ip_reass.c revision 135275
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 135275 2004-09-15 20:17:03Z andre $ 31 */ 32 33#include "opt_bootp.h" 34#include "opt_ipfw.h" 35#include "opt_ipstealth.h" 36#include "opt_ipsec.h" 37#include "opt_mac.h" 38 39#include <sys/param.h> 40#include <sys/systm.h> 41#include <sys/mac.h> 42#include <sys/mbuf.h> 43#include <sys/malloc.h> 44#include <sys/domain.h> 45#include <sys/protosw.h> 46#include <sys/socket.h> 47#include <sys/time.h> 48#include <sys/kernel.h> 49#include <sys/syslog.h> 50#include <sys/sysctl.h> 51 52#include <net/pfil.h> 53#include <net/if.h> 54#include <net/if_types.h> 55#include <net/if_var.h> 56#include <net/if_dl.h> 57#include <net/route.h> 58#include <net/netisr.h> 59 60#include <netinet/in.h> 61#include <netinet/in_systm.h> 62#include <netinet/in_var.h> 63#include <netinet/ip.h> 64#include <netinet/in_pcb.h> 65#include <netinet/ip_var.h> 66#include <netinet/ip_icmp.h> 67#include <machine/in_cksum.h> 68 69#include <sys/socketvar.h> 70 71/* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */ 72#include <netinet/ip_fw.h> 73#include <netinet/ip_dummynet.h> 74 75#ifdef IPSEC 76#include <netinet6/ipsec.h> 77#include <netkey/key.h> 78#endif 79 80#ifdef FAST_IPSEC 81#include <netipsec/ipsec.h> 82#include <netipsec/key.h> 83#endif 84 85int rsvp_on = 0; 86 87int ipforwarding = 0; 88SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 89 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 90 91static int ipsendredirects = 1; /* XXX */ 92SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 93 &ipsendredirects, 0, "Enable sending IP redirects"); 94 95int ip_defttl = IPDEFTTL; 96SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 97 &ip_defttl, 0, "Maximum TTL on IP packets"); 98 99static int ip_dosourceroute = 0; 100SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 101 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 102 103static int ip_acceptsourceroute = 0; 104SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 105 CTLFLAG_RW, &ip_acceptsourceroute, 0, 106 "Enable accepting source routed IP packets"); 107 108int ip_doopts = 1; /* 0 = ignore, 1 = process, 2 = reject */ 109SYSCTL_INT(_net_inet_ip, OID_AUTO, process_options, CTLFLAG_RW, 110 &ip_doopts, 0, "Enable IP options processing ([LS]SRR, RR, TS)"); 111 112static int ip_keepfaith = 0; 113SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 114 &ip_keepfaith, 0, 115 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 116 117static int nipq = 0; /* total # of reass queues */ 118static int maxnipq; 119SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 120 &maxnipq, 0, 121 "Maximum number of IPv4 fragment reassembly queue entries"); 122 123static int maxfragsperpacket; 124SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 125 &maxfragsperpacket, 0, 126 "Maximum number of IPv4 fragments allowed per packet"); 127 128static int ip_sendsourcequench = 0; 129SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 130 &ip_sendsourcequench, 0, 131 "Enable the transmission of source quench packets"); 132 133int ip_do_randomid = 0; 134SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 135 &ip_do_randomid, 0, 136 "Assign random ip_id values"); 137 138/* 139 * XXX - Setting ip_checkinterface mostly implements the receive side of 140 * the Strong ES model described in RFC 1122, but since the routing table 141 * and transmit implementation do not implement the Strong ES model, 142 * setting this to 1 results in an odd hybrid. 143 * 144 * XXX - ip_checkinterface currently must be disabled if you use ipnat 145 * to translate the destination address to another local interface. 146 * 147 * XXX - ip_checkinterface must be disabled if you add IP aliases 148 * to the loopback interface instead of the interface where the 149 * packets for those addresses are received. 150 */ 151static int ip_checkinterface = 1; 152SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 153 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 154 155#ifdef DIAGNOSTIC 156static int ipprintfs = 0; 157#endif 158 159struct pfil_head inet_pfil_hook; /* Packet filter hooks */ 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 * ipfw_ether and ipfw_bridge hooks. 208 * XXX: Temporary until those are converted to pfil_hooks as well. 209 */ 210ip_fw_chk_t *ip_fw_chk_ptr = NULL; 211ip_dn_io_t *ip_dn_io_ptr = NULL; 212int fw_enable = 1; 213int fw_one_pass = 1; 214 215/* 216 * XXX this is ugly. IP options source routing magic. 217 */ 218struct ipoptrt { 219 struct in_addr dst; /* final destination */ 220 char nop; /* one NOP to align */ 221 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 222 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 223}; 224 225struct ipopt_tag { 226 struct m_tag tag; 227 int ip_nhops; 228 struct ipoptrt ip_srcrt; 229}; 230 231static void save_rte(struct mbuf *, u_char *, struct in_addr); 232static int ip_dooptions(struct mbuf *m, int); 233static void ip_forward(struct mbuf *m, int srcrt); 234static void ip_freef(struct ipqhead *, struct ipq *); 235 236/* 237 * IP initialization: fill in IP protocol switch table. 238 * All protocols not implemented in kernel go to raw IP protocol handler. 239 */ 240void 241ip_init() 242{ 243 register struct protosw *pr; 244 register int i; 245 246 TAILQ_INIT(&in_ifaddrhead); 247 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 248 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 249 if (pr == 0) 250 panic("ip_init"); 251 for (i = 0; i < IPPROTO_MAX; i++) 252 ip_protox[i] = pr - inetsw; 253 for (pr = inetdomain.dom_protosw; 254 pr < inetdomain.dom_protoswNPROTOSW; pr++) 255 if (pr->pr_domain->dom_family == PF_INET && 256 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 257 ip_protox[pr->pr_protocol] = pr - inetsw; 258 259 /* Initialize packet filter hooks. */ 260 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 261 inet_pfil_hook.ph_af = AF_INET; 262 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 263 printf("%s: WARNING: unable to register pfil hook, " 264 "error %d\n", __func__, i); 265 266 IPQ_LOCK_INIT(); 267 for (i = 0; i < IPREASS_NHASH; i++) 268 TAILQ_INIT(&ipq[i]); 269 270 maxnipq = nmbclusters / 32; 271 maxfragsperpacket = 16; 272 273 ip_id = time_second & 0xffff; 274 ipintrq.ifq_maxlen = ipqmaxlen; 275 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF); 276 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE); 277} 278 279/* 280 * Ip input routine. Checksum and byte swap header. If fragmented 281 * try to reassemble. Process options. Pass to next level. 282 */ 283void 284ip_input(struct mbuf *m) 285{ 286 struct ip *ip = NULL; 287 struct in_ifaddr *ia = NULL; 288 struct ifaddr *ifa; 289 int checkif, hlen = 0; 290 u_short sum; 291 int dchg = 0; /* dest changed after fw */ 292 struct in_addr odst; /* original dst address */ 293#ifdef FAST_IPSEC 294 struct m_tag *mtag; 295 struct tdb_ident *tdbi; 296 struct secpolicy *sp; 297 int s, error; 298#endif /* FAST_IPSEC */ 299 300 M_ASSERTPKTHDR(m); 301 302 if (m->m_flags & M_FASTFWD_OURS) { 303 /* 304 * Firewall or NAT changed destination to local. 305 * We expect ip_len and ip_off to be in host byte order. 306 */ 307 m->m_flags &= ~M_FASTFWD_OURS; 308 /* Set up some basics that will be used later. */ 309 ip = mtod(m, struct ip *); 310 hlen = ip->ip_hl << 2; 311 goto ours; 312 } 313 314 ipstat.ips_total++; 315 316 if (m->m_pkthdr.len < sizeof(struct ip)) 317 goto tooshort; 318 319 if (m->m_len < sizeof (struct ip) && 320 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 321 ipstat.ips_toosmall++; 322 return; 323 } 324 ip = mtod(m, struct ip *); 325 326 if (ip->ip_v != IPVERSION) { 327 ipstat.ips_badvers++; 328 goto bad; 329 } 330 331 hlen = ip->ip_hl << 2; 332 if (hlen < sizeof(struct ip)) { /* minimum header length */ 333 ipstat.ips_badhlen++; 334 goto bad; 335 } 336 if (hlen > m->m_len) { 337 if ((m = m_pullup(m, hlen)) == NULL) { 338 ipstat.ips_badhlen++; 339 return; 340 } 341 ip = mtod(m, struct ip *); 342 } 343 344 /* 127/8 must not appear on wire - RFC1122 */ 345 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 346 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 347 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 348 ipstat.ips_badaddr++; 349 goto bad; 350 } 351 } 352 353 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 354 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 355 } else { 356 if (hlen == sizeof(struct ip)) { 357 sum = in_cksum_hdr(ip); 358 } else { 359 sum = in_cksum(m, hlen); 360 } 361 } 362 if (sum) { 363 ipstat.ips_badsum++; 364 goto bad; 365 } 366 367#ifdef ALTQ 368 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 369 /* packet is dropped by traffic conditioner */ 370 return; 371#endif 372 373 /* 374 * Convert fields to host representation. 375 */ 376 ip->ip_len = ntohs(ip->ip_len); 377 if (ip->ip_len < hlen) { 378 ipstat.ips_badlen++; 379 goto bad; 380 } 381 ip->ip_off = ntohs(ip->ip_off); 382 383 /* 384 * Check that the amount of data in the buffers 385 * is as at least much as the IP header would have us expect. 386 * Trim mbufs if longer than we expect. 387 * Drop packet if shorter than we expect. 388 */ 389 if (m->m_pkthdr.len < ip->ip_len) { 390tooshort: 391 ipstat.ips_tooshort++; 392 goto bad; 393 } 394 if (m->m_pkthdr.len > ip->ip_len) { 395 if (m->m_len == m->m_pkthdr.len) { 396 m->m_len = ip->ip_len; 397 m->m_pkthdr.len = ip->ip_len; 398 } else 399 m_adj(m, ip->ip_len - m->m_pkthdr.len); 400 } 401#if defined(IPSEC) && !defined(IPSEC_FILTERGIF) 402 /* 403 * Bypass packet filtering for packets from a tunnel (gif). 404 */ 405 if (ipsec_getnhist(m)) 406 goto passin; 407#endif 408#if defined(FAST_IPSEC) && !defined(IPSEC_FILTERGIF) 409 /* 410 * Bypass packet filtering for packets from a tunnel (gif). 411 */ 412 if (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL) 413 goto passin; 414#endif 415 416 /* 417 * Run through list of hooks for input packets. 418 * 419 * NB: Beware of the destination address changing (e.g. 420 * by NAT rewriting). When this happens, tell 421 * ip_forward to do the right thing. 422 */ 423 424 /* Jump over all PFIL processing if hooks are not active. */ 425 if (inet_pfil_hook.ph_busy_count == -1) 426 goto passin; 427 428 odst = ip->ip_dst; 429 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 430 PFIL_IN) != 0) 431 return; 432 if (m == NULL) /* consumed by filter */ 433 return; 434 435 ip = mtod(m, struct ip *); 436 dchg = (odst.s_addr != ip->ip_dst.s_addr); 437 438#ifdef IPFIREWALL_FORWARD 439 if (m->m_flags & M_FASTFWD_OURS) { 440 m->m_flags &= ~M_FASTFWD_OURS; 441 goto ours; 442 } 443 dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL); 444#endif /* IPFIREWALL_FORWARD */ 445 446passin: 447 /* 448 * Process options and, if not destined for us, 449 * ship it on. ip_dooptions returns 1 when an 450 * error was detected (causing an icmp message 451 * to be sent and the original packet to be freed). 452 */ 453 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 454 return; 455 456 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 457 * matter if it is destined to another node, or whether it is 458 * a multicast one, RSVP wants it! and prevents it from being forwarded 459 * anywhere else. Also checks if the rsvp daemon is running before 460 * grabbing the packet. 461 */ 462 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 463 goto ours; 464 465 /* 466 * Check our list of addresses, to see if the packet is for us. 467 * If we don't have any addresses, assume any unicast packet 468 * we receive might be for us (and let the upper layers deal 469 * with it). 470 */ 471 if (TAILQ_EMPTY(&in_ifaddrhead) && 472 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 473 goto ours; 474 475 /* 476 * Enable a consistency check between the destination address 477 * and the arrival interface for a unicast packet (the RFC 1122 478 * strong ES model) if IP forwarding is disabled and the packet 479 * is not locally generated and the packet is not subject to 480 * 'ipfw fwd'. 481 * 482 * XXX - Checking also should be disabled if the destination 483 * address is ipnat'ed to a different interface. 484 * 485 * XXX - Checking is incompatible with IP aliases added 486 * to the loopback interface instead of the interface where 487 * the packets are received. 488 */ 489 checkif = ip_checkinterface && (ipforwarding == 0) && 490 m->m_pkthdr.rcvif != NULL && 491 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 492 (dchg == 0); 493 494 /* 495 * Check for exact addresses in the hash bucket. 496 */ 497 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 498 /* 499 * If the address matches, verify that the packet 500 * arrived via the correct interface if checking is 501 * enabled. 502 */ 503 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 504 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 505 goto ours; 506 } 507 /* 508 * Check for broadcast addresses. 509 * 510 * Only accept broadcast packets that arrive via the matching 511 * interface. Reception of forwarded directed broadcasts would 512 * be handled via ip_forward() and ether_output() with the loopback 513 * into the stack for SIMPLEX interfaces handled by ether_output(). 514 */ 515 if (m->m_pkthdr.rcvif != NULL && 516 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 517 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 518 if (ifa->ifa_addr->sa_family != AF_INET) 519 continue; 520 ia = ifatoia(ifa); 521 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 522 ip->ip_dst.s_addr) 523 goto ours; 524 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) 525 goto ours; 526#ifdef BOOTP_COMPAT 527 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 528 goto ours; 529#endif 530 } 531 } 532 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 533 struct in_multi *inm; 534 if (ip_mrouter) { 535 /* 536 * If we are acting as a multicast router, all 537 * incoming multicast packets are passed to the 538 * kernel-level multicast forwarding function. 539 * The packet is returned (relatively) intact; if 540 * ip_mforward() returns a non-zero value, the packet 541 * must be discarded, else it may be accepted below. 542 */ 543 if (ip_mforward && 544 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 545 ipstat.ips_cantforward++; 546 m_freem(m); 547 return; 548 } 549 550 /* 551 * The process-level routing daemon needs to receive 552 * all multicast IGMP packets, whether or not this 553 * host belongs to their destination groups. 554 */ 555 if (ip->ip_p == IPPROTO_IGMP) 556 goto ours; 557 ipstat.ips_forward++; 558 } 559 /* 560 * See if we belong to the destination multicast group on the 561 * arrival interface. 562 */ 563 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 564 if (inm == NULL) { 565 ipstat.ips_notmember++; 566 m_freem(m); 567 return; 568 } 569 goto ours; 570 } 571 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 572 goto ours; 573 if (ip->ip_dst.s_addr == INADDR_ANY) 574 goto ours; 575 576 /* 577 * FAITH(Firewall Aided Internet Translator) 578 */ 579 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 580 if (ip_keepfaith) { 581 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 582 goto ours; 583 } 584 m_freem(m); 585 return; 586 } 587 588 /* 589 * Not for us; forward if possible and desirable. 590 */ 591 if (ipforwarding == 0) { 592 ipstat.ips_cantforward++; 593 m_freem(m); 594 } else { 595#ifdef IPSEC 596 /* 597 * Enforce inbound IPsec SPD. 598 */ 599 if (ipsec4_in_reject(m, NULL)) { 600 ipsecstat.in_polvio++; 601 goto bad; 602 } 603#endif /* IPSEC */ 604#ifdef FAST_IPSEC 605 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 606 s = splnet(); 607 if (mtag != NULL) { 608 tdbi = (struct tdb_ident *)(mtag + 1); 609 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 610 } else { 611 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 612 IP_FORWARDING, &error); 613 } 614 if (sp == NULL) { /* NB: can happen if error */ 615 splx(s); 616 /*XXX error stat???*/ 617 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 618 goto bad; 619 } 620 621 /* 622 * Check security policy against packet attributes. 623 */ 624 error = ipsec_in_reject(sp, m); 625 KEY_FREESP(&sp); 626 splx(s); 627 if (error) { 628 ipstat.ips_cantforward++; 629 goto bad; 630 } 631#endif /* FAST_IPSEC */ 632 ip_forward(m, dchg); 633 } 634 return; 635 636ours: 637#ifdef IPSTEALTH 638 /* 639 * IPSTEALTH: Process non-routing options only 640 * if the packet is destined for us. 641 */ 642 if (ipstealth && hlen > sizeof (struct ip) && 643 ip_dooptions(m, 1)) 644 return; 645#endif /* IPSTEALTH */ 646 647 /* Count the packet in the ip address stats */ 648 if (ia != NULL) { 649 ia->ia_ifa.if_ipackets++; 650 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 651 } 652 653 /* 654 * Attempt reassembly; if it succeeds, proceed. 655 * ip_reass() will return a different mbuf. 656 */ 657 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 658 m = ip_reass(m); 659 if (m == NULL) 660 return; 661 ip = mtod(m, struct ip *); 662 /* Get the header length of the reassembled packet */ 663 hlen = ip->ip_hl << 2; 664 } 665 666 /* 667 * Further protocols expect the packet length to be w/o the 668 * IP header. 669 */ 670 ip->ip_len -= hlen; 671 672#ifdef IPSEC 673 /* 674 * enforce IPsec policy checking if we are seeing last header. 675 * note that we do not visit this with protocols with pcb layer 676 * code - like udp/tcp/raw ip. 677 */ 678 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 679 ipsec4_in_reject(m, NULL)) { 680 ipsecstat.in_polvio++; 681 goto bad; 682 } 683#endif 684#if FAST_IPSEC 685 /* 686 * enforce IPsec policy checking if we are seeing last header. 687 * note that we do not visit this with protocols with pcb layer 688 * code - like udp/tcp/raw ip. 689 */ 690 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 691 /* 692 * Check if the packet has already had IPsec processing 693 * done. If so, then just pass it along. This tag gets 694 * set during AH, ESP, etc. input handling, before the 695 * packet is returned to the ip input queue for delivery. 696 */ 697 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 698 s = splnet(); 699 if (mtag != NULL) { 700 tdbi = (struct tdb_ident *)(mtag + 1); 701 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 702 } else { 703 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 704 IP_FORWARDING, &error); 705 } 706 if (sp != NULL) { 707 /* 708 * Check security policy against packet attributes. 709 */ 710 error = ipsec_in_reject(sp, m); 711 KEY_FREESP(&sp); 712 } else { 713 /* XXX error stat??? */ 714 error = EINVAL; 715DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 716 goto bad; 717 } 718 splx(s); 719 if (error) 720 goto bad; 721 } 722#endif /* FAST_IPSEC */ 723 724 /* 725 * Switch out to protocol's input routine. 726 */ 727 ipstat.ips_delivered++; 728 729 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 730 return; 731bad: 732 m_freem(m); 733} 734 735/* 736 * Take incoming datagram fragment and try to reassemble it into 737 * whole datagram. If the argument is the first fragment or one 738 * in between the function will return NULL and store the mbuf 739 * in the fragment chain. If the argument is the last fragment 740 * the packet will be reassembled and the pointer to the new 741 * mbuf returned for further processing. Only m_tags attached 742 * to the first packet/fragment are preserved. 743 * The IP header is *NOT* adjusted out of iplen. 744 */ 745 746struct mbuf * 747ip_reass(struct mbuf *m) 748{ 749 struct ip *ip; 750 struct mbuf *p, *q, *nq, *t; 751 struct ipq *fp = NULL; 752 struct ipqhead *head; 753 int i, hlen, next; 754 u_int8_t ecn, ecn0; 755 u_short hash; 756 757 /* If maxnipq is 0, never accept fragments. */ 758 if (maxnipq == 0) { 759 ipstat.ips_fragments++; 760 ipstat.ips_fragdropped++; 761 m_freem(m); 762 return (NULL); 763 } 764 765 ip = mtod(m, struct ip *); 766 hlen = ip->ip_hl << 2; 767 768 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 769 head = &ipq[hash]; 770 IPQ_LOCK(); 771 772 /* 773 * Look for queue of fragments 774 * of this datagram. 775 */ 776 TAILQ_FOREACH(fp, head, ipq_list) 777 if (ip->ip_id == fp->ipq_id && 778 ip->ip_src.s_addr == fp->ipq_src.s_addr && 779 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 780#ifdef MAC 781 mac_fragment_match(m, fp) && 782#endif 783 ip->ip_p == fp->ipq_p) 784 goto found; 785 786 fp = NULL; 787 788 /* 789 * Enforce upper bound on number of fragmented packets 790 * for which we attempt reassembly; 791 * If maxnipq is -1, accept all fragments without limitation. 792 */ 793 if ((nipq > maxnipq) && (maxnipq > 0)) { 794 /* 795 * drop something from the tail of the current queue 796 * before proceeding further 797 */ 798 struct ipq *q = TAILQ_LAST(head, ipqhead); 799 if (q == NULL) { /* gak */ 800 for (i = 0; i < IPREASS_NHASH; i++) { 801 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 802 if (r) { 803 ipstat.ips_fragtimeout += r->ipq_nfrags; 804 ip_freef(&ipq[i], r); 805 break; 806 } 807 } 808 } else { 809 ipstat.ips_fragtimeout += q->ipq_nfrags; 810 ip_freef(head, q); 811 } 812 } 813 814found: 815 /* 816 * Adjust ip_len to not reflect header, 817 * convert offset of this to bytes. 818 */ 819 ip->ip_len -= hlen; 820 if (ip->ip_off & IP_MF) { 821 /* 822 * Make sure that fragments have a data length 823 * that's a non-zero multiple of 8 bytes. 824 */ 825 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 826 ipstat.ips_toosmall++; /* XXX */ 827 goto dropfrag; 828 } 829 m->m_flags |= M_FRAG; 830 } else 831 m->m_flags &= ~M_FRAG; 832 ip->ip_off <<= 3; 833 834 835 /* 836 * Attempt reassembly; if it succeeds, proceed. 837 * ip_reass() will return a different mbuf. 838 */ 839 ipstat.ips_fragments++; 840 m->m_pkthdr.header = ip; 841 842 /* Previous ip_reass() started here. */ 843 /* 844 * Presence of header sizes in mbufs 845 * would confuse code below. 846 */ 847 m->m_data += hlen; 848 m->m_len -= hlen; 849 850 /* 851 * If first fragment to arrive, create a reassembly queue. 852 */ 853 if (fp == NULL) { 854 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 855 goto dropfrag; 856 fp = mtod(t, struct ipq *); 857#ifdef MAC 858 if (mac_init_ipq(fp, M_NOWAIT) != 0) { 859 m_free(t); 860 goto dropfrag; 861 } 862 mac_create_ipq(m, fp); 863#endif 864 TAILQ_INSERT_HEAD(head, fp, ipq_list); 865 nipq++; 866 fp->ipq_nfrags = 1; 867 fp->ipq_ttl = IPFRAGTTL; 868 fp->ipq_p = ip->ip_p; 869 fp->ipq_id = ip->ip_id; 870 fp->ipq_src = ip->ip_src; 871 fp->ipq_dst = ip->ip_dst; 872 fp->ipq_frags = m; 873 m->m_nextpkt = NULL; 874 goto inserted; 875 } else { 876 fp->ipq_nfrags++; 877#ifdef MAC 878 mac_update_ipq(m, fp); 879#endif 880 } 881 882#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 883 884 /* 885 * Handle ECN by comparing this segment with the first one; 886 * if CE is set, do not lose CE. 887 * drop if CE and not-ECT are mixed for the same packet. 888 */ 889 ecn = ip->ip_tos & IPTOS_ECN_MASK; 890 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 891 if (ecn == IPTOS_ECN_CE) { 892 if (ecn0 == IPTOS_ECN_NOTECT) 893 goto dropfrag; 894 if (ecn0 != IPTOS_ECN_CE) 895 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 896 } 897 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 898 goto dropfrag; 899 900 /* 901 * Find a segment which begins after this one does. 902 */ 903 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 904 if (GETIP(q)->ip_off > ip->ip_off) 905 break; 906 907 /* 908 * If there is a preceding segment, it may provide some of 909 * our data already. If so, drop the data from the incoming 910 * segment. If it provides all of our data, drop us, otherwise 911 * stick new segment in the proper place. 912 * 913 * If some of the data is dropped from the the preceding 914 * segment, then it's checksum is invalidated. 915 */ 916 if (p) { 917 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 918 if (i > 0) { 919 if (i >= ip->ip_len) 920 goto dropfrag; 921 m_adj(m, i); 922 m->m_pkthdr.csum_flags = 0; 923 ip->ip_off += i; 924 ip->ip_len -= i; 925 } 926 m->m_nextpkt = p->m_nextpkt; 927 p->m_nextpkt = m; 928 } else { 929 m->m_nextpkt = fp->ipq_frags; 930 fp->ipq_frags = m; 931 } 932 933 /* 934 * While we overlap succeeding segments trim them or, 935 * if they are completely covered, dequeue them. 936 */ 937 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 938 q = nq) { 939 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 940 if (i < GETIP(q)->ip_len) { 941 GETIP(q)->ip_len -= i; 942 GETIP(q)->ip_off += i; 943 m_adj(q, i); 944 q->m_pkthdr.csum_flags = 0; 945 break; 946 } 947 nq = q->m_nextpkt; 948 m->m_nextpkt = nq; 949 ipstat.ips_fragdropped++; 950 fp->ipq_nfrags--; 951 m_freem(q); 952 } 953 954inserted: 955 956 /* 957 * Check for complete reassembly and perform frag per packet 958 * limiting. 959 * 960 * Frag limiting is performed here so that the nth frag has 961 * a chance to complete the packet before we drop the packet. 962 * As a result, n+1 frags are actually allowed per packet, but 963 * only n will ever be stored. (n = maxfragsperpacket.) 964 * 965 */ 966 next = 0; 967 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 968 if (GETIP(q)->ip_off != next) { 969 if (fp->ipq_nfrags > maxfragsperpacket) { 970 ipstat.ips_fragdropped += fp->ipq_nfrags; 971 ip_freef(head, fp); 972 } 973 goto done; 974 } 975 next += GETIP(q)->ip_len; 976 } 977 /* Make sure the last packet didn't have the IP_MF flag */ 978 if (p->m_flags & M_FRAG) { 979 if (fp->ipq_nfrags > maxfragsperpacket) { 980 ipstat.ips_fragdropped += fp->ipq_nfrags; 981 ip_freef(head, fp); 982 } 983 goto done; 984 } 985 986 /* 987 * Reassembly is complete. Make sure the packet is a sane size. 988 */ 989 q = fp->ipq_frags; 990 ip = GETIP(q); 991 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 992 ipstat.ips_toolong++; 993 ipstat.ips_fragdropped += fp->ipq_nfrags; 994 ip_freef(head, fp); 995 goto done; 996 } 997 998 /* 999 * Concatenate fragments. 1000 */ 1001 m = q; 1002 t = m->m_next; 1003 m->m_next = 0; 1004 m_cat(m, t); 1005 nq = q->m_nextpkt; 1006 q->m_nextpkt = 0; 1007 for (q = nq; q != NULL; q = nq) { 1008 nq = q->m_nextpkt; 1009 q->m_nextpkt = NULL; 1010 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1011 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1012 m_cat(m, q); 1013 } 1014#ifdef MAC 1015 mac_create_datagram_from_ipq(fp, m); 1016 mac_destroy_ipq(fp); 1017#endif 1018 1019 /* 1020 * Create header for new ip packet by modifying header of first 1021 * packet; dequeue and discard fragment reassembly header. 1022 * Make header visible. 1023 */ 1024 ip->ip_len = (ip->ip_hl << 2) + next; 1025 ip->ip_src = fp->ipq_src; 1026 ip->ip_dst = fp->ipq_dst; 1027 TAILQ_REMOVE(head, fp, ipq_list); 1028 nipq--; 1029 (void) m_free(dtom(fp)); 1030 m->m_len += (ip->ip_hl << 2); 1031 m->m_data -= (ip->ip_hl << 2); 1032 /* some debugging cruft by sklower, below, will go away soon */ 1033 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1034 m_fixhdr(m); 1035 ipstat.ips_reassembled++; 1036 IPQ_UNLOCK(); 1037 return (m); 1038 1039dropfrag: 1040 ipstat.ips_fragdropped++; 1041 if (fp != NULL) 1042 fp->ipq_nfrags--; 1043 m_freem(m); 1044done: 1045 IPQ_UNLOCK(); 1046 return (NULL); 1047 1048#undef GETIP 1049} 1050 1051/* 1052 * Free a fragment reassembly header and all 1053 * associated datagrams. 1054 */ 1055static void 1056ip_freef(fhp, fp) 1057 struct ipqhead *fhp; 1058 struct ipq *fp; 1059{ 1060 register struct mbuf *q; 1061 1062 IPQ_LOCK_ASSERT(); 1063 1064 while (fp->ipq_frags) { 1065 q = fp->ipq_frags; 1066 fp->ipq_frags = q->m_nextpkt; 1067 m_freem(q); 1068 } 1069 TAILQ_REMOVE(fhp, fp, ipq_list); 1070 (void) m_free(dtom(fp)); 1071 nipq--; 1072} 1073 1074/* 1075 * IP timer processing; 1076 * if a timer expires on a reassembly 1077 * queue, discard it. 1078 */ 1079void 1080ip_slowtimo() 1081{ 1082 register struct ipq *fp; 1083 int s = splnet(); 1084 int i; 1085 1086 IPQ_LOCK(); 1087 for (i = 0; i < IPREASS_NHASH; i++) { 1088 for(fp = TAILQ_FIRST(&ipq[i]); fp;) { 1089 struct ipq *fpp; 1090 1091 fpp = fp; 1092 fp = TAILQ_NEXT(fp, ipq_list); 1093 if(--fpp->ipq_ttl == 0) { 1094 ipstat.ips_fragtimeout += fpp->ipq_nfrags; 1095 ip_freef(&ipq[i], fpp); 1096 } 1097 } 1098 } 1099 /* 1100 * If we are over the maximum number of fragments 1101 * (due to the limit being lowered), drain off 1102 * enough to get down to the new limit. 1103 */ 1104 if (maxnipq >= 0 && nipq > maxnipq) { 1105 for (i = 0; i < IPREASS_NHASH; i++) { 1106 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) { 1107 ipstat.ips_fragdropped += 1108 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1109 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1110 } 1111 } 1112 } 1113 IPQ_UNLOCK(); 1114 splx(s); 1115} 1116 1117/* 1118 * Drain off all datagram fragments. 1119 */ 1120void 1121ip_drain() 1122{ 1123 int i; 1124 1125 IPQ_LOCK(); 1126 for (i = 0; i < IPREASS_NHASH; i++) { 1127 while(!TAILQ_EMPTY(&ipq[i])) { 1128 ipstat.ips_fragdropped += 1129 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1130 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1131 } 1132 } 1133 IPQ_UNLOCK(); 1134 in_rtqdrain(); 1135} 1136 1137/* 1138 * Do option processing on a datagram, 1139 * possibly discarding it if bad options are encountered, 1140 * or forwarding it if source-routed. 1141 * The pass argument is used when operating in the IPSTEALTH 1142 * mode to tell what options to process: 1143 * [LS]SRR (pass 0) or the others (pass 1). 1144 * The reason for as many as two passes is that when doing IPSTEALTH, 1145 * non-routing options should be processed only if the packet is for us. 1146 * Returns 1 if packet has been forwarded/freed, 1147 * 0 if the packet should be processed further. 1148 */ 1149static int 1150ip_dooptions(struct mbuf *m, int pass) 1151{ 1152 struct ip *ip = mtod(m, struct ip *); 1153 u_char *cp; 1154 struct in_ifaddr *ia; 1155 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1156 struct in_addr *sin, dst; 1157 n_time ntime; 1158 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 1159 1160 /* ignore or reject packets with IP options */ 1161 if (ip_doopts == 0) 1162 return 0; 1163 else if (ip_doopts == 2) { 1164 type = ICMP_UNREACH; 1165 code = ICMP_UNREACH_FILTER_PROHIB; 1166 goto bad; 1167 } 1168 1169 dst = ip->ip_dst; 1170 cp = (u_char *)(ip + 1); 1171 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1172 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1173 opt = cp[IPOPT_OPTVAL]; 1174 if (opt == IPOPT_EOL) 1175 break; 1176 if (opt == IPOPT_NOP) 1177 optlen = 1; 1178 else { 1179 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1180 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1181 goto bad; 1182 } 1183 optlen = cp[IPOPT_OLEN]; 1184 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1185 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1186 goto bad; 1187 } 1188 } 1189 switch (opt) { 1190 1191 default: 1192 break; 1193 1194 /* 1195 * Source routing with record. 1196 * Find interface with current destination address. 1197 * If none on this machine then drop if strictly routed, 1198 * or do nothing if loosely routed. 1199 * Record interface address and bring up next address 1200 * component. If strictly routed make sure next 1201 * address is on directly accessible net. 1202 */ 1203 case IPOPT_LSRR: 1204 case IPOPT_SSRR: 1205#ifdef IPSTEALTH 1206 if (ipstealth && pass > 0) 1207 break; 1208#endif 1209 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1210 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1211 goto bad; 1212 } 1213 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1214 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1215 goto bad; 1216 } 1217 ipaddr.sin_addr = ip->ip_dst; 1218 ia = (struct in_ifaddr *) 1219 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1220 if (ia == NULL) { 1221 if (opt == IPOPT_SSRR) { 1222 type = ICMP_UNREACH; 1223 code = ICMP_UNREACH_SRCFAIL; 1224 goto bad; 1225 } 1226 if (!ip_dosourceroute) 1227 goto nosourcerouting; 1228 /* 1229 * Loose routing, and not at next destination 1230 * yet; nothing to do except forward. 1231 */ 1232 break; 1233 } 1234 off--; /* 0 origin */ 1235 if (off > optlen - (int)sizeof(struct in_addr)) { 1236 /* 1237 * End of source route. Should be for us. 1238 */ 1239 if (!ip_acceptsourceroute) 1240 goto nosourcerouting; 1241 save_rte(m, cp, ip->ip_src); 1242 break; 1243 } 1244#ifdef IPSTEALTH 1245 if (ipstealth) 1246 goto dropit; 1247#endif 1248 if (!ip_dosourceroute) { 1249 if (ipforwarding) { 1250 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1251 /* 1252 * Acting as a router, so generate ICMP 1253 */ 1254nosourcerouting: 1255 strcpy(buf, inet_ntoa(ip->ip_dst)); 1256 log(LOG_WARNING, 1257 "attempted source route from %s to %s\n", 1258 inet_ntoa(ip->ip_src), buf); 1259 type = ICMP_UNREACH; 1260 code = ICMP_UNREACH_SRCFAIL; 1261 goto bad; 1262 } else { 1263 /* 1264 * Not acting as a router, so silently drop. 1265 */ 1266#ifdef IPSTEALTH 1267dropit: 1268#endif 1269 ipstat.ips_cantforward++; 1270 m_freem(m); 1271 return (1); 1272 } 1273 } 1274 1275 /* 1276 * locate outgoing interface 1277 */ 1278 (void)memcpy(&ipaddr.sin_addr, cp + off, 1279 sizeof(ipaddr.sin_addr)); 1280 1281 if (opt == IPOPT_SSRR) { 1282#define INA struct in_ifaddr * 1283#define SA struct sockaddr * 1284 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == NULL) 1285 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1286 } else 1287 ia = ip_rtaddr(ipaddr.sin_addr); 1288 if (ia == NULL) { 1289 type = ICMP_UNREACH; 1290 code = ICMP_UNREACH_SRCFAIL; 1291 goto bad; 1292 } 1293 ip->ip_dst = ipaddr.sin_addr; 1294 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1295 sizeof(struct in_addr)); 1296 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1297 /* 1298 * Let ip_intr's mcast routing check handle mcast pkts 1299 */ 1300 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1301 break; 1302 1303 case IPOPT_RR: 1304#ifdef IPSTEALTH 1305 if (ipstealth && pass == 0) 1306 break; 1307#endif 1308 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1309 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1310 goto bad; 1311 } 1312 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1313 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1314 goto bad; 1315 } 1316 /* 1317 * If no space remains, ignore. 1318 */ 1319 off--; /* 0 origin */ 1320 if (off > optlen - (int)sizeof(struct in_addr)) 1321 break; 1322 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1323 sizeof(ipaddr.sin_addr)); 1324 /* 1325 * locate outgoing interface; if we're the destination, 1326 * use the incoming interface (should be same). 1327 */ 1328 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL && 1329 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1330 type = ICMP_UNREACH; 1331 code = ICMP_UNREACH_HOST; 1332 goto bad; 1333 } 1334 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1335 sizeof(struct in_addr)); 1336 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1337 break; 1338 1339 case IPOPT_TS: 1340#ifdef IPSTEALTH 1341 if (ipstealth && pass == 0) 1342 break; 1343#endif 1344 code = cp - (u_char *)ip; 1345 if (optlen < 4 || optlen > 40) { 1346 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1347 goto bad; 1348 } 1349 if ((off = cp[IPOPT_OFFSET]) < 5) { 1350 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1351 goto bad; 1352 } 1353 if (off > optlen - (int)sizeof(int32_t)) { 1354 cp[IPOPT_OFFSET + 1] += (1 << 4); 1355 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1356 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1357 goto bad; 1358 } 1359 break; 1360 } 1361 off--; /* 0 origin */ 1362 sin = (struct in_addr *)(cp + off); 1363 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1364 1365 case IPOPT_TS_TSONLY: 1366 break; 1367 1368 case IPOPT_TS_TSANDADDR: 1369 if (off + sizeof(n_time) + 1370 sizeof(struct in_addr) > optlen) { 1371 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1372 goto bad; 1373 } 1374 ipaddr.sin_addr = dst; 1375 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1376 m->m_pkthdr.rcvif); 1377 if (ia == NULL) 1378 continue; 1379 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1380 sizeof(struct in_addr)); 1381 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1382 off += sizeof(struct in_addr); 1383 break; 1384 1385 case IPOPT_TS_PRESPEC: 1386 if (off + sizeof(n_time) + 1387 sizeof(struct in_addr) > optlen) { 1388 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1389 goto bad; 1390 } 1391 (void)memcpy(&ipaddr.sin_addr, sin, 1392 sizeof(struct in_addr)); 1393 if (ifa_ifwithaddr((SA)&ipaddr) == NULL) 1394 continue; 1395 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1396 off += sizeof(struct in_addr); 1397 break; 1398 1399 default: 1400 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1401 goto bad; 1402 } 1403 ntime = iptime(); 1404 (void)memcpy(cp + off, &ntime, sizeof(n_time)); 1405 cp[IPOPT_OFFSET] += sizeof(n_time); 1406 } 1407 } 1408 if (forward && ipforwarding) { 1409 ip_forward(m, 1); 1410 return (1); 1411 } 1412 return (0); 1413bad: 1414 icmp_error(m, type, code, 0, 0); 1415 ipstat.ips_badoptions++; 1416 return (1); 1417} 1418 1419/* 1420 * Given address of next destination (final or next hop), 1421 * return internet address info of interface to be used to get there. 1422 */ 1423struct in_ifaddr * 1424ip_rtaddr(dst) 1425 struct in_addr dst; 1426{ 1427 struct route sro; 1428 struct sockaddr_in *sin; 1429 struct in_ifaddr *ifa; 1430 1431 bzero(&sro, sizeof(sro)); 1432 sin = (struct sockaddr_in *)&sro.ro_dst; 1433 sin->sin_family = AF_INET; 1434 sin->sin_len = sizeof(*sin); 1435 sin->sin_addr = dst; 1436 rtalloc_ign(&sro, RTF_CLONING); 1437 1438 if (sro.ro_rt == NULL) 1439 return ((struct in_ifaddr *)0); 1440 1441 ifa = ifatoia(sro.ro_rt->rt_ifa); 1442 RTFREE(sro.ro_rt); 1443 return ifa; 1444} 1445 1446/* 1447 * Save incoming source route for use in replies, 1448 * to be picked up later by ip_srcroute if the receiver is interested. 1449 */ 1450static void 1451save_rte(m, option, dst) 1452 struct mbuf *m; 1453 u_char *option; 1454 struct in_addr dst; 1455{ 1456 unsigned olen; 1457 struct ipopt_tag *opts; 1458 1459 opts = (struct ipopt_tag *)m_tag_get(PACKET_TAG_IPOPTIONS, 1460 sizeof(struct ipopt_tag), M_NOWAIT); 1461 if (opts == NULL) 1462 return; 1463 1464 olen = option[IPOPT_OLEN]; 1465#ifdef DIAGNOSTIC 1466 if (ipprintfs) 1467 printf("save_rte: olen %d\n", olen); 1468#endif 1469 if (olen > sizeof(opts->ip_srcrt) - (1 + sizeof(dst))) 1470 return; 1471 bcopy(option, opts->ip_srcrt.srcopt, olen); 1472 opts->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1473 opts->ip_srcrt.dst = dst; 1474 m_tag_prepend(m, (struct m_tag *)opts); 1475} 1476 1477/* 1478 * Retrieve incoming source route for use in replies, 1479 * in the same form used by setsockopt. 1480 * The first hop is placed before the options, will be removed later. 1481 */ 1482struct mbuf * 1483ip_srcroute(m0) 1484 struct mbuf *m0; 1485{ 1486 register struct in_addr *p, *q; 1487 register struct mbuf *m; 1488 struct ipopt_tag *opts; 1489 1490 opts = (struct ipopt_tag *)m_tag_find(m0, PACKET_TAG_IPOPTIONS, NULL); 1491 if (opts == NULL) 1492 return ((struct mbuf *)0); 1493 1494 if (opts->ip_nhops == 0) 1495 return ((struct mbuf *)0); 1496 m = m_get(M_DONTWAIT, MT_HEADER); 1497 if (m == NULL) 1498 return ((struct mbuf *)0); 1499 1500#define OPTSIZ (sizeof(opts->ip_srcrt.nop) + sizeof(opts->ip_srcrt.srcopt)) 1501 1502 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1503 m->m_len = opts->ip_nhops * sizeof(struct in_addr) + 1504 sizeof(struct in_addr) + OPTSIZ; 1505#ifdef DIAGNOSTIC 1506 if (ipprintfs) 1507 printf("ip_srcroute: nhops %d mlen %d", opts->ip_nhops, m->m_len); 1508#endif 1509 1510 /* 1511 * First save first hop for return route 1512 */ 1513 p = &(opts->ip_srcrt.route[opts->ip_nhops - 1]); 1514 *(mtod(m, struct in_addr *)) = *p--; 1515#ifdef DIAGNOSTIC 1516 if (ipprintfs) 1517 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1518#endif 1519 1520 /* 1521 * Copy option fields and padding (nop) to mbuf. 1522 */ 1523 opts->ip_srcrt.nop = IPOPT_NOP; 1524 opts->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1525 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1526 &(opts->ip_srcrt.nop), OPTSIZ); 1527 q = (struct in_addr *)(mtod(m, caddr_t) + 1528 sizeof(struct in_addr) + OPTSIZ); 1529#undef OPTSIZ 1530 /* 1531 * Record return path as an IP source route, 1532 * reversing the path (pointers are now aligned). 1533 */ 1534 while (p >= opts->ip_srcrt.route) { 1535#ifdef DIAGNOSTIC 1536 if (ipprintfs) 1537 printf(" %lx", (u_long)ntohl(q->s_addr)); 1538#endif 1539 *q++ = *p--; 1540 } 1541 /* 1542 * Last hop goes to final destination. 1543 */ 1544 *q = opts->ip_srcrt.dst; 1545#ifdef DIAGNOSTIC 1546 if (ipprintfs) 1547 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1548#endif 1549 m_tag_delete(m0, (struct m_tag *)opts); 1550 return (m); 1551} 1552 1553/* 1554 * Strip out IP options, at higher 1555 * level protocol in the kernel. 1556 * Second argument is buffer to which options 1557 * will be moved, and return value is their length. 1558 * XXX should be deleted; last arg currently ignored. 1559 */ 1560void 1561ip_stripoptions(m, mopt) 1562 register struct mbuf *m; 1563 struct mbuf *mopt; 1564{ 1565 register int i; 1566 struct ip *ip = mtod(m, struct ip *); 1567 register caddr_t opts; 1568 int olen; 1569 1570 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1571 opts = (caddr_t)(ip + 1); 1572 i = m->m_len - (sizeof (struct ip) + olen); 1573 bcopy(opts + olen, opts, (unsigned)i); 1574 m->m_len -= olen; 1575 if (m->m_flags & M_PKTHDR) 1576 m->m_pkthdr.len -= olen; 1577 ip->ip_v = IPVERSION; 1578 ip->ip_hl = sizeof(struct ip) >> 2; 1579} 1580 1581u_char inetctlerrmap[PRC_NCMDS] = { 1582 0, 0, 0, 0, 1583 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1584 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1585 EMSGSIZE, EHOSTUNREACH, 0, 0, 1586 0, 0, EHOSTUNREACH, 0, 1587 ENOPROTOOPT, ECONNREFUSED 1588}; 1589 1590/* 1591 * Forward a packet. If some error occurs return the sender 1592 * an icmp packet. Note we can't always generate a meaningful 1593 * icmp message because icmp doesn't have a large enough repertoire 1594 * of codes and types. 1595 * 1596 * If not forwarding, just drop the packet. This could be confusing 1597 * if ipforwarding was zero but some routing protocol was advancing 1598 * us as a gateway to somewhere. However, we must let the routing 1599 * protocol deal with that. 1600 * 1601 * The srcrt parameter indicates whether the packet is being forwarded 1602 * via a source route. 1603 */ 1604void 1605ip_forward(struct mbuf *m, int srcrt) 1606{ 1607 struct ip *ip = mtod(m, struct ip *); 1608 struct in_ifaddr *ia = NULL; 1609 int error, type = 0, code = 0; 1610 struct mbuf *mcopy; 1611 struct in_addr dest; 1612 struct ifnet *destifp, dummyifp; 1613 1614#ifdef DIAGNOSTIC 1615 if (ipprintfs) 1616 printf("forward: src %lx dst %lx ttl %x\n", 1617 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr, 1618 ip->ip_ttl); 1619#endif 1620 1621 1622 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1623 ipstat.ips_cantforward++; 1624 m_freem(m); 1625 return; 1626 } 1627#ifdef IPSTEALTH 1628 if (!ipstealth) { 1629#endif 1630 if (ip->ip_ttl <= IPTTLDEC) { 1631 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1632 0, 0); 1633 return; 1634 } 1635#ifdef IPSTEALTH 1636 } 1637#endif 1638 1639 if (!srcrt && (ia = ip_rtaddr(ip->ip_dst)) == NULL) { 1640 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1641 return; 1642 } 1643 1644 /* 1645 * Save the IP header and at most 8 bytes of the payload, 1646 * in case we need to generate an ICMP message to the src. 1647 * 1648 * XXX this can be optimized a lot by saving the data in a local 1649 * buffer on the stack (72 bytes at most), and only allocating the 1650 * mbuf if really necessary. The vast majority of the packets 1651 * are forwarded without having to send an ICMP back (either 1652 * because unnecessary, or because rate limited), so we are 1653 * really we are wasting a lot of work here. 1654 * 1655 * We don't use m_copy() because it might return a reference 1656 * to a shared cluster. Both this function and ip_output() 1657 * assume exclusive access to the IP header in `m', so any 1658 * data in a cluster may change before we reach icmp_error(). 1659 */ 1660 MGET(mcopy, M_DONTWAIT, m->m_type); 1661 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1662 /* 1663 * It's probably ok if the pkthdr dup fails (because 1664 * the deep copy of the tag chain failed), but for now 1665 * be conservative and just discard the copy since 1666 * code below may some day want the tags. 1667 */ 1668 m_free(mcopy); 1669 mcopy = NULL; 1670 } 1671 if (mcopy != NULL) { 1672 mcopy->m_len = imin((ip->ip_hl << 2) + 8, 1673 (int)ip->ip_len); 1674 mcopy->m_pkthdr.len = mcopy->m_len; 1675 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1676 } 1677 1678#ifdef IPSTEALTH 1679 if (!ipstealth) { 1680#endif 1681 ip->ip_ttl -= IPTTLDEC; 1682#ifdef IPSTEALTH 1683 } 1684#endif 1685 1686 /* 1687 * If forwarding packet using same interface that it came in on, 1688 * perhaps should send a redirect to sender to shortcut a hop. 1689 * Only send redirect if source is sending directly to us, 1690 * and if packet was not source routed (or has any options). 1691 * Also, don't send redirect if forwarding using a default route 1692 * or a route modified by a redirect. 1693 */ 1694 dest.s_addr = 0; 1695 if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) { 1696 struct sockaddr_in *sin; 1697 struct route ro; 1698 struct rtentry *rt; 1699 1700 bzero(&ro, sizeof(ro)); 1701 sin = (struct sockaddr_in *)&ro.ro_dst; 1702 sin->sin_family = AF_INET; 1703 sin->sin_len = sizeof(*sin); 1704 sin->sin_addr = ip->ip_dst; 1705 rtalloc_ign(&ro, RTF_CLONING); 1706 1707 rt = ro.ro_rt; 1708 1709 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1710 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1711#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1712 u_long src = ntohl(ip->ip_src.s_addr); 1713 1714 if (RTA(rt) && 1715 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1716 if (rt->rt_flags & RTF_GATEWAY) 1717 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1718 else 1719 dest.s_addr = ip->ip_dst.s_addr; 1720 /* Router requirements says to only send host redirects */ 1721 type = ICMP_REDIRECT; 1722 code = ICMP_REDIRECT_HOST; 1723#ifdef DIAGNOSTIC 1724 if (ipprintfs) 1725 printf("redirect (%d) to %lx\n", code, (u_long)dest.s_addr); 1726#endif 1727 } 1728 } 1729 if (rt) 1730 RTFREE(rt); 1731 } 1732 1733 error = ip_output(m, (struct mbuf *)0, NULL, IP_FORWARDING, 0, NULL); 1734 if (error) 1735 ipstat.ips_cantforward++; 1736 else { 1737 ipstat.ips_forward++; 1738 if (type) 1739 ipstat.ips_redirectsent++; 1740 else { 1741 if (mcopy) 1742 m_freem(mcopy); 1743 return; 1744 } 1745 } 1746 if (mcopy == NULL) 1747 return; 1748 destifp = NULL; 1749 1750 switch (error) { 1751 1752 case 0: /* forwarded, but need redirect */ 1753 /* type, code set above */ 1754 break; 1755 1756 case ENETUNREACH: /* shouldn't happen, checked above */ 1757 case EHOSTUNREACH: 1758 case ENETDOWN: 1759 case EHOSTDOWN: 1760 default: 1761 type = ICMP_UNREACH; 1762 code = ICMP_UNREACH_HOST; 1763 break; 1764 1765 case EMSGSIZE: 1766 type = ICMP_UNREACH; 1767 code = ICMP_UNREACH_NEEDFRAG; 1768#if defined(IPSEC) || defined(FAST_IPSEC) 1769 /* 1770 * If the packet is routed over IPsec tunnel, tell the 1771 * originator the tunnel MTU. 1772 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1773 * XXX quickhack!!! 1774 */ 1775 { 1776 struct secpolicy *sp = NULL; 1777 int ipsecerror; 1778 int ipsechdr; 1779 struct route *ro; 1780 1781#ifdef IPSEC 1782 sp = ipsec4_getpolicybyaddr(mcopy, 1783 IPSEC_DIR_OUTBOUND, 1784 IP_FORWARDING, 1785 &ipsecerror); 1786#else /* FAST_IPSEC */ 1787 sp = ipsec_getpolicybyaddr(mcopy, 1788 IPSEC_DIR_OUTBOUND, 1789 IP_FORWARDING, 1790 &ipsecerror); 1791#endif 1792 if (sp != NULL) { 1793 /* count IPsec header size */ 1794 ipsechdr = ipsec4_hdrsiz(mcopy, 1795 IPSEC_DIR_OUTBOUND, 1796 NULL); 1797 1798 /* 1799 * find the correct route for outer IPv4 1800 * header, compute tunnel MTU. 1801 * 1802 * XXX BUG ALERT 1803 * The "dummyifp" code relies upon the fact 1804 * that icmp_error() touches only ifp->if_mtu. 1805 */ 1806 /*XXX*/ 1807 destifp = NULL; 1808 if (sp->req != NULL 1809 && sp->req->sav != NULL 1810 && sp->req->sav->sah != NULL) { 1811 ro = &sp->req->sav->sah->sa_route; 1812 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1813 dummyifp.if_mtu = 1814 ro->ro_rt->rt_rmx.rmx_mtu ? 1815 ro->ro_rt->rt_rmx.rmx_mtu : 1816 ro->ro_rt->rt_ifp->if_mtu; 1817 dummyifp.if_mtu -= ipsechdr; 1818 destifp = &dummyifp; 1819 } 1820 } 1821 1822#ifdef IPSEC 1823 key_freesp(sp); 1824#else /* FAST_IPSEC */ 1825 KEY_FREESP(&sp); 1826#endif 1827 ipstat.ips_cantfrag++; 1828 break; 1829 } else 1830#endif /*IPSEC || FAST_IPSEC*/ 1831 /* 1832 * When doing source routing 'ia' can be NULL. Fall back 1833 * to the minimum guaranteed routeable packet size and use 1834 * the same hack as IPSEC to setup a dummyifp for icmp. 1835 */ 1836 if (ia == NULL) { 1837 dummyifp.if_mtu = IP_MSS; 1838 destifp = &dummyifp; 1839 } else 1840 destifp = ia->ia_ifp; 1841#if defined(IPSEC) || defined(FAST_IPSEC) 1842 } 1843#endif /*IPSEC || FAST_IPSEC*/ 1844 ipstat.ips_cantfrag++; 1845 break; 1846 1847 case ENOBUFS: 1848 /* 1849 * A router should not generate ICMP_SOURCEQUENCH as 1850 * required in RFC1812 Requirements for IP Version 4 Routers. 1851 * Source quench could be a big problem under DoS attacks, 1852 * or if the underlying interface is rate-limited. 1853 * Those who need source quench packets may re-enable them 1854 * via the net.inet.ip.sendsourcequench sysctl. 1855 */ 1856 if (ip_sendsourcequench == 0) { 1857 m_freem(mcopy); 1858 return; 1859 } else { 1860 type = ICMP_SOURCEQUENCH; 1861 code = 0; 1862 } 1863 break; 1864 1865 case EACCES: /* ipfw denied packet */ 1866 m_freem(mcopy); 1867 return; 1868 } 1869 icmp_error(mcopy, type, code, dest.s_addr, destifp); 1870} 1871 1872void 1873ip_savecontrol(inp, mp, ip, m) 1874 register struct inpcb *inp; 1875 register struct mbuf **mp; 1876 register struct ip *ip; 1877 register struct mbuf *m; 1878{ 1879 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1880 struct bintime bt; 1881 1882 bintime(&bt); 1883 if (inp->inp_socket->so_options & SO_BINTIME) { 1884 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1885 SCM_BINTIME, SOL_SOCKET); 1886 if (*mp) 1887 mp = &(*mp)->m_next; 1888 } 1889 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1890 struct timeval tv; 1891 1892 bintime2timeval(&bt, &tv); 1893 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1894 SCM_TIMESTAMP, SOL_SOCKET); 1895 if (*mp) 1896 mp = &(*mp)->m_next; 1897 } 1898 } 1899 if (inp->inp_flags & INP_RECVDSTADDR) { 1900 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1901 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1902 if (*mp) 1903 mp = &(*mp)->m_next; 1904 } 1905 if (inp->inp_flags & INP_RECVTTL) { 1906 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1907 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1908 if (*mp) 1909 mp = &(*mp)->m_next; 1910 } 1911#ifdef notyet 1912 /* XXX 1913 * Moving these out of udp_input() made them even more broken 1914 * than they already were. 1915 */ 1916 /* options were tossed already */ 1917 if (inp->inp_flags & INP_RECVOPTS) { 1918 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1919 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1920 if (*mp) 1921 mp = &(*mp)->m_next; 1922 } 1923 /* ip_srcroute doesn't do what we want here, need to fix */ 1924 if (inp->inp_flags & INP_RECVRETOPTS) { 1925 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1926 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1927 if (*mp) 1928 mp = &(*mp)->m_next; 1929 } 1930#endif 1931 if (inp->inp_flags & INP_RECVIF) { 1932 struct ifnet *ifp; 1933 struct sdlbuf { 1934 struct sockaddr_dl sdl; 1935 u_char pad[32]; 1936 } sdlbuf; 1937 struct sockaddr_dl *sdp; 1938 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1939 1940 if (((ifp = m->m_pkthdr.rcvif)) 1941 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1942 sdp = (struct sockaddr_dl *) 1943 (ifaddr_byindex(ifp->if_index)->ifa_addr); 1944 /* 1945 * Change our mind and don't try copy. 1946 */ 1947 if ((sdp->sdl_family != AF_LINK) 1948 || (sdp->sdl_len > sizeof(sdlbuf))) { 1949 goto makedummy; 1950 } 1951 bcopy(sdp, sdl2, sdp->sdl_len); 1952 } else { 1953makedummy: 1954 sdl2->sdl_len 1955 = offsetof(struct sockaddr_dl, sdl_data[0]); 1956 sdl2->sdl_family = AF_LINK; 1957 sdl2->sdl_index = 0; 1958 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1959 } 1960 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1961 IP_RECVIF, IPPROTO_IP); 1962 if (*mp) 1963 mp = &(*mp)->m_next; 1964 } 1965} 1966 1967/* 1968 * XXX these routines are called from the upper part of the kernel. 1969 * They need to be locked when we remove Giant. 1970 * 1971 * They could also be moved to ip_mroute.c, since all the RSVP 1972 * handling is done there already. 1973 */ 1974static int ip_rsvp_on; 1975struct socket *ip_rsvpd; 1976int 1977ip_rsvp_init(struct socket *so) 1978{ 1979 if (so->so_type != SOCK_RAW || 1980 so->so_proto->pr_protocol != IPPROTO_RSVP) 1981 return EOPNOTSUPP; 1982 1983 if (ip_rsvpd != NULL) 1984 return EADDRINUSE; 1985 1986 ip_rsvpd = so; 1987 /* 1988 * This may seem silly, but we need to be sure we don't over-increment 1989 * the RSVP counter, in case something slips up. 1990 */ 1991 if (!ip_rsvp_on) { 1992 ip_rsvp_on = 1; 1993 rsvp_on++; 1994 } 1995 1996 return 0; 1997} 1998 1999int 2000ip_rsvp_done(void) 2001{ 2002 ip_rsvpd = NULL; 2003 /* 2004 * This may seem silly, but we need to be sure we don't over-decrement 2005 * the RSVP counter, in case something slips up. 2006 */ 2007 if (ip_rsvp_on) { 2008 ip_rsvp_on = 0; 2009 rsvp_on--; 2010 } 2011 return 0; 2012} 2013 2014void 2015rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 2016{ 2017 if (rsvp_input_p) { /* call the real one if loaded */ 2018 rsvp_input_p(m, off); 2019 return; 2020 } 2021 2022 /* Can still get packets with rsvp_on = 0 if there is a local member 2023 * of the group to which the RSVP packet is addressed. But in this 2024 * case we want to throw the packet away. 2025 */ 2026 2027 if (!rsvp_on) { 2028 m_freem(m); 2029 return; 2030 } 2031 2032 if (ip_rsvpd != NULL) { 2033 rip_input(m, off); 2034 return; 2035 } 2036 /* Drop the packet */ 2037 m_freem(m); 2038} 2039