ip_input.c revision 91271
1/* 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 34 * $FreeBSD: head/sys/netinet/ip_input.c 91271 2002-02-26 02:11:13Z jedgar $ 35 */ 36 37#define _IP_VHL 38 39#include "opt_bootp.h" 40#include "opt_ipfw.h" 41#include "opt_ipdn.h" 42#include "opt_ipdivert.h" 43#include "opt_ipfilter.h" 44#include "opt_ipstealth.h" 45#include "opt_ipsec.h" 46#include "opt_pfil_hooks.h" 47#include "opt_random_ip_id.h" 48 49#include <sys/param.h> 50#include <sys/systm.h> 51#include <sys/mbuf.h> 52#include <sys/malloc.h> 53#include <sys/domain.h> 54#include <sys/protosw.h> 55#include <sys/socket.h> 56#include <sys/time.h> 57#include <sys/kernel.h> 58#include <sys/syslog.h> 59#include <sys/sysctl.h> 60 61#include <net/pfil.h> 62#include <net/if.h> 63#include <net/if_types.h> 64#include <net/if_var.h> 65#include <net/if_dl.h> 66#include <net/route.h> 67#include <net/netisr.h> 68#include <net/intrq.h> 69 70#include <netinet/in.h> 71#include <netinet/in_systm.h> 72#include <netinet/in_var.h> 73#include <netinet/ip.h> 74#include <netinet/in_pcb.h> 75#include <netinet/ip_var.h> 76#include <netinet/ip_icmp.h> 77#include <machine/in_cksum.h> 78 79#include <sys/socketvar.h> 80 81#include <netinet/ip_fw.h> 82#include <netinet/ip_dummynet.h> 83 84#ifdef IPSEC 85#include <netinet6/ipsec.h> 86#include <netkey/key.h> 87#endif 88 89int rsvp_on = 0; 90static int ip_rsvp_on; 91struct socket *ip_rsvpd; 92 93int ipforwarding = 0; 94SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 95 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 96 97static int ipsendredirects = 1; /* XXX */ 98SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 99 &ipsendredirects, 0, "Enable sending IP redirects"); 100 101int ip_defttl = IPDEFTTL; 102SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 103 &ip_defttl, 0, "Maximum TTL on IP packets"); 104 105static int ip_dosourceroute = 0; 106SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 107 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 108 109static int ip_acceptsourceroute = 0; 110SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 111 CTLFLAG_RW, &ip_acceptsourceroute, 0, 112 "Enable accepting source routed IP packets"); 113 114static int ip_keepfaith = 0; 115SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 116 &ip_keepfaith, 0, 117 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 118 119static int ip_nfragpackets = 0; 120static int ip_maxfragpackets; /* initialized in ip_init() */ 121SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 122 &ip_maxfragpackets, 0, 123 "Maximum number of IPv4 fragment reassembly queue entries"); 124 125/* 126 * XXX - Setting ip_checkinterface mostly implements the receive side of 127 * the Strong ES model described in RFC 1122, but since the routing table 128 * and transmit implementation do not implement the Strong ES model, 129 * setting this to 1 results in an odd hybrid. 130 * 131 * XXX - ip_checkinterface currently must be disabled if you use ipnat 132 * to translate the destination address to another local interface. 133 * 134 * XXX - ip_checkinterface must be disabled if you add IP aliases 135 * to the loopback interface instead of the interface where the 136 * packets for those addresses are received. 137 */ 138static int ip_checkinterface = 1; 139SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 140 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 141 142#ifdef DIAGNOSTIC 143static int ipprintfs = 0; 144#endif 145 146static int ipqmaxlen = IFQ_MAXLEN; 147 148extern struct domain inetdomain; 149extern struct protosw inetsw[]; 150u_char ip_protox[IPPROTO_MAX]; 151struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 152struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 153u_long in_ifaddrhmask; /* mask for hash table */ 154 155SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 156 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 157SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 158 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 159 160struct ipstat ipstat; 161SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 162 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 163 164/* Packet reassembly stuff */ 165#define IPREASS_NHASH_LOG2 6 166#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 167#define IPREASS_HMASK (IPREASS_NHASH - 1) 168#define IPREASS_HASH(x,y) \ 169 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 170 171static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 172static int nipq = 0; /* total # of reass queues */ 173static int maxnipq; 174 175#ifdef IPCTL_DEFMTU 176SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 177 &ip_mtu, 0, "Default MTU"); 178#endif 179 180#ifdef IPSTEALTH 181static int ipstealth = 0; 182SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 183 &ipstealth, 0, ""); 184#endif 185 186 187/* Firewall hooks */ 188ip_fw_chk_t *ip_fw_chk_ptr; 189int fw_enable = 1 ; 190 191/* Dummynet hooks */ 192ip_dn_io_t *ip_dn_io_ptr; 193 194 195/* 196 * We need to save the IP options in case a protocol wants to respond 197 * to an incoming packet over the same route if the packet got here 198 * using IP source routing. This allows connection establishment and 199 * maintenance when the remote end is on a network that is not known 200 * to us. 201 */ 202static int ip_nhops = 0; 203static struct ip_srcrt { 204 struct in_addr dst; /* final destination */ 205 char nop; /* one NOP to align */ 206 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 207 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 208} ip_srcrt; 209 210struct sockaddr_in *ip_fw_fwd_addr; 211 212static void save_rte __P((u_char *, struct in_addr)); 213static int ip_dooptions __P((struct mbuf *, int)); 214static void ip_forward __P((struct mbuf *, int)); 215static void ip_freef __P((struct ipqhead *, struct ipq *)); 216#ifdef IPDIVERT 217static struct mbuf *ip_reass __P((struct mbuf *, struct ipqhead *, struct ipq *, u_int32_t *, u_int16_t *)); 218#else 219static struct mbuf *ip_reass __P((struct mbuf *, struct ipqhead *, struct ipq *)); 220#endif 221static void ipintr __P((void)); 222 223/* 224 * IP initialization: fill in IP protocol switch table. 225 * All protocols not implemented in kernel go to raw IP protocol handler. 226 */ 227void 228ip_init() 229{ 230 register struct protosw *pr; 231 register int i; 232 233 TAILQ_INIT(&in_ifaddrhead); 234 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 235 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 236 if (pr == 0) 237 panic("ip_init"); 238 for (i = 0; i < IPPROTO_MAX; i++) 239 ip_protox[i] = pr - inetsw; 240 for (pr = inetdomain.dom_protosw; 241 pr < inetdomain.dom_protoswNPROTOSW; pr++) 242 if (pr->pr_domain->dom_family == PF_INET && 243 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 244 ip_protox[pr->pr_protocol] = pr - inetsw; 245 246 for (i = 0; i < IPREASS_NHASH; i++) 247 TAILQ_INIT(&ipq[i]); 248 249 maxnipq = nmbclusters / 4; 250 ip_maxfragpackets = nmbclusters / 4; 251 252#ifndef RANDOM_IP_ID 253 ip_id = time_second & 0xffff; 254#endif 255 ipintrq.ifq_maxlen = ipqmaxlen; 256 mtx_init(&ipintrq.ifq_mtx, "ip_inq", MTX_DEF); 257 ipintrq_present = 1; 258 259 register_netisr(NETISR_IP, ipintr); 260} 261 262static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 263struct route ipforward_rt; 264 265/* 266 * Ip input routine. Checksum and byte swap header. If fragmented 267 * try to reassemble. Process options. Pass to next level. 268 */ 269void 270ip_input(struct mbuf *m) 271{ 272 struct ip *ip; 273 struct ipq *fp; 274 struct in_ifaddr *ia = NULL; 275 struct ifaddr *ifa; 276 int i, hlen, checkif; 277 u_short sum; 278 u_int16_t divert_cookie; /* firewall cookie */ 279 struct in_addr pkt_dst; 280#ifdef IPDIVERT 281 u_int32_t divert_info = 0; /* packet divert/tee info */ 282#endif 283 struct ip_fw *rule = NULL; 284#ifdef PFIL_HOOKS 285 struct packet_filter_hook *pfh; 286 struct mbuf *m0; 287 int rv; 288#endif /* PFIL_HOOKS */ 289 290#ifdef IPDIVERT 291 /* Get and reset firewall cookie */ 292 divert_cookie = ip_divert_cookie; 293 ip_divert_cookie = 0; 294#else 295 divert_cookie = 0; 296#endif 297 298 /* 299 * dummynet packet are prepended a vestigial mbuf with 300 * m_type = MT_DUMMYNET and m_data pointing to the matching 301 * rule. 302 */ 303 if (m->m_type == MT_DUMMYNET) { 304 rule = (struct ip_fw *)(m->m_data) ; 305 m = m->m_next ; 306 ip = mtod(m, struct ip *); 307 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 308 goto iphack ; 309 } else 310 rule = NULL ; 311 312#ifdef DIAGNOSTIC 313 if (m == NULL || (m->m_flags & M_PKTHDR) == 0) 314 panic("ip_input no HDR"); 315#endif 316 ipstat.ips_total++; 317 318 if (m->m_pkthdr.len < sizeof(struct ip)) 319 goto tooshort; 320 321 if (m->m_len < sizeof (struct ip) && 322 (m = m_pullup(m, sizeof (struct ip))) == 0) { 323 ipstat.ips_toosmall++; 324 return; 325 } 326 ip = mtod(m, struct ip *); 327 328 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 329 ipstat.ips_badvers++; 330 goto bad; 331 } 332 333 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 334 if (hlen < sizeof(struct ip)) { /* minimum header length */ 335 ipstat.ips_badhlen++; 336 goto bad; 337 } 338 if (hlen > m->m_len) { 339 if ((m = m_pullup(m, hlen)) == 0) { 340 ipstat.ips_badhlen++; 341 return; 342 } 343 ip = mtod(m, struct ip *); 344 } 345 346 /* 127/8 must not appear on wire - RFC1122 */ 347 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 348 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 349 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 350 ipstat.ips_badaddr++; 351 goto bad; 352 } 353 } 354 355 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 356 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 357 } else { 358 if (hlen == sizeof(struct ip)) { 359 sum = in_cksum_hdr(ip); 360 } else { 361 sum = in_cksum(m, hlen); 362 } 363 } 364 if (sum) { 365 ipstat.ips_badsum++; 366 goto bad; 367 } 368 369 /* 370 * Convert fields to host representation. 371 */ 372 ip->ip_len = ntohs(ip->ip_len); 373 if (ip->ip_len < hlen) { 374 ipstat.ips_badlen++; 375 goto bad; 376 } 377 ip->ip_off = ntohs(ip->ip_off); 378 379 /* 380 * Check that the amount of data in the buffers 381 * is as at least much as the IP header would have us expect. 382 * Trim mbufs if longer than we expect. 383 * Drop packet if shorter than we expect. 384 */ 385 if (m->m_pkthdr.len < ip->ip_len) { 386tooshort: 387 ipstat.ips_tooshort++; 388 goto bad; 389 } 390 if (m->m_pkthdr.len > ip->ip_len) { 391 if (m->m_len == m->m_pkthdr.len) { 392 m->m_len = ip->ip_len; 393 m->m_pkthdr.len = ip->ip_len; 394 } else 395 m_adj(m, ip->ip_len - m->m_pkthdr.len); 396 } 397 398#ifdef IPSEC 399 if (ipsec_gethist(m, NULL)) 400 goto pass; 401#endif 402 403 /* 404 * IpHack's section. 405 * Right now when no processing on packet has done 406 * and it is still fresh out of network we do our black 407 * deals with it. 408 * - Firewall: deny/allow/divert 409 * - Xlate: translate packet's addr/port (NAT). 410 * - Pipe: pass pkt through dummynet. 411 * - Wrap: fake packet's addr/port <unimpl.> 412 * - Encapsulate: put it in another IP and send out. <unimp.> 413 */ 414 415iphack: 416 417#ifdef PFIL_HOOKS 418 /* 419 * Run through list of hooks for input packets. If there are any 420 * filters which require that additional packets in the flow are 421 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 422 * Note that filters must _never_ set this flag, as another filter 423 * in the list may have previously cleared it. 424 */ 425 m0 = m; 426 pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 427 for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link)) 428 if (pfh->pfil_func) { 429 rv = pfh->pfil_func(ip, hlen, 430 m->m_pkthdr.rcvif, 0, &m0); 431 if (rv) 432 return; 433 m = m0; 434 if (m == NULL) 435 return; 436 ip = mtod(m, struct ip *); 437 } 438#endif /* PFIL_HOOKS */ 439 440 if (fw_enable && IPFW_LOADED) { 441#ifdef IPFIREWALL_FORWARD 442 /* 443 * If we've been forwarded from the output side, then 444 * skip the firewall a second time 445 */ 446 if (ip_fw_fwd_addr) 447 goto ours; 448#endif /* IPFIREWALL_FORWARD */ 449 /* 450 * See the comment in ip_output for the return values 451 * produced by the firewall. 452 */ 453 i = ip_fw_chk_ptr(&ip, hlen, NULL, 454 &divert_cookie, &m, &rule, &ip_fw_fwd_addr); 455 if (i & IP_FW_PORT_DENY_FLAG) { /* XXX new interface-denied */ 456 if (m) 457 m_freem(m); 458 return; 459 } 460 if (m == NULL) { /* Packet discarded by firewall */ 461 static int __debug=10; 462 if (__debug > 0) { 463 printf( 464 "firewall returns NULL, please update!\n"); 465 __debug--; 466 } 467 return; 468 } 469 if (i == 0 && ip_fw_fwd_addr == NULL) /* common case */ 470 goto pass; 471 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) { 472 /* Send packet to the appropriate pipe */ 473 ip_dn_io_ptr(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule, 474 0); 475 return; 476 } 477#ifdef IPDIVERT 478 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 479 /* Divert or tee packet */ 480 divert_info = i; 481 goto ours; 482 } 483#endif 484#ifdef IPFIREWALL_FORWARD 485 if (i == 0 && ip_fw_fwd_addr != NULL) 486 goto pass; 487#endif 488 /* 489 * if we get here, the packet must be dropped 490 */ 491 m_freem(m); 492 return; 493 } 494pass: 495 496 /* 497 * Process options and, if not destined for us, 498 * ship it on. ip_dooptions returns 1 when an 499 * error was detected (causing an icmp message 500 * to be sent and the original packet to be freed). 501 */ 502 ip_nhops = 0; /* for source routed packets */ 503 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) { 504#ifdef IPFIREWALL_FORWARD 505 ip_fw_fwd_addr = NULL; 506#endif 507 return; 508 } 509 510 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 511 * matter if it is destined to another node, or whether it is 512 * a multicast one, RSVP wants it! and prevents it from being forwarded 513 * anywhere else. Also checks if the rsvp daemon is running before 514 * grabbing the packet. 515 */ 516 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 517 goto ours; 518 519 /* 520 * Check our list of addresses, to see if the packet is for us. 521 * If we don't have any addresses, assume any unicast packet 522 * we receive might be for us (and let the upper layers deal 523 * with it). 524 */ 525 if (TAILQ_EMPTY(&in_ifaddrhead) && 526 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 527 goto ours; 528 529 /* 530 * Cache the destination address of the packet; this may be 531 * changed by use of 'ipfw fwd'. 532 */ 533 pkt_dst = ip_fw_fwd_addr == NULL ? 534 ip->ip_dst : ip_fw_fwd_addr->sin_addr; 535 536 /* 537 * Enable a consistency check between the destination address 538 * and the arrival interface for a unicast packet (the RFC 1122 539 * strong ES model) if IP forwarding is disabled and the packet 540 * is not locally generated and the packet is not subject to 541 * 'ipfw fwd'. 542 * 543 * XXX - Checking also should be disabled if the destination 544 * address is ipnat'ed to a different interface. 545 * 546 * XXX - Checking is incompatible with IP aliases added 547 * to the loopback interface instead of the interface where 548 * the packets are received. 549 */ 550 checkif = ip_checkinterface && (ipforwarding == 0) && 551 m->m_pkthdr.rcvif != NULL && 552 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 553 (ip_fw_fwd_addr == NULL); 554 555 /* 556 * Check for exact addresses in the hash bucket. 557 */ 558 LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) { 559 /* 560 * If the address matches, verify that the packet 561 * arrived via the correct interface if checking is 562 * enabled. 563 */ 564 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 565 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 566 goto ours; 567 } 568 /* 569 * Check for broadcast addresses. 570 * 571 * Only accept broadcast packets that arrive via the matching 572 * interface. Reception of forwarded directed broadcasts would 573 * be handled via ip_forward() and ether_output() with the loopback 574 * into the stack for SIMPLEX interfaces handled by ether_output(). 575 */ 576 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 577 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 578 if (ifa->ifa_addr->sa_family != AF_INET) 579 continue; 580 ia = ifatoia(ifa); 581 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 582 pkt_dst.s_addr) 583 goto ours; 584 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 585 goto ours; 586#ifdef BOOTP_COMPAT 587 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 588 goto ours; 589#endif 590 } 591 } 592 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 593 struct in_multi *inm; 594 if (ip_mrouter) { 595 /* 596 * If we are acting as a multicast router, all 597 * incoming multicast packets are passed to the 598 * kernel-level multicast forwarding function. 599 * The packet is returned (relatively) intact; if 600 * ip_mforward() returns a non-zero value, the packet 601 * must be discarded, else it may be accepted below. 602 */ 603 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 604 ipstat.ips_cantforward++; 605 m_freem(m); 606 return; 607 } 608 609 /* 610 * The process-level routing demon needs to receive 611 * all multicast IGMP packets, whether or not this 612 * host belongs to their destination groups. 613 */ 614 if (ip->ip_p == IPPROTO_IGMP) 615 goto ours; 616 ipstat.ips_forward++; 617 } 618 /* 619 * See if we belong to the destination multicast group on the 620 * arrival interface. 621 */ 622 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 623 if (inm == NULL) { 624 ipstat.ips_notmember++; 625 m_freem(m); 626 return; 627 } 628 goto ours; 629 } 630 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 631 goto ours; 632 if (ip->ip_dst.s_addr == INADDR_ANY) 633 goto ours; 634 635 /* 636 * FAITH(Firewall Aided Internet Translator) 637 */ 638 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 639 if (ip_keepfaith) { 640 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 641 goto ours; 642 } 643 m_freem(m); 644 return; 645 } 646 647 /* 648 * Not for us; forward if possible and desirable. 649 */ 650 if (ipforwarding == 0) { 651 ipstat.ips_cantforward++; 652 m_freem(m); 653 } else { 654#ifdef IPSEC 655 /* 656 * Enforce inbound IPsec SPD. 657 */ 658 if (ipsec4_in_reject(m, NULL)) { 659 ipsecstat.in_polvio++; 660 goto bad; 661 } 662#endif /* IPSEC */ 663 ip_forward(m, 0); 664 } 665#ifdef IPFIREWALL_FORWARD 666 ip_fw_fwd_addr = NULL; 667#endif 668 return; 669 670ours: 671#ifdef IPSTEALTH 672 /* 673 * IPSTEALTH: Process non-routing options only 674 * if the packet is destined for us. 675 */ 676 if (ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) { 677#ifdef IPFIREWALL_FORWARD 678 ip_fw_fwd_addr = NULL; 679#endif 680 return; 681 } 682#endif /* IPSTEALTH */ 683 684 /* Count the packet in the ip address stats */ 685 if (ia != NULL) { 686 ia->ia_ifa.if_ipackets++; 687 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 688 } 689 690 /* 691 * If offset or IP_MF are set, must reassemble. 692 * Otherwise, nothing need be done. 693 * (We could look in the reassembly queue to see 694 * if the packet was previously fragmented, 695 * but it's not worth the time; just let them time out.) 696 */ 697 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 698 699 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 700 /* 701 * Look for queue of fragments 702 * of this datagram. 703 */ 704 TAILQ_FOREACH(fp, &ipq[sum], ipq_list) 705 if (ip->ip_id == fp->ipq_id && 706 ip->ip_src.s_addr == fp->ipq_src.s_addr && 707 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 708 ip->ip_p == fp->ipq_p) 709 goto found; 710 711 fp = 0; 712 713 /* check if there's a place for the new queue */ 714 if (nipq > maxnipq) { 715 /* 716 * drop something from the tail of the current queue 717 * before proceeding further 718 */ 719 struct ipq *q = TAILQ_LAST(&ipq[sum], ipqhead); 720 if (q == NULL) { /* gak */ 721 for (i = 0; i < IPREASS_NHASH; i++) { 722 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 723 if (r) { 724 ip_freef(&ipq[i], r); 725 break; 726 } 727 } 728 } else 729 ip_freef(&ipq[sum], q); 730 } 731found: 732 /* 733 * Adjust ip_len to not reflect header, 734 * convert offset of this to bytes. 735 */ 736 ip->ip_len -= hlen; 737 if (ip->ip_off & IP_MF) { 738 /* 739 * Make sure that fragments have a data length 740 * that's a non-zero multiple of 8 bytes. 741 */ 742 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 743 ipstat.ips_toosmall++; /* XXX */ 744 goto bad; 745 } 746 m->m_flags |= M_FRAG; 747 } 748 ip->ip_off <<= 3; 749 750 /* 751 * Attempt reassembly; if it succeeds, proceed. 752 */ 753 ipstat.ips_fragments++; 754 m->m_pkthdr.header = ip; 755#ifdef IPDIVERT 756 m = ip_reass(m, 757 &ipq[sum], fp, &divert_info, &divert_cookie); 758#else 759 m = ip_reass(m, &ipq[sum], fp); 760#endif 761 if (m == 0) { 762#ifdef IPFIREWALL_FORWARD 763 ip_fw_fwd_addr = NULL; 764#endif 765 return; 766 } 767 ipstat.ips_reassembled++; 768 ip = mtod(m, struct ip *); 769 /* Get the header length of the reassembled packet */ 770 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 771#ifdef IPDIVERT 772 /* Restore original checksum before diverting packet */ 773 if (divert_info != 0) { 774 ip->ip_len += hlen; 775 ip->ip_len = htons(ip->ip_len); 776 ip->ip_off = htons(ip->ip_off); 777 ip->ip_sum = 0; 778 if (hlen == sizeof(struct ip)) 779 ip->ip_sum = in_cksum_hdr(ip); 780 else 781 ip->ip_sum = in_cksum(m, hlen); 782 ip->ip_off = ntohs(ip->ip_off); 783 ip->ip_len = ntohs(ip->ip_len); 784 ip->ip_len -= hlen; 785 } 786#endif 787 } else 788 ip->ip_len -= hlen; 789 790#ifdef IPDIVERT 791 /* 792 * Divert or tee packet to the divert protocol if required. 793 * 794 * If divert_info is zero then cookie should be too, so we shouldn't 795 * need to clear them here. Assume divert_packet() does so also. 796 */ 797 if (divert_info != 0) { 798 struct mbuf *clone = NULL; 799 800 /* Clone packet if we're doing a 'tee' */ 801 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 802 clone = m_dup(m, M_DONTWAIT); 803 804 /* Restore packet header fields to original values */ 805 ip->ip_len += hlen; 806 ip->ip_len = htons(ip->ip_len); 807 ip->ip_off = htons(ip->ip_off); 808 809 /* Deliver packet to divert input routine */ 810 ip_divert_cookie = divert_cookie; 811 divert_packet(m, 1, divert_info & 0xffff); 812 ipstat.ips_delivered++; 813 814 /* If 'tee', continue with original packet */ 815 if (clone == NULL) 816 return; 817 m = clone; 818 ip = mtod(m, struct ip *); 819 ip->ip_len += hlen; 820 divert_info = 0; 821 goto pass; 822 } 823#endif 824 825#ifdef IPSEC 826 /* 827 * enforce IPsec policy checking if we are seeing last header. 828 * note that we do not visit this with protocols with pcb layer 829 * code - like udp/tcp/raw ip. 830 */ 831 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 832 ipsec4_in_reject(m, NULL)) { 833 ipsecstat.in_polvio++; 834 goto bad; 835 } 836#endif 837 838 /* 839 * Switch out to protocol's input routine. 840 */ 841 ipstat.ips_delivered++; 842 { 843 int off = hlen; 844 845 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off); 846#ifdef IPFIREWALL_FORWARD 847 ip_fw_fwd_addr = NULL; /* tcp needed it */ 848#endif 849 return; 850 } 851bad: 852#ifdef IPFIREWALL_FORWARD 853 ip_fw_fwd_addr = NULL; 854#endif 855 m_freem(m); 856} 857 858/* 859 * IP software interrupt routine - to go away sometime soon 860 */ 861static void 862ipintr(void) 863{ 864 struct mbuf *m; 865 866 while (1) { 867 IF_DEQUEUE(&ipintrq, m); 868 if (m == 0) 869 return; 870 ip_input(m); 871 } 872} 873 874/* 875 * Take incoming datagram fragment and try to reassemble it into 876 * whole datagram. If a chain for reassembly of this datagram already 877 * exists, then it is given as fp; otherwise have to make a chain. 878 * 879 * When IPDIVERT enabled, keep additional state with each packet that 880 * tells us if we need to divert or tee the packet we're building. 881 */ 882 883static struct mbuf * 884#ifdef IPDIVERT 885ip_reass(m, head, fp, divinfo, divcookie) 886#else 887ip_reass(m, head, fp) 888#endif 889 struct mbuf *m; 890 struct ipqhead *head; 891 struct ipq *fp; 892#ifdef IPDIVERT 893 u_int32_t *divinfo; 894 u_int16_t *divcookie; 895#endif 896{ 897 struct ip *ip = mtod(m, struct ip *); 898 register struct mbuf *p, *q, *nq; 899 struct mbuf *t; 900 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 901 int i, next; 902 903 /* 904 * Presence of header sizes in mbufs 905 * would confuse code below. 906 */ 907 m->m_data += hlen; 908 m->m_len -= hlen; 909 910 /* 911 * If first fragment to arrive, create a reassembly queue. 912 */ 913 if (fp == 0) { 914 /* 915 * Enforce upper bound on number of fragmented packets 916 * for which we attempt reassembly; 917 * If maxfrag is 0, never accept fragments. 918 * If maxfrag is -1, accept all fragments without limitation. 919 */ 920 if ((ip_maxfragpackets >= 0) && (ip_nfragpackets >= ip_maxfragpackets)) 921 goto dropfrag; 922 ip_nfragpackets++; 923 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 924 goto dropfrag; 925 fp = mtod(t, struct ipq *); 926 TAILQ_INSERT_HEAD(head, fp, ipq_list); 927 nipq++; 928 fp->ipq_ttl = IPFRAGTTL; 929 fp->ipq_p = ip->ip_p; 930 fp->ipq_id = ip->ip_id; 931 fp->ipq_src = ip->ip_src; 932 fp->ipq_dst = ip->ip_dst; 933 fp->ipq_frags = m; 934 m->m_nextpkt = NULL; 935#ifdef IPDIVERT 936 fp->ipq_div_info = 0; 937 fp->ipq_div_cookie = 0; 938#endif 939 goto inserted; 940 } 941 942#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 943 944 /* 945 * Find a segment which begins after this one does. 946 */ 947 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 948 if (GETIP(q)->ip_off > ip->ip_off) 949 break; 950 951 /* 952 * If there is a preceding segment, it may provide some of 953 * our data already. If so, drop the data from the incoming 954 * segment. If it provides all of our data, drop us, otherwise 955 * stick new segment in the proper place. 956 * 957 * If some of the data is dropped from the the preceding 958 * segment, then it's checksum is invalidated. 959 */ 960 if (p) { 961 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 962 if (i > 0) { 963 if (i >= ip->ip_len) 964 goto dropfrag; 965 m_adj(m, i); 966 m->m_pkthdr.csum_flags = 0; 967 ip->ip_off += i; 968 ip->ip_len -= i; 969 } 970 m->m_nextpkt = p->m_nextpkt; 971 p->m_nextpkt = m; 972 } else { 973 m->m_nextpkt = fp->ipq_frags; 974 fp->ipq_frags = m; 975 } 976 977 /* 978 * While we overlap succeeding segments trim them or, 979 * if they are completely covered, dequeue them. 980 */ 981 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 982 q = nq) { 983 i = (ip->ip_off + ip->ip_len) - 984 GETIP(q)->ip_off; 985 if (i < GETIP(q)->ip_len) { 986 GETIP(q)->ip_len -= i; 987 GETIP(q)->ip_off += i; 988 m_adj(q, i); 989 q->m_pkthdr.csum_flags = 0; 990 break; 991 } 992 nq = q->m_nextpkt; 993 m->m_nextpkt = nq; 994 m_freem(q); 995 } 996 997inserted: 998 999#ifdef IPDIVERT 1000 /* 1001 * Transfer firewall instructions to the fragment structure. 1002 * Any fragment diverting causes the whole packet to divert. 1003 */ 1004 fp->ipq_div_info = *divinfo; 1005 fp->ipq_div_cookie = *divcookie; 1006 *divinfo = 0; 1007 *divcookie = 0; 1008#endif 1009 1010 /* 1011 * Check for complete reassembly. 1012 */ 1013 next = 0; 1014 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1015 if (GETIP(q)->ip_off != next) 1016 return (0); 1017 next += GETIP(q)->ip_len; 1018 } 1019 /* Make sure the last packet didn't have the IP_MF flag */ 1020 if (p->m_flags & M_FRAG) 1021 return (0); 1022 1023 /* 1024 * Reassembly is complete. Make sure the packet is a sane size. 1025 */ 1026 q = fp->ipq_frags; 1027 ip = GETIP(q); 1028 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 1029 ipstat.ips_toolong++; 1030 ip_freef(head, fp); 1031 return (0); 1032 } 1033 1034 /* 1035 * Concatenate fragments. 1036 */ 1037 m = q; 1038 t = m->m_next; 1039 m->m_next = 0; 1040 m_cat(m, t); 1041 nq = q->m_nextpkt; 1042 q->m_nextpkt = 0; 1043 for (q = nq; q != NULL; q = nq) { 1044 nq = q->m_nextpkt; 1045 q->m_nextpkt = NULL; 1046 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1047 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1048 m_cat(m, q); 1049 } 1050 1051#ifdef IPDIVERT 1052 /* 1053 * Extract firewall instructions from the fragment structure. 1054 */ 1055 *divinfo = fp->ipq_div_info; 1056 *divcookie = fp->ipq_div_cookie; 1057#endif 1058 1059 /* 1060 * Create header for new ip packet by 1061 * modifying header of first packet; 1062 * dequeue and discard fragment reassembly header. 1063 * Make header visible. 1064 */ 1065 ip->ip_len = next; 1066 ip->ip_src = fp->ipq_src; 1067 ip->ip_dst = fp->ipq_dst; 1068 TAILQ_REMOVE(head, fp, ipq_list); 1069 nipq--; 1070 (void) m_free(dtom(fp)); 1071 ip_nfragpackets--; 1072 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 1073 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 1074 /* some debugging cruft by sklower, below, will go away soon */ 1075 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1076 register int plen = 0; 1077 for (t = m; t; t = t->m_next) 1078 plen += t->m_len; 1079 m->m_pkthdr.len = plen; 1080 } 1081 return (m); 1082 1083dropfrag: 1084#ifdef IPDIVERT 1085 *divinfo = 0; 1086 *divcookie = 0; 1087#endif 1088 ipstat.ips_fragdropped++; 1089 m_freem(m); 1090 return (0); 1091 1092#undef GETIP 1093} 1094 1095/* 1096 * Free a fragment reassembly header and all 1097 * associated datagrams. 1098 */ 1099static void 1100ip_freef(fhp, fp) 1101 struct ipqhead *fhp; 1102 struct ipq *fp; 1103{ 1104 register struct mbuf *q; 1105 1106 while (fp->ipq_frags) { 1107 q = fp->ipq_frags; 1108 fp->ipq_frags = q->m_nextpkt; 1109 m_freem(q); 1110 } 1111 TAILQ_REMOVE(fhp, fp, ipq_list); 1112 (void) m_free(dtom(fp)); 1113 ip_nfragpackets--; 1114 nipq--; 1115} 1116 1117/* 1118 * IP timer processing; 1119 * if a timer expires on a reassembly 1120 * queue, discard it. 1121 */ 1122void 1123ip_slowtimo() 1124{ 1125 register struct ipq *fp; 1126 int s = splnet(); 1127 int i; 1128 1129 for (i = 0; i < IPREASS_NHASH; i++) { 1130 for(fp = TAILQ_FIRST(&ipq[i]); fp;) { 1131 struct ipq *fpp; 1132 1133 fpp = fp; 1134 fp = TAILQ_NEXT(fp, ipq_list); 1135 if(--fpp->ipq_ttl == 0) { 1136 ipstat.ips_fragtimeout++; 1137 ip_freef(&ipq[i], fpp); 1138 } 1139 } 1140 } 1141 /* 1142 * If we are over the maximum number of fragments 1143 * (due to the limit being lowered), drain off 1144 * enough to get down to the new limit. 1145 */ 1146 for (i = 0; i < IPREASS_NHASH; i++) { 1147 if (ip_maxfragpackets >= 0) { 1148 while (ip_nfragpackets > ip_maxfragpackets && 1149 !TAILQ_EMPTY(&ipq[i])) { 1150 ipstat.ips_fragdropped++; 1151 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1152 } 1153 } 1154 } 1155 ipflow_slowtimo(); 1156 splx(s); 1157} 1158 1159/* 1160 * Drain off all datagram fragments. 1161 */ 1162void 1163ip_drain() 1164{ 1165 int i; 1166 1167 for (i = 0; i < IPREASS_NHASH; i++) { 1168 while(!TAILQ_EMPTY(&ipq[i])) { 1169 ipstat.ips_fragdropped++; 1170 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1171 } 1172 } 1173 in_rtqdrain(); 1174} 1175 1176/* 1177 * Do option processing on a datagram, 1178 * possibly discarding it if bad options are encountered, 1179 * or forwarding it if source-routed. 1180 * The pass argument is used when operating in the IPSTEALTH 1181 * mode to tell what options to process: 1182 * [LS]SRR (pass 0) or the others (pass 1). 1183 * The reason for as many as two passes is that when doing IPSTEALTH, 1184 * non-routing options should be processed only if the packet is for us. 1185 * Returns 1 if packet has been forwarded/freed, 1186 * 0 if the packet should be processed further. 1187 */ 1188static int 1189ip_dooptions(m, pass) 1190 struct mbuf *m; 1191 int pass; 1192{ 1193 register struct ip *ip = mtod(m, struct ip *); 1194 register u_char *cp; 1195 register struct in_ifaddr *ia; 1196 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1197 struct in_addr *sin, dst; 1198 n_time ntime; 1199 1200 dst = ip->ip_dst; 1201 cp = (u_char *)(ip + 1); 1202 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1203 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1204 opt = cp[IPOPT_OPTVAL]; 1205 if (opt == IPOPT_EOL) 1206 break; 1207 if (opt == IPOPT_NOP) 1208 optlen = 1; 1209 else { 1210 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1211 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1212 goto bad; 1213 } 1214 optlen = cp[IPOPT_OLEN]; 1215 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1216 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1217 goto bad; 1218 } 1219 } 1220 switch (opt) { 1221 1222 default: 1223 break; 1224 1225 /* 1226 * Source routing with record. 1227 * Find interface with current destination address. 1228 * If none on this machine then drop if strictly routed, 1229 * or do nothing if loosely routed. 1230 * Record interface address and bring up next address 1231 * component. If strictly routed make sure next 1232 * address is on directly accessible net. 1233 */ 1234 case IPOPT_LSRR: 1235 case IPOPT_SSRR: 1236#ifdef IPSTEALTH 1237 if (ipstealth && pass > 0) 1238 break; 1239#endif 1240 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1241 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1242 goto bad; 1243 } 1244 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1245 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1246 goto bad; 1247 } 1248 ipaddr.sin_addr = ip->ip_dst; 1249 ia = (struct in_ifaddr *) 1250 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1251 if (ia == 0) { 1252 if (opt == IPOPT_SSRR) { 1253 type = ICMP_UNREACH; 1254 code = ICMP_UNREACH_SRCFAIL; 1255 goto bad; 1256 } 1257 if (!ip_dosourceroute) 1258 goto nosourcerouting; 1259 /* 1260 * Loose routing, and not at next destination 1261 * yet; nothing to do except forward. 1262 */ 1263 break; 1264 } 1265 off--; /* 0 origin */ 1266 if (off > optlen - (int)sizeof(struct in_addr)) { 1267 /* 1268 * End of source route. Should be for us. 1269 */ 1270 if (!ip_acceptsourceroute) 1271 goto nosourcerouting; 1272 save_rte(cp, ip->ip_src); 1273 break; 1274 } 1275#ifdef IPSTEALTH 1276 if (ipstealth) 1277 goto dropit; 1278#endif 1279 if (!ip_dosourceroute) { 1280 if (ipforwarding) { 1281 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1282 /* 1283 * Acting as a router, so generate ICMP 1284 */ 1285nosourcerouting: 1286 strcpy(buf, inet_ntoa(ip->ip_dst)); 1287 log(LOG_WARNING, 1288 "attempted source route from %s to %s\n", 1289 inet_ntoa(ip->ip_src), buf); 1290 type = ICMP_UNREACH; 1291 code = ICMP_UNREACH_SRCFAIL; 1292 goto bad; 1293 } else { 1294 /* 1295 * Not acting as a router, so silently drop. 1296 */ 1297#ifdef IPSTEALTH 1298dropit: 1299#endif 1300 ipstat.ips_cantforward++; 1301 m_freem(m); 1302 return (1); 1303 } 1304 } 1305 1306 /* 1307 * locate outgoing interface 1308 */ 1309 (void)memcpy(&ipaddr.sin_addr, cp + off, 1310 sizeof(ipaddr.sin_addr)); 1311 1312 if (opt == IPOPT_SSRR) { 1313#define INA struct in_ifaddr * 1314#define SA struct sockaddr * 1315 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1316 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1317 } else 1318 ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt); 1319 if (ia == 0) { 1320 type = ICMP_UNREACH; 1321 code = ICMP_UNREACH_SRCFAIL; 1322 goto bad; 1323 } 1324 ip->ip_dst = ipaddr.sin_addr; 1325 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1326 sizeof(struct in_addr)); 1327 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1328 /* 1329 * Let ip_intr's mcast routing check handle mcast pkts 1330 */ 1331 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1332 break; 1333 1334 case IPOPT_RR: 1335#ifdef IPSTEALTH 1336 if (ipstealth && pass == 0) 1337 break; 1338#endif 1339 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1340 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1341 goto bad; 1342 } 1343 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1344 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1345 goto bad; 1346 } 1347 /* 1348 * If no space remains, ignore. 1349 */ 1350 off--; /* 0 origin */ 1351 if (off > optlen - (int)sizeof(struct in_addr)) 1352 break; 1353 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1354 sizeof(ipaddr.sin_addr)); 1355 /* 1356 * locate outgoing interface; if we're the destination, 1357 * use the incoming interface (should be same). 1358 */ 1359 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1360 (ia = ip_rtaddr(ipaddr.sin_addr, 1361 &ipforward_rt)) == 0) { 1362 type = ICMP_UNREACH; 1363 code = ICMP_UNREACH_HOST; 1364 goto bad; 1365 } 1366 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1367 sizeof(struct in_addr)); 1368 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1369 break; 1370 1371 case IPOPT_TS: 1372#ifdef IPSTEALTH 1373 if (ipstealth && pass == 0) 1374 break; 1375#endif 1376 code = cp - (u_char *)ip; 1377 if (optlen < 4 || optlen > 40) { 1378 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1379 goto bad; 1380 } 1381 if ((off = cp[IPOPT_OFFSET]) < 5) { 1382 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1383 goto bad; 1384 } 1385 if (off > optlen - (int)sizeof(int32_t)) { 1386 cp[IPOPT_OFFSET + 1] += (1 << 4); 1387 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1388 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1389 goto bad; 1390 } 1391 break; 1392 } 1393 off--; /* 0 origin */ 1394 sin = (struct in_addr *)(cp + off); 1395 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1396 1397 case IPOPT_TS_TSONLY: 1398 break; 1399 1400 case IPOPT_TS_TSANDADDR: 1401 if (off + sizeof(n_time) + 1402 sizeof(struct in_addr) > optlen) { 1403 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1404 goto bad; 1405 } 1406 ipaddr.sin_addr = dst; 1407 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1408 m->m_pkthdr.rcvif); 1409 if (ia == 0) 1410 continue; 1411 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1412 sizeof(struct in_addr)); 1413 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1414 break; 1415 1416 case IPOPT_TS_PRESPEC: 1417 if (off + sizeof(n_time) + 1418 sizeof(struct in_addr) > optlen) { 1419 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1420 goto bad; 1421 } 1422 (void)memcpy(&ipaddr.sin_addr, sin, 1423 sizeof(struct in_addr)); 1424 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1425 continue; 1426 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1427 break; 1428 1429 default: 1430 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1431 goto bad; 1432 } 1433 ntime = iptime(); 1434 (void)memcpy(cp + off, &ntime, sizeof(n_time)); 1435 cp[IPOPT_OFFSET] += sizeof(n_time); 1436 } 1437 } 1438 if (forward && ipforwarding) { 1439 ip_forward(m, 1); 1440 return (1); 1441 } 1442 return (0); 1443bad: 1444 icmp_error(m, type, code, 0, 0); 1445 ipstat.ips_badoptions++; 1446 return (1); 1447} 1448 1449/* 1450 * Given address of next destination (final or next hop), 1451 * return internet address info of interface to be used to get there. 1452 */ 1453struct in_ifaddr * 1454ip_rtaddr(dst, rt) 1455 struct in_addr dst; 1456 struct route *rt; 1457{ 1458 register struct sockaddr_in *sin; 1459 1460 sin = (struct sockaddr_in *)&rt->ro_dst; 1461 1462 if (rt->ro_rt == 0 || 1463 !(rt->ro_rt->rt_flags & RTF_UP) || 1464 dst.s_addr != sin->sin_addr.s_addr) { 1465 if (rt->ro_rt) { 1466 RTFREE(rt->ro_rt); 1467 rt->ro_rt = 0; 1468 } 1469 sin->sin_family = AF_INET; 1470 sin->sin_len = sizeof(*sin); 1471 sin->sin_addr = dst; 1472 1473 rtalloc_ign(rt, RTF_PRCLONING); 1474 } 1475 if (rt->ro_rt == 0) 1476 return ((struct in_ifaddr *)0); 1477 return (ifatoia(rt->ro_rt->rt_ifa)); 1478} 1479 1480/* 1481 * Save incoming source route for use in replies, 1482 * to be picked up later by ip_srcroute if the receiver is interested. 1483 */ 1484void 1485save_rte(option, dst) 1486 u_char *option; 1487 struct in_addr dst; 1488{ 1489 unsigned olen; 1490 1491 olen = option[IPOPT_OLEN]; 1492#ifdef DIAGNOSTIC 1493 if (ipprintfs) 1494 printf("save_rte: olen %d\n", olen); 1495#endif 1496 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1497 return; 1498 bcopy(option, ip_srcrt.srcopt, olen); 1499 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1500 ip_srcrt.dst = dst; 1501} 1502 1503/* 1504 * Retrieve incoming source route for use in replies, 1505 * in the same form used by setsockopt. 1506 * The first hop is placed before the options, will be removed later. 1507 */ 1508struct mbuf * 1509ip_srcroute() 1510{ 1511 register struct in_addr *p, *q; 1512 register struct mbuf *m; 1513 1514 if (ip_nhops == 0) 1515 return ((struct mbuf *)0); 1516 m = m_get(M_DONTWAIT, MT_HEADER); 1517 if (m == 0) 1518 return ((struct mbuf *)0); 1519 1520#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1521 1522 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1523 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1524 OPTSIZ; 1525#ifdef DIAGNOSTIC 1526 if (ipprintfs) 1527 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1528#endif 1529 1530 /* 1531 * First save first hop for return route 1532 */ 1533 p = &ip_srcrt.route[ip_nhops - 1]; 1534 *(mtod(m, struct in_addr *)) = *p--; 1535#ifdef DIAGNOSTIC 1536 if (ipprintfs) 1537 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1538#endif 1539 1540 /* 1541 * Copy option fields and padding (nop) to mbuf. 1542 */ 1543 ip_srcrt.nop = IPOPT_NOP; 1544 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1545 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1546 &ip_srcrt.nop, OPTSIZ); 1547 q = (struct in_addr *)(mtod(m, caddr_t) + 1548 sizeof(struct in_addr) + OPTSIZ); 1549#undef OPTSIZ 1550 /* 1551 * Record return path as an IP source route, 1552 * reversing the path (pointers are now aligned). 1553 */ 1554 while (p >= ip_srcrt.route) { 1555#ifdef DIAGNOSTIC 1556 if (ipprintfs) 1557 printf(" %lx", (u_long)ntohl(q->s_addr)); 1558#endif 1559 *q++ = *p--; 1560 } 1561 /* 1562 * Last hop goes to final destination. 1563 */ 1564 *q = ip_srcrt.dst; 1565#ifdef DIAGNOSTIC 1566 if (ipprintfs) 1567 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1568#endif 1569 return (m); 1570} 1571 1572/* 1573 * Strip out IP options, at higher 1574 * level protocol in the kernel. 1575 * Second argument is buffer to which options 1576 * will be moved, and return value is their length. 1577 * XXX should be deleted; last arg currently ignored. 1578 */ 1579void 1580ip_stripoptions(m, mopt) 1581 register struct mbuf *m; 1582 struct mbuf *mopt; 1583{ 1584 register int i; 1585 struct ip *ip = mtod(m, struct ip *); 1586 register caddr_t opts; 1587 int olen; 1588 1589 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1590 opts = (caddr_t)(ip + 1); 1591 i = m->m_len - (sizeof (struct ip) + olen); 1592 bcopy(opts + olen, opts, (unsigned)i); 1593 m->m_len -= olen; 1594 if (m->m_flags & M_PKTHDR) 1595 m->m_pkthdr.len -= olen; 1596 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1597} 1598 1599u_char inetctlerrmap[PRC_NCMDS] = { 1600 0, 0, 0, 0, 1601 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1602 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1603 EMSGSIZE, EHOSTUNREACH, 0, 0, 1604 0, 0, 0, 0, 1605 ENOPROTOOPT, ECONNREFUSED 1606}; 1607 1608/* 1609 * Forward a packet. If some error occurs return the sender 1610 * an icmp packet. Note we can't always generate a meaningful 1611 * icmp message because icmp doesn't have a large enough repertoire 1612 * of codes and types. 1613 * 1614 * If not forwarding, just drop the packet. This could be confusing 1615 * if ipforwarding was zero but some routing protocol was advancing 1616 * us as a gateway to somewhere. However, we must let the routing 1617 * protocol deal with that. 1618 * 1619 * The srcrt parameter indicates whether the packet is being forwarded 1620 * via a source route. 1621 */ 1622static void 1623ip_forward(m, srcrt) 1624 struct mbuf *m; 1625 int srcrt; 1626{ 1627 register struct ip *ip = mtod(m, struct ip *); 1628 register struct rtentry *rt; 1629 int error, type = 0, code = 0; 1630 struct mbuf *mcopy; 1631 n_long dest; 1632 struct in_addr pkt_dst; 1633 struct ifnet *destifp; 1634#ifdef IPSEC 1635 struct ifnet dummyifp; 1636#endif 1637 1638 dest = 0; 1639 /* 1640 * Cache the destination address of the packet; this may be 1641 * changed by use of 'ipfw fwd'. 1642 */ 1643 pkt_dst = ip_fw_fwd_addr == NULL ? 1644 ip->ip_dst : ip_fw_fwd_addr->sin_addr; 1645 1646#ifdef DIAGNOSTIC 1647 if (ipprintfs) 1648 printf("forward: src %lx dst %lx ttl %x\n", 1649 (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr, 1650 ip->ip_ttl); 1651#endif 1652 1653 1654 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) { 1655 ipstat.ips_cantforward++; 1656 m_freem(m); 1657 return; 1658 } 1659#ifdef IPSTEALTH 1660 if (!ipstealth) { 1661#endif 1662 if (ip->ip_ttl <= IPTTLDEC) { 1663 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1664 dest, 0); 1665 return; 1666 } 1667#ifdef IPSTEALTH 1668 } 1669#endif 1670 1671 if (ip_rtaddr(pkt_dst, &ipforward_rt) == 0) { 1672 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1673 return; 1674 } else 1675 rt = ipforward_rt.ro_rt; 1676 1677 /* 1678 * Save the IP header and at most 8 bytes of the payload, 1679 * in case we need to generate an ICMP message to the src. 1680 * 1681 * We don't use m_copy() because it might return a reference 1682 * to a shared cluster. Both this function and ip_output() 1683 * assume exclusive access to the IP header in `m', so any 1684 * data in a cluster may change before we reach icmp_error(). 1685 */ 1686 MGET(mcopy, M_DONTWAIT, m->m_type); 1687 if (mcopy != NULL) { 1688 M_COPY_PKTHDR(mcopy, m); 1689 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8, 1690 (int)ip->ip_len); 1691 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1692 } 1693 1694#ifdef IPSTEALTH 1695 if (!ipstealth) { 1696#endif 1697 ip->ip_ttl -= IPTTLDEC; 1698#ifdef IPSTEALTH 1699 } 1700#endif 1701 1702 /* 1703 * If forwarding packet using same interface that it came in on, 1704 * perhaps should send a redirect to sender to shortcut a hop. 1705 * Only send redirect if source is sending directly to us, 1706 * and if packet was not source routed (or has any options). 1707 * Also, don't send redirect if forwarding using a default route 1708 * or a route modified by a redirect. 1709 */ 1710 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1711 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1712 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1713 ipsendredirects && !srcrt && !ip_fw_fwd_addr) { 1714#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1715 u_long src = ntohl(ip->ip_src.s_addr); 1716 1717 if (RTA(rt) && 1718 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1719 if (rt->rt_flags & RTF_GATEWAY) 1720 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1721 else 1722 dest = pkt_dst.s_addr; 1723 /* Router requirements says to only send host redirects */ 1724 type = ICMP_REDIRECT; 1725 code = ICMP_REDIRECT_HOST; 1726#ifdef DIAGNOSTIC 1727 if (ipprintfs) 1728 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1729#endif 1730 } 1731 } 1732 1733 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1734 IP_FORWARDING, 0); 1735 if (error) 1736 ipstat.ips_cantforward++; 1737 else { 1738 ipstat.ips_forward++; 1739 if (type) 1740 ipstat.ips_redirectsent++; 1741 else { 1742 if (mcopy) { 1743 ipflow_create(&ipforward_rt, mcopy); 1744 m_freem(mcopy); 1745 } 1746 return; 1747 } 1748 } 1749 if (mcopy == NULL) 1750 return; 1751 destifp = NULL; 1752 1753 switch (error) { 1754 1755 case 0: /* forwarded, but need redirect */ 1756 /* type, code set above */ 1757 break; 1758 1759 case ENETUNREACH: /* shouldn't happen, checked above */ 1760 case EHOSTUNREACH: 1761 case ENETDOWN: 1762 case EHOSTDOWN: 1763 default: 1764 type = ICMP_UNREACH; 1765 code = ICMP_UNREACH_HOST; 1766 break; 1767 1768 case EMSGSIZE: 1769 type = ICMP_UNREACH; 1770 code = ICMP_UNREACH_NEEDFRAG; 1771#ifndef IPSEC 1772 if (ipforward_rt.ro_rt) 1773 destifp = ipforward_rt.ro_rt->rt_ifp; 1774#else 1775 /* 1776 * If the packet is routed over IPsec tunnel, tell the 1777 * originator the tunnel MTU. 1778 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1779 * XXX quickhack!!! 1780 */ 1781 if (ipforward_rt.ro_rt) { 1782 struct secpolicy *sp = NULL; 1783 int ipsecerror; 1784 int ipsechdr; 1785 struct route *ro; 1786 1787 sp = ipsec4_getpolicybyaddr(mcopy, 1788 IPSEC_DIR_OUTBOUND, 1789 IP_FORWARDING, 1790 &ipsecerror); 1791 1792 if (sp == NULL) 1793 destifp = ipforward_rt.ro_rt->rt_ifp; 1794 else { 1795 /* count IPsec header size */ 1796 ipsechdr = ipsec4_hdrsiz(mcopy, 1797 IPSEC_DIR_OUTBOUND, 1798 NULL); 1799 1800 /* 1801 * find the correct route for outer IPv4 1802 * header, compute tunnel MTU. 1803 * 1804 * XXX BUG ALERT 1805 * The "dummyifp" code relies upon the fact 1806 * that icmp_error() touches only ifp->if_mtu. 1807 */ 1808 /*XXX*/ 1809 destifp = NULL; 1810 if (sp->req != NULL 1811 && sp->req->sav != NULL 1812 && sp->req->sav->sah != NULL) { 1813 ro = &sp->req->sav->sah->sa_route; 1814 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1815 dummyifp.if_mtu = 1816 ro->ro_rt->rt_ifp->if_mtu; 1817 dummyifp.if_mtu -= ipsechdr; 1818 destifp = &dummyifp; 1819 } 1820 } 1821 1822 key_freesp(sp); 1823 } 1824 } 1825#endif /*IPSEC*/ 1826 ipstat.ips_cantfrag++; 1827 break; 1828 1829 case ENOBUFS: 1830 type = ICMP_SOURCEQUENCH; 1831 code = 0; 1832 break; 1833 1834 case EACCES: /* ipfw denied packet */ 1835 m_freem(mcopy); 1836 return; 1837 } 1838 icmp_error(mcopy, type, code, dest, destifp); 1839} 1840 1841void 1842ip_savecontrol(inp, mp, ip, m) 1843 register struct inpcb *inp; 1844 register struct mbuf **mp; 1845 register struct ip *ip; 1846 register struct mbuf *m; 1847{ 1848 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1849 struct timeval tv; 1850 1851 microtime(&tv); 1852 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1853 SCM_TIMESTAMP, SOL_SOCKET); 1854 if (*mp) 1855 mp = &(*mp)->m_next; 1856 } 1857 if (inp->inp_flags & INP_RECVDSTADDR) { 1858 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1859 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1860 if (*mp) 1861 mp = &(*mp)->m_next; 1862 } 1863#ifdef notyet 1864 /* XXX 1865 * Moving these out of udp_input() made them even more broken 1866 * than they already were. 1867 */ 1868 /* options were tossed already */ 1869 if (inp->inp_flags & INP_RECVOPTS) { 1870 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1871 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1872 if (*mp) 1873 mp = &(*mp)->m_next; 1874 } 1875 /* ip_srcroute doesn't do what we want here, need to fix */ 1876 if (inp->inp_flags & INP_RECVRETOPTS) { 1877 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1878 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1879 if (*mp) 1880 mp = &(*mp)->m_next; 1881 } 1882#endif 1883 if (inp->inp_flags & INP_RECVIF) { 1884 struct ifnet *ifp; 1885 struct sdlbuf { 1886 struct sockaddr_dl sdl; 1887 u_char pad[32]; 1888 } sdlbuf; 1889 struct sockaddr_dl *sdp; 1890 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1891 1892 if (((ifp = m->m_pkthdr.rcvif)) 1893 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1894 sdp = (struct sockaddr_dl *) 1895 (ifaddr_byindex(ifp->if_index)->ifa_addr); 1896 /* 1897 * Change our mind and don't try copy. 1898 */ 1899 if ((sdp->sdl_family != AF_LINK) 1900 || (sdp->sdl_len > sizeof(sdlbuf))) { 1901 goto makedummy; 1902 } 1903 bcopy(sdp, sdl2, sdp->sdl_len); 1904 } else { 1905makedummy: 1906 sdl2->sdl_len 1907 = offsetof(struct sockaddr_dl, sdl_data[0]); 1908 sdl2->sdl_family = AF_LINK; 1909 sdl2->sdl_index = 0; 1910 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1911 } 1912 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1913 IP_RECVIF, IPPROTO_IP); 1914 if (*mp) 1915 mp = &(*mp)->m_next; 1916 } 1917} 1918 1919int 1920ip_rsvp_init(struct socket *so) 1921{ 1922 if (so->so_type != SOCK_RAW || 1923 so->so_proto->pr_protocol != IPPROTO_RSVP) 1924 return EOPNOTSUPP; 1925 1926 if (ip_rsvpd != NULL) 1927 return EADDRINUSE; 1928 1929 ip_rsvpd = so; 1930 /* 1931 * This may seem silly, but we need to be sure we don't over-increment 1932 * the RSVP counter, in case something slips up. 1933 */ 1934 if (!ip_rsvp_on) { 1935 ip_rsvp_on = 1; 1936 rsvp_on++; 1937 } 1938 1939 return 0; 1940} 1941 1942int 1943ip_rsvp_done(void) 1944{ 1945 ip_rsvpd = NULL; 1946 /* 1947 * This may seem silly, but we need to be sure we don't over-decrement 1948 * the RSVP counter, in case something slips up. 1949 */ 1950 if (ip_rsvp_on) { 1951 ip_rsvp_on = 0; 1952 rsvp_on--; 1953 } 1954 return 0; 1955} 1956