1 /* $NetBSD: ip_input.c,v 1.94 1999/10/26 09:53:17 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" 108 109 #include <sys/param.h> 110 #include <sys/systm.h> 111 #include <sys/malloc.h> 112 #include <sys/mbuf.h> 113 #include <sys/domain.h> 114 #include <sys/protosw.h> 115 #include <sys/socket.h> 116 #include <sys/socketvar.h> 117 #include <sys/errno.h> 118 #include <sys/time.h> 119 #include <sys/kernel.h> 120 #include <sys/proc.h> 121 #include <sys/pool.h> 122 123 #include <vm/vm.h> 124 #include <sys/sysctl.h> 125 126 #include <net/if.h> 127 #include <net/if_dl.h> 128 #include <net/route.h> 129 #include <net/pfil.h> 130 131 #include <netinet/in.h> 132 #include <netinet/in_systm.h> 133 #include <netinet/ip.h> 134 #include <netinet/in_pcb.h> 135 #include <netinet/in_var.h> 136 #include <netinet/ip_var.h> 137 #include <netinet/ip_icmp.h> 138 /* just for gif_ttl */ 139 #include <netinet/in_gif.h> 140 #include "gif.h" 141 142 #ifdef IPSEC 143 #include <netinet6/ipsec.h> 144 #include <netkey/key.h> 145 #include <netkey/key_debug.h> 146 #endif 147 148 #ifndef IPFORWARDING 149 #ifdef GATEWAY 150 #define IPFORWARDING 1 /* forward IP packets not for us */ 151 #else /* GATEWAY */ 152 #define IPFORWARDING 0 /* don't forward IP packets not for us */ 153 #endif /* GATEWAY */ 154 #endif /* IPFORWARDING */ 155 #ifndef IPSENDREDIRECTS 156 #define IPSENDREDIRECTS 1 157 #endif 158 #ifndef IPFORWSRCRT 159 #define IPFORWSRCRT 1 /* forward source-routed packets */ 160 #endif 161 #ifndef IPALLOWSRCRT 162 #define IPALLOWSRCRT 1 /* allow source-routed packets */ 163 #endif 164 #ifndef IPMTUDISC 165 #define IPMTUDISC 0 166 #endif 167 #ifndef IPMTUDISCTIMEOUT 168 #define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 169 #endif 170 171 /* 172 * Note: DIRECTED_BROADCAST is handled this way so that previous 173 * configuration using this option will Just Work. 174 */ 175 #ifndef IPDIRECTEDBCAST 176 #ifdef DIRECTED_BROADCAST 177 #define IPDIRECTEDBCAST 1 178 #else 179 #define IPDIRECTEDBCAST 0 180 #endif /* DIRECTED_BROADCAST */ 181 #endif /* IPDIRECTEDBCAST */ 182 int ipforwarding = IPFORWARDING; 183 int ipsendredirects = IPSENDREDIRECTS; 184 int ip_defttl = IPDEFTTL; 185 int ip_forwsrcrt = IPFORWSRCRT; 186 int ip_directedbcast = IPDIRECTEDBCAST; 187 int ip_allowsrcrt = IPALLOWSRCRT; 188 int ip_mtudisc = IPMTUDISC; 189 u_int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 190 #ifdef DIAGNOSTIC 191 int ipprintfs = 0; 192 #endif 193 194 struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 195 196 extern struct domain inetdomain; 197 extern struct protosw inetsw[]; 198 u_char ip_protox[IPPROTO_MAX]; 199 int ipqmaxlen = IFQ_MAXLEN; 200 struct in_ifaddrhead in_ifaddr; 201 struct in_ifaddrhashhead *in_ifaddrhashtbl; 202 struct ifqueue ipintrq; 203 struct ipstat ipstat; 204 u_int16_t ip_id; 205 int ip_defttl; 206 207 struct ipqhead ipq; 208 int ipq_locked; 209 210 static __inline int ipq_lock_try __P((void)); 211 static __inline void ipq_unlock __P((void)); 212 213 static __inline int 214 ipq_lock_try() 215 { 216 int s; 217 218 s = splimp(); 219 if (ipq_locked) { 220 splx(s); 221 return (0); 222 } 223 ipq_locked = 1; 224 splx(s); 225 return (1); 226 } 227 228 static __inline void 229 ipq_unlock() 230 { 231 int s; 232 233 s = splimp(); 234 ipq_locked = 0; 235 splx(s); 236 } 237 238 #ifdef DIAGNOSTIC 239 #define IPQ_LOCK() \ 240 do { \ 241 if (ipq_lock_try() == 0) { \ 242 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 243 panic("ipq_lock"); \ 244 } \ 245 } while (0) 246 #define IPQ_LOCK_CHECK() \ 247 do { \ 248 if (ipq_locked == 0) { \ 249 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 250 panic("ipq lock check"); \ 251 } \ 252 } while (0) 253 #else 254 #define IPQ_LOCK() (void) ipq_lock_try() 255 #define IPQ_LOCK_CHECK() /* nothing */ 256 #endif 257 258 #define IPQ_UNLOCK() ipq_unlock() 259 260 struct pool ipqent_pool; 261 262 /* 263 * We need to save the IP options in case a protocol wants to respond 264 * to an incoming packet over the same route if the packet got here 265 * using IP source routing. This allows connection establishment and 266 * maintenance when the remote end is on a network that is not known 267 * to us. 268 */ 269 int ip_nhops = 0; 270 static struct ip_srcrt { 271 struct in_addr dst; /* final destination */ 272 char nop; /* one NOP to align */ 273 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 274 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 275 } ip_srcrt; 276 277 static void save_rte __P((u_char *, struct in_addr)); 278 279 /* 280 * IP initialization: fill in IP protocol switch table. 281 * All protocols not implemented in kernel go to raw IP protocol handler. 282 */ 283 void 284 ip_init() 285 { 286 register struct protosw *pr; 287 register int i; 288 289 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 290 0, NULL, NULL, M_IPQ); 291 292 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 293 if (pr == 0) 294 panic("ip_init"); 295 for (i = 0; i < IPPROTO_MAX; i++) 296 ip_protox[i] = pr - inetsw; 297 for (pr = inetdomain.dom_protosw; 298 pr < inetdomain.dom_protoswNPROTOSW; pr++) 299 if (pr->pr_domain->dom_family == PF_INET && 300 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 301 ip_protox[pr->pr_protocol] = pr - inetsw; 302 LIST_INIT(&ipq); 303 ip_id = time.tv_sec & 0xffff; 304 ipintrq.ifq_maxlen = ipqmaxlen; 305 TAILQ_INIT(&in_ifaddr); 306 in_ifaddrhashtbl = 307 hashinit(IN_IFADDR_HASH_SIZE, M_IFADDR, M_WAITOK, &in_ifaddrhash); 308 if (ip_mtudisc != 0) 309 ip_mtudisc_timeout_q = 310 rt_timer_queue_create(ip_mtudisc_timeout); 311 #ifdef GATEWAY 312 ipflow_init(); 313 #endif 314 } 315 316 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 317 struct route ipforward_rt; 318 319 /* 320 * IP software interrupt routine 321 */ 322 void 323 ipintr() 324 { 325 int s; 326 struct mbuf *m; 327 328 while (1) { 329 s = splimp(); 330 IF_DEQUEUE(&ipintrq, m); 331 splx(s); 332 if (m == 0) 333 return; 334 ip_input(m); 335 } 336 } 337 338 /* 339 * Ip input routine. Checksum and byte swap header. If fragmented 340 * try to reassemble. Process options. Pass to next level. 341 */ 342 void 343 ip_input(struct mbuf *m) 344 { 345 register struct ip *ip = NULL; 346 register struct ipq *fp; 347 register struct in_ifaddr *ia; 348 register struct ifaddr *ifa; 349 struct ipqent *ipqe; 350 int hlen = 0, mff, len; 351 #ifdef PFIL_HOOKS 352 struct packet_filter_hook *pfh; 353 struct mbuf *m0; 354 int rv; 355 #endif /* PFIL_HOOKS */ 356 357 #ifdef DIAGNOSTIC 358 if ((m->m_flags & M_PKTHDR) == 0) 359 panic("ipintr no HDR"); 360 #endif 361 #ifdef IPSEC 362 /* 363 * should the inner packet be considered authentic? 364 * see comment in ah4_input(). 365 */ 366 if (m) { 367 m->m_flags &= ~M_AUTHIPHDR; 368 m->m_flags &= ~M_AUTHIPDGM; 369 } 370 #endif 371 /* 372 * If no IP addresses have been set yet but the interfaces 373 * are receiving, can't do anything with incoming packets yet. 374 */ 375 if (in_ifaddr.tqh_first == 0) 376 goto bad; 377 ipstat.ips_total++; 378 if (m->m_len < sizeof (struct ip) && 379 (m = m_pullup(m, sizeof (struct ip))) == 0) { 380 ipstat.ips_toosmall++; 381 return; 382 } 383 ip = mtod(m, struct ip *); 384 if (ip->ip_v != IPVERSION) { 385 ipstat.ips_badvers++; 386 goto bad; 387 } 388 hlen = ip->ip_hl << 2; 389 if (hlen < sizeof(struct ip)) { /* minimum header length */ 390 ipstat.ips_badhlen++; 391 goto bad; 392 } 393 if (hlen > m->m_len) { 394 if ((m = m_pullup(m, hlen)) == 0) { 395 ipstat.ips_badhlen++; 396 return; 397 } 398 ip = mtod(m, struct ip *); 399 } 400 /* 401 * we drop packets that have a multicast address as source 402 * as wanted by rfc 1112 403 */ 404 if (IN_MULTICAST(ip->ip_src.s_addr)) { 405 goto bad; 406 } 407 408 if (in_cksum(m, hlen) != 0) { 409 ipstat.ips_badsum++; 410 goto bad; 411 } 412 413 /* 414 * Convert fields to host representation. 415 */ 416 NTOHS(ip->ip_len); 417 NTOHS(ip->ip_off); 418 len = ip->ip_len; 419 420 /* 421 * Check for additional length bogosity 422 */ 423 if (len < hlen) { 424 ipstat.ips_badlen++; 425 goto bad; 426 } 427 428 /* 429 * Check that the amount of data in the buffers 430 * is as at least much as the IP header would have us expect. 431 * Trim mbufs if longer than we expect. 432 * Drop packet if shorter than we expect. 433 */ 434 if (m->m_pkthdr.len < len) { 435 ipstat.ips_tooshort++; 436 goto bad; 437 } 438 if (m->m_pkthdr.len > len) { 439 if (m->m_len == m->m_pkthdr.len) { 440 m->m_len = len; 441 m->m_pkthdr.len = len; 442 } else 443 m_adj(m, len - m->m_pkthdr.len); 444 } 445 446 #ifdef IPSEC 447 /* ipflow (IP fast fowarding) is not compatible with IPsec. */ 448 m->m_flags &= ~M_CANFASTFWD; 449 #else 450 /* 451 * Assume that we can create a fast-forward IP flow entry 452 * based on this packet. 453 */ 454 m->m_flags |= M_CANFASTFWD; 455 #endif 456 457 #ifdef PFIL_HOOKS 458 /* 459 * Run through list of hooks for input packets. If there are any 460 * filters which require that additional packets in the flow are 461 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 462 * Note that filters must _never_ set this flag, as another filter 463 * in the list may have previously cleared it. 464 */ 465 m0 = m; 466 for (pfh = pfil_hook_get(PFIL_IN); pfh; pfh = pfh->pfil_link.tqe_next) 467 if (pfh->pfil_func) { 468 rv = pfh->pfil_func(ip, hlen, m->m_pkthdr.rcvif, 0, &m0); 469 if (rv) 470 return; 471 m = m0; 472 if (m == NULL) 473 return; 474 ip = mtod(m, struct ip *); 475 } 476 #endif /* PFIL_HOOKS */ 477 478 /* 479 * Process options and, if not destined for us, 480 * ship it on. ip_dooptions returns 1 when an 481 * error was detected (causing an icmp message 482 * to be sent and the original packet to be freed). 483 */ 484 ip_nhops = 0; /* for source routed packets */ 485 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 486 return; 487 488 /* 489 * Check our list of addresses, to see if the packet is for us. 490 */ 491 INADDR_TO_IA(ip->ip_dst, ia); 492 if (ia != NULL) 493 goto ours; 494 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 495 for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first; 496 ifa != NULL; ifa = ifa->ifa_list.tqe_next) { 497 if (ifa->ifa_addr->sa_family != AF_INET) continue; 498 ia = ifatoia(ifa); 499 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 500 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 501 /* 502 * Look for all-0's host part (old broadcast addr), 503 * either for subnet or net. 504 */ 505 ip->ip_dst.s_addr == ia->ia_subnet || 506 ip->ip_dst.s_addr == ia->ia_net) 507 goto ours; 508 /* 509 * An interface with IP address zero accepts 510 * all packets that arrive on that interface. 511 */ 512 if (in_nullhost(ia->ia_addr.sin_addr)) 513 goto ours; 514 } 515 } 516 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 517 struct in_multi *inm; 518 #ifdef MROUTING 519 extern struct socket *ip_mrouter; 520 521 if (m->m_flags & M_EXT) { 522 if ((m = m_pullup(m, hlen)) == 0) { 523 ipstat.ips_toosmall++; 524 return; 525 } 526 ip = mtod(m, struct ip *); 527 } 528 529 if (ip_mrouter) { 530 /* 531 * If we are acting as a multicast router, all 532 * incoming multicast packets are passed to the 533 * kernel-level multicast forwarding function. 534 * The packet is returned (relatively) intact; if 535 * ip_mforward() returns a non-zero value, the packet 536 * must be discarded, else it may be accepted below. 537 * 538 * (The IP ident field is put in the same byte order 539 * as expected when ip_mforward() is called from 540 * ip_output().) 541 */ 542 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 543 ipstat.ips_cantforward++; 544 m_freem(m); 545 return; 546 } 547 548 /* 549 * The process-level routing demon needs to receive 550 * all multicast IGMP packets, whether or not this 551 * host belongs to their destination groups. 552 */ 553 if (ip->ip_p == IPPROTO_IGMP) 554 goto ours; 555 ipstat.ips_forward++; 556 } 557 #endif 558 /* 559 * See if we belong to the destination multicast group on the 560 * arrival interface. 561 */ 562 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 563 if (inm == NULL) { 564 ipstat.ips_cantforward++; 565 m_freem(m); 566 return; 567 } 568 goto ours; 569 } 570 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 571 in_nullhost(ip->ip_dst)) 572 goto ours; 573 574 /* 575 * Not for us; forward if possible and desirable. 576 */ 577 if (ipforwarding == 0) { 578 ipstat.ips_cantforward++; 579 m_freem(m); 580 } else 581 ip_forward(m, 0); 582 return; 583 584 ours: 585 /* 586 * If offset or IP_MF are set, must reassemble. 587 * Otherwise, nothing need be done. 588 * (We could look in the reassembly queue to see 589 * if the packet was previously fragmented, 590 * but it's not worth the time; just let them time out.) 591 */ 592 if (ip->ip_off & ~(IP_DF|IP_RF)) { 593 /* 594 * Look for queue of fragments 595 * of this datagram. 596 */ 597 IPQ_LOCK(); 598 for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next) 599 if (ip->ip_id == fp->ipq_id && 600 in_hosteq(ip->ip_src, fp->ipq_src) && 601 in_hosteq(ip->ip_dst, fp->ipq_dst) && 602 ip->ip_p == fp->ipq_p) 603 goto found; 604 fp = 0; 605 found: 606 607 /* 608 * Adjust ip_len to not reflect header, 609 * set ipqe_mff if more fragments are expected, 610 * convert offset of this to bytes. 611 */ 612 ip->ip_len -= hlen; 613 mff = (ip->ip_off & IP_MF) != 0; 614 if (mff) { 615 /* 616 * Make sure that fragments have a data length 617 * that's a non-zero multiple of 8 bytes. 618 */ 619 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 620 ipstat.ips_badfrags++; 621 IPQ_UNLOCK(); 622 goto bad; 623 } 624 } 625 ip->ip_off <<= 3; 626 627 /* 628 * If datagram marked as having more fragments 629 * or if this is not the first fragment, 630 * attempt reassembly; if it succeeds, proceed. 631 */ 632 if (mff || ip->ip_off) { 633 ipstat.ips_fragments++; 634 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 635 if (ipqe == NULL) { 636 ipstat.ips_rcvmemdrop++; 637 IPQ_UNLOCK(); 638 goto bad; 639 } 640 ipqe->ipqe_mff = mff; 641 ipqe->ipqe_m = m; 642 ipqe->ipqe_ip = ip; 643 m = ip_reass(ipqe, fp); 644 if (m == 0) { 645 IPQ_UNLOCK(); 646 return; 647 } 648 ipstat.ips_reassembled++; 649 ip = mtod(m, struct ip *); 650 hlen = ip->ip_hl << 2; 651 ip->ip_len += hlen; 652 } else 653 if (fp) 654 ip_freef(fp); 655 IPQ_UNLOCK(); 656 } 657 658 /* 659 * Switch out to protocol's input routine. 660 */ 661 #if IFA_STATS 662 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 663 #endif 664 ipstat.ips_delivered++; 665 { 666 int off = hlen, nh = ip->ip_p; 667 668 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 669 return; 670 } 671 bad: 672 m_freem(m); 673 } 674 675 /* 676 * Take incoming datagram fragment and try to 677 * reassemble it into whole datagram. If a chain for 678 * reassembly of this datagram already exists, then it 679 * is given as fp; otherwise have to make a chain. 680 */ 681 struct mbuf * 682 ip_reass(ipqe, fp) 683 register struct ipqent *ipqe; 684 register struct ipq *fp; 685 { 686 register struct mbuf *m = ipqe->ipqe_m; 687 register struct ipqent *nq, *p, *q; 688 struct ip *ip; 689 struct mbuf *t; 690 int hlen = ipqe->ipqe_ip->ip_hl << 2; 691 int i, next; 692 693 IPQ_LOCK_CHECK(); 694 695 /* 696 * Presence of header sizes in mbufs 697 * would confuse code below. 698 */ 699 m->m_data += hlen; 700 m->m_len -= hlen; 701 702 /* 703 * If first fragment to arrive, create a reassembly queue. 704 */ 705 if (fp == 0) { 706 MALLOC(fp, struct ipq *, sizeof (struct ipq), 707 M_FTABLE, M_NOWAIT); 708 if (fp == NULL) 709 goto dropfrag; 710 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 711 fp->ipq_ttl = IPFRAGTTL; 712 fp->ipq_p = ipqe->ipqe_ip->ip_p; 713 fp->ipq_id = ipqe->ipqe_ip->ip_id; 714 LIST_INIT(&fp->ipq_fragq); 715 fp->ipq_src = ipqe->ipqe_ip->ip_src; 716 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 717 p = NULL; 718 goto insert; 719 } 720 721 /* 722 * Find a segment which begins after this one does. 723 */ 724 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 725 p = q, q = q->ipqe_q.le_next) 726 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 727 break; 728 729 /* 730 * If there is a preceding segment, it may provide some of 731 * our data already. If so, drop the data from the incoming 732 * segment. If it provides all of our data, drop us. 733 */ 734 if (p != NULL) { 735 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 736 ipqe->ipqe_ip->ip_off; 737 if (i > 0) { 738 if (i >= ipqe->ipqe_ip->ip_len) 739 goto dropfrag; 740 m_adj(ipqe->ipqe_m, i); 741 ipqe->ipqe_ip->ip_off += i; 742 ipqe->ipqe_ip->ip_len -= i; 743 } 744 } 745 746 /* 747 * While we overlap succeeding segments trim them or, 748 * if they are completely covered, dequeue them. 749 */ 750 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 751 q->ipqe_ip->ip_off; q = nq) { 752 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 753 q->ipqe_ip->ip_off; 754 if (i < q->ipqe_ip->ip_len) { 755 q->ipqe_ip->ip_len -= i; 756 q->ipqe_ip->ip_off += i; 757 m_adj(q->ipqe_m, i); 758 break; 759 } 760 nq = q->ipqe_q.le_next; 761 m_freem(q->ipqe_m); 762 LIST_REMOVE(q, ipqe_q); 763 pool_put(&ipqent_pool, q); 764 } 765 766 insert: 767 /* 768 * Stick new segment in its place; 769 * check for complete reassembly. 770 */ 771 if (p == NULL) { 772 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 773 } else { 774 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 775 } 776 next = 0; 777 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 778 p = q, q = q->ipqe_q.le_next) { 779 if (q->ipqe_ip->ip_off != next) 780 return (0); 781 next += q->ipqe_ip->ip_len; 782 } 783 if (p->ipqe_mff) 784 return (0); 785 786 /* 787 * Reassembly is complete. Check for a bogus message size and 788 * concatenate fragments. 789 */ 790 q = fp->ipq_fragq.lh_first; 791 ip = q->ipqe_ip; 792 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 793 ipstat.ips_toolong++; 794 ip_freef(fp); 795 return (0); 796 } 797 m = q->ipqe_m; 798 t = m->m_next; 799 m->m_next = 0; 800 m_cat(m, t); 801 nq = q->ipqe_q.le_next; 802 pool_put(&ipqent_pool, q); 803 for (q = nq; q != NULL; q = nq) { 804 t = q->ipqe_m; 805 nq = q->ipqe_q.le_next; 806 pool_put(&ipqent_pool, q); 807 m_cat(m, t); 808 } 809 810 /* 811 * Create header for new ip packet by 812 * modifying header of first packet; 813 * dequeue and discard fragment reassembly header. 814 * Make header visible. 815 */ 816 ip->ip_len = next; 817 ip->ip_ttl = 0; /* xxx */ 818 ip->ip_sum = 0; 819 ip->ip_src = fp->ipq_src; 820 ip->ip_dst = fp->ipq_dst; 821 LIST_REMOVE(fp, ipq_q); 822 FREE(fp, M_FTABLE); 823 m->m_len += (ip->ip_hl << 2); 824 m->m_data -= (ip->ip_hl << 2); 825 /* some debugging cruft by sklower, below, will go away soon */ 826 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 827 register int plen = 0; 828 for (t = m; t; t = t->m_next) 829 plen += t->m_len; 830 m->m_pkthdr.len = plen; 831 } 832 return (m); 833 834 dropfrag: 835 ipstat.ips_fragdropped++; 836 m_freem(m); 837 pool_put(&ipqent_pool, ipqe); 838 return (0); 839 } 840 841 /* 842 * Free a fragment reassembly header and all 843 * associated datagrams. 844 */ 845 void 846 ip_freef(fp) 847 struct ipq *fp; 848 { 849 register struct ipqent *q, *p; 850 851 IPQ_LOCK_CHECK(); 852 853 for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { 854 p = q->ipqe_q.le_next; 855 m_freem(q->ipqe_m); 856 LIST_REMOVE(q, ipqe_q); 857 pool_put(&ipqent_pool, q); 858 } 859 LIST_REMOVE(fp, ipq_q); 860 FREE(fp, M_FTABLE); 861 } 862 863 /* 864 * IP timer processing; 865 * if a timer expires on a reassembly 866 * queue, discard it. 867 */ 868 void 869 ip_slowtimo() 870 { 871 register struct ipq *fp, *nfp; 872 int s = splsoftnet(); 873 874 IPQ_LOCK(); 875 for (fp = ipq.lh_first; fp != NULL; fp = nfp) { 876 nfp = fp->ipq_q.le_next; 877 if (--fp->ipq_ttl == 0) { 878 ipstat.ips_fragtimeout++; 879 ip_freef(fp); 880 } 881 } 882 IPQ_UNLOCK(); 883 #ifdef GATEWAY 884 ipflow_slowtimo(); 885 #endif 886 splx(s); 887 } 888 889 /* 890 * Drain off all datagram fragments. 891 */ 892 void 893 ip_drain() 894 { 895 896 /* 897 * We may be called from a device's interrupt context. If 898 * the ipq is already busy, just bail out now. 899 */ 900 if (ipq_lock_try() == 0) 901 return; 902 903 while (ipq.lh_first != NULL) { 904 ipstat.ips_fragdropped++; 905 ip_freef(ipq.lh_first); 906 } 907 908 IPQ_UNLOCK(); 909 } 910 911 /* 912 * Do option processing on a datagram, 913 * possibly discarding it if bad options are encountered, 914 * or forwarding it if source-routed. 915 * Returns 1 if packet has been forwarded/freed, 916 * 0 if the packet should be processed further. 917 */ 918 int 919 ip_dooptions(m) 920 struct mbuf *m; 921 { 922 register struct ip *ip = mtod(m, struct ip *); 923 register u_char *cp; 924 register struct ip_timestamp *ipt; 925 register struct in_ifaddr *ia; 926 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 927 struct in_addr *sin, dst; 928 n_time ntime; 929 930 dst = ip->ip_dst; 931 cp = (u_char *)(ip + 1); 932 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 933 for (; cnt > 0; cnt -= optlen, cp += optlen) { 934 opt = cp[IPOPT_OPTVAL]; 935 if (opt == IPOPT_EOL) 936 break; 937 if (opt == IPOPT_NOP) 938 optlen = 1; 939 else { 940 optlen = cp[IPOPT_OLEN]; 941 if (optlen <= 0 || optlen > cnt) { 942 code = &cp[IPOPT_OLEN] - (u_char *)ip; 943 goto bad; 944 } 945 } 946 switch (opt) { 947 948 default: 949 break; 950 951 /* 952 * Source routing with record. 953 * Find interface with current destination address. 954 * If none on this machine then drop if strictly routed, 955 * or do nothing if loosely routed. 956 * Record interface address and bring up next address 957 * component. If strictly routed make sure next 958 * address is on directly accessible net. 959 */ 960 case IPOPT_LSRR: 961 case IPOPT_SSRR: 962 if (ip_allowsrcrt == 0) { 963 type = ICMP_UNREACH; 964 code = ICMP_UNREACH_NET_PROHIB; 965 goto bad; 966 } 967 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 968 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 969 goto bad; 970 } 971 ipaddr.sin_addr = ip->ip_dst; 972 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 973 if (ia == 0) { 974 if (opt == IPOPT_SSRR) { 975 type = ICMP_UNREACH; 976 code = ICMP_UNREACH_SRCFAIL; 977 goto bad; 978 } 979 /* 980 * Loose routing, and not at next destination 981 * yet; nothing to do except forward. 982 */ 983 break; 984 } 985 off--; /* 0 origin */ 986 if (off > optlen - sizeof(struct in_addr)) { 987 /* 988 * End of source route. Should be for us. 989 */ 990 save_rte(cp, ip->ip_src); 991 break; 992 } 993 /* 994 * locate outgoing interface 995 */ 996 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 997 sizeof(ipaddr.sin_addr)); 998 if (opt == IPOPT_SSRR) { 999 #define INA struct in_ifaddr * 1000 #define SA struct sockaddr * 1001 ia = (INA)ifa_ifwithladdr((SA)&ipaddr); 1002 } else 1003 ia = ip_rtaddr(ipaddr.sin_addr); 1004 if (ia == 0) { 1005 type = ICMP_UNREACH; 1006 code = ICMP_UNREACH_SRCFAIL; 1007 goto bad; 1008 } 1009 ip->ip_dst = ipaddr.sin_addr; 1010 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1011 (caddr_t)(cp + off), sizeof(struct in_addr)); 1012 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1013 /* 1014 * Let ip_intr's mcast routing check handle mcast pkts 1015 */ 1016 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1017 break; 1018 1019 case IPOPT_RR: 1020 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1021 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1022 goto bad; 1023 } 1024 /* 1025 * If no space remains, ignore. 1026 */ 1027 off--; /* 0 origin */ 1028 if (off > optlen - sizeof(struct in_addr)) 1029 break; 1030 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1031 sizeof(ipaddr.sin_addr)); 1032 /* 1033 * locate outgoing interface; if we're the destination, 1034 * use the incoming interface (should be same). 1035 */ 1036 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1037 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1038 type = ICMP_UNREACH; 1039 code = ICMP_UNREACH_HOST; 1040 goto bad; 1041 } 1042 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1043 (caddr_t)(cp + off), sizeof(struct in_addr)); 1044 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1045 break; 1046 1047 case IPOPT_TS: 1048 code = cp - (u_char *)ip; 1049 ipt = (struct ip_timestamp *)cp; 1050 if (ipt->ipt_len < 5) 1051 goto bad; 1052 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1053 if (++ipt->ipt_oflw == 0) 1054 goto bad; 1055 break; 1056 } 1057 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1058 switch (ipt->ipt_flg) { 1059 1060 case IPOPT_TS_TSONLY: 1061 break; 1062 1063 case IPOPT_TS_TSANDADDR: 1064 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1065 sizeof(struct in_addr) > ipt->ipt_len) 1066 goto bad; 1067 ipaddr.sin_addr = dst; 1068 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1069 m->m_pkthdr.rcvif); 1070 if (ia == 0) 1071 continue; 1072 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1073 (caddr_t)sin, sizeof(struct in_addr)); 1074 ipt->ipt_ptr += sizeof(struct in_addr); 1075 break; 1076 1077 case IPOPT_TS_PRESPEC: 1078 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1079 sizeof(struct in_addr) > ipt->ipt_len) 1080 goto bad; 1081 bcopy((caddr_t)sin, (caddr_t)&ipaddr.sin_addr, 1082 sizeof(struct in_addr)); 1083 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1084 continue; 1085 ipt->ipt_ptr += sizeof(struct in_addr); 1086 break; 1087 1088 default: 1089 goto bad; 1090 } 1091 ntime = iptime(); 1092 bcopy((caddr_t)&ntime, (caddr_t)cp + ipt->ipt_ptr - 1, 1093 sizeof(n_time)); 1094 ipt->ipt_ptr += sizeof(n_time); 1095 } 1096 } 1097 if (forward) { 1098 if (ip_forwsrcrt == 0) { 1099 type = ICMP_UNREACH; 1100 code = ICMP_UNREACH_SRCFAIL; 1101 goto bad; 1102 } 1103 ip_forward(m, 1); 1104 return (1); 1105 } 1106 return (0); 1107 bad: 1108 icmp_error(m, type, code, 0, 0); 1109 ipstat.ips_badoptions++; 1110 return (1); 1111 } 1112 1113 /* 1114 * Given address of next destination (final or next hop), 1115 * return internet address info of interface to be used to get there. 1116 */ 1117 struct in_ifaddr * 1118 ip_rtaddr(dst) 1119 struct in_addr dst; 1120 { 1121 register struct sockaddr_in *sin; 1122 1123 sin = satosin(&ipforward_rt.ro_dst); 1124 1125 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1126 if (ipforward_rt.ro_rt) { 1127 RTFREE(ipforward_rt.ro_rt); 1128 ipforward_rt.ro_rt = 0; 1129 } 1130 sin->sin_family = AF_INET; 1131 sin->sin_len = sizeof(*sin); 1132 sin->sin_addr = dst; 1133 1134 rtalloc(&ipforward_rt); 1135 } 1136 if (ipforward_rt.ro_rt == 0) 1137 return ((struct in_ifaddr *)0); 1138 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1139 } 1140 1141 /* 1142 * Save incoming source route for use in replies, 1143 * to be picked up later by ip_srcroute if the receiver is interested. 1144 */ 1145 void 1146 save_rte(option, dst) 1147 u_char *option; 1148 struct in_addr dst; 1149 { 1150 unsigned olen; 1151 1152 olen = option[IPOPT_OLEN]; 1153 #ifdef DIAGNOSTIC 1154 if (ipprintfs) 1155 printf("save_rte: olen %d\n", olen); 1156 #endif /* 0 */ 1157 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1158 return; 1159 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1160 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1161 ip_srcrt.dst = dst; 1162 } 1163 1164 /* 1165 * Retrieve incoming source route for use in replies, 1166 * in the same form used by setsockopt. 1167 * The first hop is placed before the options, will be removed later. 1168 */ 1169 struct mbuf * 1170 ip_srcroute() 1171 { 1172 register struct in_addr *p, *q; 1173 register struct mbuf *m; 1174 1175 if (ip_nhops == 0) 1176 return ((struct mbuf *)0); 1177 m = m_get(M_DONTWAIT, MT_SOOPTS); 1178 if (m == 0) 1179 return ((struct mbuf *)0); 1180 1181 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1182 1183 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1184 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1185 OPTSIZ; 1186 #ifdef DIAGNOSTIC 1187 if (ipprintfs) 1188 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1189 #endif 1190 1191 /* 1192 * First save first hop for return route 1193 */ 1194 p = &ip_srcrt.route[ip_nhops - 1]; 1195 *(mtod(m, struct in_addr *)) = *p--; 1196 #ifdef DIAGNOSTIC 1197 if (ipprintfs) 1198 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1199 #endif 1200 1201 /* 1202 * Copy option fields and padding (nop) to mbuf. 1203 */ 1204 ip_srcrt.nop = IPOPT_NOP; 1205 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1206 bcopy((caddr_t)&ip_srcrt.nop, 1207 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1208 q = (struct in_addr *)(mtod(m, caddr_t) + 1209 sizeof(struct in_addr) + OPTSIZ); 1210 #undef OPTSIZ 1211 /* 1212 * Record return path as an IP source route, 1213 * reversing the path (pointers are now aligned). 1214 */ 1215 while (p >= ip_srcrt.route) { 1216 #ifdef DIAGNOSTIC 1217 if (ipprintfs) 1218 printf(" %x", ntohl(q->s_addr)); 1219 #endif 1220 *q++ = *p--; 1221 } 1222 /* 1223 * Last hop goes to final destination. 1224 */ 1225 *q = ip_srcrt.dst; 1226 #ifdef DIAGNOSTIC 1227 if (ipprintfs) 1228 printf(" %x\n", ntohl(q->s_addr)); 1229 #endif 1230 return (m); 1231 } 1232 1233 /* 1234 * Strip out IP options, at higher 1235 * level protocol in the kernel. 1236 * Second argument is buffer to which options 1237 * will be moved, and return value is their length. 1238 * XXX should be deleted; last arg currently ignored. 1239 */ 1240 void 1241 ip_stripoptions(m, mopt) 1242 register struct mbuf *m; 1243 struct mbuf *mopt; 1244 { 1245 register int i; 1246 struct ip *ip = mtod(m, struct ip *); 1247 register caddr_t opts; 1248 int olen; 1249 1250 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1251 opts = (caddr_t)(ip + 1); 1252 i = m->m_len - (sizeof (struct ip) + olen); 1253 bcopy(opts + olen, opts, (unsigned)i); 1254 m->m_len -= olen; 1255 if (m->m_flags & M_PKTHDR) 1256 m->m_pkthdr.len -= olen; 1257 ip->ip_len -= olen; 1258 ip->ip_hl = sizeof (struct ip) >> 2; 1259 } 1260 1261 int inetctlerrmap[PRC_NCMDS] = { 1262 0, 0, 0, 0, 1263 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1264 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1265 EMSGSIZE, EHOSTUNREACH, 0, 0, 1266 0, 0, 0, 0, 1267 ENOPROTOOPT 1268 }; 1269 1270 /* 1271 * Forward a packet. If some error occurs return the sender 1272 * an icmp packet. Note we can't always generate a meaningful 1273 * icmp message because icmp doesn't have a large enough repertoire 1274 * of codes and types. 1275 * 1276 * If not forwarding, just drop the packet. This could be confusing 1277 * if ipforwarding was zero but some routing protocol was advancing 1278 * us as a gateway to somewhere. However, we must let the routing 1279 * protocol deal with that. 1280 * 1281 * The srcrt parameter indicates whether the packet is being forwarded 1282 * via a source route. 1283 */ 1284 void 1285 ip_forward(m, srcrt) 1286 struct mbuf *m; 1287 int srcrt; 1288 { 1289 register struct ip *ip = mtod(m, struct ip *); 1290 register struct sockaddr_in *sin; 1291 register struct rtentry *rt; 1292 int error, type = 0, code = 0; 1293 struct mbuf *mcopy; 1294 n_long dest; 1295 struct ifnet *destifp; 1296 #ifdef IPSEC 1297 struct ifnet dummyifp; 1298 #endif 1299 1300 dest = 0; 1301 #ifdef DIAGNOSTIC 1302 if (ipprintfs) 1303 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1304 ntohl(ip->ip_src.s_addr), 1305 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1306 #endif 1307 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1308 ipstat.ips_cantforward++; 1309 m_freem(m); 1310 return; 1311 } 1312 if (ip->ip_ttl <= IPTTLDEC) { 1313 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1314 return; 1315 } 1316 ip->ip_ttl -= IPTTLDEC; 1317 1318 sin = satosin(&ipforward_rt.ro_dst); 1319 if ((rt = ipforward_rt.ro_rt) == 0 || 1320 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1321 if (ipforward_rt.ro_rt) { 1322 RTFREE(ipforward_rt.ro_rt); 1323 ipforward_rt.ro_rt = 0; 1324 } 1325 sin->sin_family = AF_INET; 1326 sin->sin_len = sizeof(struct sockaddr_in); 1327 sin->sin_addr = ip->ip_dst; 1328 1329 rtalloc(&ipforward_rt); 1330 if (ipforward_rt.ro_rt == 0) { 1331 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1332 return; 1333 } 1334 rt = ipforward_rt.ro_rt; 1335 } 1336 1337 /* 1338 * Save at most 68 bytes of the packet in case 1339 * we need to generate an ICMP message to the src. 1340 */ 1341 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 68)); 1342 1343 /* 1344 * If forwarding packet using same interface that it came in on, 1345 * perhaps should send a redirect to sender to shortcut a hop. 1346 * Only send redirect if source is sending directly to us, 1347 * and if packet was not source routed (or has any options). 1348 * Also, don't send redirect if forwarding using a default route 1349 * or a route modified by a redirect. 1350 */ 1351 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1352 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1353 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1354 ipsendredirects && !srcrt) { 1355 if (rt->rt_ifa && 1356 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1357 ifatoia(rt->rt_ifa)->ia_subnet) { 1358 if (rt->rt_flags & RTF_GATEWAY) 1359 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1360 else 1361 dest = ip->ip_dst.s_addr; 1362 /* 1363 * Router requirements says to only send host 1364 * redirects. 1365 */ 1366 type = ICMP_REDIRECT; 1367 code = ICMP_REDIRECT_HOST; 1368 #ifdef DIAGNOSTIC 1369 if (ipprintfs) 1370 printf("redirect (%d) to %x\n", code, 1371 (u_int32_t)dest); 1372 #endif 1373 } 1374 } 1375 1376 #ifdef IPSEC 1377 m->m_pkthdr.rcvif = NULL; 1378 #endif /*IPSEC*/ 1379 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1380 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1381 if (error) 1382 ipstat.ips_cantforward++; 1383 else { 1384 ipstat.ips_forward++; 1385 if (type) 1386 ipstat.ips_redirectsent++; 1387 else { 1388 if (mcopy) { 1389 #ifdef GATEWAY 1390 if (mcopy->m_flags & M_CANFASTFWD) 1391 ipflow_create(&ipforward_rt, mcopy); 1392 #endif 1393 m_freem(mcopy); 1394 } 1395 return; 1396 } 1397 } 1398 if (mcopy == NULL) 1399 return; 1400 destifp = NULL; 1401 1402 switch (error) { 1403 1404 case 0: /* forwarded, but need redirect */ 1405 /* type, code set above */ 1406 break; 1407 1408 case ENETUNREACH: /* shouldn't happen, checked above */ 1409 case EHOSTUNREACH: 1410 case ENETDOWN: 1411 case EHOSTDOWN: 1412 default: 1413 type = ICMP_UNREACH; 1414 code = ICMP_UNREACH_HOST; 1415 break; 1416 1417 case EMSGSIZE: 1418 type = ICMP_UNREACH; 1419 code = ICMP_UNREACH_NEEDFRAG; 1420 #ifndef IPSEC 1421 if (ipforward_rt.ro_rt) 1422 destifp = ipforward_rt.ro_rt->rt_ifp; 1423 #else 1424 /* 1425 * If the packet is routed over IPsec tunnel, tell the 1426 * originator the tunnel MTU. 1427 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1428 * XXX quickhack!!! 1429 */ 1430 if (ipforward_rt.ro_rt) { 1431 struct secpolicy *sp; 1432 int ipsecerror; 1433 int ipsechdr; 1434 struct route *ro; 1435 1436 sp = ipsec4_getpolicybyaddr(mcopy, 1437 IP_FORWARDING, 1438 &ipsecerror); 1439 1440 if (sp == NULL) 1441 destifp = ipforward_rt.ro_rt->rt_ifp; 1442 else { 1443 /* count IPsec header size */ 1444 ipsechdr = ipsec4_hdrsiz(mcopy, NULL); 1445 1446 /* 1447 * find the correct route for outer IPv4 1448 * header, compute tunnel MTU. 1449 * 1450 * XXX BUG ALERT 1451 * The "dummyifp" code relies upon the fact 1452 * that icmp_error() touches only ifp->if_mtu. 1453 */ 1454 /*XXX*/ 1455 destifp = NULL; 1456 if (sp->req != NULL 1457 && sp->req->sa != NULL) { 1458 ro = &sp->req->sa->saidx->sa_route; 1459 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1460 dummyifp.if_mtu = 1461 ro->ro_rt->rt_ifp->if_mtu; 1462 dummyifp.if_mtu -= ipsechdr; 1463 destifp = &dummyifp; 1464 } 1465 } 1466 1467 key_freesp(sp); 1468 } 1469 } 1470 #endif /*IPSEC*/ 1471 ipstat.ips_cantfrag++; 1472 break; 1473 1474 case ENOBUFS: 1475 type = ICMP_SOURCEQUENCH; 1476 code = 0; 1477 break; 1478 } 1479 icmp_error(mcopy, type, code, dest, destifp); 1480 } 1481 1482 void 1483 ip_savecontrol(inp, mp, ip, m) 1484 register struct inpcb *inp; 1485 register struct mbuf **mp; 1486 register struct ip *ip; 1487 register struct mbuf *m; 1488 { 1489 1490 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1491 struct timeval tv; 1492 1493 microtime(&tv); 1494 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1495 SCM_TIMESTAMP, SOL_SOCKET); 1496 if (*mp) 1497 mp = &(*mp)->m_next; 1498 } 1499 if (inp->inp_flags & INP_RECVDSTADDR) { 1500 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1501 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1502 if (*mp) 1503 mp = &(*mp)->m_next; 1504 } 1505 #ifdef notyet 1506 /* 1507 * XXX 1508 * Moving these out of udp_input() made them even more broken 1509 * than they already were. 1510 * - fenner@parc.xerox.com 1511 */ 1512 /* options were tossed already */ 1513 if (inp->inp_flags & INP_RECVOPTS) { 1514 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1515 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1516 if (*mp) 1517 mp = &(*mp)->m_next; 1518 } 1519 /* ip_srcroute doesn't do what we want here, need to fix */ 1520 if (inp->inp_flags & INP_RECVRETOPTS) { 1521 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1522 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1523 if (*mp) 1524 mp = &(*mp)->m_next; 1525 } 1526 #endif 1527 if (inp->inp_flags & INP_RECVIF) { 1528 struct sockaddr_dl sdl; 1529 1530 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1531 sdl.sdl_family = AF_LINK; 1532 sdl.sdl_index = m->m_pkthdr.rcvif ? 1533 m->m_pkthdr.rcvif->if_index : 0; 1534 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1535 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1536 IP_RECVIF, IPPROTO_IP); 1537 if (*mp) 1538 mp = &(*mp)->m_next; 1539 } 1540 } 1541 1542 int 1543 ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1544 int *name; 1545 u_int namelen; 1546 void *oldp; 1547 size_t *oldlenp; 1548 void *newp; 1549 size_t newlen; 1550 { 1551 extern int subnetsarelocal, hostzeroisbroadcast; 1552 1553 int error, old; 1554 1555 /* All sysctl names at this level are terminal. */ 1556 if (namelen != 1) 1557 return (ENOTDIR); 1558 1559 switch (name[0]) { 1560 case IPCTL_FORWARDING: 1561 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1562 case IPCTL_SENDREDIRECTS: 1563 return (sysctl_int(oldp, oldlenp, newp, newlen, 1564 &ipsendredirects)); 1565 case IPCTL_DEFTTL: 1566 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1567 #ifdef notyet 1568 case IPCTL_DEFMTU: 1569 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1570 #endif 1571 case IPCTL_FORWSRCRT: 1572 /* Don't allow this to change in a secure environment. */ 1573 if (securelevel > 0) 1574 return (sysctl_rdint(oldp, oldlenp, newp, 1575 ip_forwsrcrt)); 1576 else 1577 return (sysctl_int(oldp, oldlenp, newp, newlen, 1578 &ip_forwsrcrt)); 1579 case IPCTL_DIRECTEDBCAST: 1580 return (sysctl_int(oldp, oldlenp, newp, newlen, 1581 &ip_directedbcast)); 1582 case IPCTL_ALLOWSRCRT: 1583 return (sysctl_int(oldp, oldlenp, newp, newlen, 1584 &ip_allowsrcrt)); 1585 case IPCTL_SUBNETSARELOCAL: 1586 return (sysctl_int(oldp, oldlenp, newp, newlen, 1587 &subnetsarelocal)); 1588 case IPCTL_MTUDISC: 1589 error = sysctl_int(oldp, oldlenp, newp, newlen, 1590 &ip_mtudisc); 1591 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1592 ip_mtudisc_timeout_q = 1593 rt_timer_queue_create(ip_mtudisc_timeout); 1594 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1595 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1596 ip_mtudisc_timeout_q = NULL; 1597 } 1598 return error; 1599 case IPCTL_ANONPORTMIN: 1600 old = anonportmin; 1601 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1602 if (anonportmin >= anonportmax || anonportmin > 65535 1603 #ifndef IPNOPRIVPORTS 1604 || anonportmin < IPPORT_RESERVED 1605 #endif 1606 ) { 1607 anonportmin = old; 1608 return (EINVAL); 1609 } 1610 return (error); 1611 case IPCTL_ANONPORTMAX: 1612 old = anonportmax; 1613 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1614 if (anonportmin >= anonportmax || anonportmax > 65535 1615 #ifndef IPNOPRIVPORTS 1616 || anonportmax < IPPORT_RESERVED 1617 #endif 1618 ) { 1619 anonportmax = old; 1620 return (EINVAL); 1621 } 1622 return (error); 1623 case IPCTL_MTUDISCTIMEOUT: 1624 error = sysctl_int(oldp, oldlenp, newp, newlen, 1625 &ip_mtudisc_timeout); 1626 if (ip_mtudisc_timeout_q != NULL) 1627 rt_timer_queue_change(ip_mtudisc_timeout_q, 1628 ip_mtudisc_timeout); 1629 return (error); 1630 #ifdef GATEWAY 1631 case IPCTL_MAXFLOWS: 1632 { 1633 int s; 1634 1635 error = sysctl_int(oldp, oldlenp, newp, newlen, 1636 &ip_maxflows); 1637 s = splsoftnet(); 1638 ipflow_reap(0); 1639 splx(s); 1640 return (error); 1641 } 1642 #endif 1643 case IPCTL_HOSTZEROBROADCAST: 1644 return (sysctl_int(oldp, oldlenp, newp, newlen, 1645 &hostzeroisbroadcast)); 1646 #if NGIF > 0 1647 case IPCTL_GIF_TTL: 1648 return(sysctl_int(oldp, oldlenp, newp, newlen, 1649 &ip_gif_ttl)); 1650 #endif 1651 1652 default: 1653 return (EOPNOTSUPP); 1654 } 1655 /* NOTREACHED */ 1656 } 1657