1 /* $NetBSD: ip_input.c,v 1.148 2002/05/07 02:59:38 matt 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 <sys/cdefs.h> 105 __KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.148 2002/05/07 02:59:38 matt Exp $"); 106 107 #include "opt_gateway.h" 108 #include "opt_pfil_hooks.h" 109 #include "opt_ipsec.h" 110 #include "opt_mrouting.h" 111 #include "opt_inet_csum.h" 112 113 #include <sys/param.h> 114 #include <sys/systm.h> 115 #include <sys/malloc.h> 116 #include <sys/mbuf.h> 117 #include <sys/domain.h> 118 #include <sys/protosw.h> 119 #include <sys/socket.h> 120 #include <sys/socketvar.h> 121 #include <sys/errno.h> 122 #include <sys/time.h> 123 #include <sys/kernel.h> 124 #include <sys/pool.h> 125 #include <sys/sysctl.h> 126 127 #include <net/if.h> 128 #include <net/if_dl.h> 129 #include <net/route.h> 130 #include <net/pfil.h> 131 132 #include <netinet/in.h> 133 #include <netinet/in_systm.h> 134 #include <netinet/ip.h> 135 #include <netinet/in_pcb.h> 136 #include <netinet/in_var.h> 137 #include <netinet/ip_var.h> 138 #include <netinet/ip_icmp.h> 139 /* just for gif_ttl */ 140 #include <netinet/in_gif.h> 141 #include "gif.h" 142 #include <net/if_gre.h> 143 #include "gre.h" 144 145 #ifdef MROUTING 146 #include <netinet/ip_mroute.h> 147 #endif 148 149 #ifdef IPSEC 150 #include <netinet6/ipsec.h> 151 #include <netkey/key.h> 152 #endif 153 154 #ifndef IPFORWARDING 155 #ifdef GATEWAY 156 #define IPFORWARDING 1 /* forward IP packets not for us */ 157 #else /* GATEWAY */ 158 #define IPFORWARDING 0 /* don't forward IP packets not for us */ 159 #endif /* GATEWAY */ 160 #endif /* IPFORWARDING */ 161 #ifndef IPSENDREDIRECTS 162 #define IPSENDREDIRECTS 1 163 #endif 164 #ifndef IPFORWSRCRT 165 #define IPFORWSRCRT 1 /* forward source-routed packets */ 166 #endif 167 #ifndef IPALLOWSRCRT 168 #define IPALLOWSRCRT 1 /* allow source-routed packets */ 169 #endif 170 #ifndef IPMTUDISC 171 #define IPMTUDISC 0 172 #endif 173 #ifndef IPMTUDISCTIMEOUT 174 #define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 175 #endif 176 177 /* 178 * Note: DIRECTED_BROADCAST is handled this way so that previous 179 * configuration using this option will Just Work. 180 */ 181 #ifndef IPDIRECTEDBCAST 182 #ifdef DIRECTED_BROADCAST 183 #define IPDIRECTEDBCAST 1 184 #else 185 #define IPDIRECTEDBCAST 0 186 #endif /* DIRECTED_BROADCAST */ 187 #endif /* IPDIRECTEDBCAST */ 188 int ipforwarding = IPFORWARDING; 189 int ipsendredirects = IPSENDREDIRECTS; 190 int ip_defttl = IPDEFTTL; 191 int ip_forwsrcrt = IPFORWSRCRT; 192 int ip_directedbcast = IPDIRECTEDBCAST; 193 int ip_allowsrcrt = IPALLOWSRCRT; 194 int ip_mtudisc = IPMTUDISC; 195 u_int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 196 #ifdef DIAGNOSTIC 197 int ipprintfs = 0; 198 #endif 199 200 struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 201 202 extern struct domain inetdomain; 203 int ipqmaxlen = IFQ_MAXLEN; 204 struct in_ifaddrhead in_ifaddr; 205 struct in_ifaddrhashhead *in_ifaddrhashtbl; 206 struct ifqueue ipintrq; 207 struct ipstat ipstat; 208 u_int16_t ip_id; 209 210 #ifdef PFIL_HOOKS 211 struct pfil_head inet_pfil_hook; 212 #endif 213 214 struct ipqhead ipq; 215 int ipq_locked; 216 int ip_nfragpackets = 0; 217 int ip_maxfragpackets = 200; 218 219 static __inline int ipq_lock_try __P((void)); 220 static __inline void ipq_unlock __P((void)); 221 222 static __inline int 223 ipq_lock_try() 224 { 225 int s; 226 227 /* 228 * Use splvm() -- we're bloking things that would cause 229 * mbuf allocation. 230 */ 231 s = splvm(); 232 if (ipq_locked) { 233 splx(s); 234 return (0); 235 } 236 ipq_locked = 1; 237 splx(s); 238 return (1); 239 } 240 241 static __inline void 242 ipq_unlock() 243 { 244 int s; 245 246 s = splvm(); 247 ipq_locked = 0; 248 splx(s); 249 } 250 251 #ifdef DIAGNOSTIC 252 #define IPQ_LOCK() \ 253 do { \ 254 if (ipq_lock_try() == 0) { \ 255 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 256 panic("ipq_lock"); \ 257 } \ 258 } while (0) 259 #define IPQ_LOCK_CHECK() \ 260 do { \ 261 if (ipq_locked == 0) { \ 262 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 263 panic("ipq lock check"); \ 264 } \ 265 } while (0) 266 #else 267 #define IPQ_LOCK() (void) ipq_lock_try() 268 #define IPQ_LOCK_CHECK() /* nothing */ 269 #endif 270 271 #define IPQ_UNLOCK() ipq_unlock() 272 273 struct pool ipqent_pool; 274 275 #ifdef INET_CSUM_COUNTERS 276 #include <sys/device.h> 277 278 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 279 NULL, "inet", "hwcsum bad"); 280 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 281 NULL, "inet", "hwcsum ok"); 282 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 283 NULL, "inet", "swcsum"); 284 285 #define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 286 287 #else 288 289 #define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 290 291 #endif /* INET_CSUM_COUNTERS */ 292 293 /* 294 * We need to save the IP options in case a protocol wants to respond 295 * to an incoming packet over the same route if the packet got here 296 * using IP source routing. This allows connection establishment and 297 * maintenance when the remote end is on a network that is not known 298 * to us. 299 */ 300 int ip_nhops = 0; 301 static struct ip_srcrt { 302 struct in_addr dst; /* final destination */ 303 char nop; /* one NOP to align */ 304 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 305 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 306 } ip_srcrt; 307 308 static void save_rte __P((u_char *, struct in_addr)); 309 310 /* 311 * IP initialization: fill in IP protocol switch table. 312 * All protocols not implemented in kernel go to raw IP protocol handler. 313 */ 314 void 315 ip_init() 316 { 317 struct protosw *pr; 318 int i; 319 320 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 321 NULL); 322 323 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 324 if (pr == 0) 325 panic("ip_init"); 326 for (i = 0; i < IPPROTO_MAX; i++) 327 ip_protox[i] = pr - inetsw; 328 for (pr = inetdomain.dom_protosw; 329 pr < inetdomain.dom_protoswNPROTOSW; pr++) 330 if (pr->pr_domain->dom_family == PF_INET && 331 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 332 ip_protox[pr->pr_protocol] = pr - inetsw; 333 LIST_INIT(&ipq); 334 ip_id = time.tv_sec & 0xffff; 335 ipintrq.ifq_maxlen = ipqmaxlen; 336 TAILQ_INIT(&in_ifaddr); 337 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 338 M_WAITOK, &in_ifaddrhash); 339 if (ip_mtudisc != 0) 340 ip_mtudisc_timeout_q = 341 rt_timer_queue_create(ip_mtudisc_timeout); 342 #ifdef GATEWAY 343 ipflow_init(); 344 #endif 345 346 #ifdef PFIL_HOOKS 347 /* Register our Packet Filter hook. */ 348 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 349 inet_pfil_hook.ph_af = AF_INET; 350 i = pfil_head_register(&inet_pfil_hook); 351 if (i != 0) 352 printf("ip_init: WARNING: unable to register pfil hook, " 353 "error %d\n", i); 354 #endif /* PFIL_HOOKS */ 355 356 #ifdef INET_CSUM_COUNTERS 357 evcnt_attach_static(&ip_hwcsum_bad); 358 evcnt_attach_static(&ip_hwcsum_ok); 359 evcnt_attach_static(&ip_swcsum); 360 #endif /* INET_CSUM_COUNTERS */ 361 } 362 363 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 364 struct route ipforward_rt; 365 366 /* 367 * IP software interrupt routine 368 */ 369 void 370 ipintr() 371 { 372 int s; 373 struct mbuf *m; 374 375 while (1) { 376 s = splnet(); 377 IF_DEQUEUE(&ipintrq, m); 378 splx(s); 379 if (m == 0) 380 return; 381 ip_input(m); 382 } 383 } 384 385 /* 386 * Ip input routine. Checksum and byte swap header. If fragmented 387 * try to reassemble. Process options. Pass to next level. 388 */ 389 void 390 ip_input(struct mbuf *m) 391 { 392 struct ip *ip = NULL; 393 struct ipq *fp; 394 struct in_ifaddr *ia; 395 struct ifaddr *ifa; 396 struct ipqent *ipqe; 397 int hlen = 0, mff, len; 398 int downmatch; 399 400 #ifdef DIAGNOSTIC 401 if ((m->m_flags & M_PKTHDR) == 0) 402 panic("ipintr no HDR"); 403 #endif 404 #ifdef IPSEC 405 /* 406 * should the inner packet be considered authentic? 407 * see comment in ah4_input(). 408 */ 409 if (m) { 410 m->m_flags &= ~M_AUTHIPHDR; 411 m->m_flags &= ~M_AUTHIPDGM; 412 } 413 #endif 414 /* 415 * If no IP addresses have been set yet but the interfaces 416 * are receiving, can't do anything with incoming packets yet. 417 */ 418 if (TAILQ_FIRST(&in_ifaddr) == 0) 419 goto bad; 420 ipstat.ips_total++; 421 if (m->m_len < sizeof (struct ip) && 422 (m = m_pullup(m, sizeof (struct ip))) == 0) { 423 ipstat.ips_toosmall++; 424 return; 425 } 426 ip = mtod(m, struct ip *); 427 if (ip->ip_v != IPVERSION) { 428 ipstat.ips_badvers++; 429 goto bad; 430 } 431 hlen = ip->ip_hl << 2; 432 if (hlen < sizeof(struct ip)) { /* minimum header length */ 433 ipstat.ips_badhlen++; 434 goto bad; 435 } 436 if (hlen > m->m_len) { 437 if ((m = m_pullup(m, hlen)) == 0) { 438 ipstat.ips_badhlen++; 439 return; 440 } 441 ip = mtod(m, struct ip *); 442 } 443 444 /* 445 * RFC1122: packets with a multicast source address are 446 * not allowed. 447 */ 448 if (IN_MULTICAST(ip->ip_src.s_addr)) { 449 ipstat.ips_badaddr++; 450 goto bad; 451 } 452 453 /* 127/8 must not appear on wire - RFC1122 */ 454 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 455 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 456 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 457 ipstat.ips_badaddr++; 458 goto bad; 459 } 460 } 461 462 switch (m->m_pkthdr.csum_flags & 463 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) | 464 M_CSUM_IPv4_BAD)) { 465 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 466 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 467 goto badcsum; 468 469 case M_CSUM_IPv4: 470 /* Checksum was okay. */ 471 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 472 break; 473 474 default: 475 /* Must compute it ourselves. */ 476 INET_CSUM_COUNTER_INCR(&ip_swcsum); 477 if (in_cksum(m, hlen) != 0) 478 goto bad; 479 break; 480 } 481 482 /* Retrieve the packet length. */ 483 len = ntohs(ip->ip_len); 484 485 /* 486 * Check for additional length bogosity 487 */ 488 if (len < hlen) { 489 ipstat.ips_badlen++; 490 goto bad; 491 } 492 493 /* 494 * Check that the amount of data in the buffers 495 * is as at least much as the IP header would have us expect. 496 * Trim mbufs if longer than we expect. 497 * Drop packet if shorter than we expect. 498 */ 499 if (m->m_pkthdr.len < len) { 500 ipstat.ips_tooshort++; 501 goto bad; 502 } 503 if (m->m_pkthdr.len > len) { 504 if (m->m_len == m->m_pkthdr.len) { 505 m->m_len = len; 506 m->m_pkthdr.len = len; 507 } else 508 m_adj(m, len - m->m_pkthdr.len); 509 } 510 511 #ifdef IPSEC 512 /* ipflow (IP fast fowarding) is not compatible with IPsec. */ 513 m->m_flags &= ~M_CANFASTFWD; 514 #else 515 /* 516 * Assume that we can create a fast-forward IP flow entry 517 * based on this packet. 518 */ 519 m->m_flags |= M_CANFASTFWD; 520 #endif 521 522 #ifdef PFIL_HOOKS 523 /* 524 * Run through list of hooks for input packets. If there are any 525 * filters which require that additional packets in the flow are 526 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 527 * Note that filters must _never_ set this flag, as another filter 528 * in the list may have previously cleared it. 529 */ 530 /* 531 * let ipfilter look at packet on the wire, 532 * not the decapsulated packet. 533 */ 534 #ifdef IPSEC 535 if (!ipsec_getnhist(m)) 536 #else 537 if (1) 538 #endif 539 { 540 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 541 PFIL_IN) != 0) 542 return; 543 if (m == NULL) 544 return; 545 ip = mtod(m, struct ip *); 546 hlen = ip->ip_hl << 2; 547 } 548 #endif /* PFIL_HOOKS */ 549 550 #ifdef ALTQ 551 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 552 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 553 /* packet dropped by traffic conditioner */ 554 return; 555 } 556 #endif 557 558 /* 559 * Convert fields to host representation. 560 */ 561 NTOHS(ip->ip_len); 562 NTOHS(ip->ip_off); 563 564 /* 565 * Process options and, if not destined for us, 566 * ship it on. ip_dooptions returns 1 when an 567 * error was detected (causing an icmp message 568 * to be sent and the original packet to be freed). 569 */ 570 ip_nhops = 0; /* for source routed packets */ 571 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 572 return; 573 574 /* 575 * Check our list of addresses, to see if the packet is for us. 576 * 577 * Traditional 4.4BSD did not consult IFF_UP at all. 578 * The behavior here is to treat addresses on !IFF_UP interface 579 * as not mine. 580 */ 581 downmatch = 0; 582 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 583 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 584 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 585 break; 586 else 587 downmatch++; 588 } 589 } 590 if (ia != NULL) 591 goto ours; 592 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 593 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrlist, ifa_list) { 594 if (ifa->ifa_addr->sa_family != AF_INET) 595 continue; 596 ia = ifatoia(ifa); 597 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 598 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 599 /* 600 * Look for all-0's host part (old broadcast addr), 601 * either for subnet or net. 602 */ 603 ip->ip_dst.s_addr == ia->ia_subnet || 604 ip->ip_dst.s_addr == ia->ia_net) 605 goto ours; 606 /* 607 * An interface with IP address zero accepts 608 * all packets that arrive on that interface. 609 */ 610 if (in_nullhost(ia->ia_addr.sin_addr)) 611 goto ours; 612 } 613 } 614 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 615 struct in_multi *inm; 616 #ifdef MROUTING 617 extern struct socket *ip_mrouter; 618 619 if (M_READONLY(m)) { 620 if ((m = m_pullup(m, hlen)) == 0) { 621 ipstat.ips_toosmall++; 622 return; 623 } 624 ip = mtod(m, struct ip *); 625 } 626 627 if (ip_mrouter) { 628 /* 629 * If we are acting as a multicast router, all 630 * incoming multicast packets are passed to the 631 * kernel-level multicast forwarding function. 632 * The packet is returned (relatively) intact; if 633 * ip_mforward() returns a non-zero value, the packet 634 * must be discarded, else it may be accepted below. 635 * 636 * (The IP ident field is put in the same byte order 637 * as expected when ip_mforward() is called from 638 * ip_output().) 639 */ 640 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 641 ipstat.ips_cantforward++; 642 m_freem(m); 643 return; 644 } 645 646 /* 647 * The process-level routing demon needs to receive 648 * all multicast IGMP packets, whether or not this 649 * host belongs to their destination groups. 650 */ 651 if (ip->ip_p == IPPROTO_IGMP) 652 goto ours; 653 ipstat.ips_forward++; 654 } 655 #endif 656 /* 657 * See if we belong to the destination multicast group on the 658 * arrival interface. 659 */ 660 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 661 if (inm == NULL) { 662 ipstat.ips_cantforward++; 663 m_freem(m); 664 return; 665 } 666 goto ours; 667 } 668 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 669 in_nullhost(ip->ip_dst)) 670 goto ours; 671 672 /* 673 * Not for us; forward if possible and desirable. 674 */ 675 if (ipforwarding == 0) { 676 ipstat.ips_cantforward++; 677 m_freem(m); 678 } else { 679 /* 680 * If ip_dst matched any of my address on !IFF_UP interface, 681 * and there's no IFF_UP interface that matches ip_dst, 682 * send icmp unreach. Forwarding it will result in in-kernel 683 * forwarding loop till TTL goes to 0. 684 */ 685 if (downmatch) { 686 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 687 ipstat.ips_cantforward++; 688 return; 689 } 690 #ifdef IPSEC 691 if (ipsec4_in_reject(m, NULL)) { 692 ipsecstat.in_polvio++; 693 goto bad; 694 } 695 #endif 696 697 ip_forward(m, 0); 698 } 699 return; 700 701 ours: 702 /* 703 * If offset or IP_MF are set, must reassemble. 704 * Otherwise, nothing need be done. 705 * (We could look in the reassembly queue to see 706 * if the packet was previously fragmented, 707 * but it's not worth the time; just let them time out.) 708 */ 709 if (ip->ip_off & ~(IP_DF|IP_RF)) { 710 /* 711 * Look for queue of fragments 712 * of this datagram. 713 */ 714 IPQ_LOCK(); 715 LIST_FOREACH(fp, &ipq, ipq_q) 716 if (ip->ip_id == fp->ipq_id && 717 in_hosteq(ip->ip_src, fp->ipq_src) && 718 in_hosteq(ip->ip_dst, fp->ipq_dst) && 719 ip->ip_p == fp->ipq_p) 720 goto found; 721 fp = 0; 722 found: 723 724 /* 725 * Adjust ip_len to not reflect header, 726 * set ipqe_mff if more fragments are expected, 727 * convert offset of this to bytes. 728 */ 729 ip->ip_len -= hlen; 730 mff = (ip->ip_off & IP_MF) != 0; 731 if (mff) { 732 /* 733 * Make sure that fragments have a data length 734 * that's a non-zero multiple of 8 bytes. 735 */ 736 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 737 ipstat.ips_badfrags++; 738 IPQ_UNLOCK(); 739 goto bad; 740 } 741 } 742 ip->ip_off <<= 3; 743 744 /* 745 * If datagram marked as having more fragments 746 * or if this is not the first fragment, 747 * attempt reassembly; if it succeeds, proceed. 748 */ 749 if (mff || ip->ip_off) { 750 ipstat.ips_fragments++; 751 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 752 if (ipqe == NULL) { 753 ipstat.ips_rcvmemdrop++; 754 IPQ_UNLOCK(); 755 goto bad; 756 } 757 ipqe->ipqe_mff = mff; 758 ipqe->ipqe_m = m; 759 ipqe->ipqe_ip = ip; 760 m = ip_reass(ipqe, fp); 761 if (m == 0) { 762 IPQ_UNLOCK(); 763 return; 764 } 765 ipstat.ips_reassembled++; 766 ip = mtod(m, struct ip *); 767 hlen = ip->ip_hl << 2; 768 ip->ip_len += hlen; 769 } else 770 if (fp) 771 ip_freef(fp); 772 IPQ_UNLOCK(); 773 } 774 775 #ifdef IPSEC 776 /* 777 * enforce IPsec policy checking if we are seeing last header. 778 * note that we do not visit this with protocols with pcb layer 779 * code - like udp/tcp/raw ip. 780 */ 781 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 782 ipsec4_in_reject(m, NULL)) { 783 ipsecstat.in_polvio++; 784 goto bad; 785 } 786 #endif 787 788 /* 789 * Switch out to protocol's input routine. 790 */ 791 #if IFA_STATS 792 if (ia && ip) 793 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 794 #endif 795 ipstat.ips_delivered++; 796 { 797 int off = hlen, nh = ip->ip_p; 798 799 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 800 return; 801 } 802 bad: 803 m_freem(m); 804 return; 805 806 badcsum: 807 ipstat.ips_badsum++; 808 m_freem(m); 809 } 810 811 /* 812 * Take incoming datagram fragment and try to 813 * reassemble it into whole datagram. If a chain for 814 * reassembly of this datagram already exists, then it 815 * is given as fp; otherwise have to make a chain. 816 */ 817 struct mbuf * 818 ip_reass(ipqe, fp) 819 struct ipqent *ipqe; 820 struct ipq *fp; 821 { 822 struct mbuf *m = ipqe->ipqe_m; 823 struct ipqent *nq, *p, *q; 824 struct ip *ip; 825 struct mbuf *t; 826 int hlen = ipqe->ipqe_ip->ip_hl << 2; 827 int i, next; 828 829 IPQ_LOCK_CHECK(); 830 831 /* 832 * Presence of header sizes in mbufs 833 * would confuse code below. 834 */ 835 m->m_data += hlen; 836 m->m_len -= hlen; 837 838 /* 839 * If first fragment to arrive, create a reassembly queue. 840 */ 841 if (fp == 0) { 842 /* 843 * Enforce upper bound on number of fragmented packets 844 * for which we attempt reassembly; 845 * If maxfrag is 0, never accept fragments. 846 * If maxfrag is -1, accept all fragments without limitation. 847 */ 848 if (ip_maxfragpackets < 0) 849 ; 850 else if (ip_nfragpackets >= ip_maxfragpackets) 851 goto dropfrag; 852 ip_nfragpackets++; 853 MALLOC(fp, struct ipq *, sizeof (struct ipq), 854 M_FTABLE, M_NOWAIT); 855 if (fp == NULL) 856 goto dropfrag; 857 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 858 fp->ipq_ttl = IPFRAGTTL; 859 fp->ipq_p = ipqe->ipqe_ip->ip_p; 860 fp->ipq_id = ipqe->ipqe_ip->ip_id; 861 TAILQ_INIT(&fp->ipq_fragq); 862 fp->ipq_src = ipqe->ipqe_ip->ip_src; 863 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 864 p = NULL; 865 goto insert; 866 } 867 868 /* 869 * Find a segment which begins after this one does. 870 */ 871 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 872 p = q, q = TAILQ_NEXT(q, ipqe_q)) 873 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 874 break; 875 876 /* 877 * If there is a preceding segment, it may provide some of 878 * our data already. If so, drop the data from the incoming 879 * segment. If it provides all of our data, drop us. 880 */ 881 if (p != NULL) { 882 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 883 ipqe->ipqe_ip->ip_off; 884 if (i > 0) { 885 if (i >= ipqe->ipqe_ip->ip_len) 886 goto dropfrag; 887 m_adj(ipqe->ipqe_m, i); 888 ipqe->ipqe_ip->ip_off += i; 889 ipqe->ipqe_ip->ip_len -= i; 890 } 891 } 892 893 /* 894 * While we overlap succeeding segments trim them or, 895 * if they are completely covered, dequeue them. 896 */ 897 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 898 q->ipqe_ip->ip_off; q = nq) { 899 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 900 q->ipqe_ip->ip_off; 901 if (i < q->ipqe_ip->ip_len) { 902 q->ipqe_ip->ip_len -= i; 903 q->ipqe_ip->ip_off += i; 904 m_adj(q->ipqe_m, i); 905 break; 906 } 907 nq = TAILQ_NEXT(q, ipqe_q); 908 m_freem(q->ipqe_m); 909 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 910 pool_put(&ipqent_pool, q); 911 } 912 913 insert: 914 /* 915 * Stick new segment in its place; 916 * check for complete reassembly. 917 */ 918 if (p == NULL) { 919 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 920 } else { 921 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q); 922 } 923 next = 0; 924 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; 925 p = q, q = TAILQ_NEXT(q, ipqe_q)) { 926 if (q->ipqe_ip->ip_off != next) 927 return (0); 928 next += q->ipqe_ip->ip_len; 929 } 930 if (p->ipqe_mff) 931 return (0); 932 933 /* 934 * Reassembly is complete. Check for a bogus message size and 935 * concatenate fragments. 936 */ 937 q = TAILQ_FIRST(&fp->ipq_fragq); 938 ip = q->ipqe_ip; 939 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 940 ipstat.ips_toolong++; 941 ip_freef(fp); 942 return (0); 943 } 944 m = q->ipqe_m; 945 t = m->m_next; 946 m->m_next = 0; 947 m_cat(m, t); 948 nq = TAILQ_NEXT(q, ipqe_q); 949 pool_put(&ipqent_pool, q); 950 for (q = nq; q != NULL; q = nq) { 951 t = q->ipqe_m; 952 nq = TAILQ_NEXT(q, ipqe_q); 953 pool_put(&ipqent_pool, q); 954 m_cat(m, t); 955 } 956 957 /* 958 * Create header for new ip packet by 959 * modifying header of first packet; 960 * dequeue and discard fragment reassembly header. 961 * Make header visible. 962 */ 963 ip->ip_len = next; 964 ip->ip_src = fp->ipq_src; 965 ip->ip_dst = fp->ipq_dst; 966 LIST_REMOVE(fp, ipq_q); 967 FREE(fp, M_FTABLE); 968 ip_nfragpackets--; 969 m->m_len += (ip->ip_hl << 2); 970 m->m_data -= (ip->ip_hl << 2); 971 /* some debugging cruft by sklower, below, will go away soon */ 972 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 973 int plen = 0; 974 for (t = m; t; t = t->m_next) 975 plen += t->m_len; 976 m->m_pkthdr.len = plen; 977 } 978 return (m); 979 980 dropfrag: 981 ipstat.ips_fragdropped++; 982 m_freem(m); 983 pool_put(&ipqent_pool, ipqe); 984 return (0); 985 } 986 987 /* 988 * Free a fragment reassembly header and all 989 * associated datagrams. 990 */ 991 void 992 ip_freef(fp) 993 struct ipq *fp; 994 { 995 struct ipqent *q, *p; 996 997 IPQ_LOCK_CHECK(); 998 999 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) { 1000 p = TAILQ_NEXT(q, ipqe_q); 1001 m_freem(q->ipqe_m); 1002 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); 1003 pool_put(&ipqent_pool, q); 1004 } 1005 LIST_REMOVE(fp, ipq_q); 1006 FREE(fp, M_FTABLE); 1007 ip_nfragpackets--; 1008 } 1009 1010 /* 1011 * IP timer processing; 1012 * if a timer expires on a reassembly 1013 * queue, discard it. 1014 */ 1015 void 1016 ip_slowtimo() 1017 { 1018 struct ipq *fp, *nfp; 1019 int s = splsoftnet(); 1020 1021 IPQ_LOCK(); 1022 for (fp = LIST_FIRST(&ipq); fp != NULL; fp = nfp) { 1023 nfp = LIST_NEXT(fp, ipq_q); 1024 if (--fp->ipq_ttl == 0) { 1025 ipstat.ips_fragtimeout++; 1026 ip_freef(fp); 1027 } 1028 } 1029 /* 1030 * If we are over the maximum number of fragments 1031 * (due to the limit being lowered), drain off 1032 * enough to get down to the new limit. 1033 */ 1034 if (ip_maxfragpackets < 0) 1035 ; 1036 else { 1037 while (ip_nfragpackets > ip_maxfragpackets && LIST_FIRST(&ipq)) 1038 ip_freef(LIST_FIRST(&ipq)); 1039 } 1040 IPQ_UNLOCK(); 1041 #ifdef GATEWAY 1042 ipflow_slowtimo(); 1043 #endif 1044 splx(s); 1045 } 1046 1047 /* 1048 * Drain off all datagram fragments. 1049 */ 1050 void 1051 ip_drain() 1052 { 1053 1054 /* 1055 * We may be called from a device's interrupt context. If 1056 * the ipq is already busy, just bail out now. 1057 */ 1058 if (ipq_lock_try() == 0) 1059 return; 1060 1061 while (LIST_FIRST(&ipq) != NULL) { 1062 ipstat.ips_fragdropped++; 1063 ip_freef(LIST_FIRST(&ipq)); 1064 } 1065 1066 IPQ_UNLOCK(); 1067 } 1068 1069 /* 1070 * Do option processing on a datagram, 1071 * possibly discarding it if bad options are encountered, 1072 * or forwarding it if source-routed. 1073 * Returns 1 if packet has been forwarded/freed, 1074 * 0 if the packet should be processed further. 1075 */ 1076 int 1077 ip_dooptions(m) 1078 struct mbuf *m; 1079 { 1080 struct ip *ip = mtod(m, struct ip *); 1081 u_char *cp, *cp0; 1082 struct ip_timestamp *ipt; 1083 struct in_ifaddr *ia; 1084 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1085 struct in_addr dst; 1086 n_time ntime; 1087 1088 dst = ip->ip_dst; 1089 cp = (u_char *)(ip + 1); 1090 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1091 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1092 opt = cp[IPOPT_OPTVAL]; 1093 if (opt == IPOPT_EOL) 1094 break; 1095 if (opt == IPOPT_NOP) 1096 optlen = 1; 1097 else { 1098 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1099 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1100 goto bad; 1101 } 1102 optlen = cp[IPOPT_OLEN]; 1103 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1104 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1105 goto bad; 1106 } 1107 } 1108 switch (opt) { 1109 1110 default: 1111 break; 1112 1113 /* 1114 * Source routing with record. 1115 * Find interface with current destination address. 1116 * If none on this machine then drop if strictly routed, 1117 * or do nothing if loosely routed. 1118 * Record interface address and bring up next address 1119 * component. If strictly routed make sure next 1120 * address is on directly accessible net. 1121 */ 1122 case IPOPT_LSRR: 1123 case IPOPT_SSRR: 1124 if (ip_allowsrcrt == 0) { 1125 type = ICMP_UNREACH; 1126 code = ICMP_UNREACH_NET_PROHIB; 1127 goto bad; 1128 } 1129 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1130 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1131 goto bad; 1132 } 1133 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1134 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1135 goto bad; 1136 } 1137 ipaddr.sin_addr = ip->ip_dst; 1138 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1139 if (ia == 0) { 1140 if (opt == IPOPT_SSRR) { 1141 type = ICMP_UNREACH; 1142 code = ICMP_UNREACH_SRCFAIL; 1143 goto bad; 1144 } 1145 /* 1146 * Loose routing, and not at next destination 1147 * yet; nothing to do except forward. 1148 */ 1149 break; 1150 } 1151 off--; /* 0 origin */ 1152 if ((off + sizeof(struct in_addr)) > optlen) { 1153 /* 1154 * End of source route. Should be for us. 1155 */ 1156 save_rte(cp, ip->ip_src); 1157 break; 1158 } 1159 /* 1160 * locate outgoing interface 1161 */ 1162 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1163 sizeof(ipaddr.sin_addr)); 1164 if (opt == IPOPT_SSRR) 1165 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1166 else 1167 ia = ip_rtaddr(ipaddr.sin_addr); 1168 if (ia == 0) { 1169 type = ICMP_UNREACH; 1170 code = ICMP_UNREACH_SRCFAIL; 1171 goto bad; 1172 } 1173 ip->ip_dst = ipaddr.sin_addr; 1174 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1175 (caddr_t)(cp + off), sizeof(struct in_addr)); 1176 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1177 /* 1178 * Let ip_intr's mcast routing check handle mcast pkts 1179 */ 1180 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1181 break; 1182 1183 case IPOPT_RR: 1184 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1185 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1186 goto bad; 1187 } 1188 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1189 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1190 goto bad; 1191 } 1192 /* 1193 * If no space remains, ignore. 1194 */ 1195 off--; /* 0 origin */ 1196 if ((off + sizeof(struct in_addr)) > optlen) 1197 break; 1198 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1199 sizeof(ipaddr.sin_addr)); 1200 /* 1201 * locate outgoing interface; if we're the destination, 1202 * use the incoming interface (should be same). 1203 */ 1204 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1205 == NULL && 1206 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1207 type = ICMP_UNREACH; 1208 code = ICMP_UNREACH_HOST; 1209 goto bad; 1210 } 1211 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1212 (caddr_t)(cp + off), sizeof(struct in_addr)); 1213 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1214 break; 1215 1216 case IPOPT_TS: 1217 code = cp - (u_char *)ip; 1218 ipt = (struct ip_timestamp *)cp; 1219 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1220 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1221 goto bad; 1222 } 1223 if (ipt->ipt_ptr < 5) { 1224 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1225 goto bad; 1226 } 1227 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1228 if (++ipt->ipt_oflw == 0) { 1229 code = (u_char *)&ipt->ipt_ptr - 1230 (u_char *)ip; 1231 goto bad; 1232 } 1233 break; 1234 } 1235 cp0 = (cp + ipt->ipt_ptr - 1); 1236 switch (ipt->ipt_flg) { 1237 1238 case IPOPT_TS_TSONLY: 1239 break; 1240 1241 case IPOPT_TS_TSANDADDR: 1242 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1243 sizeof(struct in_addr) > ipt->ipt_len) { 1244 code = (u_char *)&ipt->ipt_ptr - 1245 (u_char *)ip; 1246 goto bad; 1247 } 1248 ipaddr.sin_addr = dst; 1249 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1250 m->m_pkthdr.rcvif)); 1251 if (ia == 0) 1252 continue; 1253 bcopy(&ia->ia_addr.sin_addr, 1254 cp0, sizeof(struct in_addr)); 1255 ipt->ipt_ptr += sizeof(struct in_addr); 1256 break; 1257 1258 case IPOPT_TS_PRESPEC: 1259 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1260 sizeof(struct in_addr) > ipt->ipt_len) { 1261 code = (u_char *)&ipt->ipt_ptr - 1262 (u_char *)ip; 1263 goto bad; 1264 } 1265 bcopy(cp0, &ipaddr.sin_addr, 1266 sizeof(struct in_addr)); 1267 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1268 == NULL) 1269 continue; 1270 ipt->ipt_ptr += sizeof(struct in_addr); 1271 break; 1272 1273 default: 1274 /* XXX can't take &ipt->ipt_flg */ 1275 code = (u_char *)&ipt->ipt_ptr - 1276 (u_char *)ip + 1; 1277 goto bad; 1278 } 1279 ntime = iptime(); 1280 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1281 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1282 sizeof(n_time)); 1283 ipt->ipt_ptr += sizeof(n_time); 1284 } 1285 } 1286 if (forward) { 1287 if (ip_forwsrcrt == 0) { 1288 type = ICMP_UNREACH; 1289 code = ICMP_UNREACH_SRCFAIL; 1290 goto bad; 1291 } 1292 ip_forward(m, 1); 1293 return (1); 1294 } 1295 return (0); 1296 bad: 1297 icmp_error(m, type, code, 0, 0); 1298 ipstat.ips_badoptions++; 1299 return (1); 1300 } 1301 1302 /* 1303 * Given address of next destination (final or next hop), 1304 * return internet address info of interface to be used to get there. 1305 */ 1306 struct in_ifaddr * 1307 ip_rtaddr(dst) 1308 struct in_addr dst; 1309 { 1310 struct sockaddr_in *sin; 1311 1312 sin = satosin(&ipforward_rt.ro_dst); 1313 1314 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1315 if (ipforward_rt.ro_rt) { 1316 RTFREE(ipforward_rt.ro_rt); 1317 ipforward_rt.ro_rt = 0; 1318 } 1319 sin->sin_family = AF_INET; 1320 sin->sin_len = sizeof(*sin); 1321 sin->sin_addr = dst; 1322 1323 rtalloc(&ipforward_rt); 1324 } 1325 if (ipforward_rt.ro_rt == 0) 1326 return ((struct in_ifaddr *)0); 1327 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1328 } 1329 1330 /* 1331 * Save incoming source route for use in replies, 1332 * to be picked up later by ip_srcroute if the receiver is interested. 1333 */ 1334 void 1335 save_rte(option, dst) 1336 u_char *option; 1337 struct in_addr dst; 1338 { 1339 unsigned olen; 1340 1341 olen = option[IPOPT_OLEN]; 1342 #ifdef DIAGNOSTIC 1343 if (ipprintfs) 1344 printf("save_rte: olen %d\n", olen); 1345 #endif /* 0 */ 1346 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1347 return; 1348 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1349 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1350 ip_srcrt.dst = dst; 1351 } 1352 1353 /* 1354 * Retrieve incoming source route for use in replies, 1355 * in the same form used by setsockopt. 1356 * The first hop is placed before the options, will be removed later. 1357 */ 1358 struct mbuf * 1359 ip_srcroute() 1360 { 1361 struct in_addr *p, *q; 1362 struct mbuf *m; 1363 1364 if (ip_nhops == 0) 1365 return ((struct mbuf *)0); 1366 m = m_get(M_DONTWAIT, MT_SOOPTS); 1367 if (m == 0) 1368 return ((struct mbuf *)0); 1369 1370 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1371 1372 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1373 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1374 OPTSIZ; 1375 #ifdef DIAGNOSTIC 1376 if (ipprintfs) 1377 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1378 #endif 1379 1380 /* 1381 * First save first hop for return route 1382 */ 1383 p = &ip_srcrt.route[ip_nhops - 1]; 1384 *(mtod(m, struct in_addr *)) = *p--; 1385 #ifdef DIAGNOSTIC 1386 if (ipprintfs) 1387 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1388 #endif 1389 1390 /* 1391 * Copy option fields and padding (nop) to mbuf. 1392 */ 1393 ip_srcrt.nop = IPOPT_NOP; 1394 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1395 bcopy((caddr_t)&ip_srcrt.nop, 1396 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1397 q = (struct in_addr *)(mtod(m, caddr_t) + 1398 sizeof(struct in_addr) + OPTSIZ); 1399 #undef OPTSIZ 1400 /* 1401 * Record return path as an IP source route, 1402 * reversing the path (pointers are now aligned). 1403 */ 1404 while (p >= ip_srcrt.route) { 1405 #ifdef DIAGNOSTIC 1406 if (ipprintfs) 1407 printf(" %x", ntohl(q->s_addr)); 1408 #endif 1409 *q++ = *p--; 1410 } 1411 /* 1412 * Last hop goes to final destination. 1413 */ 1414 *q = ip_srcrt.dst; 1415 #ifdef DIAGNOSTIC 1416 if (ipprintfs) 1417 printf(" %x\n", ntohl(q->s_addr)); 1418 #endif 1419 return (m); 1420 } 1421 1422 /* 1423 * Strip out IP options, at higher 1424 * level protocol in the kernel. 1425 * Second argument is buffer to which options 1426 * will be moved, and return value is their length. 1427 * XXX should be deleted; last arg currently ignored. 1428 */ 1429 void 1430 ip_stripoptions(m, mopt) 1431 struct mbuf *m; 1432 struct mbuf *mopt; 1433 { 1434 int i; 1435 struct ip *ip = mtod(m, struct ip *); 1436 caddr_t opts; 1437 int olen; 1438 1439 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1440 opts = (caddr_t)(ip + 1); 1441 i = m->m_len - (sizeof (struct ip) + olen); 1442 bcopy(opts + olen, opts, (unsigned)i); 1443 m->m_len -= olen; 1444 if (m->m_flags & M_PKTHDR) 1445 m->m_pkthdr.len -= olen; 1446 ip->ip_len -= olen; 1447 ip->ip_hl = sizeof (struct ip) >> 2; 1448 } 1449 1450 const int inetctlerrmap[PRC_NCMDS] = { 1451 0, 0, 0, 0, 1452 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1453 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1454 EMSGSIZE, EHOSTUNREACH, 0, 0, 1455 0, 0, 0, 0, 1456 ENOPROTOOPT 1457 }; 1458 1459 /* 1460 * Forward a packet. If some error occurs return the sender 1461 * an icmp packet. Note we can't always generate a meaningful 1462 * icmp message because icmp doesn't have a large enough repertoire 1463 * of codes and types. 1464 * 1465 * If not forwarding, just drop the packet. This could be confusing 1466 * if ipforwarding was zero but some routing protocol was advancing 1467 * us as a gateway to somewhere. However, we must let the routing 1468 * protocol deal with that. 1469 * 1470 * The srcrt parameter indicates whether the packet is being forwarded 1471 * via a source route. 1472 */ 1473 void 1474 ip_forward(m, srcrt) 1475 struct mbuf *m; 1476 int srcrt; 1477 { 1478 struct ip *ip = mtod(m, struct ip *); 1479 struct sockaddr_in *sin; 1480 struct rtentry *rt; 1481 int error, type = 0, code = 0; 1482 struct mbuf *mcopy; 1483 n_long dest; 1484 struct ifnet *destifp; 1485 #ifdef IPSEC 1486 struct ifnet dummyifp; 1487 #endif 1488 1489 /* 1490 * Clear any in-bound checksum flags for this packet. 1491 */ 1492 m->m_pkthdr.csum_flags = 0; 1493 1494 dest = 0; 1495 #ifdef DIAGNOSTIC 1496 if (ipprintfs) 1497 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1498 ntohl(ip->ip_src.s_addr), 1499 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1500 #endif 1501 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1502 ipstat.ips_cantforward++; 1503 m_freem(m); 1504 return; 1505 } 1506 if (ip->ip_ttl <= IPTTLDEC) { 1507 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1508 return; 1509 } 1510 ip->ip_ttl -= IPTTLDEC; 1511 1512 sin = satosin(&ipforward_rt.ro_dst); 1513 if ((rt = ipforward_rt.ro_rt) == 0 || 1514 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1515 if (ipforward_rt.ro_rt) { 1516 RTFREE(ipforward_rt.ro_rt); 1517 ipforward_rt.ro_rt = 0; 1518 } 1519 sin->sin_family = AF_INET; 1520 sin->sin_len = sizeof(struct sockaddr_in); 1521 sin->sin_addr = ip->ip_dst; 1522 1523 rtalloc(&ipforward_rt); 1524 if (ipforward_rt.ro_rt == 0) { 1525 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1526 return; 1527 } 1528 rt = ipforward_rt.ro_rt; 1529 } 1530 1531 /* 1532 * Save at most 68 bytes of the packet in case 1533 * we need to generate an ICMP message to the src. 1534 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1535 */ 1536 mcopy = m_copym(m, 0, imin((int)ip->ip_len, 68), M_DONTWAIT); 1537 if (mcopy) 1538 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1539 1540 /* 1541 * If forwarding packet using same interface that it came in on, 1542 * perhaps should send a redirect to sender to shortcut a hop. 1543 * Only send redirect if source is sending directly to us, 1544 * and if packet was not source routed (or has any options). 1545 * Also, don't send redirect if forwarding using a default route 1546 * or a route modified by a redirect. 1547 */ 1548 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1549 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1550 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1551 ipsendredirects && !srcrt) { 1552 if (rt->rt_ifa && 1553 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1554 ifatoia(rt->rt_ifa)->ia_subnet) { 1555 if (rt->rt_flags & RTF_GATEWAY) 1556 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1557 else 1558 dest = ip->ip_dst.s_addr; 1559 /* 1560 * Router requirements says to only send host 1561 * redirects. 1562 */ 1563 type = ICMP_REDIRECT; 1564 code = ICMP_REDIRECT_HOST; 1565 #ifdef DIAGNOSTIC 1566 if (ipprintfs) 1567 printf("redirect (%d) to %x\n", code, 1568 (u_int32_t)dest); 1569 #endif 1570 } 1571 } 1572 1573 #ifdef IPSEC 1574 /* Don't lookup socket in forwarding case */ 1575 (void)ipsec_setsocket(m, NULL); 1576 #endif 1577 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1578 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1579 if (error) 1580 ipstat.ips_cantforward++; 1581 else { 1582 ipstat.ips_forward++; 1583 if (type) 1584 ipstat.ips_redirectsent++; 1585 else { 1586 if (mcopy) { 1587 #ifdef GATEWAY 1588 if (mcopy->m_flags & M_CANFASTFWD) 1589 ipflow_create(&ipforward_rt, mcopy); 1590 #endif 1591 m_freem(mcopy); 1592 } 1593 return; 1594 } 1595 } 1596 if (mcopy == NULL) 1597 return; 1598 destifp = NULL; 1599 1600 switch (error) { 1601 1602 case 0: /* forwarded, but need redirect */ 1603 /* type, code set above */ 1604 break; 1605 1606 case ENETUNREACH: /* shouldn't happen, checked above */ 1607 case EHOSTUNREACH: 1608 case ENETDOWN: 1609 case EHOSTDOWN: 1610 default: 1611 type = ICMP_UNREACH; 1612 code = ICMP_UNREACH_HOST; 1613 break; 1614 1615 case EMSGSIZE: 1616 type = ICMP_UNREACH; 1617 code = ICMP_UNREACH_NEEDFRAG; 1618 #ifndef IPSEC 1619 if (ipforward_rt.ro_rt) 1620 destifp = ipforward_rt.ro_rt->rt_ifp; 1621 #else 1622 /* 1623 * If the packet is routed over IPsec tunnel, tell the 1624 * originator the tunnel MTU. 1625 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1626 * XXX quickhack!!! 1627 */ 1628 if (ipforward_rt.ro_rt) { 1629 struct secpolicy *sp; 1630 int ipsecerror; 1631 size_t ipsechdr; 1632 struct route *ro; 1633 1634 sp = ipsec4_getpolicybyaddr(mcopy, 1635 IPSEC_DIR_OUTBOUND, 1636 IP_FORWARDING, 1637 &ipsecerror); 1638 1639 if (sp == NULL) 1640 destifp = ipforward_rt.ro_rt->rt_ifp; 1641 else { 1642 /* count IPsec header size */ 1643 ipsechdr = ipsec4_hdrsiz(mcopy, 1644 IPSEC_DIR_OUTBOUND, 1645 NULL); 1646 1647 /* 1648 * find the correct route for outer IPv4 1649 * header, compute tunnel MTU. 1650 * 1651 * XXX BUG ALERT 1652 * The "dummyifp" code relies upon the fact 1653 * that icmp_error() touches only ifp->if_mtu. 1654 */ 1655 /*XXX*/ 1656 destifp = NULL; 1657 if (sp->req != NULL 1658 && sp->req->sav != NULL 1659 && sp->req->sav->sah != NULL) { 1660 ro = &sp->req->sav->sah->sa_route; 1661 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1662 dummyifp.if_mtu = 1663 ro->ro_rt->rt_ifp->if_mtu; 1664 dummyifp.if_mtu -= ipsechdr; 1665 destifp = &dummyifp; 1666 } 1667 } 1668 1669 key_freesp(sp); 1670 } 1671 } 1672 #endif /*IPSEC*/ 1673 ipstat.ips_cantfrag++; 1674 break; 1675 1676 case ENOBUFS: 1677 #if 1 1678 /* 1679 * a router should not generate ICMP_SOURCEQUENCH as 1680 * required in RFC1812 Requirements for IP Version 4 Routers. 1681 * source quench could be a big problem under DoS attacks, 1682 * or the underlying interface is rate-limited. 1683 */ 1684 if (mcopy) 1685 m_freem(mcopy); 1686 return; 1687 #else 1688 type = ICMP_SOURCEQUENCH; 1689 code = 0; 1690 break; 1691 #endif 1692 } 1693 icmp_error(mcopy, type, code, dest, destifp); 1694 } 1695 1696 void 1697 ip_savecontrol(inp, mp, ip, m) 1698 struct inpcb *inp; 1699 struct mbuf **mp; 1700 struct ip *ip; 1701 struct mbuf *m; 1702 { 1703 1704 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1705 struct timeval tv; 1706 1707 microtime(&tv); 1708 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1709 SCM_TIMESTAMP, SOL_SOCKET); 1710 if (*mp) 1711 mp = &(*mp)->m_next; 1712 } 1713 if (inp->inp_flags & INP_RECVDSTADDR) { 1714 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1715 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1716 if (*mp) 1717 mp = &(*mp)->m_next; 1718 } 1719 #ifdef notyet 1720 /* 1721 * XXX 1722 * Moving these out of udp_input() made them even more broken 1723 * than they already were. 1724 * - fenner@parc.xerox.com 1725 */ 1726 /* options were tossed already */ 1727 if (inp->inp_flags & INP_RECVOPTS) { 1728 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1729 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1730 if (*mp) 1731 mp = &(*mp)->m_next; 1732 } 1733 /* ip_srcroute doesn't do what we want here, need to fix */ 1734 if (inp->inp_flags & INP_RECVRETOPTS) { 1735 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1736 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1737 if (*mp) 1738 mp = &(*mp)->m_next; 1739 } 1740 #endif 1741 if (inp->inp_flags & INP_RECVIF) { 1742 struct sockaddr_dl sdl; 1743 1744 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1745 sdl.sdl_family = AF_LINK; 1746 sdl.sdl_index = m->m_pkthdr.rcvif ? 1747 m->m_pkthdr.rcvif->if_index : 0; 1748 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1749 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1750 IP_RECVIF, IPPROTO_IP); 1751 if (*mp) 1752 mp = &(*mp)->m_next; 1753 } 1754 } 1755 1756 int 1757 ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1758 int *name; 1759 u_int namelen; 1760 void *oldp; 1761 size_t *oldlenp; 1762 void *newp; 1763 size_t newlen; 1764 { 1765 extern int subnetsarelocal, hostzeroisbroadcast; 1766 1767 int error, old; 1768 1769 /* All sysctl names at this level are terminal. */ 1770 if (namelen != 1) 1771 return (ENOTDIR); 1772 1773 switch (name[0]) { 1774 case IPCTL_FORWARDING: 1775 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1776 case IPCTL_SENDREDIRECTS: 1777 return (sysctl_int(oldp, oldlenp, newp, newlen, 1778 &ipsendredirects)); 1779 case IPCTL_DEFTTL: 1780 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1781 #ifdef notyet 1782 case IPCTL_DEFMTU: 1783 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1784 #endif 1785 case IPCTL_FORWSRCRT: 1786 /* Don't allow this to change in a secure environment. */ 1787 if (securelevel > 0) 1788 return (sysctl_rdint(oldp, oldlenp, newp, 1789 ip_forwsrcrt)); 1790 else 1791 return (sysctl_int(oldp, oldlenp, newp, newlen, 1792 &ip_forwsrcrt)); 1793 case IPCTL_DIRECTEDBCAST: 1794 return (sysctl_int(oldp, oldlenp, newp, newlen, 1795 &ip_directedbcast)); 1796 case IPCTL_ALLOWSRCRT: 1797 return (sysctl_int(oldp, oldlenp, newp, newlen, 1798 &ip_allowsrcrt)); 1799 case IPCTL_SUBNETSARELOCAL: 1800 return (sysctl_int(oldp, oldlenp, newp, newlen, 1801 &subnetsarelocal)); 1802 case IPCTL_MTUDISC: 1803 error = sysctl_int(oldp, oldlenp, newp, newlen, 1804 &ip_mtudisc); 1805 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1806 ip_mtudisc_timeout_q = 1807 rt_timer_queue_create(ip_mtudisc_timeout); 1808 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1809 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1810 ip_mtudisc_timeout_q = NULL; 1811 } 1812 return error; 1813 case IPCTL_ANONPORTMIN: 1814 old = anonportmin; 1815 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1816 if (anonportmin >= anonportmax || anonportmin < 0 1817 || anonportmin > 65535 1818 #ifndef IPNOPRIVPORTS 1819 || anonportmin < IPPORT_RESERVED 1820 #endif 1821 ) { 1822 anonportmin = old; 1823 return (EINVAL); 1824 } 1825 return (error); 1826 case IPCTL_ANONPORTMAX: 1827 old = anonportmax; 1828 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1829 if (anonportmin >= anonportmax || anonportmax < 0 1830 || anonportmax > 65535 1831 #ifndef IPNOPRIVPORTS 1832 || anonportmax < IPPORT_RESERVED 1833 #endif 1834 ) { 1835 anonportmax = old; 1836 return (EINVAL); 1837 } 1838 return (error); 1839 case IPCTL_MTUDISCTIMEOUT: 1840 error = sysctl_int(oldp, oldlenp, newp, newlen, 1841 &ip_mtudisc_timeout); 1842 if (ip_mtudisc_timeout_q != NULL) 1843 rt_timer_queue_change(ip_mtudisc_timeout_q, 1844 ip_mtudisc_timeout); 1845 return (error); 1846 #ifdef GATEWAY 1847 case IPCTL_MAXFLOWS: 1848 { 1849 int s; 1850 1851 error = sysctl_int(oldp, oldlenp, newp, newlen, 1852 &ip_maxflows); 1853 s = splsoftnet(); 1854 ipflow_reap(0); 1855 splx(s); 1856 return (error); 1857 } 1858 #endif 1859 case IPCTL_HOSTZEROBROADCAST: 1860 return (sysctl_int(oldp, oldlenp, newp, newlen, 1861 &hostzeroisbroadcast)); 1862 #if NGIF > 0 1863 case IPCTL_GIF_TTL: 1864 return(sysctl_int(oldp, oldlenp, newp, newlen, 1865 &ip_gif_ttl)); 1866 #endif 1867 1868 #if NGRE > 0 1869 case IPCTL_GRE_TTL: 1870 return(sysctl_int(oldp, oldlenp, newp, newlen, 1871 &ip_gre_ttl)); 1872 #endif 1873 1874 #ifndef IPNOPRIVPORTS 1875 case IPCTL_LOWPORTMIN: 1876 old = lowportmin; 1877 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmin); 1878 if (lowportmin >= lowportmax 1879 || lowportmin > IPPORT_RESERVEDMAX 1880 || lowportmin < IPPORT_RESERVEDMIN 1881 ) { 1882 lowportmin = old; 1883 return (EINVAL); 1884 } 1885 return (error); 1886 case IPCTL_LOWPORTMAX: 1887 old = lowportmax; 1888 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmax); 1889 if (lowportmin >= lowportmax 1890 || lowportmax > IPPORT_RESERVEDMAX 1891 || lowportmax < IPPORT_RESERVEDMIN 1892 ) { 1893 lowportmax = old; 1894 return (EINVAL); 1895 } 1896 return (error); 1897 #endif 1898 1899 case IPCTL_MAXFRAGPACKETS: 1900 return (sysctl_int(oldp, oldlenp, newp, newlen, 1901 &ip_maxfragpackets)); 1902 1903 default: 1904 return (EOPNOTSUPP); 1905 } 1906 /* NOTREACHED */ 1907 } 1908