ip_input.c revision 1.128
1/* $NetBSD: ip_input.c,v 1.128 2001/03/01 16:31:39 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_type = PFIL_TYPE_AF; 323 inet_pfil_hook.ph_af = AF_INET; 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 /* 475 * let ipfilter look at packet on the wire, 476 * not the decapsulated packet. 477 */ 478#ifdef IPSEC 479 if (!ipsec_gethist(m, NULL)) 480#else 481 if (1) 482#endif 483 { 484 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 485 PFIL_IN) != 0) 486 return; 487 if (m == NULL) 488 return; 489 ip = mtod(m, struct ip *); 490 } 491#endif /* PFIL_HOOKS */ 492 493#ifdef ALTQ 494 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 495 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 496 /* packet dropped by traffic conditioner */ 497 return; 498 } 499#endif 500 501 /* 502 * Convert fields to host representation. 503 */ 504 NTOHS(ip->ip_len); 505 NTOHS(ip->ip_off); 506 507 /* 508 * Process options and, if not destined for us, 509 * ship it on. ip_dooptions returns 1 when an 510 * error was detected (causing an icmp message 511 * to be sent and the original packet to be freed). 512 */ 513 ip_nhops = 0; /* for source routed packets */ 514 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 515 return; 516 517 /* 518 * Check our list of addresses, to see if the packet is for us. 519 * 520 * Traditional 4.4BSD did not consult IFF_UP at all. 521 * The behavior here is to treat addresses on !IFF_UP interface 522 * as not mine. 523 */ 524 downmatch = 0; 525 for (ia = IN_IFADDR_HASH(ip->ip_dst.s_addr).lh_first; 526 ia != NULL; 527 ia = ia->ia_hash.le_next) { 528 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 529 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 530 break; 531 else 532 downmatch++; 533 } 534 } 535 if (ia != NULL) 536 goto ours; 537 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 538 for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first; 539 ifa != NULL; ifa = ifa->ifa_list.tqe_next) { 540 if (ifa->ifa_addr->sa_family != AF_INET) continue; 541 ia = ifatoia(ifa); 542 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 543 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 544 /* 545 * Look for all-0's host part (old broadcast addr), 546 * either for subnet or net. 547 */ 548 ip->ip_dst.s_addr == ia->ia_subnet || 549 ip->ip_dst.s_addr == ia->ia_net) 550 goto ours; 551 /* 552 * An interface with IP address zero accepts 553 * all packets that arrive on that interface. 554 */ 555 if (in_nullhost(ia->ia_addr.sin_addr)) 556 goto ours; 557 } 558 } 559 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 560 struct in_multi *inm; 561#ifdef MROUTING 562 extern struct socket *ip_mrouter; 563 564 if (m->m_flags & M_EXT) { 565 if ((m = m_pullup(m, hlen)) == 0) { 566 ipstat.ips_toosmall++; 567 return; 568 } 569 ip = mtod(m, struct ip *); 570 } 571 572 if (ip_mrouter) { 573 /* 574 * If we are acting as a multicast router, all 575 * incoming multicast packets are passed to the 576 * kernel-level multicast forwarding function. 577 * The packet is returned (relatively) intact; if 578 * ip_mforward() returns a non-zero value, the packet 579 * must be discarded, else it may be accepted below. 580 * 581 * (The IP ident field is put in the same byte order 582 * as expected when ip_mforward() is called from 583 * ip_output().) 584 */ 585 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 586 ipstat.ips_cantforward++; 587 m_freem(m); 588 return; 589 } 590 591 /* 592 * The process-level routing demon needs to receive 593 * all multicast IGMP packets, whether or not this 594 * host belongs to their destination groups. 595 */ 596 if (ip->ip_p == IPPROTO_IGMP) 597 goto ours; 598 ipstat.ips_forward++; 599 } 600#endif 601 /* 602 * See if we belong to the destination multicast group on the 603 * arrival interface. 604 */ 605 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 606 if (inm == NULL) { 607 ipstat.ips_cantforward++; 608 m_freem(m); 609 return; 610 } 611 goto ours; 612 } 613 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 614 in_nullhost(ip->ip_dst)) 615 goto ours; 616 617 /* 618 * Not for us; forward if possible and desirable. 619 */ 620 if (ipforwarding == 0) { 621 ipstat.ips_cantforward++; 622 m_freem(m); 623 } else { 624 /* 625 * If ip_dst matched any of my address on !IFF_UP interface, 626 * and there's no IFF_UP interface that matches ip_dst, 627 * send icmp unreach. Forwarding it will result in in-kernel 628 * forwarding loop till TTL goes to 0. 629 */ 630 if (downmatch) { 631 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 632 ipstat.ips_cantforward++; 633 return; 634 } 635 ip_forward(m, 0); 636 } 637 return; 638 639ours: 640 /* 641 * If offset or IP_MF are set, must reassemble. 642 * Otherwise, nothing need be done. 643 * (We could look in the reassembly queue to see 644 * if the packet was previously fragmented, 645 * but it's not worth the time; just let them time out.) 646 */ 647 if (ip->ip_off & ~(IP_DF|IP_RF)) { 648 /* 649 * Look for queue of fragments 650 * of this datagram. 651 */ 652 IPQ_LOCK(); 653 for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next) 654 if (ip->ip_id == fp->ipq_id && 655 in_hosteq(ip->ip_src, fp->ipq_src) && 656 in_hosteq(ip->ip_dst, fp->ipq_dst) && 657 ip->ip_p == fp->ipq_p) 658 goto found; 659 fp = 0; 660found: 661 662 /* 663 * Adjust ip_len to not reflect header, 664 * set ipqe_mff if more fragments are expected, 665 * convert offset of this to bytes. 666 */ 667 ip->ip_len -= hlen; 668 mff = (ip->ip_off & IP_MF) != 0; 669 if (mff) { 670 /* 671 * Make sure that fragments have a data length 672 * that's a non-zero multiple of 8 bytes. 673 */ 674 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 675 ipstat.ips_badfrags++; 676 IPQ_UNLOCK(); 677 goto bad; 678 } 679 } 680 ip->ip_off <<= 3; 681 682 /* 683 * If datagram marked as having more fragments 684 * or if this is not the first fragment, 685 * attempt reassembly; if it succeeds, proceed. 686 */ 687 if (mff || ip->ip_off) { 688 ipstat.ips_fragments++; 689 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 690 if (ipqe == NULL) { 691 ipstat.ips_rcvmemdrop++; 692 IPQ_UNLOCK(); 693 goto bad; 694 } 695 ipqe->ipqe_mff = mff; 696 ipqe->ipqe_m = m; 697 ipqe->ipqe_ip = ip; 698 m = ip_reass(ipqe, fp); 699 if (m == 0) { 700 IPQ_UNLOCK(); 701 return; 702 } 703 ipstat.ips_reassembled++; 704 ip = mtod(m, struct ip *); 705 hlen = ip->ip_hl << 2; 706 ip->ip_len += hlen; 707 } else 708 if (fp) 709 ip_freef(fp); 710 IPQ_UNLOCK(); 711 } 712 713#ifdef IPSEC 714 /* 715 * enforce IPsec policy checking if we are seeing last header. 716 * note that we do not visit this with protocols with pcb layer 717 * code - like udp/tcp/raw ip. 718 */ 719 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 720 ipsec4_in_reject(m, NULL)) { 721 ipsecstat.in_polvio++; 722 goto bad; 723 } 724#endif 725 726 /* 727 * Switch out to protocol's input routine. 728 */ 729#if IFA_STATS 730 if (ia && ip) 731 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 732#endif 733 ipstat.ips_delivered++; 734 { 735 int off = hlen, nh = ip->ip_p; 736 737 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 738 return; 739 } 740bad: 741 m_freem(m); 742} 743 744/* 745 * Take incoming datagram fragment and try to 746 * reassemble it into whole datagram. If a chain for 747 * reassembly of this datagram already exists, then it 748 * is given as fp; otherwise have to make a chain. 749 */ 750struct mbuf * 751ip_reass(ipqe, fp) 752 struct ipqent *ipqe; 753 struct ipq *fp; 754{ 755 struct mbuf *m = ipqe->ipqe_m; 756 struct ipqent *nq, *p, *q; 757 struct ip *ip; 758 struct mbuf *t; 759 int hlen = ipqe->ipqe_ip->ip_hl << 2; 760 int i, next; 761 762 IPQ_LOCK_CHECK(); 763 764 /* 765 * Presence of header sizes in mbufs 766 * would confuse code below. 767 */ 768 m->m_data += hlen; 769 m->m_len -= hlen; 770 771 /* 772 * If first fragment to arrive, create a reassembly queue. 773 */ 774 if (fp == 0) { 775 MALLOC(fp, struct ipq *, sizeof (struct ipq), 776 M_FTABLE, M_NOWAIT); 777 if (fp == NULL) 778 goto dropfrag; 779 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 780 fp->ipq_ttl = IPFRAGTTL; 781 fp->ipq_p = ipqe->ipqe_ip->ip_p; 782 fp->ipq_id = ipqe->ipqe_ip->ip_id; 783 LIST_INIT(&fp->ipq_fragq); 784 fp->ipq_src = ipqe->ipqe_ip->ip_src; 785 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 786 p = NULL; 787 goto insert; 788 } 789 790 /* 791 * Find a segment which begins after this one does. 792 */ 793 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 794 p = q, q = q->ipqe_q.le_next) 795 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 796 break; 797 798 /* 799 * If there is a preceding segment, it may provide some of 800 * our data already. If so, drop the data from the incoming 801 * segment. If it provides all of our data, drop us. 802 */ 803 if (p != NULL) { 804 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 805 ipqe->ipqe_ip->ip_off; 806 if (i > 0) { 807 if (i >= ipqe->ipqe_ip->ip_len) 808 goto dropfrag; 809 m_adj(ipqe->ipqe_m, i); 810 ipqe->ipqe_ip->ip_off += i; 811 ipqe->ipqe_ip->ip_len -= i; 812 } 813 } 814 815 /* 816 * While we overlap succeeding segments trim them or, 817 * if they are completely covered, dequeue them. 818 */ 819 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 820 q->ipqe_ip->ip_off; q = nq) { 821 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 822 q->ipqe_ip->ip_off; 823 if (i < q->ipqe_ip->ip_len) { 824 q->ipqe_ip->ip_len -= i; 825 q->ipqe_ip->ip_off += i; 826 m_adj(q->ipqe_m, i); 827 break; 828 } 829 nq = q->ipqe_q.le_next; 830 m_freem(q->ipqe_m); 831 LIST_REMOVE(q, ipqe_q); 832 pool_put(&ipqent_pool, q); 833 } 834 835insert: 836 /* 837 * Stick new segment in its place; 838 * check for complete reassembly. 839 */ 840 if (p == NULL) { 841 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 842 } else { 843 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 844 } 845 next = 0; 846 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 847 p = q, q = q->ipqe_q.le_next) { 848 if (q->ipqe_ip->ip_off != next) 849 return (0); 850 next += q->ipqe_ip->ip_len; 851 } 852 if (p->ipqe_mff) 853 return (0); 854 855 /* 856 * Reassembly is complete. Check for a bogus message size and 857 * concatenate fragments. 858 */ 859 q = fp->ipq_fragq.lh_first; 860 ip = q->ipqe_ip; 861 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 862 ipstat.ips_toolong++; 863 ip_freef(fp); 864 return (0); 865 } 866 m = q->ipqe_m; 867 t = m->m_next; 868 m->m_next = 0; 869 m_cat(m, t); 870 nq = q->ipqe_q.le_next; 871 pool_put(&ipqent_pool, q); 872 for (q = nq; q != NULL; q = nq) { 873 t = q->ipqe_m; 874 nq = q->ipqe_q.le_next; 875 pool_put(&ipqent_pool, q); 876 m_cat(m, t); 877 } 878 879 /* 880 * Create header for new ip packet by 881 * modifying header of first packet; 882 * dequeue and discard fragment reassembly header. 883 * Make header visible. 884 */ 885 ip->ip_len = next; 886 ip->ip_src = fp->ipq_src; 887 ip->ip_dst = fp->ipq_dst; 888 LIST_REMOVE(fp, ipq_q); 889 FREE(fp, M_FTABLE); 890 m->m_len += (ip->ip_hl << 2); 891 m->m_data -= (ip->ip_hl << 2); 892 /* some debugging cruft by sklower, below, will go away soon */ 893 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 894 int plen = 0; 895 for (t = m; t; t = t->m_next) 896 plen += t->m_len; 897 m->m_pkthdr.len = plen; 898 } 899 return (m); 900 901dropfrag: 902 ipstat.ips_fragdropped++; 903 m_freem(m); 904 pool_put(&ipqent_pool, ipqe); 905 return (0); 906} 907 908/* 909 * Free a fragment reassembly header and all 910 * associated datagrams. 911 */ 912void 913ip_freef(fp) 914 struct ipq *fp; 915{ 916 struct ipqent *q, *p; 917 918 IPQ_LOCK_CHECK(); 919 920 for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { 921 p = q->ipqe_q.le_next; 922 m_freem(q->ipqe_m); 923 LIST_REMOVE(q, ipqe_q); 924 pool_put(&ipqent_pool, q); 925 } 926 LIST_REMOVE(fp, ipq_q); 927 FREE(fp, M_FTABLE); 928} 929 930/* 931 * IP timer processing; 932 * if a timer expires on a reassembly 933 * queue, discard it. 934 */ 935void 936ip_slowtimo() 937{ 938 struct ipq *fp, *nfp; 939 int s = splsoftnet(); 940 941 IPQ_LOCK(); 942 for (fp = ipq.lh_first; fp != NULL; fp = nfp) { 943 nfp = fp->ipq_q.le_next; 944 if (--fp->ipq_ttl == 0) { 945 ipstat.ips_fragtimeout++; 946 ip_freef(fp); 947 } 948 } 949 IPQ_UNLOCK(); 950#ifdef GATEWAY 951 ipflow_slowtimo(); 952#endif 953 splx(s); 954} 955 956/* 957 * Drain off all datagram fragments. 958 */ 959void 960ip_drain() 961{ 962 963 /* 964 * We may be called from a device's interrupt context. If 965 * the ipq is already busy, just bail out now. 966 */ 967 if (ipq_lock_try() == 0) 968 return; 969 970 while (ipq.lh_first != NULL) { 971 ipstat.ips_fragdropped++; 972 ip_freef(ipq.lh_first); 973 } 974 975 IPQ_UNLOCK(); 976} 977 978/* 979 * Do option processing on a datagram, 980 * possibly discarding it if bad options are encountered, 981 * or forwarding it if source-routed. 982 * Returns 1 if packet has been forwarded/freed, 983 * 0 if the packet should be processed further. 984 */ 985int 986ip_dooptions(m) 987 struct mbuf *m; 988{ 989 struct ip *ip = mtod(m, struct ip *); 990 u_char *cp, *cp0; 991 struct ip_timestamp *ipt; 992 struct in_ifaddr *ia; 993 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 994 struct in_addr dst; 995 n_time ntime; 996 997 dst = ip->ip_dst; 998 cp = (u_char *)(ip + 1); 999 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1000 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1001 opt = cp[IPOPT_OPTVAL]; 1002 if (opt == IPOPT_EOL) 1003 break; 1004 if (opt == IPOPT_NOP) 1005 optlen = 1; 1006 else { 1007 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1008 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1009 goto bad; 1010 } 1011 optlen = cp[IPOPT_OLEN]; 1012 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1013 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1014 goto bad; 1015 } 1016 } 1017 switch (opt) { 1018 1019 default: 1020 break; 1021 1022 /* 1023 * Source routing with record. 1024 * Find interface with current destination address. 1025 * If none on this machine then drop if strictly routed, 1026 * or do nothing if loosely routed. 1027 * Record interface address and bring up next address 1028 * component. If strictly routed make sure next 1029 * address is on directly accessible net. 1030 */ 1031 case IPOPT_LSRR: 1032 case IPOPT_SSRR: 1033 if (ip_allowsrcrt == 0) { 1034 type = ICMP_UNREACH; 1035 code = ICMP_UNREACH_NET_PROHIB; 1036 goto bad; 1037 } 1038 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1039 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1040 goto bad; 1041 } 1042 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1043 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1044 goto bad; 1045 } 1046 ipaddr.sin_addr = ip->ip_dst; 1047 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1048 if (ia == 0) { 1049 if (opt == IPOPT_SSRR) { 1050 type = ICMP_UNREACH; 1051 code = ICMP_UNREACH_SRCFAIL; 1052 goto bad; 1053 } 1054 /* 1055 * Loose routing, and not at next destination 1056 * yet; nothing to do except forward. 1057 */ 1058 break; 1059 } 1060 off--; /* 0 origin */ 1061 if ((off + sizeof(struct in_addr)) > optlen) { 1062 /* 1063 * End of source route. Should be for us. 1064 */ 1065 save_rte(cp, ip->ip_src); 1066 break; 1067 } 1068 /* 1069 * locate outgoing interface 1070 */ 1071 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1072 sizeof(ipaddr.sin_addr)); 1073 if (opt == IPOPT_SSRR) 1074 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1075 else 1076 ia = ip_rtaddr(ipaddr.sin_addr); 1077 if (ia == 0) { 1078 type = ICMP_UNREACH; 1079 code = ICMP_UNREACH_SRCFAIL; 1080 goto bad; 1081 } 1082 ip->ip_dst = ipaddr.sin_addr; 1083 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1084 (caddr_t)(cp + off), sizeof(struct in_addr)); 1085 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1086 /* 1087 * Let ip_intr's mcast routing check handle mcast pkts 1088 */ 1089 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1090 break; 1091 1092 case IPOPT_RR: 1093 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1094 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1095 goto bad; 1096 } 1097 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1098 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1099 goto bad; 1100 } 1101 /* 1102 * If no space remains, ignore. 1103 */ 1104 off--; /* 0 origin */ 1105 if ((off + sizeof(struct in_addr)) > optlen) 1106 break; 1107 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1108 sizeof(ipaddr.sin_addr)); 1109 /* 1110 * locate outgoing interface; if we're the destination, 1111 * use the incoming interface (should be same). 1112 */ 1113 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1114 == NULL && 1115 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1116 type = ICMP_UNREACH; 1117 code = ICMP_UNREACH_HOST; 1118 goto bad; 1119 } 1120 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1121 (caddr_t)(cp + off), sizeof(struct in_addr)); 1122 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1123 break; 1124 1125 case IPOPT_TS: 1126 code = cp - (u_char *)ip; 1127 ipt = (struct ip_timestamp *)cp; 1128 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1129 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1130 goto bad; 1131 } 1132 if (ipt->ipt_ptr < 5) { 1133 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1134 goto bad; 1135 } 1136 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1137 if (++ipt->ipt_oflw == 0) { 1138 code = (u_char *)&ipt->ipt_ptr - 1139 (u_char *)ip; 1140 goto bad; 1141 } 1142 break; 1143 } 1144 cp0 = (cp + ipt->ipt_ptr - 1); 1145 switch (ipt->ipt_flg) { 1146 1147 case IPOPT_TS_TSONLY: 1148 break; 1149 1150 case IPOPT_TS_TSANDADDR: 1151 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1152 sizeof(struct in_addr) > ipt->ipt_len) { 1153 code = (u_char *)&ipt->ipt_ptr - 1154 (u_char *)ip; 1155 goto bad; 1156 } 1157 ipaddr.sin_addr = dst; 1158 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1159 m->m_pkthdr.rcvif)); 1160 if (ia == 0) 1161 continue; 1162 bcopy(&ia->ia_addr.sin_addr, 1163 cp0, sizeof(struct in_addr)); 1164 ipt->ipt_ptr += sizeof(struct in_addr); 1165 break; 1166 1167 case IPOPT_TS_PRESPEC: 1168 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1169 sizeof(struct in_addr) > ipt->ipt_len) { 1170 code = (u_char *)&ipt->ipt_ptr - 1171 (u_char *)ip; 1172 goto bad; 1173 } 1174 bcopy(cp0, &ipaddr.sin_addr, 1175 sizeof(struct in_addr)); 1176 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1177 == NULL) 1178 continue; 1179 ipt->ipt_ptr += sizeof(struct in_addr); 1180 break; 1181 1182 default: 1183 /* XXX can't take &ipt->ipt_flg */ 1184 code = (u_char *)&ipt->ipt_ptr - 1185 (u_char *)ip + 1; 1186 goto bad; 1187 } 1188 ntime = iptime(); 1189 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1190 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1191 sizeof(n_time)); 1192 ipt->ipt_ptr += sizeof(n_time); 1193 } 1194 } 1195 if (forward) { 1196 if (ip_forwsrcrt == 0) { 1197 type = ICMP_UNREACH; 1198 code = ICMP_UNREACH_SRCFAIL; 1199 goto bad; 1200 } 1201 ip_forward(m, 1); 1202 return (1); 1203 } 1204 return (0); 1205bad: 1206 icmp_error(m, type, code, 0, 0); 1207 ipstat.ips_badoptions++; 1208 return (1); 1209} 1210 1211/* 1212 * Given address of next destination (final or next hop), 1213 * return internet address info of interface to be used to get there. 1214 */ 1215struct in_ifaddr * 1216ip_rtaddr(dst) 1217 struct in_addr dst; 1218{ 1219 struct sockaddr_in *sin; 1220 1221 sin = satosin(&ipforward_rt.ro_dst); 1222 1223 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1224 if (ipforward_rt.ro_rt) { 1225 RTFREE(ipforward_rt.ro_rt); 1226 ipforward_rt.ro_rt = 0; 1227 } 1228 sin->sin_family = AF_INET; 1229 sin->sin_len = sizeof(*sin); 1230 sin->sin_addr = dst; 1231 1232 rtalloc(&ipforward_rt); 1233 } 1234 if (ipforward_rt.ro_rt == 0) 1235 return ((struct in_ifaddr *)0); 1236 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1237} 1238 1239/* 1240 * Save incoming source route for use in replies, 1241 * to be picked up later by ip_srcroute if the receiver is interested. 1242 */ 1243void 1244save_rte(option, dst) 1245 u_char *option; 1246 struct in_addr dst; 1247{ 1248 unsigned olen; 1249 1250 olen = option[IPOPT_OLEN]; 1251#ifdef DIAGNOSTIC 1252 if (ipprintfs) 1253 printf("save_rte: olen %d\n", olen); 1254#endif /* 0 */ 1255 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1256 return; 1257 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1258 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1259 ip_srcrt.dst = dst; 1260} 1261 1262/* 1263 * Retrieve incoming source route for use in replies, 1264 * in the same form used by setsockopt. 1265 * The first hop is placed before the options, will be removed later. 1266 */ 1267struct mbuf * 1268ip_srcroute() 1269{ 1270 struct in_addr *p, *q; 1271 struct mbuf *m; 1272 1273 if (ip_nhops == 0) 1274 return ((struct mbuf *)0); 1275 m = m_get(M_DONTWAIT, MT_SOOPTS); 1276 if (m == 0) 1277 return ((struct mbuf *)0); 1278 1279#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1280 1281 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1282 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1283 OPTSIZ; 1284#ifdef DIAGNOSTIC 1285 if (ipprintfs) 1286 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1287#endif 1288 1289 /* 1290 * First save first hop for return route 1291 */ 1292 p = &ip_srcrt.route[ip_nhops - 1]; 1293 *(mtod(m, struct in_addr *)) = *p--; 1294#ifdef DIAGNOSTIC 1295 if (ipprintfs) 1296 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1297#endif 1298 1299 /* 1300 * Copy option fields and padding (nop) to mbuf. 1301 */ 1302 ip_srcrt.nop = IPOPT_NOP; 1303 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1304 bcopy((caddr_t)&ip_srcrt.nop, 1305 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1306 q = (struct in_addr *)(mtod(m, caddr_t) + 1307 sizeof(struct in_addr) + OPTSIZ); 1308#undef OPTSIZ 1309 /* 1310 * Record return path as an IP source route, 1311 * reversing the path (pointers are now aligned). 1312 */ 1313 while (p >= ip_srcrt.route) { 1314#ifdef DIAGNOSTIC 1315 if (ipprintfs) 1316 printf(" %x", ntohl(q->s_addr)); 1317#endif 1318 *q++ = *p--; 1319 } 1320 /* 1321 * Last hop goes to final destination. 1322 */ 1323 *q = ip_srcrt.dst; 1324#ifdef DIAGNOSTIC 1325 if (ipprintfs) 1326 printf(" %x\n", ntohl(q->s_addr)); 1327#endif 1328 return (m); 1329} 1330 1331/* 1332 * Strip out IP options, at higher 1333 * level protocol in the kernel. 1334 * Second argument is buffer to which options 1335 * will be moved, and return value is their length. 1336 * XXX should be deleted; last arg currently ignored. 1337 */ 1338void 1339ip_stripoptions(m, mopt) 1340 struct mbuf *m; 1341 struct mbuf *mopt; 1342{ 1343 int i; 1344 struct ip *ip = mtod(m, struct ip *); 1345 caddr_t opts; 1346 int olen; 1347 1348 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1349 opts = (caddr_t)(ip + 1); 1350 i = m->m_len - (sizeof (struct ip) + olen); 1351 bcopy(opts + olen, opts, (unsigned)i); 1352 m->m_len -= olen; 1353 if (m->m_flags & M_PKTHDR) 1354 m->m_pkthdr.len -= olen; 1355 ip->ip_len -= olen; 1356 ip->ip_hl = sizeof (struct ip) >> 2; 1357} 1358 1359int inetctlerrmap[PRC_NCMDS] = { 1360 0, 0, 0, 0, 1361 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1362 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1363 EMSGSIZE, EHOSTUNREACH, 0, 0, 1364 0, 0, 0, 0, 1365 ENOPROTOOPT 1366}; 1367 1368/* 1369 * Forward a packet. If some error occurs return the sender 1370 * an icmp packet. Note we can't always generate a meaningful 1371 * icmp message because icmp doesn't have a large enough repertoire 1372 * of codes and types. 1373 * 1374 * If not forwarding, just drop the packet. This could be confusing 1375 * if ipforwarding was zero but some routing protocol was advancing 1376 * us as a gateway to somewhere. However, we must let the routing 1377 * protocol deal with that. 1378 * 1379 * The srcrt parameter indicates whether the packet is being forwarded 1380 * via a source route. 1381 */ 1382void 1383ip_forward(m, srcrt) 1384 struct mbuf *m; 1385 int srcrt; 1386{ 1387 struct ip *ip = mtod(m, struct ip *); 1388 struct sockaddr_in *sin; 1389 struct rtentry *rt; 1390 int error, type = 0, code = 0; 1391 struct mbuf *mcopy; 1392 n_long dest; 1393 struct ifnet *destifp; 1394#ifdef IPSEC 1395 struct ifnet dummyifp; 1396#endif 1397 1398 dest = 0; 1399#ifdef DIAGNOSTIC 1400 if (ipprintfs) 1401 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1402 ntohl(ip->ip_src.s_addr), 1403 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1404#endif 1405 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1406 ipstat.ips_cantforward++; 1407 m_freem(m); 1408 return; 1409 } 1410 if (ip->ip_ttl <= IPTTLDEC) { 1411 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1412 return; 1413 } 1414 ip->ip_ttl -= IPTTLDEC; 1415 1416 sin = satosin(&ipforward_rt.ro_dst); 1417 if ((rt = ipforward_rt.ro_rt) == 0 || 1418 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1419 if (ipforward_rt.ro_rt) { 1420 RTFREE(ipforward_rt.ro_rt); 1421 ipforward_rt.ro_rt = 0; 1422 } 1423 sin->sin_family = AF_INET; 1424 sin->sin_len = sizeof(struct sockaddr_in); 1425 sin->sin_addr = ip->ip_dst; 1426 1427 rtalloc(&ipforward_rt); 1428 if (ipforward_rt.ro_rt == 0) { 1429 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1430 return; 1431 } 1432 rt = ipforward_rt.ro_rt; 1433 } 1434 1435 /* 1436 * Save at most 68 bytes of the packet in case 1437 * we need to generate an ICMP message to the src. 1438 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1439 */ 1440 mcopy = m_copym(m, 0, imin((int)ip->ip_len, 68), M_DONTWAIT); 1441 if (mcopy) 1442 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1443 1444 /* 1445 * If forwarding packet using same interface that it came in on, 1446 * perhaps should send a redirect to sender to shortcut a hop. 1447 * Only send redirect if source is sending directly to us, 1448 * and if packet was not source routed (or has any options). 1449 * Also, don't send redirect if forwarding using a default route 1450 * or a route modified by a redirect. 1451 */ 1452 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1453 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1454 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1455 ipsendredirects && !srcrt) { 1456 if (rt->rt_ifa && 1457 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1458 ifatoia(rt->rt_ifa)->ia_subnet) { 1459 if (rt->rt_flags & RTF_GATEWAY) 1460 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1461 else 1462 dest = ip->ip_dst.s_addr; 1463 /* 1464 * Router requirements says to only send host 1465 * redirects. 1466 */ 1467 type = ICMP_REDIRECT; 1468 code = ICMP_REDIRECT_HOST; 1469#ifdef DIAGNOSTIC 1470 if (ipprintfs) 1471 printf("redirect (%d) to %x\n", code, 1472 (u_int32_t)dest); 1473#endif 1474 } 1475 } 1476 1477#ifdef IPSEC 1478 /* Don't lookup socket in forwading case */ 1479 (void)ipsec_setsocket(m, NULL); 1480#endif 1481 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1482 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1483 if (error) 1484 ipstat.ips_cantforward++; 1485 else { 1486 ipstat.ips_forward++; 1487 if (type) 1488 ipstat.ips_redirectsent++; 1489 else { 1490 if (mcopy) { 1491#ifdef GATEWAY 1492 if (mcopy->m_flags & M_CANFASTFWD) 1493 ipflow_create(&ipforward_rt, mcopy); 1494#endif 1495 m_freem(mcopy); 1496 } 1497 return; 1498 } 1499 } 1500 if (mcopy == NULL) 1501 return; 1502 destifp = NULL; 1503 1504 switch (error) { 1505 1506 case 0: /* forwarded, but need redirect */ 1507 /* type, code set above */ 1508 break; 1509 1510 case ENETUNREACH: /* shouldn't happen, checked above */ 1511 case EHOSTUNREACH: 1512 case ENETDOWN: 1513 case EHOSTDOWN: 1514 default: 1515 type = ICMP_UNREACH; 1516 code = ICMP_UNREACH_HOST; 1517 break; 1518 1519 case EMSGSIZE: 1520 type = ICMP_UNREACH; 1521 code = ICMP_UNREACH_NEEDFRAG; 1522#ifndef IPSEC 1523 if (ipforward_rt.ro_rt) 1524 destifp = ipforward_rt.ro_rt->rt_ifp; 1525#else 1526 /* 1527 * If the packet is routed over IPsec tunnel, tell the 1528 * originator the tunnel MTU. 1529 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1530 * XXX quickhack!!! 1531 */ 1532 if (ipforward_rt.ro_rt) { 1533 struct secpolicy *sp; 1534 int ipsecerror; 1535 size_t ipsechdr; 1536 struct route *ro; 1537 1538 sp = ipsec4_getpolicybyaddr(mcopy, 1539 IPSEC_DIR_OUTBOUND, 1540 IP_FORWARDING, 1541 &ipsecerror); 1542 1543 if (sp == NULL) 1544 destifp = ipforward_rt.ro_rt->rt_ifp; 1545 else { 1546 /* count IPsec header size */ 1547 ipsechdr = ipsec4_hdrsiz(mcopy, 1548 IPSEC_DIR_OUTBOUND, 1549 NULL); 1550 1551 /* 1552 * find the correct route for outer IPv4 1553 * header, compute tunnel MTU. 1554 * 1555 * XXX BUG ALERT 1556 * The "dummyifp" code relies upon the fact 1557 * that icmp_error() touches only ifp->if_mtu. 1558 */ 1559 /*XXX*/ 1560 destifp = NULL; 1561 if (sp->req != NULL 1562 && sp->req->sav != NULL 1563 && sp->req->sav->sah != NULL) { 1564 ro = &sp->req->sav->sah->sa_route; 1565 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1566 dummyifp.if_mtu = 1567 ro->ro_rt->rt_ifp->if_mtu; 1568 dummyifp.if_mtu -= ipsechdr; 1569 destifp = &dummyifp; 1570 } 1571 } 1572 1573 key_freesp(sp); 1574 } 1575 } 1576#endif /*IPSEC*/ 1577 ipstat.ips_cantfrag++; 1578 break; 1579 1580 case ENOBUFS: 1581 type = ICMP_SOURCEQUENCH; 1582 code = 0; 1583 break; 1584 } 1585 icmp_error(mcopy, type, code, dest, destifp); 1586} 1587 1588void 1589ip_savecontrol(inp, mp, ip, m) 1590 struct inpcb *inp; 1591 struct mbuf **mp; 1592 struct ip *ip; 1593 struct mbuf *m; 1594{ 1595 1596 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1597 struct timeval tv; 1598 1599 microtime(&tv); 1600 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1601 SCM_TIMESTAMP, SOL_SOCKET); 1602 if (*mp) 1603 mp = &(*mp)->m_next; 1604 } 1605 if (inp->inp_flags & INP_RECVDSTADDR) { 1606 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1607 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1608 if (*mp) 1609 mp = &(*mp)->m_next; 1610 } 1611#ifdef notyet 1612 /* 1613 * XXX 1614 * Moving these out of udp_input() made them even more broken 1615 * than they already were. 1616 * - fenner@parc.xerox.com 1617 */ 1618 /* options were tossed already */ 1619 if (inp->inp_flags & INP_RECVOPTS) { 1620 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1621 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1622 if (*mp) 1623 mp = &(*mp)->m_next; 1624 } 1625 /* ip_srcroute doesn't do what we want here, need to fix */ 1626 if (inp->inp_flags & INP_RECVRETOPTS) { 1627 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1628 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1629 if (*mp) 1630 mp = &(*mp)->m_next; 1631 } 1632#endif 1633 if (inp->inp_flags & INP_RECVIF) { 1634 struct sockaddr_dl sdl; 1635 1636 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1637 sdl.sdl_family = AF_LINK; 1638 sdl.sdl_index = m->m_pkthdr.rcvif ? 1639 m->m_pkthdr.rcvif->if_index : 0; 1640 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1641 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1642 IP_RECVIF, IPPROTO_IP); 1643 if (*mp) 1644 mp = &(*mp)->m_next; 1645 } 1646} 1647 1648int 1649ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1650 int *name; 1651 u_int namelen; 1652 void *oldp; 1653 size_t *oldlenp; 1654 void *newp; 1655 size_t newlen; 1656{ 1657 extern int subnetsarelocal, hostzeroisbroadcast; 1658 1659 int error, old; 1660 1661 /* All sysctl names at this level are terminal. */ 1662 if (namelen != 1) 1663 return (ENOTDIR); 1664 1665 switch (name[0]) { 1666 case IPCTL_FORWARDING: 1667 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1668 case IPCTL_SENDREDIRECTS: 1669 return (sysctl_int(oldp, oldlenp, newp, newlen, 1670 &ipsendredirects)); 1671 case IPCTL_DEFTTL: 1672 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1673#ifdef notyet 1674 case IPCTL_DEFMTU: 1675 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1676#endif 1677 case IPCTL_FORWSRCRT: 1678 /* Don't allow this to change in a secure environment. */ 1679 if (securelevel > 0) 1680 return (sysctl_rdint(oldp, oldlenp, newp, 1681 ip_forwsrcrt)); 1682 else 1683 return (sysctl_int(oldp, oldlenp, newp, newlen, 1684 &ip_forwsrcrt)); 1685 case IPCTL_DIRECTEDBCAST: 1686 return (sysctl_int(oldp, oldlenp, newp, newlen, 1687 &ip_directedbcast)); 1688 case IPCTL_ALLOWSRCRT: 1689 return (sysctl_int(oldp, oldlenp, newp, newlen, 1690 &ip_allowsrcrt)); 1691 case IPCTL_SUBNETSARELOCAL: 1692 return (sysctl_int(oldp, oldlenp, newp, newlen, 1693 &subnetsarelocal)); 1694 case IPCTL_MTUDISC: 1695 error = sysctl_int(oldp, oldlenp, newp, newlen, 1696 &ip_mtudisc); 1697 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1698 ip_mtudisc_timeout_q = 1699 rt_timer_queue_create(ip_mtudisc_timeout); 1700 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1701 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1702 ip_mtudisc_timeout_q = NULL; 1703 } 1704 return error; 1705 case IPCTL_ANONPORTMIN: 1706 old = anonportmin; 1707 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1708 if (anonportmin >= anonportmax || anonportmin < 0 1709 || anonportmin > 65535 1710#ifndef IPNOPRIVPORTS 1711 || anonportmin < IPPORT_RESERVED 1712#endif 1713 ) { 1714 anonportmin = old; 1715 return (EINVAL); 1716 } 1717 return (error); 1718 case IPCTL_ANONPORTMAX: 1719 old = anonportmax; 1720 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1721 if (anonportmin >= anonportmax || anonportmax < 0 1722 || anonportmax > 65535 1723#ifndef IPNOPRIVPORTS 1724 || anonportmax < IPPORT_RESERVED 1725#endif 1726 ) { 1727 anonportmax = old; 1728 return (EINVAL); 1729 } 1730 return (error); 1731 case IPCTL_MTUDISCTIMEOUT: 1732 error = sysctl_int(oldp, oldlenp, newp, newlen, 1733 &ip_mtudisc_timeout); 1734 if (ip_mtudisc_timeout_q != NULL) 1735 rt_timer_queue_change(ip_mtudisc_timeout_q, 1736 ip_mtudisc_timeout); 1737 return (error); 1738#ifdef GATEWAY 1739 case IPCTL_MAXFLOWS: 1740 { 1741 int s; 1742 1743 error = sysctl_int(oldp, oldlenp, newp, newlen, 1744 &ip_maxflows); 1745 s = splsoftnet(); 1746 ipflow_reap(0); 1747 splx(s); 1748 return (error); 1749 } 1750#endif 1751 case IPCTL_HOSTZEROBROADCAST: 1752 return (sysctl_int(oldp, oldlenp, newp, newlen, 1753 &hostzeroisbroadcast)); 1754#if NGIF > 0 1755 case IPCTL_GIF_TTL: 1756 return(sysctl_int(oldp, oldlenp, newp, newlen, 1757 &ip_gif_ttl)); 1758#endif 1759 1760#ifndef IPNOPRIVPORTS 1761 case IPCTL_LOWPORTMIN: 1762 old = lowportmin; 1763 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmin); 1764 if (lowportmin >= lowportmax 1765 || lowportmin > IPPORT_RESERVEDMAX 1766 || lowportmin < IPPORT_RESERVEDMIN 1767 ) { 1768 lowportmin = old; 1769 return (EINVAL); 1770 } 1771 return (error); 1772 case IPCTL_LOWPORTMAX: 1773 old = lowportmax; 1774 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmax); 1775 if (lowportmin >= lowportmax 1776 || lowportmax > IPPORT_RESERVEDMAX 1777 || lowportmax < IPPORT_RESERVEDMIN 1778 ) { 1779 lowportmax = old; 1780 return (EINVAL); 1781 } 1782 return (error); 1783#endif 1784 1785 default: 1786 return (EOPNOTSUPP); 1787 } 1788 /* NOTREACHED */ 1789} 1790