ip_reass.c revision 131814
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 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 30 * $FreeBSD: head/sys/netinet/ip_input.c 131814 2004-07-08 13:40:33Z brian $ 31 */ 32 33#include "opt_ipfw.h" 34#include "opt_ipdn.h" 35#include "opt_ipdivert.h" 36#include "opt_ipfilter.h" 37#include "opt_ipstealth.h" 38#include "opt_ipsec.h" 39#include "opt_mac.h" 40#include "opt_pfil_hooks.h" 41#include "opt_random_ip_id.h" 42 43#include <sys/param.h> 44#include <sys/systm.h> 45#include <sys/mac.h> 46#include <sys/mbuf.h> 47#include <sys/malloc.h> 48#include <sys/domain.h> 49#include <sys/protosw.h> 50#include <sys/socket.h> 51#include <sys/time.h> 52#include <sys/kernel.h> 53#include <sys/syslog.h> 54#include <sys/sysctl.h> 55 56#include <net/pfil.h> 57#include <net/if.h> 58#include <net/if_types.h> 59#include <net/if_var.h> 60#include <net/if_dl.h> 61#include <net/route.h> 62#include <net/netisr.h> 63 64#include <netinet/in.h> 65#include <netinet/in_systm.h> 66#include <netinet/in_var.h> 67#include <netinet/ip.h> 68#include <netinet/in_pcb.h> 69#include <netinet/ip_var.h> 70#include <netinet/ip_icmp.h> 71#include <machine/in_cksum.h> 72 73#include <sys/socketvar.h> 74 75#include <netinet/ip_fw.h> 76#include <netinet/ip_divert.h> 77#include <netinet/ip_dummynet.h> 78 79#ifdef IPSEC 80#include <netinet6/ipsec.h> 81#include <netkey/key.h> 82#endif 83 84#ifdef FAST_IPSEC 85#include <netipsec/ipsec.h> 86#include <netipsec/key.h> 87#endif 88 89int rsvp_on = 0; 90 91int ipforwarding = 0; 92SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 93 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 94 95static int ipsendredirects = 1; /* XXX */ 96SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 97 &ipsendredirects, 0, "Enable sending IP redirects"); 98 99int ip_defttl = IPDEFTTL; 100SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 101 &ip_defttl, 0, "Maximum TTL on IP packets"); 102 103static int ip_dosourceroute = 0; 104SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 105 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 106 107static int ip_acceptsourceroute = 0; 108SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 109 CTLFLAG_RW, &ip_acceptsourceroute, 0, 110 "Enable accepting source routed IP packets"); 111 112int ip_doopts = 1; /* 0 = ignore, 1 = process, 2 = reject */ 113SYSCTL_INT(_net_inet_ip, OID_AUTO, process_options, CTLFLAG_RW, 114 &ip_doopts, 0, "Enable IP options processing ([LS]SRR, RR, TS)"); 115 116static int ip_keepfaith = 0; 117SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 118 &ip_keepfaith, 0, 119 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 120 121static int nipq = 0; /* total # of reass queues */ 122static int maxnipq; 123SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 124 &maxnipq, 0, 125 "Maximum number of IPv4 fragment reassembly queue entries"); 126 127static int maxfragsperpacket; 128SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 129 &maxfragsperpacket, 0, 130 "Maximum number of IPv4 fragments allowed per packet"); 131 132static int ip_sendsourcequench = 0; 133SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 134 &ip_sendsourcequench, 0, 135 "Enable the transmission of source quench packets"); 136 137/* 138 * XXX - Setting ip_checkinterface mostly implements the receive side of 139 * the Strong ES model described in RFC 1122, but since the routing table 140 * and transmit implementation do not implement the Strong ES model, 141 * setting this to 1 results in an odd hybrid. 142 * 143 * XXX - ip_checkinterface currently must be disabled if you use ipnat 144 * to translate the destination address to another local interface. 145 * 146 * XXX - ip_checkinterface must be disabled if you add IP aliases 147 * to the loopback interface instead of the interface where the 148 * packets for those addresses are received. 149 */ 150static int ip_checkinterface = 1; 151SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 152 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 153 154#ifdef DIAGNOSTIC 155static int ipprintfs = 0; 156#endif 157#ifdef PFIL_HOOKS 158struct pfil_head inet_pfil_hook; 159#endif 160 161static struct ifqueue ipintrq; 162static int ipqmaxlen = IFQ_MAXLEN; 163 164extern struct domain inetdomain; 165extern struct protosw inetsw[]; 166u_char ip_protox[IPPROTO_MAX]; 167struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 168struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 169u_long in_ifaddrhmask; /* mask for hash table */ 170 171SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 172 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 173SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 174 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 175 176struct ipstat ipstat; 177SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 178 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 179 180/* Packet reassembly stuff */ 181#define IPREASS_NHASH_LOG2 6 182#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 183#define IPREASS_HMASK (IPREASS_NHASH - 1) 184#define IPREASS_HASH(x,y) \ 185 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 186 187static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 188struct mtx ipqlock; 189 190#define IPQ_LOCK() mtx_lock(&ipqlock) 191#define IPQ_UNLOCK() mtx_unlock(&ipqlock) 192#define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 193#define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 194 195#ifdef IPCTL_DEFMTU 196SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 197 &ip_mtu, 0, "Default MTU"); 198#endif 199 200#ifdef IPSTEALTH 201int ipstealth = 0; 202SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 203 &ipstealth, 0, ""); 204#endif 205 206 207/* Firewall hooks */ 208ip_fw_chk_t *ip_fw_chk_ptr; 209int fw_enable = 1 ; 210int fw_one_pass = 1; 211 212/* Dummynet hooks */ 213ip_dn_io_t *ip_dn_io_ptr; 214 215/* 216 * XXX this is ugly -- the following two global variables are 217 * used to store packet state while it travels through the stack. 218 * Note that the code even makes assumptions on the size and 219 * alignment of fields inside struct ip_srcrt so e.g. adding some 220 * fields will break the code. This needs to be fixed. 221 * 222 * We need to save the IP options in case a protocol wants to respond 223 * to an incoming packet over the same route if the packet got here 224 * using IP source routing. This allows connection establishment and 225 * maintenance when the remote end is on a network that is not known 226 * to us. 227 */ 228static int ip_nhops = 0; 229static struct ip_srcrt { 230 struct in_addr dst; /* final destination */ 231 char nop; /* one NOP to align */ 232 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 233 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 234} ip_srcrt; 235 236static void save_rte(u_char *, struct in_addr); 237static int ip_dooptions(struct mbuf *m, int, 238 struct sockaddr_in *next_hop); 239static void ip_forward(struct mbuf *m, int srcrt, 240 struct sockaddr_in *next_hop); 241static void ip_freef(struct ipqhead *, struct ipq *); 242static struct mbuf *ip_reass(struct mbuf *, struct ipqhead *, struct ipq *); 243 244/* 245 * IP initialization: fill in IP protocol switch table. 246 * All protocols not implemented in kernel go to raw IP protocol handler. 247 */ 248void 249ip_init() 250{ 251 register struct protosw *pr; 252 register int i; 253 254 TAILQ_INIT(&in_ifaddrhead); 255 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 256 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 257 if (pr == 0) 258 panic("ip_init"); 259 for (i = 0; i < IPPROTO_MAX; i++) 260 ip_protox[i] = pr - inetsw; 261 for (pr = inetdomain.dom_protosw; 262 pr < inetdomain.dom_protoswNPROTOSW; pr++) 263 if (pr->pr_domain->dom_family == PF_INET && 264 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 265 ip_protox[pr->pr_protocol] = pr - inetsw; 266 267#ifdef PFIL_HOOKS 268 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 269 inet_pfil_hook.ph_af = AF_INET; 270 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 271 printf("%s: WARNING: unable to register pfil hook, " 272 "error %d\n", __func__, i); 273#endif /* PFIL_HOOKS */ 274 275 IPQ_LOCK_INIT(); 276 for (i = 0; i < IPREASS_NHASH; i++) 277 TAILQ_INIT(&ipq[i]); 278 279 maxnipq = nmbclusters / 32; 280 maxfragsperpacket = 16; 281 282#ifndef RANDOM_IP_ID 283 ip_id = time_second & 0xffff; 284#endif 285 ipintrq.ifq_maxlen = ipqmaxlen; 286 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF); 287 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE); 288} 289 290/* 291 * Ip input routine. Checksum and byte swap header. If fragmented 292 * try to reassemble. Process options. Pass to next level. 293 */ 294void 295ip_input(struct mbuf *m) 296{ 297 struct ip *ip = NULL; 298 struct ipq *fp; 299 struct in_ifaddr *ia = NULL; 300 struct ifaddr *ifa; 301 char *cp; 302 int i, checkif, hlen = 0; 303 u_short sum; 304 struct in_addr pkt_dst; 305#ifdef IPDIVERT 306 u_int32_t divert_info; /* packet divert/tee info */ 307#endif 308 struct ip_fw_args args; 309 int dchg = 0; /* dest changed after fw */ 310#ifdef PFIL_HOOKS 311 struct in_addr odst; /* original dst address */ 312#endif 313#ifdef FAST_IPSEC 314 struct m_tag *mtag; 315 struct tdb_ident *tdbi; 316 struct secpolicy *sp; 317 int s, error; 318#endif /* FAST_IPSEC */ 319 320 args.eh = NULL; 321 args.oif = NULL; 322 323 M_ASSERTPKTHDR(m); 324 325 args.next_hop = m_claim_next(m, PACKET_TAG_IPFORWARD); 326 args.rule = ip_dn_claim_rule(m); 327 328 if (m->m_flags & M_FASTFWD_OURS) { 329 /* ip_fastforward firewall changed dest to local */ 330 m->m_flags &= ~M_FASTFWD_OURS; /* for reflected mbufs */ 331 goto ours; 332 } 333 334 if (args.rule) { /* dummynet already filtered us */ 335 ip = mtod(m, struct ip *); 336 hlen = ip->ip_hl << 2; 337 goto iphack ; 338 } 339 340 ipstat.ips_total++; 341 342 if (m->m_pkthdr.len < sizeof(struct ip)) 343 goto tooshort; 344 345 if (m->m_len < sizeof (struct ip) && 346 (m = m_pullup(m, sizeof (struct ip))) == 0) { 347 ipstat.ips_toosmall++; 348 return; 349 } 350 ip = mtod(m, struct ip *); 351 352 if (ip->ip_v != IPVERSION) { 353 ipstat.ips_badvers++; 354 goto bad; 355 } 356 357 hlen = ip->ip_hl << 2; 358 if (hlen < sizeof(struct ip)) { /* minimum header length */ 359 ipstat.ips_badhlen++; 360 goto bad; 361 } 362 if (hlen > m->m_len) { 363 if ((m = m_pullup(m, hlen)) == 0) { 364 ipstat.ips_badhlen++; 365 return; 366 } 367 ip = mtod(m, struct ip *); 368 } 369 370 /* 127/8 must not appear on wire - RFC1122 */ 371 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 372 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 373 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 374 ipstat.ips_badaddr++; 375 goto bad; 376 } 377 } 378 379 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 380 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 381 } else { 382 if (hlen == sizeof(struct ip)) { 383 sum = in_cksum_hdr(ip); 384 } else { 385 sum = in_cksum(m, hlen); 386 } 387 } 388 if (sum) { 389 ipstat.ips_badsum++; 390 goto bad; 391 } 392 393#ifdef ALTQ 394 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 395 /* packet is dropped by traffic conditioner */ 396 return; 397#endif 398 399 /* 400 * Convert fields to host representation. 401 */ 402 ip->ip_len = ntohs(ip->ip_len); 403 if (ip->ip_len < hlen) { 404 ipstat.ips_badlen++; 405 goto bad; 406 } 407 ip->ip_off = ntohs(ip->ip_off); 408 409 /* 410 * Check that the amount of data in the buffers 411 * is as at least much as the IP header would have us expect. 412 * Trim mbufs if longer than we expect. 413 * Drop packet if shorter than we expect. 414 */ 415 if (m->m_pkthdr.len < ip->ip_len) { 416tooshort: 417 ipstat.ips_tooshort++; 418 goto bad; 419 } 420 if (m->m_pkthdr.len > ip->ip_len) { 421 if (m->m_len == m->m_pkthdr.len) { 422 m->m_len = ip->ip_len; 423 m->m_pkthdr.len = ip->ip_len; 424 } else 425 m_adj(m, ip->ip_len - m->m_pkthdr.len); 426 } 427#if defined(IPSEC) && !defined(IPSEC_FILTERGIF) 428 /* 429 * Bypass packet filtering for packets from a tunnel (gif). 430 */ 431 if (ipsec_getnhist(m)) 432 goto pass; 433#endif 434#if defined(FAST_IPSEC) && !defined(IPSEC_FILTERGIF) 435 /* 436 * Bypass packet filtering for packets from a tunnel (gif). 437 */ 438 if (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL) 439 goto pass; 440#endif 441 442 /* 443 * IpHack's section. 444 * Right now when no processing on packet has done 445 * and it is still fresh out of network we do our black 446 * deals with it. 447 * - Firewall: deny/allow/divert 448 * - Xlate: translate packet's addr/port (NAT). 449 * - Pipe: pass pkt through dummynet. 450 * - Wrap: fake packet's addr/port <unimpl.> 451 * - Encapsulate: put it in another IP and send out. <unimp.> 452 */ 453 454iphack: 455 456#ifdef PFIL_HOOKS 457 /* 458 * Run through list of hooks for input packets. 459 * 460 * NB: Beware of the destination address changing (e.g. 461 * by NAT rewriting). When this happens, tell 462 * ip_forward to do the right thing. 463 */ 464 odst = ip->ip_dst; 465 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 466 PFIL_IN) != 0) 467 return; 468 if (m == NULL) /* consumed by filter */ 469 return; 470 ip = mtod(m, struct ip *); 471 dchg = (odst.s_addr != ip->ip_dst.s_addr); 472#endif /* PFIL_HOOKS */ 473 474 if (fw_enable && IPFW_LOADED) { 475 /* 476 * If we've been forwarded from the output side, then 477 * skip the firewall a second time 478 */ 479 if (args.next_hop) 480 goto ours; 481 482 args.m = m; 483 i = ip_fw_chk_ptr(&args); 484 m = args.m; 485 486 if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */ 487 if (m) 488 m_freem(m); 489 return; 490 } 491 ip = mtod(m, struct ip *); /* just in case m changed */ 492 if (i == 0 && args.next_hop == NULL) /* common case */ 493 goto pass; 494 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) { 495 /* Send packet to the appropriate pipe */ 496 ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args); 497 return; 498 } 499#ifdef IPDIVERT 500 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 501 /* Divert or tee packet */ 502 goto ours; 503 } 504#endif 505 if (i == 0 && args.next_hop != NULL) 506 goto pass; 507 /* 508 * if we get here, the packet must be dropped 509 */ 510 m_freem(m); 511 return; 512 } 513pass: 514 515 /* 516 * Process options and, if not destined for us, 517 * ship it on. ip_dooptions returns 1 when an 518 * error was detected (causing an icmp message 519 * to be sent and the original packet to be freed). 520 */ 521 ip_nhops = 0; /* for source routed packets */ 522 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.next_hop)) 523 return; 524 525 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 526 * matter if it is destined to another node, or whether it is 527 * a multicast one, RSVP wants it! and prevents it from being forwarded 528 * anywhere else. Also checks if the rsvp daemon is running before 529 * grabbing the packet. 530 */ 531 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 532 goto ours; 533 534 /* 535 * Check our list of addresses, to see if the packet is for us. 536 * If we don't have any addresses, assume any unicast packet 537 * we receive might be for us (and let the upper layers deal 538 * with it). 539 */ 540 if (TAILQ_EMPTY(&in_ifaddrhead) && 541 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 542 goto ours; 543 544 /* 545 * Cache the destination address of the packet; this may be 546 * changed by use of 'ipfw fwd'. 547 */ 548 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst; 549 550 /* 551 * Enable a consistency check between the destination address 552 * and the arrival interface for a unicast packet (the RFC 1122 553 * strong ES model) if IP forwarding is disabled and the packet 554 * is not locally generated and the packet is not subject to 555 * 'ipfw fwd'. 556 * 557 * XXX - Checking also should be disabled if the destination 558 * address is ipnat'ed to a different interface. 559 * 560 * XXX - Checking is incompatible with IP aliases added 561 * to the loopback interface instead of the interface where 562 * the packets are received. 563 */ 564 checkif = ip_checkinterface && (ipforwarding == 0) && 565 m->m_pkthdr.rcvif != NULL && 566 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 567 (args.next_hop == NULL) && (dchg == 0); 568 569 /* 570 * Check for exact addresses in the hash bucket. 571 */ 572 LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) { 573 /* 574 * If the address matches, verify that the packet 575 * arrived via the correct interface if checking is 576 * enabled. 577 */ 578 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 579 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 580 goto ours; 581 } 582 /* 583 * Check for broadcast addresses. 584 * 585 * Only accept broadcast packets that arrive via the matching 586 * interface. Reception of forwarded directed broadcasts would 587 * be handled via ip_forward() and ether_output() with the loopback 588 * into the stack for SIMPLEX interfaces handled by ether_output(). 589 */ 590 if (m->m_pkthdr.rcvif != NULL && 591 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 592 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 593 if (ifa->ifa_addr->sa_family != AF_INET) 594 continue; 595 ia = ifatoia(ifa); 596 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 597 pkt_dst.s_addr) 598 goto ours; 599 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 600 goto ours; 601 if ((cp = getenv("bootp.compat")) != NULL) { 602 freeenv(cp); 603 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 604 goto ours; 605 } 606 } 607 } 608 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 609 struct in_multi *inm; 610 if (ip_mrouter) { 611 /* 612 * If we are acting as a multicast router, all 613 * incoming multicast packets are passed to the 614 * kernel-level multicast forwarding function. 615 * The packet is returned (relatively) intact; if 616 * ip_mforward() returns a non-zero value, the packet 617 * must be discarded, else it may be accepted below. 618 */ 619 if (ip_mforward && 620 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 621 ipstat.ips_cantforward++; 622 m_freem(m); 623 return; 624 } 625 626 /* 627 * The process-level routing daemon needs to receive 628 * all multicast IGMP packets, whether or not this 629 * host belongs to their destination groups. 630 */ 631 if (ip->ip_p == IPPROTO_IGMP) 632 goto ours; 633 ipstat.ips_forward++; 634 } 635 /* 636 * See if we belong to the destination multicast group on the 637 * arrival interface. 638 */ 639 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 640 if (inm == NULL) { 641 ipstat.ips_notmember++; 642 m_freem(m); 643 return; 644 } 645 goto ours; 646 } 647 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 648 goto ours; 649 if (ip->ip_dst.s_addr == INADDR_ANY) 650 goto ours; 651 652 /* 653 * FAITH(Firewall Aided Internet Translator) 654 */ 655 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 656 if (ip_keepfaith) { 657 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 658 goto ours; 659 } 660 m_freem(m); 661 return; 662 } 663 664 /* 665 * Not for us; forward if possible and desirable. 666 */ 667 if (ipforwarding == 0) { 668 ipstat.ips_cantforward++; 669 m_freem(m); 670 } else { 671#ifdef IPSEC 672 /* 673 * Enforce inbound IPsec SPD. 674 */ 675 if (ipsec4_in_reject(m, NULL)) { 676 ipsecstat.in_polvio++; 677 goto bad; 678 } 679#endif /* IPSEC */ 680#ifdef FAST_IPSEC 681 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 682 s = splnet(); 683 if (mtag != NULL) { 684 tdbi = (struct tdb_ident *)(mtag + 1); 685 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 686 } else { 687 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 688 IP_FORWARDING, &error); 689 } 690 if (sp == NULL) { /* NB: can happen if error */ 691 splx(s); 692 /*XXX error stat???*/ 693 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 694 goto bad; 695 } 696 697 /* 698 * Check security policy against packet attributes. 699 */ 700 error = ipsec_in_reject(sp, m); 701 KEY_FREESP(&sp); 702 splx(s); 703 if (error) { 704 ipstat.ips_cantforward++; 705 goto bad; 706 } 707#endif /* FAST_IPSEC */ 708 ip_forward(m, dchg, args.next_hop); 709 } 710 return; 711 712ours: 713#ifdef IPSTEALTH 714 /* 715 * IPSTEALTH: Process non-routing options only 716 * if the packet is destined for us. 717 */ 718 if (ipstealth && hlen > sizeof (struct ip) && 719 ip_dooptions(m, 1, args.next_hop)) 720 return; 721#endif /* IPSTEALTH */ 722 723 /* Count the packet in the ip address stats */ 724 if (ia != NULL) { 725 ia->ia_ifa.if_ipackets++; 726 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 727 } 728 729 /* 730 * If offset or IP_MF are set, must reassemble. 731 * Otherwise, nothing need be done. 732 * (We could look in the reassembly queue to see 733 * if the packet was previously fragmented, 734 * but it's not worth the time; just let them time out.) 735 */ 736 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 737 738 /* If maxnipq is 0, never accept fragments. */ 739 if (maxnipq == 0) { 740 ipstat.ips_fragments++; 741 ipstat.ips_fragdropped++; 742 goto bad; 743 } 744 745 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 746 IPQ_LOCK(); 747 /* 748 * Look for queue of fragments 749 * of this datagram. 750 */ 751 TAILQ_FOREACH(fp, &ipq[sum], ipq_list) 752 if (ip->ip_id == fp->ipq_id && 753 ip->ip_src.s_addr == fp->ipq_src.s_addr && 754 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 755#ifdef MAC 756 mac_fragment_match(m, fp) && 757#endif 758 ip->ip_p == fp->ipq_p) 759 goto found; 760 761 fp = NULL; 762 763 /* 764 * Enforce upper bound on number of fragmented packets 765 * for which we attempt reassembly; 766 * If maxnipq is -1, accept all fragments without limitation. 767 */ 768 if ((nipq > maxnipq) && (maxnipq > 0)) { 769 /* 770 * drop something from the tail of the current queue 771 * before proceeding further 772 */ 773 struct ipq *q = TAILQ_LAST(&ipq[sum], ipqhead); 774 if (q == NULL) { /* gak */ 775 for (i = 0; i < IPREASS_NHASH; i++) { 776 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 777 if (r) { 778 ipstat.ips_fragtimeout += r->ipq_nfrags; 779 ip_freef(&ipq[i], r); 780 break; 781 } 782 } 783 } else { 784 ipstat.ips_fragtimeout += q->ipq_nfrags; 785 ip_freef(&ipq[sum], q); 786 } 787 } 788found: 789 /* 790 * Adjust ip_len to not reflect header, 791 * convert offset of this to bytes. 792 */ 793 ip->ip_len -= hlen; 794 if (ip->ip_off & IP_MF) { 795 /* 796 * Make sure that fragments have a data length 797 * that's a non-zero multiple of 8 bytes. 798 */ 799 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 800 IPQ_UNLOCK(); 801 ipstat.ips_toosmall++; /* XXX */ 802 goto bad; 803 } 804 m->m_flags |= M_FRAG; 805 } else 806 m->m_flags &= ~M_FRAG; 807 ip->ip_off <<= 3; 808 809 /* 810 * Attempt reassembly; if it succeeds, proceed. 811 * ip_reass() will return a different mbuf. 812 */ 813 ipstat.ips_fragments++; 814 m->m_pkthdr.header = ip; 815 m = ip_reass(m, &ipq[sum], fp); 816 IPQ_UNLOCK(); 817 if (m == 0) 818 return; 819 ipstat.ips_reassembled++; 820 ip = mtod(m, struct ip *); 821 /* Get the header length of the reassembled packet */ 822 hlen = ip->ip_hl << 2; 823#ifdef IPDIVERT 824 /* Restore original checksum before diverting packet */ 825 if (divert_find_info(m) != 0) { 826 ip->ip_len += hlen; 827 ip->ip_len = htons(ip->ip_len); 828 ip->ip_off = htons(ip->ip_off); 829 ip->ip_sum = 0; 830 if (hlen == sizeof(struct ip)) 831 ip->ip_sum = in_cksum_hdr(ip); 832 else 833 ip->ip_sum = in_cksum(m, hlen); 834 ip->ip_off = ntohs(ip->ip_off); 835 ip->ip_len = ntohs(ip->ip_len); 836 ip->ip_len -= hlen; 837 } 838#endif 839 } else 840 ip->ip_len -= hlen; 841 842#ifdef IPDIVERT 843 /* 844 * Divert or tee packet to the divert protocol if required. 845 */ 846 divert_info = divert_find_info(m); 847 if (divert_info != 0) { 848 struct mbuf *clone; 849 850 /* Clone packet if we're doing a 'tee' */ 851 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 852 clone = divert_clone(m); 853 else 854 clone = NULL; 855 856 /* Restore packet header fields to original values */ 857 ip->ip_len += hlen; 858 ip->ip_len = htons(ip->ip_len); 859 ip->ip_off = htons(ip->ip_off); 860 861 /* Deliver packet to divert input routine */ 862 divert_packet(m, 1); 863 ipstat.ips_delivered++; 864 865 /* If 'tee', continue with original packet */ 866 if (clone == NULL) 867 return; 868 m = clone; 869 ip = mtod(m, struct ip *); 870 ip->ip_len += hlen; 871 /* 872 * Jump backwards to complete processing of the 873 * packet. We do not need to clear args.next_hop 874 * as that will not be used again and the cloned packet 875 * doesn't contain a divert packet tag so we won't 876 * re-entry this block. 877 */ 878 goto pass; 879 } 880#endif 881 882#ifdef IPSEC 883 /* 884 * enforce IPsec policy checking if we are seeing last header. 885 * note that we do not visit this with protocols with pcb layer 886 * code - like udp/tcp/raw ip. 887 */ 888 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 889 ipsec4_in_reject(m, NULL)) { 890 ipsecstat.in_polvio++; 891 goto bad; 892 } 893#endif 894#if FAST_IPSEC 895 /* 896 * enforce IPsec policy checking if we are seeing last header. 897 * note that we do not visit this with protocols with pcb layer 898 * code - like udp/tcp/raw ip. 899 */ 900 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 901 /* 902 * Check if the packet has already had IPsec processing 903 * done. If so, then just pass it along. This tag gets 904 * set during AH, ESP, etc. input handling, before the 905 * packet is returned to the ip input queue for delivery. 906 */ 907 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 908 s = splnet(); 909 if (mtag != NULL) { 910 tdbi = (struct tdb_ident *)(mtag + 1); 911 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 912 } else { 913 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 914 IP_FORWARDING, &error); 915 } 916 if (sp != NULL) { 917 /* 918 * Check security policy against packet attributes. 919 */ 920 error = ipsec_in_reject(sp, m); 921 KEY_FREESP(&sp); 922 } else { 923 /* XXX error stat??? */ 924 error = EINVAL; 925DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 926 goto bad; 927 } 928 splx(s); 929 if (error) 930 goto bad; 931 } 932#endif /* FAST_IPSEC */ 933 934 /* 935 * Switch out to protocol's input routine. 936 */ 937 ipstat.ips_delivered++; 938 if (args.next_hop && ip->ip_p == IPPROTO_TCP) { 939 /* attach next hop info for TCP */ 940 struct m_tag *mtag = m_tag_get(PACKET_TAG_IPFORWARD, 941 sizeof(struct sockaddr_in *), M_NOWAIT); 942 if (mtag == NULL) 943 goto bad; 944 *(struct sockaddr_in **)(mtag+1) = args.next_hop; 945 m_tag_prepend(m, mtag); 946 } 947 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 948 return; 949bad: 950 m_freem(m); 951} 952 953/* 954 * Take incoming datagram fragment and try to reassemble it into 955 * whole datagram. If a chain for reassembly of this datagram already 956 * exists, then it is given as fp; otherwise have to make a chain. 957 * 958 * When IPDIVERT enabled, keep additional state with each packet that 959 * tells us if we need to divert or tee the packet we're building. 960 * In particular, *divinfo includes the port and TEE flag, 961 * *divert_rule is the number of the matching rule. 962 */ 963 964static struct mbuf * 965ip_reass(struct mbuf *m, struct ipqhead *head, struct ipq *fp) 966{ 967 struct ip *ip = mtod(m, struct ip *); 968 register struct mbuf *p, *q, *nq; 969 struct mbuf *t; 970 int hlen = ip->ip_hl << 2; 971 int i, next; 972 u_int8_t ecn, ecn0; 973 974 IPQ_LOCK_ASSERT(); 975 976 /* 977 * Presence of header sizes in mbufs 978 * would confuse code below. 979 */ 980 m->m_data += hlen; 981 m->m_len -= hlen; 982 983 /* 984 * If first fragment to arrive, create a reassembly queue. 985 */ 986 if (fp == NULL) { 987 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 988 goto dropfrag; 989 fp = mtod(t, struct ipq *); 990#ifdef MAC 991 if (mac_init_ipq(fp, M_NOWAIT) != 0) { 992 m_free(t); 993 goto dropfrag; 994 } 995 mac_create_ipq(m, fp); 996#endif 997 TAILQ_INSERT_HEAD(head, fp, ipq_list); 998 nipq++; 999 fp->ipq_nfrags = 1; 1000 fp->ipq_ttl = IPFRAGTTL; 1001 fp->ipq_p = ip->ip_p; 1002 fp->ipq_id = ip->ip_id; 1003 fp->ipq_src = ip->ip_src; 1004 fp->ipq_dst = ip->ip_dst; 1005 fp->ipq_frags = m; 1006 m->m_nextpkt = NULL; 1007 goto inserted; 1008 } else { 1009 fp->ipq_nfrags++; 1010#ifdef MAC 1011 mac_update_ipq(m, fp); 1012#endif 1013 } 1014 1015#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1016 1017 /* 1018 * Handle ECN by comparing this segment with the first one; 1019 * if CE is set, do not lose CE. 1020 * drop if CE and not-ECT are mixed for the same packet. 1021 */ 1022 ecn = ip->ip_tos & IPTOS_ECN_MASK; 1023 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 1024 if (ecn == IPTOS_ECN_CE) { 1025 if (ecn0 == IPTOS_ECN_NOTECT) 1026 goto dropfrag; 1027 if (ecn0 != IPTOS_ECN_CE) 1028 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 1029 } 1030 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 1031 goto dropfrag; 1032 1033 /* 1034 * Find a segment which begins after this one does. 1035 */ 1036 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1037 if (GETIP(q)->ip_off > ip->ip_off) 1038 break; 1039 1040 /* 1041 * If there is a preceding segment, it may provide some of 1042 * our data already. If so, drop the data from the incoming 1043 * segment. If it provides all of our data, drop us, otherwise 1044 * stick new segment in the proper place. 1045 * 1046 * If some of the data is dropped from the the preceding 1047 * segment, then it's checksum is invalidated. 1048 */ 1049 if (p) { 1050 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1051 if (i > 0) { 1052 if (i >= ip->ip_len) 1053 goto dropfrag; 1054 m_adj(m, i); 1055 m->m_pkthdr.csum_flags = 0; 1056 ip->ip_off += i; 1057 ip->ip_len -= i; 1058 } 1059 m->m_nextpkt = p->m_nextpkt; 1060 p->m_nextpkt = m; 1061 } else { 1062 m->m_nextpkt = fp->ipq_frags; 1063 fp->ipq_frags = m; 1064 } 1065 1066 /* 1067 * While we overlap succeeding segments trim them or, 1068 * if they are completely covered, dequeue them. 1069 */ 1070 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1071 q = nq) { 1072 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1073 if (i < GETIP(q)->ip_len) { 1074 GETIP(q)->ip_len -= i; 1075 GETIP(q)->ip_off += i; 1076 m_adj(q, i); 1077 q->m_pkthdr.csum_flags = 0; 1078 break; 1079 } 1080 nq = q->m_nextpkt; 1081 m->m_nextpkt = nq; 1082 ipstat.ips_fragdropped++; 1083 fp->ipq_nfrags--; 1084 m_freem(q); 1085 } 1086 1087inserted: 1088 1089#ifdef IPDIVERT 1090 if (ip->ip_off != 0) { 1091 /* 1092 * Strip any divert information; only the info 1093 * on the first fragment is used/kept. 1094 */ 1095 struct m_tag *mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 1096 if (mtag) 1097 m_tag_delete(m, mtag); 1098 } 1099#endif 1100 1101 /* 1102 * Check for complete reassembly and perform frag per packet 1103 * limiting. 1104 * 1105 * Frag limiting is performed here so that the nth frag has 1106 * a chance to complete the packet before we drop the packet. 1107 * As a result, n+1 frags are actually allowed per packet, but 1108 * only n will ever be stored. (n = maxfragsperpacket.) 1109 * 1110 */ 1111 next = 0; 1112 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1113 if (GETIP(q)->ip_off != next) { 1114 if (fp->ipq_nfrags > maxfragsperpacket) { 1115 ipstat.ips_fragdropped += fp->ipq_nfrags; 1116 ip_freef(head, fp); 1117 } 1118 return (0); 1119 } 1120 next += GETIP(q)->ip_len; 1121 } 1122 /* Make sure the last packet didn't have the IP_MF flag */ 1123 if (p->m_flags & M_FRAG) { 1124 if (fp->ipq_nfrags > maxfragsperpacket) { 1125 ipstat.ips_fragdropped += fp->ipq_nfrags; 1126 ip_freef(head, fp); 1127 } 1128 return (0); 1129 } 1130 1131 /* 1132 * Reassembly is complete. Make sure the packet is a sane size. 1133 */ 1134 q = fp->ipq_frags; 1135 ip = GETIP(q); 1136 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1137 ipstat.ips_toolong++; 1138 ipstat.ips_fragdropped += fp->ipq_nfrags; 1139 ip_freef(head, fp); 1140 return (0); 1141 } 1142 1143 /* 1144 * Concatenate fragments. 1145 */ 1146 m = q; 1147 t = m->m_next; 1148 m->m_next = 0; 1149 m_cat(m, t); 1150 nq = q->m_nextpkt; 1151 q->m_nextpkt = 0; 1152 for (q = nq; q != NULL; q = nq) { 1153 nq = q->m_nextpkt; 1154 q->m_nextpkt = NULL; 1155 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1156 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1157 m_cat(m, q); 1158 } 1159#ifdef MAC 1160 mac_create_datagram_from_ipq(fp, m); 1161 mac_destroy_ipq(fp); 1162#endif 1163 1164 /* 1165 * Create header for new ip packet by 1166 * modifying header of first packet; 1167 * dequeue and discard fragment reassembly header. 1168 * Make header visible. 1169 */ 1170 ip->ip_len = next; 1171 ip->ip_src = fp->ipq_src; 1172 ip->ip_dst = fp->ipq_dst; 1173 TAILQ_REMOVE(head, fp, ipq_list); 1174 nipq--; 1175 (void) m_free(dtom(fp)); 1176 m->m_len += (ip->ip_hl << 2); 1177 m->m_data -= (ip->ip_hl << 2); 1178 /* some debugging cruft by sklower, below, will go away soon */ 1179 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1180 m_fixhdr(m); 1181 return (m); 1182 1183dropfrag: 1184 ipstat.ips_fragdropped++; 1185 if (fp != NULL) 1186 fp->ipq_nfrags--; 1187 m_freem(m); 1188 return (0); 1189 1190#undef GETIP 1191} 1192 1193/* 1194 * Free a fragment reassembly header and all 1195 * associated datagrams. 1196 */ 1197static void 1198ip_freef(fhp, fp) 1199 struct ipqhead *fhp; 1200 struct ipq *fp; 1201{ 1202 register struct mbuf *q; 1203 1204 IPQ_LOCK_ASSERT(); 1205 1206 while (fp->ipq_frags) { 1207 q = fp->ipq_frags; 1208 fp->ipq_frags = q->m_nextpkt; 1209 m_freem(q); 1210 } 1211 TAILQ_REMOVE(fhp, fp, ipq_list); 1212 (void) m_free(dtom(fp)); 1213 nipq--; 1214} 1215 1216/* 1217 * IP timer processing; 1218 * if a timer expires on a reassembly 1219 * queue, discard it. 1220 */ 1221void 1222ip_slowtimo() 1223{ 1224 register struct ipq *fp; 1225 int s = splnet(); 1226 int i; 1227 1228 IPQ_LOCK(); 1229 for (i = 0; i < IPREASS_NHASH; i++) { 1230 for(fp = TAILQ_FIRST(&ipq[i]); fp;) { 1231 struct ipq *fpp; 1232 1233 fpp = fp; 1234 fp = TAILQ_NEXT(fp, ipq_list); 1235 if(--fpp->ipq_ttl == 0) { 1236 ipstat.ips_fragtimeout += fpp->ipq_nfrags; 1237 ip_freef(&ipq[i], fpp); 1238 } 1239 } 1240 } 1241 /* 1242 * If we are over the maximum number of fragments 1243 * (due to the limit being lowered), drain off 1244 * enough to get down to the new limit. 1245 */ 1246 if (maxnipq >= 0 && nipq > maxnipq) { 1247 for (i = 0; i < IPREASS_NHASH; i++) { 1248 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) { 1249 ipstat.ips_fragdropped += 1250 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1251 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1252 } 1253 } 1254 } 1255 IPQ_UNLOCK(); 1256 splx(s); 1257} 1258 1259/* 1260 * Drain off all datagram fragments. 1261 */ 1262void 1263ip_drain() 1264{ 1265 int i; 1266 1267 IPQ_LOCK(); 1268 for (i = 0; i < IPREASS_NHASH; i++) { 1269 while(!TAILQ_EMPTY(&ipq[i])) { 1270 ipstat.ips_fragdropped += 1271 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1272 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1273 } 1274 } 1275 IPQ_UNLOCK(); 1276 in_rtqdrain(); 1277} 1278 1279/* 1280 * Do option processing on a datagram, 1281 * possibly discarding it if bad options are encountered, 1282 * or forwarding it if source-routed. 1283 * The pass argument is used when operating in the IPSTEALTH 1284 * mode to tell what options to process: 1285 * [LS]SRR (pass 0) or the others (pass 1). 1286 * The reason for as many as two passes is that when doing IPSTEALTH, 1287 * non-routing options should be processed only if the packet is for us. 1288 * Returns 1 if packet has been forwarded/freed, 1289 * 0 if the packet should be processed further. 1290 */ 1291static int 1292ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop) 1293{ 1294 struct ip *ip = mtod(m, struct ip *); 1295 u_char *cp; 1296 struct in_ifaddr *ia; 1297 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1298 struct in_addr *sin, dst; 1299 n_time ntime; 1300 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 1301 1302 /* ignore or reject packets with IP options */ 1303 if (ip_doopts == 0) 1304 return 0; 1305 else if (ip_doopts == 2) { 1306 type = ICMP_UNREACH; 1307 code = ICMP_UNREACH_FILTER_PROHIB; 1308 goto bad; 1309 } 1310 1311 dst = ip->ip_dst; 1312 cp = (u_char *)(ip + 1); 1313 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1314 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1315 opt = cp[IPOPT_OPTVAL]; 1316 if (opt == IPOPT_EOL) 1317 break; 1318 if (opt == IPOPT_NOP) 1319 optlen = 1; 1320 else { 1321 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1322 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1323 goto bad; 1324 } 1325 optlen = cp[IPOPT_OLEN]; 1326 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1327 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1328 goto bad; 1329 } 1330 } 1331 switch (opt) { 1332 1333 default: 1334 break; 1335 1336 /* 1337 * Source routing with record. 1338 * Find interface with current destination address. 1339 * If none on this machine then drop if strictly routed, 1340 * or do nothing if loosely routed. 1341 * Record interface address and bring up next address 1342 * component. If strictly routed make sure next 1343 * address is on directly accessible net. 1344 */ 1345 case IPOPT_LSRR: 1346 case IPOPT_SSRR: 1347#ifdef IPSTEALTH 1348 if (ipstealth && pass > 0) 1349 break; 1350#endif 1351 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1352 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1353 goto bad; 1354 } 1355 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1356 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1357 goto bad; 1358 } 1359 ipaddr.sin_addr = ip->ip_dst; 1360 ia = (struct in_ifaddr *) 1361 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1362 if (ia == 0) { 1363 if (opt == IPOPT_SSRR) { 1364 type = ICMP_UNREACH; 1365 code = ICMP_UNREACH_SRCFAIL; 1366 goto bad; 1367 } 1368 if (!ip_dosourceroute) 1369 goto nosourcerouting; 1370 /* 1371 * Loose routing, and not at next destination 1372 * yet; nothing to do except forward. 1373 */ 1374 break; 1375 } 1376 off--; /* 0 origin */ 1377 if (off > optlen - (int)sizeof(struct in_addr)) { 1378 /* 1379 * End of source route. Should be for us. 1380 */ 1381 if (!ip_acceptsourceroute) 1382 goto nosourcerouting; 1383 save_rte(cp, ip->ip_src); 1384 break; 1385 } 1386#ifdef IPSTEALTH 1387 if (ipstealth) 1388 goto dropit; 1389#endif 1390 if (!ip_dosourceroute) { 1391 if (ipforwarding) { 1392 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1393 /* 1394 * Acting as a router, so generate ICMP 1395 */ 1396nosourcerouting: 1397 strcpy(buf, inet_ntoa(ip->ip_dst)); 1398 log(LOG_WARNING, 1399 "attempted source route from %s to %s\n", 1400 inet_ntoa(ip->ip_src), buf); 1401 type = ICMP_UNREACH; 1402 code = ICMP_UNREACH_SRCFAIL; 1403 goto bad; 1404 } else { 1405 /* 1406 * Not acting as a router, so silently drop. 1407 */ 1408#ifdef IPSTEALTH 1409dropit: 1410#endif 1411 ipstat.ips_cantforward++; 1412 m_freem(m); 1413 return (1); 1414 } 1415 } 1416 1417 /* 1418 * locate outgoing interface 1419 */ 1420 (void)memcpy(&ipaddr.sin_addr, cp + off, 1421 sizeof(ipaddr.sin_addr)); 1422 1423 if (opt == IPOPT_SSRR) { 1424#define INA struct in_ifaddr * 1425#define SA struct sockaddr * 1426 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1427 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1428 } else 1429 ia = ip_rtaddr(ipaddr.sin_addr); 1430 if (ia == 0) { 1431 type = ICMP_UNREACH; 1432 code = ICMP_UNREACH_SRCFAIL; 1433 goto bad; 1434 } 1435 ip->ip_dst = ipaddr.sin_addr; 1436 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1437 sizeof(struct in_addr)); 1438 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1439 /* 1440 * Let ip_intr's mcast routing check handle mcast pkts 1441 */ 1442 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1443 break; 1444 1445 case IPOPT_RR: 1446#ifdef IPSTEALTH 1447 if (ipstealth && pass == 0) 1448 break; 1449#endif 1450 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1451 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1452 goto bad; 1453 } 1454 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1455 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1456 goto bad; 1457 } 1458 /* 1459 * If no space remains, ignore. 1460 */ 1461 off--; /* 0 origin */ 1462 if (off > optlen - (int)sizeof(struct in_addr)) 1463 break; 1464 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1465 sizeof(ipaddr.sin_addr)); 1466 /* 1467 * locate outgoing interface; if we're the destination, 1468 * use the incoming interface (should be same). 1469 */ 1470 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1471 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1472 type = ICMP_UNREACH; 1473 code = ICMP_UNREACH_HOST; 1474 goto bad; 1475 } 1476 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1477 sizeof(struct in_addr)); 1478 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1479 break; 1480 1481 case IPOPT_TS: 1482#ifdef IPSTEALTH 1483 if (ipstealth && pass == 0) 1484 break; 1485#endif 1486 code = cp - (u_char *)ip; 1487 if (optlen < 4 || optlen > 40) { 1488 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1489 goto bad; 1490 } 1491 if ((off = cp[IPOPT_OFFSET]) < 5) { 1492 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1493 goto bad; 1494 } 1495 if (off > optlen - (int)sizeof(int32_t)) { 1496 cp[IPOPT_OFFSET + 1] += (1 << 4); 1497 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1498 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1499 goto bad; 1500 } 1501 break; 1502 } 1503 off--; /* 0 origin */ 1504 sin = (struct in_addr *)(cp + off); 1505 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1506 1507 case IPOPT_TS_TSONLY: 1508 break; 1509 1510 case IPOPT_TS_TSANDADDR: 1511 if (off + sizeof(n_time) + 1512 sizeof(struct in_addr) > optlen) { 1513 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1514 goto bad; 1515 } 1516 ipaddr.sin_addr = dst; 1517 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1518 m->m_pkthdr.rcvif); 1519 if (ia == 0) 1520 continue; 1521 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1522 sizeof(struct in_addr)); 1523 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1524 off += sizeof(struct in_addr); 1525 break; 1526 1527 case IPOPT_TS_PRESPEC: 1528 if (off + sizeof(n_time) + 1529 sizeof(struct in_addr) > optlen) { 1530 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1531 goto bad; 1532 } 1533 (void)memcpy(&ipaddr.sin_addr, sin, 1534 sizeof(struct in_addr)); 1535 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1536 continue; 1537 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1538 off += sizeof(struct in_addr); 1539 break; 1540 1541 default: 1542 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1543 goto bad; 1544 } 1545 ntime = iptime(); 1546 (void)memcpy(cp + off, &ntime, sizeof(n_time)); 1547 cp[IPOPT_OFFSET] += sizeof(n_time); 1548 } 1549 } 1550 if (forward && ipforwarding) { 1551 ip_forward(m, 1, next_hop); 1552 return (1); 1553 } 1554 return (0); 1555bad: 1556 icmp_error(m, type, code, 0, 0); 1557 ipstat.ips_badoptions++; 1558 return (1); 1559} 1560 1561/* 1562 * Given address of next destination (final or next hop), 1563 * return internet address info of interface to be used to get there. 1564 */ 1565struct in_ifaddr * 1566ip_rtaddr(dst) 1567 struct in_addr dst; 1568{ 1569 struct route sro; 1570 struct sockaddr_in *sin; 1571 struct in_ifaddr *ifa; 1572 1573 bzero(&sro, sizeof(sro)); 1574 sin = (struct sockaddr_in *)&sro.ro_dst; 1575 sin->sin_family = AF_INET; 1576 sin->sin_len = sizeof(*sin); 1577 sin->sin_addr = dst; 1578 rtalloc_ign(&sro, RTF_CLONING); 1579 1580 if (sro.ro_rt == NULL) 1581 return ((struct in_ifaddr *)0); 1582 1583 ifa = ifatoia(sro.ro_rt->rt_ifa); 1584 RTFREE(sro.ro_rt); 1585 return ifa; 1586} 1587 1588/* 1589 * Save incoming source route for use in replies, 1590 * to be picked up later by ip_srcroute if the receiver is interested. 1591 */ 1592static void 1593save_rte(option, dst) 1594 u_char *option; 1595 struct in_addr dst; 1596{ 1597 unsigned olen; 1598 1599 olen = option[IPOPT_OLEN]; 1600#ifdef DIAGNOSTIC 1601 if (ipprintfs) 1602 printf("save_rte: olen %d\n", olen); 1603#endif 1604 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1605 return; 1606 bcopy(option, ip_srcrt.srcopt, olen); 1607 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1608 ip_srcrt.dst = dst; 1609} 1610 1611/* 1612 * Retrieve incoming source route for use in replies, 1613 * in the same form used by setsockopt. 1614 * The first hop is placed before the options, will be removed later. 1615 */ 1616struct mbuf * 1617ip_srcroute() 1618{ 1619 register struct in_addr *p, *q; 1620 register struct mbuf *m; 1621 1622 if (ip_nhops == 0) 1623 return ((struct mbuf *)0); 1624 m = m_get(M_DONTWAIT, MT_HEADER); 1625 if (m == 0) 1626 return ((struct mbuf *)0); 1627 1628#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1629 1630 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1631 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1632 OPTSIZ; 1633#ifdef DIAGNOSTIC 1634 if (ipprintfs) 1635 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1636#endif 1637 1638 /* 1639 * First save first hop for return route 1640 */ 1641 p = &ip_srcrt.route[ip_nhops - 1]; 1642 *(mtod(m, struct in_addr *)) = *p--; 1643#ifdef DIAGNOSTIC 1644 if (ipprintfs) 1645 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1646#endif 1647 1648 /* 1649 * Copy option fields and padding (nop) to mbuf. 1650 */ 1651 ip_srcrt.nop = IPOPT_NOP; 1652 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1653 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1654 &ip_srcrt.nop, OPTSIZ); 1655 q = (struct in_addr *)(mtod(m, caddr_t) + 1656 sizeof(struct in_addr) + OPTSIZ); 1657#undef OPTSIZ 1658 /* 1659 * Record return path as an IP source route, 1660 * reversing the path (pointers are now aligned). 1661 */ 1662 while (p >= ip_srcrt.route) { 1663#ifdef DIAGNOSTIC 1664 if (ipprintfs) 1665 printf(" %lx", (u_long)ntohl(q->s_addr)); 1666#endif 1667 *q++ = *p--; 1668 } 1669 /* 1670 * Last hop goes to final destination. 1671 */ 1672 *q = ip_srcrt.dst; 1673#ifdef DIAGNOSTIC 1674 if (ipprintfs) 1675 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1676#endif 1677 return (m); 1678} 1679 1680/* 1681 * Strip out IP options, at higher 1682 * level protocol in the kernel. 1683 * Second argument is buffer to which options 1684 * will be moved, and return value is their length. 1685 * XXX should be deleted; last arg currently ignored. 1686 */ 1687void 1688ip_stripoptions(m, mopt) 1689 register struct mbuf *m; 1690 struct mbuf *mopt; 1691{ 1692 register int i; 1693 struct ip *ip = mtod(m, struct ip *); 1694 register caddr_t opts; 1695 int olen; 1696 1697 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1698 opts = (caddr_t)(ip + 1); 1699 i = m->m_len - (sizeof (struct ip) + olen); 1700 bcopy(opts + olen, opts, (unsigned)i); 1701 m->m_len -= olen; 1702 if (m->m_flags & M_PKTHDR) 1703 m->m_pkthdr.len -= olen; 1704 ip->ip_v = IPVERSION; 1705 ip->ip_hl = sizeof(struct ip) >> 2; 1706} 1707 1708u_char inetctlerrmap[PRC_NCMDS] = { 1709 0, 0, 0, 0, 1710 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1711 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1712 EMSGSIZE, EHOSTUNREACH, 0, 0, 1713 0, 0, EHOSTUNREACH, 0, 1714 ENOPROTOOPT, ECONNREFUSED 1715}; 1716 1717/* 1718 * Forward a packet. If some error occurs return the sender 1719 * an icmp packet. Note we can't always generate a meaningful 1720 * icmp message because icmp doesn't have a large enough repertoire 1721 * of codes and types. 1722 * 1723 * If not forwarding, just drop the packet. This could be confusing 1724 * if ipforwarding was zero but some routing protocol was advancing 1725 * us as a gateway to somewhere. However, we must let the routing 1726 * protocol deal with that. 1727 * 1728 * The srcrt parameter indicates whether the packet is being forwarded 1729 * via a source route. 1730 */ 1731static void 1732ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop) 1733{ 1734 struct ip *ip = mtod(m, struct ip *); 1735 struct in_ifaddr *ia; 1736 int error, type = 0, code = 0; 1737 struct mbuf *mcopy; 1738 n_long dest; 1739 struct in_addr pkt_dst; 1740 struct ifnet *destifp; 1741#if defined(IPSEC) || defined(FAST_IPSEC) 1742 struct ifnet dummyifp; 1743#endif 1744 1745 /* 1746 * Cache the destination address of the packet; this may be 1747 * changed by use of 'ipfw fwd'. 1748 */ 1749 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst; 1750 1751#ifdef DIAGNOSTIC 1752 if (ipprintfs) 1753 printf("forward: src %lx dst %lx ttl %x\n", 1754 (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr, 1755 ip->ip_ttl); 1756#endif 1757 1758 1759 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) { 1760 ipstat.ips_cantforward++; 1761 m_freem(m); 1762 return; 1763 } 1764#ifdef IPSTEALTH 1765 if (!ipstealth) { 1766#endif 1767 if (ip->ip_ttl <= IPTTLDEC) { 1768 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1769 0, 0); 1770 return; 1771 } 1772#ifdef IPSTEALTH 1773 } 1774#endif 1775 1776 if ((ia = ip_rtaddr(pkt_dst)) == 0) { 1777 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1778 return; 1779 } 1780 1781 /* 1782 * Save the IP header and at most 8 bytes of the payload, 1783 * in case we need to generate an ICMP message to the src. 1784 * 1785 * XXX this can be optimized a lot by saving the data in a local 1786 * buffer on the stack (72 bytes at most), and only allocating the 1787 * mbuf if really necessary. The vast majority of the packets 1788 * are forwarded without having to send an ICMP back (either 1789 * because unnecessary, or because rate limited), so we are 1790 * really we are wasting a lot of work here. 1791 * 1792 * We don't use m_copy() because it might return a reference 1793 * to a shared cluster. Both this function and ip_output() 1794 * assume exclusive access to the IP header in `m', so any 1795 * data in a cluster may change before we reach icmp_error(). 1796 */ 1797 MGET(mcopy, M_DONTWAIT, m->m_type); 1798 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1799 /* 1800 * It's probably ok if the pkthdr dup fails (because 1801 * the deep copy of the tag chain failed), but for now 1802 * be conservative and just discard the copy since 1803 * code below may some day want the tags. 1804 */ 1805 m_free(mcopy); 1806 mcopy = NULL; 1807 } 1808 if (mcopy != NULL) { 1809 mcopy->m_len = imin((ip->ip_hl << 2) + 8, 1810 (int)ip->ip_len); 1811 mcopy->m_pkthdr.len = mcopy->m_len; 1812 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1813 } 1814 1815#ifdef IPSTEALTH 1816 if (!ipstealth) { 1817#endif 1818 ip->ip_ttl -= IPTTLDEC; 1819#ifdef IPSTEALTH 1820 } 1821#endif 1822 1823 /* 1824 * If forwarding packet using same interface that it came in on, 1825 * perhaps should send a redirect to sender to shortcut a hop. 1826 * Only send redirect if source is sending directly to us, 1827 * and if packet was not source routed (or has any options). 1828 * Also, don't send redirect if forwarding using a default route 1829 * or a route modified by a redirect. 1830 */ 1831 dest = 0; 1832 if (ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) { 1833 struct sockaddr_in *sin; 1834 struct route ro; 1835 struct rtentry *rt; 1836 1837 bzero(&ro, sizeof(ro)); 1838 sin = (struct sockaddr_in *)&ro.ro_dst; 1839 sin->sin_family = AF_INET; 1840 sin->sin_len = sizeof(*sin); 1841 sin->sin_addr = pkt_dst; 1842 rtalloc_ign(&ro, RTF_CLONING); 1843 1844 rt = ro.ro_rt; 1845 1846 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1847 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1848 ipsendredirects && !srcrt && !next_hop) { 1849#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1850 u_long src = ntohl(ip->ip_src.s_addr); 1851 1852 if (RTA(rt) && 1853 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1854 if (rt->rt_flags & RTF_GATEWAY) 1855 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1856 else 1857 dest = pkt_dst.s_addr; 1858 /* Router requirements says to only send host redirects */ 1859 type = ICMP_REDIRECT; 1860 code = ICMP_REDIRECT_HOST; 1861#ifdef DIAGNOSTIC 1862 if (ipprintfs) 1863 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1864#endif 1865 } 1866 } 1867 if (rt) 1868 RTFREE(rt); 1869 } 1870 1871 if (next_hop) { 1872 struct m_tag *mtag = m_tag_get(PACKET_TAG_IPFORWARD, 1873 sizeof(struct sockaddr_in *), M_NOWAIT); 1874 if (mtag == NULL) { 1875 m_freem(m); 1876 return; 1877 } 1878 *(struct sockaddr_in **)(mtag+1) = next_hop; 1879 m_tag_prepend(m, mtag); 1880 } 1881 error = ip_output(m, (struct mbuf *)0, NULL, IP_FORWARDING, 0, NULL); 1882 if (error) 1883 ipstat.ips_cantforward++; 1884 else { 1885 ipstat.ips_forward++; 1886 if (type) 1887 ipstat.ips_redirectsent++; 1888 else { 1889 if (mcopy) 1890 m_freem(mcopy); 1891 return; 1892 } 1893 } 1894 if (mcopy == NULL) 1895 return; 1896 destifp = NULL; 1897 1898 switch (error) { 1899 1900 case 0: /* forwarded, but need redirect */ 1901 /* type, code set above */ 1902 break; 1903 1904 case ENETUNREACH: /* shouldn't happen, checked above */ 1905 case EHOSTUNREACH: 1906 case ENETDOWN: 1907 case EHOSTDOWN: 1908 default: 1909 type = ICMP_UNREACH; 1910 code = ICMP_UNREACH_HOST; 1911 break; 1912 1913 case EMSGSIZE: 1914 type = ICMP_UNREACH; 1915 code = ICMP_UNREACH_NEEDFRAG; 1916#if defined(IPSEC) || defined(FAST_IPSEC) 1917 /* 1918 * If the packet is routed over IPsec tunnel, tell the 1919 * originator the tunnel MTU. 1920 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1921 * XXX quickhack!!! 1922 */ 1923 { 1924 struct secpolicy *sp = NULL; 1925 int ipsecerror; 1926 int ipsechdr; 1927 struct route *ro; 1928 1929#ifdef IPSEC 1930 sp = ipsec4_getpolicybyaddr(mcopy, 1931 IPSEC_DIR_OUTBOUND, 1932 IP_FORWARDING, 1933 &ipsecerror); 1934#else /* FAST_IPSEC */ 1935 sp = ipsec_getpolicybyaddr(mcopy, 1936 IPSEC_DIR_OUTBOUND, 1937 IP_FORWARDING, 1938 &ipsecerror); 1939#endif 1940 if (sp != NULL) { 1941 /* count IPsec header size */ 1942 ipsechdr = ipsec4_hdrsiz(mcopy, 1943 IPSEC_DIR_OUTBOUND, 1944 NULL); 1945 1946 /* 1947 * find the correct route for outer IPv4 1948 * header, compute tunnel MTU. 1949 * 1950 * XXX BUG ALERT 1951 * The "dummyifp" code relies upon the fact 1952 * that icmp_error() touches only ifp->if_mtu. 1953 */ 1954 /*XXX*/ 1955 destifp = NULL; 1956 if (sp->req != NULL 1957 && sp->req->sav != NULL 1958 && sp->req->sav->sah != NULL) { 1959 ro = &sp->req->sav->sah->sa_route; 1960 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1961 dummyifp.if_mtu = 1962 ro->ro_rt->rt_rmx.rmx_mtu ? 1963 ro->ro_rt->rt_rmx.rmx_mtu : 1964 ro->ro_rt->rt_ifp->if_mtu; 1965 dummyifp.if_mtu -= ipsechdr; 1966 destifp = &dummyifp; 1967 } 1968 } 1969 1970#ifdef IPSEC 1971 key_freesp(sp); 1972#else /* FAST_IPSEC */ 1973 KEY_FREESP(&sp); 1974#endif 1975 ipstat.ips_cantfrag++; 1976 break; 1977 } else 1978#endif /*IPSEC || FAST_IPSEC*/ 1979 destifp = ia->ia_ifp; 1980#if defined(IPSEC) || defined(FAST_IPSEC) 1981 } 1982#endif /*IPSEC || FAST_IPSEC*/ 1983 ipstat.ips_cantfrag++; 1984 break; 1985 1986 case ENOBUFS: 1987 /* 1988 * A router should not generate ICMP_SOURCEQUENCH as 1989 * required in RFC1812 Requirements for IP Version 4 Routers. 1990 * Source quench could be a big problem under DoS attacks, 1991 * or if the underlying interface is rate-limited. 1992 * Those who need source quench packets may re-enable them 1993 * via the net.inet.ip.sendsourcequench sysctl. 1994 */ 1995 if (ip_sendsourcequench == 0) { 1996 m_freem(mcopy); 1997 return; 1998 } else { 1999 type = ICMP_SOURCEQUENCH; 2000 code = 0; 2001 } 2002 break; 2003 2004 case EACCES: /* ipfw denied packet */ 2005 m_freem(mcopy); 2006 return; 2007 } 2008 icmp_error(mcopy, type, code, dest, destifp); 2009} 2010 2011void 2012ip_savecontrol(inp, mp, ip, m) 2013 register struct inpcb *inp; 2014 register struct mbuf **mp; 2015 register struct ip *ip; 2016 register struct mbuf *m; 2017{ 2018 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 2019 struct bintime bt; 2020 2021 bintime(&bt); 2022 if (inp->inp_socket->so_options & SO_BINTIME) { 2023 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 2024 SCM_BINTIME, SOL_SOCKET); 2025 if (*mp) 2026 mp = &(*mp)->m_next; 2027 } 2028 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2029 struct timeval tv; 2030 2031 bintime2timeval(&bt, &tv); 2032 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2033 SCM_TIMESTAMP, SOL_SOCKET); 2034 if (*mp) 2035 mp = &(*mp)->m_next; 2036 } 2037 } 2038 if (inp->inp_flags & INP_RECVDSTADDR) { 2039 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2040 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2041 if (*mp) 2042 mp = &(*mp)->m_next; 2043 } 2044 if (inp->inp_flags & INP_RECVTTL) { 2045 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 2046 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 2047 if (*mp) 2048 mp = &(*mp)->m_next; 2049 } 2050#ifdef notyet 2051 /* XXX 2052 * Moving these out of udp_input() made them even more broken 2053 * than they already were. 2054 */ 2055 /* options were tossed already */ 2056 if (inp->inp_flags & INP_RECVOPTS) { 2057 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2058 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2059 if (*mp) 2060 mp = &(*mp)->m_next; 2061 } 2062 /* ip_srcroute doesn't do what we want here, need to fix */ 2063 if (inp->inp_flags & INP_RECVRETOPTS) { 2064 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 2065 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2066 if (*mp) 2067 mp = &(*mp)->m_next; 2068 } 2069#endif 2070 if (inp->inp_flags & INP_RECVIF) { 2071 struct ifnet *ifp; 2072 struct sdlbuf { 2073 struct sockaddr_dl sdl; 2074 u_char pad[32]; 2075 } sdlbuf; 2076 struct sockaddr_dl *sdp; 2077 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2078 2079 if (((ifp = m->m_pkthdr.rcvif)) 2080 && ( ifp->if_index && (ifp->if_index <= if_index))) { 2081 sdp = (struct sockaddr_dl *) 2082 (ifaddr_byindex(ifp->if_index)->ifa_addr); 2083 /* 2084 * Change our mind and don't try copy. 2085 */ 2086 if ((sdp->sdl_family != AF_LINK) 2087 || (sdp->sdl_len > sizeof(sdlbuf))) { 2088 goto makedummy; 2089 } 2090 bcopy(sdp, sdl2, sdp->sdl_len); 2091 } else { 2092makedummy: 2093 sdl2->sdl_len 2094 = offsetof(struct sockaddr_dl, sdl_data[0]); 2095 sdl2->sdl_family = AF_LINK; 2096 sdl2->sdl_index = 0; 2097 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2098 } 2099 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 2100 IP_RECVIF, IPPROTO_IP); 2101 if (*mp) 2102 mp = &(*mp)->m_next; 2103 } 2104} 2105 2106/* 2107 * XXX these routines are called from the upper part of the kernel. 2108 * They need to be locked when we remove Giant. 2109 * 2110 * They could also be moved to ip_mroute.c, since all the RSVP 2111 * handling is done there already. 2112 */ 2113static int ip_rsvp_on; 2114struct socket *ip_rsvpd; 2115int 2116ip_rsvp_init(struct socket *so) 2117{ 2118 if (so->so_type != SOCK_RAW || 2119 so->so_proto->pr_protocol != IPPROTO_RSVP) 2120 return EOPNOTSUPP; 2121 2122 if (ip_rsvpd != NULL) 2123 return EADDRINUSE; 2124 2125 ip_rsvpd = so; 2126 /* 2127 * This may seem silly, but we need to be sure we don't over-increment 2128 * the RSVP counter, in case something slips up. 2129 */ 2130 if (!ip_rsvp_on) { 2131 ip_rsvp_on = 1; 2132 rsvp_on++; 2133 } 2134 2135 return 0; 2136} 2137 2138int 2139ip_rsvp_done(void) 2140{ 2141 ip_rsvpd = NULL; 2142 /* 2143 * This may seem silly, but we need to be sure we don't over-decrement 2144 * the RSVP counter, in case something slips up. 2145 */ 2146 if (ip_rsvp_on) { 2147 ip_rsvp_on = 0; 2148 rsvp_on--; 2149 } 2150 return 0; 2151} 2152 2153void 2154rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 2155{ 2156 if (rsvp_input_p) { /* call the real one if loaded */ 2157 rsvp_input_p(m, off); 2158 return; 2159 } 2160 2161 /* Can still get packets with rsvp_on = 0 if there is a local member 2162 * of the group to which the RSVP packet is addressed. But in this 2163 * case we want to throw the packet away. 2164 */ 2165 2166 if (!rsvp_on) { 2167 m_freem(m); 2168 return; 2169 } 2170 2171 if (ip_rsvpd != NULL) { 2172 rip_input(m, off); 2173 return; 2174 } 2175 /* Drop the packet */ 2176 m_freem(m); 2177} 2178