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