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 * XXX: Broken on SMP and possibly preemption! 229 */ 230 static int ip_nhops = 0; 231 static struct ip_srcrt { 232 struct in_addr dst; /* final destination */ 233 char nop; /* one NOP to align */ 234 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 235 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 236 } ip_srcrt; 237 238 static void save_rte(u_char *, struct in_addr); 239 static int ip_dooptions(struct mbuf *m, int, 240 struct sockaddr_in *next_hop); 241 static void ip_forward(struct mbuf *m, int srcrt, 242 struct sockaddr_in *next_hop); 243 static void ip_freef(struct ipqhead *, struct ipq *); 244 static struct mbuf *ip_reass(struct mbuf *); 245 246 /* 247 * IP initialization: fill in IP protocol switch table. 248 * All protocols not implemented in kernel go to raw IP protocol handler. 249 */ 250 void 251 ip_init() 252 { 253 register struct protosw *pr; 254 register int i; 255 256 TAILQ_INIT(&in_ifaddrhead); 257 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 258 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 259 if (pr == 0) 260 panic("ip_init"); 261 for (i = 0; i < IPPROTO_MAX; i++) 262 ip_protox[i] = pr - inetsw; 263 for (pr = inetdomain.dom_protosw; 264 pr < inetdomain.dom_protoswNPROTOSW; pr++) 265 if (pr->pr_domain->dom_family == PF_INET && 266 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 267 ip_protox[pr->pr_protocol] = pr - inetsw; 268 269 #ifdef PFIL_HOOKS 270 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 271 inet_pfil_hook.ph_af = AF_INET; 272 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 273 printf("%s: WARNING: unable to register pfil hook, " 274 "error %d\n", __func__, i); 275 #endif /* PFIL_HOOKS */ 276 277 IPQ_LOCK_INIT(); 278 for (i = 0; i < IPREASS_NHASH; i++) 279 TAILQ_INIT(&ipq[i]); 280 281 maxnipq = nmbclusters / 32; 282 maxfragsperpacket = 16; 283 284 #ifndef RANDOM_IP_ID 285 ip_id = time_second & 0xffff; 286 #endif 287 ipintrq.ifq_maxlen = ipqmaxlen; 288 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF); 289 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE); 290 } 291 292 /* 293 * Ip input routine. Checksum and byte swap header. If fragmented 294 * try to reassemble. Process options. Pass to next level. 295 */ 296 void 297 ip_input(struct mbuf *m) 298 { 299 struct ip *ip = NULL; 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))) == NULL) { 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)) == NULL) { 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 != NULL && 591 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 592 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 593 if (ifa->ifa_addr->sa_family != AF_INET) 594 continue; 595 ia = ifatoia(ifa); 596 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 597 pkt_dst.s_addr) 598 goto ours; 599 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 600 goto ours; 601 #ifdef BOOTP_COMPAT 602 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 603 goto ours; 604 #endif 605 } 606 } 607 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 608 struct in_multi *inm; 609 if (ip_mrouter) { 610 /* 611 * If we are acting as a multicast router, all 612 * incoming multicast packets are passed to the 613 * kernel-level multicast forwarding function. 614 * The packet is returned (relatively) intact; if 615 * ip_mforward() returns a non-zero value, the packet 616 * must be discarded, else it may be accepted below. 617 */ 618 if (ip_mforward && 619 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 620 ipstat.ips_cantforward++; 621 m_freem(m); 622 return; 623 } 624 625 /* 626 * The process-level routing daemon needs to receive 627 * all multicast IGMP packets, whether or not this 628 * host belongs to their destination groups. 629 */ 630 if (ip->ip_p == IPPROTO_IGMP) 631 goto ours; 632 ipstat.ips_forward++; 633 } 634 /* 635 * See if we belong to the destination multicast group on the 636 * arrival interface. 637 */ 638 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 639 if (inm == NULL) { 640 ipstat.ips_notmember++; 641 m_freem(m); 642 return; 643 } 644 goto ours; 645 } 646 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 647 goto ours; 648 if (ip->ip_dst.s_addr == INADDR_ANY) 649 goto ours; 650 651 /* 652 * FAITH(Firewall Aided Internet Translator) 653 */ 654 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 655 if (ip_keepfaith) { 656 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 657 goto ours; 658 } 659 m_freem(m); 660 return; 661 } 662 663 /* 664 * Not for us; forward if possible and desirable. 665 */ 666 if (ipforwarding == 0) { 667 ipstat.ips_cantforward++; 668 m_freem(m); 669 } else { 670 #ifdef IPSEC 671 /* 672 * Enforce inbound IPsec SPD. 673 */ 674 if (ipsec4_in_reject(m, NULL)) { 675 ipsecstat.in_polvio++; 676 goto bad; 677 } 678 #endif /* IPSEC */ 679 #ifdef FAST_IPSEC 680 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 681 s = splnet(); 682 if (mtag != NULL) { 683 tdbi = (struct tdb_ident *)(mtag + 1); 684 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 685 } else { 686 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 687 IP_FORWARDING, &error); 688 } 689 if (sp == NULL) { /* NB: can happen if error */ 690 splx(s); 691 /*XXX error stat???*/ 692 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 693 goto bad; 694 } 695 696 /* 697 * Check security policy against packet attributes. 698 */ 699 error = ipsec_in_reject(sp, m); 700 KEY_FREESP(&sp); 701 splx(s); 702 if (error) { 703 ipstat.ips_cantforward++; 704 goto bad; 705 } 706 #endif /* FAST_IPSEC */ 707 ip_forward(m, dchg, args.next_hop); 708 } 709 return; 710 711 ours: 712 #ifdef IPSTEALTH 713 /* 714 * IPSTEALTH: Process non-routing options only 715 * if the packet is destined for us. 716 */ 717 if (ipstealth && hlen > sizeof (struct ip) && 718 ip_dooptions(m, 1, args.next_hop)) 719 return; 720 #endif /* IPSTEALTH */ 721 722 /* Count the packet in the ip address stats */ 723 if (ia != NULL) { 724 ia->ia_ifa.if_ipackets++; 725 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 726 } 727 728 /* 729 * Attempt reassembly; if it succeeds, proceed. 730 * ip_reass() will return a different mbuf. 731 */ 732 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 733 m = ip_reass(m); 734 if (m == NULL) 735 return; 736 ip = mtod(m, struct ip *); 737 /* Get the header length of the reassembled packet */ 738 hlen = ip->ip_hl << 2; 739 #ifdef IPDIVERT 740 /* Restore original checksum before diverting packet */ 741 if (divert_find_info(m) != 0) { 742 ip->ip_len = htons(ip->ip_len); 743 ip->ip_off = htons(ip->ip_off); 744 ip->ip_sum = 0; 745 if (hlen == sizeof(struct ip)) 746 ip->ip_sum = in_cksum_hdr(ip); 747 else 748 ip->ip_sum = in_cksum(m, hlen); 749 ip->ip_off = ntohs(ip->ip_off); 750 ip->ip_len = ntohs(ip->ip_len); 751 } 752 #endif 753 } 754 755 /* 756 * Further protocols expect the packet length to be w/o the 757 * IP header. 758 */ 759 ip->ip_len -= hlen; 760 761 #ifdef IPDIVERT 762 /* 763 * Divert or tee packet to the divert protocol if required. 764 */ 765 divert_info = divert_find_info(m); 766 if (divert_info != 0) { 767 struct mbuf *clone; 768 769 /* Clone packet if we're doing a 'tee' */ 770 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 771 clone = divert_clone(m); 772 else 773 clone = NULL; 774 775 /* Restore packet header fields to original values */ 776 ip->ip_len += hlen; 777 ip->ip_len = htons(ip->ip_len); 778 ip->ip_off = htons(ip->ip_off); 779 780 /* Deliver packet to divert input routine */ 781 divert_packet(m, 1); 782 ipstat.ips_delivered++; 783 784 /* If 'tee', continue with original packet */ 785 if (clone == NULL) 786 return; 787 m = clone; 788 ip = mtod(m, struct ip *); 789 ip->ip_len += hlen; 790 /* 791 * Jump backwards to complete processing of the 792 * packet. We do not need to clear args.next_hop 793 * as that will not be used again and the cloned packet 794 * doesn't contain a divert packet tag so we won't 795 * re-entry this block. 796 */ 797 goto pass; 798 } 799 #endif 800 801 #ifdef IPSEC 802 /* 803 * enforce IPsec policy checking if we are seeing last header. 804 * note that we do not visit this with protocols with pcb layer 805 * code - like udp/tcp/raw ip. 806 */ 807 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 808 ipsec4_in_reject(m, NULL)) { 809 ipsecstat.in_polvio++; 810 goto bad; 811 } 812 #endif 813 #if FAST_IPSEC 814 /* 815 * enforce IPsec policy checking if we are seeing last header. 816 * note that we do not visit this with protocols with pcb layer 817 * code - like udp/tcp/raw ip. 818 */ 819 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 820 /* 821 * Check if the packet has already had IPsec processing 822 * done. If so, then just pass it along. This tag gets 823 * set during AH, ESP, etc. input handling, before the 824 * packet is returned to the ip input queue for delivery. 825 */ 826 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 827 s = splnet(); 828 if (mtag != NULL) { 829 tdbi = (struct tdb_ident *)(mtag + 1); 830 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 831 } else { 832 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 833 IP_FORWARDING, &error); 834 } 835 if (sp != NULL) { 836 /* 837 * Check security policy against packet attributes. 838 */ 839 error = ipsec_in_reject(sp, m); 840 KEY_FREESP(&sp); 841 } else { 842 /* XXX error stat??? */ 843 error = EINVAL; 844 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 845 goto bad; 846 } 847 splx(s); 848 if (error) 849 goto bad; 850 } 851 #endif /* FAST_IPSEC */ 852 853 /* 854 * Switch out to protocol's input routine. 855 */ 856 ipstat.ips_delivered++; 857 if (args.next_hop && ip->ip_p == IPPROTO_TCP) { 858 /* attach next hop info for TCP */ 859 struct m_tag *mtag = m_tag_get(PACKET_TAG_IPFORWARD, 860 sizeof(struct sockaddr_in *), M_NOWAIT); 861 if (mtag == NULL) 862 goto bad; 863 *(struct sockaddr_in **)(mtag+1) = args.next_hop; 864 m_tag_prepend(m, mtag); 865 } 866 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 867 return; 868 bad: 869 m_freem(m); 870 } 871 872 /* 873 * Take incoming datagram fragment and try to reassemble it into 874 * whole datagram. If the argument is the first fragment or one 875 * in between the function will return NULL and store the mbuf 876 * in the fragment chain. If the argument is the last fragment 877 * the packet will be reassembled and the pointer to the new 878 * mbuf returned for further processing. Only m_tags attached 879 * to the first packet/fragment are preserved. 880 * The IP header is *NOT* adjusted out of iplen. 881 */ 882 883 struct mbuf * 884 ip_reass(struct mbuf *m) 885 { 886 struct ip *ip; 887 struct mbuf *p, *q, *nq, *t; 888 struct ipq *fp = NULL; 889 struct ipqhead *head; 890 int i, hlen, next; 891 u_int8_t ecn, ecn0; 892 u_short hash; 893 894 /* If maxnipq is 0, never accept fragments. */ 895 if (maxnipq == 0) { 896 ipstat.ips_fragments++; 897 ipstat.ips_fragdropped++; 898 m_freem(m); 899 return (NULL); 900 } 901 902 ip = mtod(m, struct ip *); 903 hlen = ip->ip_hl << 2; 904 905 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 906 head = &ipq[hash]; 907 IPQ_LOCK(); 908 909 /* 910 * Look for queue of fragments 911 * of this datagram. 912 */ 913 TAILQ_FOREACH(fp, head, ipq_list) 914 if (ip->ip_id == fp->ipq_id && 915 ip->ip_src.s_addr == fp->ipq_src.s_addr && 916 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 917 #ifdef MAC 918 mac_fragment_match(m, fp) && 919 #endif 920 ip->ip_p == fp->ipq_p) 921 goto found; 922 923 fp = NULL; 924 925 /* 926 * Enforce upper bound on number of fragmented packets 927 * for which we attempt reassembly; 928 * If maxnipq is -1, accept all fragments without limitation. 929 */ 930 if ((nipq > maxnipq) && (maxnipq > 0)) { 931 /* 932 * drop something from the tail of the current queue 933 * before proceeding further 934 */ 935 struct ipq *q = TAILQ_LAST(head, ipqhead); 936 if (q == NULL) { /* gak */ 937 for (i = 0; i < IPREASS_NHASH; i++) { 938 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 939 if (r) { 940 ipstat.ips_fragtimeout += r->ipq_nfrags; 941 ip_freef(&ipq[i], r); 942 break; 943 } 944 } 945 } else { 946 ipstat.ips_fragtimeout += q->ipq_nfrags; 947 ip_freef(head, q); 948 } 949 } 950 951 found: 952 /* 953 * Adjust ip_len to not reflect header, 954 * convert offset of this to bytes. 955 */ 956 ip->ip_len -= hlen; 957 if (ip->ip_off & IP_MF) { 958 /* 959 * Make sure that fragments have a data length 960 * that's a non-zero multiple of 8 bytes. 961 */ 962 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 963 ipstat.ips_toosmall++; /* XXX */ 964 goto dropfrag; 965 } 966 m->m_flags |= M_FRAG; 967 } else 968 m->m_flags &= ~M_FRAG; 969 ip->ip_off <<= 3; 970 971 972 /* 973 * Attempt reassembly; if it succeeds, proceed. 974 * ip_reass() will return a different mbuf. 975 */ 976 ipstat.ips_fragments++; 977 m->m_pkthdr.header = ip; 978 979 /* Previous ip_reass() started here. */ 980 /* 981 * Presence of header sizes in mbufs 982 * would confuse code below. 983 */ 984 m->m_data += hlen; 985 m->m_len -= hlen; 986 987 /* 988 * If first fragment to arrive, create a reassembly queue. 989 */ 990 if (fp == NULL) { 991 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 992 goto dropfrag; 993 fp = mtod(t, struct ipq *); 994 #ifdef MAC 995 if (mac_init_ipq(fp, M_NOWAIT) != 0) { 996 m_free(t); 997 goto dropfrag; 998 } 999 mac_create_ipq(m, fp); 1000 #endif 1001 TAILQ_INSERT_HEAD(head, fp, ipq_list); 1002 nipq++; 1003 fp->ipq_nfrags = 1; 1004 fp->ipq_ttl = IPFRAGTTL; 1005 fp->ipq_p = ip->ip_p; 1006 fp->ipq_id = ip->ip_id; 1007 fp->ipq_src = ip->ip_src; 1008 fp->ipq_dst = ip->ip_dst; 1009 fp->ipq_frags = m; 1010 m->m_nextpkt = NULL; 1011 goto inserted; 1012 } else { 1013 fp->ipq_nfrags++; 1014 #ifdef MAC 1015 mac_update_ipq(m, fp); 1016 #endif 1017 } 1018 1019 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1020 1021 /* 1022 * Handle ECN by comparing this segment with the first one; 1023 * if CE is set, do not lose CE. 1024 * drop if CE and not-ECT are mixed for the same packet. 1025 */ 1026 ecn = ip->ip_tos & IPTOS_ECN_MASK; 1027 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 1028 if (ecn == IPTOS_ECN_CE) { 1029 if (ecn0 == IPTOS_ECN_NOTECT) 1030 goto dropfrag; 1031 if (ecn0 != IPTOS_ECN_CE) 1032 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 1033 } 1034 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 1035 goto dropfrag; 1036 1037 /* 1038 * Find a segment which begins after this one does. 1039 */ 1040 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1041 if (GETIP(q)->ip_off > ip->ip_off) 1042 break; 1043 1044 /* 1045 * If there is a preceding segment, it may provide some of 1046 * our data already. If so, drop the data from the incoming 1047 * segment. If it provides all of our data, drop us, otherwise 1048 * stick new segment in the proper place. 1049 * 1050 * If some of the data is dropped from the the preceding 1051 * segment, then it's checksum is invalidated. 1052 */ 1053 if (p) { 1054 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1055 if (i > 0) { 1056 if (i >= ip->ip_len) 1057 goto dropfrag; 1058 m_adj(m, i); 1059 m->m_pkthdr.csum_flags = 0; 1060 ip->ip_off += i; 1061 ip->ip_len -= i; 1062 } 1063 m->m_nextpkt = p->m_nextpkt; 1064 p->m_nextpkt = m; 1065 } else { 1066 m->m_nextpkt = fp->ipq_frags; 1067 fp->ipq_frags = m; 1068 } 1069 1070 /* 1071 * While we overlap succeeding segments trim them or, 1072 * if they are completely covered, dequeue them. 1073 */ 1074 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1075 q = nq) { 1076 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1077 if (i < GETIP(q)->ip_len) { 1078 GETIP(q)->ip_len -= i; 1079 GETIP(q)->ip_off += i; 1080 m_adj(q, i); 1081 q->m_pkthdr.csum_flags = 0; 1082 break; 1083 } 1084 nq = q->m_nextpkt; 1085 m->m_nextpkt = nq; 1086 ipstat.ips_fragdropped++; 1087 fp->ipq_nfrags--; 1088 m_freem(q); 1089 } 1090 1091 inserted: 1092 1093 #ifdef IPDIVERT 1094 if (ip->ip_off != 0) { 1095 /* 1096 * Strip any divert information; only the info 1097 * on the first fragment is used/kept. 1098 */ 1099 struct m_tag *mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 1100 if (mtag) 1101 m_tag_delete(m, mtag); 1102 } 1103 #endif 1104 1105 /* 1106 * Check for complete reassembly and perform frag per packet 1107 * limiting. 1108 * 1109 * Frag limiting is performed here so that the nth frag has 1110 * a chance to complete the packet before we drop the packet. 1111 * As a result, n+1 frags are actually allowed per packet, but 1112 * only n will ever be stored. (n = maxfragsperpacket.) 1113 * 1114 */ 1115 next = 0; 1116 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1117 if (GETIP(q)->ip_off != next) { 1118 if (fp->ipq_nfrags > maxfragsperpacket) { 1119 ipstat.ips_fragdropped += fp->ipq_nfrags; 1120 ip_freef(head, fp); 1121 } 1122 goto done; 1123 } 1124 next += GETIP(q)->ip_len; 1125 } 1126 /* Make sure the last packet didn't have the IP_MF flag */ 1127 if (p->m_flags & M_FRAG) { 1128 if (fp->ipq_nfrags > maxfragsperpacket) { 1129 ipstat.ips_fragdropped += fp->ipq_nfrags; 1130 ip_freef(head, fp); 1131 } 1132 goto done; 1133 } 1134 1135 /* 1136 * Reassembly is complete. Make sure the packet is a sane size. 1137 */ 1138 q = fp->ipq_frags; 1139 ip = GETIP(q); 1140 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1141 ipstat.ips_toolong++; 1142 ipstat.ips_fragdropped += fp->ipq_nfrags; 1143 ip_freef(head, fp); 1144 goto done; 1145 } 1146 1147 /* 1148 * Concatenate fragments. 1149 */ 1150 m = q; 1151 t = m->m_next; 1152 m->m_next = 0; 1153 m_cat(m, t); 1154 nq = q->m_nextpkt; 1155 q->m_nextpkt = 0; 1156 for (q = nq; q != NULL; q = nq) { 1157 nq = q->m_nextpkt; 1158 q->m_nextpkt = NULL; 1159 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1160 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1161 m_cat(m, q); 1162 } 1163 #ifdef MAC 1164 mac_create_datagram_from_ipq(fp, m); 1165 mac_destroy_ipq(fp); 1166 #endif 1167 1168 /* 1169 * Create header for new ip packet by modifying header of first 1170 * packet; dequeue and discard fragment reassembly header. 1171 * Make header visible. 1172 */ 1173 ip->ip_len = (ip->ip_hl << 2) + next; 1174 ip->ip_src = fp->ipq_src; 1175 ip->ip_dst = fp->ipq_dst; 1176 TAILQ_REMOVE(head, fp, ipq_list); 1177 nipq--; 1178 (void) m_free(dtom(fp)); 1179 m->m_len += (ip->ip_hl << 2); 1180 m->m_data -= (ip->ip_hl << 2); 1181 /* some debugging cruft by sklower, below, will go away soon */ 1182 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1183 m_fixhdr(m); 1184 ipstat.ips_reassembled++; 1185 IPQ_UNLOCK(); 1186 return (m); 1187 1188 dropfrag: 1189 ipstat.ips_fragdropped++; 1190 if (fp != NULL) 1191 fp->ipq_nfrags--; 1192 m_freem(m); 1193 done: 1194 IPQ_UNLOCK(); 1195 return (NULL); 1196 1197 #undef GETIP 1198 } 1199 1200 /* 1201 * Free a fragment reassembly header and all 1202 * associated datagrams. 1203 */ 1204 static void 1205 ip_freef(fhp, fp) 1206 struct ipqhead *fhp; 1207 struct ipq *fp; 1208 { 1209 register struct mbuf *q; 1210 1211 IPQ_LOCK_ASSERT(); 1212 1213 while (fp->ipq_frags) { 1214 q = fp->ipq_frags; 1215 fp->ipq_frags = q->m_nextpkt; 1216 m_freem(q); 1217 } 1218 TAILQ_REMOVE(fhp, fp, ipq_list); 1219 (void) m_free(dtom(fp)); 1220 nipq--; 1221 } 1222 1223 /* 1224 * IP timer processing; 1225 * if a timer expires on a reassembly 1226 * queue, discard it. 1227 */ 1228 void 1229 ip_slowtimo() 1230 { 1231 register struct ipq *fp; 1232 int s = splnet(); 1233 int i; 1234 1235 IPQ_LOCK(); 1236 for (i = 0; i < IPREASS_NHASH; i++) { 1237 for(fp = TAILQ_FIRST(&ipq[i]); fp;) { 1238 struct ipq *fpp; 1239 1240 fpp = fp; 1241 fp = TAILQ_NEXT(fp, ipq_list); 1242 if(--fpp->ipq_ttl == 0) { 1243 ipstat.ips_fragtimeout += fpp->ipq_nfrags; 1244 ip_freef(&ipq[i], fpp); 1245 } 1246 } 1247 } 1248 /* 1249 * If we are over the maximum number of fragments 1250 * (due to the limit being lowered), drain off 1251 * enough to get down to the new limit. 1252 */ 1253 if (maxnipq >= 0 && nipq > maxnipq) { 1254 for (i = 0; i < IPREASS_NHASH; i++) { 1255 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) { 1256 ipstat.ips_fragdropped += 1257 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1258 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1259 } 1260 } 1261 } 1262 IPQ_UNLOCK(); 1263 splx(s); 1264 } 1265 1266 /* 1267 * Drain off all datagram fragments. 1268 */ 1269 void 1270 ip_drain() 1271 { 1272 int i; 1273 1274 IPQ_LOCK(); 1275 for (i = 0; i < IPREASS_NHASH; i++) { 1276 while(!TAILQ_EMPTY(&ipq[i])) { 1277 ipstat.ips_fragdropped += 1278 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1279 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1280 } 1281 } 1282 IPQ_UNLOCK(); 1283 in_rtqdrain(); 1284 } 1285 1286 /* 1287 * Do option processing on a datagram, 1288 * possibly discarding it if bad options are encountered, 1289 * or forwarding it if source-routed. 1290 * The pass argument is used when operating in the IPSTEALTH 1291 * mode to tell what options to process: 1292 * [LS]SRR (pass 0) or the others (pass 1). 1293 * The reason for as many as two passes is that when doing IPSTEALTH, 1294 * non-routing options should be processed only if the packet is for us. 1295 * Returns 1 if packet has been forwarded/freed, 1296 * 0 if the packet should be processed further. 1297 */ 1298 static int 1299 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop) 1300 { 1301 struct ip *ip = mtod(m, struct ip *); 1302 u_char *cp; 1303 struct in_ifaddr *ia; 1304 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1305 struct in_addr *sin, dst; 1306 n_time ntime; 1307 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 1308 1309 /* ignore or reject packets with IP options */ 1310 if (ip_doopts == 0) 1311 return 0; 1312 else if (ip_doopts == 2) { 1313 type = ICMP_UNREACH; 1314 code = ICMP_UNREACH_FILTER_PROHIB; 1315 goto bad; 1316 } 1317 1318 dst = ip->ip_dst; 1319 cp = (u_char *)(ip + 1); 1320 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1321 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1322 opt = cp[IPOPT_OPTVAL]; 1323 if (opt == IPOPT_EOL) 1324 break; 1325 if (opt == IPOPT_NOP) 1326 optlen = 1; 1327 else { 1328 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1329 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1330 goto bad; 1331 } 1332 optlen = cp[IPOPT_OLEN]; 1333 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1334 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1335 goto bad; 1336 } 1337 } 1338 switch (opt) { 1339 1340 default: 1341 break; 1342 1343 /* 1344 * Source routing with record. 1345 * Find interface with current destination address. 1346 * If none on this machine then drop if strictly routed, 1347 * or do nothing if loosely routed. 1348 * Record interface address and bring up next address 1349 * component. If strictly routed make sure next 1350 * address is on directly accessible net. 1351 */ 1352 case IPOPT_LSRR: 1353 case IPOPT_SSRR: 1354 #ifdef IPSTEALTH 1355 if (ipstealth && pass > 0) 1356 break; 1357 #endif 1358 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1359 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1360 goto bad; 1361 } 1362 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1363 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1364 goto bad; 1365 } 1366 ipaddr.sin_addr = ip->ip_dst; 1367 ia = (struct in_ifaddr *) 1368 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1369 if (ia == NULL) { 1370 if (opt == IPOPT_SSRR) { 1371 type = ICMP_UNREACH; 1372 code = ICMP_UNREACH_SRCFAIL; 1373 goto bad; 1374 } 1375 if (!ip_dosourceroute) 1376 goto nosourcerouting; 1377 /* 1378 * Loose routing, and not at next destination 1379 * yet; nothing to do except forward. 1380 */ 1381 break; 1382 } 1383 off--; /* 0 origin */ 1384 if (off > optlen - (int)sizeof(struct in_addr)) { 1385 /* 1386 * End of source route. Should be for us. 1387 */ 1388 if (!ip_acceptsourceroute) 1389 goto nosourcerouting; 1390 save_rte(cp, ip->ip_src); 1391 break; 1392 } 1393 #ifdef IPSTEALTH 1394 if (ipstealth) 1395 goto dropit; 1396 #endif 1397 if (!ip_dosourceroute) { 1398 if (ipforwarding) { 1399 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1400 /* 1401 * Acting as a router, so generate ICMP 1402 */ 1403 nosourcerouting: 1404 strcpy(buf, inet_ntoa(ip->ip_dst)); 1405 log(LOG_WARNING, 1406 "attempted source route from %s to %s\n", 1407 inet_ntoa(ip->ip_src), buf); 1408 type = ICMP_UNREACH; 1409 code = ICMP_UNREACH_SRCFAIL; 1410 goto bad; 1411 } else { 1412 /* 1413 * Not acting as a router, so silently drop. 1414 */ 1415 #ifdef IPSTEALTH 1416 dropit: 1417 #endif 1418 ipstat.ips_cantforward++; 1419 m_freem(m); 1420 return (1); 1421 } 1422 } 1423 1424 /* 1425 * locate outgoing interface 1426 */ 1427 (void)memcpy(&ipaddr.sin_addr, cp + off, 1428 sizeof(ipaddr.sin_addr)); 1429 1430 if (opt == IPOPT_SSRR) { 1431 #define INA struct in_ifaddr * 1432 #define SA struct sockaddr * 1433 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == NULL) 1434 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1435 } else 1436 ia = ip_rtaddr(ipaddr.sin_addr); 1437 if (ia == NULL) { 1438 type = ICMP_UNREACH; 1439 code = ICMP_UNREACH_SRCFAIL; 1440 goto bad; 1441 } 1442 ip->ip_dst = ipaddr.sin_addr; 1443 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1444 sizeof(struct in_addr)); 1445 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1446 /* 1447 * Let ip_intr's mcast routing check handle mcast pkts 1448 */ 1449 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1450 break; 1451 1452 case IPOPT_RR: 1453 #ifdef IPSTEALTH 1454 if (ipstealth && pass == 0) 1455 break; 1456 #endif 1457 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1458 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1459 goto bad; 1460 } 1461 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1462 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1463 goto bad; 1464 } 1465 /* 1466 * If no space remains, ignore. 1467 */ 1468 off--; /* 0 origin */ 1469 if (off > optlen - (int)sizeof(struct in_addr)) 1470 break; 1471 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1472 sizeof(ipaddr.sin_addr)); 1473 /* 1474 * locate outgoing interface; if we're the destination, 1475 * use the incoming interface (should be same). 1476 */ 1477 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL && 1478 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1479 type = ICMP_UNREACH; 1480 code = ICMP_UNREACH_HOST; 1481 goto bad; 1482 } 1483 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1484 sizeof(struct in_addr)); 1485 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1486 break; 1487 1488 case IPOPT_TS: 1489 #ifdef IPSTEALTH 1490 if (ipstealth && pass == 0) 1491 break; 1492 #endif 1493 code = cp - (u_char *)ip; 1494 if (optlen < 4 || optlen > 40) { 1495 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1496 goto bad; 1497 } 1498 if ((off = cp[IPOPT_OFFSET]) < 5) { 1499 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1500 goto bad; 1501 } 1502 if (off > optlen - (int)sizeof(int32_t)) { 1503 cp[IPOPT_OFFSET + 1] += (1 << 4); 1504 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1505 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1506 goto bad; 1507 } 1508 break; 1509 } 1510 off--; /* 0 origin */ 1511 sin = (struct in_addr *)(cp + off); 1512 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1513 1514 case IPOPT_TS_TSONLY: 1515 break; 1516 1517 case IPOPT_TS_TSANDADDR: 1518 if (off + sizeof(n_time) + 1519 sizeof(struct in_addr) > optlen) { 1520 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1521 goto bad; 1522 } 1523 ipaddr.sin_addr = dst; 1524 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1525 m->m_pkthdr.rcvif); 1526 if (ia == NULL) 1527 continue; 1528 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1529 sizeof(struct in_addr)); 1530 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1531 off += sizeof(struct in_addr); 1532 break; 1533 1534 case IPOPT_TS_PRESPEC: 1535 if (off + sizeof(n_time) + 1536 sizeof(struct in_addr) > optlen) { 1537 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1538 goto bad; 1539 } 1540 (void)memcpy(&ipaddr.sin_addr, sin, 1541 sizeof(struct in_addr)); 1542 if (ifa_ifwithaddr((SA)&ipaddr) == NULL) 1543 continue; 1544 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1545 off += sizeof(struct in_addr); 1546 break; 1547 1548 default: 1549 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1550 goto bad; 1551 } 1552 ntime = iptime(); 1553 (void)memcpy(cp + off, &ntime, sizeof(n_time)); 1554 cp[IPOPT_OFFSET] += sizeof(n_time); 1555 } 1556 } 1557 if (forward && ipforwarding) { 1558 ip_forward(m, 1, next_hop); 1559 return (1); 1560 } 1561 return (0); 1562 bad: 1563 icmp_error(m, type, code, 0, 0); 1564 ipstat.ips_badoptions++; 1565 return (1); 1566 } 1567 1568 /* 1569 * Given address of next destination (final or next hop), 1570 * return internet address info of interface to be used to get there. 1571 */ 1572 struct in_ifaddr * 1573 ip_rtaddr(dst) 1574 struct in_addr dst; 1575 { 1576 struct route sro; 1577 struct sockaddr_in *sin; 1578 struct in_ifaddr *ifa; 1579 1580 bzero(&sro, sizeof(sro)); 1581 sin = (struct sockaddr_in *)&sro.ro_dst; 1582 sin->sin_family = AF_INET; 1583 sin->sin_len = sizeof(*sin); 1584 sin->sin_addr = dst; 1585 rtalloc_ign(&sro, RTF_CLONING); 1586 1587 if (sro.ro_rt == NULL) 1588 return ((struct in_ifaddr *)0); 1589 1590 ifa = ifatoia(sro.ro_rt->rt_ifa); 1591 RTFREE(sro.ro_rt); 1592 return ifa; 1593 } 1594 1595 /* 1596 * Save incoming source route for use in replies, 1597 * to be picked up later by ip_srcroute if the receiver is interested. 1598 */ 1599 static void 1600 save_rte(option, dst) 1601 u_char *option; 1602 struct in_addr dst; 1603 { 1604 unsigned olen; 1605 1606 olen = option[IPOPT_OLEN]; 1607 #ifdef DIAGNOSTIC 1608 if (ipprintfs) 1609 printf("save_rte: olen %d\n", olen); 1610 #endif 1611 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1612 return; 1613 bcopy(option, ip_srcrt.srcopt, olen); 1614 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1615 ip_srcrt.dst = dst; 1616 } 1617 1618 /* 1619 * Retrieve incoming source route for use in replies, 1620 * in the same form used by setsockopt. 1621 * The first hop is placed before the options, will be removed later. 1622 */ 1623 struct mbuf * 1624 ip_srcroute() 1625 { 1626 register struct in_addr *p, *q; 1627 register struct mbuf *m; 1628 1629 if (ip_nhops == 0) 1630 return ((struct mbuf *)0); 1631 m = m_get(M_DONTWAIT, MT_HEADER); 1632 if (m == NULL) 1633 return ((struct mbuf *)0); 1634 1635 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1636 1637 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1638 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1639 OPTSIZ; 1640 #ifdef DIAGNOSTIC 1641 if (ipprintfs) 1642 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1643 #endif 1644 1645 /* 1646 * First save first hop for return route 1647 */ 1648 p = &ip_srcrt.route[ip_nhops - 1]; 1649 *(mtod(m, struct in_addr *)) = *p--; 1650 #ifdef DIAGNOSTIC 1651 if (ipprintfs) 1652 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1653 #endif 1654 1655 /* 1656 * Copy option fields and padding (nop) to mbuf. 1657 */ 1658 ip_srcrt.nop = IPOPT_NOP; 1659 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1660 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1661 &ip_srcrt.nop, OPTSIZ); 1662 q = (struct in_addr *)(mtod(m, caddr_t) + 1663 sizeof(struct in_addr) + OPTSIZ); 1664 #undef OPTSIZ 1665 /* 1666 * Record return path as an IP source route, 1667 * reversing the path (pointers are now aligned). 1668 */ 1669 while (p >= ip_srcrt.route) { 1670 #ifdef DIAGNOSTIC 1671 if (ipprintfs) 1672 printf(" %lx", (u_long)ntohl(q->s_addr)); 1673 #endif 1674 *q++ = *p--; 1675 } 1676 /* 1677 * Last hop goes to final destination. 1678 */ 1679 *q = ip_srcrt.dst; 1680 #ifdef DIAGNOSTIC 1681 if (ipprintfs) 1682 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1683 #endif 1684 return (m); 1685 } 1686 1687 /* 1688 * Strip out IP options, at higher 1689 * level protocol in the kernel. 1690 * Second argument is buffer to which options 1691 * will be moved, and return value is their length. 1692 * XXX should be deleted; last arg currently ignored. 1693 */ 1694 void 1695 ip_stripoptions(m, mopt) 1696 register struct mbuf *m; 1697 struct mbuf *mopt; 1698 { 1699 register int i; 1700 struct ip *ip = mtod(m, struct ip *); 1701 register caddr_t opts; 1702 int olen; 1703 1704 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1705 opts = (caddr_t)(ip + 1); 1706 i = m->m_len - (sizeof (struct ip) + olen); 1707 bcopy(opts + olen, opts, (unsigned)i); 1708 m->m_len -= olen; 1709 if (m->m_flags & M_PKTHDR) 1710 m->m_pkthdr.len -= olen; 1711 ip->ip_v = IPVERSION; 1712 ip->ip_hl = sizeof(struct ip) >> 2; 1713 } 1714 1715 u_char inetctlerrmap[PRC_NCMDS] = { 1716 0, 0, 0, 0, 1717 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1718 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1719 EMSGSIZE, EHOSTUNREACH, 0, 0, 1720 0, 0, EHOSTUNREACH, 0, 1721 ENOPROTOOPT, ECONNREFUSED 1722 }; 1723 1724 /* 1725 * Forward a packet. If some error occurs return the sender 1726 * an icmp packet. Note we can't always generate a meaningful 1727 * icmp message because icmp doesn't have a large enough repertoire 1728 * of codes and types. 1729 * 1730 * If not forwarding, just drop the packet. This could be confusing 1731 * if ipforwarding was zero but some routing protocol was advancing 1732 * us as a gateway to somewhere. However, we must let the routing 1733 * protocol deal with that. 1734 * 1735 * The srcrt parameter indicates whether the packet is being forwarded 1736 * via a source route. 1737 */ 1738 static void 1739 ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop) 1740 { 1741 struct ip *ip = mtod(m, struct ip *); 1742 struct in_ifaddr *ia; 1743 int error, type = 0, code = 0; 1744 struct mbuf *mcopy; 1745 n_long dest; 1746 struct in_addr pkt_dst; 1747 struct ifnet *destifp; 1748 #if defined(IPSEC) || defined(FAST_IPSEC) 1749 struct ifnet dummyifp; 1750 #endif 1751 1752 /* 1753 * Cache the destination address of the packet; this may be 1754 * changed by use of 'ipfw fwd'. 1755 */ 1756 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst; 1757 1758 #ifdef DIAGNOSTIC 1759 if (ipprintfs) 1760 printf("forward: src %lx dst %lx ttl %x\n", 1761 (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr, 1762 ip->ip_ttl); 1763 #endif 1764 1765 1766 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) { 1767 ipstat.ips_cantforward++; 1768 m_freem(m); 1769 return; 1770 } 1771 #ifdef IPSTEALTH 1772 if (!ipstealth) { 1773 #endif 1774 if (ip->ip_ttl <= IPTTLDEC) { 1775 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1776 0, 0); 1777 return; 1778 } 1779 #ifdef IPSTEALTH 1780 } 1781 #endif 1782 1783 if ((ia = ip_rtaddr(pkt_dst)) == NULL) { 1784 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1785 return; 1786 } 1787 1788 /* 1789 * Save the IP header and at most 8 bytes of the payload, 1790 * in case we need to generate an ICMP message to the src. 1791 * 1792 * XXX this can be optimized a lot by saving the data in a local 1793 * buffer on the stack (72 bytes at most), and only allocating the 1794 * mbuf if really necessary. The vast majority of the packets 1795 * are forwarded without having to send an ICMP back (either 1796 * because unnecessary, or because rate limited), so we are 1797 * really we are wasting a lot of work here. 1798 * 1799 * We don't use m_copy() because it might return a reference 1800 * to a shared cluster. Both this function and ip_output() 1801 * assume exclusive access to the IP header in `m', so any 1802 * data in a cluster may change before we reach icmp_error(). 1803 */ 1804 MGET(mcopy, M_DONTWAIT, m->m_type); 1805 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1806 /* 1807 * It's probably ok if the pkthdr dup fails (because 1808 * the deep copy of the tag chain failed), but for now 1809 * be conservative and just discard the copy since 1810 * code below may some day want the tags. 1811 */ 1812 m_free(mcopy); 1813 mcopy = NULL; 1814 } 1815 if (mcopy != NULL) { 1816 mcopy->m_len = imin((ip->ip_hl << 2) + 8, 1817 (int)ip->ip_len); 1818 mcopy->m_pkthdr.len = mcopy->m_len; 1819 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1820 } 1821 1822 #ifdef IPSTEALTH 1823 if (!ipstealth) { 1824 #endif 1825 ip->ip_ttl -= IPTTLDEC; 1826 #ifdef IPSTEALTH 1827 } 1828 #endif 1829 1830 /* 1831 * If forwarding packet using same interface that it came in on, 1832 * perhaps should send a redirect to sender to shortcut a hop. 1833 * Only send redirect if source is sending directly to us, 1834 * and if packet was not source routed (or has any options). 1835 * Also, don't send redirect if forwarding using a default route 1836 * or a route modified by a redirect. 1837 */ 1838 dest = 0; 1839 if (ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) { 1840 struct sockaddr_in *sin; 1841 struct route ro; 1842 struct rtentry *rt; 1843 1844 bzero(&ro, sizeof(ro)); 1845 sin = (struct sockaddr_in *)&ro.ro_dst; 1846 sin->sin_family = AF_INET; 1847 sin->sin_len = sizeof(*sin); 1848 sin->sin_addr = pkt_dst; 1849 rtalloc_ign(&ro, RTF_CLONING); 1850 1851 rt = ro.ro_rt; 1852 1853 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1854 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1855 ipsendredirects && !srcrt && !next_hop) { 1856 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1857 u_long src = ntohl(ip->ip_src.s_addr); 1858 1859 if (RTA(rt) && 1860 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1861 if (rt->rt_flags & RTF_GATEWAY) 1862 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1863 else 1864 dest = pkt_dst.s_addr; 1865 /* Router requirements says to only send host redirects */ 1866 type = ICMP_REDIRECT; 1867 code = ICMP_REDIRECT_HOST; 1868 #ifdef DIAGNOSTIC 1869 if (ipprintfs) 1870 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1871 #endif 1872 } 1873 } 1874 if (rt) 1875 RTFREE(rt); 1876 } 1877 1878 if (next_hop) { 1879 struct m_tag *mtag = m_tag_get(PACKET_TAG_IPFORWARD, 1880 sizeof(struct sockaddr_in *), M_NOWAIT); 1881 if (mtag == NULL) { 1882 m_freem(m); 1883 return; 1884 } 1885 *(struct sockaddr_in **)(mtag+1) = next_hop; 1886 m_tag_prepend(m, mtag); 1887 } 1888 error = ip_output(m, (struct mbuf *)0, NULL, IP_FORWARDING, 0, NULL); 1889 if (error) 1890 ipstat.ips_cantforward++; 1891 else { 1892 ipstat.ips_forward++; 1893 if (type) 1894 ipstat.ips_redirectsent++; 1895 else { 1896 if (mcopy) 1897 m_freem(mcopy); 1898 return; 1899 } 1900 } 1901 if (mcopy == NULL) 1902 return; 1903 destifp = NULL; 1904 1905 switch (error) { 1906 1907 case 0: /* forwarded, but need redirect */ 1908 /* type, code set above */ 1909 break; 1910 1911 case ENETUNREACH: /* shouldn't happen, checked above */ 1912 case EHOSTUNREACH: 1913 case ENETDOWN: 1914 case EHOSTDOWN: 1915 default: 1916 type = ICMP_UNREACH; 1917 code = ICMP_UNREACH_HOST; 1918 break; 1919 1920 case EMSGSIZE: 1921 type = ICMP_UNREACH; 1922 code = ICMP_UNREACH_NEEDFRAG; 1923 #if defined(IPSEC) || defined(FAST_IPSEC) 1924 /* 1925 * If the packet is routed over IPsec tunnel, tell the 1926 * originator the tunnel MTU. 1927 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1928 * XXX quickhack!!! 1929 */ 1930 { 1931 struct secpolicy *sp = NULL; 1932 int ipsecerror; 1933 int ipsechdr; 1934 struct route *ro; 1935 1936 #ifdef IPSEC 1937 sp = ipsec4_getpolicybyaddr(mcopy, 1938 IPSEC_DIR_OUTBOUND, 1939 IP_FORWARDING, 1940 &ipsecerror); 1941 #else /* FAST_IPSEC */ 1942 sp = ipsec_getpolicybyaddr(mcopy, 1943 IPSEC_DIR_OUTBOUND, 1944 IP_FORWARDING, 1945 &ipsecerror); 1946 #endif 1947 if (sp != NULL) { 1948 /* count IPsec header size */ 1949 ipsechdr = ipsec4_hdrsiz(mcopy, 1950 IPSEC_DIR_OUTBOUND, 1951 NULL); 1952 1953 /* 1954 * find the correct route for outer IPv4 1955 * header, compute tunnel MTU. 1956 * 1957 * XXX BUG ALERT 1958 * The "dummyifp" code relies upon the fact 1959 * that icmp_error() touches only ifp->if_mtu. 1960 */ 1961 /*XXX*/ 1962 destifp = NULL; 1963 if (sp->req != NULL 1964 && sp->req->sav != NULL 1965 && sp->req->sav->sah != NULL) { 1966 ro = &sp->req->sav->sah->sa_route; 1967 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1968 dummyifp.if_mtu = 1969 ro->ro_rt->rt_rmx.rmx_mtu ? 1970 ro->ro_rt->rt_rmx.rmx_mtu : 1971 ro->ro_rt->rt_ifp->if_mtu; 1972 dummyifp.if_mtu -= ipsechdr; 1973 destifp = &dummyifp; 1974 } 1975 } 1976 1977 #ifdef IPSEC 1978 key_freesp(sp); 1979 #else /* FAST_IPSEC */ 1980 KEY_FREESP(&sp); 1981 #endif 1982 ipstat.ips_cantfrag++; 1983 break; 1984 } else 1985 #endif /*IPSEC || FAST_IPSEC*/ 1986 destifp = ia->ia_ifp; 1987 #if defined(IPSEC) || defined(FAST_IPSEC) 1988 } 1989 #endif /*IPSEC || FAST_IPSEC*/ 1990 ipstat.ips_cantfrag++; 1991 break; 1992 1993 case ENOBUFS: 1994 /* 1995 * A router should not generate ICMP_SOURCEQUENCH as 1996 * required in RFC1812 Requirements for IP Version 4 Routers. 1997 * Source quench could be a big problem under DoS attacks, 1998 * or if the underlying interface is rate-limited. 1999 * Those who need source quench packets may re-enable them 2000 * via the net.inet.ip.sendsourcequench sysctl. 2001 */ 2002 if (ip_sendsourcequench == 0) { 2003 m_freem(mcopy); 2004 return; 2005 } else { 2006 type = ICMP_SOURCEQUENCH; 2007 code = 0; 2008 } 2009 break; 2010 2011 case EACCES: /* ipfw denied packet */ 2012 m_freem(mcopy); 2013 return; 2014 } 2015 icmp_error(mcopy, type, code, dest, destifp); 2016 } 2017 2018 void 2019 ip_savecontrol(inp, mp, ip, m) 2020 register struct inpcb *inp; 2021 register struct mbuf **mp; 2022 register struct ip *ip; 2023 register struct mbuf *m; 2024 { 2025 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 2026 struct bintime bt; 2027 2028 bintime(&bt); 2029 if (inp->inp_socket->so_options & SO_BINTIME) { 2030 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 2031 SCM_BINTIME, SOL_SOCKET); 2032 if (*mp) 2033 mp = &(*mp)->m_next; 2034 } 2035 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2036 struct timeval tv; 2037 2038 bintime2timeval(&bt, &tv); 2039 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2040 SCM_TIMESTAMP, SOL_SOCKET); 2041 if (*mp) 2042 mp = &(*mp)->m_next; 2043 } 2044 } 2045 if (inp->inp_flags & INP_RECVDSTADDR) { 2046 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2047 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2048 if (*mp) 2049 mp = &(*mp)->m_next; 2050 } 2051 if (inp->inp_flags & INP_RECVTTL) { 2052 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 2053 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 2054 if (*mp) 2055 mp = &(*mp)->m_next; 2056 } 2057 #ifdef notyet 2058 /* XXX 2059 * Moving these out of udp_input() made them even more broken 2060 * than they already were. 2061 */ 2062 /* options were tossed already */ 2063 if (inp->inp_flags & INP_RECVOPTS) { 2064 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2065 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2066 if (*mp) 2067 mp = &(*mp)->m_next; 2068 } 2069 /* ip_srcroute doesn't do what we want here, need to fix */ 2070 if (inp->inp_flags & INP_RECVRETOPTS) { 2071 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 2072 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2073 if (*mp) 2074 mp = &(*mp)->m_next; 2075 } 2076 #endif 2077 if (inp->inp_flags & INP_RECVIF) { 2078 struct ifnet *ifp; 2079 struct sdlbuf { 2080 struct sockaddr_dl sdl; 2081 u_char pad[32]; 2082 } sdlbuf; 2083 struct sockaddr_dl *sdp; 2084 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2085 2086 if (((ifp = m->m_pkthdr.rcvif)) 2087 && ( ifp->if_index && (ifp->if_index <= if_index))) { 2088 sdp = (struct sockaddr_dl *) 2089 (ifaddr_byindex(ifp->if_index)->ifa_addr); 2090 /* 2091 * Change our mind and don't try copy. 2092 */ 2093 if ((sdp->sdl_family != AF_LINK) 2094 || (sdp->sdl_len > sizeof(sdlbuf))) { 2095 goto makedummy; 2096 } 2097 bcopy(sdp, sdl2, sdp->sdl_len); 2098 } else { 2099 makedummy: 2100 sdl2->sdl_len 2101 = offsetof(struct sockaddr_dl, sdl_data[0]); 2102 sdl2->sdl_family = AF_LINK; 2103 sdl2->sdl_index = 0; 2104 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2105 } 2106 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 2107 IP_RECVIF, IPPROTO_IP); 2108 if (*mp) 2109 mp = &(*mp)->m_next; 2110 } 2111 } 2112 2113 /* 2114 * XXX these routines are called from the upper part of the kernel. 2115 * They need to be locked when we remove Giant. 2116 * 2117 * They could also be moved to ip_mroute.c, since all the RSVP 2118 * handling is done there already. 2119 */ 2120 static int ip_rsvp_on; 2121 struct socket *ip_rsvpd; 2122 int 2123 ip_rsvp_init(struct socket *so) 2124 { 2125 if (so->so_type != SOCK_RAW || 2126 so->so_proto->pr_protocol != IPPROTO_RSVP) 2127 return EOPNOTSUPP; 2128 2129 if (ip_rsvpd != NULL) 2130 return EADDRINUSE; 2131 2132 ip_rsvpd = so; 2133 /* 2134 * This may seem silly, but we need to be sure we don't over-increment 2135 * the RSVP counter, in case something slips up. 2136 */ 2137 if (!ip_rsvp_on) { 2138 ip_rsvp_on = 1; 2139 rsvp_on++; 2140 } 2141 2142 return 0; 2143 } 2144 2145 int 2146 ip_rsvp_done(void) 2147 { 2148 ip_rsvpd = NULL; 2149 /* 2150 * This may seem silly, but we need to be sure we don't over-decrement 2151 * the RSVP counter, in case something slips up. 2152 */ 2153 if (ip_rsvp_on) { 2154 ip_rsvp_on = 0; 2155 rsvp_on--; 2156 } 2157 return 0; 2158 } 2159 2160 void 2161 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 2162 { 2163 if (rsvp_input_p) { /* call the real one if loaded */ 2164 rsvp_input_p(m, off); 2165 return; 2166 } 2167 2168 /* Can still get packets with rsvp_on = 0 if there is a local member 2169 * of the group to which the RSVP packet is addressed. But in this 2170 * case we want to throw the packet away. 2171 */ 2172 2173 if (!rsvp_on) { 2174 m_freem(m); 2175 return; 2176 } 2177 2178 if (ip_rsvpd != NULL) { 2179 rip_input(m, off); 2180 return; 2181 } 2182 /* Drop the packet */ 2183 m_freem(m); 2184 } 2185