ip_input.c revision 1.98
1/* $NetBSD: ip_input.c,v 1.98 2000/02/12 17:45:44 thorpej Exp $ */ 2 3/* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32/*- 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 3. All advertising materials mentioning features or use of this software 49 * must display the following acknowledgement: 50 * This product includes software developed by the NetBSD 51 * Foundation, Inc. and its contributors. 52 * 4. Neither the name of The NetBSD Foundation nor the names of its 53 * contributors may be used to endorse or promote products derived 54 * from this software without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 57 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 58 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 59 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 60 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 61 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 62 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 63 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 64 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 65 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 66 * POSSIBILITY OF SUCH DAMAGE. 67 */ 68 69/* 70 * Copyright (c) 1982, 1986, 1988, 1993 71 * The Regents of the University of California. All rights reserved. 72 * 73 * Redistribution and use in source and binary forms, with or without 74 * modification, are permitted provided that the following conditions 75 * are met: 76 * 1. Redistributions of source code must retain the above copyright 77 * notice, this list of conditions and the following disclaimer. 78 * 2. Redistributions in binary form must reproduce the above copyright 79 * notice, this list of conditions and the following disclaimer in the 80 * documentation and/or other materials provided with the distribution. 81 * 3. All advertising materials mentioning features or use of this software 82 * must display the following acknowledgement: 83 * This product includes software developed by the University of 84 * California, Berkeley and its contributors. 85 * 4. Neither the name of the University nor the names of its contributors 86 * may be used to endorse or promote products derived from this software 87 * without specific prior written permission. 88 * 89 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 90 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 91 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 92 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 93 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 94 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 95 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 96 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 97 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 98 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 99 * SUCH DAMAGE. 100 * 101 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 102 */ 103 104#include "opt_gateway.h" 105#include "opt_pfil_hooks.h" 106#include "opt_ipsec.h" 107#include "opt_mrouting.h" 108 109#include <sys/param.h> 110#include <sys/systm.h> 111#include <sys/malloc.h> 112#include <sys/mbuf.h> 113#include <sys/domain.h> 114#include <sys/protosw.h> 115#include <sys/socket.h> 116#include <sys/socketvar.h> 117#include <sys/errno.h> 118#include <sys/time.h> 119#include <sys/kernel.h> 120#include <sys/proc.h> 121#include <sys/pool.h> 122 123#include <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 /* 402 * RFC1112: packets with a multicast source address are 403 * not allowed. 404 */ 405 if (IN_MULTICAST(ip->ip_src.s_addr)) { 406 /* XXX stat */ 407 goto bad; 408 } 409 410 if (in_cksum(m, hlen) != 0) { 411 ipstat.ips_badsum++; 412 goto bad; 413 } 414 415 /* 416 * Convert fields to host representation. 417 */ 418 NTOHS(ip->ip_len); 419 NTOHS(ip->ip_off); 420 len = ip->ip_len; 421 422 /* 423 * Check for additional length bogosity 424 */ 425 if (len < hlen) { 426 ipstat.ips_badlen++; 427 goto bad; 428 } 429 430 /* 431 * Check that the amount of data in the buffers 432 * is as at least much as the IP header would have us expect. 433 * Trim mbufs if longer than we expect. 434 * Drop packet if shorter than we expect. 435 */ 436 if (m->m_pkthdr.len < len) { 437 ipstat.ips_tooshort++; 438 goto bad; 439 } 440 if (m->m_pkthdr.len > len) { 441 if (m->m_len == m->m_pkthdr.len) { 442 m->m_len = len; 443 m->m_pkthdr.len = len; 444 } else 445 m_adj(m, len - m->m_pkthdr.len); 446 } 447 448#ifdef IPSEC 449 /* ipflow (IP fast fowarding) is not compatible with IPsec. */ 450 m->m_flags &= ~M_CANFASTFWD; 451#else 452 /* 453 * Assume that we can create a fast-forward IP flow entry 454 * based on this packet. 455 */ 456 m->m_flags |= M_CANFASTFWD; 457#endif 458 459#ifdef PFIL_HOOKS 460 /* 461 * Run through list of hooks for input packets. If there are any 462 * filters which require that additional packets in the flow are 463 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 464 * Note that filters must _never_ set this flag, as another filter 465 * in the list may have previously cleared it. 466 */ 467 m0 = m; 468 for (pfh = pfil_hook_get(PFIL_IN); pfh; pfh = pfh->pfil_link.tqe_next) 469 if (pfh->pfil_func) { 470 rv = pfh->pfil_func(ip, hlen, m->m_pkthdr.rcvif, 0, &m0); 471 if (rv) 472 return; 473 m = m0; 474 if (m == NULL) 475 return; 476 ip = mtod(m, struct ip *); 477 } 478#endif /* PFIL_HOOKS */ 479 480 /* 481 * Process options and, if not destined for us, 482 * ship it on. ip_dooptions returns 1 when an 483 * error was detected (causing an icmp message 484 * to be sent and the original packet to be freed). 485 */ 486 ip_nhops = 0; /* for source routed packets */ 487 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 488 return; 489 490 /* 491 * Check our list of addresses, to see if the packet is for us. 492 */ 493 for (ia = IN_IFADDR_HASH(ip->ip_dst.s_addr).lh_first; 494 ia != NULL; 495 ia = ia->ia_hash.le_next) { 496 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 497 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 498 break; 499 else { 500 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 501 0, m->m_pkthdr.rcvif); 502 return; 503 } 504 } 505 } 506 if (ia != NULL) 507 goto ours; 508 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 509 for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first; 510 ifa != NULL; ifa = ifa->ifa_list.tqe_next) { 511 if (ifa->ifa_addr->sa_family != AF_INET) continue; 512 ia = ifatoia(ifa); 513 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 514 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 515 /* 516 * Look for all-0's host part (old broadcast addr), 517 * either for subnet or net. 518 */ 519 ip->ip_dst.s_addr == ia->ia_subnet || 520 ip->ip_dst.s_addr == ia->ia_net) 521 goto ours; 522 /* 523 * An interface with IP address zero accepts 524 * all packets that arrive on that interface. 525 */ 526 if (in_nullhost(ia->ia_addr.sin_addr)) 527 goto ours; 528 } 529 } 530 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 531 struct in_multi *inm; 532#ifdef MROUTING 533 extern struct socket *ip_mrouter; 534 535 if (m->m_flags & M_EXT) { 536 if ((m = m_pullup(m, hlen)) == 0) { 537 ipstat.ips_toosmall++; 538 return; 539 } 540 ip = mtod(m, struct ip *); 541 } 542 543 if (ip_mrouter) { 544 /* 545 * If we are acting as a multicast router, all 546 * incoming multicast packets are passed to the 547 * kernel-level multicast forwarding function. 548 * The packet is returned (relatively) intact; if 549 * ip_mforward() returns a non-zero value, the packet 550 * must be discarded, else it may be accepted below. 551 * 552 * (The IP ident field is put in the same byte order 553 * as expected when ip_mforward() is called from 554 * ip_output().) 555 */ 556 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 557 ipstat.ips_cantforward++; 558 m_freem(m); 559 return; 560 } 561 562 /* 563 * The process-level routing demon needs to receive 564 * all multicast IGMP packets, whether or not this 565 * host belongs to their destination groups. 566 */ 567 if (ip->ip_p == IPPROTO_IGMP) 568 goto ours; 569 ipstat.ips_forward++; 570 } 571#endif 572 /* 573 * See if we belong to the destination multicast group on the 574 * arrival interface. 575 */ 576 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 577 if (inm == NULL) { 578 ipstat.ips_cantforward++; 579 m_freem(m); 580 return; 581 } 582 goto ours; 583 } 584 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 585 in_nullhost(ip->ip_dst)) 586 goto ours; 587 588 /* 589 * Not for us; forward if possible and desirable. 590 */ 591 if (ipforwarding == 0) { 592 ipstat.ips_cantforward++; 593 m_freem(m); 594 } else 595 ip_forward(m, 0); 596 return; 597 598ours: 599 /* 600 * If offset or IP_MF are set, must reassemble. 601 * Otherwise, nothing need be done. 602 * (We could look in the reassembly queue to see 603 * if the packet was previously fragmented, 604 * but it's not worth the time; just let them time out.) 605 */ 606 if (ip->ip_off & ~(IP_DF|IP_RF)) { 607 /* 608 * Look for queue of fragments 609 * of this datagram. 610 */ 611 IPQ_LOCK(); 612 for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next) 613 if (ip->ip_id == fp->ipq_id && 614 in_hosteq(ip->ip_src, fp->ipq_src) && 615 in_hosteq(ip->ip_dst, fp->ipq_dst) && 616 ip->ip_p == fp->ipq_p) 617 goto found; 618 fp = 0; 619found: 620 621 /* 622 * Adjust ip_len to not reflect header, 623 * set ipqe_mff if more fragments are expected, 624 * convert offset of this to bytes. 625 */ 626 ip->ip_len -= hlen; 627 mff = (ip->ip_off & IP_MF) != 0; 628 if (mff) { 629 /* 630 * Make sure that fragments have a data length 631 * that's a non-zero multiple of 8 bytes. 632 */ 633 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 634 ipstat.ips_badfrags++; 635 IPQ_UNLOCK(); 636 goto bad; 637 } 638 } 639 ip->ip_off <<= 3; 640 641 /* 642 * If datagram marked as having more fragments 643 * or if this is not the first fragment, 644 * attempt reassembly; if it succeeds, proceed. 645 */ 646 if (mff || ip->ip_off) { 647 ipstat.ips_fragments++; 648 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 649 if (ipqe == NULL) { 650 ipstat.ips_rcvmemdrop++; 651 IPQ_UNLOCK(); 652 goto bad; 653 } 654 ipqe->ipqe_mff = mff; 655 ipqe->ipqe_m = m; 656 ipqe->ipqe_ip = ip; 657 m = ip_reass(ipqe, fp); 658 if (m == 0) { 659 IPQ_UNLOCK(); 660 return; 661 } 662 ipstat.ips_reassembled++; 663 ip = mtod(m, struct ip *); 664 hlen = ip->ip_hl << 2; 665 ip->ip_len += hlen; 666 } else 667 if (fp) 668 ip_freef(fp); 669 IPQ_UNLOCK(); 670 } 671 672 /* 673 * Switch out to protocol's input routine. 674 */ 675#if IFA_STATS 676 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 677#endif 678 ipstat.ips_delivered++; 679 { 680 int off = hlen, nh = ip->ip_p; 681 682 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 683 return; 684 } 685bad: 686 m_freem(m); 687} 688 689/* 690 * Take incoming datagram fragment and try to 691 * reassemble it into whole datagram. If a chain for 692 * reassembly of this datagram already exists, then it 693 * is given as fp; otherwise have to make a chain. 694 */ 695struct mbuf * 696ip_reass(ipqe, fp) 697 register struct ipqent *ipqe; 698 register struct ipq *fp; 699{ 700 register struct mbuf *m = ipqe->ipqe_m; 701 register struct ipqent *nq, *p, *q; 702 struct ip *ip; 703 struct mbuf *t; 704 int hlen = ipqe->ipqe_ip->ip_hl << 2; 705 int i, next; 706 707 IPQ_LOCK_CHECK(); 708 709 /* 710 * Presence of header sizes in mbufs 711 * would confuse code below. 712 */ 713 m->m_data += hlen; 714 m->m_len -= hlen; 715 716 /* 717 * If first fragment to arrive, create a reassembly queue. 718 */ 719 if (fp == 0) { 720 MALLOC(fp, struct ipq *, sizeof (struct ipq), 721 M_FTABLE, M_NOWAIT); 722 if (fp == NULL) 723 goto dropfrag; 724 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 725 fp->ipq_ttl = IPFRAGTTL; 726 fp->ipq_p = ipqe->ipqe_ip->ip_p; 727 fp->ipq_id = ipqe->ipqe_ip->ip_id; 728 LIST_INIT(&fp->ipq_fragq); 729 fp->ipq_src = ipqe->ipqe_ip->ip_src; 730 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 731 p = NULL; 732 goto insert; 733 } 734 735 /* 736 * Find a segment which begins after this one does. 737 */ 738 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 739 p = q, q = q->ipqe_q.le_next) 740 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 741 break; 742 743 /* 744 * If there is a preceding segment, it may provide some of 745 * our data already. If so, drop the data from the incoming 746 * segment. If it provides all of our data, drop us. 747 */ 748 if (p != NULL) { 749 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 750 ipqe->ipqe_ip->ip_off; 751 if (i > 0) { 752 if (i >= ipqe->ipqe_ip->ip_len) 753 goto dropfrag; 754 m_adj(ipqe->ipqe_m, i); 755 ipqe->ipqe_ip->ip_off += i; 756 ipqe->ipqe_ip->ip_len -= i; 757 } 758 } 759 760 /* 761 * While we overlap succeeding segments trim them or, 762 * if they are completely covered, dequeue them. 763 */ 764 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 765 q->ipqe_ip->ip_off; q = nq) { 766 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 767 q->ipqe_ip->ip_off; 768 if (i < q->ipqe_ip->ip_len) { 769 q->ipqe_ip->ip_len -= i; 770 q->ipqe_ip->ip_off += i; 771 m_adj(q->ipqe_m, i); 772 break; 773 } 774 nq = q->ipqe_q.le_next; 775 m_freem(q->ipqe_m); 776 LIST_REMOVE(q, ipqe_q); 777 pool_put(&ipqent_pool, q); 778 } 779 780insert: 781 /* 782 * Stick new segment in its place; 783 * check for complete reassembly. 784 */ 785 if (p == NULL) { 786 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 787 } else { 788 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 789 } 790 next = 0; 791 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 792 p = q, q = q->ipqe_q.le_next) { 793 if (q->ipqe_ip->ip_off != next) 794 return (0); 795 next += q->ipqe_ip->ip_len; 796 } 797 if (p->ipqe_mff) 798 return (0); 799 800 /* 801 * Reassembly is complete. Check for a bogus message size and 802 * concatenate fragments. 803 */ 804 q = fp->ipq_fragq.lh_first; 805 ip = q->ipqe_ip; 806 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 807 ipstat.ips_toolong++; 808 ip_freef(fp); 809 return (0); 810 } 811 m = q->ipqe_m; 812 t = m->m_next; 813 m->m_next = 0; 814 m_cat(m, t); 815 nq = q->ipqe_q.le_next; 816 pool_put(&ipqent_pool, q); 817 for (q = nq; q != NULL; q = nq) { 818 t = q->ipqe_m; 819 nq = q->ipqe_q.le_next; 820 pool_put(&ipqent_pool, q); 821 m_cat(m, t); 822 } 823 824 /* 825 * Create header for new ip packet by 826 * modifying header of first packet; 827 * dequeue and discard fragment reassembly header. 828 * Make header visible. 829 */ 830 ip->ip_len = next; 831 ip->ip_ttl = 0; /* xxx */ 832 ip->ip_sum = 0; 833 ip->ip_src = fp->ipq_src; 834 ip->ip_dst = fp->ipq_dst; 835 LIST_REMOVE(fp, ipq_q); 836 FREE(fp, M_FTABLE); 837 m->m_len += (ip->ip_hl << 2); 838 m->m_data -= (ip->ip_hl << 2); 839 /* some debugging cruft by sklower, below, will go away soon */ 840 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 841 register int plen = 0; 842 for (t = m; t; t = t->m_next) 843 plen += t->m_len; 844 m->m_pkthdr.len = plen; 845 } 846 return (m); 847 848dropfrag: 849 ipstat.ips_fragdropped++; 850 m_freem(m); 851 pool_put(&ipqent_pool, ipqe); 852 return (0); 853} 854 855/* 856 * Free a fragment reassembly header and all 857 * associated datagrams. 858 */ 859void 860ip_freef(fp) 861 struct ipq *fp; 862{ 863 register struct ipqent *q, *p; 864 865 IPQ_LOCK_CHECK(); 866 867 for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { 868 p = q->ipqe_q.le_next; 869 m_freem(q->ipqe_m); 870 LIST_REMOVE(q, ipqe_q); 871 pool_put(&ipqent_pool, q); 872 } 873 LIST_REMOVE(fp, ipq_q); 874 FREE(fp, M_FTABLE); 875} 876 877/* 878 * IP timer processing; 879 * if a timer expires on a reassembly 880 * queue, discard it. 881 */ 882void 883ip_slowtimo() 884{ 885 register struct ipq *fp, *nfp; 886 int s = splsoftnet(); 887 888 IPQ_LOCK(); 889 for (fp = ipq.lh_first; fp != NULL; fp = nfp) { 890 nfp = fp->ipq_q.le_next; 891 if (--fp->ipq_ttl == 0) { 892 ipstat.ips_fragtimeout++; 893 ip_freef(fp); 894 } 895 } 896 IPQ_UNLOCK(); 897#ifdef GATEWAY 898 ipflow_slowtimo(); 899#endif 900 splx(s); 901} 902 903/* 904 * Drain off all datagram fragments. 905 */ 906void 907ip_drain() 908{ 909 910 /* 911 * We may be called from a device's interrupt context. If 912 * the ipq is already busy, just bail out now. 913 */ 914 if (ipq_lock_try() == 0) 915 return; 916 917 while (ipq.lh_first != NULL) { 918 ipstat.ips_fragdropped++; 919 ip_freef(ipq.lh_first); 920 } 921 922 IPQ_UNLOCK(); 923} 924 925/* 926 * Do option processing on a datagram, 927 * possibly discarding it if bad options are encountered, 928 * or forwarding it if source-routed. 929 * Returns 1 if packet has been forwarded/freed, 930 * 0 if the packet should be processed further. 931 */ 932int 933ip_dooptions(m) 934 struct mbuf *m; 935{ 936 register struct ip *ip = mtod(m, struct ip *); 937 register u_char *cp; 938 register struct ip_timestamp *ipt; 939 register struct in_ifaddr *ia; 940 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 941 struct in_addr *sin, dst; 942 n_time ntime; 943 944 dst = ip->ip_dst; 945 cp = (u_char *)(ip + 1); 946 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 947 for (; cnt > 0; cnt -= optlen, cp += optlen) { 948 opt = cp[IPOPT_OPTVAL]; 949 if (opt == IPOPT_EOL) 950 break; 951 if (opt == IPOPT_NOP) 952 optlen = 1; 953 else { 954 optlen = cp[IPOPT_OLEN]; 955 if (optlen <= 0 || optlen > cnt) { 956 code = &cp[IPOPT_OLEN] - (u_char *)ip; 957 goto bad; 958 } 959 } 960 switch (opt) { 961 962 default: 963 break; 964 965 /* 966 * Source routing with record. 967 * Find interface with current destination address. 968 * If none on this machine then drop if strictly routed, 969 * or do nothing if loosely routed. 970 * Record interface address and bring up next address 971 * component. If strictly routed make sure next 972 * address is on directly accessible net. 973 */ 974 case IPOPT_LSRR: 975 case IPOPT_SSRR: 976 if (ip_allowsrcrt == 0) { 977 type = ICMP_UNREACH; 978 code = ICMP_UNREACH_NET_PROHIB; 979 goto bad; 980 } 981 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 982 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 983 goto bad; 984 } 985 ipaddr.sin_addr = ip->ip_dst; 986 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 987 if (ia == 0) { 988 if (opt == IPOPT_SSRR) { 989 type = ICMP_UNREACH; 990 code = ICMP_UNREACH_SRCFAIL; 991 goto bad; 992 } 993 /* 994 * Loose routing, and not at next destination 995 * yet; nothing to do except forward. 996 */ 997 break; 998 } 999 off--; /* 0 origin */ 1000 if (off > optlen - sizeof(struct in_addr)) { 1001 /* 1002 * End of source route. Should be for us. 1003 */ 1004 save_rte(cp, ip->ip_src); 1005 break; 1006 } 1007 /* 1008 * locate outgoing interface 1009 */ 1010 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1011 sizeof(ipaddr.sin_addr)); 1012 if (opt == IPOPT_SSRR) 1013 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1014 else 1015 ia = ip_rtaddr(ipaddr.sin_addr); 1016 if (ia == 0) { 1017 type = ICMP_UNREACH; 1018 code = ICMP_UNREACH_SRCFAIL; 1019 goto bad; 1020 } 1021 ip->ip_dst = ipaddr.sin_addr; 1022 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1023 (caddr_t)(cp + off), sizeof(struct in_addr)); 1024 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1025 /* 1026 * Let ip_intr's mcast routing check handle mcast pkts 1027 */ 1028 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1029 break; 1030 1031 case IPOPT_RR: 1032 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1033 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1034 goto bad; 1035 } 1036 /* 1037 * If no space remains, ignore. 1038 */ 1039 off--; /* 0 origin */ 1040 if (off > optlen - sizeof(struct in_addr)) 1041 break; 1042 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1043 sizeof(ipaddr.sin_addr)); 1044 /* 1045 * locate outgoing interface; if we're the destination, 1046 * use the incoming interface (should be same). 1047 */ 1048 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1049 == NULL && 1050 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1051 type = ICMP_UNREACH; 1052 code = ICMP_UNREACH_HOST; 1053 goto bad; 1054 } 1055 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1056 (caddr_t)(cp + off), sizeof(struct in_addr)); 1057 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1058 break; 1059 1060 case IPOPT_TS: 1061 code = cp - (u_char *)ip; 1062 ipt = (struct ip_timestamp *)cp; 1063 if (ipt->ipt_len < 5) 1064 goto bad; 1065 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1066 if (++ipt->ipt_oflw == 0) 1067 goto bad; 1068 break; 1069 } 1070 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1071 switch (ipt->ipt_flg) { 1072 1073 case IPOPT_TS_TSONLY: 1074 break; 1075 1076 case IPOPT_TS_TSANDADDR: 1077 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1078 sizeof(struct in_addr) > ipt->ipt_len) 1079 goto bad; 1080 ipaddr.sin_addr = dst; 1081 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1082 m->m_pkthdr.rcvif)); 1083 if (ia == 0) 1084 continue; 1085 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1086 (caddr_t)sin, sizeof(struct in_addr)); 1087 ipt->ipt_ptr += sizeof(struct in_addr); 1088 break; 1089 1090 case IPOPT_TS_PRESPEC: 1091 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1092 sizeof(struct in_addr) > ipt->ipt_len) 1093 goto bad; 1094 bcopy((caddr_t)sin, (caddr_t)&ipaddr.sin_addr, 1095 sizeof(struct in_addr)); 1096 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1097 == NULL) 1098 continue; 1099 ipt->ipt_ptr += sizeof(struct in_addr); 1100 break; 1101 1102 default: 1103 goto bad; 1104 } 1105 ntime = iptime(); 1106 bcopy((caddr_t)&ntime, (caddr_t)cp + ipt->ipt_ptr - 1, 1107 sizeof(n_time)); 1108 ipt->ipt_ptr += sizeof(n_time); 1109 } 1110 } 1111 if (forward) { 1112 if (ip_forwsrcrt == 0) { 1113 type = ICMP_UNREACH; 1114 code = ICMP_UNREACH_SRCFAIL; 1115 goto bad; 1116 } 1117 ip_forward(m, 1); 1118 return (1); 1119 } 1120 return (0); 1121bad: 1122 icmp_error(m, type, code, 0, 0); 1123 ipstat.ips_badoptions++; 1124 return (1); 1125} 1126 1127/* 1128 * Given address of next destination (final or next hop), 1129 * return internet address info of interface to be used to get there. 1130 */ 1131struct in_ifaddr * 1132ip_rtaddr(dst) 1133 struct in_addr dst; 1134{ 1135 register struct sockaddr_in *sin; 1136 1137 sin = satosin(&ipforward_rt.ro_dst); 1138 1139 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1140 if (ipforward_rt.ro_rt) { 1141 RTFREE(ipforward_rt.ro_rt); 1142 ipforward_rt.ro_rt = 0; 1143 } 1144 sin->sin_family = AF_INET; 1145 sin->sin_len = sizeof(*sin); 1146 sin->sin_addr = dst; 1147 1148 rtalloc(&ipforward_rt); 1149 } 1150 if (ipforward_rt.ro_rt == 0) 1151 return ((struct in_ifaddr *)0); 1152 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1153} 1154 1155/* 1156 * Save incoming source route for use in replies, 1157 * to be picked up later by ip_srcroute if the receiver is interested. 1158 */ 1159void 1160save_rte(option, dst) 1161 u_char *option; 1162 struct in_addr dst; 1163{ 1164 unsigned olen; 1165 1166 olen = option[IPOPT_OLEN]; 1167#ifdef DIAGNOSTIC 1168 if (ipprintfs) 1169 printf("save_rte: olen %d\n", olen); 1170#endif /* 0 */ 1171 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1172 return; 1173 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1174 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1175 ip_srcrt.dst = dst; 1176} 1177 1178/* 1179 * Retrieve incoming source route for use in replies, 1180 * in the same form used by setsockopt. 1181 * The first hop is placed before the options, will be removed later. 1182 */ 1183struct mbuf * 1184ip_srcroute() 1185{ 1186 register struct in_addr *p, *q; 1187 register struct mbuf *m; 1188 1189 if (ip_nhops == 0) 1190 return ((struct mbuf *)0); 1191 m = m_get(M_DONTWAIT, MT_SOOPTS); 1192 if (m == 0) 1193 return ((struct mbuf *)0); 1194 1195#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1196 1197 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1198 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1199 OPTSIZ; 1200#ifdef DIAGNOSTIC 1201 if (ipprintfs) 1202 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1203#endif 1204 1205 /* 1206 * First save first hop for return route 1207 */ 1208 p = &ip_srcrt.route[ip_nhops - 1]; 1209 *(mtod(m, struct in_addr *)) = *p--; 1210#ifdef DIAGNOSTIC 1211 if (ipprintfs) 1212 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1213#endif 1214 1215 /* 1216 * Copy option fields and padding (nop) to mbuf. 1217 */ 1218 ip_srcrt.nop = IPOPT_NOP; 1219 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1220 bcopy((caddr_t)&ip_srcrt.nop, 1221 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1222 q = (struct in_addr *)(mtod(m, caddr_t) + 1223 sizeof(struct in_addr) + OPTSIZ); 1224#undef OPTSIZ 1225 /* 1226 * Record return path as an IP source route, 1227 * reversing the path (pointers are now aligned). 1228 */ 1229 while (p >= ip_srcrt.route) { 1230#ifdef DIAGNOSTIC 1231 if (ipprintfs) 1232 printf(" %x", ntohl(q->s_addr)); 1233#endif 1234 *q++ = *p--; 1235 } 1236 /* 1237 * Last hop goes to final destination. 1238 */ 1239 *q = ip_srcrt.dst; 1240#ifdef DIAGNOSTIC 1241 if (ipprintfs) 1242 printf(" %x\n", ntohl(q->s_addr)); 1243#endif 1244 return (m); 1245} 1246 1247/* 1248 * Strip out IP options, at higher 1249 * level protocol in the kernel. 1250 * Second argument is buffer to which options 1251 * will be moved, and return value is their length. 1252 * XXX should be deleted; last arg currently ignored. 1253 */ 1254void 1255ip_stripoptions(m, mopt) 1256 register struct mbuf *m; 1257 struct mbuf *mopt; 1258{ 1259 register int i; 1260 struct ip *ip = mtod(m, struct ip *); 1261 register caddr_t opts; 1262 int olen; 1263 1264 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1265 opts = (caddr_t)(ip + 1); 1266 i = m->m_len - (sizeof (struct ip) + olen); 1267 bcopy(opts + olen, opts, (unsigned)i); 1268 m->m_len -= olen; 1269 if (m->m_flags & M_PKTHDR) 1270 m->m_pkthdr.len -= olen; 1271 ip->ip_len -= olen; 1272 ip->ip_hl = sizeof (struct ip) >> 2; 1273} 1274 1275int inetctlerrmap[PRC_NCMDS] = { 1276 0, 0, 0, 0, 1277 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1278 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1279 EMSGSIZE, EHOSTUNREACH, 0, 0, 1280 0, 0, 0, 0, 1281 ENOPROTOOPT 1282}; 1283 1284/* 1285 * Forward a packet. If some error occurs return the sender 1286 * an icmp packet. Note we can't always generate a meaningful 1287 * icmp message because icmp doesn't have a large enough repertoire 1288 * of codes and types. 1289 * 1290 * If not forwarding, just drop the packet. This could be confusing 1291 * if ipforwarding was zero but some routing protocol was advancing 1292 * us as a gateway to somewhere. However, we must let the routing 1293 * protocol deal with that. 1294 * 1295 * The srcrt parameter indicates whether the packet is being forwarded 1296 * via a source route. 1297 */ 1298void 1299ip_forward(m, srcrt) 1300 struct mbuf *m; 1301 int srcrt; 1302{ 1303 register struct ip *ip = mtod(m, struct ip *); 1304 register struct sockaddr_in *sin; 1305 register struct rtentry *rt; 1306 int error, type = 0, code = 0; 1307 struct mbuf *mcopy; 1308 n_long dest; 1309 struct ifnet *destifp; 1310#ifdef IPSEC 1311 struct ifnet dummyifp; 1312#endif 1313 1314 dest = 0; 1315#ifdef DIAGNOSTIC 1316 if (ipprintfs) 1317 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1318 ntohl(ip->ip_src.s_addr), 1319 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1320#endif 1321 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1322 ipstat.ips_cantforward++; 1323 m_freem(m); 1324 return; 1325 } 1326 if (ip->ip_ttl <= IPTTLDEC) { 1327 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1328 return; 1329 } 1330 ip->ip_ttl -= IPTTLDEC; 1331 1332 sin = satosin(&ipforward_rt.ro_dst); 1333 if ((rt = ipforward_rt.ro_rt) == 0 || 1334 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1335 if (ipforward_rt.ro_rt) { 1336 RTFREE(ipforward_rt.ro_rt); 1337 ipforward_rt.ro_rt = 0; 1338 } 1339 sin->sin_family = AF_INET; 1340 sin->sin_len = sizeof(struct sockaddr_in); 1341 sin->sin_addr = ip->ip_dst; 1342 1343 rtalloc(&ipforward_rt); 1344 if (ipforward_rt.ro_rt == 0) { 1345 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1346 return; 1347 } 1348 rt = ipforward_rt.ro_rt; 1349 } 1350 1351 /* 1352 * Save at most 68 bytes of the packet in case 1353 * we need to generate an ICMP message to the src. 1354 */ 1355 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 68)); 1356 1357 /* 1358 * If forwarding packet using same interface that it came in on, 1359 * perhaps should send a redirect to sender to shortcut a hop. 1360 * Only send redirect if source is sending directly to us, 1361 * and if packet was not source routed (or has any options). 1362 * Also, don't send redirect if forwarding using a default route 1363 * or a route modified by a redirect. 1364 */ 1365 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1366 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1367 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1368 ipsendredirects && !srcrt) { 1369 if (rt->rt_ifa && 1370 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1371 ifatoia(rt->rt_ifa)->ia_subnet) { 1372 if (rt->rt_flags & RTF_GATEWAY) 1373 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1374 else 1375 dest = ip->ip_dst.s_addr; 1376 /* 1377 * Router requirements says to only send host 1378 * redirects. 1379 */ 1380 type = ICMP_REDIRECT; 1381 code = ICMP_REDIRECT_HOST; 1382#ifdef DIAGNOSTIC 1383 if (ipprintfs) 1384 printf("redirect (%d) to %x\n", code, 1385 (u_int32_t)dest); 1386#endif 1387 } 1388 } 1389 1390#ifdef IPSEC 1391 m->m_pkthdr.rcvif = NULL; 1392#endif /*IPSEC*/ 1393 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1394 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1395 if (error) 1396 ipstat.ips_cantforward++; 1397 else { 1398 ipstat.ips_forward++; 1399 if (type) 1400 ipstat.ips_redirectsent++; 1401 else { 1402 if (mcopy) { 1403#ifdef GATEWAY 1404 if (mcopy->m_flags & M_CANFASTFWD) 1405 ipflow_create(&ipforward_rt, mcopy); 1406#endif 1407 m_freem(mcopy); 1408 } 1409 return; 1410 } 1411 } 1412 if (mcopy == NULL) 1413 return; 1414 destifp = NULL; 1415 1416 switch (error) { 1417 1418 case 0: /* forwarded, but need redirect */ 1419 /* type, code set above */ 1420 break; 1421 1422 case ENETUNREACH: /* shouldn't happen, checked above */ 1423 case EHOSTUNREACH: 1424 case ENETDOWN: 1425 case EHOSTDOWN: 1426 default: 1427 type = ICMP_UNREACH; 1428 code = ICMP_UNREACH_HOST; 1429 break; 1430 1431 case EMSGSIZE: 1432 type = ICMP_UNREACH; 1433 code = ICMP_UNREACH_NEEDFRAG; 1434#ifndef IPSEC 1435 if (ipforward_rt.ro_rt) 1436 destifp = ipforward_rt.ro_rt->rt_ifp; 1437#else 1438 /* 1439 * If the packet is routed over IPsec tunnel, tell the 1440 * originator the tunnel MTU. 1441 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1442 * XXX quickhack!!! 1443 */ 1444 if (ipforward_rt.ro_rt) { 1445 struct secpolicy *sp; 1446 int ipsecerror; 1447 size_t ipsechdr; 1448 struct route *ro; 1449 1450 sp = ipsec4_getpolicybyaddr(mcopy, 1451 IPSEC_DIR_OUTBOUND, 1452 IP_FORWARDING, 1453 &ipsecerror); 1454 1455 if (sp == NULL) 1456 destifp = ipforward_rt.ro_rt->rt_ifp; 1457 else { 1458 /* count IPsec header size */ 1459 ipsechdr = ipsec4_hdrsiz(mcopy, 1460 IPSEC_DIR_OUTBOUND, 1461 NULL); 1462 1463 /* 1464 * find the correct route for outer IPv4 1465 * header, compute tunnel MTU. 1466 * 1467 * XXX BUG ALERT 1468 * The "dummyifp" code relies upon the fact 1469 * that icmp_error() touches only ifp->if_mtu. 1470 */ 1471 /*XXX*/ 1472 destifp = NULL; 1473 if (sp->req != NULL 1474 && sp->req->sav != NULL 1475 && sp->req->sav->sah != NULL) { 1476 ro = &sp->req->sav->sah->sa_route; 1477 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1478 dummyifp.if_mtu = 1479 ro->ro_rt->rt_ifp->if_mtu; 1480 dummyifp.if_mtu -= ipsechdr; 1481 destifp = &dummyifp; 1482 } 1483 } 1484 1485 key_freesp(sp); 1486 } 1487 } 1488#endif /*IPSEC*/ 1489 ipstat.ips_cantfrag++; 1490 break; 1491 1492 case ENOBUFS: 1493 type = ICMP_SOURCEQUENCH; 1494 code = 0; 1495 break; 1496 } 1497 icmp_error(mcopy, type, code, dest, destifp); 1498} 1499 1500void 1501ip_savecontrol(inp, mp, ip, m) 1502 register struct inpcb *inp; 1503 register struct mbuf **mp; 1504 register struct ip *ip; 1505 register struct mbuf *m; 1506{ 1507 1508 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1509 struct timeval tv; 1510 1511 microtime(&tv); 1512 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1513 SCM_TIMESTAMP, SOL_SOCKET); 1514 if (*mp) 1515 mp = &(*mp)->m_next; 1516 } 1517 if (inp->inp_flags & INP_RECVDSTADDR) { 1518 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1519 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1520 if (*mp) 1521 mp = &(*mp)->m_next; 1522 } 1523#ifdef notyet 1524 /* 1525 * XXX 1526 * Moving these out of udp_input() made them even more broken 1527 * than they already were. 1528 * - fenner@parc.xerox.com 1529 */ 1530 /* options were tossed already */ 1531 if (inp->inp_flags & INP_RECVOPTS) { 1532 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1533 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1534 if (*mp) 1535 mp = &(*mp)->m_next; 1536 } 1537 /* ip_srcroute doesn't do what we want here, need to fix */ 1538 if (inp->inp_flags & INP_RECVRETOPTS) { 1539 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1540 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1541 if (*mp) 1542 mp = &(*mp)->m_next; 1543 } 1544#endif 1545 if (inp->inp_flags & INP_RECVIF) { 1546 struct sockaddr_dl sdl; 1547 1548 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1549 sdl.sdl_family = AF_LINK; 1550 sdl.sdl_index = m->m_pkthdr.rcvif ? 1551 m->m_pkthdr.rcvif->if_index : 0; 1552 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1553 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1554 IP_RECVIF, IPPROTO_IP); 1555 if (*mp) 1556 mp = &(*mp)->m_next; 1557 } 1558} 1559 1560int 1561ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1562 int *name; 1563 u_int namelen; 1564 void *oldp; 1565 size_t *oldlenp; 1566 void *newp; 1567 size_t newlen; 1568{ 1569 extern int subnetsarelocal, hostzeroisbroadcast; 1570 1571 int error, old; 1572 1573 /* All sysctl names at this level are terminal. */ 1574 if (namelen != 1) 1575 return (ENOTDIR); 1576 1577 switch (name[0]) { 1578 case IPCTL_FORWARDING: 1579 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1580 case IPCTL_SENDREDIRECTS: 1581 return (sysctl_int(oldp, oldlenp, newp, newlen, 1582 &ipsendredirects)); 1583 case IPCTL_DEFTTL: 1584 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1585#ifdef notyet 1586 case IPCTL_DEFMTU: 1587 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1588#endif 1589 case IPCTL_FORWSRCRT: 1590 /* Don't allow this to change in a secure environment. */ 1591 if (securelevel > 0) 1592 return (sysctl_rdint(oldp, oldlenp, newp, 1593 ip_forwsrcrt)); 1594 else 1595 return (sysctl_int(oldp, oldlenp, newp, newlen, 1596 &ip_forwsrcrt)); 1597 case IPCTL_DIRECTEDBCAST: 1598 return (sysctl_int(oldp, oldlenp, newp, newlen, 1599 &ip_directedbcast)); 1600 case IPCTL_ALLOWSRCRT: 1601 return (sysctl_int(oldp, oldlenp, newp, newlen, 1602 &ip_allowsrcrt)); 1603 case IPCTL_SUBNETSARELOCAL: 1604 return (sysctl_int(oldp, oldlenp, newp, newlen, 1605 &subnetsarelocal)); 1606 case IPCTL_MTUDISC: 1607 error = sysctl_int(oldp, oldlenp, newp, newlen, 1608 &ip_mtudisc); 1609 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1610 ip_mtudisc_timeout_q = 1611 rt_timer_queue_create(ip_mtudisc_timeout); 1612 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1613 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1614 ip_mtudisc_timeout_q = NULL; 1615 } 1616 return error; 1617 case IPCTL_ANONPORTMIN: 1618 old = anonportmin; 1619 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1620 if (anonportmin >= anonportmax || anonportmin > 65535 1621#ifndef IPNOPRIVPORTS 1622 || anonportmin < IPPORT_RESERVED 1623#endif 1624 ) { 1625 anonportmin = old; 1626 return (EINVAL); 1627 } 1628 return (error); 1629 case IPCTL_ANONPORTMAX: 1630 old = anonportmax; 1631 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1632 if (anonportmin >= anonportmax || anonportmax > 65535 1633#ifndef IPNOPRIVPORTS 1634 || anonportmax < IPPORT_RESERVED 1635#endif 1636 ) { 1637 anonportmax = old; 1638 return (EINVAL); 1639 } 1640 return (error); 1641 case IPCTL_MTUDISCTIMEOUT: 1642 error = sysctl_int(oldp, oldlenp, newp, newlen, 1643 &ip_mtudisc_timeout); 1644 if (ip_mtudisc_timeout_q != NULL) 1645 rt_timer_queue_change(ip_mtudisc_timeout_q, 1646 ip_mtudisc_timeout); 1647 return (error); 1648#ifdef GATEWAY 1649 case IPCTL_MAXFLOWS: 1650 { 1651 int s; 1652 1653 error = sysctl_int(oldp, oldlenp, newp, newlen, 1654 &ip_maxflows); 1655 s = splsoftnet(); 1656 ipflow_reap(0); 1657 splx(s); 1658 return (error); 1659 } 1660#endif 1661 case IPCTL_HOSTZEROBROADCAST: 1662 return (sysctl_int(oldp, oldlenp, newp, newlen, 1663 &hostzeroisbroadcast)); 1664#if NGIF > 0 1665 case IPCTL_GIF_TTL: 1666 return(sysctl_int(oldp, oldlenp, newp, newlen, 1667 &ip_gif_ttl)); 1668#endif 1669 1670 default: 1671 return (EOPNOTSUPP); 1672 } 1673 /* NOTREACHED */ 1674} 1675