ip_input.c revision 1.97
1/* $NetBSD: ip_input.c,v 1.97 2000/02/11 05:57:58 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 <vm/vm.h> 124#include <sys/sysctl.h> 125 126#include <net/if.h> 127#include <net/if_dl.h> 128#include <net/route.h> 129#include <net/pfil.h> 130 131#include <netinet/in.h> 132#include <netinet/in_systm.h> 133#include <netinet/ip.h> 134#include <netinet/in_pcb.h> 135#include <netinet/in_var.h> 136#include <netinet/ip_var.h> 137#include <netinet/ip_icmp.h> 138/* just for gif_ttl */ 139#include <netinet/in_gif.h> 140#include "gif.h" 141 142#ifdef IPSEC 143#include <netinet6/ipsec.h> 144#include <netkey/key.h> 145#include <netkey/key_debug.h> 146#endif 147 148#ifndef IPFORWARDING 149#ifdef GATEWAY 150#define IPFORWARDING 1 /* forward IP packets not for us */ 151#else /* GATEWAY */ 152#define IPFORWARDING 0 /* don't forward IP packets not for us */ 153#endif /* GATEWAY */ 154#endif /* IPFORWARDING */ 155#ifndef IPSENDREDIRECTS 156#define IPSENDREDIRECTS 1 157#endif 158#ifndef IPFORWSRCRT 159#define IPFORWSRCRT 1 /* forward source-routed packets */ 160#endif 161#ifndef IPALLOWSRCRT 162#define IPALLOWSRCRT 1 /* allow source-routed packets */ 163#endif 164#ifndef IPMTUDISC 165#define IPMTUDISC 0 166#endif 167#ifndef IPMTUDISCTIMEOUT 168#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 169#endif 170 171/* 172 * Note: DIRECTED_BROADCAST is handled this way so that previous 173 * configuration using this option will Just Work. 174 */ 175#ifndef IPDIRECTEDBCAST 176#ifdef DIRECTED_BROADCAST 177#define IPDIRECTEDBCAST 1 178#else 179#define IPDIRECTEDBCAST 0 180#endif /* DIRECTED_BROADCAST */ 181#endif /* IPDIRECTEDBCAST */ 182int ipforwarding = IPFORWARDING; 183int ipsendredirects = IPSENDREDIRECTS; 184int ip_defttl = IPDEFTTL; 185int ip_forwsrcrt = IPFORWSRCRT; 186int ip_directedbcast = IPDIRECTEDBCAST; 187int ip_allowsrcrt = IPALLOWSRCRT; 188int ip_mtudisc = IPMTUDISC; 189u_int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 190#ifdef DIAGNOSTIC 191int ipprintfs = 0; 192#endif 193 194struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 195 196extern struct domain inetdomain; 197extern struct protosw inetsw[]; 198u_char ip_protox[IPPROTO_MAX]; 199int ipqmaxlen = IFQ_MAXLEN; 200struct in_ifaddrhead in_ifaddr; 201struct in_ifaddrhashhead *in_ifaddrhashtbl; 202struct ifqueue ipintrq; 203struct ipstat ipstat; 204u_int16_t ip_id; 205int ip_defttl; 206 207struct ipqhead ipq; 208int ipq_locked; 209 210static __inline int ipq_lock_try __P((void)); 211static __inline void ipq_unlock __P((void)); 212 213static __inline int 214ipq_lock_try() 215{ 216 int s; 217 218 s = splimp(); 219 if (ipq_locked) { 220 splx(s); 221 return (0); 222 } 223 ipq_locked = 1; 224 splx(s); 225 return (1); 226} 227 228static __inline void 229ipq_unlock() 230{ 231 int s; 232 233 s = splimp(); 234 ipq_locked = 0; 235 splx(s); 236} 237 238#ifdef DIAGNOSTIC 239#define IPQ_LOCK() \ 240do { \ 241 if (ipq_lock_try() == 0) { \ 242 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 243 panic("ipq_lock"); \ 244 } \ 245} while (0) 246#define IPQ_LOCK_CHECK() \ 247do { \ 248 if (ipq_locked == 0) { \ 249 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 250 panic("ipq lock check"); \ 251 } \ 252} while (0) 253#else 254#define IPQ_LOCK() (void) ipq_lock_try() 255#define IPQ_LOCK_CHECK() /* nothing */ 256#endif 257 258#define IPQ_UNLOCK() ipq_unlock() 259 260struct pool ipqent_pool; 261 262/* 263 * We need to save the IP options in case a protocol wants to respond 264 * to an incoming packet over the same route if the packet got here 265 * using IP source routing. This allows connection establishment and 266 * maintenance when the remote end is on a network that is not known 267 * to us. 268 */ 269int ip_nhops = 0; 270static struct ip_srcrt { 271 struct in_addr dst; /* final destination */ 272 char nop; /* one NOP to align */ 273 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 274 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 275} ip_srcrt; 276 277static void save_rte __P((u_char *, struct in_addr)); 278 279/* 280 * IP initialization: fill in IP protocol switch table. 281 * All protocols not implemented in kernel go to raw IP protocol handler. 282 */ 283void 284ip_init() 285{ 286 register struct protosw *pr; 287 register int i; 288 289 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 290 0, NULL, NULL, M_IPQ); 291 292 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 293 if (pr == 0) 294 panic("ip_init"); 295 for (i = 0; i < IPPROTO_MAX; i++) 296 ip_protox[i] = pr - inetsw; 297 for (pr = inetdomain.dom_protosw; 298 pr < inetdomain.dom_protoswNPROTOSW; pr++) 299 if (pr->pr_domain->dom_family == PF_INET && 300 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 301 ip_protox[pr->pr_protocol] = pr - inetsw; 302 LIST_INIT(&ipq); 303 ip_id = time.tv_sec & 0xffff; 304 ipintrq.ifq_maxlen = ipqmaxlen; 305 TAILQ_INIT(&in_ifaddr); 306 in_ifaddrhashtbl = 307 hashinit(IN_IFADDR_HASH_SIZE, M_IFADDR, M_WAITOK, &in_ifaddrhash); 308 if (ip_mtudisc != 0) 309 ip_mtudisc_timeout_q = 310 rt_timer_queue_create(ip_mtudisc_timeout); 311#ifdef GATEWAY 312 ipflow_init(); 313#endif 314} 315 316struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 317struct route ipforward_rt; 318 319/* 320 * IP software interrupt routine 321 */ 322void 323ipintr() 324{ 325 int s; 326 struct mbuf *m; 327 328 while (1) { 329 s = splimp(); 330 IF_DEQUEUE(&ipintrq, m); 331 splx(s); 332 if (m == 0) 333 return; 334 ip_input(m); 335 } 336} 337 338/* 339 * Ip input routine. Checksum and byte swap header. If fragmented 340 * try to reassemble. Process options. Pass to next level. 341 */ 342void 343ip_input(struct mbuf *m) 344{ 345 register struct ip *ip = NULL; 346 register struct ipq *fp; 347 register struct in_ifaddr *ia; 348 register struct ifaddr *ifa; 349 struct ipqent *ipqe; 350 int hlen = 0, mff, len; 351#ifdef PFIL_HOOKS 352 struct packet_filter_hook *pfh; 353 struct mbuf *m0; 354 int rv; 355#endif /* PFIL_HOOKS */ 356 357#ifdef DIAGNOSTIC 358 if ((m->m_flags & M_PKTHDR) == 0) 359 panic("ipintr no HDR"); 360#endif 361#ifdef IPSEC 362 /* 363 * should the inner packet be considered authentic? 364 * see comment in ah4_input(). 365 */ 366 if (m) { 367 m->m_flags &= ~M_AUTHIPHDR; 368 m->m_flags &= ~M_AUTHIPDGM; 369 } 370#endif 371 /* 372 * If no IP addresses have been set yet but the interfaces 373 * are receiving, can't do anything with incoming packets yet. 374 */ 375 if (in_ifaddr.tqh_first == 0) 376 goto bad; 377 ipstat.ips_total++; 378 if (m->m_len < sizeof (struct ip) && 379 (m = m_pullup(m, sizeof (struct ip))) == 0) { 380 ipstat.ips_toosmall++; 381 return; 382 } 383 ip = mtod(m, struct ip *); 384 if (ip->ip_v != IPVERSION) { 385 ipstat.ips_badvers++; 386 goto bad; 387 } 388 hlen = ip->ip_hl << 2; 389 if (hlen < sizeof(struct ip)) { /* minimum header length */ 390 ipstat.ips_badhlen++; 391 goto bad; 392 } 393 if (hlen > m->m_len) { 394 if ((m = m_pullup(m, hlen)) == 0) { 395 ipstat.ips_badhlen++; 396 return; 397 } 398 ip = mtod(m, struct ip *); 399 } 400 /* 401 * we drop packets that have a multicast address as source 402 * as wanted by rfc 1112 403 */ 404 if (IN_MULTICAST(ip->ip_src.s_addr)) { 405 goto bad; 406 } 407 408 if (in_cksum(m, hlen) != 0) { 409 ipstat.ips_badsum++; 410 goto bad; 411 } 412 413 /* 414 * Convert fields to host representation. 415 */ 416 NTOHS(ip->ip_len); 417 NTOHS(ip->ip_off); 418 len = ip->ip_len; 419 420 /* 421 * Check for additional length bogosity 422 */ 423 if (len < hlen) { 424 ipstat.ips_badlen++; 425 goto bad; 426 } 427 428 /* 429 * Check that the amount of data in the buffers 430 * is as at least much as the IP header would have us expect. 431 * Trim mbufs if longer than we expect. 432 * Drop packet if shorter than we expect. 433 */ 434 if (m->m_pkthdr.len < len) { 435 ipstat.ips_tooshort++; 436 goto bad; 437 } 438 if (m->m_pkthdr.len > len) { 439 if (m->m_len == m->m_pkthdr.len) { 440 m->m_len = len; 441 m->m_pkthdr.len = len; 442 } else 443 m_adj(m, len - m->m_pkthdr.len); 444 } 445 446#ifdef IPSEC 447 /* ipflow (IP fast fowarding) is not compatible with IPsec. */ 448 m->m_flags &= ~M_CANFASTFWD; 449#else 450 /* 451 * Assume that we can create a fast-forward IP flow entry 452 * based on this packet. 453 */ 454 m->m_flags |= M_CANFASTFWD; 455#endif 456 457#ifdef PFIL_HOOKS 458 /* 459 * Run through list of hooks for input packets. If there are any 460 * filters which require that additional packets in the flow are 461 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 462 * Note that filters must _never_ set this flag, as another filter 463 * in the list may have previously cleared it. 464 */ 465 m0 = m; 466 for (pfh = pfil_hook_get(PFIL_IN); pfh; pfh = pfh->pfil_link.tqe_next) 467 if (pfh->pfil_func) { 468 rv = pfh->pfil_func(ip, hlen, m->m_pkthdr.rcvif, 0, &m0); 469 if (rv) 470 return; 471 m = m0; 472 if (m == NULL) 473 return; 474 ip = mtod(m, struct ip *); 475 } 476#endif /* PFIL_HOOKS */ 477 478 /* 479 * Process options and, if not destined for us, 480 * ship it on. ip_dooptions returns 1 when an 481 * error was detected (causing an icmp message 482 * to be sent and the original packet to be freed). 483 */ 484 ip_nhops = 0; /* for source routed packets */ 485 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 486 return; 487 488 /* 489 * Check our list of addresses, to see if the packet is for us. 490 */ 491 for (ia = IN_IFADDR_HASH(ip->ip_dst.s_addr).lh_first; 492 ia != NULL; 493 ia = ia->ia_hash.le_next) { 494 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 495 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 496 break; 497 else { 498 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 499 0, m->m_pkthdr.rcvif); 500 return; 501 } 502 } 503 } 504 if (ia != NULL) 505 goto ours; 506 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 507 for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first; 508 ifa != NULL; ifa = ifa->ifa_list.tqe_next) { 509 if (ifa->ifa_addr->sa_family != AF_INET) continue; 510 ia = ifatoia(ifa); 511 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 512 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 513 /* 514 * Look for all-0's host part (old broadcast addr), 515 * either for subnet or net. 516 */ 517 ip->ip_dst.s_addr == ia->ia_subnet || 518 ip->ip_dst.s_addr == ia->ia_net) 519 goto ours; 520 /* 521 * An interface with IP address zero accepts 522 * all packets that arrive on that interface. 523 */ 524 if (in_nullhost(ia->ia_addr.sin_addr)) 525 goto ours; 526 } 527 } 528 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 529 struct in_multi *inm; 530#ifdef MROUTING 531 extern struct socket *ip_mrouter; 532 533 if (m->m_flags & M_EXT) { 534 if ((m = m_pullup(m, hlen)) == 0) { 535 ipstat.ips_toosmall++; 536 return; 537 } 538 ip = mtod(m, struct ip *); 539 } 540 541 if (ip_mrouter) { 542 /* 543 * If we are acting as a multicast router, all 544 * incoming multicast packets are passed to the 545 * kernel-level multicast forwarding function. 546 * The packet is returned (relatively) intact; if 547 * ip_mforward() returns a non-zero value, the packet 548 * must be discarded, else it may be accepted below. 549 * 550 * (The IP ident field is put in the same byte order 551 * as expected when ip_mforward() is called from 552 * ip_output().) 553 */ 554 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 555 ipstat.ips_cantforward++; 556 m_freem(m); 557 return; 558 } 559 560 /* 561 * The process-level routing demon needs to receive 562 * all multicast IGMP packets, whether or not this 563 * host belongs to their destination groups. 564 */ 565 if (ip->ip_p == IPPROTO_IGMP) 566 goto ours; 567 ipstat.ips_forward++; 568 } 569#endif 570 /* 571 * See if we belong to the destination multicast group on the 572 * arrival interface. 573 */ 574 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 575 if (inm == NULL) { 576 ipstat.ips_cantforward++; 577 m_freem(m); 578 return; 579 } 580 goto ours; 581 } 582 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 583 in_nullhost(ip->ip_dst)) 584 goto ours; 585 586 /* 587 * Not for us; forward if possible and desirable. 588 */ 589 if (ipforwarding == 0) { 590 ipstat.ips_cantforward++; 591 m_freem(m); 592 } else 593 ip_forward(m, 0); 594 return; 595 596ours: 597 /* 598 * If offset or IP_MF are set, must reassemble. 599 * Otherwise, nothing need be done. 600 * (We could look in the reassembly queue to see 601 * if the packet was previously fragmented, 602 * but it's not worth the time; just let them time out.) 603 */ 604 if (ip->ip_off & ~(IP_DF|IP_RF)) { 605 /* 606 * Look for queue of fragments 607 * of this datagram. 608 */ 609 IPQ_LOCK(); 610 for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next) 611 if (ip->ip_id == fp->ipq_id && 612 in_hosteq(ip->ip_src, fp->ipq_src) && 613 in_hosteq(ip->ip_dst, fp->ipq_dst) && 614 ip->ip_p == fp->ipq_p) 615 goto found; 616 fp = 0; 617found: 618 619 /* 620 * Adjust ip_len to not reflect header, 621 * set ipqe_mff if more fragments are expected, 622 * convert offset of this to bytes. 623 */ 624 ip->ip_len -= hlen; 625 mff = (ip->ip_off & IP_MF) != 0; 626 if (mff) { 627 /* 628 * Make sure that fragments have a data length 629 * that's a non-zero multiple of 8 bytes. 630 */ 631 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 632 ipstat.ips_badfrags++; 633 IPQ_UNLOCK(); 634 goto bad; 635 } 636 } 637 ip->ip_off <<= 3; 638 639 /* 640 * If datagram marked as having more fragments 641 * or if this is not the first fragment, 642 * attempt reassembly; if it succeeds, proceed. 643 */ 644 if (mff || ip->ip_off) { 645 ipstat.ips_fragments++; 646 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 647 if (ipqe == NULL) { 648 ipstat.ips_rcvmemdrop++; 649 IPQ_UNLOCK(); 650 goto bad; 651 } 652 ipqe->ipqe_mff = mff; 653 ipqe->ipqe_m = m; 654 ipqe->ipqe_ip = ip; 655 m = ip_reass(ipqe, fp); 656 if (m == 0) { 657 IPQ_UNLOCK(); 658 return; 659 } 660 ipstat.ips_reassembled++; 661 ip = mtod(m, struct ip *); 662 hlen = ip->ip_hl << 2; 663 ip->ip_len += hlen; 664 } else 665 if (fp) 666 ip_freef(fp); 667 IPQ_UNLOCK(); 668 } 669 670 /* 671 * Switch out to protocol's input routine. 672 */ 673#if IFA_STATS 674 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 675#endif 676 ipstat.ips_delivered++; 677 { 678 int off = hlen, nh = ip->ip_p; 679 680 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 681 return; 682 } 683bad: 684 m_freem(m); 685} 686 687/* 688 * Take incoming datagram fragment and try to 689 * reassemble it into whole datagram. If a chain for 690 * reassembly of this datagram already exists, then it 691 * is given as fp; otherwise have to make a chain. 692 */ 693struct mbuf * 694ip_reass(ipqe, fp) 695 register struct ipqent *ipqe; 696 register struct ipq *fp; 697{ 698 register struct mbuf *m = ipqe->ipqe_m; 699 register struct ipqent *nq, *p, *q; 700 struct ip *ip; 701 struct mbuf *t; 702 int hlen = ipqe->ipqe_ip->ip_hl << 2; 703 int i, next; 704 705 IPQ_LOCK_CHECK(); 706 707 /* 708 * Presence of header sizes in mbufs 709 * would confuse code below. 710 */ 711 m->m_data += hlen; 712 m->m_len -= hlen; 713 714 /* 715 * If first fragment to arrive, create a reassembly queue. 716 */ 717 if (fp == 0) { 718 MALLOC(fp, struct ipq *, sizeof (struct ipq), 719 M_FTABLE, M_NOWAIT); 720 if (fp == NULL) 721 goto dropfrag; 722 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 723 fp->ipq_ttl = IPFRAGTTL; 724 fp->ipq_p = ipqe->ipqe_ip->ip_p; 725 fp->ipq_id = ipqe->ipqe_ip->ip_id; 726 LIST_INIT(&fp->ipq_fragq); 727 fp->ipq_src = ipqe->ipqe_ip->ip_src; 728 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 729 p = NULL; 730 goto insert; 731 } 732 733 /* 734 * Find a segment which begins after this one does. 735 */ 736 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 737 p = q, q = q->ipqe_q.le_next) 738 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 739 break; 740 741 /* 742 * If there is a preceding segment, it may provide some of 743 * our data already. If so, drop the data from the incoming 744 * segment. If it provides all of our data, drop us. 745 */ 746 if (p != NULL) { 747 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 748 ipqe->ipqe_ip->ip_off; 749 if (i > 0) { 750 if (i >= ipqe->ipqe_ip->ip_len) 751 goto dropfrag; 752 m_adj(ipqe->ipqe_m, i); 753 ipqe->ipqe_ip->ip_off += i; 754 ipqe->ipqe_ip->ip_len -= i; 755 } 756 } 757 758 /* 759 * While we overlap succeeding segments trim them or, 760 * if they are completely covered, dequeue them. 761 */ 762 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 763 q->ipqe_ip->ip_off; q = nq) { 764 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 765 q->ipqe_ip->ip_off; 766 if (i < q->ipqe_ip->ip_len) { 767 q->ipqe_ip->ip_len -= i; 768 q->ipqe_ip->ip_off += i; 769 m_adj(q->ipqe_m, i); 770 break; 771 } 772 nq = q->ipqe_q.le_next; 773 m_freem(q->ipqe_m); 774 LIST_REMOVE(q, ipqe_q); 775 pool_put(&ipqent_pool, q); 776 } 777 778insert: 779 /* 780 * Stick new segment in its place; 781 * check for complete reassembly. 782 */ 783 if (p == NULL) { 784 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 785 } else { 786 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 787 } 788 next = 0; 789 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 790 p = q, q = q->ipqe_q.le_next) { 791 if (q->ipqe_ip->ip_off != next) 792 return (0); 793 next += q->ipqe_ip->ip_len; 794 } 795 if (p->ipqe_mff) 796 return (0); 797 798 /* 799 * Reassembly is complete. Check for a bogus message size and 800 * concatenate fragments. 801 */ 802 q = fp->ipq_fragq.lh_first; 803 ip = q->ipqe_ip; 804 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 805 ipstat.ips_toolong++; 806 ip_freef(fp); 807 return (0); 808 } 809 m = q->ipqe_m; 810 t = m->m_next; 811 m->m_next = 0; 812 m_cat(m, t); 813 nq = q->ipqe_q.le_next; 814 pool_put(&ipqent_pool, q); 815 for (q = nq; q != NULL; q = nq) { 816 t = q->ipqe_m; 817 nq = q->ipqe_q.le_next; 818 pool_put(&ipqent_pool, q); 819 m_cat(m, t); 820 } 821 822 /* 823 * Create header for new ip packet by 824 * modifying header of first packet; 825 * dequeue and discard fragment reassembly header. 826 * Make header visible. 827 */ 828 ip->ip_len = next; 829 ip->ip_ttl = 0; /* xxx */ 830 ip->ip_sum = 0; 831 ip->ip_src = fp->ipq_src; 832 ip->ip_dst = fp->ipq_dst; 833 LIST_REMOVE(fp, ipq_q); 834 FREE(fp, M_FTABLE); 835 m->m_len += (ip->ip_hl << 2); 836 m->m_data -= (ip->ip_hl << 2); 837 /* some debugging cruft by sklower, below, will go away soon */ 838 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 839 register int plen = 0; 840 for (t = m; t; t = t->m_next) 841 plen += t->m_len; 842 m->m_pkthdr.len = plen; 843 } 844 return (m); 845 846dropfrag: 847 ipstat.ips_fragdropped++; 848 m_freem(m); 849 pool_put(&ipqent_pool, ipqe); 850 return (0); 851} 852 853/* 854 * Free a fragment reassembly header and all 855 * associated datagrams. 856 */ 857void 858ip_freef(fp) 859 struct ipq *fp; 860{ 861 register struct ipqent *q, *p; 862 863 IPQ_LOCK_CHECK(); 864 865 for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { 866 p = q->ipqe_q.le_next; 867 m_freem(q->ipqe_m); 868 LIST_REMOVE(q, ipqe_q); 869 pool_put(&ipqent_pool, q); 870 } 871 LIST_REMOVE(fp, ipq_q); 872 FREE(fp, M_FTABLE); 873} 874 875/* 876 * IP timer processing; 877 * if a timer expires on a reassembly 878 * queue, discard it. 879 */ 880void 881ip_slowtimo() 882{ 883 register struct ipq *fp, *nfp; 884 int s = splsoftnet(); 885 886 IPQ_LOCK(); 887 for (fp = ipq.lh_first; fp != NULL; fp = nfp) { 888 nfp = fp->ipq_q.le_next; 889 if (--fp->ipq_ttl == 0) { 890 ipstat.ips_fragtimeout++; 891 ip_freef(fp); 892 } 893 } 894 IPQ_UNLOCK(); 895#ifdef GATEWAY 896 ipflow_slowtimo(); 897#endif 898 splx(s); 899} 900 901/* 902 * Drain off all datagram fragments. 903 */ 904void 905ip_drain() 906{ 907 908 /* 909 * We may be called from a device's interrupt context. If 910 * the ipq is already busy, just bail out now. 911 */ 912 if (ipq_lock_try() == 0) 913 return; 914 915 while (ipq.lh_first != NULL) { 916 ipstat.ips_fragdropped++; 917 ip_freef(ipq.lh_first); 918 } 919 920 IPQ_UNLOCK(); 921} 922 923/* 924 * Do option processing on a datagram, 925 * possibly discarding it if bad options are encountered, 926 * or forwarding it if source-routed. 927 * Returns 1 if packet has been forwarded/freed, 928 * 0 if the packet should be processed further. 929 */ 930int 931ip_dooptions(m) 932 struct mbuf *m; 933{ 934 register struct ip *ip = mtod(m, struct ip *); 935 register u_char *cp; 936 register struct ip_timestamp *ipt; 937 register struct in_ifaddr *ia; 938 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 939 struct in_addr *sin, dst; 940 n_time ntime; 941 942 dst = ip->ip_dst; 943 cp = (u_char *)(ip + 1); 944 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 945 for (; cnt > 0; cnt -= optlen, cp += optlen) { 946 opt = cp[IPOPT_OPTVAL]; 947 if (opt == IPOPT_EOL) 948 break; 949 if (opt == IPOPT_NOP) 950 optlen = 1; 951 else { 952 optlen = cp[IPOPT_OLEN]; 953 if (optlen <= 0 || optlen > cnt) { 954 code = &cp[IPOPT_OLEN] - (u_char *)ip; 955 goto bad; 956 } 957 } 958 switch (opt) { 959 960 default: 961 break; 962 963 /* 964 * Source routing with record. 965 * Find interface with current destination address. 966 * If none on this machine then drop if strictly routed, 967 * or do nothing if loosely routed. 968 * Record interface address and bring up next address 969 * component. If strictly routed make sure next 970 * address is on directly accessible net. 971 */ 972 case IPOPT_LSRR: 973 case IPOPT_SSRR: 974 if (ip_allowsrcrt == 0) { 975 type = ICMP_UNREACH; 976 code = ICMP_UNREACH_NET_PROHIB; 977 goto bad; 978 } 979 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 980 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 981 goto bad; 982 } 983 ipaddr.sin_addr = ip->ip_dst; 984 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 985 if (ia == 0) { 986 if (opt == IPOPT_SSRR) { 987 type = ICMP_UNREACH; 988 code = ICMP_UNREACH_SRCFAIL; 989 goto bad; 990 } 991 /* 992 * Loose routing, and not at next destination 993 * yet; nothing to do except forward. 994 */ 995 break; 996 } 997 off--; /* 0 origin */ 998 if (off > optlen - sizeof(struct in_addr)) { 999 /* 1000 * End of source route. Should be for us. 1001 */ 1002 save_rte(cp, ip->ip_src); 1003 break; 1004 } 1005 /* 1006 * locate outgoing interface 1007 */ 1008 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1009 sizeof(ipaddr.sin_addr)); 1010 if (opt == IPOPT_SSRR) 1011 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1012 else 1013 ia = ip_rtaddr(ipaddr.sin_addr); 1014 if (ia == 0) { 1015 type = ICMP_UNREACH; 1016 code = ICMP_UNREACH_SRCFAIL; 1017 goto bad; 1018 } 1019 ip->ip_dst = ipaddr.sin_addr; 1020 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1021 (caddr_t)(cp + off), sizeof(struct in_addr)); 1022 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1023 /* 1024 * Let ip_intr's mcast routing check handle mcast pkts 1025 */ 1026 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1027 break; 1028 1029 case IPOPT_RR: 1030 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1031 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1032 goto bad; 1033 } 1034 /* 1035 * If no space remains, ignore. 1036 */ 1037 off--; /* 0 origin */ 1038 if (off > optlen - sizeof(struct in_addr)) 1039 break; 1040 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1041 sizeof(ipaddr.sin_addr)); 1042 /* 1043 * locate outgoing interface; if we're the destination, 1044 * use the incoming interface (should be same). 1045 */ 1046 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1047 == NULL && 1048 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1049 type = ICMP_UNREACH; 1050 code = ICMP_UNREACH_HOST; 1051 goto bad; 1052 } 1053 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1054 (caddr_t)(cp + off), sizeof(struct in_addr)); 1055 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1056 break; 1057 1058 case IPOPT_TS: 1059 code = cp - (u_char *)ip; 1060 ipt = (struct ip_timestamp *)cp; 1061 if (ipt->ipt_len < 5) 1062 goto bad; 1063 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1064 if (++ipt->ipt_oflw == 0) 1065 goto bad; 1066 break; 1067 } 1068 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1069 switch (ipt->ipt_flg) { 1070 1071 case IPOPT_TS_TSONLY: 1072 break; 1073 1074 case IPOPT_TS_TSANDADDR: 1075 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1076 sizeof(struct in_addr) > ipt->ipt_len) 1077 goto bad; 1078 ipaddr.sin_addr = dst; 1079 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1080 m->m_pkthdr.rcvif)); 1081 if (ia == 0) 1082 continue; 1083 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1084 (caddr_t)sin, sizeof(struct in_addr)); 1085 ipt->ipt_ptr += sizeof(struct in_addr); 1086 break; 1087 1088 case IPOPT_TS_PRESPEC: 1089 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1090 sizeof(struct in_addr) > ipt->ipt_len) 1091 goto bad; 1092 bcopy((caddr_t)sin, (caddr_t)&ipaddr.sin_addr, 1093 sizeof(struct in_addr)); 1094 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1095 == NULL) 1096 continue; 1097 ipt->ipt_ptr += sizeof(struct in_addr); 1098 break; 1099 1100 default: 1101 goto bad; 1102 } 1103 ntime = iptime(); 1104 bcopy((caddr_t)&ntime, (caddr_t)cp + ipt->ipt_ptr - 1, 1105 sizeof(n_time)); 1106 ipt->ipt_ptr += sizeof(n_time); 1107 } 1108 } 1109 if (forward) { 1110 if (ip_forwsrcrt == 0) { 1111 type = ICMP_UNREACH; 1112 code = ICMP_UNREACH_SRCFAIL; 1113 goto bad; 1114 } 1115 ip_forward(m, 1); 1116 return (1); 1117 } 1118 return (0); 1119bad: 1120 icmp_error(m, type, code, 0, 0); 1121 ipstat.ips_badoptions++; 1122 return (1); 1123} 1124 1125/* 1126 * Given address of next destination (final or next hop), 1127 * return internet address info of interface to be used to get there. 1128 */ 1129struct in_ifaddr * 1130ip_rtaddr(dst) 1131 struct in_addr dst; 1132{ 1133 register struct sockaddr_in *sin; 1134 1135 sin = satosin(&ipforward_rt.ro_dst); 1136 1137 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1138 if (ipforward_rt.ro_rt) { 1139 RTFREE(ipforward_rt.ro_rt); 1140 ipforward_rt.ro_rt = 0; 1141 } 1142 sin->sin_family = AF_INET; 1143 sin->sin_len = sizeof(*sin); 1144 sin->sin_addr = dst; 1145 1146 rtalloc(&ipforward_rt); 1147 } 1148 if (ipforward_rt.ro_rt == 0) 1149 return ((struct in_ifaddr *)0); 1150 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1151} 1152 1153/* 1154 * Save incoming source route for use in replies, 1155 * to be picked up later by ip_srcroute if the receiver is interested. 1156 */ 1157void 1158save_rte(option, dst) 1159 u_char *option; 1160 struct in_addr dst; 1161{ 1162 unsigned olen; 1163 1164 olen = option[IPOPT_OLEN]; 1165#ifdef DIAGNOSTIC 1166 if (ipprintfs) 1167 printf("save_rte: olen %d\n", olen); 1168#endif /* 0 */ 1169 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1170 return; 1171 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1172 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1173 ip_srcrt.dst = dst; 1174} 1175 1176/* 1177 * Retrieve incoming source route for use in replies, 1178 * in the same form used by setsockopt. 1179 * The first hop is placed before the options, will be removed later. 1180 */ 1181struct mbuf * 1182ip_srcroute() 1183{ 1184 register struct in_addr *p, *q; 1185 register struct mbuf *m; 1186 1187 if (ip_nhops == 0) 1188 return ((struct mbuf *)0); 1189 m = m_get(M_DONTWAIT, MT_SOOPTS); 1190 if (m == 0) 1191 return ((struct mbuf *)0); 1192 1193#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1194 1195 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1196 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1197 OPTSIZ; 1198#ifdef DIAGNOSTIC 1199 if (ipprintfs) 1200 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1201#endif 1202 1203 /* 1204 * First save first hop for return route 1205 */ 1206 p = &ip_srcrt.route[ip_nhops - 1]; 1207 *(mtod(m, struct in_addr *)) = *p--; 1208#ifdef DIAGNOSTIC 1209 if (ipprintfs) 1210 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1211#endif 1212 1213 /* 1214 * Copy option fields and padding (nop) to mbuf. 1215 */ 1216 ip_srcrt.nop = IPOPT_NOP; 1217 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1218 bcopy((caddr_t)&ip_srcrt.nop, 1219 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1220 q = (struct in_addr *)(mtod(m, caddr_t) + 1221 sizeof(struct in_addr) + OPTSIZ); 1222#undef OPTSIZ 1223 /* 1224 * Record return path as an IP source route, 1225 * reversing the path (pointers are now aligned). 1226 */ 1227 while (p >= ip_srcrt.route) { 1228#ifdef DIAGNOSTIC 1229 if (ipprintfs) 1230 printf(" %x", ntohl(q->s_addr)); 1231#endif 1232 *q++ = *p--; 1233 } 1234 /* 1235 * Last hop goes to final destination. 1236 */ 1237 *q = ip_srcrt.dst; 1238#ifdef DIAGNOSTIC 1239 if (ipprintfs) 1240 printf(" %x\n", ntohl(q->s_addr)); 1241#endif 1242 return (m); 1243} 1244 1245/* 1246 * Strip out IP options, at higher 1247 * level protocol in the kernel. 1248 * Second argument is buffer to which options 1249 * will be moved, and return value is their length. 1250 * XXX should be deleted; last arg currently ignored. 1251 */ 1252void 1253ip_stripoptions(m, mopt) 1254 register struct mbuf *m; 1255 struct mbuf *mopt; 1256{ 1257 register int i; 1258 struct ip *ip = mtod(m, struct ip *); 1259 register caddr_t opts; 1260 int olen; 1261 1262 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1263 opts = (caddr_t)(ip + 1); 1264 i = m->m_len - (sizeof (struct ip) + olen); 1265 bcopy(opts + olen, opts, (unsigned)i); 1266 m->m_len -= olen; 1267 if (m->m_flags & M_PKTHDR) 1268 m->m_pkthdr.len -= olen; 1269 ip->ip_len -= olen; 1270 ip->ip_hl = sizeof (struct ip) >> 2; 1271} 1272 1273int inetctlerrmap[PRC_NCMDS] = { 1274 0, 0, 0, 0, 1275 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1276 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1277 EMSGSIZE, EHOSTUNREACH, 0, 0, 1278 0, 0, 0, 0, 1279 ENOPROTOOPT 1280}; 1281 1282/* 1283 * Forward a packet. If some error occurs return the sender 1284 * an icmp packet. Note we can't always generate a meaningful 1285 * icmp message because icmp doesn't have a large enough repertoire 1286 * of codes and types. 1287 * 1288 * If not forwarding, just drop the packet. This could be confusing 1289 * if ipforwarding was zero but some routing protocol was advancing 1290 * us as a gateway to somewhere. However, we must let the routing 1291 * protocol deal with that. 1292 * 1293 * The srcrt parameter indicates whether the packet is being forwarded 1294 * via a source route. 1295 */ 1296void 1297ip_forward(m, srcrt) 1298 struct mbuf *m; 1299 int srcrt; 1300{ 1301 register struct ip *ip = mtod(m, struct ip *); 1302 register struct sockaddr_in *sin; 1303 register struct rtentry *rt; 1304 int error, type = 0, code = 0; 1305 struct mbuf *mcopy; 1306 n_long dest; 1307 struct ifnet *destifp; 1308#ifdef IPSEC 1309 struct ifnet dummyifp; 1310#endif 1311 1312 dest = 0; 1313#ifdef DIAGNOSTIC 1314 if (ipprintfs) 1315 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1316 ntohl(ip->ip_src.s_addr), 1317 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1318#endif 1319 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1320 ipstat.ips_cantforward++; 1321 m_freem(m); 1322 return; 1323 } 1324 if (ip->ip_ttl <= IPTTLDEC) { 1325 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1326 return; 1327 } 1328 ip->ip_ttl -= IPTTLDEC; 1329 1330 sin = satosin(&ipforward_rt.ro_dst); 1331 if ((rt = ipforward_rt.ro_rt) == 0 || 1332 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1333 if (ipforward_rt.ro_rt) { 1334 RTFREE(ipforward_rt.ro_rt); 1335 ipforward_rt.ro_rt = 0; 1336 } 1337 sin->sin_family = AF_INET; 1338 sin->sin_len = sizeof(struct sockaddr_in); 1339 sin->sin_addr = ip->ip_dst; 1340 1341 rtalloc(&ipforward_rt); 1342 if (ipforward_rt.ro_rt == 0) { 1343 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1344 return; 1345 } 1346 rt = ipforward_rt.ro_rt; 1347 } 1348 1349 /* 1350 * Save at most 68 bytes of the packet in case 1351 * we need to generate an ICMP message to the src. 1352 */ 1353 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 68)); 1354 1355 /* 1356 * If forwarding packet using same interface that it came in on, 1357 * perhaps should send a redirect to sender to shortcut a hop. 1358 * Only send redirect if source is sending directly to us, 1359 * and if packet was not source routed (or has any options). 1360 * Also, don't send redirect if forwarding using a default route 1361 * or a route modified by a redirect. 1362 */ 1363 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1364 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1365 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1366 ipsendredirects && !srcrt) { 1367 if (rt->rt_ifa && 1368 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1369 ifatoia(rt->rt_ifa)->ia_subnet) { 1370 if (rt->rt_flags & RTF_GATEWAY) 1371 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1372 else 1373 dest = ip->ip_dst.s_addr; 1374 /* 1375 * Router requirements says to only send host 1376 * redirects. 1377 */ 1378 type = ICMP_REDIRECT; 1379 code = ICMP_REDIRECT_HOST; 1380#ifdef DIAGNOSTIC 1381 if (ipprintfs) 1382 printf("redirect (%d) to %x\n", code, 1383 (u_int32_t)dest); 1384#endif 1385 } 1386 } 1387 1388#ifdef IPSEC 1389 m->m_pkthdr.rcvif = NULL; 1390#endif /*IPSEC*/ 1391 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1392 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1393 if (error) 1394 ipstat.ips_cantforward++; 1395 else { 1396 ipstat.ips_forward++; 1397 if (type) 1398 ipstat.ips_redirectsent++; 1399 else { 1400 if (mcopy) { 1401#ifdef GATEWAY 1402 if (mcopy->m_flags & M_CANFASTFWD) 1403 ipflow_create(&ipforward_rt, mcopy); 1404#endif 1405 m_freem(mcopy); 1406 } 1407 return; 1408 } 1409 } 1410 if (mcopy == NULL) 1411 return; 1412 destifp = NULL; 1413 1414 switch (error) { 1415 1416 case 0: /* forwarded, but need redirect */ 1417 /* type, code set above */ 1418 break; 1419 1420 case ENETUNREACH: /* shouldn't happen, checked above */ 1421 case EHOSTUNREACH: 1422 case ENETDOWN: 1423 case EHOSTDOWN: 1424 default: 1425 type = ICMP_UNREACH; 1426 code = ICMP_UNREACH_HOST; 1427 break; 1428 1429 case EMSGSIZE: 1430 type = ICMP_UNREACH; 1431 code = ICMP_UNREACH_NEEDFRAG; 1432#ifndef IPSEC 1433 if (ipforward_rt.ro_rt) 1434 destifp = ipforward_rt.ro_rt->rt_ifp; 1435#else 1436 /* 1437 * If the packet is routed over IPsec tunnel, tell the 1438 * originator the tunnel MTU. 1439 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1440 * XXX quickhack!!! 1441 */ 1442 if (ipforward_rt.ro_rt) { 1443 struct secpolicy *sp; 1444 int ipsecerror; 1445 size_t ipsechdr; 1446 struct route *ro; 1447 1448 sp = ipsec4_getpolicybyaddr(mcopy, 1449 IPSEC_DIR_OUTBOUND, 1450 IP_FORWARDING, 1451 &ipsecerror); 1452 1453 if (sp == NULL) 1454 destifp = ipforward_rt.ro_rt->rt_ifp; 1455 else { 1456 /* count IPsec header size */ 1457 ipsechdr = ipsec4_hdrsiz(mcopy, 1458 IPSEC_DIR_OUTBOUND, 1459 NULL); 1460 1461 /* 1462 * find the correct route for outer IPv4 1463 * header, compute tunnel MTU. 1464 * 1465 * XXX BUG ALERT 1466 * The "dummyifp" code relies upon the fact 1467 * that icmp_error() touches only ifp->if_mtu. 1468 */ 1469 /*XXX*/ 1470 destifp = NULL; 1471 if (sp->req != NULL 1472 && sp->req->sav != NULL 1473 && sp->req->sav->sah != NULL) { 1474 ro = &sp->req->sav->sah->sa_route; 1475 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1476 dummyifp.if_mtu = 1477 ro->ro_rt->rt_ifp->if_mtu; 1478 dummyifp.if_mtu -= ipsechdr; 1479 destifp = &dummyifp; 1480 } 1481 } 1482 1483 key_freesp(sp); 1484 } 1485 } 1486#endif /*IPSEC*/ 1487 ipstat.ips_cantfrag++; 1488 break; 1489 1490 case ENOBUFS: 1491 type = ICMP_SOURCEQUENCH; 1492 code = 0; 1493 break; 1494 } 1495 icmp_error(mcopy, type, code, dest, destifp); 1496} 1497 1498void 1499ip_savecontrol(inp, mp, ip, m) 1500 register struct inpcb *inp; 1501 register struct mbuf **mp; 1502 register struct ip *ip; 1503 register struct mbuf *m; 1504{ 1505 1506 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1507 struct timeval tv; 1508 1509 microtime(&tv); 1510 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1511 SCM_TIMESTAMP, SOL_SOCKET); 1512 if (*mp) 1513 mp = &(*mp)->m_next; 1514 } 1515 if (inp->inp_flags & INP_RECVDSTADDR) { 1516 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1517 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1518 if (*mp) 1519 mp = &(*mp)->m_next; 1520 } 1521#ifdef notyet 1522 /* 1523 * XXX 1524 * Moving these out of udp_input() made them even more broken 1525 * than they already were. 1526 * - fenner@parc.xerox.com 1527 */ 1528 /* options were tossed already */ 1529 if (inp->inp_flags & INP_RECVOPTS) { 1530 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1531 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1532 if (*mp) 1533 mp = &(*mp)->m_next; 1534 } 1535 /* ip_srcroute doesn't do what we want here, need to fix */ 1536 if (inp->inp_flags & INP_RECVRETOPTS) { 1537 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1538 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1539 if (*mp) 1540 mp = &(*mp)->m_next; 1541 } 1542#endif 1543 if (inp->inp_flags & INP_RECVIF) { 1544 struct sockaddr_dl sdl; 1545 1546 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1547 sdl.sdl_family = AF_LINK; 1548 sdl.sdl_index = m->m_pkthdr.rcvif ? 1549 m->m_pkthdr.rcvif->if_index : 0; 1550 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1551 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1552 IP_RECVIF, IPPROTO_IP); 1553 if (*mp) 1554 mp = &(*mp)->m_next; 1555 } 1556} 1557 1558int 1559ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1560 int *name; 1561 u_int namelen; 1562 void *oldp; 1563 size_t *oldlenp; 1564 void *newp; 1565 size_t newlen; 1566{ 1567 extern int subnetsarelocal, hostzeroisbroadcast; 1568 1569 int error, old; 1570 1571 /* All sysctl names at this level are terminal. */ 1572 if (namelen != 1) 1573 return (ENOTDIR); 1574 1575 switch (name[0]) { 1576 case IPCTL_FORWARDING: 1577 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1578 case IPCTL_SENDREDIRECTS: 1579 return (sysctl_int(oldp, oldlenp, newp, newlen, 1580 &ipsendredirects)); 1581 case IPCTL_DEFTTL: 1582 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1583#ifdef notyet 1584 case IPCTL_DEFMTU: 1585 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1586#endif 1587 case IPCTL_FORWSRCRT: 1588 /* Don't allow this to change in a secure environment. */ 1589 if (securelevel > 0) 1590 return (sysctl_rdint(oldp, oldlenp, newp, 1591 ip_forwsrcrt)); 1592 else 1593 return (sysctl_int(oldp, oldlenp, newp, newlen, 1594 &ip_forwsrcrt)); 1595 case IPCTL_DIRECTEDBCAST: 1596 return (sysctl_int(oldp, oldlenp, newp, newlen, 1597 &ip_directedbcast)); 1598 case IPCTL_ALLOWSRCRT: 1599 return (sysctl_int(oldp, oldlenp, newp, newlen, 1600 &ip_allowsrcrt)); 1601 case IPCTL_SUBNETSARELOCAL: 1602 return (sysctl_int(oldp, oldlenp, newp, newlen, 1603 &subnetsarelocal)); 1604 case IPCTL_MTUDISC: 1605 error = sysctl_int(oldp, oldlenp, newp, newlen, 1606 &ip_mtudisc); 1607 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1608 ip_mtudisc_timeout_q = 1609 rt_timer_queue_create(ip_mtudisc_timeout); 1610 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1611 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1612 ip_mtudisc_timeout_q = NULL; 1613 } 1614 return error; 1615 case IPCTL_ANONPORTMIN: 1616 old = anonportmin; 1617 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1618 if (anonportmin >= anonportmax || anonportmin > 65535 1619#ifndef IPNOPRIVPORTS 1620 || anonportmin < IPPORT_RESERVED 1621#endif 1622 ) { 1623 anonportmin = old; 1624 return (EINVAL); 1625 } 1626 return (error); 1627 case IPCTL_ANONPORTMAX: 1628 old = anonportmax; 1629 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1630 if (anonportmin >= anonportmax || anonportmax > 65535 1631#ifndef IPNOPRIVPORTS 1632 || anonportmax < IPPORT_RESERVED 1633#endif 1634 ) { 1635 anonportmax = old; 1636 return (EINVAL); 1637 } 1638 return (error); 1639 case IPCTL_MTUDISCTIMEOUT: 1640 error = sysctl_int(oldp, oldlenp, newp, newlen, 1641 &ip_mtudisc_timeout); 1642 if (ip_mtudisc_timeout_q != NULL) 1643 rt_timer_queue_change(ip_mtudisc_timeout_q, 1644 ip_mtudisc_timeout); 1645 return (error); 1646#ifdef GATEWAY 1647 case IPCTL_MAXFLOWS: 1648 { 1649 int s; 1650 1651 error = sysctl_int(oldp, oldlenp, newp, newlen, 1652 &ip_maxflows); 1653 s = splsoftnet(); 1654 ipflow_reap(0); 1655 splx(s); 1656 return (error); 1657 } 1658#endif 1659 case IPCTL_HOSTZEROBROADCAST: 1660 return (sysctl_int(oldp, oldlenp, newp, newlen, 1661 &hostzeroisbroadcast)); 1662#if NGIF > 0 1663 case IPCTL_GIF_TTL: 1664 return(sysctl_int(oldp, oldlenp, newp, newlen, 1665 &ip_gif_ttl)); 1666#endif 1667 1668 default: 1669 return (EOPNOTSUPP); 1670 } 1671 /* NOTREACHED */ 1672} 1673