2130 splx(s); 2131 2132 return (0); 2133} 2134 2135static int 2136sysctl_net_inet_ip_hashsize(SYSCTLFN_ARGS) 2137{ 2138 int error, tmp; 2139 struct sysctlnode node; 2140 2141 node = *rnode; 2142 tmp = ip_hashsize; 2143 node.sysctl_data = &tmp; 2144 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2145 if (error || newp == NULL) 2146 return (error); 2147 2148 if ((tmp & (tmp - 1)) == 0 && tmp != 0) { 2149 /* 2150 * Can only fail due to malloc() 2151 */ 2152 if (ipflow_invalidate_all(tmp)) 2153 return ENOMEM; 2154 } else { 2155 /* 2156 * EINVAL if not a power of 2 2157 */ 2158 return EINVAL; 2159 } 2160 2161 return (0); 2162} 2163#endif /* GATEWAY */ 2164 2165 2166SYSCTL_SETUP(sysctl_net_inet_ip_setup, "sysctl net.inet.ip subtree setup") 2167{ 2168 extern int subnetsarelocal, hostzeroisbroadcast; 2169 2170 sysctl_createv(clog, 0, NULL, NULL, 2171 CTLFLAG_PERMANENT, 2172 CTLTYPE_NODE, "net", NULL, 2173 NULL, 0, NULL, 0, 2174 CTL_NET, CTL_EOL); 2175 sysctl_createv(clog, 0, NULL, NULL, 2176 CTLFLAG_PERMANENT, 2177 CTLTYPE_NODE, "inet", 2178 SYSCTL_DESCR("PF_INET related settings"), 2179 NULL, 0, NULL, 0, 2180 CTL_NET, PF_INET, CTL_EOL); 2181 sysctl_createv(clog, 0, NULL, NULL, 2182 CTLFLAG_PERMANENT, 2183 CTLTYPE_NODE, "ip", 2184 SYSCTL_DESCR("IPv4 related settings"), 2185 NULL, 0, NULL, 0, 2186 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 2187 2188 sysctl_createv(clog, 0, NULL, NULL, 2189 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2190 CTLTYPE_INT, "forwarding", 2191 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 2192 NULL, 0, &ipforwarding, 0, 2193 CTL_NET, PF_INET, IPPROTO_IP, 2194 IPCTL_FORWARDING, CTL_EOL); 2195 sysctl_createv(clog, 0, NULL, NULL, 2196 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2197 CTLTYPE_INT, "redirect", 2198 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 2199 NULL, 0, &ipsendredirects, 0, 2200 CTL_NET, PF_INET, IPPROTO_IP, 2201 IPCTL_SENDREDIRECTS, CTL_EOL); 2202 sysctl_createv(clog, 0, NULL, NULL, 2203 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2204 CTLTYPE_INT, "ttl", 2205 SYSCTL_DESCR("Default TTL for an INET datagram"), 2206 NULL, 0, &ip_defttl, 0, 2207 CTL_NET, PF_INET, IPPROTO_IP, 2208 IPCTL_DEFTTL, CTL_EOL); 2209#ifdef IPCTL_DEFMTU 2210 sysctl_createv(clog, 0, NULL, NULL, 2211 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 2212 CTLTYPE_INT, "mtu", 2213 SYSCTL_DESCR("Default MTA for an INET route"), 2214 NULL, 0, &ip_mtu, 0, 2215 CTL_NET, PF_INET, IPPROTO_IP, 2216 IPCTL_DEFMTU, CTL_EOL); 2217#endif /* IPCTL_DEFMTU */ 2218 sysctl_createv(clog, 0, NULL, NULL, 2219 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2220 CTLTYPE_INT, "forwsrcrt", 2221 SYSCTL_DESCR("Enable forwarding of source-routed " 2222 "datagrams"), 2223 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 2224 CTL_NET, PF_INET, IPPROTO_IP, 2225 IPCTL_FORWSRCRT, CTL_EOL); 2226 sysctl_createv(clog, 0, NULL, NULL, 2227 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2228 CTLTYPE_INT, "directed-broadcast", 2229 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 2230 NULL, 0, &ip_directedbcast, 0, 2231 CTL_NET, PF_INET, IPPROTO_IP, 2232 IPCTL_DIRECTEDBCAST, CTL_EOL); 2233 sysctl_createv(clog, 0, NULL, NULL, 2234 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2235 CTLTYPE_INT, "allowsrcrt", 2236 SYSCTL_DESCR("Accept source-routed datagrams"), 2237 NULL, 0, &ip_allowsrcrt, 0, 2238 CTL_NET, PF_INET, IPPROTO_IP, 2239 IPCTL_ALLOWSRCRT, CTL_EOL); 2240 sysctl_createv(clog, 0, NULL, NULL, 2241 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2242 CTLTYPE_INT, "subnetsarelocal", 2243 SYSCTL_DESCR("Whether logical subnets are considered " 2244 "local"), 2245 NULL, 0, &subnetsarelocal, 0, 2246 CTL_NET, PF_INET, IPPROTO_IP, 2247 IPCTL_SUBNETSARELOCAL, CTL_EOL); 2248 sysctl_createv(clog, 0, NULL, NULL, 2249 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2250 CTLTYPE_INT, "mtudisc", 2251 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 2252 NULL, 0, &ip_mtudisc, 0, 2253 CTL_NET, PF_INET, IPPROTO_IP, 2254 IPCTL_MTUDISC, CTL_EOL); 2255 sysctl_createv(clog, 0, NULL, NULL, 2256 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2257 CTLTYPE_INT, "anonportmin", 2258 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 2259 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 2260 CTL_NET, PF_INET, IPPROTO_IP, 2261 IPCTL_ANONPORTMIN, CTL_EOL); 2262 sysctl_createv(clog, 0, NULL, NULL, 2263 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2264 CTLTYPE_INT, "anonportmax", 2265 SYSCTL_DESCR("Highest ephemeral port number to assign"), 2266 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 2267 CTL_NET, PF_INET, IPPROTO_IP, 2268 IPCTL_ANONPORTMAX, CTL_EOL); 2269 sysctl_createv(clog, 0, NULL, NULL, 2270 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2271 CTLTYPE_INT, "mtudisctimeout", 2272 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 2273 sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0, 2274 CTL_NET, PF_INET, IPPROTO_IP, 2275 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 2276#ifdef GATEWAY 2277 sysctl_createv(clog, 0, NULL, NULL, 2278 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2279 CTLTYPE_INT, "maxflows", 2280 SYSCTL_DESCR("Number of flows for fast forwarding"), 2281 sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0, 2282 CTL_NET, PF_INET, IPPROTO_IP, 2283 IPCTL_MAXFLOWS, CTL_EOL); 2284 sysctl_createv(clog, 0, NULL, NULL, 2285 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2286 CTLTYPE_INT, "hashsize", 2287 SYSCTL_DESCR("Size of hash table for fast forwarding (IPv4)"), 2288 sysctl_net_inet_ip_hashsize, 0, &ip_hashsize, 0, 2289 CTL_NET, PF_INET, IPPROTO_IP, 2290 CTL_CREATE, CTL_EOL); 2291#endif /* GATEWAY */ 2292 sysctl_createv(clog, 0, NULL, NULL, 2293 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2294 CTLTYPE_INT, "hostzerobroadcast", 2295 SYSCTL_DESCR("All zeroes address is broadcast address"), 2296 NULL, 0, &hostzeroisbroadcast, 0, 2297 CTL_NET, PF_INET, IPPROTO_IP, 2298 IPCTL_HOSTZEROBROADCAST, CTL_EOL); 2299#if NGIF > 0 2300 sysctl_createv(clog, 0, NULL, NULL, 2301 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2302 CTLTYPE_INT, "gifttl", 2303 SYSCTL_DESCR("Default TTL for a gif tunnel datagram"), 2304 NULL, 0, &ip_gif_ttl, 0, 2305 CTL_NET, PF_INET, IPPROTO_IP, 2306 IPCTL_GIF_TTL, CTL_EOL); 2307#endif /* NGIF */ 2308#ifndef IPNOPRIVPORTS 2309 sysctl_createv(clog, 0, NULL, NULL, 2310 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2311 CTLTYPE_INT, "lowportmin", 2312 SYSCTL_DESCR("Lowest privileged ephemeral port number " 2313 "to assign"), 2314 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 2315 CTL_NET, PF_INET, IPPROTO_IP, 2316 IPCTL_LOWPORTMIN, CTL_EOL); 2317 sysctl_createv(clog, 0, NULL, NULL, 2318 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2319 CTLTYPE_INT, "lowportmax", 2320 SYSCTL_DESCR("Highest privileged ephemeral port number " 2321 "to assign"), 2322 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 2323 CTL_NET, PF_INET, IPPROTO_IP, 2324 IPCTL_LOWPORTMAX, CTL_EOL); 2325#endif /* IPNOPRIVPORTS */ 2326 sysctl_createv(clog, 0, NULL, NULL, 2327 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2328 CTLTYPE_INT, "maxfragpackets", 2329 SYSCTL_DESCR("Maximum number of fragments to retain for " 2330 "possible reassembly"), 2331 NULL, 0, &ip_maxfragpackets, 0, 2332 CTL_NET, PF_INET, IPPROTO_IP, 2333 IPCTL_MAXFRAGPACKETS, CTL_EOL); 2334#if NGRE > 0 2335 sysctl_createv(clog, 0, NULL, NULL, 2336 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2337 CTLTYPE_INT, "grettl", 2338 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 2339 NULL, 0, &ip_gre_ttl, 0, 2340 CTL_NET, PF_INET, IPPROTO_IP, 2341 IPCTL_GRE_TTL, CTL_EOL); 2342#endif /* NGRE */ 2343 sysctl_createv(clog, 0, NULL, NULL, 2344 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2345 CTLTYPE_INT, "checkinterface", 2346 SYSCTL_DESCR("Enable receive side of Strong ES model " 2347 "from RFC1122"), 2348 NULL, 0, &ip_checkinterface, 0, 2349 CTL_NET, PF_INET, IPPROTO_IP, 2350 IPCTL_CHECKINTERFACE, CTL_EOL); 2351 sysctl_createv(clog, 0, NULL, NULL, 2352 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2353 CTLTYPE_INT, "random_id", 2354 SYSCTL_DESCR("Assign random ip_id values"), 2355 NULL, 0, &ip_do_randomid, 0, 2356 CTL_NET, PF_INET, IPPROTO_IP, 2357 IPCTL_RANDOMID, CTL_EOL); 2358 sysctl_createv(clog, 0, NULL, NULL, 2359 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2360 CTLTYPE_INT, "do_loopback_cksum", 2361 SYSCTL_DESCR("Perform IP checksum on loopback"), 2362 NULL, 0, &ip_do_loopback_cksum, 0, 2363 CTL_NET, PF_INET, IPPROTO_IP, 2364 IPCTL_LOOPBACKCKSUM, CTL_EOL); 2365 sysctl_createv(clog, 0, NULL, NULL, 2366 CTLFLAG_PERMANENT, 2367 CTLTYPE_STRUCT, "stats", 2368 SYSCTL_DESCR("IP statistics"), 2369 NULL, 0, &ipstat, sizeof(ipstat), 2370 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 2371 CTL_EOL); 2372} 2373