ip_input.c revision 1.121
1/* $NetBSD: ip_input.c,v 1.121 2000/11/11 00:52:38 thorpej 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; 214 215static __inline int ipq_lock_try __P((void)); 216static __inline void ipq_unlock __P((void)); 217 218static __inline int 219ipq_lock_try() 220{ 221 int s; 222 223 s = splimp(); 224 if (ipq_locked) { 225 splx(s); 226 return (0); 227 } 228 ipq_locked = 1; 229 splx(s); 230 return (1); 231} 232 233static __inline void 234ipq_unlock() 235{ 236 int s; 237 238 s = splimp(); 239 ipq_locked = 0; 240 splx(s); 241} 242 243#ifdef DIAGNOSTIC 244#define IPQ_LOCK() \ 245do { \ 246 if (ipq_lock_try() == 0) { \ 247 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 248 panic("ipq_lock"); \ 249 } \ 250} while (0) 251#define IPQ_LOCK_CHECK() \ 252do { \ 253 if (ipq_locked == 0) { \ 254 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 255 panic("ipq lock check"); \ 256 } \ 257} while (0) 258#else 259#define IPQ_LOCK() (void) ipq_lock_try() 260#define IPQ_LOCK_CHECK() /* nothing */ 261#endif 262 263#define IPQ_UNLOCK() ipq_unlock() 264 265struct pool ipqent_pool; 266 267/* 268 * We need to save the IP options in case a protocol wants to respond 269 * to an incoming packet over the same route if the packet got here 270 * using IP source routing. This allows connection establishment and 271 * maintenance when the remote end is on a network that is not known 272 * to us. 273 */ 274int ip_nhops = 0; 275static struct ip_srcrt { 276 struct in_addr dst; /* final destination */ 277 char nop; /* one NOP to align */ 278 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 279 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 280} ip_srcrt; 281 282static void save_rte __P((u_char *, struct in_addr)); 283 284/* 285 * IP initialization: fill in IP protocol switch table. 286 * All protocols not implemented in kernel go to raw IP protocol handler. 287 */ 288void 289ip_init() 290{ 291 struct protosw *pr; 292 int i; 293 294 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 295 0, NULL, NULL, M_IPQ); 296 297 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 298 if (pr == 0) 299 panic("ip_init"); 300 for (i = 0; i < IPPROTO_MAX; i++) 301 ip_protox[i] = pr - inetsw; 302 for (pr = inetdomain.dom_protosw; 303 pr < inetdomain.dom_protoswNPROTOSW; pr++) 304 if (pr->pr_domain->dom_family == PF_INET && 305 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 306 ip_protox[pr->pr_protocol] = pr - inetsw; 307 LIST_INIT(&ipq); 308 ip_id = time.tv_sec & 0xffff; 309 ipintrq.ifq_maxlen = ipqmaxlen; 310 TAILQ_INIT(&in_ifaddr); 311 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 312 M_WAITOK, &in_ifaddrhash); 313 if (ip_mtudisc != 0) 314 ip_mtudisc_timeout_q = 315 rt_timer_queue_create(ip_mtudisc_timeout); 316#ifdef GATEWAY 317 ipflow_init(); 318#endif 319 320#ifdef PFIL_HOOKS 321 /* Register our Packet Filter hook. */ 322 inet_pfil_hook.ph_key = (void *)(u_long) AF_INET; 323 inet_pfil_hook.ph_dlt = DLT_RAW; 324 i = pfil_head_register(&inet_pfil_hook); 325 if (i != 0) 326 printf("ip_init: WARNING: unable to register pfil hook, " 327 "error %d\n", i); 328#endif /* PFIL_HOOKS */ 329} 330 331struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 332struct route ipforward_rt; 333 334/* 335 * IP software interrupt routine 336 */ 337void 338ipintr() 339{ 340 int s; 341 struct mbuf *m; 342 343 while (1) { 344 s = splimp(); 345 IF_DEQUEUE(&ipintrq, m); 346 splx(s); 347 if (m == 0) 348 return; 349 ip_input(m); 350 } 351} 352 353/* 354 * Ip input routine. Checksum and byte swap header. If fragmented 355 * try to reassemble. Process options. Pass to next level. 356 */ 357void 358ip_input(struct mbuf *m) 359{ 360 struct ip *ip = NULL; 361 struct ipq *fp; 362 struct in_ifaddr *ia; 363 struct ifaddr *ifa; 364 struct ipqent *ipqe; 365 int hlen = 0, mff, len; 366 int downmatch; 367 368#ifdef DIAGNOSTIC 369 if ((m->m_flags & M_PKTHDR) == 0) 370 panic("ipintr no HDR"); 371#endif 372#ifdef IPSEC 373 /* 374 * should the inner packet be considered authentic? 375 * see comment in ah4_input(). 376 */ 377 if (m) { 378 m->m_flags &= ~M_AUTHIPHDR; 379 m->m_flags &= ~M_AUTHIPDGM; 380 } 381#endif 382 /* 383 * If no IP addresses have been set yet but the interfaces 384 * are receiving, can't do anything with incoming packets yet. 385 */ 386 if (in_ifaddr.tqh_first == 0) 387 goto bad; 388 ipstat.ips_total++; 389 if (m->m_len < sizeof (struct ip) && 390 (m = m_pullup(m, sizeof (struct ip))) == 0) { 391 ipstat.ips_toosmall++; 392 return; 393 } 394 ip = mtod(m, struct ip *); 395 if (ip->ip_v != IPVERSION) { 396 ipstat.ips_badvers++; 397 goto bad; 398 } 399 hlen = ip->ip_hl << 2; 400 if (hlen < sizeof(struct ip)) { /* minimum header length */ 401 ipstat.ips_badhlen++; 402 goto bad; 403 } 404 if (hlen > m->m_len) { 405 if ((m = m_pullup(m, hlen)) == 0) { 406 ipstat.ips_badhlen++; 407 return; 408 } 409 ip = mtod(m, struct ip *); 410 } 411 412 /* 413 * RFC1122: packets with a multicast source address are 414 * not allowed. 415 */ 416 if (IN_MULTICAST(ip->ip_src.s_addr)) { 417 /* XXX stat */ 418 goto bad; 419 } 420 421 if (in_cksum(m, hlen) != 0) { 422 ipstat.ips_badsum++; 423 goto bad; 424 } 425 426 /* Retrieve the packet length. */ 427 len = ntohs(ip->ip_len); 428 429 /* 430 * Check for additional length bogosity 431 */ 432 if (len < hlen) { 433 ipstat.ips_badlen++; 434 goto bad; 435 } 436 437 /* 438 * Check that the amount of data in the buffers 439 * is as at least much as the IP header would have us expect. 440 * Trim mbufs if longer than we expect. 441 * Drop packet if shorter than we expect. 442 */ 443 if (m->m_pkthdr.len < len) { 444 ipstat.ips_tooshort++; 445 goto bad; 446 } 447 if (m->m_pkthdr.len > len) { 448 if (m->m_len == m->m_pkthdr.len) { 449 m->m_len = len; 450 m->m_pkthdr.len = len; 451 } else 452 m_adj(m, len - m->m_pkthdr.len); 453 } 454 455#ifdef IPSEC 456 /* ipflow (IP fast fowarding) is not compatible with IPsec. */ 457 m->m_flags &= ~M_CANFASTFWD; 458#else 459 /* 460 * Assume that we can create a fast-forward IP flow entry 461 * based on this packet. 462 */ 463 m->m_flags |= M_CANFASTFWD; 464#endif 465 466#ifdef PFIL_HOOKS 467 /* 468 * Run through list of hooks for input packets. If there are any 469 * filters which require that additional packets in the flow are 470 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 471 * Note that filters must _never_ set this flag, as another filter 472 * in the list may have previously cleared it. 473 */ 474 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 475 PFIL_IN) != 0) 476 return; 477 if (m == NULL) 478 return; 479 ip = mtod(m, struct ip *); 480#endif /* PFIL_HOOKS */ 481 482 /* 483 * Convert fields to host representation. 484 */ 485 NTOHS(ip->ip_len); 486 NTOHS(ip->ip_off); 487 488 /* 489 * Process options and, if not destined for us, 490 * ship it on. ip_dooptions returns 1 when an 491 * error was detected (causing an icmp message 492 * to be sent and the original packet to be freed). 493 */ 494 ip_nhops = 0; /* for source routed packets */ 495 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 496 return; 497 498 /* 499 * Check our list of addresses, to see if the packet is for us. 500 * 501 * Traditional 4.4BSD did not consult IFF_UP at all. 502 * The behavior here is to treat addresses on !IFF_UP interface 503 * as not mine. 504 */ 505 downmatch = 0; 506 for (ia = IN_IFADDR_HASH(ip->ip_dst.s_addr).lh_first; 507 ia != NULL; 508 ia = ia->ia_hash.le_next) { 509 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 510 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 511 break; 512 else 513 downmatch++; 514 } 515 } 516 if (ia != NULL) 517 goto ours; 518 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 519 for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first; 520 ifa != NULL; ifa = ifa->ifa_list.tqe_next) { 521 if (ifa->ifa_addr->sa_family != AF_INET) continue; 522 ia = ifatoia(ifa); 523 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 524 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 525 /* 526 * Look for all-0's host part (old broadcast addr), 527 * either for subnet or net. 528 */ 529 ip->ip_dst.s_addr == ia->ia_subnet || 530 ip->ip_dst.s_addr == ia->ia_net) 531 goto ours; 532 /* 533 * An interface with IP address zero accepts 534 * all packets that arrive on that interface. 535 */ 536 if (in_nullhost(ia->ia_addr.sin_addr)) 537 goto ours; 538 } 539 } 540 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 541 struct in_multi *inm; 542#ifdef MROUTING 543 extern struct socket *ip_mrouter; 544 545 if (m->m_flags & M_EXT) { 546 if ((m = m_pullup(m, hlen)) == 0) { 547 ipstat.ips_toosmall++; 548 return; 549 } 550 ip = mtod(m, struct ip *); 551 } 552 553 if (ip_mrouter) { 554 /* 555 * If we are acting as a multicast router, all 556 * incoming multicast packets are passed to the 557 * kernel-level multicast forwarding function. 558 * The packet is returned (relatively) intact; if 559 * ip_mforward() returns a non-zero value, the packet 560 * must be discarded, else it may be accepted below. 561 * 562 * (The IP ident field is put in the same byte order 563 * as expected when ip_mforward() is called from 564 * ip_output().) 565 */ 566 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 567 ipstat.ips_cantforward++; 568 m_freem(m); 569 return; 570 } 571 572 /* 573 * The process-level routing demon needs to receive 574 * all multicast IGMP packets, whether or not this 575 * host belongs to their destination groups. 576 */ 577 if (ip->ip_p == IPPROTO_IGMP) 578 goto ours; 579 ipstat.ips_forward++; 580 } 581#endif 582 /* 583 * See if we belong to the destination multicast group on the 584 * arrival interface. 585 */ 586 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 587 if (inm == NULL) { 588 ipstat.ips_cantforward++; 589 m_freem(m); 590 return; 591 } 592 goto ours; 593 } 594 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 595 in_nullhost(ip->ip_dst)) 596 goto ours; 597 598 /* 599 * Not for us; forward if possible and desirable. 600 */ 601 if (ipforwarding == 0) { 602 ipstat.ips_cantforward++; 603 m_freem(m); 604 } else { 605 /* 606 * If ip_dst matched any of my address on !IFF_UP interface, 607 * and there's no IFF_UP interface that matches ip_dst, 608 * send icmp unreach. Forwarding it will result in in-kernel 609 * forwarding loop till TTL goes to 0. 610 */ 611 if (downmatch) { 612 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 613 ipstat.ips_cantforward++; 614 return; 615 } 616 ip_forward(m, 0); 617 } 618 return; 619 620ours: 621 /* 622 * If offset or IP_MF are set, must reassemble. 623 * Otherwise, nothing need be done. 624 * (We could look in the reassembly queue to see 625 * if the packet was previously fragmented, 626 * but it's not worth the time; just let them time out.) 627 */ 628 if (ip->ip_off & ~(IP_DF|IP_RF)) { 629 /* 630 * Look for queue of fragments 631 * of this datagram. 632 */ 633 IPQ_LOCK(); 634 for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next) 635 if (ip->ip_id == fp->ipq_id && 636 in_hosteq(ip->ip_src, fp->ipq_src) && 637 in_hosteq(ip->ip_dst, fp->ipq_dst) && 638 ip->ip_p == fp->ipq_p) 639 goto found; 640 fp = 0; 641found: 642 643 /* 644 * Adjust ip_len to not reflect header, 645 * set ipqe_mff if more fragments are expected, 646 * convert offset of this to bytes. 647 */ 648 ip->ip_len -= hlen; 649 mff = (ip->ip_off & IP_MF) != 0; 650 if (mff) { 651 /* 652 * Make sure that fragments have a data length 653 * that's a non-zero multiple of 8 bytes. 654 */ 655 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 656 ipstat.ips_badfrags++; 657 IPQ_UNLOCK(); 658 goto bad; 659 } 660 } 661 ip->ip_off <<= 3; 662 663 /* 664 * If datagram marked as having more fragments 665 * or if this is not the first fragment, 666 * attempt reassembly; if it succeeds, proceed. 667 */ 668 if (mff || ip->ip_off) { 669 ipstat.ips_fragments++; 670 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 671 if (ipqe == NULL) { 672 ipstat.ips_rcvmemdrop++; 673 IPQ_UNLOCK(); 674 goto bad; 675 } 676 ipqe->ipqe_mff = mff; 677 ipqe->ipqe_m = m; 678 ipqe->ipqe_ip = ip; 679 m = ip_reass(ipqe, fp); 680 if (m == 0) { 681 IPQ_UNLOCK(); 682 return; 683 } 684 ipstat.ips_reassembled++; 685 ip = mtod(m, struct ip *); 686 hlen = ip->ip_hl << 2; 687 ip->ip_len += hlen; 688 } else 689 if (fp) 690 ip_freef(fp); 691 IPQ_UNLOCK(); 692 } 693 694 /* 695 * Switch out to protocol's input routine. 696 */ 697#if IFA_STATS 698 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 699#endif 700 ipstat.ips_delivered++; 701 { 702 int off = hlen, nh = ip->ip_p; 703 704 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 705 return; 706 } 707bad: 708 m_freem(m); 709} 710 711/* 712 * Take incoming datagram fragment and try to 713 * reassemble it into whole datagram. If a chain for 714 * reassembly of this datagram already exists, then it 715 * is given as fp; otherwise have to make a chain. 716 */ 717struct mbuf * 718ip_reass(ipqe, fp) 719 struct ipqent *ipqe; 720 struct ipq *fp; 721{ 722 struct mbuf *m = ipqe->ipqe_m; 723 struct ipqent *nq, *p, *q; 724 struct ip *ip; 725 struct mbuf *t; 726 int hlen = ipqe->ipqe_ip->ip_hl << 2; 727 int i, next; 728 729 IPQ_LOCK_CHECK(); 730 731 /* 732 * Presence of header sizes in mbufs 733 * would confuse code below. 734 */ 735 m->m_data += hlen; 736 m->m_len -= hlen; 737 738 /* 739 * If first fragment to arrive, create a reassembly queue. 740 */ 741 if (fp == 0) { 742 MALLOC(fp, struct ipq *, sizeof (struct ipq), 743 M_FTABLE, M_NOWAIT); 744 if (fp == NULL) 745 goto dropfrag; 746 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 747 fp->ipq_ttl = IPFRAGTTL; 748 fp->ipq_p = ipqe->ipqe_ip->ip_p; 749 fp->ipq_id = ipqe->ipqe_ip->ip_id; 750 LIST_INIT(&fp->ipq_fragq); 751 fp->ipq_src = ipqe->ipqe_ip->ip_src; 752 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 753 p = NULL; 754 goto insert; 755 } 756 757 /* 758 * Find a segment which begins after this one does. 759 */ 760 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 761 p = q, q = q->ipqe_q.le_next) 762 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 763 break; 764 765 /* 766 * If there is a preceding segment, it may provide some of 767 * our data already. If so, drop the data from the incoming 768 * segment. If it provides all of our data, drop us. 769 */ 770 if (p != NULL) { 771 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 772 ipqe->ipqe_ip->ip_off; 773 if (i > 0) { 774 if (i >= ipqe->ipqe_ip->ip_len) 775 goto dropfrag; 776 m_adj(ipqe->ipqe_m, i); 777 ipqe->ipqe_ip->ip_off += i; 778 ipqe->ipqe_ip->ip_len -= i; 779 } 780 } 781 782 /* 783 * While we overlap succeeding segments trim them or, 784 * if they are completely covered, dequeue them. 785 */ 786 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 787 q->ipqe_ip->ip_off; q = nq) { 788 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 789 q->ipqe_ip->ip_off; 790 if (i < q->ipqe_ip->ip_len) { 791 q->ipqe_ip->ip_len -= i; 792 q->ipqe_ip->ip_off += i; 793 m_adj(q->ipqe_m, i); 794 break; 795 } 796 nq = q->ipqe_q.le_next; 797 m_freem(q->ipqe_m); 798 LIST_REMOVE(q, ipqe_q); 799 pool_put(&ipqent_pool, q); 800 } 801 802insert: 803 /* 804 * Stick new segment in its place; 805 * check for complete reassembly. 806 */ 807 if (p == NULL) { 808 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 809 } else { 810 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 811 } 812 next = 0; 813 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 814 p = q, q = q->ipqe_q.le_next) { 815 if (q->ipqe_ip->ip_off != next) 816 return (0); 817 next += q->ipqe_ip->ip_len; 818 } 819 if (p->ipqe_mff) 820 return (0); 821 822 /* 823 * Reassembly is complete. Check for a bogus message size and 824 * concatenate fragments. 825 */ 826 q = fp->ipq_fragq.lh_first; 827 ip = q->ipqe_ip; 828 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 829 ipstat.ips_toolong++; 830 ip_freef(fp); 831 return (0); 832 } 833 m = q->ipqe_m; 834 t = m->m_next; 835 m->m_next = 0; 836 m_cat(m, t); 837 nq = q->ipqe_q.le_next; 838 pool_put(&ipqent_pool, q); 839 for (q = nq; q != NULL; q = nq) { 840 t = q->ipqe_m; 841 nq = q->ipqe_q.le_next; 842 pool_put(&ipqent_pool, q); 843 m_cat(m, t); 844 } 845 846 /* 847 * Create header for new ip packet by 848 * modifying header of first packet; 849 * dequeue and discard fragment reassembly header. 850 * Make header visible. 851 */ 852 ip->ip_len = next; 853 ip->ip_src = fp->ipq_src; 854 ip->ip_dst = fp->ipq_dst; 855 LIST_REMOVE(fp, ipq_q); 856 FREE(fp, M_FTABLE); 857 m->m_len += (ip->ip_hl << 2); 858 m->m_data -= (ip->ip_hl << 2); 859 /* some debugging cruft by sklower, below, will go away soon */ 860 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 861 int plen = 0; 862 for (t = m; t; t = t->m_next) 863 plen += t->m_len; 864 m->m_pkthdr.len = plen; 865 } 866 return (m); 867 868dropfrag: 869 ipstat.ips_fragdropped++; 870 m_freem(m); 871 pool_put(&ipqent_pool, ipqe); 872 return (0); 873} 874 875/* 876 * Free a fragment reassembly header and all 877 * associated datagrams. 878 */ 879void 880ip_freef(fp) 881 struct ipq *fp; 882{ 883 struct ipqent *q, *p; 884 885 IPQ_LOCK_CHECK(); 886 887 for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { 888 p = q->ipqe_q.le_next; 889 m_freem(q->ipqe_m); 890 LIST_REMOVE(q, ipqe_q); 891 pool_put(&ipqent_pool, q); 892 } 893 LIST_REMOVE(fp, ipq_q); 894 FREE(fp, M_FTABLE); 895} 896 897/* 898 * IP timer processing; 899 * if a timer expires on a reassembly 900 * queue, discard it. 901 */ 902void 903ip_slowtimo() 904{ 905 struct ipq *fp, *nfp; 906 int s = splsoftnet(); 907 908 IPQ_LOCK(); 909 for (fp = ipq.lh_first; fp != NULL; fp = nfp) { 910 nfp = fp->ipq_q.le_next; 911 if (--fp->ipq_ttl == 0) { 912 ipstat.ips_fragtimeout++; 913 ip_freef(fp); 914 } 915 } 916 IPQ_UNLOCK(); 917#ifdef GATEWAY 918 ipflow_slowtimo(); 919#endif 920 splx(s); 921} 922 923/* 924 * Drain off all datagram fragments. 925 */ 926void 927ip_drain() 928{ 929 930 /* 931 * We may be called from a device's interrupt context. If 932 * the ipq is already busy, just bail out now. 933 */ 934 if (ipq_lock_try() == 0) 935 return; 936 937 while (ipq.lh_first != NULL) { 938 ipstat.ips_fragdropped++; 939 ip_freef(ipq.lh_first); 940 } 941 942 IPQ_UNLOCK(); 943} 944 945/* 946 * Do option processing on a datagram, 947 * possibly discarding it if bad options are encountered, 948 * or forwarding it if source-routed. 949 * Returns 1 if packet has been forwarded/freed, 950 * 0 if the packet should be processed further. 951 */ 952int 953ip_dooptions(m) 954 struct mbuf *m; 955{ 956 struct ip *ip = mtod(m, struct ip *); 957 u_char *cp, *cp0; 958 struct ip_timestamp *ipt; 959 struct in_ifaddr *ia; 960 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 961 struct in_addr dst; 962 n_time ntime; 963 964 dst = ip->ip_dst; 965 cp = (u_char *)(ip + 1); 966 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 967 for (; cnt > 0; cnt -= optlen, cp += optlen) { 968 opt = cp[IPOPT_OPTVAL]; 969 if (opt == IPOPT_EOL) 970 break; 971 if (opt == IPOPT_NOP) 972 optlen = 1; 973 else { 974 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 975 code = &cp[IPOPT_OLEN] - (u_char *)ip; 976 goto bad; 977 } 978 optlen = cp[IPOPT_OLEN]; 979 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 980 code = &cp[IPOPT_OLEN] - (u_char *)ip; 981 goto bad; 982 } 983 } 984 switch (opt) { 985 986 default: 987 break; 988 989 /* 990 * Source routing with record. 991 * Find interface with current destination address. 992 * If none on this machine then drop if strictly routed, 993 * or do nothing if loosely routed. 994 * Record interface address and bring up next address 995 * component. If strictly routed make sure next 996 * address is on directly accessible net. 997 */ 998 case IPOPT_LSRR: 999 case IPOPT_SSRR: 1000 if (ip_allowsrcrt == 0) { 1001 type = ICMP_UNREACH; 1002 code = ICMP_UNREACH_NET_PROHIB; 1003 goto bad; 1004 } 1005 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1006 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1007 goto bad; 1008 } 1009 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1010 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1011 goto bad; 1012 } 1013 ipaddr.sin_addr = ip->ip_dst; 1014 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1015 if (ia == 0) { 1016 if (opt == IPOPT_SSRR) { 1017 type = ICMP_UNREACH; 1018 code = ICMP_UNREACH_SRCFAIL; 1019 goto bad; 1020 } 1021 /* 1022 * Loose routing, and not at next destination 1023 * yet; nothing to do except forward. 1024 */ 1025 break; 1026 } 1027 off--; /* 0 origin */ 1028 if ((off + sizeof(struct in_addr)) > optlen) { 1029 /* 1030 * End of source route. Should be for us. 1031 */ 1032 save_rte(cp, ip->ip_src); 1033 break; 1034 } 1035 /* 1036 * locate outgoing interface 1037 */ 1038 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1039 sizeof(ipaddr.sin_addr)); 1040 if (opt == IPOPT_SSRR) 1041 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1042 else 1043 ia = ip_rtaddr(ipaddr.sin_addr); 1044 if (ia == 0) { 1045 type = ICMP_UNREACH; 1046 code = ICMP_UNREACH_SRCFAIL; 1047 goto bad; 1048 } 1049 ip->ip_dst = ipaddr.sin_addr; 1050 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1051 (caddr_t)(cp + off), sizeof(struct in_addr)); 1052 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1053 /* 1054 * Let ip_intr's mcast routing check handle mcast pkts 1055 */ 1056 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1057 break; 1058 1059 case IPOPT_RR: 1060 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1061 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1062 goto bad; 1063 } 1064 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1065 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1066 goto bad; 1067 } 1068 /* 1069 * If no space remains, ignore. 1070 */ 1071 off--; /* 0 origin */ 1072 if ((off + sizeof(struct in_addr)) > optlen) 1073 break; 1074 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1075 sizeof(ipaddr.sin_addr)); 1076 /* 1077 * locate outgoing interface; if we're the destination, 1078 * use the incoming interface (should be same). 1079 */ 1080 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1081 == NULL && 1082 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1083 type = ICMP_UNREACH; 1084 code = ICMP_UNREACH_HOST; 1085 goto bad; 1086 } 1087 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1088 (caddr_t)(cp + off), sizeof(struct in_addr)); 1089 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1090 break; 1091 1092 case IPOPT_TS: 1093 code = cp - (u_char *)ip; 1094 ipt = (struct ip_timestamp *)cp; 1095 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1096 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1097 goto bad; 1098 } 1099 if (ipt->ipt_ptr < 5) { 1100 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1101 goto bad; 1102 } 1103 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1104 if (++ipt->ipt_oflw == 0) { 1105 code = (u_char *)&ipt->ipt_ptr - 1106 (u_char *)ip; 1107 goto bad; 1108 } 1109 break; 1110 } 1111 cp0 = (cp + ipt->ipt_ptr - 1); 1112 switch (ipt->ipt_flg) { 1113 1114 case IPOPT_TS_TSONLY: 1115 break; 1116 1117 case IPOPT_TS_TSANDADDR: 1118 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1119 sizeof(struct in_addr) > ipt->ipt_len) { 1120 code = (u_char *)&ipt->ipt_ptr - 1121 (u_char *)ip; 1122 goto bad; 1123 } 1124 ipaddr.sin_addr = dst; 1125 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1126 m->m_pkthdr.rcvif)); 1127 if (ia == 0) 1128 continue; 1129 bcopy(&ia->ia_addr.sin_addr, 1130 cp0, sizeof(struct in_addr)); 1131 ipt->ipt_ptr += sizeof(struct in_addr); 1132 break; 1133 1134 case IPOPT_TS_PRESPEC: 1135 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1136 sizeof(struct in_addr) > ipt->ipt_len) { 1137 code = (u_char *)&ipt->ipt_ptr - 1138 (u_char *)ip; 1139 goto bad; 1140 } 1141 bcopy(cp0, &ipaddr.sin_addr, 1142 sizeof(struct in_addr)); 1143 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1144 == NULL) 1145 continue; 1146 ipt->ipt_ptr += sizeof(struct in_addr); 1147 break; 1148 1149 default: 1150 /* XXX can't take &ipt->ipt_flg */ 1151 code = (u_char *)&ipt->ipt_ptr - 1152 (u_char *)ip + 1; 1153 goto bad; 1154 } 1155 ntime = iptime(); 1156 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1157 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1158 sizeof(n_time)); 1159 ipt->ipt_ptr += sizeof(n_time); 1160 } 1161 } 1162 if (forward) { 1163 if (ip_forwsrcrt == 0) { 1164 type = ICMP_UNREACH; 1165 code = ICMP_UNREACH_SRCFAIL; 1166 goto bad; 1167 } 1168 ip_forward(m, 1); 1169 return (1); 1170 } 1171 return (0); 1172bad: 1173 icmp_error(m, type, code, 0, 0); 1174 ipstat.ips_badoptions++; 1175 return (1); 1176} 1177 1178/* 1179 * Given address of next destination (final or next hop), 1180 * return internet address info of interface to be used to get there. 1181 */ 1182struct in_ifaddr * 1183ip_rtaddr(dst) 1184 struct in_addr dst; 1185{ 1186 struct sockaddr_in *sin; 1187 1188 sin = satosin(&ipforward_rt.ro_dst); 1189 1190 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1191 if (ipforward_rt.ro_rt) { 1192 RTFREE(ipforward_rt.ro_rt); 1193 ipforward_rt.ro_rt = 0; 1194 } 1195 sin->sin_family = AF_INET; 1196 sin->sin_len = sizeof(*sin); 1197 sin->sin_addr = dst; 1198 1199 rtalloc(&ipforward_rt); 1200 } 1201 if (ipforward_rt.ro_rt == 0) 1202 return ((struct in_ifaddr *)0); 1203 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1204} 1205 1206/* 1207 * Save incoming source route for use in replies, 1208 * to be picked up later by ip_srcroute if the receiver is interested. 1209 */ 1210void 1211save_rte(option, dst) 1212 u_char *option; 1213 struct in_addr dst; 1214{ 1215 unsigned olen; 1216 1217 olen = option[IPOPT_OLEN]; 1218#ifdef DIAGNOSTIC 1219 if (ipprintfs) 1220 printf("save_rte: olen %d\n", olen); 1221#endif /* 0 */ 1222 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1223 return; 1224 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1225 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1226 ip_srcrt.dst = dst; 1227} 1228 1229/* 1230 * Retrieve incoming source route for use in replies, 1231 * in the same form used by setsockopt. 1232 * The first hop is placed before the options, will be removed later. 1233 */ 1234struct mbuf * 1235ip_srcroute() 1236{ 1237 struct in_addr *p, *q; 1238 struct mbuf *m; 1239 1240 if (ip_nhops == 0) 1241 return ((struct mbuf *)0); 1242 m = m_get(M_DONTWAIT, MT_SOOPTS); 1243 if (m == 0) 1244 return ((struct mbuf *)0); 1245 1246#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1247 1248 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1249 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1250 OPTSIZ; 1251#ifdef DIAGNOSTIC 1252 if (ipprintfs) 1253 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1254#endif 1255 1256 /* 1257 * First save first hop for return route 1258 */ 1259 p = &ip_srcrt.route[ip_nhops - 1]; 1260 *(mtod(m, struct in_addr *)) = *p--; 1261#ifdef DIAGNOSTIC 1262 if (ipprintfs) 1263 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1264#endif 1265 1266 /* 1267 * Copy option fields and padding (nop) to mbuf. 1268 */ 1269 ip_srcrt.nop = IPOPT_NOP; 1270 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1271 bcopy((caddr_t)&ip_srcrt.nop, 1272 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1273 q = (struct in_addr *)(mtod(m, caddr_t) + 1274 sizeof(struct in_addr) + OPTSIZ); 1275#undef OPTSIZ 1276 /* 1277 * Record return path as an IP source route, 1278 * reversing the path (pointers are now aligned). 1279 */ 1280 while (p >= ip_srcrt.route) { 1281#ifdef DIAGNOSTIC 1282 if (ipprintfs) 1283 printf(" %x", ntohl(q->s_addr)); 1284#endif 1285 *q++ = *p--; 1286 } 1287 /* 1288 * Last hop goes to final destination. 1289 */ 1290 *q = ip_srcrt.dst; 1291#ifdef DIAGNOSTIC 1292 if (ipprintfs) 1293 printf(" %x\n", ntohl(q->s_addr)); 1294#endif 1295 return (m); 1296} 1297 1298/* 1299 * Strip out IP options, at higher 1300 * level protocol in the kernel. 1301 * Second argument is buffer to which options 1302 * will be moved, and return value is their length. 1303 * XXX should be deleted; last arg currently ignored. 1304 */ 1305void 1306ip_stripoptions(m, mopt) 1307 struct mbuf *m; 1308 struct mbuf *mopt; 1309{ 1310 int i; 1311 struct ip *ip = mtod(m, struct ip *); 1312 caddr_t opts; 1313 int olen; 1314 1315 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1316 opts = (caddr_t)(ip + 1); 1317 i = m->m_len - (sizeof (struct ip) + olen); 1318 bcopy(opts + olen, opts, (unsigned)i); 1319 m->m_len -= olen; 1320 if (m->m_flags & M_PKTHDR) 1321 m->m_pkthdr.len -= olen; 1322 ip->ip_len -= olen; 1323 ip->ip_hl = sizeof (struct ip) >> 2; 1324} 1325 1326int inetctlerrmap[PRC_NCMDS] = { 1327 0, 0, 0, 0, 1328 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1329 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1330 EMSGSIZE, EHOSTUNREACH, 0, 0, 1331 0, 0, 0, 0, 1332 ENOPROTOOPT 1333}; 1334 1335/* 1336 * Forward a packet. If some error occurs return the sender 1337 * an icmp packet. Note we can't always generate a meaningful 1338 * icmp message because icmp doesn't have a large enough repertoire 1339 * of codes and types. 1340 * 1341 * If not forwarding, just drop the packet. This could be confusing 1342 * if ipforwarding was zero but some routing protocol was advancing 1343 * us as a gateway to somewhere. However, we must let the routing 1344 * protocol deal with that. 1345 * 1346 * The srcrt parameter indicates whether the packet is being forwarded 1347 * via a source route. 1348 */ 1349void 1350ip_forward(m, srcrt) 1351 struct mbuf *m; 1352 int srcrt; 1353{ 1354 struct ip *ip = mtod(m, struct ip *); 1355 struct sockaddr_in *sin; 1356 struct rtentry *rt; 1357 int error, type = 0, code = 0; 1358 struct mbuf *mcopy; 1359 n_long dest; 1360 struct ifnet *destifp; 1361#ifdef IPSEC 1362 struct ifnet dummyifp; 1363#endif 1364 1365 dest = 0; 1366#ifdef DIAGNOSTIC 1367 if (ipprintfs) 1368 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1369 ntohl(ip->ip_src.s_addr), 1370 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1371#endif 1372 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1373 ipstat.ips_cantforward++; 1374 m_freem(m); 1375 return; 1376 } 1377 if (ip->ip_ttl <= IPTTLDEC) { 1378 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1379 return; 1380 } 1381 ip->ip_ttl -= IPTTLDEC; 1382 1383 sin = satosin(&ipforward_rt.ro_dst); 1384 if ((rt = ipforward_rt.ro_rt) == 0 || 1385 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1386 if (ipforward_rt.ro_rt) { 1387 RTFREE(ipforward_rt.ro_rt); 1388 ipforward_rt.ro_rt = 0; 1389 } 1390 sin->sin_family = AF_INET; 1391 sin->sin_len = sizeof(struct sockaddr_in); 1392 sin->sin_addr = ip->ip_dst; 1393 1394 rtalloc(&ipforward_rt); 1395 if (ipforward_rt.ro_rt == 0) { 1396 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1397 return; 1398 } 1399 rt = ipforward_rt.ro_rt; 1400 } 1401 1402 /* 1403 * Save at most 68 bytes of the packet in case 1404 * we need to generate an ICMP message to the src. 1405 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1406 */ 1407 mcopy = m_copym(m, 0, imin((int)ip->ip_len, 68), M_DONTWAIT); 1408 if (mcopy) 1409 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1410 1411 /* 1412 * If forwarding packet using same interface that it came in on, 1413 * perhaps should send a redirect to sender to shortcut a hop. 1414 * Only send redirect if source is sending directly to us, 1415 * and if packet was not source routed (or has any options). 1416 * Also, don't send redirect if forwarding using a default route 1417 * or a route modified by a redirect. 1418 */ 1419 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1420 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1421 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1422 ipsendredirects && !srcrt) { 1423 if (rt->rt_ifa && 1424 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1425 ifatoia(rt->rt_ifa)->ia_subnet) { 1426 if (rt->rt_flags & RTF_GATEWAY) 1427 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1428 else 1429 dest = ip->ip_dst.s_addr; 1430 /* 1431 * Router requirements says to only send host 1432 * redirects. 1433 */ 1434 type = ICMP_REDIRECT; 1435 code = ICMP_REDIRECT_HOST; 1436#ifdef DIAGNOSTIC 1437 if (ipprintfs) 1438 printf("redirect (%d) to %x\n", code, 1439 (u_int32_t)dest); 1440#endif 1441 } 1442 } 1443 1444#ifdef IPSEC 1445 /* Don't lookup socket in forwading case */ 1446 ipsec_setsocket(m, NULL); 1447#endif 1448 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1449 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1450 if (error) 1451 ipstat.ips_cantforward++; 1452 else { 1453 ipstat.ips_forward++; 1454 if (type) 1455 ipstat.ips_redirectsent++; 1456 else { 1457 if (mcopy) { 1458#ifdef GATEWAY 1459 if (mcopy->m_flags & M_CANFASTFWD) 1460 ipflow_create(&ipforward_rt, mcopy); 1461#endif 1462 m_freem(mcopy); 1463 } 1464 return; 1465 } 1466 } 1467 if (mcopy == NULL) 1468 return; 1469 destifp = NULL; 1470 1471 switch (error) { 1472 1473 case 0: /* forwarded, but need redirect */ 1474 /* type, code set above */ 1475 break; 1476 1477 case ENETUNREACH: /* shouldn't happen, checked above */ 1478 case EHOSTUNREACH: 1479 case ENETDOWN: 1480 case EHOSTDOWN: 1481 default: 1482 type = ICMP_UNREACH; 1483 code = ICMP_UNREACH_HOST; 1484 break; 1485 1486 case EMSGSIZE: 1487 type = ICMP_UNREACH; 1488 code = ICMP_UNREACH_NEEDFRAG; 1489#ifndef IPSEC 1490 if (ipforward_rt.ro_rt) 1491 destifp = ipforward_rt.ro_rt->rt_ifp; 1492#else 1493 /* 1494 * If the packet is routed over IPsec tunnel, tell the 1495 * originator the tunnel MTU. 1496 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1497 * XXX quickhack!!! 1498 */ 1499 if (ipforward_rt.ro_rt) { 1500 struct secpolicy *sp; 1501 int ipsecerror; 1502 size_t ipsechdr; 1503 struct route *ro; 1504 1505 sp = ipsec4_getpolicybyaddr(mcopy, 1506 IPSEC_DIR_OUTBOUND, 1507 IP_FORWARDING, 1508 &ipsecerror); 1509 1510 if (sp == NULL) 1511 destifp = ipforward_rt.ro_rt->rt_ifp; 1512 else { 1513 /* count IPsec header size */ 1514 ipsechdr = ipsec4_hdrsiz(mcopy, 1515 IPSEC_DIR_OUTBOUND, 1516 NULL); 1517 1518 /* 1519 * find the correct route for outer IPv4 1520 * header, compute tunnel MTU. 1521 * 1522 * XXX BUG ALERT 1523 * The "dummyifp" code relies upon the fact 1524 * that icmp_error() touches only ifp->if_mtu. 1525 */ 1526 /*XXX*/ 1527 destifp = NULL; 1528 if (sp->req != NULL 1529 && sp->req->sav != NULL 1530 && sp->req->sav->sah != NULL) { 1531 ro = &sp->req->sav->sah->sa_route; 1532 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1533 dummyifp.if_mtu = 1534 ro->ro_rt->rt_ifp->if_mtu; 1535 dummyifp.if_mtu -= ipsechdr; 1536 destifp = &dummyifp; 1537 } 1538 } 1539 1540 key_freesp(sp); 1541 } 1542 } 1543#endif /*IPSEC*/ 1544 ipstat.ips_cantfrag++; 1545 break; 1546 1547 case ENOBUFS: 1548 type = ICMP_SOURCEQUENCH; 1549 code = 0; 1550 break; 1551 } 1552 icmp_error(mcopy, type, code, dest, destifp); 1553} 1554 1555void 1556ip_savecontrol(inp, mp, ip, m) 1557 struct inpcb *inp; 1558 struct mbuf **mp; 1559 struct ip *ip; 1560 struct mbuf *m; 1561{ 1562 1563 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1564 struct timeval tv; 1565 1566 microtime(&tv); 1567 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1568 SCM_TIMESTAMP, SOL_SOCKET); 1569 if (*mp) 1570 mp = &(*mp)->m_next; 1571 } 1572 if (inp->inp_flags & INP_RECVDSTADDR) { 1573 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1574 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1575 if (*mp) 1576 mp = &(*mp)->m_next; 1577 } 1578#ifdef notyet 1579 /* 1580 * XXX 1581 * Moving these out of udp_input() made them even more broken 1582 * than they already were. 1583 * - fenner@parc.xerox.com 1584 */ 1585 /* options were tossed already */ 1586 if (inp->inp_flags & INP_RECVOPTS) { 1587 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1588 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1589 if (*mp) 1590 mp = &(*mp)->m_next; 1591 } 1592 /* ip_srcroute doesn't do what we want here, need to fix */ 1593 if (inp->inp_flags & INP_RECVRETOPTS) { 1594 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1595 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1596 if (*mp) 1597 mp = &(*mp)->m_next; 1598 } 1599#endif 1600 if (inp->inp_flags & INP_RECVIF) { 1601 struct sockaddr_dl sdl; 1602 1603 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1604 sdl.sdl_family = AF_LINK; 1605 sdl.sdl_index = m->m_pkthdr.rcvif ? 1606 m->m_pkthdr.rcvif->if_index : 0; 1607 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1608 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1609 IP_RECVIF, IPPROTO_IP); 1610 if (*mp) 1611 mp = &(*mp)->m_next; 1612 } 1613} 1614 1615int 1616ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1617 int *name; 1618 u_int namelen; 1619 void *oldp; 1620 size_t *oldlenp; 1621 void *newp; 1622 size_t newlen; 1623{ 1624 extern int subnetsarelocal, hostzeroisbroadcast; 1625 1626 int error, old; 1627 1628 /* All sysctl names at this level are terminal. */ 1629 if (namelen != 1) 1630 return (ENOTDIR); 1631 1632 switch (name[0]) { 1633 case IPCTL_FORWARDING: 1634 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1635 case IPCTL_SENDREDIRECTS: 1636 return (sysctl_int(oldp, oldlenp, newp, newlen, 1637 &ipsendredirects)); 1638 case IPCTL_DEFTTL: 1639 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1640#ifdef notyet 1641 case IPCTL_DEFMTU: 1642 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1643#endif 1644 case IPCTL_FORWSRCRT: 1645 /* Don't allow this to change in a secure environment. */ 1646 if (securelevel > 0) 1647 return (sysctl_rdint(oldp, oldlenp, newp, 1648 ip_forwsrcrt)); 1649 else 1650 return (sysctl_int(oldp, oldlenp, newp, newlen, 1651 &ip_forwsrcrt)); 1652 case IPCTL_DIRECTEDBCAST: 1653 return (sysctl_int(oldp, oldlenp, newp, newlen, 1654 &ip_directedbcast)); 1655 case IPCTL_ALLOWSRCRT: 1656 return (sysctl_int(oldp, oldlenp, newp, newlen, 1657 &ip_allowsrcrt)); 1658 case IPCTL_SUBNETSARELOCAL: 1659 return (sysctl_int(oldp, oldlenp, newp, newlen, 1660 &subnetsarelocal)); 1661 case IPCTL_MTUDISC: 1662 error = sysctl_int(oldp, oldlenp, newp, newlen, 1663 &ip_mtudisc); 1664 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1665 ip_mtudisc_timeout_q = 1666 rt_timer_queue_create(ip_mtudisc_timeout); 1667 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1668 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1669 ip_mtudisc_timeout_q = NULL; 1670 } 1671 return error; 1672 case IPCTL_ANONPORTMIN: 1673 old = anonportmin; 1674 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1675 if (anonportmin >= anonportmax || anonportmin < 0 1676 || anonportmin > 65535 1677#ifndef IPNOPRIVPORTS 1678 || anonportmin < IPPORT_RESERVED 1679#endif 1680 ) { 1681 anonportmin = old; 1682 return (EINVAL); 1683 } 1684 return (error); 1685 case IPCTL_ANONPORTMAX: 1686 old = anonportmax; 1687 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1688 if (anonportmin >= anonportmax || anonportmax < 0 1689 || anonportmax > 65535 1690#ifndef IPNOPRIVPORTS 1691 || anonportmax < IPPORT_RESERVED 1692#endif 1693 ) { 1694 anonportmax = old; 1695 return (EINVAL); 1696 } 1697 return (error); 1698 case IPCTL_MTUDISCTIMEOUT: 1699 error = sysctl_int(oldp, oldlenp, newp, newlen, 1700 &ip_mtudisc_timeout); 1701 if (ip_mtudisc_timeout_q != NULL) 1702 rt_timer_queue_change(ip_mtudisc_timeout_q, 1703 ip_mtudisc_timeout); 1704 return (error); 1705#ifdef GATEWAY 1706 case IPCTL_MAXFLOWS: 1707 { 1708 int s; 1709 1710 error = sysctl_int(oldp, oldlenp, newp, newlen, 1711 &ip_maxflows); 1712 s = splsoftnet(); 1713 ipflow_reap(0); 1714 splx(s); 1715 return (error); 1716 } 1717#endif 1718 case IPCTL_HOSTZEROBROADCAST: 1719 return (sysctl_int(oldp, oldlenp, newp, newlen, 1720 &hostzeroisbroadcast)); 1721#if NGIF > 0 1722 case IPCTL_GIF_TTL: 1723 return(sysctl_int(oldp, oldlenp, newp, newlen, 1724 &ip_gif_ttl)); 1725#endif 1726 1727#ifndef IPNOPRIVPORTS 1728 case IPCTL_LOWPORTMIN: 1729 old = lowportmin; 1730 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmin); 1731 if (lowportmin >= lowportmax 1732 || lowportmin > IPPORT_RESERVEDMAX 1733 || lowportmin < IPPORT_RESERVEDMIN 1734 ) { 1735 lowportmin = old; 1736 return (EINVAL); 1737 } 1738 return (error); 1739 case IPCTL_LOWPORTMAX: 1740 old = lowportmax; 1741 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmax); 1742 if (lowportmin >= lowportmax 1743 || lowportmax > IPPORT_RESERVEDMAX 1744 || lowportmax < IPPORT_RESERVEDMIN 1745 ) { 1746 lowportmax = old; 1747 return (EINVAL); 1748 } 1749 return (error); 1750#endif 1751 1752 default: 1753 return (EOPNOTSUPP); 1754 } 1755 /* NOTREACHED */ 1756} 1757