ip_input.c revision 1.91
1/* $NetBSD: ip_input.c,v 1.91 1999/07/09 22:57:18 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 <netinet6/ah.h> 145#ifdef IPSEC_ESP 146#include <netinet6/esp.h> 147#endif 148#include <netkey/key.h> 149#include <netkey/key_debug.h> 150#endif 151 152#ifndef IPFORWARDING 153#ifdef GATEWAY 154#define IPFORWARDING 1 /* forward IP packets not for us */ 155#else /* GATEWAY */ 156#define IPFORWARDING 0 /* don't forward IP packets not for us */ 157#endif /* GATEWAY */ 158#endif /* IPFORWARDING */ 159#ifndef IPSENDREDIRECTS 160#define IPSENDREDIRECTS 1 161#endif 162#ifndef IPFORWSRCRT 163#define IPFORWSRCRT 1 /* forward source-routed packets */ 164#endif 165#ifndef IPALLOWSRCRT 166#define IPALLOWSRCRT 1 /* allow source-routed packets */ 167#endif 168#ifndef IPMTUDISC 169#define IPMTUDISC 0 170#endif 171#ifndef IPMTUDISCTIMEOUT 172#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 173#endif 174 175/* 176 * Note: DIRECTED_BROADCAST is handled this way so that previous 177 * configuration using this option will Just Work. 178 */ 179#ifndef IPDIRECTEDBCAST 180#ifdef DIRECTED_BROADCAST 181#define IPDIRECTEDBCAST 1 182#else 183#define IPDIRECTEDBCAST 0 184#endif /* DIRECTED_BROADCAST */ 185#endif /* IPDIRECTEDBCAST */ 186int ipforwarding = IPFORWARDING; 187int ipsendredirects = IPSENDREDIRECTS; 188int ip_defttl = IPDEFTTL; 189int ip_forwsrcrt = IPFORWSRCRT; 190int ip_directedbcast = IPDIRECTEDBCAST; 191int ip_allowsrcrt = IPALLOWSRCRT; 192int ip_mtudisc = IPMTUDISC; 193u_int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 194#ifdef DIAGNOSTIC 195int ipprintfs = 0; 196#endif 197 198struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 199 200extern struct domain inetdomain; 201extern struct protosw inetsw[]; 202u_char ip_protox[IPPROTO_MAX]; 203int ipqmaxlen = IFQ_MAXLEN; 204struct in_ifaddrhead in_ifaddr; 205struct in_ifaddrhashhead *in_ifaddrhashtbl; 206struct ifqueue ipintrq; 207struct ipstat ipstat; 208u_int16_t ip_id; 209int ip_defttl; 210 211struct ipqhead ipq; 212int ipq_locked; 213 214static __inline int ipq_lock_try __P((void)); 215static __inline void ipq_unlock __P((void)); 216 217static __inline int 218ipq_lock_try() 219{ 220 int s; 221 222 s = splimp(); 223 if (ipq_locked) { 224 splx(s); 225 return (0); 226 } 227 ipq_locked = 1; 228 splx(s); 229 return (1); 230} 231 232static __inline void 233ipq_unlock() 234{ 235 int s; 236 237 s = splimp(); 238 ipq_locked = 0; 239 splx(s); 240} 241 242#ifdef DIAGNOSTIC 243#define IPQ_LOCK() \ 244do { \ 245 if (ipq_lock_try() == 0) { \ 246 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 247 panic("ipq_lock"); \ 248 } \ 249} while (0) 250#define IPQ_LOCK_CHECK() \ 251do { \ 252 if (ipq_locked == 0) { \ 253 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 254 panic("ipq lock check"); \ 255 } \ 256} while (0) 257#else 258#define IPQ_LOCK() (void) ipq_lock_try() 259#define IPQ_LOCK_CHECK() /* nothing */ 260#endif 261 262#define IPQ_UNLOCK() ipq_unlock() 263 264struct pool ipqent_pool; 265 266/* 267 * We need to save the IP options in case a protocol wants to respond 268 * to an incoming packet over the same route if the packet got here 269 * using IP source routing. This allows connection establishment and 270 * maintenance when the remote end is on a network that is not known 271 * to us. 272 */ 273int ip_nhops = 0; 274static struct ip_srcrt { 275 struct in_addr dst; /* final destination */ 276 char nop; /* one NOP to align */ 277 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 278 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 279} ip_srcrt; 280 281static void save_rte __P((u_char *, struct in_addr)); 282 283/* 284 * IP initialization: fill in IP protocol switch table. 285 * All protocols not implemented in kernel go to raw IP protocol handler. 286 */ 287void 288ip_init() 289{ 290 register struct protosw *pr; 291 register int i; 292 293 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 294 0, NULL, NULL, M_IPQ); 295 296 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 297 if (pr == 0) 298 panic("ip_init"); 299 for (i = 0; i < IPPROTO_MAX; i++) 300 ip_protox[i] = pr - inetsw; 301 for (pr = inetdomain.dom_protosw; 302 pr < inetdomain.dom_protoswNPROTOSW; pr++) 303 if (pr->pr_domain->dom_family == PF_INET && 304 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 305 ip_protox[pr->pr_protocol] = pr - inetsw; 306 LIST_INIT(&ipq); 307 ip_id = time.tv_sec & 0xffff; 308 ipintrq.ifq_maxlen = ipqmaxlen; 309 TAILQ_INIT(&in_ifaddr); 310 in_ifaddrhashtbl = 311 hashinit(IN_IFADDR_HASH_SIZE, M_IFADDR, M_WAITOK, &in_ifaddrhash); 312 if (ip_mtudisc != 0) 313 ip_mtudisc_timeout_q = 314 rt_timer_queue_create(ip_mtudisc_timeout); 315#ifdef GATEWAY 316 ipflow_init(); 317#endif 318} 319 320struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 321struct route ipforward_rt; 322 323/* 324 * IP software interrupt routine 325 */ 326void 327ipintr() 328{ 329 int s; 330 struct mbuf *m; 331 332 while (1) { 333 s = splimp(); 334 IF_DEQUEUE(&ipintrq, m); 335 splx(s); 336 if (m == 0) 337 return; 338 ip_input(m); 339 } 340} 341 342/* 343 * Ip input routine. Checksum and byte swap header. If fragmented 344 * try to reassemble. Process options. Pass to next level. 345 */ 346void 347ip_input(struct mbuf *m) 348{ 349 register struct ip *ip = NULL; 350 register struct ipq *fp; 351 register struct in_ifaddr *ia; 352 register struct ifaddr *ifa; 353 struct ipqent *ipqe; 354 int hlen = 0, mff, len; 355#ifdef PFIL_HOOKS 356 struct packet_filter_hook *pfh; 357 struct mbuf *m0; 358 int rv; 359#endif /* PFIL_HOOKS */ 360 361#ifdef DIAGNOSTIC 362 if ((m->m_flags & M_PKTHDR) == 0) 363 panic("ipintr no HDR"); 364#endif 365#ifdef IPSEC 366 /* 367 * should the inner packet be considered authentic? 368 * see comment in ah4_input(). 369 */ 370 if (m) { 371 m->m_flags &= ~M_AUTHIPHDR; 372 m->m_flags &= ~M_AUTHIPDGM; 373 } 374#endif 375 /* 376 * If no IP addresses have been set yet but the interfaces 377 * are receiving, can't do anything with incoming packets yet. 378 */ 379 if (in_ifaddr.tqh_first == 0) 380 goto bad; 381 ipstat.ips_total++; 382 if (m->m_len < sizeof (struct ip) && 383 (m = m_pullup(m, sizeof (struct ip))) == 0) { 384 ipstat.ips_toosmall++; 385 return; 386 } 387 ip = mtod(m, struct ip *); 388 if (ip->ip_v != IPVERSION) { 389 ipstat.ips_badvers++; 390 goto bad; 391 } 392 hlen = ip->ip_hl << 2; 393 if (hlen < sizeof(struct ip)) { /* minimum header length */ 394 ipstat.ips_badhlen++; 395 goto bad; 396 } 397 if (hlen > m->m_len) { 398 if ((m = m_pullup(m, hlen)) == 0) { 399 ipstat.ips_badhlen++; 400 return; 401 } 402 ip = mtod(m, struct ip *); 403 } 404 /* 405 * we drop packets that have a multicast address as source 406 * as wanted by rfc 1112 407 */ 408 if (IN_MULTICAST(ip->ip_src.s_addr)) { 409 goto bad; 410 } 411 412 if (in_cksum(m, hlen) != 0) { 413 ipstat.ips_badsum++; 414 goto bad; 415 } 416 417 /* 418 * Convert fields to host representation. 419 */ 420 NTOHS(ip->ip_len); 421 NTOHS(ip->ip_off); 422 len = ip->ip_len; 423 424 /* 425 * Check for additional length bogosity 426 */ 427 if (len < hlen) { 428 ipstat.ips_badlen++; 429 goto bad; 430 } 431 432 /* 433 * Check that the amount of data in the buffers 434 * is as at least much as the IP header would have us expect. 435 * Trim mbufs if longer than we expect. 436 * Drop packet if shorter than we expect. 437 */ 438 if (m->m_pkthdr.len < len) { 439 ipstat.ips_tooshort++; 440 goto bad; 441 } 442 if (m->m_pkthdr.len > len) { 443 if (m->m_len == m->m_pkthdr.len) { 444 m->m_len = len; 445 m->m_pkthdr.len = len; 446 } else 447 m_adj(m, len - m->m_pkthdr.len); 448 } 449 450 /* 451 * Assume that we can create a fast-forward IP flow entry 452 * based on this packet. 453 */ 454 m->m_flags |= M_CANFASTFWD; 455 456#ifdef PFIL_HOOKS 457 /* 458 * Run through list of hooks for input packets. If there are any 459 * filters which require that additional packets in the flow are 460 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 461 * Note that filters must _never_ set this flag, as another filter 462 * in the list may have previously cleared it. 463 */ 464 m0 = m; 465 for (pfh = pfil_hook_get(PFIL_IN); pfh; pfh = pfh->pfil_link.tqe_next) 466 if (pfh->pfil_func) { 467 rv = pfh->pfil_func(ip, hlen, m->m_pkthdr.rcvif, 0, &m0); 468 if (rv) 469 return; 470 m = m0; 471 if (m == NULL) 472 return; 473 ip = mtod(m, struct ip *); 474 } 475#endif /* PFIL_HOOKS */ 476 477 /* 478 * Process options and, if not destined for us, 479 * ship it on. ip_dooptions returns 1 when an 480 * error was detected (causing an icmp message 481 * to be sent and the original packet to be freed). 482 */ 483 ip_nhops = 0; /* for source routed packets */ 484 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 485 return; 486 487 /* 488 * Check our list of addresses, to see if the packet is for us. 489 */ 490 INADDR_TO_IA(ip->ip_dst, ia); 491 if (ia != NULL) 492 goto ours; 493 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 494 for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first; 495 ifa != NULL; ifa = ifa->ifa_list.tqe_next) { 496 if (ifa->ifa_addr->sa_family != AF_INET) continue; 497 ia = ifatoia(ifa); 498 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 499 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 500 /* 501 * Look for all-0's host part (old broadcast addr), 502 * either for subnet or net. 503 */ 504 ip->ip_dst.s_addr == ia->ia_subnet || 505 ip->ip_dst.s_addr == ia->ia_net) 506 goto ours; 507 /* 508 * An interface with IP address zero accepts 509 * all packets that arrive on that interface. 510 */ 511 if (in_nullhost(ia->ia_addr.sin_addr)) 512 goto ours; 513 } 514 } 515 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 516 struct in_multi *inm; 517#ifdef MROUTING 518 extern struct socket *ip_mrouter; 519 520 if (m->m_flags & M_EXT) { 521 if ((m = m_pullup(m, hlen)) == 0) { 522 ipstat.ips_toosmall++; 523 return; 524 } 525 ip = mtod(m, struct ip *); 526 } 527 528 if (ip_mrouter) { 529 /* 530 * If we are acting as a multicast router, all 531 * incoming multicast packets are passed to the 532 * kernel-level multicast forwarding function. 533 * The packet is returned (relatively) intact; if 534 * ip_mforward() returns a non-zero value, the packet 535 * must be discarded, else it may be accepted below. 536 * 537 * (The IP ident field is put in the same byte order 538 * as expected when ip_mforward() is called from 539 * ip_output().) 540 */ 541 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 542 ipstat.ips_cantforward++; 543 m_freem(m); 544 return; 545 } 546 547 /* 548 * The process-level routing demon needs to receive 549 * all multicast IGMP packets, whether or not this 550 * host belongs to their destination groups. 551 */ 552 if (ip->ip_p == IPPROTO_IGMP) 553 goto ours; 554 ipstat.ips_forward++; 555 } 556#endif 557 /* 558 * See if we belong to the destination multicast group on the 559 * arrival interface. 560 */ 561 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 562 if (inm == NULL) { 563 ipstat.ips_cantforward++; 564 m_freem(m); 565 return; 566 } 567 goto ours; 568 } 569 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 570 in_nullhost(ip->ip_dst)) 571 goto ours; 572 573 /* 574 * Not for us; forward if possible and desirable. 575 */ 576 if (ipforwarding == 0) { 577 ipstat.ips_cantforward++; 578 m_freem(m); 579 } else 580 ip_forward(m, 0); 581 return; 582 583ours: 584 /* 585 * If offset or IP_MF are set, must reassemble. 586 * Otherwise, nothing need be done. 587 * (We could look in the reassembly queue to see 588 * if the packet was previously fragmented, 589 * but it's not worth the time; just let them time out.) 590 */ 591 if (ip->ip_off & ~(IP_DF|IP_RF)) { 592 /* 593 * Look for queue of fragments 594 * of this datagram. 595 */ 596 IPQ_LOCK(); 597 for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next) 598 if (ip->ip_id == fp->ipq_id && 599 in_hosteq(ip->ip_src, fp->ipq_src) && 600 in_hosteq(ip->ip_dst, fp->ipq_dst) && 601 ip->ip_p == fp->ipq_p) 602 goto found; 603 fp = 0; 604found: 605 606 /* 607 * Adjust ip_len to not reflect header, 608 * set ipqe_mff if more fragments are expected, 609 * convert offset of this to bytes. 610 */ 611 ip->ip_len -= hlen; 612 mff = (ip->ip_off & IP_MF) != 0; 613 if (mff) { 614 /* 615 * Make sure that fragments have a data length 616 * that's a non-zero multiple of 8 bytes. 617 */ 618 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 619 ipstat.ips_badfrags++; 620 IPQ_UNLOCK(); 621 goto bad; 622 } 623 } 624 ip->ip_off <<= 3; 625 626 /* 627 * If datagram marked as having more fragments 628 * or if this is not the first fragment, 629 * attempt reassembly; if it succeeds, proceed. 630 */ 631 if (mff || ip->ip_off) { 632 ipstat.ips_fragments++; 633 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 634 if (ipqe == NULL) { 635 ipstat.ips_rcvmemdrop++; 636 IPQ_UNLOCK(); 637 goto bad; 638 } 639 ipqe->ipqe_mff = mff; 640 ipqe->ipqe_m = m; 641 ipqe->ipqe_ip = ip; 642 m = ip_reass(ipqe, fp); 643 if (m == 0) { 644 IPQ_UNLOCK(); 645 return; 646 } 647 ipstat.ips_reassembled++; 648 ip = mtod(m, struct ip *); 649 hlen = ip->ip_hl << 2; 650 ip->ip_len += hlen; 651 } else 652 if (fp) 653 ip_freef(fp); 654 IPQ_UNLOCK(); 655 } 656 657 /* 658 * Switch out to protocol's input routine. 659 */ 660#if IFA_STATS 661 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 662#endif 663 ipstat.ips_delivered++; 664 { 665 int off = hlen, nh = ip->ip_p; 666 667 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 668 return; 669 } 670bad: 671 m_freem(m); 672} 673 674/* 675 * Take incoming datagram fragment and try to 676 * reassemble it into whole datagram. If a chain for 677 * reassembly of this datagram already exists, then it 678 * is given as fp; otherwise have to make a chain. 679 */ 680struct mbuf * 681ip_reass(ipqe, fp) 682 register struct ipqent *ipqe; 683 register struct ipq *fp; 684{ 685 register struct mbuf *m = ipqe->ipqe_m; 686 register struct ipqent *nq, *p, *q; 687 struct ip *ip; 688 struct mbuf *t; 689 int hlen = ipqe->ipqe_ip->ip_hl << 2; 690 int i, next; 691 692 IPQ_LOCK_CHECK(); 693 694 /* 695 * Presence of header sizes in mbufs 696 * would confuse code below. 697 */ 698 m->m_data += hlen; 699 m->m_len -= hlen; 700 701 /* 702 * If first fragment to arrive, create a reassembly queue. 703 */ 704 if (fp == 0) { 705 MALLOC(fp, struct ipq *, sizeof (struct ipq), 706 M_FTABLE, M_NOWAIT); 707 if (fp == NULL) 708 goto dropfrag; 709 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 710 fp->ipq_ttl = IPFRAGTTL; 711 fp->ipq_p = ipqe->ipqe_ip->ip_p; 712 fp->ipq_id = ipqe->ipqe_ip->ip_id; 713 LIST_INIT(&fp->ipq_fragq); 714 fp->ipq_src = ipqe->ipqe_ip->ip_src; 715 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 716 p = NULL; 717 goto insert; 718 } 719 720 /* 721 * Find a segment which begins after this one does. 722 */ 723 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 724 p = q, q = q->ipqe_q.le_next) 725 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 726 break; 727 728 /* 729 * If there is a preceding segment, it may provide some of 730 * our data already. If so, drop the data from the incoming 731 * segment. If it provides all of our data, drop us. 732 */ 733 if (p != NULL) { 734 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 735 ipqe->ipqe_ip->ip_off; 736 if (i > 0) { 737 if (i >= ipqe->ipqe_ip->ip_len) 738 goto dropfrag; 739 m_adj(ipqe->ipqe_m, i); 740 ipqe->ipqe_ip->ip_off += i; 741 ipqe->ipqe_ip->ip_len -= i; 742 } 743 } 744 745 /* 746 * While we overlap succeeding segments trim them or, 747 * if they are completely covered, dequeue them. 748 */ 749 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 750 q->ipqe_ip->ip_off; q = nq) { 751 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 752 q->ipqe_ip->ip_off; 753 if (i < q->ipqe_ip->ip_len) { 754 q->ipqe_ip->ip_len -= i; 755 q->ipqe_ip->ip_off += i; 756 m_adj(q->ipqe_m, i); 757 break; 758 } 759 nq = q->ipqe_q.le_next; 760 m_freem(q->ipqe_m); 761 LIST_REMOVE(q, ipqe_q); 762 pool_put(&ipqent_pool, q); 763 } 764 765insert: 766 /* 767 * Stick new segment in its place; 768 * check for complete reassembly. 769 */ 770 if (p == NULL) { 771 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 772 } else { 773 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 774 } 775 next = 0; 776 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 777 p = q, q = q->ipqe_q.le_next) { 778 if (q->ipqe_ip->ip_off != next) 779 return (0); 780 next += q->ipqe_ip->ip_len; 781 } 782 if (p->ipqe_mff) 783 return (0); 784 785 /* 786 * Reassembly is complete. Check for a bogus message size and 787 * concatenate fragments. 788 */ 789 q = fp->ipq_fragq.lh_first; 790 ip = q->ipqe_ip; 791 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 792 ipstat.ips_toolong++; 793 ip_freef(fp); 794 return (0); 795 } 796 m = q->ipqe_m; 797 t = m->m_next; 798 m->m_next = 0; 799 m_cat(m, t); 800 nq = q->ipqe_q.le_next; 801 pool_put(&ipqent_pool, q); 802 for (q = nq; q != NULL; q = nq) { 803 t = q->ipqe_m; 804 nq = q->ipqe_q.le_next; 805 pool_put(&ipqent_pool, q); 806 m_cat(m, t); 807 } 808 809 /* 810 * Create header for new ip packet by 811 * modifying header of first packet; 812 * dequeue and discard fragment reassembly header. 813 * Make header visible. 814 */ 815 ip->ip_len = next; 816 ip->ip_ttl = 0; /* xxx */ 817 ip->ip_sum = 0; 818 ip->ip_src = fp->ipq_src; 819 ip->ip_dst = fp->ipq_dst; 820 LIST_REMOVE(fp, ipq_q); 821 FREE(fp, M_FTABLE); 822 m->m_len += (ip->ip_hl << 2); 823 m->m_data -= (ip->ip_hl << 2); 824 /* some debugging cruft by sklower, below, will go away soon */ 825 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 826 register int plen = 0; 827 for (t = m; t; t = t->m_next) 828 plen += t->m_len; 829 m->m_pkthdr.len = plen; 830 } 831 return (m); 832 833dropfrag: 834 ipstat.ips_fragdropped++; 835 m_freem(m); 836 pool_put(&ipqent_pool, ipqe); 837 return (0); 838} 839 840/* 841 * Free a fragment reassembly header and all 842 * associated datagrams. 843 */ 844void 845ip_freef(fp) 846 struct ipq *fp; 847{ 848 register struct ipqent *q, *p; 849 850 IPQ_LOCK_CHECK(); 851 852 for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { 853 p = q->ipqe_q.le_next; 854 m_freem(q->ipqe_m); 855 LIST_REMOVE(q, ipqe_q); 856 pool_put(&ipqent_pool, q); 857 } 858 LIST_REMOVE(fp, ipq_q); 859 FREE(fp, M_FTABLE); 860} 861 862/* 863 * IP timer processing; 864 * if a timer expires on a reassembly 865 * queue, discard it. 866 */ 867void 868ip_slowtimo() 869{ 870 register struct ipq *fp, *nfp; 871 int s = splsoftnet(); 872 873 IPQ_LOCK(); 874 for (fp = ipq.lh_first; fp != NULL; fp = nfp) { 875 nfp = fp->ipq_q.le_next; 876 if (--fp->ipq_ttl == 0) { 877 ipstat.ips_fragtimeout++; 878 ip_freef(fp); 879 } 880 } 881 IPQ_UNLOCK(); 882#ifdef GATEWAY 883 ipflow_slowtimo(); 884#endif 885 splx(s); 886} 887 888/* 889 * Drain off all datagram fragments. 890 */ 891void 892ip_drain() 893{ 894 895 /* 896 * We may be called from a device's interrupt context. If 897 * the ipq is already busy, just bail out now. 898 */ 899 if (ipq_lock_try() == 0) 900 return; 901 902 while (ipq.lh_first != NULL) { 903 ipstat.ips_fragdropped++; 904 ip_freef(ipq.lh_first); 905 } 906 907 IPQ_UNLOCK(); 908} 909 910/* 911 * Do option processing on a datagram, 912 * possibly discarding it if bad options are encountered, 913 * or forwarding it if source-routed. 914 * Returns 1 if packet has been forwarded/freed, 915 * 0 if the packet should be processed further. 916 */ 917int 918ip_dooptions(m) 919 struct mbuf *m; 920{ 921 register struct ip *ip = mtod(m, struct ip *); 922 register u_char *cp; 923 register struct ip_timestamp *ipt; 924 register struct in_ifaddr *ia; 925 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 926 struct in_addr *sin, dst; 927 n_time ntime; 928 929 dst = ip->ip_dst; 930 cp = (u_char *)(ip + 1); 931 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 932 for (; cnt > 0; cnt -= optlen, cp += optlen) { 933 opt = cp[IPOPT_OPTVAL]; 934 if (opt == IPOPT_EOL) 935 break; 936 if (opt == IPOPT_NOP) 937 optlen = 1; 938 else { 939 optlen = cp[IPOPT_OLEN]; 940 if (optlen <= 0 || optlen > cnt) { 941 code = &cp[IPOPT_OLEN] - (u_char *)ip; 942 goto bad; 943 } 944 } 945 switch (opt) { 946 947 default: 948 break; 949 950 /* 951 * Source routing with record. 952 * Find interface with current destination address. 953 * If none on this machine then drop if strictly routed, 954 * or do nothing if loosely routed. 955 * Record interface address and bring up next address 956 * component. If strictly routed make sure next 957 * address is on directly accessible net. 958 */ 959 case IPOPT_LSRR: 960 case IPOPT_SSRR: 961 if (ip_allowsrcrt == 0) { 962 type = ICMP_UNREACH; 963 code = ICMP_UNREACH_NET_PROHIB; 964 goto bad; 965 } 966 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 967 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 968 goto bad; 969 } 970 ipaddr.sin_addr = ip->ip_dst; 971 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 972 if (ia == 0) { 973 if (opt == IPOPT_SSRR) { 974 type = ICMP_UNREACH; 975 code = ICMP_UNREACH_SRCFAIL; 976 goto bad; 977 } 978 /* 979 * Loose routing, and not at next destination 980 * yet; nothing to do except forward. 981 */ 982 break; 983 } 984 off--; /* 0 origin */ 985 if (off > optlen - sizeof(struct in_addr)) { 986 /* 987 * End of source route. Should be for us. 988 */ 989 save_rte(cp, ip->ip_src); 990 break; 991 } 992 /* 993 * locate outgoing interface 994 */ 995 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 996 sizeof(ipaddr.sin_addr)); 997 if (opt == IPOPT_SSRR) { 998#define INA struct in_ifaddr * 999#define SA struct sockaddr * 1000 ia = (INA)ifa_ifwithladdr((SA)&ipaddr); 1001 } else 1002 ia = ip_rtaddr(ipaddr.sin_addr); 1003 if (ia == 0) { 1004 type = ICMP_UNREACH; 1005 code = ICMP_UNREACH_SRCFAIL; 1006 goto bad; 1007 } 1008 ip->ip_dst = ipaddr.sin_addr; 1009 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1010 (caddr_t)(cp + off), sizeof(struct in_addr)); 1011 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1012 /* 1013 * Let ip_intr's mcast routing check handle mcast pkts 1014 */ 1015 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1016 break; 1017 1018 case IPOPT_RR: 1019 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1020 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1021 goto bad; 1022 } 1023 /* 1024 * If no space remains, ignore. 1025 */ 1026 off--; /* 0 origin */ 1027 if (off > optlen - sizeof(struct in_addr)) 1028 break; 1029 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1030 sizeof(ipaddr.sin_addr)); 1031 /* 1032 * locate outgoing interface; if we're the destination, 1033 * use the incoming interface (should be same). 1034 */ 1035 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1036 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1037 type = ICMP_UNREACH; 1038 code = ICMP_UNREACH_HOST; 1039 goto bad; 1040 } 1041 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1042 (caddr_t)(cp + off), sizeof(struct in_addr)); 1043 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1044 break; 1045 1046 case IPOPT_TS: 1047 code = cp - (u_char *)ip; 1048 ipt = (struct ip_timestamp *)cp; 1049 if (ipt->ipt_len < 5) 1050 goto bad; 1051 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1052 if (++ipt->ipt_oflw == 0) 1053 goto bad; 1054 break; 1055 } 1056 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1057 switch (ipt->ipt_flg) { 1058 1059 case IPOPT_TS_TSONLY: 1060 break; 1061 1062 case IPOPT_TS_TSANDADDR: 1063 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1064 sizeof(struct in_addr) > ipt->ipt_len) 1065 goto bad; 1066 ipaddr.sin_addr = dst; 1067 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1068 m->m_pkthdr.rcvif); 1069 if (ia == 0) 1070 continue; 1071 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1072 (caddr_t)sin, sizeof(struct in_addr)); 1073 ipt->ipt_ptr += sizeof(struct in_addr); 1074 break; 1075 1076 case IPOPT_TS_PRESPEC: 1077 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1078 sizeof(struct in_addr) > ipt->ipt_len) 1079 goto bad; 1080 bcopy((caddr_t)sin, (caddr_t)&ipaddr.sin_addr, 1081 sizeof(struct in_addr)); 1082 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1083 continue; 1084 ipt->ipt_ptr += sizeof(struct in_addr); 1085 break; 1086 1087 default: 1088 goto bad; 1089 } 1090 ntime = iptime(); 1091 bcopy((caddr_t)&ntime, (caddr_t)cp + ipt->ipt_ptr - 1, 1092 sizeof(n_time)); 1093 ipt->ipt_ptr += sizeof(n_time); 1094 } 1095 } 1096 if (forward) { 1097 if (ip_forwsrcrt == 0) { 1098 type = ICMP_UNREACH; 1099 code = ICMP_UNREACH_SRCFAIL; 1100 goto bad; 1101 } 1102 ip_forward(m, 1); 1103 return (1); 1104 } 1105 return (0); 1106bad: 1107 icmp_error(m, type, code, 0, 0); 1108 ipstat.ips_badoptions++; 1109 return (1); 1110} 1111 1112/* 1113 * Given address of next destination (final or next hop), 1114 * return internet address info of interface to be used to get there. 1115 */ 1116struct in_ifaddr * 1117ip_rtaddr(dst) 1118 struct in_addr dst; 1119{ 1120 register struct sockaddr_in *sin; 1121 1122 sin = satosin(&ipforward_rt.ro_dst); 1123 1124 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1125 if (ipforward_rt.ro_rt) { 1126 RTFREE(ipforward_rt.ro_rt); 1127 ipforward_rt.ro_rt = 0; 1128 } 1129 sin->sin_family = AF_INET; 1130 sin->sin_len = sizeof(*sin); 1131 sin->sin_addr = dst; 1132 1133 rtalloc(&ipforward_rt); 1134 } 1135 if (ipforward_rt.ro_rt == 0) 1136 return ((struct in_ifaddr *)0); 1137 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1138} 1139 1140/* 1141 * Save incoming source route for use in replies, 1142 * to be picked up later by ip_srcroute if the receiver is interested. 1143 */ 1144void 1145save_rte(option, dst) 1146 u_char *option; 1147 struct in_addr dst; 1148{ 1149 unsigned olen; 1150 1151 olen = option[IPOPT_OLEN]; 1152#ifdef DIAGNOSTIC 1153 if (ipprintfs) 1154 printf("save_rte: olen %d\n", olen); 1155#endif /* 0 */ 1156 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1157 return; 1158 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1159 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1160 ip_srcrt.dst = dst; 1161} 1162 1163/* 1164 * Retrieve incoming source route for use in replies, 1165 * in the same form used by setsockopt. 1166 * The first hop is placed before the options, will be removed later. 1167 */ 1168struct mbuf * 1169ip_srcroute() 1170{ 1171 register struct in_addr *p, *q; 1172 register struct mbuf *m; 1173 1174 if (ip_nhops == 0) 1175 return ((struct mbuf *)0); 1176 m = m_get(M_DONTWAIT, MT_SOOPTS); 1177 if (m == 0) 1178 return ((struct mbuf *)0); 1179 1180#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1181 1182 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1183 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1184 OPTSIZ; 1185#ifdef DIAGNOSTIC 1186 if (ipprintfs) 1187 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1188#endif 1189 1190 /* 1191 * First save first hop for return route 1192 */ 1193 p = &ip_srcrt.route[ip_nhops - 1]; 1194 *(mtod(m, struct in_addr *)) = *p--; 1195#ifdef DIAGNOSTIC 1196 if (ipprintfs) 1197 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1198#endif 1199 1200 /* 1201 * Copy option fields and padding (nop) to mbuf. 1202 */ 1203 ip_srcrt.nop = IPOPT_NOP; 1204 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1205 bcopy((caddr_t)&ip_srcrt.nop, 1206 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1207 q = (struct in_addr *)(mtod(m, caddr_t) + 1208 sizeof(struct in_addr) + OPTSIZ); 1209#undef OPTSIZ 1210 /* 1211 * Record return path as an IP source route, 1212 * reversing the path (pointers are now aligned). 1213 */ 1214 while (p >= ip_srcrt.route) { 1215#ifdef DIAGNOSTIC 1216 if (ipprintfs) 1217 printf(" %x", ntohl(q->s_addr)); 1218#endif 1219 *q++ = *p--; 1220 } 1221 /* 1222 * Last hop goes to final destination. 1223 */ 1224 *q = ip_srcrt.dst; 1225#ifdef DIAGNOSTIC 1226 if (ipprintfs) 1227 printf(" %x\n", ntohl(q->s_addr)); 1228#endif 1229 return (m); 1230} 1231 1232/* 1233 * Strip out IP options, at higher 1234 * level protocol in the kernel. 1235 * Second argument is buffer to which options 1236 * will be moved, and return value is their length. 1237 * XXX should be deleted; last arg currently ignored. 1238 */ 1239void 1240ip_stripoptions(m, mopt) 1241 register struct mbuf *m; 1242 struct mbuf *mopt; 1243{ 1244 register int i; 1245 struct ip *ip = mtod(m, struct ip *); 1246 register caddr_t opts; 1247 int olen; 1248 1249 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1250 opts = (caddr_t)(ip + 1); 1251 i = m->m_len - (sizeof (struct ip) + olen); 1252 bcopy(opts + olen, opts, (unsigned)i); 1253 m->m_len -= olen; 1254 if (m->m_flags & M_PKTHDR) 1255 m->m_pkthdr.len -= olen; 1256 ip->ip_len -= olen; 1257 ip->ip_hl = sizeof (struct ip) >> 2; 1258} 1259 1260int inetctlerrmap[PRC_NCMDS] = { 1261 0, 0, 0, 0, 1262 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1263 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1264 EMSGSIZE, EHOSTUNREACH, 0, 0, 1265 0, 0, 0, 0, 1266 ENOPROTOOPT 1267}; 1268 1269/* 1270 * Forward a packet. If some error occurs return the sender 1271 * an icmp packet. Note we can't always generate a meaningful 1272 * icmp message because icmp doesn't have a large enough repertoire 1273 * of codes and types. 1274 * 1275 * If not forwarding, just drop the packet. This could be confusing 1276 * if ipforwarding was zero but some routing protocol was advancing 1277 * us as a gateway to somewhere. However, we must let the routing 1278 * protocol deal with that. 1279 * 1280 * The srcrt parameter indicates whether the packet is being forwarded 1281 * via a source route. 1282 */ 1283void 1284ip_forward(m, srcrt) 1285 struct mbuf *m; 1286 int srcrt; 1287{ 1288 register struct ip *ip = mtod(m, struct ip *); 1289 register struct sockaddr_in *sin; 1290 register struct rtentry *rt; 1291 int error, type = 0, code = 0; 1292 struct mbuf *mcopy; 1293 n_long dest; 1294 struct ifnet *destifp; 1295#ifdef IPSEC 1296 struct ifnet dummyifp; 1297#endif 1298 1299 dest = 0; 1300#ifdef DIAGNOSTIC 1301 if (ipprintfs) 1302 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1303 ntohl(ip->ip_src.s_addr), 1304 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1305#endif 1306 if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) { 1307 ipstat.ips_cantforward++; 1308 m_freem(m); 1309 return; 1310 } 1311 if (ip->ip_ttl <= IPTTLDEC) { 1312 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1313 return; 1314 } 1315 ip->ip_ttl -= IPTTLDEC; 1316 1317 sin = satosin(&ipforward_rt.ro_dst); 1318 if ((rt = ipforward_rt.ro_rt) == 0 || 1319 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1320 if (ipforward_rt.ro_rt) { 1321 RTFREE(ipforward_rt.ro_rt); 1322 ipforward_rt.ro_rt = 0; 1323 } 1324 sin->sin_family = AF_INET; 1325 sin->sin_len = sizeof(struct sockaddr_in); 1326 sin->sin_addr = ip->ip_dst; 1327 1328 rtalloc(&ipforward_rt); 1329 if (ipforward_rt.ro_rt == 0) { 1330 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1331 return; 1332 } 1333 rt = ipforward_rt.ro_rt; 1334 } 1335 1336 /* 1337 * Save at most 68 bytes of the packet in case 1338 * we need to generate an ICMP message to the src. 1339 */ 1340 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 68)); 1341 1342 /* 1343 * If forwarding packet using same interface that it came in on, 1344 * perhaps should send a redirect to sender to shortcut a hop. 1345 * Only send redirect if source is sending directly to us, 1346 * and if packet was not source routed (or has any options). 1347 * Also, don't send redirect if forwarding using a default route 1348 * or a route modified by a redirect. 1349 */ 1350 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1351 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1352 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1353 ipsendredirects && !srcrt) { 1354 if (rt->rt_ifa && 1355 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1356 ifatoia(rt->rt_ifa)->ia_subnet) { 1357 if (rt->rt_flags & RTF_GATEWAY) 1358 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1359 else 1360 dest = ip->ip_dst.s_addr; 1361 /* 1362 * Router requirements says to only send host 1363 * redirects. 1364 */ 1365 type = ICMP_REDIRECT; 1366 code = ICMP_REDIRECT_HOST; 1367#ifdef DIAGNOSTIC 1368 if (ipprintfs) 1369 printf("redirect (%d) to %x\n", code, 1370 (u_int32_t)dest); 1371#endif 1372 } 1373 } 1374 1375#ifdef IPSEC 1376 m->m_pkthdr.rcvif = NULL; 1377#endif /*IPSEC*/ 1378 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1379 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1380 if (error) 1381 ipstat.ips_cantforward++; 1382 else { 1383 ipstat.ips_forward++; 1384 if (type) 1385 ipstat.ips_redirectsent++; 1386 else { 1387 if (mcopy) { 1388#ifdef GATEWAY 1389 if (mcopy->m_flags & M_CANFASTFWD) 1390 ipflow_create(&ipforward_rt, mcopy); 1391#endif 1392 m_freem(mcopy); 1393 } 1394 return; 1395 } 1396 } 1397 if (mcopy == NULL) 1398 return; 1399 destifp = NULL; 1400 1401 switch (error) { 1402 1403 case 0: /* forwarded, but need redirect */ 1404 /* type, code set above */ 1405 break; 1406 1407 case ENETUNREACH: /* shouldn't happen, checked above */ 1408 case EHOSTUNREACH: 1409 case ENETDOWN: 1410 case EHOSTDOWN: 1411 default: 1412 type = ICMP_UNREACH; 1413 code = ICMP_UNREACH_HOST; 1414 break; 1415 1416 case EMSGSIZE: 1417 type = ICMP_UNREACH; 1418 code = ICMP_UNREACH_NEEDFRAG; 1419#ifndef IPSEC 1420 if (ipforward_rt.ro_rt) 1421 destifp = ipforward_rt.ro_rt->rt_ifp; 1422#else 1423 /* 1424 * If the packet is routed over IPsec tunnel, tell the 1425 * originator the tunnel MTU. 1426 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1427 * XXX quickhack!!! 1428 */ 1429 if (ipforward_rt.ro_rt) { 1430 struct secpolicy *sp; 1431 int ipsecerror; 1432 int ipsechdr; 1433 struct route *ro; 1434 1435 sp = ipsec4_getpolicybyaddr(mcopy, 1436 IP_FORWARDING, 1437 &ipsecerror); 1438 1439 if (sp == NULL) 1440 destifp = ipforward_rt.ro_rt->rt_ifp; 1441 else { 1442 /* count IPsec header size */ 1443 ipsechdr = ipsec4_hdrsiz(mcopy, NULL); 1444 1445 /* 1446 * find the correct route for outer IPv4 1447 * header, compute tunnel MTU. 1448 * 1449 * XXX BUG ALERT 1450 * The "dummyifp" code relies upon the fact 1451 * that icmp_error() touches only ifp->if_mtu. 1452 */ 1453 /*XXX*/ 1454 destifp = NULL; 1455 if (sp->req != NULL 1456 && sp->req->sa != NULL) { 1457 ro = &sp->req->sa->saidx->sa_route; 1458 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1459 dummyifp.if_mtu = 1460 ro->ro_rt->rt_ifp->if_mtu; 1461 dummyifp.if_mtu -= ipsechdr; 1462 destifp = &dummyifp; 1463 } 1464 } 1465 1466 key_freesp(sp); 1467 } 1468 } 1469#endif /*IPSEC*/ 1470 ipstat.ips_cantfrag++; 1471 break; 1472 1473 case ENOBUFS: 1474 type = ICMP_SOURCEQUENCH; 1475 code = 0; 1476 break; 1477 } 1478 icmp_error(mcopy, type, code, dest, destifp); 1479} 1480 1481void 1482ip_savecontrol(inp, mp, ip, m) 1483 register struct inpcb *inp; 1484 register struct mbuf **mp; 1485 register struct ip *ip; 1486 register struct mbuf *m; 1487{ 1488 1489 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1490 struct timeval tv; 1491 1492 microtime(&tv); 1493 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1494 SCM_TIMESTAMP, SOL_SOCKET); 1495 if (*mp) 1496 mp = &(*mp)->m_next; 1497 } 1498 if (inp->inp_flags & INP_RECVDSTADDR) { 1499 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1500 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1501 if (*mp) 1502 mp = &(*mp)->m_next; 1503 } 1504#ifdef notyet 1505 /* 1506 * XXX 1507 * Moving these out of udp_input() made them even more broken 1508 * than they already were. 1509 * - fenner@parc.xerox.com 1510 */ 1511 /* options were tossed already */ 1512 if (inp->inp_flags & INP_RECVOPTS) { 1513 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1514 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1515 if (*mp) 1516 mp = &(*mp)->m_next; 1517 } 1518 /* ip_srcroute doesn't do what we want here, need to fix */ 1519 if (inp->inp_flags & INP_RECVRETOPTS) { 1520 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1521 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1522 if (*mp) 1523 mp = &(*mp)->m_next; 1524 } 1525#endif 1526 if (inp->inp_flags & INP_RECVIF) { 1527 struct sockaddr_dl sdl; 1528 1529 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1530 sdl.sdl_family = AF_LINK; 1531 sdl.sdl_index = m->m_pkthdr.rcvif ? 1532 m->m_pkthdr.rcvif->if_index : 0; 1533 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1534 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1535 IP_RECVIF, IPPROTO_IP); 1536 if (*mp) 1537 mp = &(*mp)->m_next; 1538 } 1539} 1540 1541int 1542ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1543 int *name; 1544 u_int namelen; 1545 void *oldp; 1546 size_t *oldlenp; 1547 void *newp; 1548 size_t newlen; 1549{ 1550 extern int subnetsarelocal, hostzeroisbroadcast; 1551 1552 int error, old; 1553 1554 /* All sysctl names at this level are terminal. */ 1555 if (namelen != 1) 1556 return (ENOTDIR); 1557 1558 switch (name[0]) { 1559 case IPCTL_FORWARDING: 1560 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1561 case IPCTL_SENDREDIRECTS: 1562 return (sysctl_int(oldp, oldlenp, newp, newlen, 1563 &ipsendredirects)); 1564 case IPCTL_DEFTTL: 1565 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1566#ifdef notyet 1567 case IPCTL_DEFMTU: 1568 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1569#endif 1570 case IPCTL_FORWSRCRT: 1571 /* Don't allow this to change in a secure environment. */ 1572 if (securelevel > 0) 1573 return (sysctl_rdint(oldp, oldlenp, newp, 1574 ip_forwsrcrt)); 1575 else 1576 return (sysctl_int(oldp, oldlenp, newp, newlen, 1577 &ip_forwsrcrt)); 1578 case IPCTL_DIRECTEDBCAST: 1579 return (sysctl_int(oldp, oldlenp, newp, newlen, 1580 &ip_directedbcast)); 1581 case IPCTL_ALLOWSRCRT: 1582 return (sysctl_int(oldp, oldlenp, newp, newlen, 1583 &ip_allowsrcrt)); 1584 case IPCTL_SUBNETSARELOCAL: 1585 return (sysctl_int(oldp, oldlenp, newp, newlen, 1586 &subnetsarelocal)); 1587 case IPCTL_MTUDISC: 1588 error = sysctl_int(oldp, oldlenp, newp, newlen, 1589 &ip_mtudisc); 1590 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1591 ip_mtudisc_timeout_q = 1592 rt_timer_queue_create(ip_mtudisc_timeout); 1593 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1594 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1595 ip_mtudisc_timeout_q = NULL; 1596 } 1597 return error; 1598 case IPCTL_ANONPORTMIN: 1599 old = anonportmin; 1600 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1601 if (anonportmin >= anonportmax || anonportmin > 65535 1602#ifndef IPNOPRIVPORTS 1603 || anonportmin < IPPORT_RESERVED 1604#endif 1605 ) { 1606 anonportmin = old; 1607 return (EINVAL); 1608 } 1609 return (error); 1610 case IPCTL_ANONPORTMAX: 1611 old = anonportmax; 1612 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1613 if (anonportmin >= anonportmax || anonportmax > 65535 1614#ifndef IPNOPRIVPORTS 1615 || anonportmax < IPPORT_RESERVED 1616#endif 1617 ) { 1618 anonportmax = old; 1619 return (EINVAL); 1620 } 1621 return (error); 1622 case IPCTL_MTUDISCTIMEOUT: 1623 error = sysctl_int(oldp, oldlenp, newp, newlen, 1624 &ip_mtudisc_timeout); 1625 if (ip_mtudisc_timeout_q != NULL) 1626 rt_timer_queue_change(ip_mtudisc_timeout_q, 1627 ip_mtudisc_timeout); 1628 return (error); 1629#ifdef GATEWAY 1630 case IPCTL_MAXFLOWS: 1631 { 1632 int s; 1633 1634 error = sysctl_int(oldp, oldlenp, newp, newlen, 1635 &ip_maxflows); 1636 s = splsoftnet(); 1637 ipflow_reap(0); 1638 splx(s); 1639 return (error); 1640 } 1641#endif 1642 case IPCTL_HOSTZEROBROADCAST: 1643 return (sysctl_int(oldp, oldlenp, newp, newlen, 1644 &hostzeroisbroadcast)); 1645#if NGIF > 0 1646 case IPCTL_GIF_TTL: 1647 return(sysctl_int(oldp, oldlenp, newp, newlen, 1648 &ip_gif_ttl)); 1649#endif 1650 1651 default: 1652 return (EOPNOTSUPP); 1653 } 1654 /* NOTREACHED */ 1655} 1656