ip_input.c revision 1.114.4.6
1 /* $NetBSD: ip_input.c,v 1.114.4.6 2001/04/24 22:21:20 he 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 int ip_nfragpackets = 0; 211 int ip_maxfragpackets = 200; 212 213 static __inline int ipq_lock_try __P((void)); 214 static __inline void ipq_unlock __P((void)); 215 216 static __inline int 217 ipq_lock_try() 218 { 219 int s; 220 221 s = splimp(); 222 if (ipq_locked) { 223 splx(s); 224 return (0); 225 } 226 ipq_locked = 1; 227 splx(s); 228 return (1); 229 } 230 231 static __inline void 232 ipq_unlock() 233 { 234 int s; 235 236 s = splimp(); 237 ipq_locked = 0; 238 splx(s); 239 } 240 241 #ifdef DIAGNOSTIC 242 #define IPQ_LOCK() \ 243 do { \ 244 if (ipq_lock_try() == 0) { \ 245 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 246 panic("ipq_lock"); \ 247 } \ 248 } while (0) 249 #define IPQ_LOCK_CHECK() \ 250 do { \ 251 if (ipq_locked == 0) { \ 252 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 253 panic("ipq lock check"); \ 254 } \ 255 } while (0) 256 #else 257 #define IPQ_LOCK() (void) ipq_lock_try() 258 #define IPQ_LOCK_CHECK() /* nothing */ 259 #endif 260 261 #define IPQ_UNLOCK() ipq_unlock() 262 263 struct pool ipqent_pool; 264 265 /* 266 * We need to save the IP options in case a protocol wants to respond 267 * to an incoming packet over the same route if the packet got here 268 * using IP source routing. This allows connection establishment and 269 * maintenance when the remote end is on a network that is not known 270 * to us. 271 */ 272 int ip_nhops = 0; 273 static struct ip_srcrt { 274 struct in_addr dst; /* final destination */ 275 char nop; /* one NOP to align */ 276 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 277 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 278 } ip_srcrt; 279 280 static void save_rte __P((u_char *, struct in_addr)); 281 282 /* 283 * IP initialization: fill in IP protocol switch table. 284 * All protocols not implemented in kernel go to raw IP protocol handler. 285 */ 286 void 287 ip_init() 288 { 289 struct protosw *pr; 290 int i; 291 292 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 293 0, NULL, NULL, M_IPQ); 294 295 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 296 if (pr == 0) 297 panic("ip_init"); 298 for (i = 0; i < IPPROTO_MAX; i++) 299 ip_protox[i] = pr - inetsw; 300 for (pr = inetdomain.dom_protosw; 301 pr < inetdomain.dom_protoswNPROTOSW; pr++) 302 if (pr->pr_domain->dom_family == PF_INET && 303 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 304 ip_protox[pr->pr_protocol] = pr - inetsw; 305 LIST_INIT(&ipq); 306 ip_id = time.tv_sec & 0xffff; 307 ipintrq.ifq_maxlen = ipqmaxlen; 308 TAILQ_INIT(&in_ifaddr); 309 in_ifaddrhashtbl = 310 hashinit(IN_IFADDR_HASH_SIZE, M_IFADDR, M_WAITOK, &in_ifaddrhash); 311 if (ip_mtudisc != 0) 312 ip_mtudisc_timeout_q = 313 rt_timer_queue_create(ip_mtudisc_timeout); 314 #ifdef GATEWAY 315 ipflow_init(); 316 #endif 317 } 318 319 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 320 struct route ipforward_rt; 321 322 /* 323 * IP software interrupt routine 324 */ 325 void 326 ipintr() 327 { 328 int s; 329 struct mbuf *m; 330 331 while (1) { 332 s = splimp(); 333 IF_DEQUEUE(&ipintrq, m); 334 splx(s); 335 if (m == 0) 336 return; 337 ip_input(m); 338 } 339 } 340 341 /* 342 * Ip input routine. Checksum and byte swap header. If fragmented 343 * try to reassemble. Process options. Pass to next level. 344 */ 345 void 346 ip_input(struct mbuf *m) 347 { 348 struct ip *ip = NULL; 349 struct ipq *fp; 350 struct in_ifaddr *ia; 351 struct ifaddr *ifa; 352 struct ipqent *ipqe; 353 int hlen = 0, mff, len; 354 int downmatch; 355 #ifdef PFIL_HOOKS 356 struct packet_filter_hook *pfh; 357 struct mbuf *m0; 358 int rv; 359 #endif /* PFIL_HOOKS */ 360 361 #ifdef DIAGNOSTIC 362 if ((m->m_flags & M_PKTHDR) == 0) 363 panic("ipintr no HDR"); 364 #endif 365 #ifdef IPSEC 366 /* 367 * should the inner packet be considered authentic? 368 * see comment in ah4_input(). 369 */ 370 if (m) { 371 m->m_flags &= ~M_AUTHIPHDR; 372 m->m_flags &= ~M_AUTHIPDGM; 373 } 374 #endif 375 /* 376 * If no IP addresses have been set yet but the interfaces 377 * are receiving, can't do anything with incoming packets yet. 378 */ 379 if (in_ifaddr.tqh_first == 0) 380 goto bad; 381 ipstat.ips_total++; 382 if (m->m_len < sizeof (struct ip) && 383 (m = m_pullup(m, sizeof (struct ip))) == 0) { 384 ipstat.ips_toosmall++; 385 return; 386 } 387 ip = mtod(m, struct ip *); 388 if (ip->ip_v != IPVERSION) { 389 ipstat.ips_badvers++; 390 goto bad; 391 } 392 hlen = ip->ip_hl << 2; 393 if (hlen < sizeof(struct ip)) { /* minimum header length */ 394 ipstat.ips_badhlen++; 395 goto bad; 396 } 397 if (hlen > m->m_len) { 398 if ((m = m_pullup(m, hlen)) == 0) { 399 ipstat.ips_badhlen++; 400 return; 401 } 402 ip = mtod(m, struct ip *); 403 } 404 405 /* 406 * RFC1122: packets with a multicast source address are 407 * not allowed. 408 */ 409 if (IN_MULTICAST(ip->ip_src.s_addr)) { 410 /* XXX stat */ 411 goto bad; 412 } 413 414 if (in_cksum(m, hlen) != 0) { 415 ipstat.ips_badsum++; 416 goto bad; 417 } 418 419 /* 420 * Convert fields to host representation. 421 */ 422 NTOHS(ip->ip_len); 423 NTOHS(ip->ip_off); 424 len = ip->ip_len; 425 426 /* 427 * Check for additional length bogosity 428 */ 429 if (len < hlen) { 430 ipstat.ips_badlen++; 431 goto bad; 432 } 433 434 /* 435 * Check that the amount of data in the buffers 436 * is as at least much as the IP header would have us expect. 437 * Trim mbufs if longer than we expect. 438 * Drop packet if shorter than we expect. 439 */ 440 if (m->m_pkthdr.len < len) { 441 ipstat.ips_tooshort++; 442 goto bad; 443 } 444 if (m->m_pkthdr.len > len) { 445 if (m->m_len == m->m_pkthdr.len) { 446 m->m_len = len; 447 m->m_pkthdr.len = len; 448 } else 449 m_adj(m, len - m->m_pkthdr.len); 450 } 451 452 #ifdef IPSEC 453 /* ipflow (IP fast fowarding) is not compatible with IPsec. */ 454 m->m_flags &= ~M_CANFASTFWD; 455 #else 456 /* 457 * Assume that we can create a fast-forward IP flow entry 458 * based on this packet. 459 */ 460 m->m_flags |= M_CANFASTFWD; 461 #endif 462 463 #ifdef PFIL_HOOKS 464 #ifdef IPSEC 465 /* 466 * let ipfilter look at packet on the wire, 467 * not the decapsulated packet. 468 */ 469 if (ipsec_gethist(m, NULL)) 470 goto nofilt; 471 #endif 472 /* 473 * Run through list of hooks for input packets. If there are any 474 * filters which require that additional packets in the flow are 475 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 476 * Note that filters must _never_ set this flag, as another filter 477 * in the list may have previously cleared it. 478 */ 479 m0 = m; 480 pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 481 for (; pfh; pfh = pfh->pfil_link.tqe_next) 482 if (pfh->pfil_func) { 483 rv = pfh->pfil_func(ip, hlen, 484 m->m_pkthdr.rcvif, 0, &m0); 485 if (rv) 486 return; 487 m = m0; 488 if (m == NULL) 489 return; 490 ip = mtod(m, struct ip *); 491 } 492 #ifdef IPSEC 493 nofilt:; 494 #endif 495 #endif /* PFIL_HOOKS */ 496 497 /* 498 * Process options and, if not destined for us, 499 * ship it on. ip_dooptions returns 1 when an 500 * error was detected (causing an icmp message 501 * to be sent and the original packet to be freed). 502 */ 503 ip_nhops = 0; /* for source routed packets */ 504 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 505 return; 506 507 /* 508 * Check our list of addresses, to see if the packet is for us. 509 * 510 * Traditional 4.4BSD did not consult IFF_UP at all. 511 * The behavior here is to treat addresses on !IFF_UP interface 512 * as not mine. 513 */ 514 downmatch = 0; 515 for (ia = IN_IFADDR_HASH(ip->ip_dst.s_addr).lh_first; 516 ia != NULL; 517 ia = ia->ia_hash.le_next) { 518 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 519 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 520 break; 521 else 522 downmatch++; 523 } 524 } 525 if (ia != NULL) 526 goto ours; 527 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 528 for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first; 529 ifa != NULL; ifa = ifa->ifa_list.tqe_next) { 530 if (ifa->ifa_addr->sa_family != AF_INET) continue; 531 ia = ifatoia(ifa); 532 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 533 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 534 /* 535 * Look for all-0's host part (old broadcast addr), 536 * either for subnet or net. 537 */ 538 ip->ip_dst.s_addr == ia->ia_subnet || 539 ip->ip_dst.s_addr == ia->ia_net) 540 goto ours; 541 /* 542 * An interface with IP address zero accepts 543 * all packets that arrive on that interface. 544 */ 545 if (in_nullhost(ia->ia_addr.sin_addr)) 546 goto ours; 547 } 548 } 549 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 550 struct in_multi *inm; 551 #ifdef MROUTING 552 extern struct socket *ip_mrouter; 553 554 if (m->m_flags & M_EXT) { 555 if ((m = m_pullup(m, hlen)) == 0) { 556 ipstat.ips_toosmall++; 557 return; 558 } 559 ip = mtod(m, struct ip *); 560 } 561 562 if (ip_mrouter) { 563 /* 564 * If we are acting as a multicast router, all 565 * incoming multicast packets are passed to the 566 * kernel-level multicast forwarding function. 567 * The packet is returned (relatively) intact; if 568 * ip_mforward() returns a non-zero value, the packet 569 * must be discarded, else it may be accepted below. 570 * 571 * (The IP ident field is put in the same byte order 572 * as expected when ip_mforward() is called from 573 * ip_output().) 574 */ 575 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 576 ipstat.ips_cantforward++; 577 m_freem(m); 578 return; 579 } 580 581 /* 582 * The process-level routing demon needs to receive 583 * all multicast IGMP packets, whether or not this 584 * host belongs to their destination groups. 585 */ 586 if (ip->ip_p == IPPROTO_IGMP) 587 goto ours; 588 ipstat.ips_forward++; 589 } 590 #endif 591 /* 592 * See if we belong to the destination multicast group on the 593 * arrival interface. 594 */ 595 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 596 if (inm == NULL) { 597 ipstat.ips_cantforward++; 598 m_freem(m); 599 return; 600 } 601 goto ours; 602 } 603 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 604 in_nullhost(ip->ip_dst)) 605 goto ours; 606 607 /* 608 * Not for us; forward if possible and desirable. 609 */ 610 if (ipforwarding == 0) { 611 ipstat.ips_cantforward++; 612 m_freem(m); 613 } else { 614 /* 615 * If ip_dst matched any of my address on !IFF_UP interface, 616 * and there's no IFF_UP interface that matches ip_dst, 617 * send icmp unreach. Forwarding it will result in in-kernel 618 * forwarding loop till TTL goes to 0. 619 */ 620 if (downmatch) { 621 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 622 ipstat.ips_cantforward++; 623 return; 624 } 625 ip_forward(m, 0); 626 } 627 return; 628 629 ours: 630 /* 631 * If offset or IP_MF are set, must reassemble. 632 * Otherwise, nothing need be done. 633 * (We could look in the reassembly queue to see 634 * if the packet was previously fragmented, 635 * but it's not worth the time; just let them time out.) 636 */ 637 if (ip->ip_off & ~(IP_DF|IP_RF)) { 638 /* 639 * Look for queue of fragments 640 * of this datagram. 641 */ 642 IPQ_LOCK(); 643 for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next) 644 if (ip->ip_id == fp->ipq_id && 645 in_hosteq(ip->ip_src, fp->ipq_src) && 646 in_hosteq(ip->ip_dst, fp->ipq_dst) && 647 ip->ip_p == fp->ipq_p) 648 goto found; 649 fp = 0; 650 found: 651 652 /* 653 * Adjust ip_len to not reflect header, 654 * set ipqe_mff if more fragments are expected, 655 * convert offset of this to bytes. 656 */ 657 ip->ip_len -= hlen; 658 mff = (ip->ip_off & IP_MF) != 0; 659 if (mff) { 660 /* 661 * Make sure that fragments have a data length 662 * that's a non-zero multiple of 8 bytes. 663 */ 664 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 665 ipstat.ips_badfrags++; 666 IPQ_UNLOCK(); 667 goto bad; 668 } 669 } 670 ip->ip_off <<= 3; 671 672 /* 673 * If datagram marked as having more fragments 674 * or if this is not the first fragment, 675 * attempt reassembly; if it succeeds, proceed. 676 */ 677 if (mff || ip->ip_off) { 678 ipstat.ips_fragments++; 679 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 680 if (ipqe == NULL) { 681 ipstat.ips_rcvmemdrop++; 682 IPQ_UNLOCK(); 683 goto bad; 684 } 685 ipqe->ipqe_mff = mff; 686 ipqe->ipqe_m = m; 687 ipqe->ipqe_ip = ip; 688 m = ip_reass(ipqe, fp); 689 if (m == 0) { 690 IPQ_UNLOCK(); 691 return; 692 } 693 ipstat.ips_reassembled++; 694 ip = mtod(m, struct ip *); 695 hlen = ip->ip_hl << 2; 696 ip->ip_len += hlen; 697 } else 698 if (fp) 699 ip_freef(fp); 700 IPQ_UNLOCK(); 701 } 702 703 #ifdef IPSEC 704 /* 705 * enforce IPsec policy checking if we are seeing last header. 706 * note that we do not visit this with protocols with pcb layer 707 * code - like udp/tcp/raw ip. 708 */ 709 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 710 ipsec4_in_reject(m, NULL)) { 711 ipsecstat.in_polvio++; 712 goto bad; 713 } 714 #endif 715 716 /* 717 * Switch out to protocol's input routine. 718 */ 719 #if IFA_STATS 720 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 721 #endif 722 ipstat.ips_delivered++; 723 { 724 int off = hlen, nh = ip->ip_p; 725 726 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 727 return; 728 } 729 bad: 730 m_freem(m); 731 } 732 733 /* 734 * Take incoming datagram fragment and try to 735 * reassemble it into whole datagram. If a chain for 736 * reassembly of this datagram already exists, then it 737 * is given as fp; otherwise have to make a chain. 738 */ 739 struct mbuf * 740 ip_reass(ipqe, fp) 741 struct ipqent *ipqe; 742 struct ipq *fp; 743 { 744 struct mbuf *m = ipqe->ipqe_m; 745 struct ipqent *nq, *p, *q; 746 struct ip *ip; 747 struct mbuf *t; 748 int hlen = ipqe->ipqe_ip->ip_hl << 2; 749 int i, next; 750 751 IPQ_LOCK_CHECK(); 752 753 /* 754 * Presence of header sizes in mbufs 755 * would confuse code below. 756 */ 757 m->m_data += hlen; 758 m->m_len -= hlen; 759 760 /* 761 * If first fragment to arrive, create a reassembly queue. 762 */ 763 if (fp == 0) { 764 /* 765 * Enforce upper bound on number of fragmented packets 766 * for which we attempt reassembly; 767 * If maxfrag is 0, never accept fragments. 768 * If maxfrag is -1, accept all fragments without limitation. 769 */ 770 if (ip_maxfragpackets < 0) 771 ; 772 else if (ip_nfragpackets >= ip_maxfragpackets) 773 goto dropfrag; 774 ip_nfragpackets++; 775 MALLOC(fp, struct ipq *, sizeof (struct ipq), 776 M_FTABLE, M_NOWAIT); 777 if (fp == NULL) 778 goto dropfrag; 779 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 780 fp->ipq_ttl = IPFRAGTTL; 781 fp->ipq_p = ipqe->ipqe_ip->ip_p; 782 fp->ipq_id = ipqe->ipqe_ip->ip_id; 783 LIST_INIT(&fp->ipq_fragq); 784 fp->ipq_src = ipqe->ipqe_ip->ip_src; 785 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 786 p = NULL; 787 goto insert; 788 } 789 790 /* 791 * Find a segment which begins after this one does. 792 */ 793 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 794 p = q, q = q->ipqe_q.le_next) 795 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 796 break; 797 798 /* 799 * If there is a preceding segment, it may provide some of 800 * our data already. If so, drop the data from the incoming 801 * segment. If it provides all of our data, drop us. 802 */ 803 if (p != NULL) { 804 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 805 ipqe->ipqe_ip->ip_off; 806 if (i > 0) { 807 if (i >= ipqe->ipqe_ip->ip_len) 808 goto dropfrag; 809 m_adj(ipqe->ipqe_m, i); 810 ipqe->ipqe_ip->ip_off += i; 811 ipqe->ipqe_ip->ip_len -= i; 812 } 813 } 814 815 /* 816 * While we overlap succeeding segments trim them or, 817 * if they are completely covered, dequeue them. 818 */ 819 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 820 q->ipqe_ip->ip_off; q = nq) { 821 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 822 q->ipqe_ip->ip_off; 823 if (i < q->ipqe_ip->ip_len) { 824 q->ipqe_ip->ip_len -= i; 825 q->ipqe_ip->ip_off += i; 826 m_adj(q->ipqe_m, i); 827 break; 828 } 829 nq = q->ipqe_q.le_next; 830 m_freem(q->ipqe_m); 831 LIST_REMOVE(q, ipqe_q); 832 pool_put(&ipqent_pool, q); 833 } 834 835 insert: 836 /* 837 * Stick new segment in its place; 838 * check for complete reassembly. 839 */ 840 if (p == NULL) { 841 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 842 } else { 843 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 844 } 845 next = 0; 846 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 847 p = q, q = q->ipqe_q.le_next) { 848 if (q->ipqe_ip->ip_off != next) 849 return (0); 850 next += q->ipqe_ip->ip_len; 851 } 852 if (p->ipqe_mff) 853 return (0); 854 855 /* 856 * Reassembly is complete. Check for a bogus message size and 857 * concatenate fragments. 858 */ 859 q = fp->ipq_fragq.lh_first; 860 ip = q->ipqe_ip; 861 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 862 ipstat.ips_toolong++; 863 ip_freef(fp); 864 return (0); 865 } 866 m = q->ipqe_m; 867 t = m->m_next; 868 m->m_next = 0; 869 m_cat(m, t); 870 nq = q->ipqe_q.le_next; 871 pool_put(&ipqent_pool, q); 872 for (q = nq; q != NULL; q = nq) { 873 t = q->ipqe_m; 874 nq = q->ipqe_q.le_next; 875 pool_put(&ipqent_pool, q); 876 m_cat(m, t); 877 } 878 879 /* 880 * Create header for new ip packet by 881 * modifying header of first packet; 882 * dequeue and discard fragment reassembly header. 883 * Make header visible. 884 */ 885 ip->ip_len = next; 886 ip->ip_src = fp->ipq_src; 887 ip->ip_dst = fp->ipq_dst; 888 LIST_REMOVE(fp, ipq_q); 889 FREE(fp, M_FTABLE); 890 ip_nfragpackets--; 891 m->m_len += (ip->ip_hl << 2); 892 m->m_data -= (ip->ip_hl << 2); 893 /* some debugging cruft by sklower, below, will go away soon */ 894 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 895 int plen = 0; 896 for (t = m; t; t = t->m_next) 897 plen += t->m_len; 898 m->m_pkthdr.len = plen; 899 } 900 return (m); 901 902 dropfrag: 903 ipstat.ips_fragdropped++; 904 m_freem(m); 905 pool_put(&ipqent_pool, ipqe); 906 return (0); 907 } 908 909 /* 910 * Free a fragment reassembly header and all 911 * associated datagrams. 912 */ 913 void 914 ip_freef(fp) 915 struct ipq *fp; 916 { 917 struct ipqent *q, *p; 918 919 IPQ_LOCK_CHECK(); 920 921 for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { 922 p = q->ipqe_q.le_next; 923 m_freem(q->ipqe_m); 924 LIST_REMOVE(q, ipqe_q); 925 pool_put(&ipqent_pool, q); 926 } 927 LIST_REMOVE(fp, ipq_q); 928 FREE(fp, M_FTABLE); 929 ip_nfragpackets--; 930 } 931 932 /* 933 * IP timer processing; 934 * if a timer expires on a reassembly 935 * queue, discard it. 936 */ 937 void 938 ip_slowtimo() 939 { 940 struct ipq *fp, *nfp; 941 int s = splsoftnet(); 942 943 IPQ_LOCK(); 944 for (fp = ipq.lh_first; fp != NULL; fp = nfp) { 945 nfp = fp->ipq_q.le_next; 946 if (--fp->ipq_ttl == 0) { 947 ipstat.ips_fragtimeout++; 948 ip_freef(fp); 949 } 950 } 951 /* 952 * If we are over the maximum number of fragments 953 * (due to the limit being lowered), drain off 954 * enough to get down to the new limit. 955 */ 956 if (ip_maxfragpackets < 0) 957 ; 958 else { 959 while (ip_nfragpackets > ip_maxfragpackets && ipq.lh_first) 960 ip_freef(ipq.lh_first); 961 } 962 IPQ_UNLOCK(); 963 #ifdef GATEWAY 964 ipflow_slowtimo(); 965 #endif 966 splx(s); 967 } 968 969 /* 970 * Drain off all datagram fragments. 971 */ 972 void 973 ip_drain() 974 { 975 976 /* 977 * We may be called from a device's interrupt context. If 978 * the ipq is already busy, just bail out now. 979 */ 980 if (ipq_lock_try() == 0) 981 return; 982 983 while (ipq.lh_first != NULL) { 984 ipstat.ips_fragdropped++; 985 ip_freef(ipq.lh_first); 986 } 987 988 IPQ_UNLOCK(); 989 } 990 991 /* 992 * Do option processing on a datagram, 993 * possibly discarding it if bad options are encountered, 994 * or forwarding it if source-routed. 995 * Returns 1 if packet has been forwarded/freed, 996 * 0 if the packet should be processed further. 997 */ 998 int 999 ip_dooptions(m) 1000 struct mbuf *m; 1001 { 1002 struct ip *ip = mtod(m, struct ip *); 1003 u_char *cp, *cp0; 1004 struct ip_timestamp *ipt; 1005 struct in_ifaddr *ia; 1006 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1007 struct in_addr dst; 1008 n_time ntime; 1009 1010 dst = ip->ip_dst; 1011 cp = (u_char *)(ip + 1); 1012 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1013 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1014 opt = cp[IPOPT_OPTVAL]; 1015 if (opt == IPOPT_EOL) 1016 break; 1017 if (opt == IPOPT_NOP) 1018 optlen = 1; 1019 else { 1020 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1021 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1022 goto bad; 1023 } 1024 optlen = cp[IPOPT_OLEN]; 1025 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1026 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1027 goto bad; 1028 } 1029 } 1030 switch (opt) { 1031 1032 default: 1033 break; 1034 1035 /* 1036 * Source routing with record. 1037 * Find interface with current destination address. 1038 * If none on this machine then drop if strictly routed, 1039 * or do nothing if loosely routed. 1040 * Record interface address and bring up next address 1041 * component. If strictly routed make sure next 1042 * address is on directly accessible net. 1043 */ 1044 case IPOPT_LSRR: 1045 case IPOPT_SSRR: 1046 if (ip_allowsrcrt == 0) { 1047 type = ICMP_UNREACH; 1048 code = ICMP_UNREACH_NET_PROHIB; 1049 goto bad; 1050 } 1051 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1052 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1053 goto bad; 1054 } 1055 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1056 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1057 goto bad; 1058 } 1059 ipaddr.sin_addr = ip->ip_dst; 1060 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1061 if (ia == 0) { 1062 if (opt == IPOPT_SSRR) { 1063 type = ICMP_UNREACH; 1064 code = ICMP_UNREACH_SRCFAIL; 1065 goto bad; 1066 } 1067 /* 1068 * Loose routing, and not at next destination 1069 * yet; nothing to do except forward. 1070 */ 1071 break; 1072 } 1073 off--; /* 0 origin */ 1074 if ((off + sizeof(struct in_addr)) > optlen) { 1075 /* 1076 * End of source route. Should be for us. 1077 */ 1078 save_rte(cp, ip->ip_src); 1079 break; 1080 } 1081 /* 1082 * locate outgoing interface 1083 */ 1084 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 1085 sizeof(ipaddr.sin_addr)); 1086 if (opt == IPOPT_SSRR) 1087 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 1088 else 1089 ia = ip_rtaddr(ipaddr.sin_addr); 1090 if (ia == 0) { 1091 type = ICMP_UNREACH; 1092 code = ICMP_UNREACH_SRCFAIL; 1093 goto bad; 1094 } 1095 ip->ip_dst = ipaddr.sin_addr; 1096 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1097 (caddr_t)(cp + off), sizeof(struct in_addr)); 1098 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1099 /* 1100 * Let ip_intr's mcast routing check handle mcast pkts 1101 */ 1102 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1103 break; 1104 1105 case IPOPT_RR: 1106 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1107 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1108 goto bad; 1109 } 1110 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1111 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1112 goto bad; 1113 } 1114 /* 1115 * If no space remains, ignore. 1116 */ 1117 off--; /* 0 origin */ 1118 if ((off + sizeof(struct in_addr)) > optlen) 1119 break; 1120 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1121 sizeof(ipaddr.sin_addr)); 1122 /* 1123 * locate outgoing interface; if we're the destination, 1124 * use the incoming interface (should be same). 1125 */ 1126 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 1127 == NULL && 1128 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1129 type = ICMP_UNREACH; 1130 code = ICMP_UNREACH_HOST; 1131 goto bad; 1132 } 1133 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1134 (caddr_t)(cp + off), sizeof(struct in_addr)); 1135 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1136 break; 1137 1138 case IPOPT_TS: 1139 code = cp - (u_char *)ip; 1140 ipt = (struct ip_timestamp *)cp; 1141 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1142 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1143 goto bad; 1144 } 1145 if (ipt->ipt_ptr < 5) { 1146 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1147 goto bad; 1148 } 1149 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1150 if (++ipt->ipt_oflw == 0) { 1151 code = (u_char *)&ipt->ipt_ptr - 1152 (u_char *)ip; 1153 goto bad; 1154 } 1155 break; 1156 } 1157 cp0 = (cp + ipt->ipt_ptr - 1); 1158 switch (ipt->ipt_flg) { 1159 1160 case IPOPT_TS_TSONLY: 1161 break; 1162 1163 case IPOPT_TS_TSANDADDR: 1164 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1165 sizeof(struct in_addr) > ipt->ipt_len) { 1166 code = (u_char *)&ipt->ipt_ptr - 1167 (u_char *)ip; 1168 goto bad; 1169 } 1170 ipaddr.sin_addr = dst; 1171 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1172 m->m_pkthdr.rcvif)); 1173 if (ia == 0) 1174 continue; 1175 bcopy(&ia->ia_addr.sin_addr, 1176 cp0, sizeof(struct in_addr)); 1177 ipt->ipt_ptr += sizeof(struct in_addr); 1178 break; 1179 1180 case IPOPT_TS_PRESPEC: 1181 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1182 sizeof(struct in_addr) > ipt->ipt_len) { 1183 code = (u_char *)&ipt->ipt_ptr - 1184 (u_char *)ip; 1185 goto bad; 1186 } 1187 bcopy(cp0, &ipaddr.sin_addr, 1188 sizeof(struct in_addr)); 1189 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1190 == NULL) 1191 continue; 1192 ipt->ipt_ptr += sizeof(struct in_addr); 1193 break; 1194 1195 default: 1196 /* XXX can't take &ipt->ipt_flg */ 1197 code = (u_char *)&ipt->ipt_ptr - 1198 (u_char *)ip + 1; 1199 goto bad; 1200 } 1201 ntime = iptime(); 1202 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1203 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, 1204 sizeof(n_time)); 1205 ipt->ipt_ptr += sizeof(n_time); 1206 } 1207 } 1208 if (forward) { 1209 if (ip_forwsrcrt == 0) { 1210 type = ICMP_UNREACH; 1211 code = ICMP_UNREACH_SRCFAIL; 1212 goto bad; 1213 } 1214 ip_forward(m, 1); 1215 return (1); 1216 } 1217 return (0); 1218 bad: 1219 icmp_error(m, type, code, 0, 0); 1220 ipstat.ips_badoptions++; 1221 return (1); 1222 } 1223 1224 /* 1225 * Given address of next destination (final or next hop), 1226 * return internet address info of interface to be used to get there. 1227 */ 1228 struct in_ifaddr * 1229 ip_rtaddr(dst) 1230 struct in_addr dst; 1231 { 1232 struct sockaddr_in *sin; 1233 1234 sin = satosin(&ipforward_rt.ro_dst); 1235 1236 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1237 if (ipforward_rt.ro_rt) { 1238 RTFREE(ipforward_rt.ro_rt); 1239 ipforward_rt.ro_rt = 0; 1240 } 1241 sin->sin_family = AF_INET; 1242 sin->sin_len = sizeof(*sin); 1243 sin->sin_addr = dst; 1244 1245 rtalloc(&ipforward_rt); 1246 } 1247 if (ipforward_rt.ro_rt == 0) 1248 return ((struct in_ifaddr *)0); 1249 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1250 } 1251 1252 /* 1253 * Save incoming source route for use in replies, 1254 * to be picked up later by ip_srcroute if the receiver is interested. 1255 */ 1256 void 1257 save_rte(option, dst) 1258 u_char *option; 1259 struct in_addr dst; 1260 { 1261 unsigned olen; 1262 1263 olen = option[IPOPT_OLEN]; 1264 #ifdef DIAGNOSTIC 1265 if (ipprintfs) 1266 printf("save_rte: olen %d\n", olen); 1267 #endif /* 0 */ 1268 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1269 return; 1270 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1271 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1272 ip_srcrt.dst = dst; 1273 } 1274 1275 /* 1276 * Retrieve incoming source route for use in replies, 1277 * in the same form used by setsockopt. 1278 * The first hop is placed before the options, will be removed later. 1279 */ 1280 struct mbuf * 1281 ip_srcroute() 1282 { 1283 struct in_addr *p, *q; 1284 struct mbuf *m; 1285 1286 if (ip_nhops == 0) 1287 return ((struct mbuf *)0); 1288 m = m_get(M_DONTWAIT, MT_SOOPTS); 1289 if (m == 0) 1290 return ((struct mbuf *)0); 1291 1292 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1293 1294 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1295 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1296 OPTSIZ; 1297 #ifdef DIAGNOSTIC 1298 if (ipprintfs) 1299 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1300 #endif 1301 1302 /* 1303 * First save first hop for return route 1304 */ 1305 p = &ip_srcrt.route[ip_nhops - 1]; 1306 *(mtod(m, struct in_addr *)) = *p--; 1307 #ifdef DIAGNOSTIC 1308 if (ipprintfs) 1309 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1310 #endif 1311 1312 /* 1313 * Copy option fields and padding (nop) to mbuf. 1314 */ 1315 ip_srcrt.nop = IPOPT_NOP; 1316 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1317 bcopy((caddr_t)&ip_srcrt.nop, 1318 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1319 q = (struct in_addr *)(mtod(m, caddr_t) + 1320 sizeof(struct in_addr) + OPTSIZ); 1321 #undef OPTSIZ 1322 /* 1323 * Record return path as an IP source route, 1324 * reversing the path (pointers are now aligned). 1325 */ 1326 while (p >= ip_srcrt.route) { 1327 #ifdef DIAGNOSTIC 1328 if (ipprintfs) 1329 printf(" %x", ntohl(q->s_addr)); 1330 #endif 1331 *q++ = *p--; 1332 } 1333 /* 1334 * Last hop goes to final destination. 1335 */ 1336 *q = ip_srcrt.dst; 1337 #ifdef DIAGNOSTIC 1338 if (ipprintfs) 1339 printf(" %x\n", ntohl(q->s_addr)); 1340 #endif 1341 return (m); 1342 } 1343 1344 /* 1345 * Strip out IP options, at higher 1346 * level protocol in the kernel. 1347 * Second argument is buffer to which options 1348 * will be moved, and return value is their length. 1349 * XXX should be deleted; last arg currently ignored. 1350 */ 1351 void 1352 ip_stripoptions(m, mopt) 1353 struct mbuf *m; 1354 struct mbuf *mopt; 1355 { 1356 int i; 1357 struct ip *ip = mtod(m, struct ip *); 1358 caddr_t opts; 1359 int olen; 1360 1361 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1362 opts = (caddr_t)(ip + 1); 1363 i = m->m_len - (sizeof (struct ip) + olen); 1364 bcopy(opts + olen, opts, (unsigned)i); 1365 m->m_len -= olen; 1366 if (m->m_flags & M_PKTHDR) 1367 m->m_pkthdr.len -= olen; 1368 ip->ip_len -= olen; 1369 ip->ip_hl = sizeof (struct ip) >> 2; 1370 } 1371 1372 int inetctlerrmap[PRC_NCMDS] = { 1373 0, 0, 0, 0, 1374 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1375 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1376 EMSGSIZE, EHOSTUNREACH, 0, 0, 1377 0, 0, 0, 0, 1378 ENOPROTOOPT 1379 }; 1380 1381 /* 1382 * Forward a packet. If some error occurs return the sender 1383 * an icmp packet. Note we can't always generate a meaningful 1384 * icmp message because icmp doesn't have a large enough repertoire 1385 * of codes and types. 1386 * 1387 * If not forwarding, just drop the packet. This could be confusing 1388 * if ipforwarding was zero but some routing protocol was advancing 1389 * us as a gateway to somewhere. However, we must let the routing 1390 * protocol deal with that. 1391 * 1392 * The srcrt parameter indicates whether the packet is being forwarded 1393 * via a source route. 1394 */ 1395 void 1396 ip_forward(m, srcrt) 1397 struct mbuf *m; 1398 int srcrt; 1399 { 1400 struct ip *ip = mtod(m, struct ip *); 1401 struct sockaddr_in *sin; 1402 struct rtentry *rt; 1403 int error, type = 0, code = 0; 1404 struct mbuf *mcopy; 1405 n_long dest; 1406 struct ifnet *destifp; 1407 #ifdef IPSEC 1408 struct ifnet dummyifp; 1409 #endif 1410 1411 dest = 0; 1412 #ifdef DIAGNOSTIC 1413 if (ipprintfs) 1414 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1415 ntohl(ip->ip_src.s_addr), 1416 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1417 #endif 1418 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1419 ipstat.ips_cantforward++; 1420 m_freem(m); 1421 return; 1422 } 1423 if (ip->ip_ttl <= IPTTLDEC) { 1424 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1425 return; 1426 } 1427 ip->ip_ttl -= IPTTLDEC; 1428 1429 sin = satosin(&ipforward_rt.ro_dst); 1430 if ((rt = ipforward_rt.ro_rt) == 0 || 1431 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1432 if (ipforward_rt.ro_rt) { 1433 RTFREE(ipforward_rt.ro_rt); 1434 ipforward_rt.ro_rt = 0; 1435 } 1436 sin->sin_family = AF_INET; 1437 sin->sin_len = sizeof(struct sockaddr_in); 1438 sin->sin_addr = ip->ip_dst; 1439 1440 rtalloc(&ipforward_rt); 1441 if (ipforward_rt.ro_rt == 0) { 1442 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1443 return; 1444 } 1445 rt = ipforward_rt.ro_rt; 1446 } 1447 1448 /* 1449 * Save at most 68 bytes of the packet in case 1450 * we need to generate an ICMP message to the src. 1451 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1452 */ 1453 mcopy = m_copym(m, 0, imin((int)ip->ip_len, 68), M_DONTWAIT); 1454 if (mcopy) 1455 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1456 1457 /* 1458 * If forwarding packet using same interface that it came in on, 1459 * perhaps should send a redirect to sender to shortcut a hop. 1460 * Only send redirect if source is sending directly to us, 1461 * and if packet was not source routed (or has any options). 1462 * Also, don't send redirect if forwarding using a default route 1463 * or a route modified by a redirect. 1464 */ 1465 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1466 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1467 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1468 ipsendredirects && !srcrt) { 1469 if (rt->rt_ifa && 1470 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1471 ifatoia(rt->rt_ifa)->ia_subnet) { 1472 if (rt->rt_flags & RTF_GATEWAY) 1473 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1474 else 1475 dest = ip->ip_dst.s_addr; 1476 /* 1477 * Router requirements says to only send host 1478 * redirects. 1479 */ 1480 type = ICMP_REDIRECT; 1481 code = ICMP_REDIRECT_HOST; 1482 #ifdef DIAGNOSTIC 1483 if (ipprintfs) 1484 printf("redirect (%d) to %x\n", code, 1485 (u_int32_t)dest); 1486 #endif 1487 } 1488 } 1489 1490 #ifdef IPSEC 1491 /* Don't lookup socket in forwading case */ 1492 (void)ipsec_setsocket(m, NULL); 1493 #endif 1494 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1495 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1496 if (error) 1497 ipstat.ips_cantforward++; 1498 else { 1499 ipstat.ips_forward++; 1500 if (type) 1501 ipstat.ips_redirectsent++; 1502 else { 1503 if (mcopy) { 1504 #ifdef GATEWAY 1505 if (mcopy->m_flags & M_CANFASTFWD) 1506 ipflow_create(&ipforward_rt, mcopy); 1507 #endif 1508 m_freem(mcopy); 1509 } 1510 return; 1511 } 1512 } 1513 if (mcopy == NULL) 1514 return; 1515 destifp = NULL; 1516 1517 switch (error) { 1518 1519 case 0: /* forwarded, but need redirect */ 1520 /* type, code set above */ 1521 break; 1522 1523 case ENETUNREACH: /* shouldn't happen, checked above */ 1524 case EHOSTUNREACH: 1525 case ENETDOWN: 1526 case EHOSTDOWN: 1527 default: 1528 type = ICMP_UNREACH; 1529 code = ICMP_UNREACH_HOST; 1530 break; 1531 1532 case EMSGSIZE: 1533 type = ICMP_UNREACH; 1534 code = ICMP_UNREACH_NEEDFRAG; 1535 #ifndef IPSEC 1536 if (ipforward_rt.ro_rt) 1537 destifp = ipforward_rt.ro_rt->rt_ifp; 1538 #else 1539 /* 1540 * If the packet is routed over IPsec tunnel, tell the 1541 * originator the tunnel MTU. 1542 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1543 * XXX quickhack!!! 1544 */ 1545 if (ipforward_rt.ro_rt) { 1546 struct secpolicy *sp; 1547 int ipsecerror; 1548 size_t ipsechdr; 1549 struct route *ro; 1550 1551 sp = ipsec4_getpolicybyaddr(mcopy, 1552 IPSEC_DIR_OUTBOUND, 1553 IP_FORWARDING, 1554 &ipsecerror); 1555 1556 if (sp == NULL) 1557 destifp = ipforward_rt.ro_rt->rt_ifp; 1558 else { 1559 /* count IPsec header size */ 1560 ipsechdr = ipsec4_hdrsiz(mcopy, 1561 IPSEC_DIR_OUTBOUND, 1562 NULL); 1563 1564 /* 1565 * find the correct route for outer IPv4 1566 * header, compute tunnel MTU. 1567 * 1568 * XXX BUG ALERT 1569 * The "dummyifp" code relies upon the fact 1570 * that icmp_error() touches only ifp->if_mtu. 1571 */ 1572 /*XXX*/ 1573 destifp = NULL; 1574 if (sp->req != NULL 1575 && sp->req->sav != NULL 1576 && sp->req->sav->sah != NULL) { 1577 ro = &sp->req->sav->sah->sa_route; 1578 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1579 dummyifp.if_mtu = 1580 ro->ro_rt->rt_ifp->if_mtu; 1581 dummyifp.if_mtu -= ipsechdr; 1582 destifp = &dummyifp; 1583 } 1584 } 1585 1586 key_freesp(sp); 1587 } 1588 } 1589 #endif /*IPSEC*/ 1590 ipstat.ips_cantfrag++; 1591 break; 1592 1593 case ENOBUFS: 1594 type = ICMP_SOURCEQUENCH; 1595 code = 0; 1596 break; 1597 } 1598 icmp_error(mcopy, type, code, dest, destifp); 1599 } 1600 1601 void 1602 ip_savecontrol(inp, mp, ip, m) 1603 struct inpcb *inp; 1604 struct mbuf **mp; 1605 struct ip *ip; 1606 struct mbuf *m; 1607 { 1608 1609 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1610 struct timeval tv; 1611 1612 microtime(&tv); 1613 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1614 SCM_TIMESTAMP, SOL_SOCKET); 1615 if (*mp) 1616 mp = &(*mp)->m_next; 1617 } 1618 if (inp->inp_flags & INP_RECVDSTADDR) { 1619 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1620 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1621 if (*mp) 1622 mp = &(*mp)->m_next; 1623 } 1624 #ifdef notyet 1625 /* 1626 * XXX 1627 * Moving these out of udp_input() made them even more broken 1628 * than they already were. 1629 * - fenner (at) parc.xerox.com 1630 */ 1631 /* options were tossed already */ 1632 if (inp->inp_flags & INP_RECVOPTS) { 1633 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1634 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1635 if (*mp) 1636 mp = &(*mp)->m_next; 1637 } 1638 /* ip_srcroute doesn't do what we want here, need to fix */ 1639 if (inp->inp_flags & INP_RECVRETOPTS) { 1640 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1641 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1642 if (*mp) 1643 mp = &(*mp)->m_next; 1644 } 1645 #endif 1646 if (inp->inp_flags & INP_RECVIF) { 1647 struct sockaddr_dl sdl; 1648 1649 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1650 sdl.sdl_family = AF_LINK; 1651 sdl.sdl_index = m->m_pkthdr.rcvif ? 1652 m->m_pkthdr.rcvif->if_index : 0; 1653 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1654 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1655 IP_RECVIF, IPPROTO_IP); 1656 if (*mp) 1657 mp = &(*mp)->m_next; 1658 } 1659 } 1660 1661 int 1662 ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1663 int *name; 1664 u_int namelen; 1665 void *oldp; 1666 size_t *oldlenp; 1667 void *newp; 1668 size_t newlen; 1669 { 1670 extern int subnetsarelocal, hostzeroisbroadcast; 1671 1672 int error, old; 1673 1674 /* All sysctl names at this level are terminal. */ 1675 if (namelen != 1) 1676 return (ENOTDIR); 1677 1678 switch (name[0]) { 1679 case IPCTL_FORWARDING: 1680 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1681 case IPCTL_SENDREDIRECTS: 1682 return (sysctl_int(oldp, oldlenp, newp, newlen, 1683 &ipsendredirects)); 1684 case IPCTL_DEFTTL: 1685 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1686 #ifdef notyet 1687 case IPCTL_DEFMTU: 1688 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1689 #endif 1690 case IPCTL_FORWSRCRT: 1691 /* Don't allow this to change in a secure environment. */ 1692 if (securelevel > 0) 1693 return (sysctl_rdint(oldp, oldlenp, newp, 1694 ip_forwsrcrt)); 1695 else 1696 return (sysctl_int(oldp, oldlenp, newp, newlen, 1697 &ip_forwsrcrt)); 1698 case IPCTL_DIRECTEDBCAST: 1699 return (sysctl_int(oldp, oldlenp, newp, newlen, 1700 &ip_directedbcast)); 1701 case IPCTL_ALLOWSRCRT: 1702 return (sysctl_int(oldp, oldlenp, newp, newlen, 1703 &ip_allowsrcrt)); 1704 case IPCTL_SUBNETSARELOCAL: 1705 return (sysctl_int(oldp, oldlenp, newp, newlen, 1706 &subnetsarelocal)); 1707 case IPCTL_MTUDISC: 1708 error = sysctl_int(oldp, oldlenp, newp, newlen, 1709 &ip_mtudisc); 1710 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1711 ip_mtudisc_timeout_q = 1712 rt_timer_queue_create(ip_mtudisc_timeout); 1713 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1714 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1715 ip_mtudisc_timeout_q = NULL; 1716 } 1717 return error; 1718 case IPCTL_ANONPORTMIN: 1719 old = anonportmin; 1720 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1721 if (anonportmin >= anonportmax || anonportmin < 0 1722 || anonportmin > 65535 1723 #ifndef IPNOPRIVPORTS 1724 || anonportmin < IPPORT_RESERVED 1725 #endif 1726 ) { 1727 anonportmin = old; 1728 return (EINVAL); 1729 } 1730 return (error); 1731 case IPCTL_ANONPORTMAX: 1732 old = anonportmax; 1733 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1734 if (anonportmin >= anonportmax || anonportmax < 0 1735 || anonportmax > 65535 1736 #ifndef IPNOPRIVPORTS 1737 || anonportmax < IPPORT_RESERVED 1738 #endif 1739 ) { 1740 anonportmax = old; 1741 return (EINVAL); 1742 } 1743 return (error); 1744 case IPCTL_MTUDISCTIMEOUT: 1745 error = sysctl_int(oldp, oldlenp, newp, newlen, 1746 &ip_mtudisc_timeout); 1747 if (ip_mtudisc_timeout_q != NULL) 1748 rt_timer_queue_change(ip_mtudisc_timeout_q, 1749 ip_mtudisc_timeout); 1750 return (error); 1751 #ifdef GATEWAY 1752 case IPCTL_MAXFLOWS: 1753 { 1754 int s; 1755 1756 error = sysctl_int(oldp, oldlenp, newp, newlen, 1757 &ip_maxflows); 1758 s = splsoftnet(); 1759 ipflow_reap(0); 1760 splx(s); 1761 return (error); 1762 } 1763 #endif 1764 case IPCTL_HOSTZEROBROADCAST: 1765 return (sysctl_int(oldp, oldlenp, newp, newlen, 1766 &hostzeroisbroadcast)); 1767 #if NGIF > 0 1768 case IPCTL_GIF_TTL: 1769 return(sysctl_int(oldp, oldlenp, newp, newlen, 1770 &ip_gif_ttl)); 1771 #endif 1772 1773 #ifndef IPNOPRIVPORTS 1774 case IPCTL_LOWPORTMIN: 1775 old = lowportmin; 1776 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmin); 1777 if (lowportmin >= lowportmax 1778 || lowportmin > IPPORT_RESERVEDMAX 1779 || lowportmin < IPPORT_RESERVEDMIN 1780 ) { 1781 lowportmin = old; 1782 return (EINVAL); 1783 } 1784 return (error); 1785 case IPCTL_LOWPORTMAX: 1786 old = lowportmax; 1787 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmax); 1788 if (lowportmin >= lowportmax 1789 || lowportmax > IPPORT_RESERVEDMAX 1790 || lowportmax < IPPORT_RESERVEDMIN 1791 ) { 1792 lowportmax = old; 1793 return (EINVAL); 1794 } 1795 return (error); 1796 #endif 1797 1798 case IPCTL_MAXFRAGPACKETS: 1799 return (sysctl_int(oldp, oldlenp, newp, newlen, 1800 &ip_maxfragpackets)); 1801 1802 default: 1803 return (EOPNOTSUPP); 1804 } 1805 /* NOTREACHED */ 1806 } 1807
Indexes created Wed Oct 22 13:09:56 GMT 2025