1 /* $NetBSD: ip_input.c,v 1.114 2000/05/10 03:31:30 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 MROUTING 143 #include <netinet/ip_mroute.h> 144 #endif 145 146 #ifdef IPSEC 147 #include <netinet6/ipsec.h> 148 #include <netkey/key.h> 149 #include <netkey/key_debug.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 struct ipqhead ipq; 209 int ipq_locked; 210 211 static __inline int ipq_lock_try __P((void)); 212 static __inline void ipq_unlock __P((void)); 213 214 static __inline int 215 ipq_lock_try() 216 { 217 int s; 218 219 s = splimp(); 220 if (ipq_locked) { 221 splx(s); 222 return (0); 223 } 224 ipq_locked = 1; 225 splx(s); 226 return (1); 227 } 228 229 static __inline void 230 ipq_unlock() 231 { 232 int s; 233 234 s = splimp(); 235 ipq_locked = 0; 236 splx(s); 237 } 238 239 #ifdef DIAGNOSTIC 240 #define IPQ_LOCK() \ 241 do { \ 242 if (ipq_lock_try() == 0) { \ 243 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 244 panic("ipq_lock"); \ 245 } \ 246 } while (0) 247 #define IPQ_LOCK_CHECK() \ 248 do { \ 249 if (ipq_locked == 0) { \ 250 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 251 panic("ipq lock check"); \ 252 } \ 253 } while (0) 254 #else 255 #define IPQ_LOCK() (void) ipq_lock_try() 256 #define IPQ_LOCK_CHECK() /* nothing */ 257 #endif 258 259 #define IPQ_UNLOCK() ipq_unlock() 260 261 struct pool ipqent_pool; 262 263 /* 264 * We need to save the IP options in case a protocol wants to respond 265 * to an incoming packet over the same route if the packet got here 266 * using IP source routing. This allows connection establishment and 267 * maintenance when the remote end is on a network that is not known 268 * to us. 269 */ 270 int ip_nhops = 0; 271 static struct ip_srcrt { 272 struct in_addr dst; /* final destination */ 273 char nop; /* one NOP to align */ 274 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 275 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 276 } ip_srcrt; 277 278 static void save_rte __P((u_char *, struct in_addr)); 279 280 /* 281 * IP initialization: fill in IP protocol switch table. 282 * All protocols not implemented in kernel go to raw IP protocol handler. 283 */ 284 void 285 ip_init() 286 { 287 struct protosw *pr; 288 int i; 289 290 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 291 0, NULL, NULL, M_IPQ); 292 293 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 294 if (pr == 0) 295 panic("ip_init"); 296 for (i = 0; i < IPPROTO_MAX; i++) 297 ip_protox[i] = pr - inetsw; 298 for (pr = inetdomain.dom_protosw; 299 pr < inetdomain.dom_protoswNPROTOSW; pr++) 300 if (pr->pr_domain->dom_family == PF_INET && 301 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 302 ip_protox[pr->pr_protocol] = pr - inetsw; 303 LIST_INIT(&ipq); 304 ip_id = time.tv_sec & 0xffff; 305 ipintrq.ifq_maxlen = ipqmaxlen; 306 TAILQ_INIT(&in_ifaddr); 307 in_ifaddrhashtbl = 308 hashinit(IN_IFADDR_HASH_SIZE, M_IFADDR, M_WAITOK, &in_ifaddrhash); 309 if (ip_mtudisc != 0) 310 ip_mtudisc_timeout_q = 311 rt_timer_queue_create(ip_mtudisc_timeout); 312 #ifdef GATEWAY 313 ipflow_init(); 314 #endif 315 } 316 317 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 318 struct route ipforward_rt; 319 320 /* 321 * IP software interrupt routine 322 */ 323 void 324 ipintr() 325 { 326 int s; 327 struct mbuf *m; 328 329 while (1) { 330 s = splimp(); 331 IF_DEQUEUE(&ipintrq, m); 332 splx(s); 333 if (m == 0) 334 return; 335 ip_input(m); 336 } 337 } 338 339 /* 340 * Ip input routine. Checksum and byte swap header. If fragmented 341 * try to reassemble. Process options. Pass to next level. 342 */ 343 void 344 ip_input(struct mbuf *m) 345 { 346 struct ip *ip = NULL; 347 struct ipq *fp; 348 struct in_ifaddr *ia; 349 struct ifaddr *ifa; 350 struct ipqent *ipqe; 351 int hlen = 0, mff, len; 352 int downmatch; 353 #ifdef PFIL_HOOKS 354 struct packet_filter_hook *pfh; 355 struct mbuf *m0; 356 int rv; 357 #endif /* PFIL_HOOKS */ 358 359 #ifdef DIAGNOSTIC 360 if ((m->m_flags & M_PKTHDR) == 0) 361 panic("ipintr no HDR"); 362 #endif 363 #ifdef IPSEC 364 /* 365 * should the inner packet be considered authentic? 366 * see comment in ah4_input(). 367 */ 368 if (m) { 369 m->m_flags &= ~M_AUTHIPHDR; 370 m->m_flags &= ~M_AUTHIPDGM; 371 } 372 #endif 373 /* 374 * If no IP addresses have been set yet but the interfaces 375 * are receiving, can't do anything with incoming packets yet. 376 */ 377 if (in_ifaddr.tqh_first == 0) 378 goto bad; 379 ipstat.ips_total++; 380 if (m->m_len < sizeof (struct ip) && 381 (m = m_pullup(m, sizeof (struct ip))) == 0) { 382 ipstat.ips_toosmall++; 383 return; 384 } 385 ip = mtod(m, struct ip *); 386 if (ip->ip_v != IPVERSION) { 387 ipstat.ips_badvers++; 388 goto bad; 389 } 390 hlen = ip->ip_hl << 2; 391 if (hlen < sizeof(struct ip)) { /* minimum header length */ 392 ipstat.ips_badhlen++; 393 goto bad; 394 } 395 if (hlen > m->m_len) { 396 if ((m = m_pullup(m, hlen)) == 0) { 397 ipstat.ips_badhlen++; 398 return; 399 } 400 ip = mtod(m, struct ip *); 401 } 402 403 /* 404 * RFC1122: packets with a multicast source address are 405 * not allowed. 406 */ 407 if (IN_MULTICAST(ip->ip_src.s_addr)) { 408 /* XXX stat */ 409 goto bad; 410 } 411 412 if (in_cksum(m, hlen) != 0) { 413 ipstat.ips_badsum++; 414 goto bad; 415 } 416 417 /* 418 * Convert fields to host representation. 419 */ 420 NTOHS(ip->ip_len); 421 NTOHS(ip->ip_off); 422 len = ip->ip_len; 423 424 /* 425 * Check for additional length bogosity 426 */ 427 if (len < hlen) { 428 ipstat.ips_badlen++; 429 goto bad; 430 } 431 432 /* 433 * Check that the amount of data in the buffers 434 * is as at least much as the IP header would have us expect. 435 * Trim mbufs if longer than we expect. 436 * Drop packet if shorter than we expect. 437 */ 438 if (m->m_pkthdr.len < len) { 439 ipstat.ips_tooshort++; 440 goto bad; 441 } 442 if (m->m_pkthdr.len > len) { 443 if (m->m_len == m->m_pkthdr.len) { 444 m->m_len = len; 445 m->m_pkthdr.len = len; 446 } else 447 m_adj(m, len - m->m_pkthdr.len); 448 } 449 450 #ifdef IPSEC 451 /* ipflow (IP fast fowarding) is not compatible with IPsec. */ 452 m->m_flags &= ~M_CANFASTFWD; 453 #else 454 /* 455 * Assume that we can create a fast-forward IP flow entry 456 * based on this packet. 457 */ 458 m->m_flags |= M_CANFASTFWD; 459 #endif 460 461 #ifdef PFIL_HOOKS 462 /* 463 * Run through list of hooks for input packets. If there are any 464 * filters which require that additional packets in the flow are 465 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 466 * Note that filters must _never_ set this flag, as another filter 467 * in the list may have previously cleared it. 468 */ 469 m0 = m; 470 pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 471 for (; pfh; pfh = pfh->pfil_link.tqe_next) 472 if (pfh->pfil_func) { 473 rv = pfh->pfil_func(ip, hlen, 474 m->m_pkthdr.rcvif, 0, &m0); 475 if (rv) 476 return; 477 m = m0; 478 if (m == NULL) 479 return; 480 ip = mtod(m, struct ip *); 481 } 482 #endif /* PFIL_HOOKS */ 483 484 /* 485 * Process options and, if not destined for us, 486 * ship it on. ip_dooptions returns 1 when an 487 * error was detected (causing an icmp message 488 * to be sent and the original packet to be freed). 489 */ 490 ip_nhops = 0; /* for source routed packets */ 491 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 492 return; 493 494 /* 495 * Check our list of addresses, to see if the packet is for us. 496 * 497 * Traditional 4.4BSD did not consult IFF_UP at all. 498 * The behavior here is to treat addresses on !IFF_UP interface 499 * as not mine. 500 */ 501 downmatch = 0; 502 for (ia = IN_IFADDR_HASH(ip->ip_dst.s_addr).lh_first; 503 ia != NULL; 504 ia = ia->ia_hash.le_next) { 505 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 506 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 507 break; 508 else 509 downmatch++; 510 } 511 } 512 if (ia != NULL) 513 goto ours; 514 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 515 for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first; 516 ifa != NULL; ifa = ifa->ifa_list.tqe_next) { 517 if (ifa->ifa_addr->sa_family != AF_INET) continue; 518 ia = ifatoia(ifa); 519 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 520 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 521 /* 522 * Look for all-0's host part (old broadcast addr), 523 * either for subnet or net. 524 */ 525 ip->ip_dst.s_addr == ia->ia_subnet || 526 ip->ip_dst.s_addr == ia->ia_net) 527 goto ours; 528 /* 529 * An interface with IP address zero accepts 530 * all packets that arrive on that interface. 531 */ 532 if (in_nullhost(ia->ia_addr.sin_addr)) 533 goto ours; 534 } 535 } 536 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 537 struct in_multi *inm; 538 #ifdef MROUTING 539 extern struct socket *ip_mrouter; 540 541 if (m->m_flags & M_EXT) { 542 if ((m = m_pullup(m, hlen)) == 0) { 543 ipstat.ips_toosmall++; 544 return; 545 } 546 ip = mtod(m, struct ip *); 547 } 548 549 if (ip_mrouter) { 550 /* 551 * If we are acting as a multicast router, all 552 * incoming multicast packets are passed to the 553 * kernel-level multicast forwarding function. 554 * The packet is returned (relatively) intact; if 555 * ip_mforward() returns a non-zero value, the packet 556 * must be discarded, else it may be accepted below. 557 * 558 * (The IP ident field is put in the same byte order 559 * as expected when ip_mforward() is called from 560 * ip_output().) 561 */ 562 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 563 ipstat.ips_cantforward++; 564 m_freem(m); 565 return; 566 } 567 568 /* 569 * The process-level routing demon needs to receive 570 * all multicast IGMP packets, whether or not this 571 * host belongs to their destination groups. 572 */ 573 if (ip->ip_p == IPPROTO_IGMP) 574 goto ours; 575 ipstat.ips_forward++; 576 } 577 #endif 578 /* 579 * See if we belong to the destination multicast group on the 580 * arrival interface. 581 */ 582 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 583 if (inm == NULL) { 584 ipstat.ips_cantforward++; 585 m_freem(m); 586 return; 587 } 588 goto ours; 589 } 590 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 591 in_nullhost(ip->ip_dst)) 592 goto ours; 593 594 /* 595 * Not for us; forward if possible and desirable. 596 */ 597 if (ipforwarding == 0) { 598 ipstat.ips_cantforward++; 599 m_freem(m); 600 } else { 601 /* 602 * If ip_dst matched any of my address on !IFF_UP interface, 603 * and there's no IFF_UP interface that matches ip_dst, 604 * send icmp unreach. Forwarding it will result in in-kernel 605 * forwarding loop till TTL goes to 0. 606 */ 607 if (downmatch) { 608 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 609 ipstat.ips_cantforward++; 610 return; 611 } 612 ip_forward(m, 0); 613 } 614 return; 615 616 ours: 617 /* 618 * If offset or IP_MF are set, must reassemble. 619 * Otherwise, nothing need be done. 620 * (We could look in the reassembly queue to see 621 * if the packet was previously fragmented, 622 * but it's not worth the time; just let them time out.) 623 */ 624 if (ip->ip_off & ~(IP_DF|IP_RF)) { 625 /* 626 * Look for queue of fragments 627 * of this datagram. 628 */ 629 IPQ_LOCK(); 630 for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next) 631 if (ip->ip_id == fp->ipq_id && 632 in_hosteq(ip->ip_src, fp->ipq_src) && 633 in_hosteq(ip->ip_dst, fp->ipq_dst) && 634 ip->ip_p == fp->ipq_p) 635 goto found; 636 fp = 0; 637 found: 638 639 /* 640 * Adjust ip_len to not reflect header, 641 * set ipqe_mff if more fragments are expected, 642 * convert offset of this to bytes. 643 */ 644 ip->ip_len -= hlen; 645 mff = (ip->ip_off & IP_MF) != 0; 646 if (mff) { 647 /* 648 * Make sure that fragments have a data length 649 * that's a non-zero multiple of 8 bytes. 650 */ 651 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 652 ipstat.ips_badfrags++; 653 IPQ_UNLOCK(); 654 goto bad; 655 } 656 } 657 ip->ip_off <<= 3; 658 659 /* 660 * If datagram marked as having more fragments 661 * or if this is not the first fragment, 662 * attempt reassembly; if it succeeds, proceed. 663 */ 664 if (mff || ip->ip_off) { 665 ipstat.ips_fragments++; 666 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 667 if (ipqe == NULL) { 668 ipstat.ips_rcvmemdrop++; 669 IPQ_UNLOCK(); 670 goto bad; 671 } 672 ipqe->ipqe_mff = mff; 673 ipqe->ipqe_m = m; 674 ipqe->ipqe_ip = ip; 675 m = ip_reass(ipqe, fp); 676 if (m == 0) { 677 IPQ_UNLOCK(); 678 return; 679 } 680 ipstat.ips_reassembled++; 681 ip = mtod(m, struct ip *); 682 hlen = ip->ip_hl << 2; 683 ip->ip_len += hlen; 684 } else 685 if (fp) 686 ip_freef(fp); 687 IPQ_UNLOCK(); 688 } 689 690 /* 691 * Switch out to protocol's input routine. 692 */ 693 #if IFA_STATS 694 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 695 #endif 696 ipstat.ips_delivered++; 697 { 698 int off = hlen, nh = ip->ip_p; 699 700 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 701 return; 702 } 703 bad: 704 m_freem(m); 705 } 706 707 /* 708 * Take incoming datagram fragment and try to 709 * reassemble it into whole datagram. If a chain for 710 * reassembly of this datagram already exists, then it 711 * is given as fp; otherwise have to make a chain. 712 */ 713 struct mbuf * 714 ip_reass(ipqe, fp) 715 struct ipqent *ipqe; 716 struct ipq *fp; 717 { 718 struct mbuf *m = ipqe->ipqe_m; 719 struct ipqent *nq, *p, *q; 720 struct ip *ip; 721 struct mbuf *t; 722 int hlen = ipqe->ipqe_ip->ip_hl << 2; 723 int i, next; 724 725 IPQ_LOCK_CHECK(); 726 727 /* 728 * Presence of header sizes in mbufs 729 * would confuse code below. 730 */ 731 m->m_data += hlen; 732 m->m_len -= hlen; 733 734 /* 735 * If first fragment to arrive, create a reassembly queue. 736 */ 737 if (fp == 0) { 738 MALLOC(fp, struct ipq *, sizeof (struct ipq), 739 M_FTABLE, M_NOWAIT); 740 if (fp == NULL) 741 goto dropfrag; 742 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 743 fp->ipq_ttl = IPFRAGTTL; 744 fp->ipq_p = ipqe->ipqe_ip->ip_p; 745 fp->ipq_id = ipqe->ipqe_ip->ip_id; 746 LIST_INIT(&fp->ipq_fragq); 747 fp->ipq_src = ipqe->ipqe_ip->ip_src; 748 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 749 p = NULL; 750 goto insert; 751 } 752 753 /* 754 * Find a segment which begins after this one does. 755 */ 756 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 757 p = q, q = q->ipqe_q.le_next) 758 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 759 break; 760 761 /* 762 * If there is a preceding segment, it may provide some of 763 * our data already. If so, drop the data from the incoming 764 * segment. If it provides all of our data, drop us. 765 */ 766 if (p != NULL) { 767 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 768 ipqe->ipqe_ip->ip_off; 769 if (i > 0) { 770 if (i >= ipqe->ipqe_ip->ip_len) 771 goto dropfrag; 772 m_adj(ipqe->ipqe_m, i); 773 ipqe->ipqe_ip->ip_off += i; 774 ipqe->ipqe_ip->ip_len -= i; 775 } 776 } 777 778 /* 779 * While we overlap succeeding segments trim them or, 780 * if they are completely covered, dequeue them. 781 */ 782 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 783 q->ipqe_ip->ip_off; q = nq) { 784 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 785 q->ipqe_ip->ip_off; 786 if (i < q->ipqe_ip->ip_len) { 787 q->ipqe_ip->ip_len -= i; 788 q->ipqe_ip->ip_off += i; 789 m_adj(q->ipqe_m, i); 790 break; 791 } 792 nq = q->ipqe_q.le_next; 793 m_freem(q->ipqe_m); 794 LIST_REMOVE(q, ipqe_q); 795 pool_put(&ipqent_pool, q); 796 } 797 798 insert: 799 /* 800 * Stick new segment in its place; 801 * check for complete reassembly. 802 */ 803 if (p == NULL) { 804 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 805 } else { 806 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 807 } 808 next = 0; 809 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 810 p = q, q = q->ipqe_q.le_next) { 811 if (q->ipqe_ip->ip_off != next) 812 return (0); 813 next += q->ipqe_ip->ip_len; 814 } 815 if (p->ipqe_mff) 816 return (0); 817 818 /* 819 * Reassembly is complete. Check for a bogus message size and 820 * concatenate fragments. 821 */ 822 q = fp->ipq_fragq.lh_first; 823 ip = q->ipqe_ip; 824 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 825 ipstat.ips_toolong++; 826 ip_freef(fp); 827 return (0); 828 } 829 m = q->ipqe_m; 830 t = m->m_next; 831 m->m_next = 0; 832 m_cat(m, t); 833 nq = q->ipqe_q.le_next; 834 pool_put(&ipqent_pool, q); 835 for (q = nq; q != NULL; q = nq) { 836 t = q->ipqe_m; 837 nq = q->ipqe_q.le_next; 838 pool_put(&ipqent_pool, q); 839 m_cat(m, t); 840 } 841 842 /* 843 * Create header for new ip packet by 844 * modifying header of first packet; 845 * dequeue and discard fragment reassembly header. 846 * Make header visible. 847 */ 848 ip->ip_len = next; 849 ip->ip_src = fp->ipq_src; 850 ip->ip_dst = fp->ipq_dst; 851 LIST_REMOVE(fp, ipq_q); 852 FREE(fp, M_FTABLE); 853 m->m_len += (ip->ip_hl << 2); 854 m->m_data -= (ip->ip_hl << 2); 855 /* some debugging cruft by sklower, below, will go away soon */ 856 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 857 int plen = 0; 858 for (t = m; t; t = t->m_next) 859 plen += t->m_len; 860 m->m_pkthdr.len = plen; 861 } 862 return (m); 863 864 dropfrag: 865 ipstat.ips_fragdropped++; 866 m_freem(m); 867 pool_put(&ipqent_pool, ipqe); 868 return (0); 869 } 870 871 /* 872 * Free a fragment reassembly header and all 873 * associated datagrams. 874 */ 875 void 876 ip_freef(fp) 877 struct ipq *fp; 878 { 879 struct ipqent *q, *p; 880 881 IPQ_LOCK_CHECK(); 882 883 for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { 884 p = q->ipqe_q.le_next; 885 m_freem(q->ipqe_m); 886 LIST_REMOVE(q, ipqe_q); 887 pool_put(&ipqent_pool, q); 888 } 889 LIST_REMOVE(fp, ipq_q); 890 FREE(fp, M_FTABLE); 891 } 892 893 /* 894 * IP timer processing; 895 * if a timer expires on a reassembly 896 * queue, discard it. 897 */ 898 void 899 ip_slowtimo() 900 { 901 struct ipq *fp, *nfp; 902 int s = splsoftnet(); 903 904 IPQ_LOCK(); 905 for (fp = ipq.lh_first; fp != NULL; fp = nfp) { 906 nfp = fp->ipq_q.le_next; 907 if (--fp->ipq_ttl == 0) { 908 ipstat.ips_fragtimeout++; 909 ip_freef(fp); 910 } 911 } 912 IPQ_UNLOCK(); 913 #ifdef GATEWAY 914 ipflow_slowtimo(); 915 #endif 916 splx(s); 917 } 918 919 /* 920 * Drain off all datagram fragments. 921 */ 922 void 923 ip_drain() 924 { 925 926 /* 927 * We may be called from a device's interrupt context. If 928 * the ipq is already busy, just bail out now. 929 */ 930 if (ipq_lock_try() == 0) 931 return; 932 933 while (ipq.lh_first != NULL) { 934 ipstat.ips_fragdropped++; 935 ip_freef(ipq.lh_first); 936 } 937 938 IPQ_UNLOCK(); 939 } 940 941 /* 942 * Do option processing on a datagram, 943 * possibly discarding it if bad options are encountered, 944 * or forwarding it if source-routed. 945 * Returns 1 if packet has been forwarded/freed, 946 * 0 if the packet should be processed further. 947 */ 948 int 949 ip_dooptions(m) 950 struct mbuf *m; 951 { 952 struct ip *ip = mtod(m, struct ip *); 953 u_char *cp, *cp0; 954 struct ip_timestamp *ipt; 955 struct in_ifaddr *ia; 956 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 957 struct in_addr dst; 958 n_time ntime; 959 960 dst = ip->ip_dst; 961 cp = (u_char *)(ip + 1); 962 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 963 for (; cnt > 0; cnt -= optlen, cp += optlen) { 964 opt = cp[IPOPT_OPTVAL]; 965 if (opt == IPOPT_EOL) 966 break; 967 if (opt == IPOPT_NOP) 968 optlen = 1; 969 else { 970 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 971 code = &cp[IPOPT_OLEN] - (u_char *)ip; 972 goto bad; 973 } 974 optlen = cp[IPOPT_OLEN]; 975 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 976 code = &cp[IPOPT_OLEN] - (u_char *)ip; 977 goto bad; 978 } 979 } 980 switch (opt) { 981 982 default: 983 break; 984 985 /* 986 * Source routing with record. 987 * Find interface with current destination address. 988 * If none on this machine then drop if strictly routed, 989 * or do nothing if loosely routed. 990 * Record interface address and bring up next address 991 * component. If strictly routed make sure next 992 * address is on directly accessible net. 993 */ 994 case IPOPT_LSRR: 995 case IPOPT_SSRR: 996 if (ip_allowsrcrt == 0) { 997 type = ICMP_UNREACH; 998 code = ICMP_UNREACH_NET_PROHIB; 999 goto bad; 1000 } 1001 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1002 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1003 goto bad; 1004 } 1005 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1006 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1007 goto bad; 1008 } 1009 ipaddr.sin_addr = ip->ip_dst; 1010 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1011 if (ia == 0) { 1012 if (opt == IPOPT_SSRR) { 1013 type = ICMP_UNREACH; 1014 code = ICMP_UNREACH_SRCFAIL; 1015 goto bad; 1016 } 1017 /* 1018 * Loose routing, and not at next destination 1019 * yet; nothing to do except forward. 1020 */ 1021 break; 1022 } 1023 off--; /* 0 origin */ 1024 if ((off + sizeof(struct in_addr)) > optlen) { 1025 /* 1026 * End of source route. Should be for us. 1027 */ 1028 save_rte(cp, ip->ip_src); 1029 break; 1030 } 1031 /* 1032 * locate outgoing interface 1033 */ 1034 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1035 sizeof(ipaddr.sin_addr)); 1036 if (opt == IPOPT_SSRR) 1037 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1038 else 1039 ia = ip_rtaddr(ipaddr.sin_addr); 1040 if (ia == 0) { 1041 type = ICMP_UNREACH; 1042 code = ICMP_UNREACH_SRCFAIL; 1043 goto bad; 1044 } 1045 ip->ip_dst = ipaddr.sin_addr; 1046 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1047 (caddr_t)(cp + off), sizeof(struct in_addr)); 1048 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1049 /* 1050 * Let ip_intr's mcast routing check handle mcast pkts 1051 */ 1052 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1053 break; 1054 1055 case IPOPT_RR: 1056 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1057 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1058 goto bad; 1059 } 1060 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1061 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1062 goto bad; 1063 } 1064 /* 1065 * If no space remains, ignore. 1066 */ 1067 off--; /* 0 origin */ 1068 if ((off + sizeof(struct in_addr)) > optlen) 1069 break; 1070 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1071 sizeof(ipaddr.sin_addr)); 1072 /* 1073 * locate outgoing interface; if we're the destination, 1074 * use the incoming interface (should be same). 1075 */ 1076 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1077 == NULL && 1078 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1079 type = ICMP_UNREACH; 1080 code = ICMP_UNREACH_HOST; 1081 goto bad; 1082 } 1083 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1084 (caddr_t)(cp + off), sizeof(struct in_addr)); 1085 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1086 break; 1087 1088 case IPOPT_TS: 1089 code = cp - (u_char *)ip; 1090 ipt = (struct ip_timestamp *)cp; 1091 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1092 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1093 goto bad; 1094 } 1095 if (ipt->ipt_ptr < 5) { 1096 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1097 goto bad; 1098 } 1099 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1100 if (++ipt->ipt_oflw == 0) { 1101 code = (u_char *)&ipt->ipt_ptr - 1102 (u_char *)ip; 1103 goto bad; 1104 } 1105 break; 1106 } 1107 cp0 = (cp + ipt->ipt_ptr - 1); 1108 switch (ipt->ipt_flg) { 1109 1110 case IPOPT_TS_TSONLY: 1111 break; 1112 1113 case IPOPT_TS_TSANDADDR: 1114 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1115 sizeof(struct in_addr) > ipt->ipt_len) { 1116 code = (u_char *)&ipt->ipt_ptr - 1117 (u_char *)ip; 1118 goto bad; 1119 } 1120 ipaddr.sin_addr = dst; 1121 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1122 m->m_pkthdr.rcvif)); 1123 if (ia == 0) 1124 continue; 1125 bcopy(&ia->ia_addr.sin_addr, 1126 cp0, sizeof(struct in_addr)); 1127 ipt->ipt_ptr += sizeof(struct in_addr); 1128 break; 1129 1130 case IPOPT_TS_PRESPEC: 1131 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1132 sizeof(struct in_addr) > ipt->ipt_len) { 1133 code = (u_char *)&ipt->ipt_ptr - 1134 (u_char *)ip; 1135 goto bad; 1136 } 1137 bcopy(cp0, &ipaddr.sin_addr, 1138 sizeof(struct in_addr)); 1139 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1140 == NULL) 1141 continue; 1142 ipt->ipt_ptr += sizeof(struct in_addr); 1143 break; 1144 1145 default: 1146 /* XXX can't take &ipt->ipt_flg */ 1147 code = (u_char *)&ipt->ipt_ptr - 1148 (u_char *)ip + 1; 1149 goto bad; 1150 } 1151 ntime = iptime(); 1152 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1153 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1154 sizeof(n_time)); 1155 ipt->ipt_ptr += sizeof(n_time); 1156 } 1157 } 1158 if (forward) { 1159 if (ip_forwsrcrt == 0) { 1160 type = ICMP_UNREACH; 1161 code = ICMP_UNREACH_SRCFAIL; 1162 goto bad; 1163 } 1164 ip_forward(m, 1); 1165 return (1); 1166 } 1167 return (0); 1168 bad: 1169 icmp_error(m, type, code, 0, 0); 1170 ipstat.ips_badoptions++; 1171 return (1); 1172 } 1173 1174 /* 1175 * Given address of next destination (final or next hop), 1176 * return internet address info of interface to be used to get there. 1177 */ 1178 struct in_ifaddr * 1179 ip_rtaddr(dst) 1180 struct in_addr dst; 1181 { 1182 struct sockaddr_in *sin; 1183 1184 sin = satosin(&ipforward_rt.ro_dst); 1185 1186 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1187 if (ipforward_rt.ro_rt) { 1188 RTFREE(ipforward_rt.ro_rt); 1189 ipforward_rt.ro_rt = 0; 1190 } 1191 sin->sin_family = AF_INET; 1192 sin->sin_len = sizeof(*sin); 1193 sin->sin_addr = dst; 1194 1195 rtalloc(&ipforward_rt); 1196 } 1197 if (ipforward_rt.ro_rt == 0) 1198 return ((struct in_ifaddr *)0); 1199 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1200 } 1201 1202 /* 1203 * Save incoming source route for use in replies, 1204 * to be picked up later by ip_srcroute if the receiver is interested. 1205 */ 1206 void 1207 save_rte(option, dst) 1208 u_char *option; 1209 struct in_addr dst; 1210 { 1211 unsigned olen; 1212 1213 olen = option[IPOPT_OLEN]; 1214 #ifdef DIAGNOSTIC 1215 if (ipprintfs) 1216 printf("save_rte: olen %d\n", olen); 1217 #endif /* 0 */ 1218 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1219 return; 1220 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1221 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1222 ip_srcrt.dst = dst; 1223 } 1224 1225 /* 1226 * Retrieve incoming source route for use in replies, 1227 * in the same form used by setsockopt. 1228 * The first hop is placed before the options, will be removed later. 1229 */ 1230 struct mbuf * 1231 ip_srcroute() 1232 { 1233 struct in_addr *p, *q; 1234 struct mbuf *m; 1235 1236 if (ip_nhops == 0) 1237 return ((struct mbuf *)0); 1238 m = m_get(M_DONTWAIT, MT_SOOPTS); 1239 if (m == 0) 1240 return ((struct mbuf *)0); 1241 1242 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1243 1244 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1245 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1246 OPTSIZ; 1247 #ifdef DIAGNOSTIC 1248 if (ipprintfs) 1249 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1250 #endif 1251 1252 /* 1253 * First save first hop for return route 1254 */ 1255 p = &ip_srcrt.route[ip_nhops - 1]; 1256 *(mtod(m, struct in_addr *)) = *p--; 1257 #ifdef DIAGNOSTIC 1258 if (ipprintfs) 1259 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1260 #endif 1261 1262 /* 1263 * Copy option fields and padding (nop) to mbuf. 1264 */ 1265 ip_srcrt.nop = IPOPT_NOP; 1266 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1267 bcopy((caddr_t)&ip_srcrt.nop, 1268 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1269 q = (struct in_addr *)(mtod(m, caddr_t) + 1270 sizeof(struct in_addr) + OPTSIZ); 1271 #undef OPTSIZ 1272 /* 1273 * Record return path as an IP source route, 1274 * reversing the path (pointers are now aligned). 1275 */ 1276 while (p >= ip_srcrt.route) { 1277 #ifdef DIAGNOSTIC 1278 if (ipprintfs) 1279 printf(" %x", ntohl(q->s_addr)); 1280 #endif 1281 *q++ = *p--; 1282 } 1283 /* 1284 * Last hop goes to final destination. 1285 */ 1286 *q = ip_srcrt.dst; 1287 #ifdef DIAGNOSTIC 1288 if (ipprintfs) 1289 printf(" %x\n", ntohl(q->s_addr)); 1290 #endif 1291 return (m); 1292 } 1293 1294 /* 1295 * Strip out IP options, at higher 1296 * level protocol in the kernel. 1297 * Second argument is buffer to which options 1298 * will be moved, and return value is their length. 1299 * XXX should be deleted; last arg currently ignored. 1300 */ 1301 void 1302 ip_stripoptions(m, mopt) 1303 struct mbuf *m; 1304 struct mbuf *mopt; 1305 { 1306 int i; 1307 struct ip *ip = mtod(m, struct ip *); 1308 caddr_t opts; 1309 int olen; 1310 1311 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1312 opts = (caddr_t)(ip + 1); 1313 i = m->m_len - (sizeof (struct ip) + olen); 1314 bcopy(opts + olen, opts, (unsigned)i); 1315 m->m_len -= olen; 1316 if (m->m_flags & M_PKTHDR) 1317 m->m_pkthdr.len -= olen; 1318 ip->ip_len -= olen; 1319 ip->ip_hl = sizeof (struct ip) >> 2; 1320 } 1321 1322 int inetctlerrmap[PRC_NCMDS] = { 1323 0, 0, 0, 0, 1324 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1325 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1326 EMSGSIZE, EHOSTUNREACH, 0, 0, 1327 0, 0, 0, 0, 1328 ENOPROTOOPT 1329 }; 1330 1331 /* 1332 * Forward a packet. If some error occurs return the sender 1333 * an icmp packet. Note we can't always generate a meaningful 1334 * icmp message because icmp doesn't have a large enough repertoire 1335 * of codes and types. 1336 * 1337 * If not forwarding, just drop the packet. This could be confusing 1338 * if ipforwarding was zero but some routing protocol was advancing 1339 * us as a gateway to somewhere. However, we must let the routing 1340 * protocol deal with that. 1341 * 1342 * The srcrt parameter indicates whether the packet is being forwarded 1343 * via a source route. 1344 */ 1345 void 1346 ip_forward(m, srcrt) 1347 struct mbuf *m; 1348 int srcrt; 1349 { 1350 struct ip *ip = mtod(m, struct ip *); 1351 struct sockaddr_in *sin; 1352 struct rtentry *rt; 1353 int error, type = 0, code = 0; 1354 struct mbuf *mcopy; 1355 n_long dest; 1356 struct ifnet *destifp; 1357 #ifdef IPSEC 1358 struct ifnet dummyifp; 1359 #endif 1360 1361 dest = 0; 1362 #ifdef DIAGNOSTIC 1363 if (ipprintfs) 1364 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1365 ntohl(ip->ip_src.s_addr), 1366 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1367 #endif 1368 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1369 ipstat.ips_cantforward++; 1370 m_freem(m); 1371 return; 1372 } 1373 if (ip->ip_ttl <= IPTTLDEC) { 1374 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1375 return; 1376 } 1377 ip->ip_ttl -= IPTTLDEC; 1378 1379 sin = satosin(&ipforward_rt.ro_dst); 1380 if ((rt = ipforward_rt.ro_rt) == 0 || 1381 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1382 if (ipforward_rt.ro_rt) { 1383 RTFREE(ipforward_rt.ro_rt); 1384 ipforward_rt.ro_rt = 0; 1385 } 1386 sin->sin_family = AF_INET; 1387 sin->sin_len = sizeof(struct sockaddr_in); 1388 sin->sin_addr = ip->ip_dst; 1389 1390 rtalloc(&ipforward_rt); 1391 if (ipforward_rt.ro_rt == 0) { 1392 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1393 return; 1394 } 1395 rt = ipforward_rt.ro_rt; 1396 } 1397 1398 /* 1399 * Save at most 68 bytes of the packet in case 1400 * we need to generate an ICMP message to the src. 1401 */ 1402 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 68)); 1403 1404 /* 1405 * If forwarding packet using same interface that it came in on, 1406 * perhaps should send a redirect to sender to shortcut a hop. 1407 * Only send redirect if source is sending directly to us, 1408 * and if packet was not source routed (or has any options). 1409 * Also, don't send redirect if forwarding using a default route 1410 * or a route modified by a redirect. 1411 */ 1412 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1413 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1414 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1415 ipsendredirects && !srcrt) { 1416 if (rt->rt_ifa && 1417 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1418 ifatoia(rt->rt_ifa)->ia_subnet) { 1419 if (rt->rt_flags & RTF_GATEWAY) 1420 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1421 else 1422 dest = ip->ip_dst.s_addr; 1423 /* 1424 * Router requirements says to only send host 1425 * redirects. 1426 */ 1427 type = ICMP_REDIRECT; 1428 code = ICMP_REDIRECT_HOST; 1429 #ifdef DIAGNOSTIC 1430 if (ipprintfs) 1431 printf("redirect (%d) to %x\n", code, 1432 (u_int32_t)dest); 1433 #endif 1434 } 1435 } 1436 1437 #ifdef IPSEC 1438 /* Don't lookup socket in forwading case */ 1439 ipsec_setsocket(m, NULL); 1440 #endif 1441 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1442 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1443 if (error) 1444 ipstat.ips_cantforward++; 1445 else { 1446 ipstat.ips_forward++; 1447 if (type) 1448 ipstat.ips_redirectsent++; 1449 else { 1450 if (mcopy) { 1451 #ifdef GATEWAY 1452 if (mcopy->m_flags & M_CANFASTFWD) 1453 ipflow_create(&ipforward_rt, mcopy); 1454 #endif 1455 m_freem(mcopy); 1456 } 1457 return; 1458 } 1459 } 1460 if (mcopy == NULL) 1461 return; 1462 destifp = NULL; 1463 1464 switch (error) { 1465 1466 case 0: /* forwarded, but need redirect */ 1467 /* type, code set above */ 1468 break; 1469 1470 case ENETUNREACH: /* shouldn't happen, checked above */ 1471 case EHOSTUNREACH: 1472 case ENETDOWN: 1473 case EHOSTDOWN: 1474 default: 1475 type = ICMP_UNREACH; 1476 code = ICMP_UNREACH_HOST; 1477 break; 1478 1479 case EMSGSIZE: 1480 type = ICMP_UNREACH; 1481 code = ICMP_UNREACH_NEEDFRAG; 1482 #ifndef IPSEC 1483 if (ipforward_rt.ro_rt) 1484 destifp = ipforward_rt.ro_rt->rt_ifp; 1485 #else 1486 /* 1487 * If the packet is routed over IPsec tunnel, tell the 1488 * originator the tunnel MTU. 1489 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1490 * XXX quickhack!!! 1491 */ 1492 if (ipforward_rt.ro_rt) { 1493 struct secpolicy *sp; 1494 int ipsecerror; 1495 size_t ipsechdr; 1496 struct route *ro; 1497 1498 sp = ipsec4_getpolicybyaddr(mcopy, 1499 IPSEC_DIR_OUTBOUND, 1500 IP_FORWARDING, 1501 &ipsecerror); 1502 1503 if (sp == NULL) 1504 destifp = ipforward_rt.ro_rt->rt_ifp; 1505 else { 1506 /* count IPsec header size */ 1507 ipsechdr = ipsec4_hdrsiz(mcopy, 1508 IPSEC_DIR_OUTBOUND, 1509 NULL); 1510 1511 /* 1512 * find the correct route for outer IPv4 1513 * header, compute tunnel MTU. 1514 * 1515 * XXX BUG ALERT 1516 * The "dummyifp" code relies upon the fact 1517 * that icmp_error() touches only ifp->if_mtu. 1518 */ 1519 /*XXX*/ 1520 destifp = NULL; 1521 if (sp->req != NULL 1522 && sp->req->sav != NULL 1523 && sp->req->sav->sah != NULL) { 1524 ro = &sp->req->sav->sah->sa_route; 1525 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1526 dummyifp.if_mtu = 1527 ro->ro_rt->rt_ifp->if_mtu; 1528 dummyifp.if_mtu -= ipsechdr; 1529 destifp = &dummyifp; 1530 } 1531 } 1532 1533 key_freesp(sp); 1534 } 1535 } 1536 #endif /*IPSEC*/ 1537 ipstat.ips_cantfrag++; 1538 break; 1539 1540 case ENOBUFS: 1541 type = ICMP_SOURCEQUENCH; 1542 code = 0; 1543 break; 1544 } 1545 icmp_error(mcopy, type, code, dest, destifp); 1546 } 1547 1548 void 1549 ip_savecontrol(inp, mp, ip, m) 1550 struct inpcb *inp; 1551 struct mbuf **mp; 1552 struct ip *ip; 1553 struct mbuf *m; 1554 { 1555 1556 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1557 struct timeval tv; 1558 1559 microtime(&tv); 1560 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1561 SCM_TIMESTAMP, SOL_SOCKET); 1562 if (*mp) 1563 mp = &(*mp)->m_next; 1564 } 1565 if (inp->inp_flags & INP_RECVDSTADDR) { 1566 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1567 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1568 if (*mp) 1569 mp = &(*mp)->m_next; 1570 } 1571 #ifdef notyet 1572 /* 1573 * XXX 1574 * Moving these out of udp_input() made them even more broken 1575 * than they already were. 1576 * - fenner@parc.xerox.com 1577 */ 1578 /* options were tossed already */ 1579 if (inp->inp_flags & INP_RECVOPTS) { 1580 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1581 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1582 if (*mp) 1583 mp = &(*mp)->m_next; 1584 } 1585 /* ip_srcroute doesn't do what we want here, need to fix */ 1586 if (inp->inp_flags & INP_RECVRETOPTS) { 1587 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1588 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1589 if (*mp) 1590 mp = &(*mp)->m_next; 1591 } 1592 #endif 1593 if (inp->inp_flags & INP_RECVIF) { 1594 struct sockaddr_dl sdl; 1595 1596 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1597 sdl.sdl_family = AF_LINK; 1598 sdl.sdl_index = m->m_pkthdr.rcvif ? 1599 m->m_pkthdr.rcvif->if_index : 0; 1600 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1601 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1602 IP_RECVIF, IPPROTO_IP); 1603 if (*mp) 1604 mp = &(*mp)->m_next; 1605 } 1606 } 1607 1608 int 1609 ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1610 int *name; 1611 u_int namelen; 1612 void *oldp; 1613 size_t *oldlenp; 1614 void *newp; 1615 size_t newlen; 1616 { 1617 extern int subnetsarelocal, hostzeroisbroadcast; 1618 1619 int error, old; 1620 1621 /* All sysctl names at this level are terminal. */ 1622 if (namelen != 1) 1623 return (ENOTDIR); 1624 1625 switch (name[0]) { 1626 case IPCTL_FORWARDING: 1627 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1628 case IPCTL_SENDREDIRECTS: 1629 return (sysctl_int(oldp, oldlenp, newp, newlen, 1630 &ipsendredirects)); 1631 case IPCTL_DEFTTL: 1632 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1633 #ifdef notyet 1634 case IPCTL_DEFMTU: 1635 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1636 #endif 1637 case IPCTL_FORWSRCRT: 1638 /* Don't allow this to change in a secure environment. */ 1639 if (securelevel > 0) 1640 return (sysctl_rdint(oldp, oldlenp, newp, 1641 ip_forwsrcrt)); 1642 else 1643 return (sysctl_int(oldp, oldlenp, newp, newlen, 1644 &ip_forwsrcrt)); 1645 case IPCTL_DIRECTEDBCAST: 1646 return (sysctl_int(oldp, oldlenp, newp, newlen, 1647 &ip_directedbcast)); 1648 case IPCTL_ALLOWSRCRT: 1649 return (sysctl_int(oldp, oldlenp, newp, newlen, 1650 &ip_allowsrcrt)); 1651 case IPCTL_SUBNETSARELOCAL: 1652 return (sysctl_int(oldp, oldlenp, newp, newlen, 1653 &subnetsarelocal)); 1654 case IPCTL_MTUDISC: 1655 error = sysctl_int(oldp, oldlenp, newp, newlen, 1656 &ip_mtudisc); 1657 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1658 ip_mtudisc_timeout_q = 1659 rt_timer_queue_create(ip_mtudisc_timeout); 1660 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1661 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1662 ip_mtudisc_timeout_q = NULL; 1663 } 1664 return error; 1665 case IPCTL_ANONPORTMIN: 1666 old = anonportmin; 1667 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1668 if (anonportmin >= anonportmax || anonportmin > 65535 1669 #ifndef IPNOPRIVPORTS 1670 || anonportmin < IPPORT_RESERVED 1671 #endif 1672 ) { 1673 anonportmin = old; 1674 return (EINVAL); 1675 } 1676 return (error); 1677 case IPCTL_ANONPORTMAX: 1678 old = anonportmax; 1679 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1680 if (anonportmin >= anonportmax || anonportmax > 65535 1681 #ifndef IPNOPRIVPORTS 1682 || anonportmax < IPPORT_RESERVED 1683 #endif 1684 ) { 1685 anonportmax = old; 1686 return (EINVAL); 1687 } 1688 return (error); 1689 case IPCTL_MTUDISCTIMEOUT: 1690 error = sysctl_int(oldp, oldlenp, newp, newlen, 1691 &ip_mtudisc_timeout); 1692 if (ip_mtudisc_timeout_q != NULL) 1693 rt_timer_queue_change(ip_mtudisc_timeout_q, 1694 ip_mtudisc_timeout); 1695 return (error); 1696 #ifdef GATEWAY 1697 case IPCTL_MAXFLOWS: 1698 { 1699 int s; 1700 1701 error = sysctl_int(oldp, oldlenp, newp, newlen, 1702 &ip_maxflows); 1703 s = splsoftnet(); 1704 ipflow_reap(0); 1705 splx(s); 1706 return (error); 1707 } 1708 #endif 1709 case IPCTL_HOSTZEROBROADCAST: 1710 return (sysctl_int(oldp, oldlenp, newp, newlen, 1711 &hostzeroisbroadcast)); 1712 #if NGIF > 0 1713 case IPCTL_GIF_TTL: 1714 return(sysctl_int(oldp, oldlenp, newp, newlen, 1715 &ip_gif_ttl)); 1716 #endif 1717 1718 default: 1719 return (EOPNOTSUPP); 1720 } 1721 /* NOTREACHED */ 1722 } 1723