ip_input.c revision 1.122
1/* $NetBSD: ip_input.c,v 1.122 2000/11/24 03:43:20 itojun 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 if (ia && ip) 699 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 700#endif 701 ipstat.ips_delivered++; 702 { 703 int off = hlen, nh = ip->ip_p; 704 705 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 706 return; 707 } 708bad: 709 m_freem(m); 710} 711 712/* 713 * Take incoming datagram fragment and try to 714 * reassemble it into whole datagram. If a chain for 715 * reassembly of this datagram already exists, then it 716 * is given as fp; otherwise have to make a chain. 717 */ 718struct mbuf * 719ip_reass(ipqe, fp) 720 struct ipqent *ipqe; 721 struct ipq *fp; 722{ 723 struct mbuf *m = ipqe->ipqe_m; 724 struct ipqent *nq, *p, *q; 725 struct ip *ip; 726 struct mbuf *t; 727 int hlen = ipqe->ipqe_ip->ip_hl << 2; 728 int i, next; 729 730 IPQ_LOCK_CHECK(); 731 732 /* 733 * Presence of header sizes in mbufs 734 * would confuse code below. 735 */ 736 m->m_data += hlen; 737 m->m_len -= hlen; 738 739 /* 740 * If first fragment to arrive, create a reassembly queue. 741 */ 742 if (fp == 0) { 743 MALLOC(fp, struct ipq *, sizeof (struct ipq), 744 M_FTABLE, M_NOWAIT); 745 if (fp == NULL) 746 goto dropfrag; 747 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 748 fp->ipq_ttl = IPFRAGTTL; 749 fp->ipq_p = ipqe->ipqe_ip->ip_p; 750 fp->ipq_id = ipqe->ipqe_ip->ip_id; 751 LIST_INIT(&fp->ipq_fragq); 752 fp->ipq_src = ipqe->ipqe_ip->ip_src; 753 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 754 p = NULL; 755 goto insert; 756 } 757 758 /* 759 * Find a segment which begins after this one does. 760 */ 761 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 762 p = q, q = q->ipqe_q.le_next) 763 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 764 break; 765 766 /* 767 * If there is a preceding segment, it may provide some of 768 * our data already. If so, drop the data from the incoming 769 * segment. If it provides all of our data, drop us. 770 */ 771 if (p != NULL) { 772 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 773 ipqe->ipqe_ip->ip_off; 774 if (i > 0) { 775 if (i >= ipqe->ipqe_ip->ip_len) 776 goto dropfrag; 777 m_adj(ipqe->ipqe_m, i); 778 ipqe->ipqe_ip->ip_off += i; 779 ipqe->ipqe_ip->ip_len -= i; 780 } 781 } 782 783 /* 784 * While we overlap succeeding segments trim them or, 785 * if they are completely covered, dequeue them. 786 */ 787 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 788 q->ipqe_ip->ip_off; q = nq) { 789 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 790 q->ipqe_ip->ip_off; 791 if (i < q->ipqe_ip->ip_len) { 792 q->ipqe_ip->ip_len -= i; 793 q->ipqe_ip->ip_off += i; 794 m_adj(q->ipqe_m, i); 795 break; 796 } 797 nq = q->ipqe_q.le_next; 798 m_freem(q->ipqe_m); 799 LIST_REMOVE(q, ipqe_q); 800 pool_put(&ipqent_pool, q); 801 } 802 803insert: 804 /* 805 * Stick new segment in its place; 806 * check for complete reassembly. 807 */ 808 if (p == NULL) { 809 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 810 } else { 811 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 812 } 813 next = 0; 814 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 815 p = q, q = q->ipqe_q.le_next) { 816 if (q->ipqe_ip->ip_off != next) 817 return (0); 818 next += q->ipqe_ip->ip_len; 819 } 820 if (p->ipqe_mff) 821 return (0); 822 823 /* 824 * Reassembly is complete. Check for a bogus message size and 825 * concatenate fragments. 826 */ 827 q = fp->ipq_fragq.lh_first; 828 ip = q->ipqe_ip; 829 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 830 ipstat.ips_toolong++; 831 ip_freef(fp); 832 return (0); 833 } 834 m = q->ipqe_m; 835 t = m->m_next; 836 m->m_next = 0; 837 m_cat(m, t); 838 nq = q->ipqe_q.le_next; 839 pool_put(&ipqent_pool, q); 840 for (q = nq; q != NULL; q = nq) { 841 t = q->ipqe_m; 842 nq = q->ipqe_q.le_next; 843 pool_put(&ipqent_pool, q); 844 m_cat(m, t); 845 } 846 847 /* 848 * Create header for new ip packet by 849 * modifying header of first packet; 850 * dequeue and discard fragment reassembly header. 851 * Make header visible. 852 */ 853 ip->ip_len = next; 854 ip->ip_src = fp->ipq_src; 855 ip->ip_dst = fp->ipq_dst; 856 LIST_REMOVE(fp, ipq_q); 857 FREE(fp, M_FTABLE); 858 m->m_len += (ip->ip_hl << 2); 859 m->m_data -= (ip->ip_hl << 2); 860 /* some debugging cruft by sklower, below, will go away soon */ 861 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 862 int plen = 0; 863 for (t = m; t; t = t->m_next) 864 plen += t->m_len; 865 m->m_pkthdr.len = plen; 866 } 867 return (m); 868 869dropfrag: 870 ipstat.ips_fragdropped++; 871 m_freem(m); 872 pool_put(&ipqent_pool, ipqe); 873 return (0); 874} 875 876/* 877 * Free a fragment reassembly header and all 878 * associated datagrams. 879 */ 880void 881ip_freef(fp) 882 struct ipq *fp; 883{ 884 struct ipqent *q, *p; 885 886 IPQ_LOCK_CHECK(); 887 888 for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { 889 p = q->ipqe_q.le_next; 890 m_freem(q->ipqe_m); 891 LIST_REMOVE(q, ipqe_q); 892 pool_put(&ipqent_pool, q); 893 } 894 LIST_REMOVE(fp, ipq_q); 895 FREE(fp, M_FTABLE); 896} 897 898/* 899 * IP timer processing; 900 * if a timer expires on a reassembly 901 * queue, discard it. 902 */ 903void 904ip_slowtimo() 905{ 906 struct ipq *fp, *nfp; 907 int s = splsoftnet(); 908 909 IPQ_LOCK(); 910 for (fp = ipq.lh_first; fp != NULL; fp = nfp) { 911 nfp = fp->ipq_q.le_next; 912 if (--fp->ipq_ttl == 0) { 913 ipstat.ips_fragtimeout++; 914 ip_freef(fp); 915 } 916 } 917 IPQ_UNLOCK(); 918#ifdef GATEWAY 919 ipflow_slowtimo(); 920#endif 921 splx(s); 922} 923 924/* 925 * Drain off all datagram fragments. 926 */ 927void 928ip_drain() 929{ 930 931 /* 932 * We may be called from a device's interrupt context. If 933 * the ipq is already busy, just bail out now. 934 */ 935 if (ipq_lock_try() == 0) 936 return; 937 938 while (ipq.lh_first != NULL) { 939 ipstat.ips_fragdropped++; 940 ip_freef(ipq.lh_first); 941 } 942 943 IPQ_UNLOCK(); 944} 945 946/* 947 * Do option processing on a datagram, 948 * possibly discarding it if bad options are encountered, 949 * or forwarding it if source-routed. 950 * Returns 1 if packet has been forwarded/freed, 951 * 0 if the packet should be processed further. 952 */ 953int 954ip_dooptions(m) 955 struct mbuf *m; 956{ 957 struct ip *ip = mtod(m, struct ip *); 958 u_char *cp, *cp0; 959 struct ip_timestamp *ipt; 960 struct in_ifaddr *ia; 961 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 962 struct in_addr dst; 963 n_time ntime; 964 965 dst = ip->ip_dst; 966 cp = (u_char *)(ip + 1); 967 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 968 for (; cnt > 0; cnt -= optlen, cp += optlen) { 969 opt = cp[IPOPT_OPTVAL]; 970 if (opt == IPOPT_EOL) 971 break; 972 if (opt == IPOPT_NOP) 973 optlen = 1; 974 else { 975 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 976 code = &cp[IPOPT_OLEN] - (u_char *)ip; 977 goto bad; 978 } 979 optlen = cp[IPOPT_OLEN]; 980 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 981 code = &cp[IPOPT_OLEN] - (u_char *)ip; 982 goto bad; 983 } 984 } 985 switch (opt) { 986 987 default: 988 break; 989 990 /* 991 * Source routing with record. 992 * Find interface with current destination address. 993 * If none on this machine then drop if strictly routed, 994 * or do nothing if loosely routed. 995 * Record interface address and bring up next address 996 * component. If strictly routed make sure next 997 * address is on directly accessible net. 998 */ 999 case IPOPT_LSRR: 1000 case IPOPT_SSRR: 1001 if (ip_allowsrcrt == 0) { 1002 type = ICMP_UNREACH; 1003 code = ICMP_UNREACH_NET_PROHIB; 1004 goto bad; 1005 } 1006 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1007 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1008 goto bad; 1009 } 1010 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1011 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1012 goto bad; 1013 } 1014 ipaddr.sin_addr = ip->ip_dst; 1015 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1016 if (ia == 0) { 1017 if (opt == IPOPT_SSRR) { 1018 type = ICMP_UNREACH; 1019 code = ICMP_UNREACH_SRCFAIL; 1020 goto bad; 1021 } 1022 /* 1023 * Loose routing, and not at next destination 1024 * yet; nothing to do except forward. 1025 */ 1026 break; 1027 } 1028 off--; /* 0 origin */ 1029 if ((off + sizeof(struct in_addr)) > optlen) { 1030 /* 1031 * End of source route. Should be for us. 1032 */ 1033 save_rte(cp, ip->ip_src); 1034 break; 1035 } 1036 /* 1037 * locate outgoing interface 1038 */ 1039 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1040 sizeof(ipaddr.sin_addr)); 1041 if (opt == IPOPT_SSRR) 1042 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1043 else 1044 ia = ip_rtaddr(ipaddr.sin_addr); 1045 if (ia == 0) { 1046 type = ICMP_UNREACH; 1047 code = ICMP_UNREACH_SRCFAIL; 1048 goto bad; 1049 } 1050 ip->ip_dst = ipaddr.sin_addr; 1051 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1052 (caddr_t)(cp + off), sizeof(struct in_addr)); 1053 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1054 /* 1055 * Let ip_intr's mcast routing check handle mcast pkts 1056 */ 1057 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1058 break; 1059 1060 case IPOPT_RR: 1061 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1062 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1063 goto bad; 1064 } 1065 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1066 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1067 goto bad; 1068 } 1069 /* 1070 * If no space remains, ignore. 1071 */ 1072 off--; /* 0 origin */ 1073 if ((off + sizeof(struct in_addr)) > optlen) 1074 break; 1075 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1076 sizeof(ipaddr.sin_addr)); 1077 /* 1078 * locate outgoing interface; if we're the destination, 1079 * use the incoming interface (should be same). 1080 */ 1081 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1082 == NULL && 1083 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1084 type = ICMP_UNREACH; 1085 code = ICMP_UNREACH_HOST; 1086 goto bad; 1087 } 1088 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1089 (caddr_t)(cp + off), sizeof(struct in_addr)); 1090 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1091 break; 1092 1093 case IPOPT_TS: 1094 code = cp - (u_char *)ip; 1095 ipt = (struct ip_timestamp *)cp; 1096 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1097 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1098 goto bad; 1099 } 1100 if (ipt->ipt_ptr < 5) { 1101 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1102 goto bad; 1103 } 1104 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1105 if (++ipt->ipt_oflw == 0) { 1106 code = (u_char *)&ipt->ipt_ptr - 1107 (u_char *)ip; 1108 goto bad; 1109 } 1110 break; 1111 } 1112 cp0 = (cp + ipt->ipt_ptr - 1); 1113 switch (ipt->ipt_flg) { 1114 1115 case IPOPT_TS_TSONLY: 1116 break; 1117 1118 case IPOPT_TS_TSANDADDR: 1119 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1120 sizeof(struct in_addr) > ipt->ipt_len) { 1121 code = (u_char *)&ipt->ipt_ptr - 1122 (u_char *)ip; 1123 goto bad; 1124 } 1125 ipaddr.sin_addr = dst; 1126 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1127 m->m_pkthdr.rcvif)); 1128 if (ia == 0) 1129 continue; 1130 bcopy(&ia->ia_addr.sin_addr, 1131 cp0, sizeof(struct in_addr)); 1132 ipt->ipt_ptr += sizeof(struct in_addr); 1133 break; 1134 1135 case IPOPT_TS_PRESPEC: 1136 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1137 sizeof(struct in_addr) > ipt->ipt_len) { 1138 code = (u_char *)&ipt->ipt_ptr - 1139 (u_char *)ip; 1140 goto bad; 1141 } 1142 bcopy(cp0, &ipaddr.sin_addr, 1143 sizeof(struct in_addr)); 1144 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1145 == NULL) 1146 continue; 1147 ipt->ipt_ptr += sizeof(struct in_addr); 1148 break; 1149 1150 default: 1151 /* XXX can't take &ipt->ipt_flg */ 1152 code = (u_char *)&ipt->ipt_ptr - 1153 (u_char *)ip + 1; 1154 goto bad; 1155 } 1156 ntime = iptime(); 1157 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1158 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1159 sizeof(n_time)); 1160 ipt->ipt_ptr += sizeof(n_time); 1161 } 1162 } 1163 if (forward) { 1164 if (ip_forwsrcrt == 0) { 1165 type = ICMP_UNREACH; 1166 code = ICMP_UNREACH_SRCFAIL; 1167 goto bad; 1168 } 1169 ip_forward(m, 1); 1170 return (1); 1171 } 1172 return (0); 1173bad: 1174 icmp_error(m, type, code, 0, 0); 1175 ipstat.ips_badoptions++; 1176 return (1); 1177} 1178 1179/* 1180 * Given address of next destination (final or next hop), 1181 * return internet address info of interface to be used to get there. 1182 */ 1183struct in_ifaddr * 1184ip_rtaddr(dst) 1185 struct in_addr dst; 1186{ 1187 struct sockaddr_in *sin; 1188 1189 sin = satosin(&ipforward_rt.ro_dst); 1190 1191 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1192 if (ipforward_rt.ro_rt) { 1193 RTFREE(ipforward_rt.ro_rt); 1194 ipforward_rt.ro_rt = 0; 1195 } 1196 sin->sin_family = AF_INET; 1197 sin->sin_len = sizeof(*sin); 1198 sin->sin_addr = dst; 1199 1200 rtalloc(&ipforward_rt); 1201 } 1202 if (ipforward_rt.ro_rt == 0) 1203 return ((struct in_ifaddr *)0); 1204 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1205} 1206 1207/* 1208 * Save incoming source route for use in replies, 1209 * to be picked up later by ip_srcroute if the receiver is interested. 1210 */ 1211void 1212save_rte(option, dst) 1213 u_char *option; 1214 struct in_addr dst; 1215{ 1216 unsigned olen; 1217 1218 olen = option[IPOPT_OLEN]; 1219#ifdef DIAGNOSTIC 1220 if (ipprintfs) 1221 printf("save_rte: olen %d\n", olen); 1222#endif /* 0 */ 1223 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1224 return; 1225 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1226 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1227 ip_srcrt.dst = dst; 1228} 1229 1230/* 1231 * Retrieve incoming source route for use in replies, 1232 * in the same form used by setsockopt. 1233 * The first hop is placed before the options, will be removed later. 1234 */ 1235struct mbuf * 1236ip_srcroute() 1237{ 1238 struct in_addr *p, *q; 1239 struct mbuf *m; 1240 1241 if (ip_nhops == 0) 1242 return ((struct mbuf *)0); 1243 m = m_get(M_DONTWAIT, MT_SOOPTS); 1244 if (m == 0) 1245 return ((struct mbuf *)0); 1246 1247#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1248 1249 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1250 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1251 OPTSIZ; 1252#ifdef DIAGNOSTIC 1253 if (ipprintfs) 1254 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1255#endif 1256 1257 /* 1258 * First save first hop for return route 1259 */ 1260 p = &ip_srcrt.route[ip_nhops - 1]; 1261 *(mtod(m, struct in_addr *)) = *p--; 1262#ifdef DIAGNOSTIC 1263 if (ipprintfs) 1264 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1265#endif 1266 1267 /* 1268 * Copy option fields and padding (nop) to mbuf. 1269 */ 1270 ip_srcrt.nop = IPOPT_NOP; 1271 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1272 bcopy((caddr_t)&ip_srcrt.nop, 1273 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1274 q = (struct in_addr *)(mtod(m, caddr_t) + 1275 sizeof(struct in_addr) + OPTSIZ); 1276#undef OPTSIZ 1277 /* 1278 * Record return path as an IP source route, 1279 * reversing the path (pointers are now aligned). 1280 */ 1281 while (p >= ip_srcrt.route) { 1282#ifdef DIAGNOSTIC 1283 if (ipprintfs) 1284 printf(" %x", ntohl(q->s_addr)); 1285#endif 1286 *q++ = *p--; 1287 } 1288 /* 1289 * Last hop goes to final destination. 1290 */ 1291 *q = ip_srcrt.dst; 1292#ifdef DIAGNOSTIC 1293 if (ipprintfs) 1294 printf(" %x\n", ntohl(q->s_addr)); 1295#endif 1296 return (m); 1297} 1298 1299/* 1300 * Strip out IP options, at higher 1301 * level protocol in the kernel. 1302 * Second argument is buffer to which options 1303 * will be moved, and return value is their length. 1304 * XXX should be deleted; last arg currently ignored. 1305 */ 1306void 1307ip_stripoptions(m, mopt) 1308 struct mbuf *m; 1309 struct mbuf *mopt; 1310{ 1311 int i; 1312 struct ip *ip = mtod(m, struct ip *); 1313 caddr_t opts; 1314 int olen; 1315 1316 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1317 opts = (caddr_t)(ip + 1); 1318 i = m->m_len - (sizeof (struct ip) + olen); 1319 bcopy(opts + olen, opts, (unsigned)i); 1320 m->m_len -= olen; 1321 if (m->m_flags & M_PKTHDR) 1322 m->m_pkthdr.len -= olen; 1323 ip->ip_len -= olen; 1324 ip->ip_hl = sizeof (struct ip) >> 2; 1325} 1326 1327int inetctlerrmap[PRC_NCMDS] = { 1328 0, 0, 0, 0, 1329 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1330 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1331 EMSGSIZE, EHOSTUNREACH, 0, 0, 1332 0, 0, 0, 0, 1333 ENOPROTOOPT 1334}; 1335 1336/* 1337 * Forward a packet. If some error occurs return the sender 1338 * an icmp packet. Note we can't always generate a meaningful 1339 * icmp message because icmp doesn't have a large enough repertoire 1340 * of codes and types. 1341 * 1342 * If not forwarding, just drop the packet. This could be confusing 1343 * if ipforwarding was zero but some routing protocol was advancing 1344 * us as a gateway to somewhere. However, we must let the routing 1345 * protocol deal with that. 1346 * 1347 * The srcrt parameter indicates whether the packet is being forwarded 1348 * via a source route. 1349 */ 1350void 1351ip_forward(m, srcrt) 1352 struct mbuf *m; 1353 int srcrt; 1354{ 1355 struct ip *ip = mtod(m, struct ip *); 1356 struct sockaddr_in *sin; 1357 struct rtentry *rt; 1358 int error, type = 0, code = 0; 1359 struct mbuf *mcopy; 1360 n_long dest; 1361 struct ifnet *destifp; 1362#ifdef IPSEC 1363 struct ifnet dummyifp; 1364#endif 1365 1366 dest = 0; 1367#ifdef DIAGNOSTIC 1368 if (ipprintfs) 1369 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1370 ntohl(ip->ip_src.s_addr), 1371 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1372#endif 1373 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1374 ipstat.ips_cantforward++; 1375 m_freem(m); 1376 return; 1377 } 1378 if (ip->ip_ttl <= IPTTLDEC) { 1379 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1380 return; 1381 } 1382 ip->ip_ttl -= IPTTLDEC; 1383 1384 sin = satosin(&ipforward_rt.ro_dst); 1385 if ((rt = ipforward_rt.ro_rt) == 0 || 1386 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1387 if (ipforward_rt.ro_rt) { 1388 RTFREE(ipforward_rt.ro_rt); 1389 ipforward_rt.ro_rt = 0; 1390 } 1391 sin->sin_family = AF_INET; 1392 sin->sin_len = sizeof(struct sockaddr_in); 1393 sin->sin_addr = ip->ip_dst; 1394 1395 rtalloc(&ipforward_rt); 1396 if (ipforward_rt.ro_rt == 0) { 1397 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1398 return; 1399 } 1400 rt = ipforward_rt.ro_rt; 1401 } 1402 1403 /* 1404 * Save at most 68 bytes of the packet in case 1405 * we need to generate an ICMP message to the src. 1406 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1407 */ 1408 mcopy = m_copym(m, 0, imin((int)ip->ip_len, 68), M_DONTWAIT); 1409 if (mcopy) 1410 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1411 1412 /* 1413 * If forwarding packet using same interface that it came in on, 1414 * perhaps should send a redirect to sender to shortcut a hop. 1415 * Only send redirect if source is sending directly to us, 1416 * and if packet was not source routed (or has any options). 1417 * Also, don't send redirect if forwarding using a default route 1418 * or a route modified by a redirect. 1419 */ 1420 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1421 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1422 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1423 ipsendredirects && !srcrt) { 1424 if (rt->rt_ifa && 1425 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1426 ifatoia(rt->rt_ifa)->ia_subnet) { 1427 if (rt->rt_flags & RTF_GATEWAY) 1428 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1429 else 1430 dest = ip->ip_dst.s_addr; 1431 /* 1432 * Router requirements says to only send host 1433 * redirects. 1434 */ 1435 type = ICMP_REDIRECT; 1436 code = ICMP_REDIRECT_HOST; 1437#ifdef DIAGNOSTIC 1438 if (ipprintfs) 1439 printf("redirect (%d) to %x\n", code, 1440 (u_int32_t)dest); 1441#endif 1442 } 1443 } 1444 1445#ifdef IPSEC 1446 /* Don't lookup socket in forwading case */ 1447 ipsec_setsocket(m, NULL); 1448#endif 1449 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1450 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1451 if (error) 1452 ipstat.ips_cantforward++; 1453 else { 1454 ipstat.ips_forward++; 1455 if (type) 1456 ipstat.ips_redirectsent++; 1457 else { 1458 if (mcopy) { 1459#ifdef GATEWAY 1460 if (mcopy->m_flags & M_CANFASTFWD) 1461 ipflow_create(&ipforward_rt, mcopy); 1462#endif 1463 m_freem(mcopy); 1464 } 1465 return; 1466 } 1467 } 1468 if (mcopy == NULL) 1469 return; 1470 destifp = NULL; 1471 1472 switch (error) { 1473 1474 case 0: /* forwarded, but need redirect */ 1475 /* type, code set above */ 1476 break; 1477 1478 case ENETUNREACH: /* shouldn't happen, checked above */ 1479 case EHOSTUNREACH: 1480 case ENETDOWN: 1481 case EHOSTDOWN: 1482 default: 1483 type = ICMP_UNREACH; 1484 code = ICMP_UNREACH_HOST; 1485 break; 1486 1487 case EMSGSIZE: 1488 type = ICMP_UNREACH; 1489 code = ICMP_UNREACH_NEEDFRAG; 1490#ifndef IPSEC 1491 if (ipforward_rt.ro_rt) 1492 destifp = ipforward_rt.ro_rt->rt_ifp; 1493#else 1494 /* 1495 * If the packet is routed over IPsec tunnel, tell the 1496 * originator the tunnel MTU. 1497 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1498 * XXX quickhack!!! 1499 */ 1500 if (ipforward_rt.ro_rt) { 1501 struct secpolicy *sp; 1502 int ipsecerror; 1503 size_t ipsechdr; 1504 struct route *ro; 1505 1506 sp = ipsec4_getpolicybyaddr(mcopy, 1507 IPSEC_DIR_OUTBOUND, 1508 IP_FORWARDING, 1509 &ipsecerror); 1510 1511 if (sp == NULL) 1512 destifp = ipforward_rt.ro_rt->rt_ifp; 1513 else { 1514 /* count IPsec header size */ 1515 ipsechdr = ipsec4_hdrsiz(mcopy, 1516 IPSEC_DIR_OUTBOUND, 1517 NULL); 1518 1519 /* 1520 * find the correct route for outer IPv4 1521 * header, compute tunnel MTU. 1522 * 1523 * XXX BUG ALERT 1524 * The "dummyifp" code relies upon the fact 1525 * that icmp_error() touches only ifp->if_mtu. 1526 */ 1527 /*XXX*/ 1528 destifp = NULL; 1529 if (sp->req != NULL 1530 && sp->req->sav != NULL 1531 && sp->req->sav->sah != NULL) { 1532 ro = &sp->req->sav->sah->sa_route; 1533 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1534 dummyifp.if_mtu = 1535 ro->ro_rt->rt_ifp->if_mtu; 1536 dummyifp.if_mtu -= ipsechdr; 1537 destifp = &dummyifp; 1538 } 1539 } 1540 1541 key_freesp(sp); 1542 } 1543 } 1544#endif /*IPSEC*/ 1545 ipstat.ips_cantfrag++; 1546 break; 1547 1548 case ENOBUFS: 1549 type = ICMP_SOURCEQUENCH; 1550 code = 0; 1551 break; 1552 } 1553 icmp_error(mcopy, type, code, dest, destifp); 1554} 1555 1556void 1557ip_savecontrol(inp, mp, ip, m) 1558 struct inpcb *inp; 1559 struct mbuf **mp; 1560 struct ip *ip; 1561 struct mbuf *m; 1562{ 1563 1564 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1565 struct timeval tv; 1566 1567 microtime(&tv); 1568 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1569 SCM_TIMESTAMP, SOL_SOCKET); 1570 if (*mp) 1571 mp = &(*mp)->m_next; 1572 } 1573 if (inp->inp_flags & INP_RECVDSTADDR) { 1574 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1575 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1576 if (*mp) 1577 mp = &(*mp)->m_next; 1578 } 1579#ifdef notyet 1580 /* 1581 * XXX 1582 * Moving these out of udp_input() made them even more broken 1583 * than they already were. 1584 * - fenner@parc.xerox.com 1585 */ 1586 /* options were tossed already */ 1587 if (inp->inp_flags & INP_RECVOPTS) { 1588 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1589 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1590 if (*mp) 1591 mp = &(*mp)->m_next; 1592 } 1593 /* ip_srcroute doesn't do what we want here, need to fix */ 1594 if (inp->inp_flags & INP_RECVRETOPTS) { 1595 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1596 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1597 if (*mp) 1598 mp = &(*mp)->m_next; 1599 } 1600#endif 1601 if (inp->inp_flags & INP_RECVIF) { 1602 struct sockaddr_dl sdl; 1603 1604 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1605 sdl.sdl_family = AF_LINK; 1606 sdl.sdl_index = m->m_pkthdr.rcvif ? 1607 m->m_pkthdr.rcvif->if_index : 0; 1608 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1609 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1610 IP_RECVIF, IPPROTO_IP); 1611 if (*mp) 1612 mp = &(*mp)->m_next; 1613 } 1614} 1615 1616int 1617ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1618 int *name; 1619 u_int namelen; 1620 void *oldp; 1621 size_t *oldlenp; 1622 void *newp; 1623 size_t newlen; 1624{ 1625 extern int subnetsarelocal, hostzeroisbroadcast; 1626 1627 int error, old; 1628 1629 /* All sysctl names at this level are terminal. */ 1630 if (namelen != 1) 1631 return (ENOTDIR); 1632 1633 switch (name[0]) { 1634 case IPCTL_FORWARDING: 1635 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1636 case IPCTL_SENDREDIRECTS: 1637 return (sysctl_int(oldp, oldlenp, newp, newlen, 1638 &ipsendredirects)); 1639 case IPCTL_DEFTTL: 1640 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1641#ifdef notyet 1642 case IPCTL_DEFMTU: 1643 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1644#endif 1645 case IPCTL_FORWSRCRT: 1646 /* Don't allow this to change in a secure environment. */ 1647 if (securelevel > 0) 1648 return (sysctl_rdint(oldp, oldlenp, newp, 1649 ip_forwsrcrt)); 1650 else 1651 return (sysctl_int(oldp, oldlenp, newp, newlen, 1652 &ip_forwsrcrt)); 1653 case IPCTL_DIRECTEDBCAST: 1654 return (sysctl_int(oldp, oldlenp, newp, newlen, 1655 &ip_directedbcast)); 1656 case IPCTL_ALLOWSRCRT: 1657 return (sysctl_int(oldp, oldlenp, newp, newlen, 1658 &ip_allowsrcrt)); 1659 case IPCTL_SUBNETSARELOCAL: 1660 return (sysctl_int(oldp, oldlenp, newp, newlen, 1661 &subnetsarelocal)); 1662 case IPCTL_MTUDISC: 1663 error = sysctl_int(oldp, oldlenp, newp, newlen, 1664 &ip_mtudisc); 1665 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1666 ip_mtudisc_timeout_q = 1667 rt_timer_queue_create(ip_mtudisc_timeout); 1668 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1669 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1670 ip_mtudisc_timeout_q = NULL; 1671 } 1672 return error; 1673 case IPCTL_ANONPORTMIN: 1674 old = anonportmin; 1675 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1676 if (anonportmin >= anonportmax || anonportmin < 0 1677 || anonportmin > 65535 1678#ifndef IPNOPRIVPORTS 1679 || anonportmin < IPPORT_RESERVED 1680#endif 1681 ) { 1682 anonportmin = old; 1683 return (EINVAL); 1684 } 1685 return (error); 1686 case IPCTL_ANONPORTMAX: 1687 old = anonportmax; 1688 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1689 if (anonportmin >= anonportmax || anonportmax < 0 1690 || anonportmax > 65535 1691#ifndef IPNOPRIVPORTS 1692 || anonportmax < IPPORT_RESERVED 1693#endif 1694 ) { 1695 anonportmax = old; 1696 return (EINVAL); 1697 } 1698 return (error); 1699 case IPCTL_MTUDISCTIMEOUT: 1700 error = sysctl_int(oldp, oldlenp, newp, newlen, 1701 &ip_mtudisc_timeout); 1702 if (ip_mtudisc_timeout_q != NULL) 1703 rt_timer_queue_change(ip_mtudisc_timeout_q, 1704 ip_mtudisc_timeout); 1705 return (error); 1706#ifdef GATEWAY 1707 case IPCTL_MAXFLOWS: 1708 { 1709 int s; 1710 1711 error = sysctl_int(oldp, oldlenp, newp, newlen, 1712 &ip_maxflows); 1713 s = splsoftnet(); 1714 ipflow_reap(0); 1715 splx(s); 1716 return (error); 1717 } 1718#endif 1719 case IPCTL_HOSTZEROBROADCAST: 1720 return (sysctl_int(oldp, oldlenp, newp, newlen, 1721 &hostzeroisbroadcast)); 1722#if NGIF > 0 1723 case IPCTL_GIF_TTL: 1724 return(sysctl_int(oldp, oldlenp, newp, newlen, 1725 &ip_gif_ttl)); 1726#endif 1727 1728#ifndef IPNOPRIVPORTS 1729 case IPCTL_LOWPORTMIN: 1730 old = lowportmin; 1731 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmin); 1732 if (lowportmin >= lowportmax 1733 || lowportmin > IPPORT_RESERVEDMAX 1734 || lowportmin < IPPORT_RESERVEDMIN 1735 ) { 1736 lowportmin = old; 1737 return (EINVAL); 1738 } 1739 return (error); 1740 case IPCTL_LOWPORTMAX: 1741 old = lowportmax; 1742 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmax); 1743 if (lowportmin >= lowportmax 1744 || lowportmax > IPPORT_RESERVEDMAX 1745 || lowportmax < IPPORT_RESERVEDMIN 1746 ) { 1747 lowportmax = old; 1748 return (EINVAL); 1749 } 1750 return (error); 1751#endif 1752 1753 default: 1754 return (EOPNOTSUPP); 1755 } 1756 /* NOTREACHED */ 1757} 1758