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