1 /*- 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 30 * $FreeBSD$ 31 */ 32 33 #include "opt_bootp.h" 34 #include "opt_ipfw.h" 35 #include "opt_ipstealth.h" 36 #include "opt_ipsec.h" 37 #include "opt_mac.h" 38 #include "opt_carp.h" 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/callout.h> 43 #include <sys/mac.h> 44 #include <sys/mbuf.h> 45 #include <sys/malloc.h> 46 #include <sys/domain.h> 47 #include <sys/protosw.h> 48 #include <sys/socket.h> 49 #include <sys/time.h> 50 #include <sys/kernel.h> 51 #include <sys/syslog.h> 52 #include <sys/sysctl.h> 53 54 #include <net/pfil.h> 55 #include <net/if.h> 56 #include <net/if_types.h> 57 #include <net/if_var.h> 58 #include <net/if_dl.h> 59 #include <net/route.h> 60 #include <net/netisr.h> 61 62 #include <netinet/in.h> 63 #include <netinet/in_systm.h> 64 #include <netinet/in_var.h> 65 #include <netinet/ip.h> 66 #include <netinet/in_pcb.h> 67 #include <netinet/ip_var.h> 68 #include <netinet/ip_icmp.h> 69 #include <machine/in_cksum.h> 70 #ifdef DEV_CARP 71 #include <netinet/ip_carp.h> 72 #endif 73 74 #include <sys/socketvar.h> 75 76 /* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */ 77 #include <netinet/ip_fw.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 int ip_do_randomid = 0; 139 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 140 &ip_do_randomid, 0, 141 "Assign random ip_id values"); 142 143 /* 144 * XXX - Setting ip_checkinterface mostly implements the receive side of 145 * the Strong ES model described in RFC 1122, but since the routing table 146 * and transmit implementation do not implement the Strong ES model, 147 * setting this to 1 results in an odd hybrid. 148 * 149 * XXX - ip_checkinterface currently must be disabled if you use ipnat 150 * to translate the destination address to another local interface. 151 * 152 * XXX - ip_checkinterface must be disabled if you add IP aliases 153 * to the loopback interface instead of the interface where the 154 * packets for those addresses are received. 155 */ 156 static int ip_checkinterface = 0; 157 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 158 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 159 160 #ifdef DIAGNOSTIC 161 static int ipprintfs = 0; 162 #endif 163 164 struct pfil_head inet_pfil_hook; /* Packet filter hooks */ 165 166 static struct ifqueue ipintrq; 167 static int ipqmaxlen = IFQ_MAXLEN; 168 169 extern struct domain inetdomain; 170 extern struct protosw inetsw[]; 171 u_char ip_protox[IPPROTO_MAX]; 172 struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 173 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 174 u_long in_ifaddrhmask; /* mask for hash table */ 175 176 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 177 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 178 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 179 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 180 181 struct ipstat ipstat; 182 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 183 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 184 185 /* Packet reassembly stuff */ 186 #define IPREASS_NHASH_LOG2 6 187 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 188 #define IPREASS_HMASK (IPREASS_NHASH - 1) 189 #define IPREASS_HASH(x,y) \ 190 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 191 192 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 193 struct mtx ipqlock; 194 struct callout ipport_tick_callout; 195 196 #define IPQ_LOCK() mtx_lock(&ipqlock) 197 #define IPQ_UNLOCK() mtx_unlock(&ipqlock) 198 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 199 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 200 201 #ifdef IPCTL_DEFMTU 202 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 203 &ip_mtu, 0, "Default MTU"); 204 #endif 205 206 #ifdef IPSTEALTH 207 int ipstealth = 0; 208 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 209 &ipstealth, 0, ""); 210 #endif 211 212 /* 213 * ipfw_ether and ipfw_bridge hooks. 214 * XXX: Temporary until those are converted to pfil_hooks as well. 215 */ 216 ip_fw_chk_t *ip_fw_chk_ptr = NULL; 217 ip_dn_io_t *ip_dn_io_ptr = NULL; 218 int fw_enable = 1; 219 int fw_one_pass = 1; 220 221 /* 222 * XXX this is ugly. IP options source routing magic. 223 */ 224 struct ipoptrt { 225 struct in_addr dst; /* final destination */ 226 char nop; /* one NOP to align */ 227 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 228 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 229 }; 230 231 struct ipopt_tag { 232 struct m_tag tag; 233 int ip_nhops; 234 struct ipoptrt ip_srcrt; 235 }; 236 237 static void save_rte(struct mbuf *, u_char *, struct in_addr); 238 static int ip_dooptions(struct mbuf *m, int); 239 static void ip_forward(struct mbuf *m, int srcrt); 240 static void ip_freef(struct ipqhead *, struct ipq *); 241 242 /* 243 * IP initialization: fill in IP protocol switch table. 244 * All protocols not implemented in kernel go to raw IP protocol handler. 245 */ 246 void 247 ip_init() 248 { 249 register struct protosw *pr; 250 register int i; 251 252 TAILQ_INIT(&in_ifaddrhead); 253 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 254 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 255 if (pr == NULL) 256 panic("ip_init: PF_INET not found"); 257 258 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 259 for (i = 0; i < IPPROTO_MAX; i++) 260 ip_protox[i] = pr - inetsw; 261 /* 262 * Cycle through IP protocols and put them into the appropriate place 263 * in ip_protox[]. 264 */ 265 for (pr = inetdomain.dom_protosw; 266 pr < inetdomain.dom_protoswNPROTOSW; pr++) 267 if (pr->pr_domain->dom_family == PF_INET && 268 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 269 /* Be careful to only index valid IP protocols. */ 270 if (pr->pr_protocol < IPPROTO_MAX) 271 ip_protox[pr->pr_protocol] = pr - inetsw; 272 } 273 274 /* Initialize packet filter hooks. */ 275 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 276 inet_pfil_hook.ph_af = AF_INET; 277 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 278 printf("%s: WARNING: unable to register pfil hook, " 279 "error %d\n", __func__, i); 280 281 /* Initialize IP reassembly queue. */ 282 IPQ_LOCK_INIT(); 283 for (i = 0; i < IPREASS_NHASH; i++) 284 TAILQ_INIT(&ipq[i]); 285 maxnipq = nmbclusters / 32; 286 maxfragsperpacket = 16; 287 288 /* Start ipport_tick. */ 289 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE); 290 ipport_tick(NULL); 291 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, 292 SHUTDOWN_PRI_DEFAULT); 293 294 /* Initialize various other remaining things. */ 295 ip_id = time_second & 0xffff; 296 ipintrq.ifq_maxlen = ipqmaxlen; 297 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF); 298 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE); 299 } 300 301 void ip_fini(xtp) 302 void *xtp; 303 { 304 callout_stop(&ipport_tick_callout); 305 } 306 307 /* 308 * Ip input routine. Checksum and byte swap header. If fragmented 309 * try to reassemble. Process options. Pass to next level. 310 */ 311 void 312 ip_input(struct mbuf *m) 313 { 314 struct ip *ip = NULL; 315 struct in_ifaddr *ia = NULL; 316 struct ifaddr *ifa; 317 int checkif, hlen = 0; 318 u_short sum; 319 int dchg = 0; /* dest changed after fw */ 320 struct in_addr odst; /* original dst address */ 321 #ifdef FAST_IPSEC 322 struct m_tag *mtag; 323 struct tdb_ident *tdbi; 324 struct secpolicy *sp; 325 int s, error; 326 #endif /* FAST_IPSEC */ 327 328 M_ASSERTPKTHDR(m); 329 330 if (m->m_flags & M_FASTFWD_OURS) { 331 /* 332 * Firewall or NAT changed destination to local. 333 * We expect ip_len and ip_off to be in host byte order. 334 */ 335 m->m_flags &= ~M_FASTFWD_OURS; 336 /* Set up some basics that will be used later. */ 337 ip = mtod(m, struct ip *); 338 hlen = ip->ip_hl << 2; 339 goto ours; 340 } 341 342 ipstat.ips_total++; 343 344 if (m->m_pkthdr.len < sizeof(struct ip)) 345 goto tooshort; 346 347 if (m->m_len < sizeof (struct ip) && 348 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 349 ipstat.ips_toosmall++; 350 return; 351 } 352 ip = mtod(m, struct ip *); 353 354 if (ip->ip_v != IPVERSION) { 355 ipstat.ips_badvers++; 356 goto bad; 357 } 358 359 hlen = ip->ip_hl << 2; 360 if (hlen < sizeof(struct ip)) { /* minimum header length */ 361 ipstat.ips_badhlen++; 362 goto bad; 363 } 364 if (hlen > m->m_len) { 365 if ((m = m_pullup(m, hlen)) == NULL) { 366 ipstat.ips_badhlen++; 367 return; 368 } 369 ip = mtod(m, struct ip *); 370 } 371 372 /* 127/8 must not appear on wire - RFC1122 */ 373 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 374 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 375 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 376 ipstat.ips_badaddr++; 377 goto bad; 378 } 379 } 380 381 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 382 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 383 } else { 384 if (hlen == sizeof(struct ip)) { 385 sum = in_cksum_hdr(ip); 386 } else { 387 sum = in_cksum(m, hlen); 388 } 389 } 390 if (sum) { 391 ipstat.ips_badsum++; 392 goto bad; 393 } 394 395 #ifdef ALTQ 396 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 397 /* packet is dropped by traffic conditioner */ 398 return; 399 #endif 400 401 /* 402 * Convert fields to host representation. 403 */ 404 ip->ip_len = ntohs(ip->ip_len); 405 if (ip->ip_len < hlen) { 406 ipstat.ips_badlen++; 407 goto bad; 408 } 409 ip->ip_off = ntohs(ip->ip_off); 410 411 /* 412 * Check that the amount of data in the buffers 413 * is as at least much as the IP header would have us expect. 414 * Trim mbufs if longer than we expect. 415 * Drop packet if shorter than we expect. 416 */ 417 if (m->m_pkthdr.len < ip->ip_len) { 418 tooshort: 419 ipstat.ips_tooshort++; 420 goto bad; 421 } 422 if (m->m_pkthdr.len > ip->ip_len) { 423 if (m->m_len == m->m_pkthdr.len) { 424 m->m_len = ip->ip_len; 425 m->m_pkthdr.len = ip->ip_len; 426 } else 427 m_adj(m, ip->ip_len - m->m_pkthdr.len); 428 } 429 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF) 430 /* 431 * Bypass packet filtering for packets from a tunnel (gif). 432 */ 433 if (ipsec_getnhist(m)) 434 goto passin; 435 #endif 436 #if defined(FAST_IPSEC) && !defined(IPSEC_FILTERGIF) 437 /* 438 * Bypass packet filtering for packets from a tunnel (gif). 439 */ 440 if (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL) 441 goto passin; 442 #endif 443 444 /* 445 * Run through list of hooks for input packets. 446 * 447 * NB: Beware of the destination address changing (e.g. 448 * by NAT rewriting). When this happens, tell 449 * ip_forward to do the right thing. 450 */ 451 452 /* Jump over all PFIL processing if hooks are not active. */ 453 if (inet_pfil_hook.ph_busy_count == -1) 454 goto passin; 455 456 odst = ip->ip_dst; 457 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 458 PFIL_IN, NULL) != 0) 459 return; 460 if (m == NULL) /* consumed by filter */ 461 return; 462 463 ip = mtod(m, struct ip *); 464 dchg = (odst.s_addr != ip->ip_dst.s_addr); 465 466 #ifdef IPFIREWALL_FORWARD 467 if (m->m_flags & M_FASTFWD_OURS) { 468 m->m_flags &= ~M_FASTFWD_OURS; 469 goto ours; 470 } 471 #ifndef IPFIREWALL_FORWARD_EXTENDED 472 dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL); 473 #else 474 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 475 /* 476 * Directly ship on the packet. This allows to forward packets 477 * that were destined for us to some other directly connected 478 * host. 479 */ 480 ip_forward(m, dchg); 481 return; 482 } 483 #endif /* IPFIREWALL_FORWARD_EXTENDED */ 484 #endif /* IPFIREWALL_FORWARD */ 485 486 passin: 487 /* 488 * Process options and, if not destined for us, 489 * ship it on. ip_dooptions returns 1 when an 490 * error was detected (causing an icmp message 491 * to be sent and the original packet to be freed). 492 */ 493 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 494 return; 495 496 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 497 * matter if it is destined to another node, or whether it is 498 * a multicast one, RSVP wants it! and prevents it from being forwarded 499 * anywhere else. Also checks if the rsvp daemon is running before 500 * grabbing the packet. 501 */ 502 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 503 goto ours; 504 505 /* 506 * Check our list of addresses, to see if the packet is for us. 507 * If we don't have any addresses, assume any unicast packet 508 * we receive might be for us (and let the upper layers deal 509 * with it). 510 */ 511 if (TAILQ_EMPTY(&in_ifaddrhead) && 512 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 513 goto ours; 514 515 /* 516 * Enable a consistency check between the destination address 517 * and the arrival interface for a unicast packet (the RFC 1122 518 * strong ES model) if IP forwarding is disabled and the packet 519 * is not locally generated and the packet is not subject to 520 * 'ipfw fwd'. 521 * 522 * XXX - Checking also should be disabled if the destination 523 * address is ipnat'ed to a different interface. 524 * 525 * XXX - Checking is incompatible with IP aliases added 526 * to the loopback interface instead of the interface where 527 * the packets are received. 528 * 529 * XXX - This is the case for carp vhost IPs as well so we 530 * insert a workaround. If the packet got here, we already 531 * checked with carp_iamatch() and carp_forus(). 532 */ 533 checkif = ip_checkinterface && (ipforwarding == 0) && 534 m->m_pkthdr.rcvif != NULL && 535 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 536 #ifdef DEV_CARP 537 !m->m_pkthdr.rcvif->if_carp && 538 #endif 539 (dchg == 0); 540 541 /* 542 * Check for exact addresses in the hash bucket. 543 */ 544 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 545 /* 546 * If the address matches, verify that the packet 547 * arrived via the correct interface if checking is 548 * enabled. 549 */ 550 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 551 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 552 goto ours; 553 } 554 /* 555 * Check for broadcast addresses. 556 * 557 * Only accept broadcast packets that arrive via the matching 558 * interface. Reception of forwarded directed broadcasts would 559 * be handled via ip_forward() and ether_output() with the loopback 560 * into the stack for SIMPLEX interfaces handled by ether_output(). 561 */ 562 if (m->m_pkthdr.rcvif != NULL && 563 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 564 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 565 if (ifa->ifa_addr->sa_family != AF_INET) 566 continue; 567 ia = ifatoia(ifa); 568 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 569 ip->ip_dst.s_addr) 570 goto ours; 571 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) 572 goto ours; 573 #ifdef BOOTP_COMPAT 574 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 575 goto ours; 576 #endif 577 } 578 } 579 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 580 struct in_multi *inm; 581 if (ip_mrouter) { 582 /* 583 * If we are acting as a multicast router, all 584 * incoming multicast packets are passed to the 585 * kernel-level multicast forwarding function. 586 * The packet is returned (relatively) intact; if 587 * ip_mforward() returns a non-zero value, the packet 588 * must be discarded, else it may be accepted below. 589 */ 590 if (ip_mforward && 591 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 592 ipstat.ips_cantforward++; 593 m_freem(m); 594 return; 595 } 596 597 /* 598 * The process-level routing daemon needs to receive 599 * all multicast IGMP packets, whether or not this 600 * host belongs to their destination groups. 601 */ 602 if (ip->ip_p == IPPROTO_IGMP) 603 goto ours; 604 ipstat.ips_forward++; 605 } 606 /* 607 * See if we belong to the destination multicast group on the 608 * arrival interface. 609 */ 610 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 611 if (inm == NULL) { 612 ipstat.ips_notmember++; 613 m_freem(m); 614 return; 615 } 616 goto ours; 617 } 618 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 619 goto ours; 620 if (ip->ip_dst.s_addr == INADDR_ANY) 621 goto ours; 622 623 /* 624 * FAITH(Firewall Aided Internet Translator) 625 */ 626 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 627 if (ip_keepfaith) { 628 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 629 goto ours; 630 } 631 m_freem(m); 632 return; 633 } 634 635 /* 636 * Not for us; forward if possible and desirable. 637 */ 638 if (ipforwarding == 0) { 639 ipstat.ips_cantforward++; 640 m_freem(m); 641 } else { 642 #ifdef IPSEC 643 /* 644 * Enforce inbound IPsec SPD. 645 */ 646 if (ipsec4_in_reject(m, NULL)) { 647 ipsecstat.in_polvio++; 648 goto bad; 649 } 650 #endif /* IPSEC */ 651 #ifdef FAST_IPSEC 652 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 653 s = splnet(); 654 if (mtag != NULL) { 655 tdbi = (struct tdb_ident *)(mtag + 1); 656 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 657 } else { 658 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 659 IP_FORWARDING, &error); 660 } 661 if (sp == NULL) { /* NB: can happen if error */ 662 splx(s); 663 /*XXX error stat???*/ 664 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 665 goto bad; 666 } 667 668 /* 669 * Check security policy against packet attributes. 670 */ 671 error = ipsec_in_reject(sp, m); 672 KEY_FREESP(&sp); 673 splx(s); 674 if (error) { 675 ipstat.ips_cantforward++; 676 goto bad; 677 } 678 #endif /* FAST_IPSEC */ 679 ip_forward(m, dchg); 680 } 681 return; 682 683 ours: 684 #ifdef IPSTEALTH 685 /* 686 * IPSTEALTH: Process non-routing options only 687 * if the packet is destined for us. 688 */ 689 if (ipstealth && hlen > sizeof (struct ip) && 690 ip_dooptions(m, 1)) 691 return; 692 #endif /* IPSTEALTH */ 693 694 /* Count the packet in the ip address stats */ 695 if (ia != NULL) { 696 ia->ia_ifa.if_ipackets++; 697 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 698 } 699 700 /* 701 * Attempt reassembly; if it succeeds, proceed. 702 * ip_reass() will return a different mbuf. 703 */ 704 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 705 m = ip_reass(m); 706 if (m == NULL) 707 return; 708 ip = mtod(m, struct ip *); 709 /* Get the header length of the reassembled packet */ 710 hlen = ip->ip_hl << 2; 711 } 712 713 /* 714 * Further protocols expect the packet length to be w/o the 715 * IP header. 716 */ 717 ip->ip_len -= hlen; 718 719 #ifdef IPSEC 720 /* 721 * enforce IPsec policy checking if we are seeing last header. 722 * note that we do not visit this with protocols with pcb layer 723 * code - like udp/tcp/raw ip. 724 */ 725 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 726 ipsec4_in_reject(m, NULL)) { 727 ipsecstat.in_polvio++; 728 goto bad; 729 } 730 #endif 731 #if FAST_IPSEC 732 /* 733 * enforce IPsec policy checking if we are seeing last header. 734 * note that we do not visit this with protocols with pcb layer 735 * code - like udp/tcp/raw ip. 736 */ 737 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 738 /* 739 * Check if the packet has already had IPsec processing 740 * done. If so, then just pass it along. This tag gets 741 * set during AH, ESP, etc. input handling, before the 742 * packet is returned to the ip input queue for delivery. 743 */ 744 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 745 s = splnet(); 746 if (mtag != NULL) { 747 tdbi = (struct tdb_ident *)(mtag + 1); 748 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 749 } else { 750 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 751 IP_FORWARDING, &error); 752 } 753 if (sp != NULL) { 754 /* 755 * Check security policy against packet attributes. 756 */ 757 error = ipsec_in_reject(sp, m); 758 KEY_FREESP(&sp); 759 } else { 760 /* XXX error stat??? */ 761 error = EINVAL; 762 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 763 goto bad; 764 } 765 splx(s); 766 if (error) 767 goto bad; 768 } 769 #endif /* FAST_IPSEC */ 770 771 /* 772 * Switch out to protocol's input routine. 773 */ 774 ipstat.ips_delivered++; 775 776 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 777 return; 778 bad: 779 m_freem(m); 780 } 781 782 /* 783 * Take incoming datagram fragment and try to reassemble it into 784 * whole datagram. If the argument is the first fragment or one 785 * in between the function will return NULL and store the mbuf 786 * in the fragment chain. If the argument is the last fragment 787 * the packet will be reassembled and the pointer to the new 788 * mbuf returned for further processing. Only m_tags attached 789 * to the first packet/fragment are preserved. 790 * The IP header is *NOT* adjusted out of iplen. 791 */ 792 793 struct mbuf * 794 ip_reass(struct mbuf *m) 795 { 796 struct ip *ip; 797 struct mbuf *p, *q, *nq, *t; 798 struct ipq *fp = NULL; 799 struct ipqhead *head; 800 int i, hlen, next; 801 u_int8_t ecn, ecn0; 802 u_short hash; 803 804 /* If maxnipq is 0, never accept fragments. */ 805 if (maxnipq == 0) { 806 ipstat.ips_fragments++; 807 ipstat.ips_fragdropped++; 808 m_freem(m); 809 return (NULL); 810 } 811 812 ip = mtod(m, struct ip *); 813 hlen = ip->ip_hl << 2; 814 815 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 816 head = &ipq[hash]; 817 IPQ_LOCK(); 818 819 /* 820 * Look for queue of fragments 821 * of this datagram. 822 */ 823 TAILQ_FOREACH(fp, head, ipq_list) 824 if (ip->ip_id == fp->ipq_id && 825 ip->ip_src.s_addr == fp->ipq_src.s_addr && 826 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 827 #ifdef MAC 828 mac_fragment_match(m, fp) && 829 #endif 830 ip->ip_p == fp->ipq_p) 831 goto found; 832 833 fp = NULL; 834 835 /* 836 * Enforce upper bound on number of fragmented packets 837 * for which we attempt reassembly; 838 * If maxnipq is -1, accept all fragments without limitation. 839 */ 840 if ((nipq > maxnipq) && (maxnipq > 0)) { 841 /* 842 * drop something from the tail of the current queue 843 * before proceeding further 844 */ 845 struct ipq *q = TAILQ_LAST(head, ipqhead); 846 if (q == NULL) { /* gak */ 847 for (i = 0; i < IPREASS_NHASH; i++) { 848 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 849 if (r) { 850 ipstat.ips_fragtimeout += r->ipq_nfrags; 851 ip_freef(&ipq[i], r); 852 break; 853 } 854 } 855 } else { 856 ipstat.ips_fragtimeout += q->ipq_nfrags; 857 ip_freef(head, q); 858 } 859 } 860 861 found: 862 /* 863 * Adjust ip_len to not reflect header, 864 * convert offset of this to bytes. 865 */ 866 ip->ip_len -= hlen; 867 if (ip->ip_off & IP_MF) { 868 /* 869 * Make sure that fragments have a data length 870 * that's a non-zero multiple of 8 bytes. 871 */ 872 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 873 ipstat.ips_toosmall++; /* XXX */ 874 goto dropfrag; 875 } 876 m->m_flags |= M_FRAG; 877 } else 878 m->m_flags &= ~M_FRAG; 879 ip->ip_off <<= 3; 880 881 882 /* 883 * Attempt reassembly; if it succeeds, proceed. 884 * ip_reass() will return a different mbuf. 885 */ 886 ipstat.ips_fragments++; 887 m->m_pkthdr.header = ip; 888 889 /* Previous ip_reass() started here. */ 890 /* 891 * Presence of header sizes in mbufs 892 * would confuse code below. 893 */ 894 m->m_data += hlen; 895 m->m_len -= hlen; 896 897 /* 898 * If first fragment to arrive, create a reassembly queue. 899 */ 900 if (fp == NULL) { 901 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 902 goto dropfrag; 903 fp = mtod(t, struct ipq *); 904 #ifdef MAC 905 if (mac_init_ipq(fp, M_NOWAIT) != 0) { 906 m_free(t); 907 goto dropfrag; 908 } 909 mac_create_ipq(m, fp); 910 #endif 911 TAILQ_INSERT_HEAD(head, fp, ipq_list); 912 nipq++; 913 fp->ipq_nfrags = 1; 914 fp->ipq_ttl = IPFRAGTTL; 915 fp->ipq_p = ip->ip_p; 916 fp->ipq_id = ip->ip_id; 917 fp->ipq_src = ip->ip_src; 918 fp->ipq_dst = ip->ip_dst; 919 fp->ipq_frags = m; 920 m->m_nextpkt = NULL; 921 goto inserted; 922 } else { 923 fp->ipq_nfrags++; 924 #ifdef MAC 925 mac_update_ipq(m, fp); 926 #endif 927 } 928 929 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 930 931 /* 932 * Handle ECN by comparing this segment with the first one; 933 * if CE is set, do not lose CE. 934 * drop if CE and not-ECT are mixed for the same packet. 935 */ 936 ecn = ip->ip_tos & IPTOS_ECN_MASK; 937 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 938 if (ecn == IPTOS_ECN_CE) { 939 if (ecn0 == IPTOS_ECN_NOTECT) 940 goto dropfrag; 941 if (ecn0 != IPTOS_ECN_CE) 942 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 943 } 944 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 945 goto dropfrag; 946 947 /* 948 * Find a segment which begins after this one does. 949 */ 950 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 951 if (GETIP(q)->ip_off > ip->ip_off) 952 break; 953 954 /* 955 * If there is a preceding segment, it may provide some of 956 * our data already. If so, drop the data from the incoming 957 * segment. If it provides all of our data, drop us, otherwise 958 * stick new segment in the proper place. 959 * 960 * If some of the data is dropped from the the preceding 961 * segment, then it's checksum is invalidated. 962 */ 963 if (p) { 964 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 965 if (i > 0) { 966 if (i >= ip->ip_len) 967 goto dropfrag; 968 m_adj(m, i); 969 m->m_pkthdr.csum_flags = 0; 970 ip->ip_off += i; 971 ip->ip_len -= i; 972 } 973 m->m_nextpkt = p->m_nextpkt; 974 p->m_nextpkt = m; 975 } else { 976 m->m_nextpkt = fp->ipq_frags; 977 fp->ipq_frags = m; 978 } 979 980 /* 981 * While we overlap succeeding segments trim them or, 982 * if they are completely covered, dequeue them. 983 */ 984 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 985 q = nq) { 986 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 987 if (i < GETIP(q)->ip_len) { 988 GETIP(q)->ip_len -= i; 989 GETIP(q)->ip_off += i; 990 m_adj(q, i); 991 q->m_pkthdr.csum_flags = 0; 992 break; 993 } 994 nq = q->m_nextpkt; 995 m->m_nextpkt = nq; 996 ipstat.ips_fragdropped++; 997 fp->ipq_nfrags--; 998 m_freem(q); 999 } 1000 1001 inserted: 1002 1003 /* 1004 * Check for complete reassembly and perform frag per packet 1005 * limiting. 1006 * 1007 * Frag limiting is performed here so that the nth frag has 1008 * a chance to complete the packet before we drop the packet. 1009 * As a result, n+1 frags are actually allowed per packet, but 1010 * only n will ever be stored. (n = maxfragsperpacket.) 1011 * 1012 */ 1013 next = 0; 1014 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1015 if (GETIP(q)->ip_off != next) { 1016 if (fp->ipq_nfrags > maxfragsperpacket) { 1017 ipstat.ips_fragdropped += fp->ipq_nfrags; 1018 ip_freef(head, fp); 1019 } 1020 goto done; 1021 } 1022 next += GETIP(q)->ip_len; 1023 } 1024 /* Make sure the last packet didn't have the IP_MF flag */ 1025 if (p->m_flags & M_FRAG) { 1026 if (fp->ipq_nfrags > maxfragsperpacket) { 1027 ipstat.ips_fragdropped += fp->ipq_nfrags; 1028 ip_freef(head, fp); 1029 } 1030 goto done; 1031 } 1032 1033 /* 1034 * Reassembly is complete. Make sure the packet is a sane size. 1035 */ 1036 q = fp->ipq_frags; 1037 ip = GETIP(q); 1038 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1039 ipstat.ips_toolong++; 1040 ipstat.ips_fragdropped += fp->ipq_nfrags; 1041 ip_freef(head, fp); 1042 goto done; 1043 } 1044 1045 /* 1046 * Concatenate fragments. 1047 */ 1048 m = q; 1049 t = m->m_next; 1050 m->m_next = NULL; 1051 m_cat(m, t); 1052 nq = q->m_nextpkt; 1053 q->m_nextpkt = NULL; 1054 for (q = nq; q != NULL; q = nq) { 1055 nq = q->m_nextpkt; 1056 q->m_nextpkt = NULL; 1057 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1058 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1059 m_cat(m, q); 1060 } 1061 #ifdef MAC 1062 mac_create_datagram_from_ipq(fp, m); 1063 mac_destroy_ipq(fp); 1064 #endif 1065 1066 /* 1067 * Create header for new ip packet by modifying header of first 1068 * packet; dequeue and discard fragment reassembly header. 1069 * Make header visible. 1070 */ 1071 ip->ip_len = (ip->ip_hl << 2) + next; 1072 ip->ip_src = fp->ipq_src; 1073 ip->ip_dst = fp->ipq_dst; 1074 TAILQ_REMOVE(head, fp, ipq_list); 1075 nipq--; 1076 (void) m_free(dtom(fp)); 1077 m->m_len += (ip->ip_hl << 2); 1078 m->m_data -= (ip->ip_hl << 2); 1079 /* some debugging cruft by sklower, below, will go away soon */ 1080 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1081 m_fixhdr(m); 1082 ipstat.ips_reassembled++; 1083 IPQ_UNLOCK(); 1084 return (m); 1085 1086 dropfrag: 1087 ipstat.ips_fragdropped++; 1088 if (fp != NULL) 1089 fp->ipq_nfrags--; 1090 m_freem(m); 1091 done: 1092 IPQ_UNLOCK(); 1093 return (NULL); 1094 1095 #undef GETIP 1096 } 1097 1098 /* 1099 * Free a fragment reassembly header and all 1100 * associated datagrams. 1101 */ 1102 static void 1103 ip_freef(fhp, fp) 1104 struct ipqhead *fhp; 1105 struct ipq *fp; 1106 { 1107 register struct mbuf *q; 1108 1109 IPQ_LOCK_ASSERT(); 1110 1111 while (fp->ipq_frags) { 1112 q = fp->ipq_frags; 1113 fp->ipq_frags = q->m_nextpkt; 1114 m_freem(q); 1115 } 1116 TAILQ_REMOVE(fhp, fp, ipq_list); 1117 (void) m_free(dtom(fp)); 1118 nipq--; 1119 } 1120 1121 /* 1122 * IP timer processing; 1123 * if a timer expires on a reassembly 1124 * queue, discard it. 1125 */ 1126 void 1127 ip_slowtimo() 1128 { 1129 register struct ipq *fp; 1130 int s = splnet(); 1131 int i; 1132 1133 IPQ_LOCK(); 1134 for (i = 0; i < IPREASS_NHASH; i++) { 1135 for(fp = TAILQ_FIRST(&ipq[i]); fp;) { 1136 struct ipq *fpp; 1137 1138 fpp = fp; 1139 fp = TAILQ_NEXT(fp, ipq_list); 1140 if(--fpp->ipq_ttl == 0) { 1141 ipstat.ips_fragtimeout += fpp->ipq_nfrags; 1142 ip_freef(&ipq[i], fpp); 1143 } 1144 } 1145 } 1146 /* 1147 * If we are over the maximum number of fragments 1148 * (due to the limit being lowered), drain off 1149 * enough to get down to the new limit. 1150 */ 1151 if (maxnipq >= 0 && nipq > maxnipq) { 1152 for (i = 0; i < IPREASS_NHASH; i++) { 1153 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) { 1154 ipstat.ips_fragdropped += 1155 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1156 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1157 } 1158 } 1159 } 1160 IPQ_UNLOCK(); 1161 splx(s); 1162 } 1163 1164 /* 1165 * Drain off all datagram fragments. 1166 */ 1167 void 1168 ip_drain() 1169 { 1170 int i; 1171 1172 IPQ_LOCK(); 1173 for (i = 0; i < IPREASS_NHASH; i++) { 1174 while(!TAILQ_EMPTY(&ipq[i])) { 1175 ipstat.ips_fragdropped += 1176 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1177 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1178 } 1179 } 1180 IPQ_UNLOCK(); 1181 in_rtqdrain(); 1182 } 1183 1184 /* 1185 * The protocol to be inserted into ip_protox[] must be already registered 1186 * in inetsw[], either statically or through pf_proto_register(). 1187 */ 1188 int 1189 ipproto_register(u_char ipproto) 1190 { 1191 struct protosw *pr; 1192 1193 /* Sanity checks. */ 1194 if (ipproto == 0) 1195 return (EPROTONOSUPPORT); 1196 1197 /* 1198 * The protocol slot must not be occupied by another protocol 1199 * already. An index pointing to IPPROTO_RAW is unused. 1200 */ 1201 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1202 if (pr == NULL) 1203 return (EPFNOSUPPORT); 1204 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1205 return (EEXIST); 1206 1207 /* Find the protocol position in inetsw[] and set the index. */ 1208 for (pr = inetdomain.dom_protosw; 1209 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1210 if (pr->pr_domain->dom_family == PF_INET && 1211 pr->pr_protocol && pr->pr_protocol == ipproto) { 1212 /* Be careful to only index valid IP protocols. */ 1213 if (pr->pr_protocol < IPPROTO_MAX) { 1214 ip_protox[pr->pr_protocol] = pr - inetsw; 1215 return (0); 1216 } else 1217 return (EINVAL); 1218 } 1219 } 1220 return (EPROTONOSUPPORT); 1221 } 1222 1223 int 1224 ipproto_unregister(u_char ipproto) 1225 { 1226 struct protosw *pr; 1227 1228 /* Sanity checks. */ 1229 if (ipproto == 0) 1230 return (EPROTONOSUPPORT); 1231 1232 /* Check if the protocol was indeed registered. */ 1233 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1234 if (pr == NULL) 1235 return (EPFNOSUPPORT); 1236 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1237 return (ENOENT); 1238 1239 /* Reset the protocol slot to IPPROTO_RAW. */ 1240 ip_protox[ipproto] = pr - inetsw; 1241 return (0); 1242 } 1243 1244 1245 /* 1246 * Do option processing on a datagram, 1247 * possibly discarding it if bad options are encountered, 1248 * or forwarding it if source-routed. 1249 * The pass argument is used when operating in the IPSTEALTH 1250 * mode to tell what options to process: 1251 * [LS]SRR (pass 0) or the others (pass 1). 1252 * The reason for as many as two passes is that when doing IPSTEALTH, 1253 * non-routing options should be processed only if the packet is for us. 1254 * Returns 1 if packet has been forwarded/freed, 1255 * 0 if the packet should be processed further. 1256 */ 1257 static int 1258 ip_dooptions(struct mbuf *m, int pass) 1259 { 1260 struct ip *ip = mtod(m, struct ip *); 1261 u_char *cp; 1262 struct in_ifaddr *ia; 1263 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1264 struct in_addr *sin, dst; 1265 n_time ntime; 1266 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 1267 1268 /* ignore or reject packets with IP options */ 1269 if (ip_doopts == 0) 1270 return 0; 1271 else if (ip_doopts == 2) { 1272 type = ICMP_UNREACH; 1273 code = ICMP_UNREACH_FILTER_PROHIB; 1274 goto bad; 1275 } 1276 1277 dst = ip->ip_dst; 1278 cp = (u_char *)(ip + 1); 1279 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1280 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1281 opt = cp[IPOPT_OPTVAL]; 1282 if (opt == IPOPT_EOL) 1283 break; 1284 if (opt == IPOPT_NOP) 1285 optlen = 1; 1286 else { 1287 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1288 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1289 goto bad; 1290 } 1291 optlen = cp[IPOPT_OLEN]; 1292 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1293 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1294 goto bad; 1295 } 1296 } 1297 switch (opt) { 1298 1299 default: 1300 break; 1301 1302 /* 1303 * Source routing with record. 1304 * Find interface with current destination address. 1305 * If none on this machine then drop if strictly routed, 1306 * or do nothing if loosely routed. 1307 * Record interface address and bring up next address 1308 * component. If strictly routed make sure next 1309 * address is on directly accessible net. 1310 */ 1311 case IPOPT_LSRR: 1312 case IPOPT_SSRR: 1313 #ifdef IPSTEALTH 1314 if (ipstealth && pass > 0) 1315 break; 1316 #endif 1317 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1318 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1319 goto bad; 1320 } 1321 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1322 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1323 goto bad; 1324 } 1325 ipaddr.sin_addr = ip->ip_dst; 1326 ia = (struct in_ifaddr *) 1327 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1328 if (ia == NULL) { 1329 if (opt == IPOPT_SSRR) { 1330 type = ICMP_UNREACH; 1331 code = ICMP_UNREACH_SRCFAIL; 1332 goto bad; 1333 } 1334 if (!ip_dosourceroute) 1335 goto nosourcerouting; 1336 /* 1337 * Loose routing, and not at next destination 1338 * yet; nothing to do except forward. 1339 */ 1340 break; 1341 } 1342 off--; /* 0 origin */ 1343 if (off > optlen - (int)sizeof(struct in_addr)) { 1344 /* 1345 * End of source route. Should be for us. 1346 */ 1347 if (!ip_acceptsourceroute) 1348 goto nosourcerouting; 1349 save_rte(m, cp, ip->ip_src); 1350 break; 1351 } 1352 #ifdef IPSTEALTH 1353 if (ipstealth) 1354 goto dropit; 1355 #endif 1356 if (!ip_dosourceroute) { 1357 if (ipforwarding) { 1358 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1359 /* 1360 * Acting as a router, so generate ICMP 1361 */ 1362 nosourcerouting: 1363 strcpy(buf, inet_ntoa(ip->ip_dst)); 1364 log(LOG_WARNING, 1365 "attempted source route from %s to %s\n", 1366 inet_ntoa(ip->ip_src), buf); 1367 type = ICMP_UNREACH; 1368 code = ICMP_UNREACH_SRCFAIL; 1369 goto bad; 1370 } else { 1371 /* 1372 * Not acting as a router, so silently drop. 1373 */ 1374 #ifdef IPSTEALTH 1375 dropit: 1376 #endif 1377 ipstat.ips_cantforward++; 1378 m_freem(m); 1379 return (1); 1380 } 1381 } 1382 1383 /* 1384 * locate outgoing interface 1385 */ 1386 (void)memcpy(&ipaddr.sin_addr, cp + off, 1387 sizeof(ipaddr.sin_addr)); 1388 1389 if (opt == IPOPT_SSRR) { 1390 #define INA struct in_ifaddr * 1391 #define SA struct sockaddr * 1392 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == NULL) 1393 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1394 } else 1395 ia = ip_rtaddr(ipaddr.sin_addr); 1396 if (ia == NULL) { 1397 type = ICMP_UNREACH; 1398 code = ICMP_UNREACH_SRCFAIL; 1399 goto bad; 1400 } 1401 ip->ip_dst = ipaddr.sin_addr; 1402 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1403 sizeof(struct in_addr)); 1404 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1405 /* 1406 * Let ip_intr's mcast routing check handle mcast pkts 1407 */ 1408 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1409 break; 1410 1411 case IPOPT_RR: 1412 #ifdef IPSTEALTH 1413 if (ipstealth && pass == 0) 1414 break; 1415 #endif 1416 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1417 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1418 goto bad; 1419 } 1420 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1421 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1422 goto bad; 1423 } 1424 /* 1425 * If no space remains, ignore. 1426 */ 1427 off--; /* 0 origin */ 1428 if (off > optlen - (int)sizeof(struct in_addr)) 1429 break; 1430 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1431 sizeof(ipaddr.sin_addr)); 1432 /* 1433 * locate outgoing interface; if we're the destination, 1434 * use the incoming interface (should be same). 1435 */ 1436 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL && 1437 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1438 type = ICMP_UNREACH; 1439 code = ICMP_UNREACH_HOST; 1440 goto bad; 1441 } 1442 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1443 sizeof(struct in_addr)); 1444 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1445 break; 1446 1447 case IPOPT_TS: 1448 #ifdef IPSTEALTH 1449 if (ipstealth && pass == 0) 1450 break; 1451 #endif 1452 code = cp - (u_char *)ip; 1453 if (optlen < 4 || optlen > 40) { 1454 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1455 goto bad; 1456 } 1457 if ((off = cp[IPOPT_OFFSET]) < 5) { 1458 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1459 goto bad; 1460 } 1461 if (off > optlen - (int)sizeof(int32_t)) { 1462 cp[IPOPT_OFFSET + 1] += (1 << 4); 1463 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1464 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1465 goto bad; 1466 } 1467 break; 1468 } 1469 off--; /* 0 origin */ 1470 sin = (struct in_addr *)(cp + off); 1471 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1472 1473 case IPOPT_TS_TSONLY: 1474 break; 1475 1476 case IPOPT_TS_TSANDADDR: 1477 if (off + sizeof(n_time) + 1478 sizeof(struct in_addr) > optlen) { 1479 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1480 goto bad; 1481 } 1482 ipaddr.sin_addr = dst; 1483 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1484 m->m_pkthdr.rcvif); 1485 if (ia == NULL) 1486 continue; 1487 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1488 sizeof(struct in_addr)); 1489 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1490 off += sizeof(struct in_addr); 1491 break; 1492 1493 case IPOPT_TS_PRESPEC: 1494 if (off + sizeof(n_time) + 1495 sizeof(struct in_addr) > optlen) { 1496 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1497 goto bad; 1498 } 1499 (void)memcpy(&ipaddr.sin_addr, sin, 1500 sizeof(struct in_addr)); 1501 if (ifa_ifwithaddr((SA)&ipaddr) == NULL) 1502 continue; 1503 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1504 off += sizeof(struct in_addr); 1505 break; 1506 1507 default: 1508 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1509 goto bad; 1510 } 1511 ntime = iptime(); 1512 (void)memcpy(cp + off, &ntime, sizeof(n_time)); 1513 cp[IPOPT_OFFSET] += sizeof(n_time); 1514 } 1515 } 1516 if (forward && ipforwarding) { 1517 ip_forward(m, 1); 1518 return (1); 1519 } 1520 return (0); 1521 bad: 1522 icmp_error(m, type, code, 0, 0); 1523 ipstat.ips_badoptions++; 1524 return (1); 1525 } 1526 1527 /* 1528 * Given address of next destination (final or next hop), 1529 * return internet address info of interface to be used to get there. 1530 */ 1531 struct in_ifaddr * 1532 ip_rtaddr(dst) 1533 struct in_addr dst; 1534 { 1535 struct route sro; 1536 struct sockaddr_in *sin; 1537 struct in_ifaddr *ifa; 1538 1539 bzero(&sro, sizeof(sro)); 1540 sin = (struct sockaddr_in *)&sro.ro_dst; 1541 sin->sin_family = AF_INET; 1542 sin->sin_len = sizeof(*sin); 1543 sin->sin_addr = dst; 1544 rtalloc_ign(&sro, RTF_CLONING); 1545 1546 if (sro.ro_rt == NULL) 1547 return (NULL); 1548 1549 ifa = ifatoia(sro.ro_rt->rt_ifa); 1550 RTFREE(sro.ro_rt); 1551 return (ifa); 1552 } 1553 1554 /* 1555 * Save incoming source route for use in replies, 1556 * to be picked up later by ip_srcroute if the receiver is interested. 1557 */ 1558 static void 1559 save_rte(m, option, dst) 1560 struct mbuf *m; 1561 u_char *option; 1562 struct in_addr dst; 1563 { 1564 unsigned olen; 1565 struct ipopt_tag *opts; 1566 1567 opts = (struct ipopt_tag *)m_tag_get(PACKET_TAG_IPOPTIONS, 1568 sizeof(struct ipopt_tag), M_NOWAIT); 1569 if (opts == NULL) 1570 return; 1571 1572 olen = option[IPOPT_OLEN]; 1573 #ifdef DIAGNOSTIC 1574 if (ipprintfs) 1575 printf("save_rte: olen %d\n", olen); 1576 #endif 1577 if (olen > sizeof(opts->ip_srcrt) - (1 + sizeof(dst))) { 1578 m_tag_free((struct m_tag *)opts); 1579 return; 1580 } 1581 bcopy(option, opts->ip_srcrt.srcopt, olen); 1582 opts->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1583 opts->ip_srcrt.dst = dst; 1584 m_tag_prepend(m, (struct m_tag *)opts); 1585 } 1586 1587 /* 1588 * Retrieve incoming source route for use in replies, 1589 * in the same form used by setsockopt. 1590 * The first hop is placed before the options, will be removed later. 1591 */ 1592 struct mbuf * 1593 ip_srcroute(m0) 1594 struct mbuf *m0; 1595 { 1596 register struct in_addr *p, *q; 1597 register struct mbuf *m; 1598 struct ipopt_tag *opts; 1599 1600 opts = (struct ipopt_tag *)m_tag_find(m0, PACKET_TAG_IPOPTIONS, NULL); 1601 if (opts == NULL) 1602 return (NULL); 1603 1604 if (opts->ip_nhops == 0) 1605 return (NULL); 1606 m = m_get(M_DONTWAIT, MT_HEADER); 1607 if (m == NULL) 1608 return (NULL); 1609 1610 #define OPTSIZ (sizeof(opts->ip_srcrt.nop) + sizeof(opts->ip_srcrt.srcopt)) 1611 1612 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1613 m->m_len = opts->ip_nhops * sizeof(struct in_addr) + 1614 sizeof(struct in_addr) + OPTSIZ; 1615 #ifdef DIAGNOSTIC 1616 if (ipprintfs) 1617 printf("ip_srcroute: nhops %d mlen %d", opts->ip_nhops, m->m_len); 1618 #endif 1619 1620 /* 1621 * First save first hop for return route 1622 */ 1623 p = &(opts->ip_srcrt.route[opts->ip_nhops - 1]); 1624 *(mtod(m, struct in_addr *)) = *p--; 1625 #ifdef DIAGNOSTIC 1626 if (ipprintfs) 1627 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1628 #endif 1629 1630 /* 1631 * Copy option fields and padding (nop) to mbuf. 1632 */ 1633 opts->ip_srcrt.nop = IPOPT_NOP; 1634 opts->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1635 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1636 &(opts->ip_srcrt.nop), OPTSIZ); 1637 q = (struct in_addr *)(mtod(m, caddr_t) + 1638 sizeof(struct in_addr) + OPTSIZ); 1639 #undef OPTSIZ 1640 /* 1641 * Record return path as an IP source route, 1642 * reversing the path (pointers are now aligned). 1643 */ 1644 while (p >= opts->ip_srcrt.route) { 1645 #ifdef DIAGNOSTIC 1646 if (ipprintfs) 1647 printf(" %lx", (u_long)ntohl(q->s_addr)); 1648 #endif 1649 *q++ = *p--; 1650 } 1651 /* 1652 * Last hop goes to final destination. 1653 */ 1654 *q = opts->ip_srcrt.dst; 1655 #ifdef DIAGNOSTIC 1656 if (ipprintfs) 1657 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1658 #endif 1659 m_tag_delete(m0, (struct m_tag *)opts); 1660 return (m); 1661 } 1662 1663 /* 1664 * Strip out IP options, at higher 1665 * level protocol in the kernel. 1666 * Second argument is buffer to which options 1667 * will be moved, and return value is their length. 1668 * XXX should be deleted; last arg currently ignored. 1669 */ 1670 void 1671 ip_stripoptions(m, mopt) 1672 register struct mbuf *m; 1673 struct mbuf *mopt; 1674 { 1675 register int i; 1676 struct ip *ip = mtod(m, struct ip *); 1677 register caddr_t opts; 1678 int olen; 1679 1680 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1681 opts = (caddr_t)(ip + 1); 1682 i = m->m_len - (sizeof (struct ip) + olen); 1683 bcopy(opts + olen, opts, (unsigned)i); 1684 m->m_len -= olen; 1685 if (m->m_flags & M_PKTHDR) 1686 m->m_pkthdr.len -= olen; 1687 ip->ip_v = IPVERSION; 1688 ip->ip_hl = sizeof(struct ip) >> 2; 1689 } 1690 1691 u_char inetctlerrmap[PRC_NCMDS] = { 1692 0, 0, 0, 0, 1693 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1694 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1695 EMSGSIZE, EHOSTUNREACH, 0, 0, 1696 0, 0, EHOSTUNREACH, 0, 1697 ENOPROTOOPT, ECONNREFUSED 1698 }; 1699 1700 /* 1701 * Forward a packet. If some error occurs return the sender 1702 * an icmp packet. Note we can't always generate a meaningful 1703 * icmp message because icmp doesn't have a large enough repertoire 1704 * of codes and types. 1705 * 1706 * If not forwarding, just drop the packet. This could be confusing 1707 * if ipforwarding was zero but some routing protocol was advancing 1708 * us as a gateway to somewhere. However, we must let the routing 1709 * protocol deal with that. 1710 * 1711 * The srcrt parameter indicates whether the packet is being forwarded 1712 * via a source route. 1713 */ 1714 void 1715 ip_forward(struct mbuf *m, int srcrt) 1716 { 1717 struct ip *ip = mtod(m, struct ip *); 1718 struct in_ifaddr *ia = NULL; 1719 int error, type = 0, code = 0; 1720 struct mbuf *mcopy; 1721 struct in_addr dest; 1722 struct ifnet *destifp, dummyifp; 1723 1724 #ifdef DIAGNOSTIC 1725 if (ipprintfs) 1726 printf("forward: src %lx dst %lx ttl %x\n", 1727 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr, 1728 ip->ip_ttl); 1729 #endif 1730 1731 1732 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1733 ipstat.ips_cantforward++; 1734 m_freem(m); 1735 return; 1736 } 1737 #ifdef IPSTEALTH 1738 if (!ipstealth) { 1739 #endif 1740 if (ip->ip_ttl <= IPTTLDEC) { 1741 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1742 0, 0); 1743 return; 1744 } 1745 #ifdef IPSTEALTH 1746 } 1747 #endif 1748 1749 if (!srcrt && (ia = ip_rtaddr(ip->ip_dst)) == NULL) { 1750 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1751 return; 1752 } 1753 1754 /* 1755 * Save the IP header and at most 8 bytes of the payload, 1756 * in case we need to generate an ICMP message to the src. 1757 * 1758 * XXX this can be optimized a lot by saving the data in a local 1759 * buffer on the stack (72 bytes at most), and only allocating the 1760 * mbuf if really necessary. The vast majority of the packets 1761 * are forwarded without having to send an ICMP back (either 1762 * because unnecessary, or because rate limited), so we are 1763 * really we are wasting a lot of work here. 1764 * 1765 * We don't use m_copy() because it might return a reference 1766 * to a shared cluster. Both this function and ip_output() 1767 * assume exclusive access to the IP header in `m', so any 1768 * data in a cluster may change before we reach icmp_error(). 1769 */ 1770 MGET(mcopy, M_DONTWAIT, m->m_type); 1771 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1772 /* 1773 * It's probably ok if the pkthdr dup fails (because 1774 * the deep copy of the tag chain failed), but for now 1775 * be conservative and just discard the copy since 1776 * code below may some day want the tags. 1777 */ 1778 m_free(mcopy); 1779 mcopy = NULL; 1780 } 1781 if (mcopy != NULL) { 1782 mcopy->m_len = imin((ip->ip_hl << 2) + 8, 1783 (int)ip->ip_len); 1784 mcopy->m_pkthdr.len = mcopy->m_len; 1785 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1786 } 1787 1788 #ifdef IPSTEALTH 1789 if (!ipstealth) { 1790 #endif 1791 ip->ip_ttl -= IPTTLDEC; 1792 #ifdef IPSTEALTH 1793 } 1794 #endif 1795 1796 /* 1797 * If forwarding packet using same interface that it came in on, 1798 * perhaps should send a redirect to sender to shortcut a hop. 1799 * Only send redirect if source is sending directly to us, 1800 * and if packet was not source routed (or has any options). 1801 * Also, don't send redirect if forwarding using a default route 1802 * or a route modified by a redirect. 1803 */ 1804 dest.s_addr = 0; 1805 if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) { 1806 struct sockaddr_in *sin; 1807 struct route ro; 1808 struct rtentry *rt; 1809 1810 bzero(&ro, sizeof(ro)); 1811 sin = (struct sockaddr_in *)&ro.ro_dst; 1812 sin->sin_family = AF_INET; 1813 sin->sin_len = sizeof(*sin); 1814 sin->sin_addr = ip->ip_dst; 1815 rtalloc_ign(&ro, RTF_CLONING); 1816 1817 rt = ro.ro_rt; 1818 1819 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1820 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1821 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1822 u_long src = ntohl(ip->ip_src.s_addr); 1823 1824 if (RTA(rt) && 1825 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1826 if (rt->rt_flags & RTF_GATEWAY) 1827 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1828 else 1829 dest.s_addr = ip->ip_dst.s_addr; 1830 /* Router requirements says to only send host redirects */ 1831 type = ICMP_REDIRECT; 1832 code = ICMP_REDIRECT_HOST; 1833 #ifdef DIAGNOSTIC 1834 if (ipprintfs) 1835 printf("redirect (%d) to %lx\n", code, (u_long)dest.s_addr); 1836 #endif 1837 } 1838 } 1839 if (rt) 1840 RTFREE(rt); 1841 } 1842 1843 error = ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL); 1844 if (error) 1845 ipstat.ips_cantforward++; 1846 else { 1847 ipstat.ips_forward++; 1848 if (type) 1849 ipstat.ips_redirectsent++; 1850 else { 1851 if (mcopy) 1852 m_freem(mcopy); 1853 return; 1854 } 1855 } 1856 if (mcopy == NULL) 1857 return; 1858 destifp = NULL; 1859 1860 switch (error) { 1861 1862 case 0: /* forwarded, but need redirect */ 1863 /* type, code set above */ 1864 break; 1865 1866 case ENETUNREACH: /* shouldn't happen, checked above */ 1867 case EHOSTUNREACH: 1868 case ENETDOWN: 1869 case EHOSTDOWN: 1870 default: 1871 type = ICMP_UNREACH; 1872 code = ICMP_UNREACH_HOST; 1873 break; 1874 1875 case EMSGSIZE: 1876 type = ICMP_UNREACH; 1877 code = ICMP_UNREACH_NEEDFRAG; 1878 #if defined(IPSEC) || defined(FAST_IPSEC) 1879 /* 1880 * If the packet is routed over IPsec tunnel, tell the 1881 * originator the tunnel MTU. 1882 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1883 * XXX quickhack!!! 1884 */ 1885 { 1886 struct secpolicy *sp = NULL; 1887 int ipsecerror; 1888 int ipsechdr; 1889 struct route *ro; 1890 1891 #ifdef IPSEC 1892 sp = ipsec4_getpolicybyaddr(mcopy, 1893 IPSEC_DIR_OUTBOUND, 1894 IP_FORWARDING, 1895 &ipsecerror); 1896 #else /* FAST_IPSEC */ 1897 sp = ipsec_getpolicybyaddr(mcopy, 1898 IPSEC_DIR_OUTBOUND, 1899 IP_FORWARDING, 1900 &ipsecerror); 1901 #endif 1902 if (sp != NULL) { 1903 /* count IPsec header size */ 1904 ipsechdr = ipsec4_hdrsiz(mcopy, 1905 IPSEC_DIR_OUTBOUND, 1906 NULL); 1907 1908 /* 1909 * find the correct route for outer IPv4 1910 * header, compute tunnel MTU. 1911 * 1912 * XXX BUG ALERT 1913 * The "dummyifp" code relies upon the fact 1914 * that icmp_error() touches only ifp->if_mtu. 1915 */ 1916 /*XXX*/ 1917 destifp = NULL; 1918 if (sp->req != NULL 1919 && sp->req->sav != NULL 1920 && sp->req->sav->sah != NULL) { 1921 ro = &sp->req->sav->sah->sa_route; 1922 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1923 dummyifp.if_mtu = 1924 ro->ro_rt->rt_rmx.rmx_mtu ? 1925 ro->ro_rt->rt_rmx.rmx_mtu : 1926 ro->ro_rt->rt_ifp->if_mtu; 1927 dummyifp.if_mtu -= ipsechdr; 1928 destifp = &dummyifp; 1929 } 1930 } 1931 1932 #ifdef IPSEC 1933 key_freesp(sp); 1934 #else /* FAST_IPSEC */ 1935 KEY_FREESP(&sp); 1936 #endif 1937 ipstat.ips_cantfrag++; 1938 break; 1939 } else 1940 #endif /*IPSEC || FAST_IPSEC*/ 1941 /* 1942 * When doing source routing 'ia' can be NULL. Fall back 1943 * to the minimum guaranteed routeable packet size and use 1944 * the same hack as IPSEC to setup a dummyifp for icmp. 1945 */ 1946 if (ia == NULL) { 1947 dummyifp.if_mtu = IP_MSS; 1948 destifp = &dummyifp; 1949 } else 1950 destifp = ia->ia_ifp; 1951 #if defined(IPSEC) || defined(FAST_IPSEC) 1952 } 1953 #endif /*IPSEC || FAST_IPSEC*/ 1954 ipstat.ips_cantfrag++; 1955 break; 1956 1957 case ENOBUFS: 1958 /* 1959 * A router should not generate ICMP_SOURCEQUENCH as 1960 * required in RFC1812 Requirements for IP Version 4 Routers. 1961 * Source quench could be a big problem under DoS attacks, 1962 * or if the underlying interface is rate-limited. 1963 * Those who need source quench packets may re-enable them 1964 * via the net.inet.ip.sendsourcequench sysctl. 1965 */ 1966 if (ip_sendsourcequench == 0) { 1967 m_freem(mcopy); 1968 return; 1969 } else { 1970 type = ICMP_SOURCEQUENCH; 1971 code = 0; 1972 } 1973 break; 1974 1975 case EACCES: /* ipfw denied packet */ 1976 m_freem(mcopy); 1977 return; 1978 } 1979 icmp_error(mcopy, type, code, dest.s_addr, destifp); 1980 } 1981 1982 void 1983 ip_savecontrol(inp, mp, ip, m) 1984 register struct inpcb *inp; 1985 register struct mbuf **mp; 1986 register struct ip *ip; 1987 register struct mbuf *m; 1988 { 1989 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1990 struct bintime bt; 1991 1992 bintime(&bt); 1993 if (inp->inp_socket->so_options & SO_BINTIME) { 1994 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1995 SCM_BINTIME, SOL_SOCKET); 1996 if (*mp) 1997 mp = &(*mp)->m_next; 1998 } 1999 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2000 struct timeval tv; 2001 2002 bintime2timeval(&bt, &tv); 2003 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2004 SCM_TIMESTAMP, SOL_SOCKET); 2005 if (*mp) 2006 mp = &(*mp)->m_next; 2007 } 2008 } 2009 if (inp->inp_flags & INP_RECVDSTADDR) { 2010 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2011 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2012 if (*mp) 2013 mp = &(*mp)->m_next; 2014 } 2015 if (inp->inp_flags & INP_RECVTTL) { 2016 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 2017 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 2018 if (*mp) 2019 mp = &(*mp)->m_next; 2020 } 2021 #ifdef notyet 2022 /* XXX 2023 * Moving these out of udp_input() made them even more broken 2024 * than they already were. 2025 */ 2026 /* options were tossed already */ 2027 if (inp->inp_flags & INP_RECVOPTS) { 2028 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2029 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2030 if (*mp) 2031 mp = &(*mp)->m_next; 2032 } 2033 /* ip_srcroute doesn't do what we want here, need to fix */ 2034 if (inp->inp_flags & INP_RECVRETOPTS) { 2035 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 2036 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2037 if (*mp) 2038 mp = &(*mp)->m_next; 2039 } 2040 #endif 2041 if (inp->inp_flags & INP_RECVIF) { 2042 struct ifnet *ifp; 2043 struct sdlbuf { 2044 struct sockaddr_dl sdl; 2045 u_char pad[32]; 2046 } sdlbuf; 2047 struct sockaddr_dl *sdp; 2048 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2049 2050 if (((ifp = m->m_pkthdr.rcvif)) 2051 && ( ifp->if_index && (ifp->if_index <= if_index))) { 2052 sdp = (struct sockaddr_dl *) 2053 (ifaddr_byindex(ifp->if_index)->ifa_addr); 2054 /* 2055 * Change our mind and don't try copy. 2056 */ 2057 if ((sdp->sdl_family != AF_LINK) 2058 || (sdp->sdl_len > sizeof(sdlbuf))) { 2059 goto makedummy; 2060 } 2061 bcopy(sdp, sdl2, sdp->sdl_len); 2062 } else { 2063 makedummy: 2064 sdl2->sdl_len 2065 = offsetof(struct sockaddr_dl, sdl_data[0]); 2066 sdl2->sdl_family = AF_LINK; 2067 sdl2->sdl_index = 0; 2068 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2069 } 2070 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 2071 IP_RECVIF, IPPROTO_IP); 2072 if (*mp) 2073 mp = &(*mp)->m_next; 2074 } 2075 } 2076 2077 /* 2078 * XXX these routines are called from the upper part of the kernel. 2079 * They need to be locked when we remove Giant. 2080 * 2081 * They could also be moved to ip_mroute.c, since all the RSVP 2082 * handling is done there already. 2083 */ 2084 static int ip_rsvp_on; 2085 struct socket *ip_rsvpd; 2086 int 2087 ip_rsvp_init(struct socket *so) 2088 { 2089 if (so->so_type != SOCK_RAW || 2090 so->so_proto->pr_protocol != IPPROTO_RSVP) 2091 return EOPNOTSUPP; 2092 2093 if (ip_rsvpd != NULL) 2094 return EADDRINUSE; 2095 2096 ip_rsvpd = so; 2097 /* 2098 * This may seem silly, but we need to be sure we don't over-increment 2099 * the RSVP counter, in case something slips up. 2100 */ 2101 if (!ip_rsvp_on) { 2102 ip_rsvp_on = 1; 2103 rsvp_on++; 2104 } 2105 2106 return 0; 2107 } 2108 2109 int 2110 ip_rsvp_done(void) 2111 { 2112 ip_rsvpd = NULL; 2113 /* 2114 * This may seem silly, but we need to be sure we don't over-decrement 2115 * the RSVP counter, in case something slips up. 2116 */ 2117 if (ip_rsvp_on) { 2118 ip_rsvp_on = 0; 2119 rsvp_on--; 2120 } 2121 return 0; 2122 } 2123 2124 void 2125 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 2126 { 2127 if (rsvp_input_p) { /* call the real one if loaded */ 2128 rsvp_input_p(m, off); 2129 return; 2130 } 2131 2132 /* Can still get packets with rsvp_on = 0 if there is a local member 2133 * of the group to which the RSVP packet is addressed. But in this 2134 * case we want to throw the packet away. 2135 */ 2136 2137 if (!rsvp_on) { 2138 m_freem(m); 2139 return; 2140 } 2141 2142 if (ip_rsvpd != NULL) { 2143 rip_input(m, off); 2144 return; 2145 } 2146 /* Drop the packet */ 2147 m_freem(m); 2148 } 2149