ip_input.c revision 1.134
1/* $NetBSD: ip_input.c,v 1.134 2001/05/21 03:31:36 lukem 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. All advertising materials mentioning features or use of this software 82 * must display the following acknowledgement: 83 * This product includes software developed by the University of 84 * California, Berkeley and its contributors. 85 * 4. Neither the name of the University nor the names of its contributors 86 * may be used to endorse or promote products derived from this software 87 * without specific prior written permission. 88 * 89 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 90 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 91 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 92 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 93 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 94 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 95 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 96 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 97 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 98 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 99 * SUCH DAMAGE. 100 * 101 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 102 */ 103 104#include "opt_gateway.h" 105#include "opt_pfil_hooks.h" 106#include "opt_ipsec.h" 107#include "opt_mrouting.h" 108 109#include <sys/param.h> 110#include <sys/systm.h> 111#include <sys/malloc.h> 112#include <sys/mbuf.h> 113#include <sys/domain.h> 114#include <sys/protosw.h> 115#include <sys/socket.h> 116#include <sys/socketvar.h> 117#include <sys/errno.h> 118#include <sys/time.h> 119#include <sys/kernel.h> 120#include <sys/proc.h> 121#include <sys/pool.h> 122 123#include <uvm/uvm_extern.h> 124 125#include <sys/sysctl.h> 126 127#include <net/if.h> 128#include <net/if_dl.h> 129#include <net/route.h> 130#include <net/pfil.h> 131 132#include <netinet/in.h> 133#include <netinet/in_systm.h> 134#include <netinet/ip.h> 135#include <netinet/in_pcb.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" 142 143#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 151 152#ifndef IPFORWARDING 153#ifdef GATEWAY 154#define IPFORWARDING 1 /* forward IP packets not for us */ 155#else /* GATEWAY */ 156#define IPFORWARDING 0 /* don't forward IP packets not for us */ 157#endif /* GATEWAY */ 158#endif /* IPFORWARDING */ 159#ifndef IPSENDREDIRECTS 160#define IPSENDREDIRECTS 1 161#endif 162#ifndef IPFORWSRCRT 163#define IPFORWSRCRT 1 /* forward source-routed packets */ 164#endif 165#ifndef IPALLOWSRCRT 166#define IPALLOWSRCRT 1 /* allow source-routed packets */ 167#endif 168#ifndef IPMTUDISC 169#define IPMTUDISC 0 170#endif 171#ifndef IPMTUDISCTIMEOUT 172#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 173#endif 174 175/* 176 * Note: DIRECTED_BROADCAST is handled this way so that previous 177 * configuration using this option will Just Work. 178 */ 179#ifndef IPDIRECTEDBCAST 180#ifdef DIRECTED_BROADCAST 181#define IPDIRECTEDBCAST 1 182#else 183#define IPDIRECTEDBCAST 0 184#endif /* DIRECTED_BROADCAST */ 185#endif /* IPDIRECTEDBCAST */ 186int ipforwarding = IPFORWARDING; 187int ipsendredirects = IPSENDREDIRECTS; 188int ip_defttl = IPDEFTTL; 189int ip_forwsrcrt = IPFORWSRCRT; 190int ip_directedbcast = IPDIRECTEDBCAST; 191int ip_allowsrcrt = IPALLOWSRCRT; 192int ip_mtudisc = IPMTUDISC; 193u_int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 194#ifdef DIAGNOSTIC 195int ipprintfs = 0; 196#endif 197 198struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 199 200extern struct domain inetdomain; 201int ipqmaxlen = IFQ_MAXLEN; 202struct in_ifaddrhead in_ifaddr; 203struct in_ifaddrhashhead *in_ifaddrhashtbl; 204struct ifqueue ipintrq; 205struct ipstat ipstat; 206u_int16_t ip_id; 207 208#ifdef PFIL_HOOKS 209struct pfil_head inet_pfil_hook; 210#endif 211 212struct ipqhead ipq; 213int ipq_locked; 214int ip_nfragpackets = 0; 215int ip_maxfragpackets = 200; 216 217static __inline int ipq_lock_try __P((void)); 218static __inline void ipq_unlock __P((void)); 219 220static __inline int 221ipq_lock_try() 222{ 223 int s; 224 225 /* 226 * Use splvm() -- we're bloking things that would cause 227 * mbuf allocation. 228 */ 229 s = splvm(); 230 if (ipq_locked) { 231 splx(s); 232 return (0); 233 } 234 ipq_locked = 1; 235 splx(s); 236 return (1); 237} 238 239static __inline void 240ipq_unlock() 241{ 242 int s; 243 244 s = splvm(); 245 ipq_locked = 0; 246 splx(s); 247} 248 249#ifdef DIAGNOSTIC 250#define IPQ_LOCK() \ 251do { \ 252 if (ipq_lock_try() == 0) { \ 253 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 254 panic("ipq_lock"); \ 255 } \ 256} while (0) 257#define IPQ_LOCK_CHECK() \ 258do { \ 259 if (ipq_locked == 0) { \ 260 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 261 panic("ipq lock check"); \ 262 } \ 263} while (0) 264#else 265#define IPQ_LOCK() (void) ipq_lock_try() 266#define IPQ_LOCK_CHECK() /* nothing */ 267#endif 268 269#define IPQ_UNLOCK() ipq_unlock() 270 271struct pool ipqent_pool; 272 273/* 274 * We need to save the IP options in case a protocol wants to respond 275 * to an incoming packet over the same route if the packet got here 276 * using IP source routing. This allows connection establishment and 277 * maintenance when the remote end is on a network that is not known 278 * to us. 279 */ 280int ip_nhops = 0; 281static struct ip_srcrt { 282 struct in_addr dst; /* final destination */ 283 char nop; /* one NOP to align */ 284 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 285 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 286} ip_srcrt; 287 288static void save_rte __P((u_char *, struct in_addr)); 289 290/* 291 * IP initialization: fill in IP protocol switch table. 292 * All protocols not implemented in kernel go to raw IP protocol handler. 293 */ 294void 295ip_init() 296{ 297 struct protosw *pr; 298 int i; 299 300 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 301 0, NULL, NULL, M_IPQ); 302 303 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 304 if (pr == 0) 305 panic("ip_init"); 306 for (i = 0; i < IPPROTO_MAX; i++) 307 ip_protox[i] = pr - inetsw; 308 for (pr = inetdomain.dom_protosw; 309 pr < inetdomain.dom_protoswNPROTOSW; pr++) 310 if (pr->pr_domain->dom_family == PF_INET && 311 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 312 ip_protox[pr->pr_protocol] = pr - inetsw; 313 LIST_INIT(&ipq); 314 ip_id = time.tv_sec & 0xffff; 315 ipintrq.ifq_maxlen = ipqmaxlen; 316 TAILQ_INIT(&in_ifaddr); 317 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 318 M_WAITOK, &in_ifaddrhash); 319 if (ip_mtudisc != 0) 320 ip_mtudisc_timeout_q = 321 rt_timer_queue_create(ip_mtudisc_timeout); 322#ifdef GATEWAY 323 ipflow_init(); 324#endif 325 326#ifdef PFIL_HOOKS 327 /* Register our Packet Filter hook. */ 328 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 329 inet_pfil_hook.ph_af = AF_INET; 330 i = pfil_head_register(&inet_pfil_hook); 331 if (i != 0) 332 printf("ip_init: WARNING: unable to register pfil hook, " 333 "error %d\n", i); 334#endif /* PFIL_HOOKS */ 335} 336 337struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 338struct route ipforward_rt; 339 340/* 341 * IP software interrupt routine 342 */ 343void 344ipintr() 345{ 346 int s; 347 struct mbuf *m; 348 349 while (1) { 350 s = splnet(); 351 IF_DEQUEUE(&ipintrq, m); 352 splx(s); 353 if (m == 0) 354 return; 355 ip_input(m); 356 } 357} 358 359/* 360 * Ip input routine. Checksum and byte swap header. If fragmented 361 * try to reassemble. Process options. Pass to next level. 362 */ 363void 364ip_input(struct mbuf *m) 365{ 366 struct ip *ip = NULL; 367 struct ipq *fp; 368 struct in_ifaddr *ia; 369 struct ifaddr *ifa; 370 struct ipqent *ipqe; 371 int hlen = 0, mff, len; 372 int downmatch; 373 374#ifdef DIAGNOSTIC 375 if ((m->m_flags & M_PKTHDR) == 0) 376 panic("ipintr no HDR"); 377#endif 378#ifdef IPSEC 379 /* 380 * should the inner packet be considered authentic? 381 * see comment in ah4_input(). 382 */ 383 if (m) { 384 m->m_flags &= ~M_AUTHIPHDR; 385 m->m_flags &= ~M_AUTHIPDGM; 386 } 387#endif 388 /* 389 * If no IP addresses have been set yet but the interfaces 390 * are receiving, can't do anything with incoming packets yet. 391 */ 392 if (in_ifaddr.tqh_first == 0) 393 goto bad; 394 ipstat.ips_total++; 395 if (m->m_len < sizeof (struct ip) && 396 (m = m_pullup(m, sizeof (struct ip))) == 0) { 397 ipstat.ips_toosmall++; 398 return; 399 } 400 ip = mtod(m, struct ip *); 401 if (ip->ip_v != IPVERSION) { 402 ipstat.ips_badvers++; 403 goto bad; 404 } 405 hlen = ip->ip_hl << 2; 406 if (hlen < sizeof(struct ip)) { /* minimum header length */ 407 ipstat.ips_badhlen++; 408 goto bad; 409 } 410 if (hlen > m->m_len) { 411 if ((m = m_pullup(m, hlen)) == 0) { 412 ipstat.ips_badhlen++; 413 return; 414 } 415 ip = mtod(m, struct ip *); 416 } 417 418 /* 419 * RFC1122: packets with a multicast source address are 420 * not allowed. 421 */ 422 if (IN_MULTICAST(ip->ip_src.s_addr)) { 423 ipstat.ips_badaddr++; 424 goto bad; 425 } 426 427 /* 127/8 must not appear on wire - RFC1122 */ 428 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 429 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 430 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 431 ipstat.ips_badaddr++; 432 goto bad; 433 } 434 } 435 436 if (in_cksum(m, hlen) != 0) { 437 ipstat.ips_badsum++; 438 goto bad; 439 } 440 441 /* Retrieve the packet length. */ 442 len = ntohs(ip->ip_len); 443 444 /* 445 * Check for additional length bogosity 446 */ 447 if (len < hlen) { 448 ipstat.ips_badlen++; 449 goto bad; 450 } 451 452 /* 453 * Check that the amount of data in the buffers 454 * is as at least much as the IP header would have us expect. 455 * Trim mbufs if longer than we expect. 456 * Drop packet if shorter than we expect. 457 */ 458 if (m->m_pkthdr.len < len) { 459 ipstat.ips_tooshort++; 460 goto bad; 461 } 462 if (m->m_pkthdr.len > len) { 463 if (m->m_len == m->m_pkthdr.len) { 464 m->m_len = len; 465 m->m_pkthdr.len = len; 466 } else 467 m_adj(m, len - m->m_pkthdr.len); 468 } 469 470#ifdef IPSEC 471 /* ipflow (IP fast fowarding) is not compatible with IPsec. */ 472 m->m_flags &= ~M_CANFASTFWD; 473#else 474 /* 475 * Assume that we can create a fast-forward IP flow entry 476 * based on this packet. 477 */ 478 m->m_flags |= M_CANFASTFWD; 479#endif 480 481#ifdef PFIL_HOOKS 482 /* 483 * Run through list of hooks for input packets. If there are any 484 * filters which require that additional packets in the flow are 485 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 486 * Note that filters must _never_ set this flag, as another filter 487 * in the list may have previously cleared it. 488 */ 489 /* 490 * let ipfilter look at packet on the wire, 491 * not the decapsulated packet. 492 */ 493#ifdef IPSEC 494 if (!ipsec_gethist(m, NULL)) 495#else 496 if (1) 497#endif 498 { 499 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 500 PFIL_IN) != 0) 501 return; 502 if (m == NULL) 503 return; 504 ip = mtod(m, struct ip *); 505 } 506#endif /* PFIL_HOOKS */ 507 508#ifdef ALTQ 509 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 510 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 511 /* packet dropped by traffic conditioner */ 512 return; 513 } 514#endif 515 516 /* 517 * Convert fields to host representation. 518 */ 519 NTOHS(ip->ip_len); 520 NTOHS(ip->ip_off); 521 522 /* 523 * Process options and, if not destined for us, 524 * ship it on. ip_dooptions returns 1 when an 525 * error was detected (causing an icmp message 526 * to be sent and the original packet to be freed). 527 */ 528 ip_nhops = 0; /* for source routed packets */ 529 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 530 return; 531 532 /* 533 * Check our list of addresses, to see if the packet is for us. 534 * 535 * Traditional 4.4BSD did not consult IFF_UP at all. 536 * The behavior here is to treat addresses on !IFF_UP interface 537 * as not mine. 538 */ 539 downmatch = 0; 540 for (ia = IN_IFADDR_HASH(ip->ip_dst.s_addr).lh_first; 541 ia != NULL; 542 ia = ia->ia_hash.le_next) { 543 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 544 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 545 break; 546 else 547 downmatch++; 548 } 549 } 550 if (ia != NULL) 551 goto ours; 552 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 553 for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first; 554 ifa != NULL; ifa = ifa->ifa_list.tqe_next) { 555 if (ifa->ifa_addr->sa_family != AF_INET) continue; 556 ia = ifatoia(ifa); 557 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 558 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 559 /* 560 * Look for all-0's host part (old broadcast addr), 561 * either for subnet or net. 562 */ 563 ip->ip_dst.s_addr == ia->ia_subnet || 564 ip->ip_dst.s_addr == ia->ia_net) 565 goto ours; 566 /* 567 * An interface with IP address zero accepts 568 * all packets that arrive on that interface. 569 */ 570 if (in_nullhost(ia->ia_addr.sin_addr)) 571 goto ours; 572 } 573 } 574 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 575 struct in_multi *inm; 576#ifdef MROUTING 577 extern struct socket *ip_mrouter; 578 579 if (m->m_flags & M_EXT) { 580 if ((m = m_pullup(m, hlen)) == 0) { 581 ipstat.ips_toosmall++; 582 return; 583 } 584 ip = mtod(m, struct ip *); 585 } 586 587 if (ip_mrouter) { 588 /* 589 * If we are acting as a multicast router, all 590 * incoming multicast packets are passed to the 591 * kernel-level multicast forwarding function. 592 * The packet is returned (relatively) intact; if 593 * ip_mforward() returns a non-zero value, the packet 594 * must be discarded, else it may be accepted below. 595 * 596 * (The IP ident field is put in the same byte order 597 * as expected when ip_mforward() is called from 598 * ip_output().) 599 */ 600 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 601 ipstat.ips_cantforward++; 602 m_freem(m); 603 return; 604 } 605 606 /* 607 * The process-level routing demon needs to receive 608 * all multicast IGMP packets, whether or not this 609 * host belongs to their destination groups. 610 */ 611 if (ip->ip_p == IPPROTO_IGMP) 612 goto ours; 613 ipstat.ips_forward++; 614 } 615#endif 616 /* 617 * See if we belong to the destination multicast group on the 618 * arrival interface. 619 */ 620 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 621 if (inm == NULL) { 622 ipstat.ips_cantforward++; 623 m_freem(m); 624 return; 625 } 626 goto ours; 627 } 628 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 629 in_nullhost(ip->ip_dst)) 630 goto ours; 631 632 /* 633 * Not for us; forward if possible and desirable. 634 */ 635 if (ipforwarding == 0) { 636 ipstat.ips_cantforward++; 637 m_freem(m); 638 } else { 639 /* 640 * If ip_dst matched any of my address on !IFF_UP interface, 641 * and there's no IFF_UP interface that matches ip_dst, 642 * send icmp unreach. Forwarding it will result in in-kernel 643 * forwarding loop till TTL goes to 0. 644 */ 645 if (downmatch) { 646 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 647 ipstat.ips_cantforward++; 648 return; 649 } 650 ip_forward(m, 0); 651 } 652 return; 653 654ours: 655 /* 656 * If offset or IP_MF are set, must reassemble. 657 * Otherwise, nothing need be done. 658 * (We could look in the reassembly queue to see 659 * if the packet was previously fragmented, 660 * but it's not worth the time; just let them time out.) 661 */ 662 if (ip->ip_off & ~(IP_DF|IP_RF)) { 663 /* 664 * Look for queue of fragments 665 * of this datagram. 666 */ 667 IPQ_LOCK(); 668 for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next) 669 if (ip->ip_id == fp->ipq_id && 670 in_hosteq(ip->ip_src, fp->ipq_src) && 671 in_hosteq(ip->ip_dst, fp->ipq_dst) && 672 ip->ip_p == fp->ipq_p) 673 goto found; 674 fp = 0; 675found: 676 677 /* 678 * Adjust ip_len to not reflect header, 679 * set ipqe_mff if more fragments are expected, 680 * convert offset of this to bytes. 681 */ 682 ip->ip_len -= hlen; 683 mff = (ip->ip_off & IP_MF) != 0; 684 if (mff) { 685 /* 686 * Make sure that fragments have a data length 687 * that's a non-zero multiple of 8 bytes. 688 */ 689 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 690 ipstat.ips_badfrags++; 691 IPQ_UNLOCK(); 692 goto bad; 693 } 694 } 695 ip->ip_off <<= 3; 696 697 /* 698 * If datagram marked as having more fragments 699 * or if this is not the first fragment, 700 * attempt reassembly; if it succeeds, proceed. 701 */ 702 if (mff || ip->ip_off) { 703 ipstat.ips_fragments++; 704 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 705 if (ipqe == NULL) { 706 ipstat.ips_rcvmemdrop++; 707 IPQ_UNLOCK(); 708 goto bad; 709 } 710 ipqe->ipqe_mff = mff; 711 ipqe->ipqe_m = m; 712 ipqe->ipqe_ip = ip; 713 m = ip_reass(ipqe, fp); 714 if (m == 0) { 715 IPQ_UNLOCK(); 716 return; 717 } 718 ipstat.ips_reassembled++; 719 ip = mtod(m, struct ip *); 720 hlen = ip->ip_hl << 2; 721 ip->ip_len += hlen; 722 } else 723 if (fp) 724 ip_freef(fp); 725 IPQ_UNLOCK(); 726 } 727 728#ifdef IPSEC 729 /* 730 * enforce IPsec policy checking if we are seeing last header. 731 * note that we do not visit this with protocols with pcb layer 732 * code - like udp/tcp/raw ip. 733 */ 734 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 735 ipsec4_in_reject(m, NULL)) { 736 ipsecstat.in_polvio++; 737 goto bad; 738 } 739#endif 740 741 /* 742 * Switch out to protocol's input routine. 743 */ 744#if IFA_STATS 745 if (ia && ip) 746 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 747#endif 748 ipstat.ips_delivered++; 749 { 750 int off = hlen, nh = ip->ip_p; 751 752 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 753 return; 754 } 755bad: 756 m_freem(m); 757} 758 759/* 760 * Take incoming datagram fragment and try to 761 * reassemble it into whole datagram. If a chain for 762 * reassembly of this datagram already exists, then it 763 * is given as fp; otherwise have to make a chain. 764 */ 765struct mbuf * 766ip_reass(ipqe, fp) 767 struct ipqent *ipqe; 768 struct ipq *fp; 769{ 770 struct mbuf *m = ipqe->ipqe_m; 771 struct ipqent *nq, *p, *q; 772 struct ip *ip; 773 struct mbuf *t; 774 int hlen = ipqe->ipqe_ip->ip_hl << 2; 775 int i, next; 776 777 IPQ_LOCK_CHECK(); 778 779 /* 780 * Presence of header sizes in mbufs 781 * would confuse code below. 782 */ 783 m->m_data += hlen; 784 m->m_len -= hlen; 785 786 /* 787 * If first fragment to arrive, create a reassembly queue. 788 */ 789 if (fp == 0) { 790 /* 791 * Enforce upper bound on number of fragmented packets 792 * for which we attempt reassembly; 793 * If maxfrag is 0, never accept fragments. 794 * If maxfrag is -1, accept all fragments without limitation. 795 */ 796 if (ip_maxfragpackets < 0) 797 ; 798 else if (ip_nfragpackets >= ip_maxfragpackets) 799 goto dropfrag; 800 ip_nfragpackets++; 801 MALLOC(fp, struct ipq *, sizeof (struct ipq), 802 M_FTABLE, M_NOWAIT); 803 if (fp == NULL) 804 goto dropfrag; 805 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 806 fp->ipq_ttl = IPFRAGTTL; 807 fp->ipq_p = ipqe->ipqe_ip->ip_p; 808 fp->ipq_id = ipqe->ipqe_ip->ip_id; 809 LIST_INIT(&fp->ipq_fragq); 810 fp->ipq_src = ipqe->ipqe_ip->ip_src; 811 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 812 p = NULL; 813 goto insert; 814 } 815 816 /* 817 * Find a segment which begins after this one does. 818 */ 819 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 820 p = q, q = q->ipqe_q.le_next) 821 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 822 break; 823 824 /* 825 * If there is a preceding segment, it may provide some of 826 * our data already. If so, drop the data from the incoming 827 * segment. If it provides all of our data, drop us. 828 */ 829 if (p != NULL) { 830 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 831 ipqe->ipqe_ip->ip_off; 832 if (i > 0) { 833 if (i >= ipqe->ipqe_ip->ip_len) 834 goto dropfrag; 835 m_adj(ipqe->ipqe_m, i); 836 ipqe->ipqe_ip->ip_off += i; 837 ipqe->ipqe_ip->ip_len -= i; 838 } 839 } 840 841 /* 842 * While we overlap succeeding segments trim them or, 843 * if they are completely covered, dequeue them. 844 */ 845 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 846 q->ipqe_ip->ip_off; q = nq) { 847 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 848 q->ipqe_ip->ip_off; 849 if (i < q->ipqe_ip->ip_len) { 850 q->ipqe_ip->ip_len -= i; 851 q->ipqe_ip->ip_off += i; 852 m_adj(q->ipqe_m, i); 853 break; 854 } 855 nq = q->ipqe_q.le_next; 856 m_freem(q->ipqe_m); 857 LIST_REMOVE(q, ipqe_q); 858 pool_put(&ipqent_pool, q); 859 } 860 861insert: 862 /* 863 * Stick new segment in its place; 864 * check for complete reassembly. 865 */ 866 if (p == NULL) { 867 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 868 } else { 869 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 870 } 871 next = 0; 872 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 873 p = q, q = q->ipqe_q.le_next) { 874 if (q->ipqe_ip->ip_off != next) 875 return (0); 876 next += q->ipqe_ip->ip_len; 877 } 878 if (p->ipqe_mff) 879 return (0); 880 881 /* 882 * Reassembly is complete. Check for a bogus message size and 883 * concatenate fragments. 884 */ 885 q = fp->ipq_fragq.lh_first; 886 ip = q->ipqe_ip; 887 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 888 ipstat.ips_toolong++; 889 ip_freef(fp); 890 return (0); 891 } 892 m = q->ipqe_m; 893 t = m->m_next; 894 m->m_next = 0; 895 m_cat(m, t); 896 nq = q->ipqe_q.le_next; 897 pool_put(&ipqent_pool, q); 898 for (q = nq; q != NULL; q = nq) { 899 t = q->ipqe_m; 900 nq = q->ipqe_q.le_next; 901 pool_put(&ipqent_pool, q); 902 m_cat(m, t); 903 } 904 905 /* 906 * Create header for new ip packet by 907 * modifying header of first packet; 908 * dequeue and discard fragment reassembly header. 909 * Make header visible. 910 */ 911 ip->ip_len = next; 912 ip->ip_src = fp->ipq_src; 913 ip->ip_dst = fp->ipq_dst; 914 LIST_REMOVE(fp, ipq_q); 915 FREE(fp, M_FTABLE); 916 ip_nfragpackets--; 917 m->m_len += (ip->ip_hl << 2); 918 m->m_data -= (ip->ip_hl << 2); 919 /* some debugging cruft by sklower, below, will go away soon */ 920 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 921 int plen = 0; 922 for (t = m; t; t = t->m_next) 923 plen += t->m_len; 924 m->m_pkthdr.len = plen; 925 } 926 return (m); 927 928dropfrag: 929 ipstat.ips_fragdropped++; 930 m_freem(m); 931 pool_put(&ipqent_pool, ipqe); 932 return (0); 933} 934 935/* 936 * Free a fragment reassembly header and all 937 * associated datagrams. 938 */ 939void 940ip_freef(fp) 941 struct ipq *fp; 942{ 943 struct ipqent *q, *p; 944 945 IPQ_LOCK_CHECK(); 946 947 for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { 948 p = q->ipqe_q.le_next; 949 m_freem(q->ipqe_m); 950 LIST_REMOVE(q, ipqe_q); 951 pool_put(&ipqent_pool, q); 952 } 953 LIST_REMOVE(fp, ipq_q); 954 FREE(fp, M_FTABLE); 955 ip_nfragpackets--; 956} 957 958/* 959 * IP timer processing; 960 * if a timer expires on a reassembly 961 * queue, discard it. 962 */ 963void 964ip_slowtimo() 965{ 966 struct ipq *fp, *nfp; 967 int s = splsoftnet(); 968 969 IPQ_LOCK(); 970 for (fp = ipq.lh_first; fp != NULL; fp = nfp) { 971 nfp = fp->ipq_q.le_next; 972 if (--fp->ipq_ttl == 0) { 973 ipstat.ips_fragtimeout++; 974 ip_freef(fp); 975 } 976 } 977 /* 978 * If we are over the maximum number of fragments 979 * (due to the limit being lowered), drain off 980 * enough to get down to the new limit. 981 */ 982 if (ip_maxfragpackets < 0) 983 ; 984 else { 985 while (ip_nfragpackets > ip_maxfragpackets && ipq.lh_first) 986 ip_freef(ipq.lh_first); 987 } 988 IPQ_UNLOCK(); 989#ifdef GATEWAY 990 ipflow_slowtimo(); 991#endif 992 splx(s); 993} 994 995/* 996 * Drain off all datagram fragments. 997 */ 998void 999ip_drain() 1000{ 1001 1002 /* 1003 * We may be called from a device's interrupt context. If 1004 * the ipq is already busy, just bail out now. 1005 */ 1006 if (ipq_lock_try() == 0) 1007 return; 1008 1009 while (ipq.lh_first != NULL) { 1010 ipstat.ips_fragdropped++; 1011 ip_freef(ipq.lh_first); 1012 } 1013 1014 IPQ_UNLOCK(); 1015} 1016 1017/* 1018 * Do option processing on a datagram, 1019 * possibly discarding it if bad options are encountered, 1020 * or forwarding it if source-routed. 1021 * Returns 1 if packet has been forwarded/freed, 1022 * 0 if the packet should be processed further. 1023 */ 1024int 1025ip_dooptions(m) 1026 struct mbuf *m; 1027{ 1028 struct ip *ip = mtod(m, struct ip *); 1029 u_char *cp, *cp0; 1030 struct ip_timestamp *ipt; 1031 struct in_ifaddr *ia; 1032 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1033 struct in_addr dst; 1034 n_time ntime; 1035 1036 dst = ip->ip_dst; 1037 cp = (u_char *)(ip + 1); 1038 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1039 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1040 opt = cp[IPOPT_OPTVAL]; 1041 if (opt == IPOPT_EOL) 1042 break; 1043 if (opt == IPOPT_NOP) 1044 optlen = 1; 1045 else { 1046 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1047 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1048 goto bad; 1049 } 1050 optlen = cp[IPOPT_OLEN]; 1051 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1052 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1053 goto bad; 1054 } 1055 } 1056 switch (opt) { 1057 1058 default: 1059 break; 1060 1061 /* 1062 * Source routing with record. 1063 * Find interface with current destination address. 1064 * If none on this machine then drop if strictly routed, 1065 * or do nothing if loosely routed. 1066 * Record interface address and bring up next address 1067 * component. If strictly routed make sure next 1068 * address is on directly accessible net. 1069 */ 1070 case IPOPT_LSRR: 1071 case IPOPT_SSRR: 1072 if (ip_allowsrcrt == 0) { 1073 type = ICMP_UNREACH; 1074 code = ICMP_UNREACH_NET_PROHIB; 1075 goto bad; 1076 } 1077 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1078 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1079 goto bad; 1080 } 1081 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1082 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1083 goto bad; 1084 } 1085 ipaddr.sin_addr = ip->ip_dst; 1086 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1087 if (ia == 0) { 1088 if (opt == IPOPT_SSRR) { 1089 type = ICMP_UNREACH; 1090 code = ICMP_UNREACH_SRCFAIL; 1091 goto bad; 1092 } 1093 /* 1094 * Loose routing, and not at next destination 1095 * yet; nothing to do except forward. 1096 */ 1097 break; 1098 } 1099 off--; /* 0 origin */ 1100 if ((off + sizeof(struct in_addr)) > optlen) { 1101 /* 1102 * End of source route. Should be for us. 1103 */ 1104 save_rte(cp, ip->ip_src); 1105 break; 1106 } 1107 /* 1108 * locate outgoing interface 1109 */ 1110 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1111 sizeof(ipaddr.sin_addr)); 1112 if (opt == IPOPT_SSRR) 1113 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1114 else 1115 ia = ip_rtaddr(ipaddr.sin_addr); 1116 if (ia == 0) { 1117 type = ICMP_UNREACH; 1118 code = ICMP_UNREACH_SRCFAIL; 1119 goto bad; 1120 } 1121 ip->ip_dst = ipaddr.sin_addr; 1122 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1123 (caddr_t)(cp + off), sizeof(struct in_addr)); 1124 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1125 /* 1126 * Let ip_intr's mcast routing check handle mcast pkts 1127 */ 1128 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1129 break; 1130 1131 case IPOPT_RR: 1132 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1133 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1134 goto bad; 1135 } 1136 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1137 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1138 goto bad; 1139 } 1140 /* 1141 * If no space remains, ignore. 1142 */ 1143 off--; /* 0 origin */ 1144 if ((off + sizeof(struct in_addr)) > optlen) 1145 break; 1146 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1147 sizeof(ipaddr.sin_addr)); 1148 /* 1149 * locate outgoing interface; if we're the destination, 1150 * use the incoming interface (should be same). 1151 */ 1152 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1153 == NULL && 1154 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1155 type = ICMP_UNREACH; 1156 code = ICMP_UNREACH_HOST; 1157 goto bad; 1158 } 1159 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1160 (caddr_t)(cp + off), sizeof(struct in_addr)); 1161 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1162 break; 1163 1164 case IPOPT_TS: 1165 code = cp - (u_char *)ip; 1166 ipt = (struct ip_timestamp *)cp; 1167 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1168 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1169 goto bad; 1170 } 1171 if (ipt->ipt_ptr < 5) { 1172 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1173 goto bad; 1174 } 1175 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1176 if (++ipt->ipt_oflw == 0) { 1177 code = (u_char *)&ipt->ipt_ptr - 1178 (u_char *)ip; 1179 goto bad; 1180 } 1181 break; 1182 } 1183 cp0 = (cp + ipt->ipt_ptr - 1); 1184 switch (ipt->ipt_flg) { 1185 1186 case IPOPT_TS_TSONLY: 1187 break; 1188 1189 case IPOPT_TS_TSANDADDR: 1190 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1191 sizeof(struct in_addr) > ipt->ipt_len) { 1192 code = (u_char *)&ipt->ipt_ptr - 1193 (u_char *)ip; 1194 goto bad; 1195 } 1196 ipaddr.sin_addr = dst; 1197 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1198 m->m_pkthdr.rcvif)); 1199 if (ia == 0) 1200 continue; 1201 bcopy(&ia->ia_addr.sin_addr, 1202 cp0, sizeof(struct in_addr)); 1203 ipt->ipt_ptr += sizeof(struct in_addr); 1204 break; 1205 1206 case IPOPT_TS_PRESPEC: 1207 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1208 sizeof(struct in_addr) > ipt->ipt_len) { 1209 code = (u_char *)&ipt->ipt_ptr - 1210 (u_char *)ip; 1211 goto bad; 1212 } 1213 bcopy(cp0, &ipaddr.sin_addr, 1214 sizeof(struct in_addr)); 1215 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1216 == NULL) 1217 continue; 1218 ipt->ipt_ptr += sizeof(struct in_addr); 1219 break; 1220 1221 default: 1222 /* XXX can't take &ipt->ipt_flg */ 1223 code = (u_char *)&ipt->ipt_ptr - 1224 (u_char *)ip + 1; 1225 goto bad; 1226 } 1227 ntime = iptime(); 1228 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1229 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1230 sizeof(n_time)); 1231 ipt->ipt_ptr += sizeof(n_time); 1232 } 1233 } 1234 if (forward) { 1235 if (ip_forwsrcrt == 0) { 1236 type = ICMP_UNREACH; 1237 code = ICMP_UNREACH_SRCFAIL; 1238 goto bad; 1239 } 1240 ip_forward(m, 1); 1241 return (1); 1242 } 1243 return (0); 1244bad: 1245 icmp_error(m, type, code, 0, 0); 1246 ipstat.ips_badoptions++; 1247 return (1); 1248} 1249 1250/* 1251 * Given address of next destination (final or next hop), 1252 * return internet address info of interface to be used to get there. 1253 */ 1254struct in_ifaddr * 1255ip_rtaddr(dst) 1256 struct in_addr dst; 1257{ 1258 struct sockaddr_in *sin; 1259 1260 sin = satosin(&ipforward_rt.ro_dst); 1261 1262 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1263 if (ipforward_rt.ro_rt) { 1264 RTFREE(ipforward_rt.ro_rt); 1265 ipforward_rt.ro_rt = 0; 1266 } 1267 sin->sin_family = AF_INET; 1268 sin->sin_len = sizeof(*sin); 1269 sin->sin_addr = dst; 1270 1271 rtalloc(&ipforward_rt); 1272 } 1273 if (ipforward_rt.ro_rt == 0) 1274 return ((struct in_ifaddr *)0); 1275 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1276} 1277 1278/* 1279 * Save incoming source route for use in replies, 1280 * to be picked up later by ip_srcroute if the receiver is interested. 1281 */ 1282void 1283save_rte(option, dst) 1284 u_char *option; 1285 struct in_addr dst; 1286{ 1287 unsigned olen; 1288 1289 olen = option[IPOPT_OLEN]; 1290#ifdef DIAGNOSTIC 1291 if (ipprintfs) 1292 printf("save_rte: olen %d\n", olen); 1293#endif /* 0 */ 1294 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1295 return; 1296 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1297 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1298 ip_srcrt.dst = dst; 1299} 1300 1301/* 1302 * Retrieve incoming source route for use in replies, 1303 * in the same form used by setsockopt. 1304 * The first hop is placed before the options, will be removed later. 1305 */ 1306struct mbuf * 1307ip_srcroute() 1308{ 1309 struct in_addr *p, *q; 1310 struct mbuf *m; 1311 1312 if (ip_nhops == 0) 1313 return ((struct mbuf *)0); 1314 m = m_get(M_DONTWAIT, MT_SOOPTS); 1315 if (m == 0) 1316 return ((struct mbuf *)0); 1317 1318#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1319 1320 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1321 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1322 OPTSIZ; 1323#ifdef DIAGNOSTIC 1324 if (ipprintfs) 1325 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1326#endif 1327 1328 /* 1329 * First save first hop for return route 1330 */ 1331 p = &ip_srcrt.route[ip_nhops - 1]; 1332 *(mtod(m, struct in_addr *)) = *p--; 1333#ifdef DIAGNOSTIC 1334 if (ipprintfs) 1335 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1336#endif 1337 1338 /* 1339 * Copy option fields and padding (nop) to mbuf. 1340 */ 1341 ip_srcrt.nop = IPOPT_NOP; 1342 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1343 bcopy((caddr_t)&ip_srcrt.nop, 1344 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1345 q = (struct in_addr *)(mtod(m, caddr_t) + 1346 sizeof(struct in_addr) + OPTSIZ); 1347#undef OPTSIZ 1348 /* 1349 * Record return path as an IP source route, 1350 * reversing the path (pointers are now aligned). 1351 */ 1352 while (p >= ip_srcrt.route) { 1353#ifdef DIAGNOSTIC 1354 if (ipprintfs) 1355 printf(" %x", ntohl(q->s_addr)); 1356#endif 1357 *q++ = *p--; 1358 } 1359 /* 1360 * Last hop goes to final destination. 1361 */ 1362 *q = ip_srcrt.dst; 1363#ifdef DIAGNOSTIC 1364 if (ipprintfs) 1365 printf(" %x\n", ntohl(q->s_addr)); 1366#endif 1367 return (m); 1368} 1369 1370/* 1371 * Strip out IP options, at higher 1372 * level protocol in the kernel. 1373 * Second argument is buffer to which options 1374 * will be moved, and return value is their length. 1375 * XXX should be deleted; last arg currently ignored. 1376 */ 1377void 1378ip_stripoptions(m, mopt) 1379 struct mbuf *m; 1380 struct mbuf *mopt; 1381{ 1382 int i; 1383 struct ip *ip = mtod(m, struct ip *); 1384 caddr_t opts; 1385 int olen; 1386 1387 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1388 opts = (caddr_t)(ip + 1); 1389 i = m->m_len - (sizeof (struct ip) + olen); 1390 bcopy(opts + olen, opts, (unsigned)i); 1391 m->m_len -= olen; 1392 if (m->m_flags & M_PKTHDR) 1393 m->m_pkthdr.len -= olen; 1394 ip->ip_len -= olen; 1395 ip->ip_hl = sizeof (struct ip) >> 2; 1396} 1397 1398int inetctlerrmap[PRC_NCMDS] = { 1399 0, 0, 0, 0, 1400 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1401 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1402 EMSGSIZE, EHOSTUNREACH, 0, 0, 1403 0, 0, 0, 0, 1404 ENOPROTOOPT 1405}; 1406 1407/* 1408 * Forward a packet. If some error occurs return the sender 1409 * an icmp packet. Note we can't always generate a meaningful 1410 * icmp message because icmp doesn't have a large enough repertoire 1411 * of codes and types. 1412 * 1413 * If not forwarding, just drop the packet. This could be confusing 1414 * if ipforwarding was zero but some routing protocol was advancing 1415 * us as a gateway to somewhere. However, we must let the routing 1416 * protocol deal with that. 1417 * 1418 * The srcrt parameter indicates whether the packet is being forwarded 1419 * via a source route. 1420 */ 1421void 1422ip_forward(m, srcrt) 1423 struct mbuf *m; 1424 int srcrt; 1425{ 1426 struct ip *ip = mtod(m, struct ip *); 1427 struct sockaddr_in *sin; 1428 struct rtentry *rt; 1429 int error, type = 0, code = 0; 1430 struct mbuf *mcopy; 1431 n_long dest; 1432 struct ifnet *destifp; 1433#ifdef IPSEC 1434 struct ifnet dummyifp; 1435#endif 1436 1437 dest = 0; 1438#ifdef DIAGNOSTIC 1439 if (ipprintfs) 1440 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1441 ntohl(ip->ip_src.s_addr), 1442 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1443#endif 1444 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1445 ipstat.ips_cantforward++; 1446 m_freem(m); 1447 return; 1448 } 1449 if (ip->ip_ttl <= IPTTLDEC) { 1450 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1451 return; 1452 } 1453 ip->ip_ttl -= IPTTLDEC; 1454 1455 sin = satosin(&ipforward_rt.ro_dst); 1456 if ((rt = ipforward_rt.ro_rt) == 0 || 1457 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1458 if (ipforward_rt.ro_rt) { 1459 RTFREE(ipforward_rt.ro_rt); 1460 ipforward_rt.ro_rt = 0; 1461 } 1462 sin->sin_family = AF_INET; 1463 sin->sin_len = sizeof(struct sockaddr_in); 1464 sin->sin_addr = ip->ip_dst; 1465 1466 rtalloc(&ipforward_rt); 1467 if (ipforward_rt.ro_rt == 0) { 1468 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1469 return; 1470 } 1471 rt = ipforward_rt.ro_rt; 1472 } 1473 1474 /* 1475 * Save at most 68 bytes of the packet in case 1476 * we need to generate an ICMP message to the src. 1477 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1478 */ 1479 mcopy = m_copym(m, 0, imin((int)ip->ip_len, 68), M_DONTWAIT); 1480 if (mcopy) 1481 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1482 1483 /* 1484 * If forwarding packet using same interface that it came in on, 1485 * perhaps should send a redirect to sender to shortcut a hop. 1486 * Only send redirect if source is sending directly to us, 1487 * and if packet was not source routed (or has any options). 1488 * Also, don't send redirect if forwarding using a default route 1489 * or a route modified by a redirect. 1490 */ 1491 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1492 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1493 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1494 ipsendredirects && !srcrt) { 1495 if (rt->rt_ifa && 1496 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1497 ifatoia(rt->rt_ifa)->ia_subnet) { 1498 if (rt->rt_flags & RTF_GATEWAY) 1499 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1500 else 1501 dest = ip->ip_dst.s_addr; 1502 /* 1503 * Router requirements says to only send host 1504 * redirects. 1505 */ 1506 type = ICMP_REDIRECT; 1507 code = ICMP_REDIRECT_HOST; 1508#ifdef DIAGNOSTIC 1509 if (ipprintfs) 1510 printf("redirect (%d) to %x\n", code, 1511 (u_int32_t)dest); 1512#endif 1513 } 1514 } 1515 1516#ifdef IPSEC 1517 /* Don't lookup socket in forwarding case */ 1518 (void)ipsec_setsocket(m, NULL); 1519#endif 1520 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1521 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1522 if (error) 1523 ipstat.ips_cantforward++; 1524 else { 1525 ipstat.ips_forward++; 1526 if (type) 1527 ipstat.ips_redirectsent++; 1528 else { 1529 if (mcopy) { 1530#ifdef GATEWAY 1531 if (mcopy->m_flags & M_CANFASTFWD) 1532 ipflow_create(&ipforward_rt, mcopy); 1533#endif 1534 m_freem(mcopy); 1535 } 1536 return; 1537 } 1538 } 1539 if (mcopy == NULL) 1540 return; 1541 destifp = NULL; 1542 1543 switch (error) { 1544 1545 case 0: /* forwarded, but need redirect */ 1546 /* type, code set above */ 1547 break; 1548 1549 case ENETUNREACH: /* shouldn't happen, checked above */ 1550 case EHOSTUNREACH: 1551 case ENETDOWN: 1552 case EHOSTDOWN: 1553 default: 1554 type = ICMP_UNREACH; 1555 code = ICMP_UNREACH_HOST; 1556 break; 1557 1558 case EMSGSIZE: 1559 type = ICMP_UNREACH; 1560 code = ICMP_UNREACH_NEEDFRAG; 1561#ifndef IPSEC 1562 if (ipforward_rt.ro_rt) 1563 destifp = ipforward_rt.ro_rt->rt_ifp; 1564#else 1565 /* 1566 * If the packet is routed over IPsec tunnel, tell the 1567 * originator the tunnel MTU. 1568 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1569 * XXX quickhack!!! 1570 */ 1571 if (ipforward_rt.ro_rt) { 1572 struct secpolicy *sp; 1573 int ipsecerror; 1574 size_t ipsechdr; 1575 struct route *ro; 1576 1577 sp = ipsec4_getpolicybyaddr(mcopy, 1578 IPSEC_DIR_OUTBOUND, 1579 IP_FORWARDING, 1580 &ipsecerror); 1581 1582 if (sp == NULL) 1583 destifp = ipforward_rt.ro_rt->rt_ifp; 1584 else { 1585 /* count IPsec header size */ 1586 ipsechdr = ipsec4_hdrsiz(mcopy, 1587 IPSEC_DIR_OUTBOUND, 1588 NULL); 1589 1590 /* 1591 * find the correct route for outer IPv4 1592 * header, compute tunnel MTU. 1593 * 1594 * XXX BUG ALERT 1595 * The "dummyifp" code relies upon the fact 1596 * that icmp_error() touches only ifp->if_mtu. 1597 */ 1598 /*XXX*/ 1599 destifp = NULL; 1600 if (sp->req != NULL 1601 && sp->req->sav != NULL 1602 && sp->req->sav->sah != NULL) { 1603 ro = &sp->req->sav->sah->sa_route; 1604 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1605 dummyifp.if_mtu = 1606 ro->ro_rt->rt_ifp->if_mtu; 1607 dummyifp.if_mtu -= ipsechdr; 1608 destifp = &dummyifp; 1609 } 1610 } 1611 1612 key_freesp(sp); 1613 } 1614 } 1615#endif /*IPSEC*/ 1616 ipstat.ips_cantfrag++; 1617 break; 1618 1619 case ENOBUFS: 1620 type = ICMP_SOURCEQUENCH; 1621 code = 0; 1622 break; 1623 } 1624 icmp_error(mcopy, type, code, dest, destifp); 1625} 1626 1627void 1628ip_savecontrol(inp, mp, ip, m) 1629 struct inpcb *inp; 1630 struct mbuf **mp; 1631 struct ip *ip; 1632 struct mbuf *m; 1633{ 1634 1635 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1636 struct timeval tv; 1637 1638 microtime(&tv); 1639 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1640 SCM_TIMESTAMP, SOL_SOCKET); 1641 if (*mp) 1642 mp = &(*mp)->m_next; 1643 } 1644 if (inp->inp_flags & INP_RECVDSTADDR) { 1645 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1646 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1647 if (*mp) 1648 mp = &(*mp)->m_next; 1649 } 1650#ifdef notyet 1651 /* 1652 * XXX 1653 * Moving these out of udp_input() made them even more broken 1654 * than they already were. 1655 * - fenner@parc.xerox.com 1656 */ 1657 /* options were tossed already */ 1658 if (inp->inp_flags & INP_RECVOPTS) { 1659 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1660 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1661 if (*mp) 1662 mp = &(*mp)->m_next; 1663 } 1664 /* ip_srcroute doesn't do what we want here, need to fix */ 1665 if (inp->inp_flags & INP_RECVRETOPTS) { 1666 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1667 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1668 if (*mp) 1669 mp = &(*mp)->m_next; 1670 } 1671#endif 1672 if (inp->inp_flags & INP_RECVIF) { 1673 struct sockaddr_dl sdl; 1674 1675 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1676 sdl.sdl_family = AF_LINK; 1677 sdl.sdl_index = m->m_pkthdr.rcvif ? 1678 m->m_pkthdr.rcvif->if_index : 0; 1679 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1680 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1681 IP_RECVIF, IPPROTO_IP); 1682 if (*mp) 1683 mp = &(*mp)->m_next; 1684 } 1685} 1686 1687int 1688ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1689 int *name; 1690 u_int namelen; 1691 void *oldp; 1692 size_t *oldlenp; 1693 void *newp; 1694 size_t newlen; 1695{ 1696 extern int subnetsarelocal, hostzeroisbroadcast; 1697 1698 int error, old; 1699 1700 /* All sysctl names at this level are terminal. */ 1701 if (namelen != 1) 1702 return (ENOTDIR); 1703 1704 switch (name[0]) { 1705 case IPCTL_FORWARDING: 1706 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1707 case IPCTL_SENDREDIRECTS: 1708 return (sysctl_int(oldp, oldlenp, newp, newlen, 1709 &ipsendredirects)); 1710 case IPCTL_DEFTTL: 1711 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1712#ifdef notyet 1713 case IPCTL_DEFMTU: 1714 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1715#endif 1716 case IPCTL_FORWSRCRT: 1717 /* Don't allow this to change in a secure environment. */ 1718 if (securelevel > 0) 1719 return (sysctl_rdint(oldp, oldlenp, newp, 1720 ip_forwsrcrt)); 1721 else 1722 return (sysctl_int(oldp, oldlenp, newp, newlen, 1723 &ip_forwsrcrt)); 1724 case IPCTL_DIRECTEDBCAST: 1725 return (sysctl_int(oldp, oldlenp, newp, newlen, 1726 &ip_directedbcast)); 1727 case IPCTL_ALLOWSRCRT: 1728 return (sysctl_int(oldp, oldlenp, newp, newlen, 1729 &ip_allowsrcrt)); 1730 case IPCTL_SUBNETSARELOCAL: 1731 return (sysctl_int(oldp, oldlenp, newp, newlen, 1732 &subnetsarelocal)); 1733 case IPCTL_MTUDISC: 1734 error = sysctl_int(oldp, oldlenp, newp, newlen, 1735 &ip_mtudisc); 1736 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1737 ip_mtudisc_timeout_q = 1738 rt_timer_queue_create(ip_mtudisc_timeout); 1739 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1740 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1741 ip_mtudisc_timeout_q = NULL; 1742 } 1743 return error; 1744 case IPCTL_ANONPORTMIN: 1745 old = anonportmin; 1746 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1747 if (anonportmin >= anonportmax || anonportmin < 0 1748 || anonportmin > 65535 1749#ifndef IPNOPRIVPORTS 1750 || anonportmin < IPPORT_RESERVED 1751#endif 1752 ) { 1753 anonportmin = old; 1754 return (EINVAL); 1755 } 1756 return (error); 1757 case IPCTL_ANONPORTMAX: 1758 old = anonportmax; 1759 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1760 if (anonportmin >= anonportmax || anonportmax < 0 1761 || anonportmax > 65535 1762#ifndef IPNOPRIVPORTS 1763 || anonportmax < IPPORT_RESERVED 1764#endif 1765 ) { 1766 anonportmax = old; 1767 return (EINVAL); 1768 } 1769 return (error); 1770 case IPCTL_MTUDISCTIMEOUT: 1771 error = sysctl_int(oldp, oldlenp, newp, newlen, 1772 &ip_mtudisc_timeout); 1773 if (ip_mtudisc_timeout_q != NULL) 1774 rt_timer_queue_change(ip_mtudisc_timeout_q, 1775 ip_mtudisc_timeout); 1776 return (error); 1777#ifdef GATEWAY 1778 case IPCTL_MAXFLOWS: 1779 { 1780 int s; 1781 1782 error = sysctl_int(oldp, oldlenp, newp, newlen, 1783 &ip_maxflows); 1784 s = splsoftnet(); 1785 ipflow_reap(0); 1786 splx(s); 1787 return (error); 1788 } 1789#endif 1790 case IPCTL_HOSTZEROBROADCAST: 1791 return (sysctl_int(oldp, oldlenp, newp, newlen, 1792 &hostzeroisbroadcast)); 1793#if NGIF > 0 1794 case IPCTL_GIF_TTL: 1795 return(sysctl_int(oldp, oldlenp, newp, newlen, 1796 &ip_gif_ttl)); 1797#endif 1798 1799#ifndef IPNOPRIVPORTS 1800 case IPCTL_LOWPORTMIN: 1801 old = lowportmin; 1802 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmin); 1803 if (lowportmin >= lowportmax 1804 || lowportmin > IPPORT_RESERVEDMAX 1805 || lowportmin < IPPORT_RESERVEDMIN 1806 ) { 1807 lowportmin = old; 1808 return (EINVAL); 1809 } 1810 return (error); 1811 case IPCTL_LOWPORTMAX: 1812 old = lowportmax; 1813 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmax); 1814 if (lowportmin >= lowportmax 1815 || lowportmax > IPPORT_RESERVEDMAX 1816 || lowportmax < IPPORT_RESERVEDMIN 1817 ) { 1818 lowportmax = old; 1819 return (EINVAL); 1820 } 1821 return (error); 1822#endif 1823 1824 case IPCTL_MAXFRAGPACKETS: 1825 return (sysctl_int(oldp, oldlenp, newp, newlen, 1826 &ip_maxfragpackets)); 1827 1828 default: 1829 return (EOPNOTSUPP); 1830 } 1831 /* NOTREACHED */ 1832} 1833