ip_input.c revision 1.90
1/* $NetBSD: ip_input.c,v 1.90 1999/07/06 12:23:20 itojun Exp $ */ 2 3/* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32/*- 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 3. All advertising materials mentioning features or use of this software 49 * must display the following acknowledgement: 50 * This product includes software developed by the NetBSD 51 * Foundation, Inc. and its contributors. 52 * 4. Neither the name of The NetBSD Foundation nor the names of its 53 * contributors may be used to endorse or promote products derived 54 * from this software without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 57 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 58 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 59 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 60 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 61 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 62 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 63 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 64 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 65 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 66 * POSSIBILITY OF SUCH DAMAGE. 67 */ 68 69/* 70 * Copyright (c) 1982, 1986, 1988, 1993 71 * The Regents of the University of California. All rights reserved. 72 * 73 * Redistribution and use in source and binary forms, with or without 74 * modification, are permitted provided that the following conditions 75 * are met: 76 * 1. Redistributions of source code must retain the above copyright 77 * notice, this list of conditions and the following disclaimer. 78 * 2. Redistributions in binary form must reproduce the above copyright 79 * notice, this list of conditions and the following disclaimer in the 80 * documentation and/or other materials provided with the distribution. 81 * 3. All advertising materials mentioning features or use of this software 82 * must display the following acknowledgement: 83 * This product includes software developed by the University of 84 * California, Berkeley and its contributors. 85 * 4. Neither the name of the University nor the names of its contributors 86 * may be used to endorse or promote products derived from this software 87 * without specific prior written permission. 88 * 89 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 90 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 91 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 92 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 93 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 94 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 95 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 96 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 97 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 98 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 99 * SUCH DAMAGE. 100 * 101 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 102 */ 103 104#include "opt_gateway.h" 105#include "opt_pfil_hooks.h" 106#include "opt_mrouting.h" 107 108#include <sys/param.h> 109#include <sys/systm.h> 110#include <sys/malloc.h> 111#include <sys/mbuf.h> 112#include <sys/domain.h> 113#include <sys/protosw.h> 114#include <sys/socket.h> 115#include <sys/socketvar.h> 116#include <sys/errno.h> 117#include <sys/time.h> 118#include <sys/kernel.h> 119#include <sys/proc.h> 120#include <sys/pool.h> 121 122#include <vm/vm.h> 123#include <sys/sysctl.h> 124 125#include <net/if.h> 126#include <net/if_dl.h> 127#include <net/route.h> 128#include <net/pfil.h> 129 130#include <netinet/in.h> 131#include <netinet/in_systm.h> 132#include <netinet/ip.h> 133#include <netinet/in_pcb.h> 134#include <netinet/in_var.h> 135#include <netinet/ip_var.h> 136#include <netinet/ip_icmp.h> 137/* just for gif_ttl */ 138#include <netinet/in_gif.h> 139#include "gif.h" 140 141#ifdef IPSEC 142#include <netinet6/ipsec.h> 143#include <netinet6/ah.h> 144#ifdef IPSEC_ESP 145#include <netinet6/esp.h> 146#endif 147#include <netkey/key.h> 148#include <netkey/key_debug.h> 149#endif 150 151#ifndef IPFORWARDING 152#ifdef GATEWAY 153#define IPFORWARDING 1 /* forward IP packets not for us */ 154#else /* GATEWAY */ 155#define IPFORWARDING 0 /* don't forward IP packets not for us */ 156#endif /* GATEWAY */ 157#endif /* IPFORWARDING */ 158#ifndef IPSENDREDIRECTS 159#define IPSENDREDIRECTS 1 160#endif 161#ifndef IPFORWSRCRT 162#define IPFORWSRCRT 1 /* forward source-routed packets */ 163#endif 164#ifndef IPALLOWSRCRT 165#define IPALLOWSRCRT 1 /* allow source-routed packets */ 166#endif 167#ifndef IPMTUDISC 168#define IPMTUDISC 0 169#endif 170#ifndef IPMTUDISCTIMEOUT 171#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 172#endif 173 174/* 175 * Note: DIRECTED_BROADCAST is handled this way so that previous 176 * configuration using this option will Just Work. 177 */ 178#ifndef IPDIRECTEDBCAST 179#ifdef DIRECTED_BROADCAST 180#define IPDIRECTEDBCAST 1 181#else 182#define IPDIRECTEDBCAST 0 183#endif /* DIRECTED_BROADCAST */ 184#endif /* IPDIRECTEDBCAST */ 185int ipforwarding = IPFORWARDING; 186int ipsendredirects = IPSENDREDIRECTS; 187int ip_defttl = IPDEFTTL; 188int ip_forwsrcrt = IPFORWSRCRT; 189int ip_directedbcast = IPDIRECTEDBCAST; 190int ip_allowsrcrt = IPALLOWSRCRT; 191int ip_mtudisc = IPMTUDISC; 192u_int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 193#ifdef DIAGNOSTIC 194int ipprintfs = 0; 195#endif 196 197struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 198 199extern struct domain inetdomain; 200extern struct protosw inetsw[]; 201u_char ip_protox[IPPROTO_MAX]; 202int ipqmaxlen = IFQ_MAXLEN; 203struct in_ifaddrhead in_ifaddr; 204struct in_ifaddrhashhead *in_ifaddrhashtbl; 205struct ifqueue ipintrq; 206struct ipstat ipstat; 207u_int16_t ip_id; 208int ip_defttl; 209 210struct ipqhead ipq; 211int ipq_locked; 212 213static __inline int ipq_lock_try __P((void)); 214static __inline void ipq_unlock __P((void)); 215 216static __inline int 217ipq_lock_try() 218{ 219 int s; 220 221 s = splimp(); 222 if (ipq_locked) { 223 splx(s); 224 return (0); 225 } 226 ipq_locked = 1; 227 splx(s); 228 return (1); 229} 230 231static __inline void 232ipq_unlock() 233{ 234 int s; 235 236 s = splimp(); 237 ipq_locked = 0; 238 splx(s); 239} 240 241#ifdef DIAGNOSTIC 242#define IPQ_LOCK() \ 243do { \ 244 if (ipq_lock_try() == 0) { \ 245 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 246 panic("ipq_lock"); \ 247 } \ 248} while (0) 249#define IPQ_LOCK_CHECK() \ 250do { \ 251 if (ipq_locked == 0) { \ 252 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 253 panic("ipq lock check"); \ 254 } \ 255} while (0) 256#else 257#define IPQ_LOCK() (void) ipq_lock_try() 258#define IPQ_LOCK_CHECK() /* nothing */ 259#endif 260 261#define IPQ_UNLOCK() ipq_unlock() 262 263struct pool ipqent_pool; 264 265/* 266 * We need to save the IP options in case a protocol wants to respond 267 * to an incoming packet over the same route if the packet got here 268 * using IP source routing. This allows connection establishment and 269 * maintenance when the remote end is on a network that is not known 270 * to us. 271 */ 272int ip_nhops = 0; 273static struct ip_srcrt { 274 struct in_addr dst; /* final destination */ 275 char nop; /* one NOP to align */ 276 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 277 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 278} ip_srcrt; 279 280static void save_rte __P((u_char *, struct in_addr)); 281 282/* 283 * IP initialization: fill in IP protocol switch table. 284 * All protocols not implemented in kernel go to raw IP protocol handler. 285 */ 286void 287ip_init() 288{ 289 register struct protosw *pr; 290 register int i; 291 292 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 293 0, NULL, NULL, M_IPQ); 294 295 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 296 if (pr == 0) 297 panic("ip_init"); 298 for (i = 0; i < IPPROTO_MAX; i++) 299 ip_protox[i] = pr - inetsw; 300 for (pr = inetdomain.dom_protosw; 301 pr < inetdomain.dom_protoswNPROTOSW; pr++) 302 if (pr->pr_domain->dom_family == PF_INET && 303 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 304 ip_protox[pr->pr_protocol] = pr - inetsw; 305 LIST_INIT(&ipq); 306 ip_id = time.tv_sec & 0xffff; 307 ipintrq.ifq_maxlen = ipqmaxlen; 308 TAILQ_INIT(&in_ifaddr); 309 in_ifaddrhashtbl = 310 hashinit(IN_IFADDR_HASH_SIZE, M_IFADDR, M_WAITOK, &in_ifaddrhash); 311 if (ip_mtudisc != 0) 312 ip_mtudisc_timeout_q = 313 rt_timer_queue_create(ip_mtudisc_timeout); 314#ifdef GATEWAY 315 ipflow_init(); 316#endif 317} 318 319struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 320struct route ipforward_rt; 321 322/* 323 * IP software interrupt routine 324 */ 325void 326ipintr() 327{ 328 int s; 329 struct mbuf *m; 330 331 while (1) { 332 s = splimp(); 333 IF_DEQUEUE(&ipintrq, m); 334 splx(s); 335 if (m == 0) 336 return; 337 ip_input(m); 338 } 339} 340 341/* 342 * Ip input routine. Checksum and byte swap header. If fragmented 343 * try to reassemble. Process options. Pass to next level. 344 */ 345void 346ip_input(struct mbuf *m) 347{ 348 register struct ip *ip = NULL; 349 register struct ipq *fp; 350 register struct in_ifaddr *ia; 351 register struct ifaddr *ifa; 352 struct ipqent *ipqe; 353 int hlen = 0, mff, len; 354#ifdef PFIL_HOOKS 355 struct packet_filter_hook *pfh; 356 struct mbuf *m0; 357 int rv; 358#endif /* PFIL_HOOKS */ 359 360#ifdef DIAGNOSTIC 361 if ((m->m_flags & M_PKTHDR) == 0) 362 panic("ipintr no HDR"); 363#endif 364#ifdef IPSEC 365 /* 366 * should the inner packet be considered authentic? 367 * see comment in ah4_input(). 368 */ 369 if (m) { 370 m->m_flags &= ~M_AUTHIPHDR; 371 m->m_flags &= ~M_AUTHIPDGM; 372 } 373#endif 374 /* 375 * If no IP addresses have been set yet but the interfaces 376 * are receiving, can't do anything with incoming packets yet. 377 */ 378 if (in_ifaddr.tqh_first == 0) 379 goto bad; 380 ipstat.ips_total++; 381 if (m->m_len < sizeof (struct ip) && 382 (m = m_pullup(m, sizeof (struct ip))) == 0) { 383 ipstat.ips_toosmall++; 384 return; 385 } 386 ip = mtod(m, struct ip *); 387 if (ip->ip_v != IPVERSION) { 388 ipstat.ips_badvers++; 389 goto bad; 390 } 391 hlen = ip->ip_hl << 2; 392 if (hlen < sizeof(struct ip)) { /* minimum header length */ 393 ipstat.ips_badhlen++; 394 goto bad; 395 } 396 if (hlen > m->m_len) { 397 if ((m = m_pullup(m, hlen)) == 0) { 398 ipstat.ips_badhlen++; 399 return; 400 } 401 ip = mtod(m, struct ip *); 402 } 403 /* 404 * we drop packets that have a multicast address as source 405 * as wanted by rfc 1112 406 */ 407 if (IN_MULTICAST(ip->ip_src.s_addr)) { 408 goto bad; 409 } 410 411 if (in_cksum(m, hlen) != 0) { 412 ipstat.ips_badsum++; 413 goto bad; 414 } 415 416 /* 417 * Convert fields to host representation. 418 */ 419 NTOHS(ip->ip_len); 420 NTOHS(ip->ip_off); 421 len = ip->ip_len; 422 423 /* 424 * Check for additional length bogosity 425 */ 426 if (len < hlen) { 427 ipstat.ips_badlen++; 428 goto bad; 429 } 430 431 /* 432 * Check that the amount of data in the buffers 433 * is as at least much as the IP header would have us expect. 434 * Trim mbufs if longer than we expect. 435 * Drop packet if shorter than we expect. 436 */ 437 if (m->m_pkthdr.len < len) { 438 ipstat.ips_tooshort++; 439 goto bad; 440 } 441 if (m->m_pkthdr.len > len) { 442 if (m->m_len == m->m_pkthdr.len) { 443 m->m_len = len; 444 m->m_pkthdr.len = len; 445 } else 446 m_adj(m, len - m->m_pkthdr.len); 447 } 448 449 /* 450 * Assume that we can create a fast-forward IP flow entry 451 * based on this packet. 452 */ 453 m->m_flags |= M_CANFASTFWD; 454 455#ifdef PFIL_HOOKS 456 /* 457 * Run through list of hooks for input packets. If there are any 458 * filters which require that additional packets in the flow are 459 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 460 * Note that filters must _never_ set this flag, as another filter 461 * in the list may have previously cleared it. 462 */ 463 m0 = m; 464 for (pfh = pfil_hook_get(PFIL_IN); pfh; pfh = pfh->pfil_link.tqe_next) 465 if (pfh->pfil_func) { 466 rv = pfh->pfil_func(ip, hlen, m->m_pkthdr.rcvif, 0, &m0); 467 if (rv) 468 return; 469 m = m0; 470 if (m == NULL) 471 return; 472 ip = mtod(m, struct ip *); 473 } 474#endif /* PFIL_HOOKS */ 475 476 /* 477 * Process options and, if not destined for us, 478 * ship it on. ip_dooptions returns 1 when an 479 * error was detected (causing an icmp message 480 * to be sent and the original packet to be freed). 481 */ 482 ip_nhops = 0; /* for source routed packets */ 483 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 484 return; 485 486 /* 487 * Check our list of addresses, to see if the packet is for us. 488 */ 489 INADDR_TO_IA(ip->ip_dst, ia); 490 if (ia != NULL) 491 goto ours; 492 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 493 for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first; 494 ifa != NULL; ifa = ifa->ifa_list.tqe_next) { 495 if (ifa->ifa_addr->sa_family != AF_INET) continue; 496 ia = ifatoia(ifa); 497 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 498 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 499 /* 500 * Look for all-0's host part (old broadcast addr), 501 * either for subnet or net. 502 */ 503 ip->ip_dst.s_addr == ia->ia_subnet || 504 ip->ip_dst.s_addr == ia->ia_net) 505 goto ours; 506 /* 507 * An interface with IP address zero accepts 508 * all packets that arrive on that interface. 509 */ 510 if (in_nullhost(ia->ia_addr.sin_addr)) 511 goto ours; 512 } 513 } 514 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 515 struct in_multi *inm; 516#ifdef MROUTING 517 extern struct socket *ip_mrouter; 518 519 if (m->m_flags & M_EXT) { 520 if ((m = m_pullup(m, hlen)) == 0) { 521 ipstat.ips_toosmall++; 522 return; 523 } 524 ip = mtod(m, struct ip *); 525 } 526 527 if (ip_mrouter) { 528 /* 529 * If we are acting as a multicast router, all 530 * incoming multicast packets are passed to the 531 * kernel-level multicast forwarding function. 532 * The packet is returned (relatively) intact; if 533 * ip_mforward() returns a non-zero value, the packet 534 * must be discarded, else it may be accepted below. 535 * 536 * (The IP ident field is put in the same byte order 537 * as expected when ip_mforward() is called from 538 * ip_output().) 539 */ 540 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 541 ipstat.ips_cantforward++; 542 m_freem(m); 543 return; 544 } 545 546 /* 547 * The process-level routing demon needs to receive 548 * all multicast IGMP packets, whether or not this 549 * host belongs to their destination groups. 550 */ 551 if (ip->ip_p == IPPROTO_IGMP) 552 goto ours; 553 ipstat.ips_forward++; 554 } 555#endif 556 /* 557 * See if we belong to the destination multicast group on the 558 * arrival interface. 559 */ 560 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 561 if (inm == NULL) { 562 ipstat.ips_cantforward++; 563 m_freem(m); 564 return; 565 } 566 goto ours; 567 } 568 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 569 in_nullhost(ip->ip_dst)) 570 goto ours; 571 572 /* 573 * Not for us; forward if possible and desirable. 574 */ 575 if (ipforwarding == 0) { 576 ipstat.ips_cantforward++; 577 m_freem(m); 578 } else 579 ip_forward(m, 0); 580 return; 581 582ours: 583 /* 584 * If offset or IP_MF are set, must reassemble. 585 * Otherwise, nothing need be done. 586 * (We could look in the reassembly queue to see 587 * if the packet was previously fragmented, 588 * but it's not worth the time; just let them time out.) 589 */ 590 if (ip->ip_off & ~(IP_DF|IP_RF)) { 591 /* 592 * Look for queue of fragments 593 * of this datagram. 594 */ 595 IPQ_LOCK(); 596 for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next) 597 if (ip->ip_id == fp->ipq_id && 598 in_hosteq(ip->ip_src, fp->ipq_src) && 599 in_hosteq(ip->ip_dst, fp->ipq_dst) && 600 ip->ip_p == fp->ipq_p) 601 goto found; 602 fp = 0; 603found: 604 605 /* 606 * Adjust ip_len to not reflect header, 607 * set ipqe_mff if more fragments are expected, 608 * convert offset of this to bytes. 609 */ 610 ip->ip_len -= hlen; 611 mff = (ip->ip_off & IP_MF) != 0; 612 if (mff) { 613 /* 614 * Make sure that fragments have a data length 615 * that's a non-zero multiple of 8 bytes. 616 */ 617 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 618 ipstat.ips_badfrags++; 619 IPQ_UNLOCK(); 620 goto bad; 621 } 622 } 623 ip->ip_off <<= 3; 624 625 /* 626 * If datagram marked as having more fragments 627 * or if this is not the first fragment, 628 * attempt reassembly; if it succeeds, proceed. 629 */ 630 if (mff || ip->ip_off) { 631 ipstat.ips_fragments++; 632 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 633 if (ipqe == NULL) { 634 ipstat.ips_rcvmemdrop++; 635 IPQ_UNLOCK(); 636 goto bad; 637 } 638 ipqe->ipqe_mff = mff; 639 ipqe->ipqe_m = m; 640 ipqe->ipqe_ip = ip; 641 m = ip_reass(ipqe, fp); 642 if (m == 0) { 643 IPQ_UNLOCK(); 644 return; 645 } 646 ipstat.ips_reassembled++; 647 ip = mtod(m, struct ip *); 648 hlen = ip->ip_hl << 2; 649 ip->ip_len += hlen; 650 } else 651 if (fp) 652 ip_freef(fp); 653 IPQ_UNLOCK(); 654 } 655 656 /* 657 * Switch out to protocol's input routine. 658 */ 659#if IFA_STATS 660 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 661#endif 662 ipstat.ips_delivered++; 663 { 664 int off = hlen, nh = ip->ip_p; 665 666 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 667 return; 668 } 669bad: 670 m_freem(m); 671} 672 673/* 674 * Take incoming datagram fragment and try to 675 * reassemble it into whole datagram. If a chain for 676 * reassembly of this datagram already exists, then it 677 * is given as fp; otherwise have to make a chain. 678 */ 679struct mbuf * 680ip_reass(ipqe, fp) 681 register struct ipqent *ipqe; 682 register struct ipq *fp; 683{ 684 register struct mbuf *m = ipqe->ipqe_m; 685 register struct ipqent *nq, *p, *q; 686 struct ip *ip; 687 struct mbuf *t; 688 int hlen = ipqe->ipqe_ip->ip_hl << 2; 689 int i, next; 690 691 IPQ_LOCK_CHECK(); 692 693 /* 694 * Presence of header sizes in mbufs 695 * would confuse code below. 696 */ 697 m->m_data += hlen; 698 m->m_len -= hlen; 699 700 /* 701 * If first fragment to arrive, create a reassembly queue. 702 */ 703 if (fp == 0) { 704 MALLOC(fp, struct ipq *, sizeof (struct ipq), 705 M_FTABLE, M_NOWAIT); 706 if (fp == NULL) 707 goto dropfrag; 708 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 709 fp->ipq_ttl = IPFRAGTTL; 710 fp->ipq_p = ipqe->ipqe_ip->ip_p; 711 fp->ipq_id = ipqe->ipqe_ip->ip_id; 712 LIST_INIT(&fp->ipq_fragq); 713 fp->ipq_src = ipqe->ipqe_ip->ip_src; 714 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 715 p = NULL; 716 goto insert; 717 } 718 719 /* 720 * Find a segment which begins after this one does. 721 */ 722 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 723 p = q, q = q->ipqe_q.le_next) 724 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 725 break; 726 727 /* 728 * If there is a preceding segment, it may provide some of 729 * our data already. If so, drop the data from the incoming 730 * segment. If it provides all of our data, drop us. 731 */ 732 if (p != NULL) { 733 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 734 ipqe->ipqe_ip->ip_off; 735 if (i > 0) { 736 if (i >= ipqe->ipqe_ip->ip_len) 737 goto dropfrag; 738 m_adj(ipqe->ipqe_m, i); 739 ipqe->ipqe_ip->ip_off += i; 740 ipqe->ipqe_ip->ip_len -= i; 741 } 742 } 743 744 /* 745 * While we overlap succeeding segments trim them or, 746 * if they are completely covered, dequeue them. 747 */ 748 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 749 q->ipqe_ip->ip_off; q = nq) { 750 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 751 q->ipqe_ip->ip_off; 752 if (i < q->ipqe_ip->ip_len) { 753 q->ipqe_ip->ip_len -= i; 754 q->ipqe_ip->ip_off += i; 755 m_adj(q->ipqe_m, i); 756 break; 757 } 758 nq = q->ipqe_q.le_next; 759 m_freem(q->ipqe_m); 760 LIST_REMOVE(q, ipqe_q); 761 pool_put(&ipqent_pool, q); 762 } 763 764insert: 765 /* 766 * Stick new segment in its place; 767 * check for complete reassembly. 768 */ 769 if (p == NULL) { 770 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 771 } else { 772 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 773 } 774 next = 0; 775 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 776 p = q, q = q->ipqe_q.le_next) { 777 if (q->ipqe_ip->ip_off != next) 778 return (0); 779 next += q->ipqe_ip->ip_len; 780 } 781 if (p->ipqe_mff) 782 return (0); 783 784 /* 785 * Reassembly is complete. Check for a bogus message size and 786 * concatenate fragments. 787 */ 788 q = fp->ipq_fragq.lh_first; 789 ip = q->ipqe_ip; 790 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 791 ipstat.ips_toolong++; 792 ip_freef(fp); 793 return (0); 794 } 795 m = q->ipqe_m; 796 t = m->m_next; 797 m->m_next = 0; 798 m_cat(m, t); 799 nq = q->ipqe_q.le_next; 800 pool_put(&ipqent_pool, q); 801 for (q = nq; q != NULL; q = nq) { 802 t = q->ipqe_m; 803 nq = q->ipqe_q.le_next; 804 pool_put(&ipqent_pool, q); 805 m_cat(m, t); 806 } 807 808 /* 809 * Create header for new ip packet by 810 * modifying header of first packet; 811 * dequeue and discard fragment reassembly header. 812 * Make header visible. 813 */ 814 ip->ip_len = next; 815 ip->ip_ttl = 0; /* xxx */ 816 ip->ip_sum = 0; 817 ip->ip_src = fp->ipq_src; 818 ip->ip_dst = fp->ipq_dst; 819 LIST_REMOVE(fp, ipq_q); 820 FREE(fp, M_FTABLE); 821 m->m_len += (ip->ip_hl << 2); 822 m->m_data -= (ip->ip_hl << 2); 823 /* some debugging cruft by sklower, below, will go away soon */ 824 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 825 register int plen = 0; 826 for (t = m; t; t = t->m_next) 827 plen += t->m_len; 828 m->m_pkthdr.len = plen; 829 } 830 return (m); 831 832dropfrag: 833 ipstat.ips_fragdropped++; 834 m_freem(m); 835 pool_put(&ipqent_pool, ipqe); 836 return (0); 837} 838 839/* 840 * Free a fragment reassembly header and all 841 * associated datagrams. 842 */ 843void 844ip_freef(fp) 845 struct ipq *fp; 846{ 847 register struct ipqent *q, *p; 848 849 IPQ_LOCK_CHECK(); 850 851 for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { 852 p = q->ipqe_q.le_next; 853 m_freem(q->ipqe_m); 854 LIST_REMOVE(q, ipqe_q); 855 pool_put(&ipqent_pool, q); 856 } 857 LIST_REMOVE(fp, ipq_q); 858 FREE(fp, M_FTABLE); 859} 860 861/* 862 * IP timer processing; 863 * if a timer expires on a reassembly 864 * queue, discard it. 865 */ 866void 867ip_slowtimo() 868{ 869 register struct ipq *fp, *nfp; 870 int s = splsoftnet(); 871 872 IPQ_LOCK(); 873 for (fp = ipq.lh_first; fp != NULL; fp = nfp) { 874 nfp = fp->ipq_q.le_next; 875 if (--fp->ipq_ttl == 0) { 876 ipstat.ips_fragtimeout++; 877 ip_freef(fp); 878 } 879 } 880 IPQ_UNLOCK(); 881#ifdef GATEWAY 882 ipflow_slowtimo(); 883#endif 884 splx(s); 885} 886 887/* 888 * Drain off all datagram fragments. 889 */ 890void 891ip_drain() 892{ 893 894 /* 895 * We may be called from a device's interrupt context. If 896 * the ipq is already busy, just bail out now. 897 */ 898 if (ipq_lock_try() == 0) 899 return; 900 901 while (ipq.lh_first != NULL) { 902 ipstat.ips_fragdropped++; 903 ip_freef(ipq.lh_first); 904 } 905 906 IPQ_UNLOCK(); 907} 908 909/* 910 * Do option processing on a datagram, 911 * possibly discarding it if bad options are encountered, 912 * or forwarding it if source-routed. 913 * Returns 1 if packet has been forwarded/freed, 914 * 0 if the packet should be processed further. 915 */ 916int 917ip_dooptions(m) 918 struct mbuf *m; 919{ 920 register struct ip *ip = mtod(m, struct ip *); 921 register u_char *cp; 922 register struct ip_timestamp *ipt; 923 register struct in_ifaddr *ia; 924 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 925 struct in_addr *sin, dst; 926 n_time ntime; 927 928 dst = ip->ip_dst; 929 cp = (u_char *)(ip + 1); 930 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 931 for (; cnt > 0; cnt -= optlen, cp += optlen) { 932 opt = cp[IPOPT_OPTVAL]; 933 if (opt == IPOPT_EOL) 934 break; 935 if (opt == IPOPT_NOP) 936 optlen = 1; 937 else { 938 optlen = cp[IPOPT_OLEN]; 939 if (optlen <= 0 || optlen > cnt) { 940 code = &cp[IPOPT_OLEN] - (u_char *)ip; 941 goto bad; 942 } 943 } 944 switch (opt) { 945 946 default: 947 break; 948 949 /* 950 * Source routing with record. 951 * Find interface with current destination address. 952 * If none on this machine then drop if strictly routed, 953 * or do nothing if loosely routed. 954 * Record interface address and bring up next address 955 * component. If strictly routed make sure next 956 * address is on directly accessible net. 957 */ 958 case IPOPT_LSRR: 959 case IPOPT_SSRR: 960 if (ip_allowsrcrt == 0) { 961 type = ICMP_UNREACH; 962 code = ICMP_UNREACH_NET_PROHIB; 963 goto bad; 964 } 965 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 966 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 967 goto bad; 968 } 969 ipaddr.sin_addr = ip->ip_dst; 970 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 971 if (ia == 0) { 972 if (opt == IPOPT_SSRR) { 973 type = ICMP_UNREACH; 974 code = ICMP_UNREACH_SRCFAIL; 975 goto bad; 976 } 977 /* 978 * Loose routing, and not at next destination 979 * yet; nothing to do except forward. 980 */ 981 break; 982 } 983 off--; /* 0 origin */ 984 if (off > optlen - sizeof(struct in_addr)) { 985 /* 986 * End of source route. Should be for us. 987 */ 988 save_rte(cp, ip->ip_src); 989 break; 990 } 991 /* 992 * locate outgoing interface 993 */ 994 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 995 sizeof(ipaddr.sin_addr)); 996 if (opt == IPOPT_SSRR) { 997#define INA struct in_ifaddr * 998#define SA struct sockaddr * 999 ia = (INA)ifa_ifwithladdr((SA)&ipaddr); 1000 } else 1001 ia = ip_rtaddr(ipaddr.sin_addr); 1002 if (ia == 0) { 1003 type = ICMP_UNREACH; 1004 code = ICMP_UNREACH_SRCFAIL; 1005 goto bad; 1006 } 1007 ip->ip_dst = ipaddr.sin_addr; 1008 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1009 (caddr_t)(cp + off), sizeof(struct in_addr)); 1010 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1011 /* 1012 * Let ip_intr's mcast routing check handle mcast pkts 1013 */ 1014 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1015 break; 1016 1017 case IPOPT_RR: 1018 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1019 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1020 goto bad; 1021 } 1022 /* 1023 * If no space remains, ignore. 1024 */ 1025 off--; /* 0 origin */ 1026 if (off > optlen - sizeof(struct in_addr)) 1027 break; 1028 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1029 sizeof(ipaddr.sin_addr)); 1030 /* 1031 * locate outgoing interface; if we're the destination, 1032 * use the incoming interface (should be same). 1033 */ 1034 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1035 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1036 type = ICMP_UNREACH; 1037 code = ICMP_UNREACH_HOST; 1038 goto bad; 1039 } 1040 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1041 (caddr_t)(cp + off), sizeof(struct in_addr)); 1042 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1043 break; 1044 1045 case IPOPT_TS: 1046 code = cp - (u_char *)ip; 1047 ipt = (struct ip_timestamp *)cp; 1048 if (ipt->ipt_len < 5) 1049 goto bad; 1050 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1051 if (++ipt->ipt_oflw == 0) 1052 goto bad; 1053 break; 1054 } 1055 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1056 switch (ipt->ipt_flg) { 1057 1058 case IPOPT_TS_TSONLY: 1059 break; 1060 1061 case IPOPT_TS_TSANDADDR: 1062 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1063 sizeof(struct in_addr) > ipt->ipt_len) 1064 goto bad; 1065 ipaddr.sin_addr = dst; 1066 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1067 m->m_pkthdr.rcvif); 1068 if (ia == 0) 1069 continue; 1070 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1071 (caddr_t)sin, sizeof(struct in_addr)); 1072 ipt->ipt_ptr += sizeof(struct in_addr); 1073 break; 1074 1075 case IPOPT_TS_PRESPEC: 1076 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1077 sizeof(struct in_addr) > ipt->ipt_len) 1078 goto bad; 1079 bcopy((caddr_t)sin, (caddr_t)&ipaddr.sin_addr, 1080 sizeof(struct in_addr)); 1081 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1082 continue; 1083 ipt->ipt_ptr += sizeof(struct in_addr); 1084 break; 1085 1086 default: 1087 goto bad; 1088 } 1089 ntime = iptime(); 1090 bcopy((caddr_t)&ntime, (caddr_t)cp + ipt->ipt_ptr - 1, 1091 sizeof(n_time)); 1092 ipt->ipt_ptr += sizeof(n_time); 1093 } 1094 } 1095 if (forward) { 1096 if (ip_forwsrcrt == 0) { 1097 type = ICMP_UNREACH; 1098 code = ICMP_UNREACH_SRCFAIL; 1099 goto bad; 1100 } 1101 ip_forward(m, 1); 1102 return (1); 1103 } 1104 return (0); 1105bad: 1106 icmp_error(m, type, code, 0, 0); 1107 ipstat.ips_badoptions++; 1108 return (1); 1109} 1110 1111/* 1112 * Given address of next destination (final or next hop), 1113 * return internet address info of interface to be used to get there. 1114 */ 1115struct in_ifaddr * 1116ip_rtaddr(dst) 1117 struct in_addr dst; 1118{ 1119 register struct sockaddr_in *sin; 1120 1121 sin = satosin(&ipforward_rt.ro_dst); 1122 1123 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1124 if (ipforward_rt.ro_rt) { 1125 RTFREE(ipforward_rt.ro_rt); 1126 ipforward_rt.ro_rt = 0; 1127 } 1128 sin->sin_family = AF_INET; 1129 sin->sin_len = sizeof(*sin); 1130 sin->sin_addr = dst; 1131 1132 rtalloc(&ipforward_rt); 1133 } 1134 if (ipforward_rt.ro_rt == 0) 1135 return ((struct in_ifaddr *)0); 1136 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1137} 1138 1139/* 1140 * Save incoming source route for use in replies, 1141 * to be picked up later by ip_srcroute if the receiver is interested. 1142 */ 1143void 1144save_rte(option, dst) 1145 u_char *option; 1146 struct in_addr dst; 1147{ 1148 unsigned olen; 1149 1150 olen = option[IPOPT_OLEN]; 1151#ifdef DIAGNOSTIC 1152 if (ipprintfs) 1153 printf("save_rte: olen %d\n", olen); 1154#endif /* 0 */ 1155 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1156 return; 1157 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1158 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1159 ip_srcrt.dst = dst; 1160} 1161 1162/* 1163 * Retrieve incoming source route for use in replies, 1164 * in the same form used by setsockopt. 1165 * The first hop is placed before the options, will be removed later. 1166 */ 1167struct mbuf * 1168ip_srcroute() 1169{ 1170 register struct in_addr *p, *q; 1171 register struct mbuf *m; 1172 1173 if (ip_nhops == 0) 1174 return ((struct mbuf *)0); 1175 m = m_get(M_DONTWAIT, MT_SOOPTS); 1176 if (m == 0) 1177 return ((struct mbuf *)0); 1178 1179#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1180 1181 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1182 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1183 OPTSIZ; 1184#ifdef DIAGNOSTIC 1185 if (ipprintfs) 1186 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1187#endif 1188 1189 /* 1190 * First save first hop for return route 1191 */ 1192 p = &ip_srcrt.route[ip_nhops - 1]; 1193 *(mtod(m, struct in_addr *)) = *p--; 1194#ifdef DIAGNOSTIC 1195 if (ipprintfs) 1196 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1197#endif 1198 1199 /* 1200 * Copy option fields and padding (nop) to mbuf. 1201 */ 1202 ip_srcrt.nop = IPOPT_NOP; 1203 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1204 bcopy((caddr_t)&ip_srcrt.nop, 1205 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1206 q = (struct in_addr *)(mtod(m, caddr_t) + 1207 sizeof(struct in_addr) + OPTSIZ); 1208#undef OPTSIZ 1209 /* 1210 * Record return path as an IP source route, 1211 * reversing the path (pointers are now aligned). 1212 */ 1213 while (p >= ip_srcrt.route) { 1214#ifdef DIAGNOSTIC 1215 if (ipprintfs) 1216 printf(" %x", ntohl(q->s_addr)); 1217#endif 1218 *q++ = *p--; 1219 } 1220 /* 1221 * Last hop goes to final destination. 1222 */ 1223 *q = ip_srcrt.dst; 1224#ifdef DIAGNOSTIC 1225 if (ipprintfs) 1226 printf(" %x\n", ntohl(q->s_addr)); 1227#endif 1228 return (m); 1229} 1230 1231/* 1232 * Strip out IP options, at higher 1233 * level protocol in the kernel. 1234 * Second argument is buffer to which options 1235 * will be moved, and return value is their length. 1236 * XXX should be deleted; last arg currently ignored. 1237 */ 1238void 1239ip_stripoptions(m, mopt) 1240 register struct mbuf *m; 1241 struct mbuf *mopt; 1242{ 1243 register int i; 1244 struct ip *ip = mtod(m, struct ip *); 1245 register caddr_t opts; 1246 int olen; 1247 1248 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1249 opts = (caddr_t)(ip + 1); 1250 i = m->m_len - (sizeof (struct ip) + olen); 1251 bcopy(opts + olen, opts, (unsigned)i); 1252 m->m_len -= olen; 1253 if (m->m_flags & M_PKTHDR) 1254 m->m_pkthdr.len -= olen; 1255 ip->ip_len -= olen; 1256 ip->ip_hl = sizeof (struct ip) >> 2; 1257} 1258 1259int inetctlerrmap[PRC_NCMDS] = { 1260 0, 0, 0, 0, 1261 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1262 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1263 EMSGSIZE, EHOSTUNREACH, 0, 0, 1264 0, 0, 0, 0, 1265 ENOPROTOOPT 1266}; 1267 1268/* 1269 * Forward a packet. If some error occurs return the sender 1270 * an icmp packet. Note we can't always generate a meaningful 1271 * icmp message because icmp doesn't have a large enough repertoire 1272 * of codes and types. 1273 * 1274 * If not forwarding, just drop the packet. This could be confusing 1275 * if ipforwarding was zero but some routing protocol was advancing 1276 * us as a gateway to somewhere. However, we must let the routing 1277 * protocol deal with that. 1278 * 1279 * The srcrt parameter indicates whether the packet is being forwarded 1280 * via a source route. 1281 */ 1282void 1283ip_forward(m, srcrt) 1284 struct mbuf *m; 1285 int srcrt; 1286{ 1287 register struct ip *ip = mtod(m, struct ip *); 1288 register struct sockaddr_in *sin; 1289 register struct rtentry *rt; 1290 int error, type = 0, code = 0; 1291 struct mbuf *mcopy; 1292 n_long dest; 1293 struct ifnet *destifp; 1294#ifdef IPSEC 1295 struct ifnet dummyifp; 1296#endif 1297 1298 dest = 0; 1299#ifdef DIAGNOSTIC 1300 if (ipprintfs) 1301 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1302 ntohl(ip->ip_src.s_addr), 1303 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1304#endif 1305 if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) { 1306 ipstat.ips_cantforward++; 1307 m_freem(m); 1308 return; 1309 } 1310 if (ip->ip_ttl <= IPTTLDEC) { 1311 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1312 return; 1313 } 1314 ip->ip_ttl -= IPTTLDEC; 1315 1316 sin = satosin(&ipforward_rt.ro_dst); 1317 if ((rt = ipforward_rt.ro_rt) == 0 || 1318 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1319 if (ipforward_rt.ro_rt) { 1320 RTFREE(ipforward_rt.ro_rt); 1321 ipforward_rt.ro_rt = 0; 1322 } 1323 sin->sin_family = AF_INET; 1324 sin->sin_len = sizeof(struct sockaddr_in); 1325 sin->sin_addr = ip->ip_dst; 1326 1327 rtalloc(&ipforward_rt); 1328 if (ipforward_rt.ro_rt == 0) { 1329 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1330 return; 1331 } 1332 rt = ipforward_rt.ro_rt; 1333 } 1334 1335 /* 1336 * Save at most 68 bytes of the packet in case 1337 * we need to generate an ICMP message to the src. 1338 */ 1339 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 68)); 1340 1341 /* 1342 * If forwarding packet using same interface that it came in on, 1343 * perhaps should send a redirect to sender to shortcut a hop. 1344 * Only send redirect if source is sending directly to us, 1345 * and if packet was not source routed (or has any options). 1346 * Also, don't send redirect if forwarding using a default route 1347 * or a route modified by a redirect. 1348 */ 1349 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1350 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1351 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1352 ipsendredirects && !srcrt) { 1353 if (rt->rt_ifa && 1354 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1355 ifatoia(rt->rt_ifa)->ia_subnet) { 1356 if (rt->rt_flags & RTF_GATEWAY) 1357 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1358 else 1359 dest = ip->ip_dst.s_addr; 1360 /* 1361 * Router requirements says to only send host 1362 * redirects. 1363 */ 1364 type = ICMP_REDIRECT; 1365 code = ICMP_REDIRECT_HOST; 1366#ifdef DIAGNOSTIC 1367 if (ipprintfs) 1368 printf("redirect (%d) to %x\n", code, 1369 (u_int32_t)dest); 1370#endif 1371 } 1372 } 1373 1374#ifdef IPSEC 1375 m->m_pkthdr.rcvif = NULL; 1376#endif /*IPSEC*/ 1377 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1378 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1379 if (error) 1380 ipstat.ips_cantforward++; 1381 else { 1382 ipstat.ips_forward++; 1383 if (type) 1384 ipstat.ips_redirectsent++; 1385 else { 1386 if (mcopy) { 1387#ifdef GATEWAY 1388 if (mcopy->m_flags & M_CANFASTFWD) 1389 ipflow_create(&ipforward_rt, mcopy); 1390#endif 1391 m_freem(mcopy); 1392 } 1393 return; 1394 } 1395 } 1396 if (mcopy == NULL) 1397 return; 1398 destifp = NULL; 1399 1400 switch (error) { 1401 1402 case 0: /* forwarded, but need redirect */ 1403 /* type, code set above */ 1404 break; 1405 1406 case ENETUNREACH: /* shouldn't happen, checked above */ 1407 case EHOSTUNREACH: 1408 case ENETDOWN: 1409 case EHOSTDOWN: 1410 default: 1411 type = ICMP_UNREACH; 1412 code = ICMP_UNREACH_HOST; 1413 break; 1414 1415 case EMSGSIZE: 1416 type = ICMP_UNREACH; 1417 code = ICMP_UNREACH_NEEDFRAG; 1418#ifndef IPSEC 1419 if (ipforward_rt.ro_rt) 1420 destifp = ipforward_rt.ro_rt->rt_ifp; 1421#else 1422 /* 1423 * If the packet is routed over IPsec tunnel, tell the 1424 * originator the tunnel MTU. 1425 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1426 * XXX quickhack!!! 1427 */ 1428 if (ipforward_rt.ro_rt) { 1429 struct secpolicy *sp; 1430 int ipsecerror; 1431 int ipsechdr; 1432 struct route *ro; 1433 1434 sp = ipsec4_getpolicybyaddr(mcopy, 1435 IP_FORWARDING, 1436 &ipsecerror); 1437 1438 if (sp == NULL) 1439 destifp = ipforward_rt.ro_rt->rt_ifp; 1440 else { 1441 /* count IPsec header size */ 1442 ipsechdr = ipsec4_hdrsiz(mcopy, NULL); 1443 1444 /* 1445 * find the correct route for outer IPv4 1446 * header, compute tunnel MTU. 1447 * 1448 * XXX BUG ALERT 1449 * The "dummyifp" code relies upon the fact 1450 * that icmp_error() touches only ifp->if_mtu. 1451 */ 1452 /*XXX*/ 1453 destifp = NULL; 1454 if (sp->req != NULL 1455 && sp->req->sa != NULL) { 1456 ro = &sp->req->sa->saidx->sa_route; 1457 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1458 dummyifp.if_mtu = 1459 ro->ro_rt->rt_ifp->if_mtu; 1460 dummyifp.if_mtu -= ipsechdr; 1461 destifp = &dummyifp; 1462 } 1463 } 1464 1465 key_freesp(sp); 1466 } 1467 } 1468#endif /*IPSEC*/ 1469 ipstat.ips_cantfrag++; 1470 break; 1471 1472 case ENOBUFS: 1473 type = ICMP_SOURCEQUENCH; 1474 code = 0; 1475 break; 1476 } 1477 icmp_error(mcopy, type, code, dest, destifp); 1478} 1479 1480void 1481ip_savecontrol(inp, mp, ip, m) 1482 register struct inpcb *inp; 1483 register struct mbuf **mp; 1484 register struct ip *ip; 1485 register struct mbuf *m; 1486{ 1487 1488 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1489 struct timeval tv; 1490 1491 microtime(&tv); 1492 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1493 SCM_TIMESTAMP, SOL_SOCKET); 1494 if (*mp) 1495 mp = &(*mp)->m_next; 1496 } 1497 if (inp->inp_flags & INP_RECVDSTADDR) { 1498 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1499 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1500 if (*mp) 1501 mp = &(*mp)->m_next; 1502 } 1503#ifdef notyet 1504 /* 1505 * XXX 1506 * Moving these out of udp_input() made them even more broken 1507 * than they already were. 1508 * - fenner@parc.xerox.com 1509 */ 1510 /* options were tossed already */ 1511 if (inp->inp_flags & INP_RECVOPTS) { 1512 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1513 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1514 if (*mp) 1515 mp = &(*mp)->m_next; 1516 } 1517 /* ip_srcroute doesn't do what we want here, need to fix */ 1518 if (inp->inp_flags & INP_RECVRETOPTS) { 1519 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1520 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1521 if (*mp) 1522 mp = &(*mp)->m_next; 1523 } 1524#endif 1525 if (inp->inp_flags & INP_RECVIF) { 1526 struct sockaddr_dl sdl; 1527 1528 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1529 sdl.sdl_family = AF_LINK; 1530 sdl.sdl_index = m->m_pkthdr.rcvif ? 1531 m->m_pkthdr.rcvif->if_index : 0; 1532 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1533 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1534 IP_RECVIF, IPPROTO_IP); 1535 if (*mp) 1536 mp = &(*mp)->m_next; 1537 } 1538} 1539 1540int 1541ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1542 int *name; 1543 u_int namelen; 1544 void *oldp; 1545 size_t *oldlenp; 1546 void *newp; 1547 size_t newlen; 1548{ 1549 extern int subnetsarelocal, hostzeroisbroadcast; 1550 1551 int error, old; 1552 1553 /* All sysctl names at this level are terminal. */ 1554 if (namelen != 1) 1555 return (ENOTDIR); 1556 1557 switch (name[0]) { 1558 case IPCTL_FORWARDING: 1559 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1560 case IPCTL_SENDREDIRECTS: 1561 return (sysctl_int(oldp, oldlenp, newp, newlen, 1562 &ipsendredirects)); 1563 case IPCTL_DEFTTL: 1564 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1565#ifdef notyet 1566 case IPCTL_DEFMTU: 1567 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1568#endif 1569 case IPCTL_FORWSRCRT: 1570 /* Don't allow this to change in a secure environment. */ 1571 if (securelevel > 0) 1572 return (sysctl_rdint(oldp, oldlenp, newp, 1573 ip_forwsrcrt)); 1574 else 1575 return (sysctl_int(oldp, oldlenp, newp, newlen, 1576 &ip_forwsrcrt)); 1577 case IPCTL_DIRECTEDBCAST: 1578 return (sysctl_int(oldp, oldlenp, newp, newlen, 1579 &ip_directedbcast)); 1580 case IPCTL_ALLOWSRCRT: 1581 return (sysctl_int(oldp, oldlenp, newp, newlen, 1582 &ip_allowsrcrt)); 1583 case IPCTL_SUBNETSARELOCAL: 1584 return (sysctl_int(oldp, oldlenp, newp, newlen, 1585 &subnetsarelocal)); 1586 case IPCTL_MTUDISC: 1587 error = sysctl_int(oldp, oldlenp, newp, newlen, 1588 &ip_mtudisc); 1589 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1590 ip_mtudisc_timeout_q = 1591 rt_timer_queue_create(ip_mtudisc_timeout); 1592 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1593 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1594 ip_mtudisc_timeout_q = NULL; 1595 } 1596 return error; 1597 case IPCTL_ANONPORTMIN: 1598 old = anonportmin; 1599 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1600 if (anonportmin >= anonportmax || anonportmin > 65535 1601#ifndef IPNOPRIVPORTS 1602 || anonportmin < IPPORT_RESERVED 1603#endif 1604 ) { 1605 anonportmin = old; 1606 return (EINVAL); 1607 } 1608 return (error); 1609 case IPCTL_ANONPORTMAX: 1610 old = anonportmax; 1611 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1612 if (anonportmin >= anonportmax || anonportmax > 65535 1613#ifndef IPNOPRIVPORTS 1614 || anonportmax < IPPORT_RESERVED 1615#endif 1616 ) { 1617 anonportmax = old; 1618 return (EINVAL); 1619 } 1620 return (error); 1621 case IPCTL_MTUDISCTIMEOUT: 1622 error = sysctl_int(oldp, oldlenp, newp, newlen, 1623 &ip_mtudisc_timeout); 1624 if (ip_mtudisc_timeout_q != NULL) 1625 rt_timer_queue_change(ip_mtudisc_timeout_q, 1626 ip_mtudisc_timeout); 1627 return (error); 1628#ifdef GATEWAY 1629 case IPCTL_MAXFLOWS: 1630 { 1631 int s; 1632 1633 error = sysctl_int(oldp, oldlenp, newp, newlen, 1634 &ip_maxflows); 1635 s = splsoftnet(); 1636 ipflow_reap(0); 1637 splx(s); 1638 return (error); 1639 } 1640#endif 1641 case IPCTL_HOSTZEROBROADCAST: 1642 return (sysctl_int(oldp, oldlenp, newp, newlen, 1643 &hostzeroisbroadcast)); 1644#if NGIF > 0 1645 case IPCTL_GIF_TTL: 1646 return(sysctl_int(oldp, oldlenp, newp, newlen, 1647 &ip_gif_ttl)); 1648#endif 1649 1650 default: 1651 return (EOPNOTSUPP); 1652 } 1653 /* NOTREACHED */ 1654} 1655