1 /* $NetBSD: ip_input.c,v 1.126 2000/12/28 21:40:59 thorpej Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 /*- 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 3. All advertising materials mentioning features or use of this software 49 * must display the following acknowledgement: 50 * This product includes software developed by the NetBSD 51 * Foundation, Inc. and its contributors. 52 * 4. Neither the name of The NetBSD Foundation nor the names of its 53 * contributors may be used to endorse or promote products derived 54 * from this software without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 57 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 58 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 59 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 60 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 61 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 62 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 63 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 64 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 65 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 66 * POSSIBILITY OF SUCH DAMAGE. 67 */ 68 69 /* 70 * Copyright (c) 1982, 1986, 1988, 1993 71 * The Regents of the University of California. All rights reserved. 72 * 73 * Redistribution and use in source and binary forms, with or without 74 * modification, are permitted provided that the following conditions 75 * are met: 76 * 1. Redistributions of source code must retain the above copyright 77 * notice, this list of conditions and the following disclaimer. 78 * 2. Redistributions in binary form must reproduce the above copyright 79 * notice, this list of conditions and the following disclaimer in the 80 * documentation and/or other materials provided with the distribution. 81 * 3. All advertising materials mentioning features or use of this software 82 * must display the following acknowledgement: 83 * This product includes software developed by the University of 84 * California, Berkeley and its contributors. 85 * 4. Neither the name of the University nor the names of its contributors 86 * may be used to endorse or promote products derived from this software 87 * without specific prior written permission. 88 * 89 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 90 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 91 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 92 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 93 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 94 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 95 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 96 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 97 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 98 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 99 * SUCH DAMAGE. 100 * 101 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 102 */ 103 104 #include "opt_gateway.h" 105 #include "opt_pfil_hooks.h" 106 #include "opt_ipsec.h" 107 #include "opt_mrouting.h" 108 109 #include <sys/param.h> 110 #include <sys/systm.h> 111 #include <sys/malloc.h> 112 #include <sys/mbuf.h> 113 #include <sys/domain.h> 114 #include <sys/protosw.h> 115 #include <sys/socket.h> 116 #include <sys/socketvar.h> 117 #include <sys/errno.h> 118 #include <sys/time.h> 119 #include <sys/kernel.h> 120 #include <sys/proc.h> 121 #include <sys/pool.h> 122 123 #include <uvm/uvm_extern.h> 124 125 #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" 142 143 #ifdef MROUTING 144 #include <netinet/ip_mroute.h> 145 #endif 146 147 #ifdef IPSEC 148 #include <netinet6/ipsec.h> 149 #include <netkey/key.h> 150 #endif 151 152 #ifndef IPFORWARDING 153 #ifdef GATEWAY 154 #define IPFORWARDING 1 /* forward IP packets not for us */ 155 #else /* GATEWAY */ 156 #define IPFORWARDING 0 /* don't forward IP packets not for us */ 157 #endif /* GATEWAY */ 158 #endif /* IPFORWARDING */ 159 #ifndef IPSENDREDIRECTS 160 #define IPSENDREDIRECTS 1 161 #endif 162 #ifndef IPFORWSRCRT 163 #define IPFORWSRCRT 1 /* forward source-routed packets */ 164 #endif 165 #ifndef IPALLOWSRCRT 166 #define IPALLOWSRCRT 1 /* allow source-routed packets */ 167 #endif 168 #ifndef IPMTUDISC 169 #define IPMTUDISC 0 170 #endif 171 #ifndef IPMTUDISCTIMEOUT 172 #define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 173 #endif 174 175 /* 176 * Note: DIRECTED_BROADCAST is handled this way so that previous 177 * configuration using this option will Just Work. 178 */ 179 #ifndef IPDIRECTEDBCAST 180 #ifdef DIRECTED_BROADCAST 181 #define IPDIRECTEDBCAST 1 182 #else 183 #define IPDIRECTEDBCAST 0 184 #endif /* DIRECTED_BROADCAST */ 185 #endif /* IPDIRECTEDBCAST */ 186 int ipforwarding = IPFORWARDING; 187 int ipsendredirects = IPSENDREDIRECTS; 188 int ip_defttl = IPDEFTTL; 189 int ip_forwsrcrt = IPFORWSRCRT; 190 int ip_directedbcast = IPDIRECTEDBCAST; 191 int ip_allowsrcrt = IPALLOWSRCRT; 192 int ip_mtudisc = IPMTUDISC; 193 u_int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 194 #ifdef DIAGNOSTIC 195 int ipprintfs = 0; 196 #endif 197 198 struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 199 200 extern struct domain inetdomain; 201 int ipqmaxlen = IFQ_MAXLEN; 202 struct in_ifaddrhead in_ifaddr; 203 struct in_ifaddrhashhead *in_ifaddrhashtbl; 204 struct ifqueue ipintrq; 205 struct ipstat ipstat; 206 u_int16_t ip_id; 207 208 #ifdef PFIL_HOOKS 209 struct pfil_head inet_pfil_hook; 210 #endif 211 212 struct ipqhead ipq; 213 int ipq_locked; 214 215 static __inline int ipq_lock_try __P((void)); 216 static __inline void ipq_unlock __P((void)); 217 218 static __inline int 219 ipq_lock_try() 220 { 221 int s; 222 223 s = splimp(); 224 if (ipq_locked) { 225 splx(s); 226 return (0); 227 } 228 ipq_locked = 1; 229 splx(s); 230 return (1); 231 } 232 233 static __inline void 234 ipq_unlock() 235 { 236 int s; 237 238 s = splimp(); 239 ipq_locked = 0; 240 splx(s); 241 } 242 243 #ifdef DIAGNOSTIC 244 #define IPQ_LOCK() \ 245 do { \ 246 if (ipq_lock_try() == 0) { \ 247 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 248 panic("ipq_lock"); \ 249 } \ 250 } while (0) 251 #define IPQ_LOCK_CHECK() \ 252 do { \ 253 if (ipq_locked == 0) { \ 254 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 255 panic("ipq lock check"); \ 256 } \ 257 } while (0) 258 #else 259 #define IPQ_LOCK() (void) ipq_lock_try() 260 #define IPQ_LOCK_CHECK() /* nothing */ 261 #endif 262 263 #define IPQ_UNLOCK() ipq_unlock() 264 265 struct pool ipqent_pool; 266 267 /* 268 * We need to save the IP options in case a protocol wants to respond 269 * to an incoming packet over the same route if the packet got here 270 * using IP source routing. This allows connection establishment and 271 * maintenance when the remote end is on a network that is not known 272 * to us. 273 */ 274 int ip_nhops = 0; 275 static struct ip_srcrt { 276 struct in_addr dst; /* final destination */ 277 char nop; /* one NOP to align */ 278 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 279 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 280 } ip_srcrt; 281 282 static void save_rte __P((u_char *, struct in_addr)); 283 284 /* 285 * IP initialization: fill in IP protocol switch table. 286 * All protocols not implemented in kernel go to raw IP protocol handler. 287 */ 288 void 289 ip_init() 290 { 291 struct protosw *pr; 292 int i; 293 294 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 295 0, NULL, NULL, M_IPQ); 296 297 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 298 if (pr == 0) 299 panic("ip_init"); 300 for (i = 0; i < IPPROTO_MAX; i++) 301 ip_protox[i] = pr - inetsw; 302 for (pr = inetdomain.dom_protosw; 303 pr < inetdomain.dom_protoswNPROTOSW; pr++) 304 if (pr->pr_domain->dom_family == PF_INET && 305 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 306 ip_protox[pr->pr_protocol] = pr - inetsw; 307 LIST_INIT(&ipq); 308 ip_id = time.tv_sec & 0xffff; 309 ipintrq.ifq_maxlen = ipqmaxlen; 310 TAILQ_INIT(&in_ifaddr); 311 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, 312 M_WAITOK, &in_ifaddrhash); 313 if (ip_mtudisc != 0) 314 ip_mtudisc_timeout_q = 315 rt_timer_queue_create(ip_mtudisc_timeout); 316 #ifdef GATEWAY 317 ipflow_init(); 318 #endif 319 320 #ifdef PFIL_HOOKS 321 /* Register our Packet Filter hook. */ 322 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 323 inet_pfil_hook.ph_af = AF_INET; 324 i = pfil_head_register(&inet_pfil_hook); 325 if (i != 0) 326 printf("ip_init: WARNING: unable to register pfil hook, " 327 "error %d\n", i); 328 #endif /* PFIL_HOOKS */ 329 } 330 331 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 332 struct route ipforward_rt; 333 334 /* 335 * IP software interrupt routine 336 */ 337 void 338 ipintr() 339 { 340 int s; 341 struct mbuf *m; 342 343 while (1) { 344 s = splimp(); 345 IF_DEQUEUE(&ipintrq, m); 346 splx(s); 347 if (m == 0) 348 return; 349 ip_input(m); 350 } 351 } 352 353 /* 354 * Ip input routine. Checksum and byte swap header. If fragmented 355 * try to reassemble. Process options. Pass to next level. 356 */ 357 void 358 ip_input(struct mbuf *m) 359 { 360 struct ip *ip = NULL; 361 struct ipq *fp; 362 struct in_ifaddr *ia; 363 struct ifaddr *ifa; 364 struct ipqent *ipqe; 365 int hlen = 0, mff, len; 366 int downmatch; 367 368 #ifdef DIAGNOSTIC 369 if ((m->m_flags & M_PKTHDR) == 0) 370 panic("ipintr no HDR"); 371 #endif 372 #ifdef IPSEC 373 /* 374 * should the inner packet be considered authentic? 375 * see comment in ah4_input(). 376 */ 377 if (m) { 378 m->m_flags &= ~M_AUTHIPHDR; 379 m->m_flags &= ~M_AUTHIPDGM; 380 } 381 #endif 382 /* 383 * If no IP addresses have been set yet but the interfaces 384 * are receiving, can't do anything with incoming packets yet. 385 */ 386 if (in_ifaddr.tqh_first == 0) 387 goto bad; 388 ipstat.ips_total++; 389 if (m->m_len < sizeof (struct ip) && 390 (m = m_pullup(m, sizeof (struct ip))) == 0) { 391 ipstat.ips_toosmall++; 392 return; 393 } 394 ip = mtod(m, struct ip *); 395 if (ip->ip_v != IPVERSION) { 396 ipstat.ips_badvers++; 397 goto bad; 398 } 399 hlen = ip->ip_hl << 2; 400 if (hlen < sizeof(struct ip)) { /* minimum header length */ 401 ipstat.ips_badhlen++; 402 goto bad; 403 } 404 if (hlen > m->m_len) { 405 if ((m = m_pullup(m, hlen)) == 0) { 406 ipstat.ips_badhlen++; 407 return; 408 } 409 ip = mtod(m, struct ip *); 410 } 411 412 /* 413 * RFC1122: packets with a multicast source address are 414 * not allowed. 415 */ 416 if (IN_MULTICAST(ip->ip_src.s_addr)) { 417 /* XXX stat */ 418 goto bad; 419 } 420 421 if (in_cksum(m, hlen) != 0) { 422 ipstat.ips_badsum++; 423 goto bad; 424 } 425 426 /* Retrieve the packet length. */ 427 len = ntohs(ip->ip_len); 428 429 /* 430 * Check for additional length bogosity 431 */ 432 if (len < hlen) { 433 ipstat.ips_badlen++; 434 goto bad; 435 } 436 437 /* 438 * Check that the amount of data in the buffers 439 * is as at least much as the IP header would have us expect. 440 * Trim mbufs if longer than we expect. 441 * Drop packet if shorter than we expect. 442 */ 443 if (m->m_pkthdr.len < len) { 444 ipstat.ips_tooshort++; 445 goto bad; 446 } 447 if (m->m_pkthdr.len > len) { 448 if (m->m_len == m->m_pkthdr.len) { 449 m->m_len = len; 450 m->m_pkthdr.len = len; 451 } else 452 m_adj(m, len - m->m_pkthdr.len); 453 } 454 455 #ifdef IPSEC 456 /* ipflow (IP fast fowarding) is not compatible with IPsec. */ 457 m->m_flags &= ~M_CANFASTFWD; 458 #else 459 /* 460 * Assume that we can create a fast-forward IP flow entry 461 * based on this packet. 462 */ 463 m->m_flags |= M_CANFASTFWD; 464 #endif 465 466 #ifdef PFIL_HOOKS 467 /* 468 * Run through list of hooks for input packets. If there are any 469 * filters which require that additional packets in the flow are 470 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 471 * Note that filters must _never_ set this flag, as another filter 472 * in the list may have previously cleared it. 473 */ 474 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 475 PFIL_IN) != 0) 476 return; 477 if (m == NULL) 478 return; 479 ip = mtod(m, struct ip *); 480 #endif /* PFIL_HOOKS */ 481 482 #ifdef ALTQ 483 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 484 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 485 /* packet dropped by traffic conditioner */ 486 return; 487 } 488 #endif 489 490 /* 491 * Convert fields to host representation. 492 */ 493 NTOHS(ip->ip_len); 494 NTOHS(ip->ip_off); 495 496 /* 497 * Process options and, if not destined for us, 498 * ship it on. ip_dooptions returns 1 when an 499 * error was detected (causing an icmp message 500 * to be sent and the original packet to be freed). 501 */ 502 ip_nhops = 0; /* for source routed packets */ 503 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 504 return; 505 506 /* 507 * Check our list of addresses, to see if the packet is for us. 508 * 509 * Traditional 4.4BSD did not consult IFF_UP at all. 510 * The behavior here is to treat addresses on !IFF_UP interface 511 * as not mine. 512 */ 513 downmatch = 0; 514 for (ia = IN_IFADDR_HASH(ip->ip_dst.s_addr).lh_first; 515 ia != NULL; 516 ia = ia->ia_hash.le_next) { 517 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 518 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 519 break; 520 else 521 downmatch++; 522 } 523 } 524 if (ia != NULL) 525 goto ours; 526 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 527 for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first; 528 ifa != NULL; ifa = ifa->ifa_list.tqe_next) { 529 if (ifa->ifa_addr->sa_family != AF_INET) continue; 530 ia = ifatoia(ifa); 531 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 532 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 533 /* 534 * Look for all-0's host part (old broadcast addr), 535 * either for subnet or net. 536 */ 537 ip->ip_dst.s_addr == ia->ia_subnet || 538 ip->ip_dst.s_addr == ia->ia_net) 539 goto ours; 540 /* 541 * An interface with IP address zero accepts 542 * all packets that arrive on that interface. 543 */ 544 if (in_nullhost(ia->ia_addr.sin_addr)) 545 goto ours; 546 } 547 } 548 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 549 struct in_multi *inm; 550 #ifdef MROUTING 551 extern struct socket *ip_mrouter; 552 553 if (m->m_flags & M_EXT) { 554 if ((m = m_pullup(m, hlen)) == 0) { 555 ipstat.ips_toosmall++; 556 return; 557 } 558 ip = mtod(m, struct ip *); 559 } 560 561 if (ip_mrouter) { 562 /* 563 * If we are acting as a multicast router, all 564 * incoming multicast packets are passed to the 565 * kernel-level multicast forwarding function. 566 * The packet is returned (relatively) intact; if 567 * ip_mforward() returns a non-zero value, the packet 568 * must be discarded, else it may be accepted below. 569 * 570 * (The IP ident field is put in the same byte order 571 * as expected when ip_mforward() is called from 572 * ip_output().) 573 */ 574 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 575 ipstat.ips_cantforward++; 576 m_freem(m); 577 return; 578 } 579 580 /* 581 * The process-level routing demon needs to receive 582 * all multicast IGMP packets, whether or not this 583 * host belongs to their destination groups. 584 */ 585 if (ip->ip_p == IPPROTO_IGMP) 586 goto ours; 587 ipstat.ips_forward++; 588 } 589 #endif 590 /* 591 * See if we belong to the destination multicast group on the 592 * arrival interface. 593 */ 594 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 595 if (inm == NULL) { 596 ipstat.ips_cantforward++; 597 m_freem(m); 598 return; 599 } 600 goto ours; 601 } 602 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 603 in_nullhost(ip->ip_dst)) 604 goto ours; 605 606 /* 607 * Not for us; forward if possible and desirable. 608 */ 609 if (ipforwarding == 0) { 610 ipstat.ips_cantforward++; 611 m_freem(m); 612 } else { 613 /* 614 * If ip_dst matched any of my address on !IFF_UP interface, 615 * and there's no IFF_UP interface that matches ip_dst, 616 * send icmp unreach. Forwarding it will result in in-kernel 617 * forwarding loop till TTL goes to 0. 618 */ 619 if (downmatch) { 620 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 621 ipstat.ips_cantforward++; 622 return; 623 } 624 ip_forward(m, 0); 625 } 626 return; 627 628 ours: 629 /* 630 * If offset or IP_MF are set, must reassemble. 631 * Otherwise, nothing need be done. 632 * (We could look in the reassembly queue to see 633 * if the packet was previously fragmented, 634 * but it's not worth the time; just let them time out.) 635 */ 636 if (ip->ip_off & ~(IP_DF|IP_RF)) { 637 /* 638 * Look for queue of fragments 639 * of this datagram. 640 */ 641 IPQ_LOCK(); 642 for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next) 643 if (ip->ip_id == fp->ipq_id && 644 in_hosteq(ip->ip_src, fp->ipq_src) && 645 in_hosteq(ip->ip_dst, fp->ipq_dst) && 646 ip->ip_p == fp->ipq_p) 647 goto found; 648 fp = 0; 649 found: 650 651 /* 652 * Adjust ip_len to not reflect header, 653 * set ipqe_mff if more fragments are expected, 654 * convert offset of this to bytes. 655 */ 656 ip->ip_len -= hlen; 657 mff = (ip->ip_off & IP_MF) != 0; 658 if (mff) { 659 /* 660 * Make sure that fragments have a data length 661 * that's a non-zero multiple of 8 bytes. 662 */ 663 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 664 ipstat.ips_badfrags++; 665 IPQ_UNLOCK(); 666 goto bad; 667 } 668 } 669 ip->ip_off <<= 3; 670 671 /* 672 * If datagram marked as having more fragments 673 * or if this is not the first fragment, 674 * attempt reassembly; if it succeeds, proceed. 675 */ 676 if (mff || ip->ip_off) { 677 ipstat.ips_fragments++; 678 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 679 if (ipqe == NULL) { 680 ipstat.ips_rcvmemdrop++; 681 IPQ_UNLOCK(); 682 goto bad; 683 } 684 ipqe->ipqe_mff = mff; 685 ipqe->ipqe_m = m; 686 ipqe->ipqe_ip = ip; 687 m = ip_reass(ipqe, fp); 688 if (m == 0) { 689 IPQ_UNLOCK(); 690 return; 691 } 692 ipstat.ips_reassembled++; 693 ip = mtod(m, struct ip *); 694 hlen = ip->ip_hl << 2; 695 ip->ip_len += hlen; 696 } else 697 if (fp) 698 ip_freef(fp); 699 IPQ_UNLOCK(); 700 } 701 702 /* 703 * Switch out to protocol's input routine. 704 */ 705 #if IFA_STATS 706 if (ia && ip) 707 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 708 #endif 709 ipstat.ips_delivered++; 710 { 711 int off = hlen, nh = ip->ip_p; 712 713 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 714 return; 715 } 716 bad: 717 m_freem(m); 718 } 719 720 /* 721 * Take incoming datagram fragment and try to 722 * reassemble it into whole datagram. If a chain for 723 * reassembly of this datagram already exists, then it 724 * is given as fp; otherwise have to make a chain. 725 */ 726 struct mbuf * 727 ip_reass(ipqe, fp) 728 struct ipqent *ipqe; 729 struct ipq *fp; 730 { 731 struct mbuf *m = ipqe->ipqe_m; 732 struct ipqent *nq, *p, *q; 733 struct ip *ip; 734 struct mbuf *t; 735 int hlen = ipqe->ipqe_ip->ip_hl << 2; 736 int i, next; 737 738 IPQ_LOCK_CHECK(); 739 740 /* 741 * Presence of header sizes in mbufs 742 * would confuse code below. 743 */ 744 m->m_data += hlen; 745 m->m_len -= hlen; 746 747 /* 748 * If first fragment to arrive, create a reassembly queue. 749 */ 750 if (fp == 0) { 751 MALLOC(fp, struct ipq *, sizeof (struct ipq), 752 M_FTABLE, M_NOWAIT); 753 if (fp == NULL) 754 goto dropfrag; 755 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 756 fp->ipq_ttl = IPFRAGTTL; 757 fp->ipq_p = ipqe->ipqe_ip->ip_p; 758 fp->ipq_id = ipqe->ipqe_ip->ip_id; 759 LIST_INIT(&fp->ipq_fragq); 760 fp->ipq_src = ipqe->ipqe_ip->ip_src; 761 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 762 p = NULL; 763 goto insert; 764 } 765 766 /* 767 * Find a segment which begins after this one does. 768 */ 769 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 770 p = q, q = q->ipqe_q.le_next) 771 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 772 break; 773 774 /* 775 * If there is a preceding segment, it may provide some of 776 * our data already. If so, drop the data from the incoming 777 * segment. If it provides all of our data, drop us. 778 */ 779 if (p != NULL) { 780 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 781 ipqe->ipqe_ip->ip_off; 782 if (i > 0) { 783 if (i >= ipqe->ipqe_ip->ip_len) 784 goto dropfrag; 785 m_adj(ipqe->ipqe_m, i); 786 ipqe->ipqe_ip->ip_off += i; 787 ipqe->ipqe_ip->ip_len -= i; 788 } 789 } 790 791 /* 792 * While we overlap succeeding segments trim them or, 793 * if they are completely covered, dequeue them. 794 */ 795 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 796 q->ipqe_ip->ip_off; q = nq) { 797 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 798 q->ipqe_ip->ip_off; 799 if (i < q->ipqe_ip->ip_len) { 800 q->ipqe_ip->ip_len -= i; 801 q->ipqe_ip->ip_off += i; 802 m_adj(q->ipqe_m, i); 803 break; 804 } 805 nq = q->ipqe_q.le_next; 806 m_freem(q->ipqe_m); 807 LIST_REMOVE(q, ipqe_q); 808 pool_put(&ipqent_pool, q); 809 } 810 811 insert: 812 /* 813 * Stick new segment in its place; 814 * check for complete reassembly. 815 */ 816 if (p == NULL) { 817 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 818 } else { 819 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 820 } 821 next = 0; 822 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 823 p = q, q = q->ipqe_q.le_next) { 824 if (q->ipqe_ip->ip_off != next) 825 return (0); 826 next += q->ipqe_ip->ip_len; 827 } 828 if (p->ipqe_mff) 829 return (0); 830 831 /* 832 * Reassembly is complete. Check for a bogus message size and 833 * concatenate fragments. 834 */ 835 q = fp->ipq_fragq.lh_first; 836 ip = q->ipqe_ip; 837 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 838 ipstat.ips_toolong++; 839 ip_freef(fp); 840 return (0); 841 } 842 m = q->ipqe_m; 843 t = m->m_next; 844 m->m_next = 0; 845 m_cat(m, t); 846 nq = q->ipqe_q.le_next; 847 pool_put(&ipqent_pool, q); 848 for (q = nq; q != NULL; q = nq) { 849 t = q->ipqe_m; 850 nq = q->ipqe_q.le_next; 851 pool_put(&ipqent_pool, q); 852 m_cat(m, t); 853 } 854 855 /* 856 * Create header for new ip packet by 857 * modifying header of first packet; 858 * dequeue and discard fragment reassembly header. 859 * Make header visible. 860 */ 861 ip->ip_len = next; 862 ip->ip_src = fp->ipq_src; 863 ip->ip_dst = fp->ipq_dst; 864 LIST_REMOVE(fp, ipq_q); 865 FREE(fp, M_FTABLE); 866 m->m_len += (ip->ip_hl << 2); 867 m->m_data -= (ip->ip_hl << 2); 868 /* some debugging cruft by sklower, below, will go away soon */ 869 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 870 int plen = 0; 871 for (t = m; t; t = t->m_next) 872 plen += t->m_len; 873 m->m_pkthdr.len = plen; 874 } 875 return (m); 876 877 dropfrag: 878 ipstat.ips_fragdropped++; 879 m_freem(m); 880 pool_put(&ipqent_pool, ipqe); 881 return (0); 882 } 883 884 /* 885 * Free a fragment reassembly header and all 886 * associated datagrams. 887 */ 888 void 889 ip_freef(fp) 890 struct ipq *fp; 891 { 892 struct ipqent *q, *p; 893 894 IPQ_LOCK_CHECK(); 895 896 for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { 897 p = q->ipqe_q.le_next; 898 m_freem(q->ipqe_m); 899 LIST_REMOVE(q, ipqe_q); 900 pool_put(&ipqent_pool, q); 901 } 902 LIST_REMOVE(fp, ipq_q); 903 FREE(fp, M_FTABLE); 904 } 905 906 /* 907 * IP timer processing; 908 * if a timer expires on a reassembly 909 * queue, discard it. 910 */ 911 void 912 ip_slowtimo() 913 { 914 struct ipq *fp, *nfp; 915 int s = splsoftnet(); 916 917 IPQ_LOCK(); 918 for (fp = ipq.lh_first; fp != NULL; fp = nfp) { 919 nfp = fp->ipq_q.le_next; 920 if (--fp->ipq_ttl == 0) { 921 ipstat.ips_fragtimeout++; 922 ip_freef(fp); 923 } 924 } 925 IPQ_UNLOCK(); 926 #ifdef GATEWAY 927 ipflow_slowtimo(); 928 #endif 929 splx(s); 930 } 931 932 /* 933 * Drain off all datagram fragments. 934 */ 935 void 936 ip_drain() 937 { 938 939 /* 940 * We may be called from a device's interrupt context. If 941 * the ipq is already busy, just bail out now. 942 */ 943 if (ipq_lock_try() == 0) 944 return; 945 946 while (ipq.lh_first != NULL) { 947 ipstat.ips_fragdropped++; 948 ip_freef(ipq.lh_first); 949 } 950 951 IPQ_UNLOCK(); 952 } 953 954 /* 955 * Do option processing on a datagram, 956 * possibly discarding it if bad options are encountered, 957 * or forwarding it if source-routed. 958 * Returns 1 if packet has been forwarded/freed, 959 * 0 if the packet should be processed further. 960 */ 961 int 962 ip_dooptions(m) 963 struct mbuf *m; 964 { 965 struct ip *ip = mtod(m, struct ip *); 966 u_char *cp, *cp0; 967 struct ip_timestamp *ipt; 968 struct in_ifaddr *ia; 969 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 970 struct in_addr dst; 971 n_time ntime; 972 973 dst = ip->ip_dst; 974 cp = (u_char *)(ip + 1); 975 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 976 for (; cnt > 0; cnt -= optlen, cp += optlen) { 977 opt = cp[IPOPT_OPTVAL]; 978 if (opt == IPOPT_EOL) 979 break; 980 if (opt == IPOPT_NOP) 981 optlen = 1; 982 else { 983 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 984 code = &cp[IPOPT_OLEN] - (u_char *)ip; 985 goto bad; 986 } 987 optlen = cp[IPOPT_OLEN]; 988 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 989 code = &cp[IPOPT_OLEN] - (u_char *)ip; 990 goto bad; 991 } 992 } 993 switch (opt) { 994 995 default: 996 break; 997 998 /* 999 * Source routing with record. 1000 * Find interface with current destination address. 1001 * If none on this machine then drop if strictly routed, 1002 * or do nothing if loosely routed. 1003 * Record interface address and bring up next address 1004 * component. If strictly routed make sure next 1005 * address is on directly accessible net. 1006 */ 1007 case IPOPT_LSRR: 1008 case IPOPT_SSRR: 1009 if (ip_allowsrcrt == 0) { 1010 type = ICMP_UNREACH; 1011 code = ICMP_UNREACH_NET_PROHIB; 1012 goto bad; 1013 } 1014 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1015 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1016 goto bad; 1017 } 1018 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1019 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1020 goto bad; 1021 } 1022 ipaddr.sin_addr = ip->ip_dst; 1023 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1024 if (ia == 0) { 1025 if (opt == IPOPT_SSRR) { 1026 type = ICMP_UNREACH; 1027 code = ICMP_UNREACH_SRCFAIL; 1028 goto bad; 1029 } 1030 /* 1031 * Loose routing, and not at next destination 1032 * yet; nothing to do except forward. 1033 */ 1034 break; 1035 } 1036 off--; /* 0 origin */ 1037 if ((off + sizeof(struct in_addr)) > optlen) { 1038 /* 1039 * End of source route. Should be for us. 1040 */ 1041 save_rte(cp, ip->ip_src); 1042 break; 1043 } 1044 /* 1045 * locate outgoing interface 1046 */ 1047 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1048 sizeof(ipaddr.sin_addr)); 1049 if (opt == IPOPT_SSRR) 1050 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1051 else 1052 ia = ip_rtaddr(ipaddr.sin_addr); 1053 if (ia == 0) { 1054 type = ICMP_UNREACH; 1055 code = ICMP_UNREACH_SRCFAIL; 1056 goto bad; 1057 } 1058 ip->ip_dst = ipaddr.sin_addr; 1059 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1060 (caddr_t)(cp + off), sizeof(struct in_addr)); 1061 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1062 /* 1063 * Let ip_intr's mcast routing check handle mcast pkts 1064 */ 1065 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1066 break; 1067 1068 case IPOPT_RR: 1069 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1070 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1071 goto bad; 1072 } 1073 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1074 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1075 goto bad; 1076 } 1077 /* 1078 * If no space remains, ignore. 1079 */ 1080 off--; /* 0 origin */ 1081 if ((off + sizeof(struct in_addr)) > optlen) 1082 break; 1083 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1084 sizeof(ipaddr.sin_addr)); 1085 /* 1086 * locate outgoing interface; if we're the destination, 1087 * use the incoming interface (should be same). 1088 */ 1089 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1090 == NULL && 1091 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1092 type = ICMP_UNREACH; 1093 code = ICMP_UNREACH_HOST; 1094 goto bad; 1095 } 1096 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1097 (caddr_t)(cp + off), sizeof(struct in_addr)); 1098 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1099 break; 1100 1101 case IPOPT_TS: 1102 code = cp - (u_char *)ip; 1103 ipt = (struct ip_timestamp *)cp; 1104 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1105 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1106 goto bad; 1107 } 1108 if (ipt->ipt_ptr < 5) { 1109 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1110 goto bad; 1111 } 1112 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1113 if (++ipt->ipt_oflw == 0) { 1114 code = (u_char *)&ipt->ipt_ptr - 1115 (u_char *)ip; 1116 goto bad; 1117 } 1118 break; 1119 } 1120 cp0 = (cp + ipt->ipt_ptr - 1); 1121 switch (ipt->ipt_flg) { 1122 1123 case IPOPT_TS_TSONLY: 1124 break; 1125 1126 case IPOPT_TS_TSANDADDR: 1127 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1128 sizeof(struct in_addr) > ipt->ipt_len) { 1129 code = (u_char *)&ipt->ipt_ptr - 1130 (u_char *)ip; 1131 goto bad; 1132 } 1133 ipaddr.sin_addr = dst; 1134 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1135 m->m_pkthdr.rcvif)); 1136 if (ia == 0) 1137 continue; 1138 bcopy(&ia->ia_addr.sin_addr, 1139 cp0, sizeof(struct in_addr)); 1140 ipt->ipt_ptr += sizeof(struct in_addr); 1141 break; 1142 1143 case IPOPT_TS_PRESPEC: 1144 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1145 sizeof(struct in_addr) > ipt->ipt_len) { 1146 code = (u_char *)&ipt->ipt_ptr - 1147 (u_char *)ip; 1148 goto bad; 1149 } 1150 bcopy(cp0, &ipaddr.sin_addr, 1151 sizeof(struct in_addr)); 1152 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1153 == NULL) 1154 continue; 1155 ipt->ipt_ptr += sizeof(struct in_addr); 1156 break; 1157 1158 default: 1159 /* XXX can't take &ipt->ipt_flg */ 1160 code = (u_char *)&ipt->ipt_ptr - 1161 (u_char *)ip + 1; 1162 goto bad; 1163 } 1164 ntime = iptime(); 1165 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1166 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1167 sizeof(n_time)); 1168 ipt->ipt_ptr += sizeof(n_time); 1169 } 1170 } 1171 if (forward) { 1172 if (ip_forwsrcrt == 0) { 1173 type = ICMP_UNREACH; 1174 code = ICMP_UNREACH_SRCFAIL; 1175 goto bad; 1176 } 1177 ip_forward(m, 1); 1178 return (1); 1179 } 1180 return (0); 1181 bad: 1182 icmp_error(m, type, code, 0, 0); 1183 ipstat.ips_badoptions++; 1184 return (1); 1185 } 1186 1187 /* 1188 * Given address of next destination (final or next hop), 1189 * return internet address info of interface to be used to get there. 1190 */ 1191 struct in_ifaddr * 1192 ip_rtaddr(dst) 1193 struct in_addr dst; 1194 { 1195 struct sockaddr_in *sin; 1196 1197 sin = satosin(&ipforward_rt.ro_dst); 1198 1199 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1200 if (ipforward_rt.ro_rt) { 1201 RTFREE(ipforward_rt.ro_rt); 1202 ipforward_rt.ro_rt = 0; 1203 } 1204 sin->sin_family = AF_INET; 1205 sin->sin_len = sizeof(*sin); 1206 sin->sin_addr = dst; 1207 1208 rtalloc(&ipforward_rt); 1209 } 1210 if (ipforward_rt.ro_rt == 0) 1211 return ((struct in_ifaddr *)0); 1212 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1213 } 1214 1215 /* 1216 * Save incoming source route for use in replies, 1217 * to be picked up later by ip_srcroute if the receiver is interested. 1218 */ 1219 void 1220 save_rte(option, dst) 1221 u_char *option; 1222 struct in_addr dst; 1223 { 1224 unsigned olen; 1225 1226 olen = option[IPOPT_OLEN]; 1227 #ifdef DIAGNOSTIC 1228 if (ipprintfs) 1229 printf("save_rte: olen %d\n", olen); 1230 #endif /* 0 */ 1231 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1232 return; 1233 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1234 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1235 ip_srcrt.dst = dst; 1236 } 1237 1238 /* 1239 * Retrieve incoming source route for use in replies, 1240 * in the same form used by setsockopt. 1241 * The first hop is placed before the options, will be removed later. 1242 */ 1243 struct mbuf * 1244 ip_srcroute() 1245 { 1246 struct in_addr *p, *q; 1247 struct mbuf *m; 1248 1249 if (ip_nhops == 0) 1250 return ((struct mbuf *)0); 1251 m = m_get(M_DONTWAIT, MT_SOOPTS); 1252 if (m == 0) 1253 return ((struct mbuf *)0); 1254 1255 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1256 1257 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1258 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1259 OPTSIZ; 1260 #ifdef DIAGNOSTIC 1261 if (ipprintfs) 1262 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1263 #endif 1264 1265 /* 1266 * First save first hop for return route 1267 */ 1268 p = &ip_srcrt.route[ip_nhops - 1]; 1269 *(mtod(m, struct in_addr *)) = *p--; 1270 #ifdef DIAGNOSTIC 1271 if (ipprintfs) 1272 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1273 #endif 1274 1275 /* 1276 * Copy option fields and padding (nop) to mbuf. 1277 */ 1278 ip_srcrt.nop = IPOPT_NOP; 1279 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1280 bcopy((caddr_t)&ip_srcrt.nop, 1281 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1282 q = (struct in_addr *)(mtod(m, caddr_t) + 1283 sizeof(struct in_addr) + OPTSIZ); 1284 #undef OPTSIZ 1285 /* 1286 * Record return path as an IP source route, 1287 * reversing the path (pointers are now aligned). 1288 */ 1289 while (p >= ip_srcrt.route) { 1290 #ifdef DIAGNOSTIC 1291 if (ipprintfs) 1292 printf(" %x", ntohl(q->s_addr)); 1293 #endif 1294 *q++ = *p--; 1295 } 1296 /* 1297 * Last hop goes to final destination. 1298 */ 1299 *q = ip_srcrt.dst; 1300 #ifdef DIAGNOSTIC 1301 if (ipprintfs) 1302 printf(" %x\n", ntohl(q->s_addr)); 1303 #endif 1304 return (m); 1305 } 1306 1307 /* 1308 * Strip out IP options, at higher 1309 * level protocol in the kernel. 1310 * Second argument is buffer to which options 1311 * will be moved, and return value is their length. 1312 * XXX should be deleted; last arg currently ignored. 1313 */ 1314 void 1315 ip_stripoptions(m, mopt) 1316 struct mbuf *m; 1317 struct mbuf *mopt; 1318 { 1319 int i; 1320 struct ip *ip = mtod(m, struct ip *); 1321 caddr_t opts; 1322 int olen; 1323 1324 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1325 opts = (caddr_t)(ip + 1); 1326 i = m->m_len - (sizeof (struct ip) + olen); 1327 bcopy(opts + olen, opts, (unsigned)i); 1328 m->m_len -= olen; 1329 if (m->m_flags & M_PKTHDR) 1330 m->m_pkthdr.len -= olen; 1331 ip->ip_len -= olen; 1332 ip->ip_hl = sizeof (struct ip) >> 2; 1333 } 1334 1335 int inetctlerrmap[PRC_NCMDS] = { 1336 0, 0, 0, 0, 1337 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1338 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1339 EMSGSIZE, EHOSTUNREACH, 0, 0, 1340 0, 0, 0, 0, 1341 ENOPROTOOPT 1342 }; 1343 1344 /* 1345 * Forward a packet. If some error occurs return the sender 1346 * an icmp packet. Note we can't always generate a meaningful 1347 * icmp message because icmp doesn't have a large enough repertoire 1348 * of codes and types. 1349 * 1350 * If not forwarding, just drop the packet. This could be confusing 1351 * if ipforwarding was zero but some routing protocol was advancing 1352 * us as a gateway to somewhere. However, we must let the routing 1353 * protocol deal with that. 1354 * 1355 * The srcrt parameter indicates whether the packet is being forwarded 1356 * via a source route. 1357 */ 1358 void 1359 ip_forward(m, srcrt) 1360 struct mbuf *m; 1361 int srcrt; 1362 { 1363 struct ip *ip = mtod(m, struct ip *); 1364 struct sockaddr_in *sin; 1365 struct rtentry *rt; 1366 int error, type = 0, code = 0; 1367 struct mbuf *mcopy; 1368 n_long dest; 1369 struct ifnet *destifp; 1370 #ifdef IPSEC 1371 struct ifnet dummyifp; 1372 #endif 1373 1374 dest = 0; 1375 #ifdef DIAGNOSTIC 1376 if (ipprintfs) 1377 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1378 ntohl(ip->ip_src.s_addr), 1379 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1380 #endif 1381 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1382 ipstat.ips_cantforward++; 1383 m_freem(m); 1384 return; 1385 } 1386 if (ip->ip_ttl <= IPTTLDEC) { 1387 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1388 return; 1389 } 1390 ip->ip_ttl -= IPTTLDEC; 1391 1392 sin = satosin(&ipforward_rt.ro_dst); 1393 if ((rt = ipforward_rt.ro_rt) == 0 || 1394 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1395 if (ipforward_rt.ro_rt) { 1396 RTFREE(ipforward_rt.ro_rt); 1397 ipforward_rt.ro_rt = 0; 1398 } 1399 sin->sin_family = AF_INET; 1400 sin->sin_len = sizeof(struct sockaddr_in); 1401 sin->sin_addr = ip->ip_dst; 1402 1403 rtalloc(&ipforward_rt); 1404 if (ipforward_rt.ro_rt == 0) { 1405 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1406 return; 1407 } 1408 rt = ipforward_rt.ro_rt; 1409 } 1410 1411 /* 1412 * Save at most 68 bytes of the packet in case 1413 * we need to generate an ICMP message to the src. 1414 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1415 */ 1416 mcopy = m_copym(m, 0, imin((int)ip->ip_len, 68), M_DONTWAIT); 1417 if (mcopy) 1418 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1419 1420 /* 1421 * If forwarding packet using same interface that it came in on, 1422 * perhaps should send a redirect to sender to shortcut a hop. 1423 * Only send redirect if source is sending directly to us, 1424 * and if packet was not source routed (or has any options). 1425 * Also, don't send redirect if forwarding using a default route 1426 * or a route modified by a redirect. 1427 */ 1428 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1429 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1430 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1431 ipsendredirects && !srcrt) { 1432 if (rt->rt_ifa && 1433 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1434 ifatoia(rt->rt_ifa)->ia_subnet) { 1435 if (rt->rt_flags & RTF_GATEWAY) 1436 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1437 else 1438 dest = ip->ip_dst.s_addr; 1439 /* 1440 * Router requirements says to only send host 1441 * redirects. 1442 */ 1443 type = ICMP_REDIRECT; 1444 code = ICMP_REDIRECT_HOST; 1445 #ifdef DIAGNOSTIC 1446 if (ipprintfs) 1447 printf("redirect (%d) to %x\n", code, 1448 (u_int32_t)dest); 1449 #endif 1450 } 1451 } 1452 1453 #ifdef IPSEC 1454 /* Don't lookup socket in forwading case */ 1455 ipsec_setsocket(m, NULL); 1456 #endif 1457 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1458 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1459 if (error) 1460 ipstat.ips_cantforward++; 1461 else { 1462 ipstat.ips_forward++; 1463 if (type) 1464 ipstat.ips_redirectsent++; 1465 else { 1466 if (mcopy) { 1467 #ifdef GATEWAY 1468 if (mcopy->m_flags & M_CANFASTFWD) 1469 ipflow_create(&ipforward_rt, mcopy); 1470 #endif 1471 m_freem(mcopy); 1472 } 1473 return; 1474 } 1475 } 1476 if (mcopy == NULL) 1477 return; 1478 destifp = NULL; 1479 1480 switch (error) { 1481 1482 case 0: /* forwarded, but need redirect */ 1483 /* type, code set above */ 1484 break; 1485 1486 case ENETUNREACH: /* shouldn't happen, checked above */ 1487 case EHOSTUNREACH: 1488 case ENETDOWN: 1489 case EHOSTDOWN: 1490 default: 1491 type = ICMP_UNREACH; 1492 code = ICMP_UNREACH_HOST; 1493 break; 1494 1495 case EMSGSIZE: 1496 type = ICMP_UNREACH; 1497 code = ICMP_UNREACH_NEEDFRAG; 1498 #ifndef IPSEC 1499 if (ipforward_rt.ro_rt) 1500 destifp = ipforward_rt.ro_rt->rt_ifp; 1501 #else 1502 /* 1503 * If the packet is routed over IPsec tunnel, tell the 1504 * originator the tunnel MTU. 1505 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1506 * XXX quickhack!!! 1507 */ 1508 if (ipforward_rt.ro_rt) { 1509 struct secpolicy *sp; 1510 int ipsecerror; 1511 size_t ipsechdr; 1512 struct route *ro; 1513 1514 sp = ipsec4_getpolicybyaddr(mcopy, 1515 IPSEC_DIR_OUTBOUND, 1516 IP_FORWARDING, 1517 &ipsecerror); 1518 1519 if (sp == NULL) 1520 destifp = ipforward_rt.ro_rt->rt_ifp; 1521 else { 1522 /* count IPsec header size */ 1523 ipsechdr = ipsec4_hdrsiz(mcopy, 1524 IPSEC_DIR_OUTBOUND, 1525 NULL); 1526 1527 /* 1528 * find the correct route for outer IPv4 1529 * header, compute tunnel MTU. 1530 * 1531 * XXX BUG ALERT 1532 * The "dummyifp" code relies upon the fact 1533 * that icmp_error() touches only ifp->if_mtu. 1534 */ 1535 /*XXX*/ 1536 destifp = NULL; 1537 if (sp->req != NULL 1538 && sp->req->sav != NULL 1539 && sp->req->sav->sah != NULL) { 1540 ro = &sp->req->sav->sah->sa_route; 1541 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1542 dummyifp.if_mtu = 1543 ro->ro_rt->rt_ifp->if_mtu; 1544 dummyifp.if_mtu -= ipsechdr; 1545 destifp = &dummyifp; 1546 } 1547 } 1548 1549 key_freesp(sp); 1550 } 1551 } 1552 #endif /*IPSEC*/ 1553 ipstat.ips_cantfrag++; 1554 break; 1555 1556 case ENOBUFS: 1557 type = ICMP_SOURCEQUENCH; 1558 code = 0; 1559 break; 1560 } 1561 icmp_error(mcopy, type, code, dest, destifp); 1562 } 1563 1564 void 1565 ip_savecontrol(inp, mp, ip, m) 1566 struct inpcb *inp; 1567 struct mbuf **mp; 1568 struct ip *ip; 1569 struct mbuf *m; 1570 { 1571 1572 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1573 struct timeval tv; 1574 1575 microtime(&tv); 1576 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1577 SCM_TIMESTAMP, SOL_SOCKET); 1578 if (*mp) 1579 mp = &(*mp)->m_next; 1580 } 1581 if (inp->inp_flags & INP_RECVDSTADDR) { 1582 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1583 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1584 if (*mp) 1585 mp = &(*mp)->m_next; 1586 } 1587 #ifdef notyet 1588 /* 1589 * XXX 1590 * Moving these out of udp_input() made them even more broken 1591 * than they already were. 1592 * - fenner@parc.xerox.com 1593 */ 1594 /* options were tossed already */ 1595 if (inp->inp_flags & INP_RECVOPTS) { 1596 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1597 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1598 if (*mp) 1599 mp = &(*mp)->m_next; 1600 } 1601 /* ip_srcroute doesn't do what we want here, need to fix */ 1602 if (inp->inp_flags & INP_RECVRETOPTS) { 1603 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1604 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1605 if (*mp) 1606 mp = &(*mp)->m_next; 1607 } 1608 #endif 1609 if (inp->inp_flags & INP_RECVIF) { 1610 struct sockaddr_dl sdl; 1611 1612 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1613 sdl.sdl_family = AF_LINK; 1614 sdl.sdl_index = m->m_pkthdr.rcvif ? 1615 m->m_pkthdr.rcvif->if_index : 0; 1616 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1617 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1618 IP_RECVIF, IPPROTO_IP); 1619 if (*mp) 1620 mp = &(*mp)->m_next; 1621 } 1622 } 1623 1624 int 1625 ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1626 int *name; 1627 u_int namelen; 1628 void *oldp; 1629 size_t *oldlenp; 1630 void *newp; 1631 size_t newlen; 1632 { 1633 extern int subnetsarelocal, hostzeroisbroadcast; 1634 1635 int error, old; 1636 1637 /* All sysctl names at this level are terminal. */ 1638 if (namelen != 1) 1639 return (ENOTDIR); 1640 1641 switch (name[0]) { 1642 case IPCTL_FORWARDING: 1643 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1644 case IPCTL_SENDREDIRECTS: 1645 return (sysctl_int(oldp, oldlenp, newp, newlen, 1646 &ipsendredirects)); 1647 case IPCTL_DEFTTL: 1648 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1649 #ifdef notyet 1650 case IPCTL_DEFMTU: 1651 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1652 #endif 1653 case IPCTL_FORWSRCRT: 1654 /* Don't allow this to change in a secure environment. */ 1655 if (securelevel > 0) 1656 return (sysctl_rdint(oldp, oldlenp, newp, 1657 ip_forwsrcrt)); 1658 else 1659 return (sysctl_int(oldp, oldlenp, newp, newlen, 1660 &ip_forwsrcrt)); 1661 case IPCTL_DIRECTEDBCAST: 1662 return (sysctl_int(oldp, oldlenp, newp, newlen, 1663 &ip_directedbcast)); 1664 case IPCTL_ALLOWSRCRT: 1665 return (sysctl_int(oldp, oldlenp, newp, newlen, 1666 &ip_allowsrcrt)); 1667 case IPCTL_SUBNETSARELOCAL: 1668 return (sysctl_int(oldp, oldlenp, newp, newlen, 1669 &subnetsarelocal)); 1670 case IPCTL_MTUDISC: 1671 error = sysctl_int(oldp, oldlenp, newp, newlen, 1672 &ip_mtudisc); 1673 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1674 ip_mtudisc_timeout_q = 1675 rt_timer_queue_create(ip_mtudisc_timeout); 1676 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1677 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1678 ip_mtudisc_timeout_q = NULL; 1679 } 1680 return error; 1681 case IPCTL_ANONPORTMIN: 1682 old = anonportmin; 1683 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1684 if (anonportmin >= anonportmax || anonportmin < 0 1685 || anonportmin > 65535 1686 #ifndef IPNOPRIVPORTS 1687 || anonportmin < IPPORT_RESERVED 1688 #endif 1689 ) { 1690 anonportmin = old; 1691 return (EINVAL); 1692 } 1693 return (error); 1694 case IPCTL_ANONPORTMAX: 1695 old = anonportmax; 1696 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1697 if (anonportmin >= anonportmax || anonportmax < 0 1698 || anonportmax > 65535 1699 #ifndef IPNOPRIVPORTS 1700 || anonportmax < IPPORT_RESERVED 1701 #endif 1702 ) { 1703 anonportmax = old; 1704 return (EINVAL); 1705 } 1706 return (error); 1707 case IPCTL_MTUDISCTIMEOUT: 1708 error = sysctl_int(oldp, oldlenp, newp, newlen, 1709 &ip_mtudisc_timeout); 1710 if (ip_mtudisc_timeout_q != NULL) 1711 rt_timer_queue_change(ip_mtudisc_timeout_q, 1712 ip_mtudisc_timeout); 1713 return (error); 1714 #ifdef GATEWAY 1715 case IPCTL_MAXFLOWS: 1716 { 1717 int s; 1718 1719 error = sysctl_int(oldp, oldlenp, newp, newlen, 1720 &ip_maxflows); 1721 s = splsoftnet(); 1722 ipflow_reap(0); 1723 splx(s); 1724 return (error); 1725 } 1726 #endif 1727 case IPCTL_HOSTZEROBROADCAST: 1728 return (sysctl_int(oldp, oldlenp, newp, newlen, 1729 &hostzeroisbroadcast)); 1730 #if NGIF > 0 1731 case IPCTL_GIF_TTL: 1732 return(sysctl_int(oldp, oldlenp, newp, newlen, 1733 &ip_gif_ttl)); 1734 #endif 1735 1736 #ifndef IPNOPRIVPORTS 1737 case IPCTL_LOWPORTMIN: 1738 old = lowportmin; 1739 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmin); 1740 if (lowportmin >= lowportmax 1741 || lowportmin > IPPORT_RESERVEDMAX 1742 || lowportmin < IPPORT_RESERVEDMIN 1743 ) { 1744 lowportmin = old; 1745 return (EINVAL); 1746 } 1747 return (error); 1748 case IPCTL_LOWPORTMAX: 1749 old = lowportmax; 1750 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmax); 1751 if (lowportmin >= lowportmax 1752 || lowportmax > IPPORT_RESERVEDMAX 1753 || lowportmax < IPPORT_RESERVEDMIN 1754 ) { 1755 lowportmax = old; 1756 return (EINVAL); 1757 } 1758 return (error); 1759 #endif 1760 1761 default: 1762 return (EOPNOTSUPP); 1763 } 1764 /* NOTREACHED */ 1765 } 1766