1 /* $NetBSD: ip_input.c,v 1.136 2001/08/06 10:25:00 itojun Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 /*- 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 3. All advertising materials mentioning features or use of this software 49 * must display the following acknowledgement: 50 * This product includes software developed by the NetBSD 51 * Foundation, Inc. and its contributors. 52 * 4. Neither the name of The NetBSD Foundation nor the names of its 53 * contributors may be used to endorse or promote products derived 54 * from this software without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 57 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 58 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 59 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 60 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 61 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 62 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 63 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 64 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 65 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 66 * POSSIBILITY OF SUCH DAMAGE. 67 */ 68 69 /* 70 * Copyright (c) 1982, 1986, 1988, 1993 71 * The Regents of the University of California. All rights reserved. 72 * 73 * Redistribution and use in source and binary forms, with or without 74 * modification, are permitted provided that the following conditions 75 * are met: 76 * 1. Redistributions of source code must retain the above copyright 77 * notice, this list of conditions and the following disclaimer. 78 * 2. Redistributions in binary form must reproduce the above copyright 79 * notice, this list of conditions and the following disclaimer in the 80 * documentation and/or other materials provided with the distribution. 81 * 3. All advertising materials mentioning features or use of this software 82 * must display the following acknowledgement: 83 * This product includes software developed by the University of 84 * California, Berkeley and its contributors. 85 * 4. Neither the name of the University nor the names of its contributors 86 * may be used to endorse or promote products derived from this software 87 * without specific prior written permission. 88 * 89 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 90 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 91 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 92 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 93 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 94 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 95 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 96 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 97 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 98 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 99 * SUCH DAMAGE. 100 * 101 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 102 */ 103 104 #include "opt_gateway.h" 105 #include "opt_pfil_hooks.h" 106 #include "opt_ipsec.h" 107 #include "opt_mrouting.h" 108 #include "opt_inet_csum.h" 109 110 #include <sys/param.h> 111 #include <sys/systm.h> 112 #include <sys/malloc.h> 113 #include <sys/mbuf.h> 114 #include <sys/domain.h> 115 #include <sys/protosw.h> 116 #include <sys/socket.h> 117 #include <sys/socketvar.h> 118 #include <sys/errno.h> 119 #include <sys/time.h> 120 #include <sys/kernel.h> 121 #include <sys/proc.h> 122 #include <sys/pool.h> 123 124 #include <uvm/uvm_extern.h> 125 126 #include <sys/sysctl.h> 127 128 #include <net/if.h> 129 #include <net/if_dl.h> 130 #include <net/route.h> 131 #include <net/pfil.h> 132 133 #include <netinet/in.h> 134 #include <netinet/in_systm.h> 135 #include <netinet/ip.h> 136 #include <netinet/in_pcb.h> 137 #include <netinet/in_var.h> 138 #include <netinet/ip_var.h> 139 #include <netinet/ip_icmp.h> 140 /* just for gif_ttl */ 141 #include <netinet/in_gif.h> 142 #include "gif.h" 143 144 #ifdef MROUTING 145 #include <netinet/ip_mroute.h> 146 #endif 147 148 #ifdef IPSEC 149 #include <netinet6/ipsec.h> 150 #include <netkey/key.h> 151 #endif 152 153 #ifndef IPFORWARDING 154 #ifdef GATEWAY 155 #define IPFORWARDING 1 /* forward IP packets not for us */ 156 #else /* GATEWAY */ 157 #define IPFORWARDING 0 /* don't forward IP packets not for us */ 158 #endif /* GATEWAY */ 159 #endif /* IPFORWARDING */ 160 #ifndef IPSENDREDIRECTS 161 #define IPSENDREDIRECTS 1 162 #endif 163 #ifndef IPFORWSRCRT 164 #define IPFORWSRCRT 1 /* forward source-routed packets */ 165 #endif 166 #ifndef IPALLOWSRCRT 167 #define IPALLOWSRCRT 1 /* allow source-routed packets */ 168 #endif 169 #ifndef IPMTUDISC 170 #define IPMTUDISC 0 171 #endif 172 #ifndef IPMTUDISCTIMEOUT 173 #define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 174 #endif 175 176 /* 177 * Note: DIRECTED_BROADCAST is handled this way so that previous 178 * configuration using this option will Just Work. 179 */ 180 #ifndef IPDIRECTEDBCAST 181 #ifdef DIRECTED_BROADCAST 182 #define IPDIRECTEDBCAST 1 183 #else 184 #define IPDIRECTEDBCAST 0 185 #endif /* DIRECTED_BROADCAST */ 186 #endif /* IPDIRECTEDBCAST */ 187 int ipforwarding = IPFORWARDING; 188 int ipsendredirects = IPSENDREDIRECTS; 189 int ip_defttl = IPDEFTTL; 190 int ip_forwsrcrt = IPFORWSRCRT; 191 int ip_directedbcast = IPDIRECTEDBCAST; 192 int ip_allowsrcrt = IPALLOWSRCRT; 193 int ip_mtudisc = IPMTUDISC; 194 u_int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 195 #ifdef DIAGNOSTIC 196 int ipprintfs = 0; 197 #endif 198 199 struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 200 201 extern struct domain inetdomain; 202 int ipqmaxlen = IFQ_MAXLEN; 203 struct in_ifaddrhead in_ifaddr; 204 struct in_ifaddrhashhead *in_ifaddrhashtbl; 205 struct ifqueue ipintrq; 206 struct ipstat ipstat; 207 u_int16_t ip_id; 208 209 #ifdef PFIL_HOOKS 210 struct pfil_head inet_pfil_hook; 211 #endif 212 213 struct ipqhead ipq; 214 int ipq_locked; 215 int ip_nfragpackets = 0; 216 int ip_maxfragpackets = 200; 217 218 static __inline int ipq_lock_try __P((void)); 219 static __inline void ipq_unlock __P((void)); 220 221 static __inline int 222 ipq_lock_try() 223 { 224 int s; 225 226 /* 227 * Use splvm() -- we're bloking things that would cause 228 * mbuf allocation. 229 */ 230 s = splvm(); 231 if (ipq_locked) { 232 splx(s); 233 return (0); 234 } 235 ipq_locked = 1; 236 splx(s); 237 return (1); 238 } 239 240 static __inline void 241 ipq_unlock() 242 { 243 int s; 244 245 s = splvm(); 246 ipq_locked = 0; 247 splx(s); 248 } 249 250 #ifdef DIAGNOSTIC 251 #define IPQ_LOCK() \ 252 do { \ 253 if (ipq_lock_try() == 0) { \ 254 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 255 panic("ipq_lock"); \ 256 } \ 257 } while (0) 258 #define IPQ_LOCK_CHECK() \ 259 do { \ 260 if (ipq_locked == 0) { \ 261 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 262 panic("ipq lock check"); \ 263 } \ 264 } while (0) 265 #else 266 #define IPQ_LOCK() (void) ipq_lock_try() 267 #define IPQ_LOCK_CHECK() /* nothing */ 268 #endif 269 270 #define IPQ_UNLOCK() ipq_unlock() 271 272 struct pool ipqent_pool; 273 274 #ifdef INET_CSUM_COUNTERS 275 #include <sys/device.h> 276 277 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 278 NULL, "inet", "hwcsum bad"); 279 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 280 NULL, "inet", "hwcsum ok"); 281 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 282 NULL, "inet", "swcsum"); 283 284 #define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 285 286 #else 287 288 #define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 289 290 #endif /* INET_CSUM_COUNTERS */ 291 292 /* 293 * We need to save the IP options in case a protocol wants to respond 294 * to an incoming packet over the same route if the packet got here 295 * using IP source routing. This allows connection establishment and 296 * maintenance when the remote end is on a network that is not known 297 * to us. 298 */ 299 int ip_nhops = 0; 300 static struct ip_srcrt { 301 struct in_addr dst; /* final destination */ 302 char nop; /* one NOP to align */ 303 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 304 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 305 } ip_srcrt; 306 307 static void save_rte __P((u_char *, struct in_addr)); 308 309 /* 310 * IP initialization: fill in IP protocol switch table. 311 * All protocols not implemented in kernel go to raw IP protocol handler. 312 */ 313 void 314 ip_init() 315 { 316 struct protosw *pr; 317 int i; 318 319 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 320 0, NULL, NULL, M_IPQ); 321 322 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 323 if (pr == 0) 324 panic("ip_init"); 325 for (i = 0; i < IPPROTO_MAX; i++) 326 ip_protox[i] = pr - inetsw; 327 for (pr = inetdomain.dom_protosw; 328 pr < inetdomain.dom_protoswNPROTOSW; pr++) 329 if (pr->pr_domain->dom_family == PF_INET && 330 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 331 ip_protox[pr->pr_protocol] = pr - inetsw; 332 LIST_INIT(&ipq); 333 ip_id = time.tv_sec & 0xffff; 334 ipintrq.ifq_maxlen = ipqmaxlen; 335 TAILQ_INIT(&in_ifaddr); 336 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 337 M_WAITOK, &in_ifaddrhash); 338 if (ip_mtudisc != 0) 339 ip_mtudisc_timeout_q = 340 rt_timer_queue_create(ip_mtudisc_timeout); 341 #ifdef GATEWAY 342 ipflow_init(); 343 #endif 344 345 #ifdef PFIL_HOOKS 346 /* Register our Packet Filter hook. */ 347 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 348 inet_pfil_hook.ph_af = AF_INET; 349 i = pfil_head_register(&inet_pfil_hook); 350 if (i != 0) 351 printf("ip_init: WARNING: unable to register pfil hook, " 352 "error %d\n", i); 353 #endif /* PFIL_HOOKS */ 354 355 #ifdef INET_CSUM_COUNTERS 356 evcnt_attach_static(&ip_hwcsum_bad); 357 evcnt_attach_static(&ip_hwcsum_ok); 358 evcnt_attach_static(&ip_swcsum); 359 #endif /* INET_CSUM_COUNTERS */ 360 } 361 362 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 363 struct route ipforward_rt; 364 365 /* 366 * IP software interrupt routine 367 */ 368 void 369 ipintr() 370 { 371 int s; 372 struct mbuf *m; 373 374 while (1) { 375 s = splnet(); 376 IF_DEQUEUE(&ipintrq, m); 377 splx(s); 378 if (m == 0) 379 return; 380 ip_input(m); 381 } 382 } 383 384 /* 385 * Ip input routine. Checksum and byte swap header. If fragmented 386 * try to reassemble. Process options. Pass to next level. 387 */ 388 void 389 ip_input(struct mbuf *m) 390 { 391 struct ip *ip = NULL; 392 struct ipq *fp; 393 struct in_ifaddr *ia; 394 struct ifaddr *ifa; 395 struct ipqent *ipqe; 396 int hlen = 0, mff, len; 397 int downmatch; 398 399 #ifdef DIAGNOSTIC 400 if ((m->m_flags & M_PKTHDR) == 0) 401 panic("ipintr no HDR"); 402 #endif 403 #ifdef IPSEC 404 /* 405 * should the inner packet be considered authentic? 406 * see comment in ah4_input(). 407 */ 408 if (m) { 409 m->m_flags &= ~M_AUTHIPHDR; 410 m->m_flags &= ~M_AUTHIPDGM; 411 } 412 #endif 413 /* 414 * If no IP addresses have been set yet but the interfaces 415 * are receiving, can't do anything with incoming packets yet. 416 */ 417 if (in_ifaddr.tqh_first == 0) 418 goto bad; 419 ipstat.ips_total++; 420 if (m->m_len < sizeof (struct ip) && 421 (m = m_pullup(m, sizeof (struct ip))) == 0) { 422 ipstat.ips_toosmall++; 423 return; 424 } 425 ip = mtod(m, struct ip *); 426 if (ip->ip_v != IPVERSION) { 427 ipstat.ips_badvers++; 428 goto bad; 429 } 430 hlen = ip->ip_hl << 2; 431 if (hlen < sizeof(struct ip)) { /* minimum header length */ 432 ipstat.ips_badhlen++; 433 goto bad; 434 } 435 if (hlen > m->m_len) { 436 if ((m = m_pullup(m, hlen)) == 0) { 437 ipstat.ips_badhlen++; 438 return; 439 } 440 ip = mtod(m, struct ip *); 441 } 442 443 /* 444 * RFC1122: packets with a multicast source address are 445 * not allowed. 446 */ 447 if (IN_MULTICAST(ip->ip_src.s_addr)) { 448 ipstat.ips_badaddr++; 449 goto bad; 450 } 451 452 /* 127/8 must not appear on wire - RFC1122 */ 453 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 454 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 455 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 456 ipstat.ips_badaddr++; 457 goto bad; 458 } 459 } 460 461 switch (m->m_pkthdr.csum_flags & 462 ((m->m_pkthdr.rcvif->if_csum_flags & M_CSUM_IPv4) | 463 M_CSUM_IPv4_BAD)) { 464 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 465 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 466 goto badcsum; 467 468 case M_CSUM_IPv4: 469 /* Checksum was okay. */ 470 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 471 break; 472 473 default: 474 /* Must compute it ourselves. */ 475 INET_CSUM_COUNTER_INCR(&ip_swcsum); 476 if (in_cksum(m, hlen) != 0) 477 goto bad; 478 break; 479 } 480 481 /* Retrieve the packet length. */ 482 len = ntohs(ip->ip_len); 483 484 /* 485 * Check for additional length bogosity 486 */ 487 if (len < hlen) { 488 ipstat.ips_badlen++; 489 goto bad; 490 } 491 492 /* 493 * Check that the amount of data in the buffers 494 * is as at least much as the IP header would have us expect. 495 * Trim mbufs if longer than we expect. 496 * Drop packet if shorter than we expect. 497 */ 498 if (m->m_pkthdr.len < len) { 499 ipstat.ips_tooshort++; 500 goto bad; 501 } 502 if (m->m_pkthdr.len > len) { 503 if (m->m_len == m->m_pkthdr.len) { 504 m->m_len = len; 505 m->m_pkthdr.len = len; 506 } else 507 m_adj(m, len - m->m_pkthdr.len); 508 } 509 510 #ifdef IPSEC 511 /* ipflow (IP fast fowarding) is not compatible with IPsec. */ 512 m->m_flags &= ~M_CANFASTFWD; 513 #else 514 /* 515 * Assume that we can create a fast-forward IP flow entry 516 * based on this packet. 517 */ 518 m->m_flags |= M_CANFASTFWD; 519 #endif 520 521 #ifdef PFIL_HOOKS 522 /* 523 * Run through list of hooks for input packets. If there are any 524 * filters which require that additional packets in the flow are 525 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 526 * Note that filters must _never_ set this flag, as another filter 527 * in the list may have previously cleared it. 528 */ 529 /* 530 * let ipfilter look at packet on the wire, 531 * not the decapsulated packet. 532 */ 533 #ifdef IPSEC 534 if (!ipsec_getnhist(m)) 535 #else 536 if (1) 537 #endif 538 { 539 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 540 PFIL_IN) != 0) 541 return; 542 if (m == NULL) 543 return; 544 ip = mtod(m, struct ip *); 545 } 546 #endif /* PFIL_HOOKS */ 547 548 #ifdef ALTQ 549 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 550 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 551 /* packet dropped by traffic conditioner */ 552 return; 553 } 554 #endif 555 556 /* 557 * Convert fields to host representation. 558 */ 559 NTOHS(ip->ip_len); 560 NTOHS(ip->ip_off); 561 562 /* 563 * Process options and, if not destined for us, 564 * ship it on. ip_dooptions returns 1 when an 565 * error was detected (causing an icmp message 566 * to be sent and the original packet to be freed). 567 */ 568 ip_nhops = 0; /* for source routed packets */ 569 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 570 return; 571 572 /* 573 * Check our list of addresses, to see if the packet is for us. 574 * 575 * Traditional 4.4BSD did not consult IFF_UP at all. 576 * The behavior here is to treat addresses on !IFF_UP interface 577 * as not mine. 578 */ 579 downmatch = 0; 580 for (ia = IN_IFADDR_HASH(ip->ip_dst.s_addr).lh_first; 581 ia != NULL; 582 ia = ia->ia_hash.le_next) { 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 for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first; 594 ifa != NULL; ifa = ifa->ifa_list.tqe_next) { 595 if (ifa->ifa_addr->sa_family != AF_INET) 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->m_flags & M_EXT) { 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 ip_forward(m, 0); 691 } 692 return; 693 694 ours: 695 /* 696 * If offset or IP_MF are set, must reassemble. 697 * Otherwise, nothing need be done. 698 * (We could look in the reassembly queue to see 699 * if the packet was previously fragmented, 700 * but it's not worth the time; just let them time out.) 701 */ 702 if (ip->ip_off & ~(IP_DF|IP_RF)) { 703 /* 704 * Look for queue of fragments 705 * of this datagram. 706 */ 707 IPQ_LOCK(); 708 for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next) 709 if (ip->ip_id == fp->ipq_id && 710 in_hosteq(ip->ip_src, fp->ipq_src) && 711 in_hosteq(ip->ip_dst, fp->ipq_dst) && 712 ip->ip_p == fp->ipq_p) 713 goto found; 714 fp = 0; 715 found: 716 717 /* 718 * Adjust ip_len to not reflect header, 719 * set ipqe_mff if more fragments are expected, 720 * convert offset of this to bytes. 721 */ 722 ip->ip_len -= hlen; 723 mff = (ip->ip_off & IP_MF) != 0; 724 if (mff) { 725 /* 726 * Make sure that fragments have a data length 727 * that's a non-zero multiple of 8 bytes. 728 */ 729 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 730 ipstat.ips_badfrags++; 731 IPQ_UNLOCK(); 732 goto bad; 733 } 734 } 735 ip->ip_off <<= 3; 736 737 /* 738 * If datagram marked as having more fragments 739 * or if this is not the first fragment, 740 * attempt reassembly; if it succeeds, proceed. 741 */ 742 if (mff || ip->ip_off) { 743 ipstat.ips_fragments++; 744 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 745 if (ipqe == NULL) { 746 ipstat.ips_rcvmemdrop++; 747 IPQ_UNLOCK(); 748 goto bad; 749 } 750 ipqe->ipqe_mff = mff; 751 ipqe->ipqe_m = m; 752 ipqe->ipqe_ip = ip; 753 m = ip_reass(ipqe, fp); 754 if (m == 0) { 755 IPQ_UNLOCK(); 756 return; 757 } 758 ipstat.ips_reassembled++; 759 ip = mtod(m, struct ip *); 760 hlen = ip->ip_hl << 2; 761 ip->ip_len += hlen; 762 } else 763 if (fp) 764 ip_freef(fp); 765 IPQ_UNLOCK(); 766 } 767 768 #ifdef IPSEC 769 /* 770 * enforce IPsec policy checking if we are seeing last header. 771 * note that we do not visit this with protocols with pcb layer 772 * code - like udp/tcp/raw ip. 773 */ 774 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 775 ipsec4_in_reject(m, NULL)) { 776 ipsecstat.in_polvio++; 777 goto bad; 778 } 779 #endif 780 781 /* 782 * Switch out to protocol's input routine. 783 */ 784 #if IFA_STATS 785 if (ia && ip) 786 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 787 #endif 788 ipstat.ips_delivered++; 789 { 790 int off = hlen, nh = ip->ip_p; 791 792 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 793 return; 794 } 795 bad: 796 m_freem(m); 797 return; 798 799 badcsum: 800 ipstat.ips_badsum++; 801 m_freem(m); 802 } 803 804 /* 805 * Take incoming datagram fragment and try to 806 * reassemble it into whole datagram. If a chain for 807 * reassembly of this datagram already exists, then it 808 * is given as fp; otherwise have to make a chain. 809 */ 810 struct mbuf * 811 ip_reass(ipqe, fp) 812 struct ipqent *ipqe; 813 struct ipq *fp; 814 { 815 struct mbuf *m = ipqe->ipqe_m; 816 struct ipqent *nq, *p, *q; 817 struct ip *ip; 818 struct mbuf *t; 819 int hlen = ipqe->ipqe_ip->ip_hl << 2; 820 int i, next; 821 822 IPQ_LOCK_CHECK(); 823 824 /* 825 * Presence of header sizes in mbufs 826 * would confuse code below. 827 */ 828 m->m_data += hlen; 829 m->m_len -= hlen; 830 831 /* 832 * If first fragment to arrive, create a reassembly queue. 833 */ 834 if (fp == 0) { 835 /* 836 * Enforce upper bound on number of fragmented packets 837 * for which we attempt reassembly; 838 * If maxfrag is 0, never accept fragments. 839 * If maxfrag is -1, accept all fragments without limitation. 840 */ 841 if (ip_maxfragpackets < 0) 842 ; 843 else if (ip_nfragpackets >= ip_maxfragpackets) 844 goto dropfrag; 845 ip_nfragpackets++; 846 MALLOC(fp, struct ipq *, sizeof (struct ipq), 847 M_FTABLE, M_NOWAIT); 848 if (fp == NULL) 849 goto dropfrag; 850 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 851 fp->ipq_ttl = IPFRAGTTL; 852 fp->ipq_p = ipqe->ipqe_ip->ip_p; 853 fp->ipq_id = ipqe->ipqe_ip->ip_id; 854 LIST_INIT(&fp->ipq_fragq); 855 fp->ipq_src = ipqe->ipqe_ip->ip_src; 856 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 857 p = NULL; 858 goto insert; 859 } 860 861 /* 862 * Find a segment which begins after this one does. 863 */ 864 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 865 p = q, q = q->ipqe_q.le_next) 866 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 867 break; 868 869 /* 870 * If there is a preceding segment, it may provide some of 871 * our data already. If so, drop the data from the incoming 872 * segment. If it provides all of our data, drop us. 873 */ 874 if (p != NULL) { 875 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 876 ipqe->ipqe_ip->ip_off; 877 if (i > 0) { 878 if (i >= ipqe->ipqe_ip->ip_len) 879 goto dropfrag; 880 m_adj(ipqe->ipqe_m, i); 881 ipqe->ipqe_ip->ip_off += i; 882 ipqe->ipqe_ip->ip_len -= i; 883 } 884 } 885 886 /* 887 * While we overlap succeeding segments trim them or, 888 * if they are completely covered, dequeue them. 889 */ 890 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 891 q->ipqe_ip->ip_off; q = nq) { 892 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 893 q->ipqe_ip->ip_off; 894 if (i < q->ipqe_ip->ip_len) { 895 q->ipqe_ip->ip_len -= i; 896 q->ipqe_ip->ip_off += i; 897 m_adj(q->ipqe_m, i); 898 break; 899 } 900 nq = q->ipqe_q.le_next; 901 m_freem(q->ipqe_m); 902 LIST_REMOVE(q, ipqe_q); 903 pool_put(&ipqent_pool, q); 904 } 905 906 insert: 907 /* 908 * Stick new segment in its place; 909 * check for complete reassembly. 910 */ 911 if (p == NULL) { 912 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 913 } else { 914 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 915 } 916 next = 0; 917 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 918 p = q, q = q->ipqe_q.le_next) { 919 if (q->ipqe_ip->ip_off != next) 920 return (0); 921 next += q->ipqe_ip->ip_len; 922 } 923 if (p->ipqe_mff) 924 return (0); 925 926 /* 927 * Reassembly is complete. Check for a bogus message size and 928 * concatenate fragments. 929 */ 930 q = fp->ipq_fragq.lh_first; 931 ip = q->ipqe_ip; 932 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 933 ipstat.ips_toolong++; 934 ip_freef(fp); 935 return (0); 936 } 937 m = q->ipqe_m; 938 t = m->m_next; 939 m->m_next = 0; 940 m_cat(m, t); 941 nq = q->ipqe_q.le_next; 942 pool_put(&ipqent_pool, q); 943 for (q = nq; q != NULL; q = nq) { 944 t = q->ipqe_m; 945 nq = q->ipqe_q.le_next; 946 pool_put(&ipqent_pool, q); 947 m_cat(m, t); 948 } 949 950 /* 951 * Create header for new ip packet by 952 * modifying header of first packet; 953 * dequeue and discard fragment reassembly header. 954 * Make header visible. 955 */ 956 ip->ip_len = next; 957 ip->ip_src = fp->ipq_src; 958 ip->ip_dst = fp->ipq_dst; 959 LIST_REMOVE(fp, ipq_q); 960 FREE(fp, M_FTABLE); 961 ip_nfragpackets--; 962 m->m_len += (ip->ip_hl << 2); 963 m->m_data -= (ip->ip_hl << 2); 964 /* some debugging cruft by sklower, below, will go away soon */ 965 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 966 int plen = 0; 967 for (t = m; t; t = t->m_next) 968 plen += t->m_len; 969 m->m_pkthdr.len = plen; 970 } 971 return (m); 972 973 dropfrag: 974 ipstat.ips_fragdropped++; 975 m_freem(m); 976 pool_put(&ipqent_pool, ipqe); 977 return (0); 978 } 979 980 /* 981 * Free a fragment reassembly header and all 982 * associated datagrams. 983 */ 984 void 985 ip_freef(fp) 986 struct ipq *fp; 987 { 988 struct ipqent *q, *p; 989 990 IPQ_LOCK_CHECK(); 991 992 for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { 993 p = q->ipqe_q.le_next; 994 m_freem(q->ipqe_m); 995 LIST_REMOVE(q, ipqe_q); 996 pool_put(&ipqent_pool, q); 997 } 998 LIST_REMOVE(fp, ipq_q); 999 FREE(fp, M_FTABLE); 1000 ip_nfragpackets--; 1001 } 1002 1003 /* 1004 * IP timer processing; 1005 * if a timer expires on a reassembly 1006 * queue, discard it. 1007 */ 1008 void 1009 ip_slowtimo() 1010 { 1011 struct ipq *fp, *nfp; 1012 int s = splsoftnet(); 1013 1014 IPQ_LOCK(); 1015 for (fp = ipq.lh_first; fp != NULL; fp = nfp) { 1016 nfp = fp->ipq_q.le_next; 1017 if (--fp->ipq_ttl == 0) { 1018 ipstat.ips_fragtimeout++; 1019 ip_freef(fp); 1020 } 1021 } 1022 /* 1023 * If we are over the maximum number of fragments 1024 * (due to the limit being lowered), drain off 1025 * enough to get down to the new limit. 1026 */ 1027 if (ip_maxfragpackets < 0) 1028 ; 1029 else { 1030 while (ip_nfragpackets > ip_maxfragpackets && ipq.lh_first) 1031 ip_freef(ipq.lh_first); 1032 } 1033 IPQ_UNLOCK(); 1034 #ifdef GATEWAY 1035 ipflow_slowtimo(); 1036 #endif 1037 splx(s); 1038 } 1039 1040 /* 1041 * Drain off all datagram fragments. 1042 */ 1043 void 1044 ip_drain() 1045 { 1046 1047 /* 1048 * We may be called from a device's interrupt context. If 1049 * the ipq is already busy, just bail out now. 1050 */ 1051 if (ipq_lock_try() == 0) 1052 return; 1053 1054 while (ipq.lh_first != NULL) { 1055 ipstat.ips_fragdropped++; 1056 ip_freef(ipq.lh_first); 1057 } 1058 1059 IPQ_UNLOCK(); 1060 } 1061 1062 /* 1063 * Do option processing on a datagram, 1064 * possibly discarding it if bad options are encountered, 1065 * or forwarding it if source-routed. 1066 * Returns 1 if packet has been forwarded/freed, 1067 * 0 if the packet should be processed further. 1068 */ 1069 int 1070 ip_dooptions(m) 1071 struct mbuf *m; 1072 { 1073 struct ip *ip = mtod(m, struct ip *); 1074 u_char *cp, *cp0; 1075 struct ip_timestamp *ipt; 1076 struct in_ifaddr *ia; 1077 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1078 struct in_addr dst; 1079 n_time ntime; 1080 1081 dst = ip->ip_dst; 1082 cp = (u_char *)(ip + 1); 1083 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1084 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1085 opt = cp[IPOPT_OPTVAL]; 1086 if (opt == IPOPT_EOL) 1087 break; 1088 if (opt == IPOPT_NOP) 1089 optlen = 1; 1090 else { 1091 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1092 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1093 goto bad; 1094 } 1095 optlen = cp[IPOPT_OLEN]; 1096 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1097 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1098 goto bad; 1099 } 1100 } 1101 switch (opt) { 1102 1103 default: 1104 break; 1105 1106 /* 1107 * Source routing with record. 1108 * Find interface with current destination address. 1109 * If none on this machine then drop if strictly routed, 1110 * or do nothing if loosely routed. 1111 * Record interface address and bring up next address 1112 * component. If strictly routed make sure next 1113 * address is on directly accessible net. 1114 */ 1115 case IPOPT_LSRR: 1116 case IPOPT_SSRR: 1117 if (ip_allowsrcrt == 0) { 1118 type = ICMP_UNREACH; 1119 code = ICMP_UNREACH_NET_PROHIB; 1120 goto bad; 1121 } 1122 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1123 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1124 goto bad; 1125 } 1126 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1127 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1128 goto bad; 1129 } 1130 ipaddr.sin_addr = ip->ip_dst; 1131 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1132 if (ia == 0) { 1133 if (opt == IPOPT_SSRR) { 1134 type = ICMP_UNREACH; 1135 code = ICMP_UNREACH_SRCFAIL; 1136 goto bad; 1137 } 1138 /* 1139 * Loose routing, and not at next destination 1140 * yet; nothing to do except forward. 1141 */ 1142 break; 1143 } 1144 off--; /* 0 origin */ 1145 if ((off + sizeof(struct in_addr)) > optlen) { 1146 /* 1147 * End of source route. Should be for us. 1148 */ 1149 save_rte(cp, ip->ip_src); 1150 break; 1151 } 1152 /* 1153 * locate outgoing interface 1154 */ 1155 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1156 sizeof(ipaddr.sin_addr)); 1157 if (opt == IPOPT_SSRR) 1158 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1159 else 1160 ia = ip_rtaddr(ipaddr.sin_addr); 1161 if (ia == 0) { 1162 type = ICMP_UNREACH; 1163 code = ICMP_UNREACH_SRCFAIL; 1164 goto bad; 1165 } 1166 ip->ip_dst = ipaddr.sin_addr; 1167 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1168 (caddr_t)(cp + off), sizeof(struct in_addr)); 1169 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1170 /* 1171 * Let ip_intr's mcast routing check handle mcast pkts 1172 */ 1173 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1174 break; 1175 1176 case IPOPT_RR: 1177 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1178 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1179 goto bad; 1180 } 1181 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1182 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1183 goto bad; 1184 } 1185 /* 1186 * If no space remains, ignore. 1187 */ 1188 off--; /* 0 origin */ 1189 if ((off + sizeof(struct in_addr)) > optlen) 1190 break; 1191 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1192 sizeof(ipaddr.sin_addr)); 1193 /* 1194 * locate outgoing interface; if we're the destination, 1195 * use the incoming interface (should be same). 1196 */ 1197 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1198 == NULL && 1199 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1200 type = ICMP_UNREACH; 1201 code = ICMP_UNREACH_HOST; 1202 goto bad; 1203 } 1204 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1205 (caddr_t)(cp + off), sizeof(struct in_addr)); 1206 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1207 break; 1208 1209 case IPOPT_TS: 1210 code = cp - (u_char *)ip; 1211 ipt = (struct ip_timestamp *)cp; 1212 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1213 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1214 goto bad; 1215 } 1216 if (ipt->ipt_ptr < 5) { 1217 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1218 goto bad; 1219 } 1220 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1221 if (++ipt->ipt_oflw == 0) { 1222 code = (u_char *)&ipt->ipt_ptr - 1223 (u_char *)ip; 1224 goto bad; 1225 } 1226 break; 1227 } 1228 cp0 = (cp + ipt->ipt_ptr - 1); 1229 switch (ipt->ipt_flg) { 1230 1231 case IPOPT_TS_TSONLY: 1232 break; 1233 1234 case IPOPT_TS_TSANDADDR: 1235 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1236 sizeof(struct in_addr) > ipt->ipt_len) { 1237 code = (u_char *)&ipt->ipt_ptr - 1238 (u_char *)ip; 1239 goto bad; 1240 } 1241 ipaddr.sin_addr = dst; 1242 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1243 m->m_pkthdr.rcvif)); 1244 if (ia == 0) 1245 continue; 1246 bcopy(&ia->ia_addr.sin_addr, 1247 cp0, sizeof(struct in_addr)); 1248 ipt->ipt_ptr += sizeof(struct in_addr); 1249 break; 1250 1251 case IPOPT_TS_PRESPEC: 1252 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1253 sizeof(struct in_addr) > ipt->ipt_len) { 1254 code = (u_char *)&ipt->ipt_ptr - 1255 (u_char *)ip; 1256 goto bad; 1257 } 1258 bcopy(cp0, &ipaddr.sin_addr, 1259 sizeof(struct in_addr)); 1260 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1261 == NULL) 1262 continue; 1263 ipt->ipt_ptr += sizeof(struct in_addr); 1264 break; 1265 1266 default: 1267 /* XXX can't take &ipt->ipt_flg */ 1268 code = (u_char *)&ipt->ipt_ptr - 1269 (u_char *)ip + 1; 1270 goto bad; 1271 } 1272 ntime = iptime(); 1273 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1274 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1275 sizeof(n_time)); 1276 ipt->ipt_ptr += sizeof(n_time); 1277 } 1278 } 1279 if (forward) { 1280 if (ip_forwsrcrt == 0) { 1281 type = ICMP_UNREACH; 1282 code = ICMP_UNREACH_SRCFAIL; 1283 goto bad; 1284 } 1285 ip_forward(m, 1); 1286 return (1); 1287 } 1288 return (0); 1289 bad: 1290 icmp_error(m, type, code, 0, 0); 1291 ipstat.ips_badoptions++; 1292 return (1); 1293 } 1294 1295 /* 1296 * Given address of next destination (final or next hop), 1297 * return internet address info of interface to be used to get there. 1298 */ 1299 struct in_ifaddr * 1300 ip_rtaddr(dst) 1301 struct in_addr dst; 1302 { 1303 struct sockaddr_in *sin; 1304 1305 sin = satosin(&ipforward_rt.ro_dst); 1306 1307 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1308 if (ipforward_rt.ro_rt) { 1309 RTFREE(ipforward_rt.ro_rt); 1310 ipforward_rt.ro_rt = 0; 1311 } 1312 sin->sin_family = AF_INET; 1313 sin->sin_len = sizeof(*sin); 1314 sin->sin_addr = dst; 1315 1316 rtalloc(&ipforward_rt); 1317 } 1318 if (ipforward_rt.ro_rt == 0) 1319 return ((struct in_ifaddr *)0); 1320 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1321 } 1322 1323 /* 1324 * Save incoming source route for use in replies, 1325 * to be picked up later by ip_srcroute if the receiver is interested. 1326 */ 1327 void 1328 save_rte(option, dst) 1329 u_char *option; 1330 struct in_addr dst; 1331 { 1332 unsigned olen; 1333 1334 olen = option[IPOPT_OLEN]; 1335 #ifdef DIAGNOSTIC 1336 if (ipprintfs) 1337 printf("save_rte: olen %d\n", olen); 1338 #endif /* 0 */ 1339 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1340 return; 1341 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1342 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1343 ip_srcrt.dst = dst; 1344 } 1345 1346 /* 1347 * Retrieve incoming source route for use in replies, 1348 * in the same form used by setsockopt. 1349 * The first hop is placed before the options, will be removed later. 1350 */ 1351 struct mbuf * 1352 ip_srcroute() 1353 { 1354 struct in_addr *p, *q; 1355 struct mbuf *m; 1356 1357 if (ip_nhops == 0) 1358 return ((struct mbuf *)0); 1359 m = m_get(M_DONTWAIT, MT_SOOPTS); 1360 if (m == 0) 1361 return ((struct mbuf *)0); 1362 1363 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1364 1365 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1366 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1367 OPTSIZ; 1368 #ifdef DIAGNOSTIC 1369 if (ipprintfs) 1370 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1371 #endif 1372 1373 /* 1374 * First save first hop for return route 1375 */ 1376 p = &ip_srcrt.route[ip_nhops - 1]; 1377 *(mtod(m, struct in_addr *)) = *p--; 1378 #ifdef DIAGNOSTIC 1379 if (ipprintfs) 1380 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1381 #endif 1382 1383 /* 1384 * Copy option fields and padding (nop) to mbuf. 1385 */ 1386 ip_srcrt.nop = IPOPT_NOP; 1387 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1388 bcopy((caddr_t)&ip_srcrt.nop, 1389 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1390 q = (struct in_addr *)(mtod(m, caddr_t) + 1391 sizeof(struct in_addr) + OPTSIZ); 1392 #undef OPTSIZ 1393 /* 1394 * Record return path as an IP source route, 1395 * reversing the path (pointers are now aligned). 1396 */ 1397 while (p >= ip_srcrt.route) { 1398 #ifdef DIAGNOSTIC 1399 if (ipprintfs) 1400 printf(" %x", ntohl(q->s_addr)); 1401 #endif 1402 *q++ = *p--; 1403 } 1404 /* 1405 * Last hop goes to final destination. 1406 */ 1407 *q = ip_srcrt.dst; 1408 #ifdef DIAGNOSTIC 1409 if (ipprintfs) 1410 printf(" %x\n", ntohl(q->s_addr)); 1411 #endif 1412 return (m); 1413 } 1414 1415 /* 1416 * Strip out IP options, at higher 1417 * level protocol in the kernel. 1418 * Second argument is buffer to which options 1419 * will be moved, and return value is their length. 1420 * XXX should be deleted; last arg currently ignored. 1421 */ 1422 void 1423 ip_stripoptions(m, mopt) 1424 struct mbuf *m; 1425 struct mbuf *mopt; 1426 { 1427 int i; 1428 struct ip *ip = mtod(m, struct ip *); 1429 caddr_t opts; 1430 int olen; 1431 1432 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1433 opts = (caddr_t)(ip + 1); 1434 i = m->m_len - (sizeof (struct ip) + olen); 1435 bcopy(opts + olen, opts, (unsigned)i); 1436 m->m_len -= olen; 1437 if (m->m_flags & M_PKTHDR) 1438 m->m_pkthdr.len -= olen; 1439 ip->ip_len -= olen; 1440 ip->ip_hl = sizeof (struct ip) >> 2; 1441 } 1442 1443 int inetctlerrmap[PRC_NCMDS] = { 1444 0, 0, 0, 0, 1445 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1446 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1447 EMSGSIZE, EHOSTUNREACH, 0, 0, 1448 0, 0, 0, 0, 1449 ENOPROTOOPT 1450 }; 1451 1452 /* 1453 * Forward a packet. If some error occurs return the sender 1454 * an icmp packet. Note we can't always generate a meaningful 1455 * icmp message because icmp doesn't have a large enough repertoire 1456 * of codes and types. 1457 * 1458 * If not forwarding, just drop the packet. This could be confusing 1459 * if ipforwarding was zero but some routing protocol was advancing 1460 * us as a gateway to somewhere. However, we must let the routing 1461 * protocol deal with that. 1462 * 1463 * The srcrt parameter indicates whether the packet is being forwarded 1464 * via a source route. 1465 */ 1466 void 1467 ip_forward(m, srcrt) 1468 struct mbuf *m; 1469 int srcrt; 1470 { 1471 struct ip *ip = mtod(m, struct ip *); 1472 struct sockaddr_in *sin; 1473 struct rtentry *rt; 1474 int error, type = 0, code = 0; 1475 struct mbuf *mcopy; 1476 n_long dest; 1477 struct ifnet *destifp; 1478 #ifdef IPSEC 1479 struct ifnet dummyifp; 1480 #endif 1481 1482 /* 1483 * Clear any in-bound checksum flags for this packet. 1484 */ 1485 m->m_pkthdr.csum_flags = 0; 1486 1487 dest = 0; 1488 #ifdef DIAGNOSTIC 1489 if (ipprintfs) 1490 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1491 ntohl(ip->ip_src.s_addr), 1492 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1493 #endif 1494 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1495 ipstat.ips_cantforward++; 1496 m_freem(m); 1497 return; 1498 } 1499 if (ip->ip_ttl <= IPTTLDEC) { 1500 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1501 return; 1502 } 1503 ip->ip_ttl -= IPTTLDEC; 1504 1505 sin = satosin(&ipforward_rt.ro_dst); 1506 if ((rt = ipforward_rt.ro_rt) == 0 || 1507 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1508 if (ipforward_rt.ro_rt) { 1509 RTFREE(ipforward_rt.ro_rt); 1510 ipforward_rt.ro_rt = 0; 1511 } 1512 sin->sin_family = AF_INET; 1513 sin->sin_len = sizeof(struct sockaddr_in); 1514 sin->sin_addr = ip->ip_dst; 1515 1516 rtalloc(&ipforward_rt); 1517 if (ipforward_rt.ro_rt == 0) { 1518 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1519 return; 1520 } 1521 rt = ipforward_rt.ro_rt; 1522 } 1523 1524 /* 1525 * Save at most 68 bytes of the packet in case 1526 * we need to generate an ICMP message to the src. 1527 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1528 */ 1529 mcopy = m_copym(m, 0, imin((int)ip->ip_len, 68), M_DONTWAIT); 1530 if (mcopy) 1531 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1532 1533 /* 1534 * If forwarding packet using same interface that it came in on, 1535 * perhaps should send a redirect to sender to shortcut a hop. 1536 * Only send redirect if source is sending directly to us, 1537 * and if packet was not source routed (or has any options). 1538 * Also, don't send redirect if forwarding using a default route 1539 * or a route modified by a redirect. 1540 */ 1541 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1542 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1543 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1544 ipsendredirects && !srcrt) { 1545 if (rt->rt_ifa && 1546 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1547 ifatoia(rt->rt_ifa)->ia_subnet) { 1548 if (rt->rt_flags & RTF_GATEWAY) 1549 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1550 else 1551 dest = ip->ip_dst.s_addr; 1552 /* 1553 * Router requirements says to only send host 1554 * redirects. 1555 */ 1556 type = ICMP_REDIRECT; 1557 code = ICMP_REDIRECT_HOST; 1558 #ifdef DIAGNOSTIC 1559 if (ipprintfs) 1560 printf("redirect (%d) to %x\n", code, 1561 (u_int32_t)dest); 1562 #endif 1563 } 1564 } 1565 1566 #ifdef IPSEC 1567 /* Don't lookup socket in forwarding case */ 1568 (void)ipsec_setsocket(m, NULL); 1569 #endif 1570 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1571 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1572 if (error) 1573 ipstat.ips_cantforward++; 1574 else { 1575 ipstat.ips_forward++; 1576 if (type) 1577 ipstat.ips_redirectsent++; 1578 else { 1579 if (mcopy) { 1580 #ifdef GATEWAY 1581 if (mcopy->m_flags & M_CANFASTFWD) 1582 ipflow_create(&ipforward_rt, mcopy); 1583 #endif 1584 m_freem(mcopy); 1585 } 1586 return; 1587 } 1588 } 1589 if (mcopy == NULL) 1590 return; 1591 destifp = NULL; 1592 1593 switch (error) { 1594 1595 case 0: /* forwarded, but need redirect */ 1596 /* type, code set above */ 1597 break; 1598 1599 case ENETUNREACH: /* shouldn't happen, checked above */ 1600 case EHOSTUNREACH: 1601 case ENETDOWN: 1602 case EHOSTDOWN: 1603 default: 1604 type = ICMP_UNREACH; 1605 code = ICMP_UNREACH_HOST; 1606 break; 1607 1608 case EMSGSIZE: 1609 type = ICMP_UNREACH; 1610 code = ICMP_UNREACH_NEEDFRAG; 1611 #ifndef IPSEC 1612 if (ipforward_rt.ro_rt) 1613 destifp = ipforward_rt.ro_rt->rt_ifp; 1614 #else 1615 /* 1616 * If the packet is routed over IPsec tunnel, tell the 1617 * originator the tunnel MTU. 1618 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1619 * XXX quickhack!!! 1620 */ 1621 if (ipforward_rt.ro_rt) { 1622 struct secpolicy *sp; 1623 int ipsecerror; 1624 size_t ipsechdr; 1625 struct route *ro; 1626 1627 sp = ipsec4_getpolicybyaddr(mcopy, 1628 IPSEC_DIR_OUTBOUND, 1629 IP_FORWARDING, 1630 &ipsecerror); 1631 1632 if (sp == NULL) 1633 destifp = ipforward_rt.ro_rt->rt_ifp; 1634 else { 1635 /* count IPsec header size */ 1636 ipsechdr = ipsec4_hdrsiz(mcopy, 1637 IPSEC_DIR_OUTBOUND, 1638 NULL); 1639 1640 /* 1641 * find the correct route for outer IPv4 1642 * header, compute tunnel MTU. 1643 * 1644 * XXX BUG ALERT 1645 * The "dummyifp" code relies upon the fact 1646 * that icmp_error() touches only ifp->if_mtu. 1647 */ 1648 /*XXX*/ 1649 destifp = NULL; 1650 if (sp->req != NULL 1651 && sp->req->sav != NULL 1652 && sp->req->sav->sah != NULL) { 1653 ro = &sp->req->sav->sah->sa_route; 1654 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1655 dummyifp.if_mtu = 1656 ro->ro_rt->rt_ifp->if_mtu; 1657 dummyifp.if_mtu -= ipsechdr; 1658 destifp = &dummyifp; 1659 } 1660 } 1661 1662 key_freesp(sp); 1663 } 1664 } 1665 #endif /*IPSEC*/ 1666 ipstat.ips_cantfrag++; 1667 break; 1668 1669 case ENOBUFS: 1670 type = ICMP_SOURCEQUENCH; 1671 code = 0; 1672 break; 1673 } 1674 icmp_error(mcopy, type, code, dest, destifp); 1675 } 1676 1677 void 1678 ip_savecontrol(inp, mp, ip, m) 1679 struct inpcb *inp; 1680 struct mbuf **mp; 1681 struct ip *ip; 1682 struct mbuf *m; 1683 { 1684 1685 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1686 struct timeval tv; 1687 1688 microtime(&tv); 1689 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1690 SCM_TIMESTAMP, SOL_SOCKET); 1691 if (*mp) 1692 mp = &(*mp)->m_next; 1693 } 1694 if (inp->inp_flags & INP_RECVDSTADDR) { 1695 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1696 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1697 if (*mp) 1698 mp = &(*mp)->m_next; 1699 } 1700 #ifdef notyet 1701 /* 1702 * XXX 1703 * Moving these out of udp_input() made them even more broken 1704 * than they already were. 1705 * - fenner@parc.xerox.com 1706 */ 1707 /* options were tossed already */ 1708 if (inp->inp_flags & INP_RECVOPTS) { 1709 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1710 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1711 if (*mp) 1712 mp = &(*mp)->m_next; 1713 } 1714 /* ip_srcroute doesn't do what we want here, need to fix */ 1715 if (inp->inp_flags & INP_RECVRETOPTS) { 1716 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1717 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1718 if (*mp) 1719 mp = &(*mp)->m_next; 1720 } 1721 #endif 1722 if (inp->inp_flags & INP_RECVIF) { 1723 struct sockaddr_dl sdl; 1724 1725 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1726 sdl.sdl_family = AF_LINK; 1727 sdl.sdl_index = m->m_pkthdr.rcvif ? 1728 m->m_pkthdr.rcvif->if_index : 0; 1729 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1730 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1731 IP_RECVIF, IPPROTO_IP); 1732 if (*mp) 1733 mp = &(*mp)->m_next; 1734 } 1735 } 1736 1737 int 1738 ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1739 int *name; 1740 u_int namelen; 1741 void *oldp; 1742 size_t *oldlenp; 1743 void *newp; 1744 size_t newlen; 1745 { 1746 extern int subnetsarelocal, hostzeroisbroadcast; 1747 1748 int error, old; 1749 1750 /* All sysctl names at this level are terminal. */ 1751 if (namelen != 1) 1752 return (ENOTDIR); 1753 1754 switch (name[0]) { 1755 case IPCTL_FORWARDING: 1756 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1757 case IPCTL_SENDREDIRECTS: 1758 return (sysctl_int(oldp, oldlenp, newp, newlen, 1759 &ipsendredirects)); 1760 case IPCTL_DEFTTL: 1761 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1762 #ifdef notyet 1763 case IPCTL_DEFMTU: 1764 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1765 #endif 1766 case IPCTL_FORWSRCRT: 1767 /* Don't allow this to change in a secure environment. */ 1768 if (securelevel > 0) 1769 return (sysctl_rdint(oldp, oldlenp, newp, 1770 ip_forwsrcrt)); 1771 else 1772 return (sysctl_int(oldp, oldlenp, newp, newlen, 1773 &ip_forwsrcrt)); 1774 case IPCTL_DIRECTEDBCAST: 1775 return (sysctl_int(oldp, oldlenp, newp, newlen, 1776 &ip_directedbcast)); 1777 case IPCTL_ALLOWSRCRT: 1778 return (sysctl_int(oldp, oldlenp, newp, newlen, 1779 &ip_allowsrcrt)); 1780 case IPCTL_SUBNETSARELOCAL: 1781 return (sysctl_int(oldp, oldlenp, newp, newlen, 1782 &subnetsarelocal)); 1783 case IPCTL_MTUDISC: 1784 error = sysctl_int(oldp, oldlenp, newp, newlen, 1785 &ip_mtudisc); 1786 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1787 ip_mtudisc_timeout_q = 1788 rt_timer_queue_create(ip_mtudisc_timeout); 1789 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1790 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1791 ip_mtudisc_timeout_q = NULL; 1792 } 1793 return error; 1794 case IPCTL_ANONPORTMIN: 1795 old = anonportmin; 1796 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1797 if (anonportmin >= anonportmax || anonportmin < 0 1798 || anonportmin > 65535 1799 #ifndef IPNOPRIVPORTS 1800 || anonportmin < IPPORT_RESERVED 1801 #endif 1802 ) { 1803 anonportmin = old; 1804 return (EINVAL); 1805 } 1806 return (error); 1807 case IPCTL_ANONPORTMAX: 1808 old = anonportmax; 1809 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1810 if (anonportmin >= anonportmax || anonportmax < 0 1811 || anonportmax > 65535 1812 #ifndef IPNOPRIVPORTS 1813 || anonportmax < IPPORT_RESERVED 1814 #endif 1815 ) { 1816 anonportmax = old; 1817 return (EINVAL); 1818 } 1819 return (error); 1820 case IPCTL_MTUDISCTIMEOUT: 1821 error = sysctl_int(oldp, oldlenp, newp, newlen, 1822 &ip_mtudisc_timeout); 1823 if (ip_mtudisc_timeout_q != NULL) 1824 rt_timer_queue_change(ip_mtudisc_timeout_q, 1825 ip_mtudisc_timeout); 1826 return (error); 1827 #ifdef GATEWAY 1828 case IPCTL_MAXFLOWS: 1829 { 1830 int s; 1831 1832 error = sysctl_int(oldp, oldlenp, newp, newlen, 1833 &ip_maxflows); 1834 s = splsoftnet(); 1835 ipflow_reap(0); 1836 splx(s); 1837 return (error); 1838 } 1839 #endif 1840 case IPCTL_HOSTZEROBROADCAST: 1841 return (sysctl_int(oldp, oldlenp, newp, newlen, 1842 &hostzeroisbroadcast)); 1843 #if NGIF > 0 1844 case IPCTL_GIF_TTL: 1845 return(sysctl_int(oldp, oldlenp, newp, newlen, 1846 &ip_gif_ttl)); 1847 #endif 1848 1849 #ifndef IPNOPRIVPORTS 1850 case IPCTL_LOWPORTMIN: 1851 old = lowportmin; 1852 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmin); 1853 if (lowportmin >= lowportmax 1854 || lowportmin > IPPORT_RESERVEDMAX 1855 || lowportmin < IPPORT_RESERVEDMIN 1856 ) { 1857 lowportmin = old; 1858 return (EINVAL); 1859 } 1860 return (error); 1861 case IPCTL_LOWPORTMAX: 1862 old = lowportmax; 1863 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmax); 1864 if (lowportmin >= lowportmax 1865 || lowportmax > IPPORT_RESERVEDMAX 1866 || lowportmax < IPPORT_RESERVEDMIN 1867 ) { 1868 lowportmax = old; 1869 return (EINVAL); 1870 } 1871 return (error); 1872 #endif 1873 1874 case IPCTL_MAXFRAGPACKETS: 1875 return (sysctl_int(oldp, oldlenp, newp, newlen, 1876 &ip_maxfragpackets)); 1877 1878 default: 1879 return (EOPNOTSUPP); 1880 } 1881 /* NOTREACHED */ 1882 } 1883