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 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 34 * $FreeBSD$ 35 */ 36 37 #include "opt_bootp.h" 38 #include "opt_ipfw.h" 39 #include "opt_ipdn.h" 40 #include "opt_ipdivert.h" 41 #include "opt_ipfilter.h" 42 #include "opt_ipstealth.h" 43 #include "opt_ipsec.h" 44 #include "opt_mac.h" 45 #include "opt_pfil_hooks.h" 46 #include "opt_random_ip_id.h" 47 48 #include <sys/param.h> 49 #include <sys/systm.h> 50 #include <sys/mac.h> 51 #include <sys/mbuf.h> 52 #include <sys/malloc.h> 53 #include <sys/domain.h> 54 #include <sys/protosw.h> 55 #include <sys/socket.h> 56 #include <sys/time.h> 57 #include <sys/kernel.h> 58 #include <sys/syslog.h> 59 #include <sys/sysctl.h> 60 61 #include <net/pfil.h> 62 #include <net/if.h> 63 #include <net/if_types.h> 64 #include <net/if_var.h> 65 #include <net/if_dl.h> 66 #include <net/route.h> 67 #include <net/netisr.h> 68 #include <net/intrq.h> 69 70 #include <netinet/in.h> 71 #include <netinet/in_systm.h> 72 #include <netinet/in_var.h> 73 #include <netinet/ip.h> 74 #include <netinet/in_pcb.h> 75 #include <netinet/ip_var.h> 76 #include <netinet/ip_icmp.h> 77 #include <machine/in_cksum.h> 78 79 #include <sys/socketvar.h> 80 81 #include <netinet/ip_fw.h> 82 #include <netinet/ip_dummynet.h> 83 84 #ifdef IPSEC 85 #include <netinet6/ipsec.h> 86 #include <netkey/key.h> 87 #endif 88 89 #ifdef FAST_IPSEC 90 #include <netipsec/ipsec.h> 91 #include <netipsec/key.h> 92 #endif 93 94 int rsvp_on = 0; 95 96 int ipforwarding = 0; 97 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 98 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 99 100 static int ipsendredirects = 1; /* XXX */ 101 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 102 &ipsendredirects, 0, "Enable sending IP redirects"); 103 104 int ip_defttl = IPDEFTTL; 105 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 106 &ip_defttl, 0, "Maximum TTL on IP packets"); 107 108 static int ip_dosourceroute = 0; 109 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 110 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 111 112 static int ip_acceptsourceroute = 0; 113 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 114 CTLFLAG_RW, &ip_acceptsourceroute, 0, 115 "Enable accepting source routed IP packets"); 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 ip_nfragpackets = 0; 123 static int ip_maxfragpackets; /* initialized in ip_init() */ 124 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 125 &ip_maxfragpackets, 0, 126 "Maximum number of IPv4 fragment reassembly queue entries"); 127 128 static int ip_sendsourcequench = 0; 129 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 130 &ip_sendsourcequench, 0, 131 "Enable the transmission of source quench packets"); 132 133 /* 134 * XXX - Setting ip_checkinterface mostly implements the receive side of 135 * the Strong ES model described in RFC 1122, but since the routing table 136 * and transmit implementation do not implement the Strong ES model, 137 * setting this to 1 results in an odd hybrid. 138 * 139 * XXX - ip_checkinterface currently must be disabled if you use ipnat 140 * to translate the destination address to another local interface. 141 * 142 * XXX - ip_checkinterface must be disabled if you add IP aliases 143 * to the loopback interface instead of the interface where the 144 * packets for those addresses are received. 145 */ 146 static int ip_checkinterface = 1; 147 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 148 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 149 150 #ifdef DIAGNOSTIC 151 static int ipprintfs = 0; 152 #endif 153 154 static int ipqmaxlen = IFQ_MAXLEN; 155 156 extern struct domain inetdomain; 157 extern struct protosw inetsw[]; 158 u_char ip_protox[IPPROTO_MAX]; 159 struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 160 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 161 u_long in_ifaddrhmask; /* mask for hash table */ 162 163 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 164 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 165 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 166 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 167 168 struct ipstat ipstat; 169 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 170 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 171 172 /* Packet reassembly stuff */ 173 #define IPREASS_NHASH_LOG2 6 174 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 175 #define IPREASS_HMASK (IPREASS_NHASH - 1) 176 #define IPREASS_HASH(x,y) \ 177 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 178 179 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 180 static int nipq = 0; /* total # of reass queues */ 181 static int maxnipq; 182 183 #ifdef IPCTL_DEFMTU 184 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 185 &ip_mtu, 0, "Default MTU"); 186 #endif 187 188 #ifdef IPSTEALTH 189 static int ipstealth = 0; 190 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 191 &ipstealth, 0, ""); 192 #endif 193 194 195 /* Firewall hooks */ 196 ip_fw_chk_t *ip_fw_chk_ptr; 197 int fw_enable = 1 ; 198 int fw_one_pass = 1; 199 200 /* Dummynet hooks */ 201 ip_dn_io_t *ip_dn_io_ptr; 202 203 204 /* 205 * XXX this is ugly -- the following two global variables are 206 * used to store packet state while it travels through the stack. 207 * Note that the code even makes assumptions on the size and 208 * alignment of fields inside struct ip_srcrt so e.g. adding some 209 * fields will break the code. This needs to be fixed. 210 * 211 * We need to save the IP options in case a protocol wants to respond 212 * to an incoming packet over the same route if the packet got here 213 * using IP source routing. This allows connection establishment and 214 * maintenance when the remote end is on a network that is not known 215 * to us. 216 */ 217 static int ip_nhops = 0; 218 static struct ip_srcrt { 219 struct in_addr dst; /* final destination */ 220 char nop; /* one NOP to align */ 221 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 222 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 223 } ip_srcrt; 224 225 static void save_rte(u_char *, struct in_addr); 226 static int ip_dooptions(struct mbuf *m, int, 227 struct sockaddr_in *next_hop); 228 static void ip_forward(struct mbuf *m, int srcrt, 229 struct sockaddr_in *next_hop); 230 static void ip_freef(struct ipqhead *, struct ipq *); 231 static struct mbuf *ip_reass(struct mbuf *, struct ipqhead *, 232 struct ipq *, u_int32_t *, u_int16_t *); 233 static void ipintr(void); 234 235 /* 236 * IP initialization: fill in IP protocol switch table. 237 * All protocols not implemented in kernel go to raw IP protocol handler. 238 */ 239 void 240 ip_init() 241 { 242 register struct protosw *pr; 243 register int i; 244 245 TAILQ_INIT(&in_ifaddrhead); 246 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 247 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 248 if (pr == 0) 249 panic("ip_init"); 250 for (i = 0; i < IPPROTO_MAX; i++) 251 ip_protox[i] = pr - inetsw; 252 for (pr = inetdomain.dom_protosw; 253 pr < inetdomain.dom_protoswNPROTOSW; pr++) 254 if (pr->pr_domain->dom_family == PF_INET && 255 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 256 ip_protox[pr->pr_protocol] = pr - inetsw; 257 258 for (i = 0; i < IPREASS_NHASH; i++) 259 TAILQ_INIT(&ipq[i]); 260 261 maxnipq = nmbclusters / 4; 262 ip_maxfragpackets = nmbclusters / 4; 263 264 #ifndef RANDOM_IP_ID 265 ip_id = time_second & 0xffff; 266 #endif 267 ipintrq.ifq_maxlen = ipqmaxlen; 268 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF); 269 ipintrq_present = 1; 270 271 register_netisr(NETISR_IP, ipintr); 272 } 273 274 /* 275 * XXX watch out this one. It is perhaps used as a cache for 276 * the most recently used route ? it is cleared in in_addroute() 277 * when a new route is successfully created. 278 */ 279 struct route ipforward_rt; 280 281 /* 282 * Ip input routine. Checksum and byte swap header. If fragmented 283 * try to reassemble. Process options. Pass to next level. 284 */ 285 void 286 ip_input(struct mbuf *m) 287 { 288 struct ip *ip; 289 struct ipq *fp; 290 struct in_ifaddr *ia = NULL; 291 struct ifaddr *ifa; 292 int i, hlen, checkif; 293 u_short sum; 294 struct in_addr pkt_dst; 295 u_int32_t divert_info = 0; /* packet divert/tee info */ 296 struct ip_fw_args args; 297 #ifdef PFIL_HOOKS 298 struct packet_filter_hook *pfh; 299 struct mbuf *m0; 300 int rv; 301 #endif /* PFIL_HOOKS */ 302 #ifdef FAST_IPSEC 303 struct m_tag *mtag; 304 struct tdb_ident *tdbi; 305 struct secpolicy *sp; 306 int s, error; 307 #endif /* FAST_IPSEC */ 308 309 args.eh = NULL; 310 args.oif = NULL; 311 args.rule = NULL; 312 args.divert_rule = 0; /* divert cookie */ 313 args.next_hop = NULL; 314 315 /* Grab info from MT_TAG mbufs prepended to the chain. */ 316 for (; m && m->m_type == MT_TAG; m = m->m_next) { 317 switch(m->_m_tag_id) { 318 default: 319 printf("ip_input: unrecognised MT_TAG tag %d\n", 320 m->_m_tag_id); 321 break; 322 323 case PACKET_TAG_DUMMYNET: 324 args.rule = ((struct dn_pkt *)m)->rule; 325 break; 326 327 case PACKET_TAG_DIVERT: 328 args.divert_rule = (intptr_t)m->m_hdr.mh_data & 0xffff; 329 break; 330 331 case PACKET_TAG_IPFORWARD: 332 args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data; 333 break; 334 } 335 } 336 337 KASSERT(m != NULL && (m->m_flags & M_PKTHDR) != 0, 338 ("ip_input: no HDR")); 339 340 if (args.rule) { /* dummynet already filtered us */ 341 ip = mtod(m, struct ip *); 342 hlen = ip->ip_hl << 2; 343 goto iphack ; 344 } 345 346 ipstat.ips_total++; 347 348 if (m->m_pkthdr.len < sizeof(struct ip)) 349 goto tooshort; 350 351 if (m->m_len < sizeof (struct ip) && 352 (m = m_pullup(m, sizeof (struct ip))) == 0) { 353 ipstat.ips_toosmall++; 354 return; 355 } 356 ip = mtod(m, struct ip *); 357 358 if (ip->ip_v != IPVERSION) { 359 ipstat.ips_badvers++; 360 goto bad; 361 } 362 363 hlen = ip->ip_hl << 2; 364 if (hlen < sizeof(struct ip)) { /* minimum header length */ 365 ipstat.ips_badhlen++; 366 goto bad; 367 } 368 if (hlen > m->m_len) { 369 if ((m = m_pullup(m, hlen)) == 0) { 370 ipstat.ips_badhlen++; 371 return; 372 } 373 ip = mtod(m, struct ip *); 374 } 375 376 /* 127/8 must not appear on wire - RFC1122 */ 377 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 378 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 379 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 380 ipstat.ips_badaddr++; 381 goto bad; 382 } 383 } 384 385 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 386 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 387 } else { 388 if (hlen == sizeof(struct ip)) { 389 sum = in_cksum_hdr(ip); 390 } else { 391 sum = in_cksum(m, hlen); 392 } 393 } 394 if (sum) { 395 ipstat.ips_badsum++; 396 goto bad; 397 } 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 } 427 428 /* 429 * IpHack's section. 430 * Right now when no processing on packet has done 431 * and it is still fresh out of network we do our black 432 * deals with it. 433 * - Firewall: deny/allow/divert 434 * - Xlate: translate packet's addr/port (NAT). 435 * - Pipe: pass pkt through dummynet. 436 * - Wrap: fake packet's addr/port <unimpl.> 437 * - Encapsulate: put it in another IP and send out. <unimp.> 438 */ 439 440 iphack: 441 442 #ifdef PFIL_HOOKS 443 /* 444 * Run through list of hooks for input packets. If there are any 445 * filters which require that additional packets in the flow are 446 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 447 * Note that filters must _never_ set this flag, as another filter 448 * in the list may have previously cleared it. 449 */ 450 m0 = m; 451 pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 452 for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link)) 453 if (pfh->pfil_func) { 454 rv = pfh->pfil_func(ip, hlen, 455 m->m_pkthdr.rcvif, 0, &m0); 456 if (rv) 457 return; 458 m = m0; 459 if (m == NULL) 460 return; 461 ip = mtod(m, struct ip *); 462 } 463 #endif /* PFIL_HOOKS */ 464 465 if (fw_enable && IPFW_LOADED) { 466 /* 467 * If we've been forwarded from the output side, then 468 * skip the firewall a second time 469 */ 470 if (args.next_hop) 471 goto ours; 472 473 args.m = m; 474 i = ip_fw_chk_ptr(&args); 475 m = args.m; 476 477 if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */ 478 if (m) 479 m_freem(m); 480 return; 481 } 482 ip = mtod(m, struct ip *); /* just in case m changed */ 483 if (i == 0 && args.next_hop == NULL) /* common case */ 484 goto pass; 485 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) { 486 /* Send packet to the appropriate pipe */ 487 ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args); 488 return; 489 } 490 #ifdef IPDIVERT 491 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 492 /* Divert or tee packet */ 493 divert_info = i; 494 goto ours; 495 } 496 #endif 497 if (i == 0 && args.next_hop != NULL) 498 goto pass; 499 /* 500 * if we get here, the packet must be dropped 501 */ 502 m_freem(m); 503 return; 504 } 505 pass: 506 507 /* 508 * Process options and, if not destined for us, 509 * ship it on. ip_dooptions returns 1 when an 510 * error was detected (causing an icmp message 511 * to be sent and the original packet to be freed). 512 */ 513 ip_nhops = 0; /* for source routed packets */ 514 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.next_hop)) 515 return; 516 517 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 518 * matter if it is destined to another node, or whether it is 519 * a multicast one, RSVP wants it! and prevents it from being forwarded 520 * anywhere else. Also checks if the rsvp daemon is running before 521 * grabbing the packet. 522 */ 523 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 524 goto ours; 525 526 /* 527 * Check our list of addresses, to see if the packet is for us. 528 * If we don't have any addresses, assume any unicast packet 529 * we receive might be for us (and let the upper layers deal 530 * with it). 531 */ 532 if (TAILQ_EMPTY(&in_ifaddrhead) && 533 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 534 goto ours; 535 536 /* 537 * Cache the destination address of the packet; this may be 538 * changed by use of 'ipfw fwd'. 539 */ 540 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst; 541 542 /* 543 * Enable a consistency check between the destination address 544 * and the arrival interface for a unicast packet (the RFC 1122 545 * strong ES model) if IP forwarding is disabled and the packet 546 * is not locally generated and the packet is not subject to 547 * 'ipfw fwd'. 548 * 549 * XXX - Checking also should be disabled if the destination 550 * address is ipnat'ed to a different interface. 551 * 552 * XXX - Checking is incompatible with IP aliases added 553 * to the loopback interface instead of the interface where 554 * the packets are received. 555 */ 556 checkif = ip_checkinterface && (ipforwarding == 0) && 557 m->m_pkthdr.rcvif != NULL && 558 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 559 (args.next_hop == NULL); 560 561 /* 562 * Check for exact addresses in the hash bucket. 563 */ 564 LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) { 565 /* 566 * If the address matches, verify that the packet 567 * arrived via the correct interface if checking is 568 * enabled. 569 */ 570 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 571 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 572 goto ours; 573 } 574 /* 575 * Check for broadcast addresses. 576 * 577 * Only accept broadcast packets that arrive via the matching 578 * interface. Reception of forwarded directed broadcasts would 579 * be handled via ip_forward() and ether_output() with the loopback 580 * into the stack for SIMPLEX interfaces handled by ether_output(). 581 */ 582 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 583 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 584 if (ifa->ifa_addr->sa_family != AF_INET) 585 continue; 586 ia = ifatoia(ifa); 587 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 588 pkt_dst.s_addr) 589 goto ours; 590 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 591 goto ours; 592 #ifdef BOOTP_COMPAT 593 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 594 goto ours; 595 #endif 596 } 597 } 598 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 599 struct in_multi *inm; 600 if (ip_mrouter) { 601 /* 602 * If we are acting as a multicast router, all 603 * incoming multicast packets are passed to the 604 * kernel-level multicast forwarding function. 605 * The packet is returned (relatively) intact; if 606 * ip_mforward() returns a non-zero value, the packet 607 * must be discarded, else it may be accepted below. 608 */ 609 if (ip_mforward && 610 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 611 ipstat.ips_cantforward++; 612 m_freem(m); 613 return; 614 } 615 616 /* 617 * The process-level routing daemon needs to receive 618 * all multicast IGMP packets, whether or not this 619 * host belongs to their destination groups. 620 */ 621 if (ip->ip_p == IPPROTO_IGMP) 622 goto ours; 623 ipstat.ips_forward++; 624 } 625 /* 626 * See if we belong to the destination multicast group on the 627 * arrival interface. 628 */ 629 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 630 if (inm == NULL) { 631 ipstat.ips_notmember++; 632 m_freem(m); 633 return; 634 } 635 goto ours; 636 } 637 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 638 goto ours; 639 if (ip->ip_dst.s_addr == INADDR_ANY) 640 goto ours; 641 642 /* 643 * FAITH(Firewall Aided Internet Translator) 644 */ 645 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 646 if (ip_keepfaith) { 647 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 648 goto ours; 649 } 650 m_freem(m); 651 return; 652 } 653 654 /* 655 * Not for us; forward if possible and desirable. 656 */ 657 if (ipforwarding == 0) { 658 ipstat.ips_cantforward++; 659 m_freem(m); 660 } else { 661 #ifdef IPSEC 662 /* 663 * Enforce inbound IPsec SPD. 664 */ 665 if (ipsec4_in_reject(m, NULL)) { 666 ipsecstat.in_polvio++; 667 goto bad; 668 } 669 #endif /* IPSEC */ 670 #ifdef FAST_IPSEC 671 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 672 s = splnet(); 673 if (mtag != NULL) { 674 tdbi = (struct tdb_ident *)(mtag + 1); 675 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 676 } else { 677 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 678 IP_FORWARDING, &error); 679 } 680 if (sp == NULL) { /* NB: can happen if error */ 681 splx(s); 682 /*XXX error stat???*/ 683 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 684 goto bad; 685 } 686 687 /* 688 * Check security policy against packet attributes. 689 */ 690 error = ipsec_in_reject(sp, m); 691 KEY_FREESP(&sp); 692 splx(s); 693 if (error) { 694 ipstat.ips_cantforward++; 695 goto bad; 696 } 697 #endif /* FAST_IPSEC */ 698 ip_forward(m, 0, args.next_hop); 699 } 700 return; 701 702 ours: 703 #ifdef IPSTEALTH 704 /* 705 * IPSTEALTH: Process non-routing options only 706 * if the packet is destined for us. 707 */ 708 if (ipstealth && hlen > sizeof (struct ip) && 709 ip_dooptions(m, 1, args.next_hop)) 710 return; 711 #endif /* IPSTEALTH */ 712 713 /* Count the packet in the ip address stats */ 714 if (ia != NULL) { 715 ia->ia_ifa.if_ipackets++; 716 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 717 } 718 719 /* 720 * If offset or IP_MF are set, must reassemble. 721 * Otherwise, nothing need be done. 722 * (We could look in the reassembly queue to see 723 * if the packet was previously fragmented, 724 * but it's not worth the time; just let them time out.) 725 */ 726 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 727 728 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 729 /* 730 * Look for queue of fragments 731 * of this datagram. 732 */ 733 TAILQ_FOREACH(fp, &ipq[sum], ipq_list) 734 if (ip->ip_id == fp->ipq_id && 735 ip->ip_src.s_addr == fp->ipq_src.s_addr && 736 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 737 #ifdef MAC 738 mac_fragment_match(m, fp) && 739 #endif 740 ip->ip_p == fp->ipq_p) 741 goto found; 742 743 fp = 0; 744 745 /* check if there's a place for the new queue */ 746 if (nipq > maxnipq) { 747 /* 748 * drop something from the tail of the current queue 749 * before proceeding further 750 */ 751 struct ipq *q = TAILQ_LAST(&ipq[sum], ipqhead); 752 if (q == NULL) { /* gak */ 753 for (i = 0; i < IPREASS_NHASH; i++) { 754 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 755 if (r) { 756 ip_freef(&ipq[i], r); 757 break; 758 } 759 } 760 } else 761 ip_freef(&ipq[sum], q); 762 } 763 found: 764 /* 765 * Adjust ip_len to not reflect header, 766 * convert offset of this to bytes. 767 */ 768 ip->ip_len -= hlen; 769 if (ip->ip_off & IP_MF) { 770 /* 771 * Make sure that fragments have a data length 772 * that's a non-zero multiple of 8 bytes. 773 */ 774 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 775 ipstat.ips_toosmall++; /* XXX */ 776 goto bad; 777 } 778 m->m_flags |= M_FRAG; 779 } else 780 m->m_flags &= ~M_FRAG; 781 ip->ip_off <<= 3; 782 783 /* 784 * Attempt reassembly; if it succeeds, proceed. 785 * ip_reass() will return a different mbuf, and update 786 * the divert info in divert_info and args.divert_rule. 787 */ 788 ipstat.ips_fragments++; 789 m->m_pkthdr.header = ip; 790 m = ip_reass(m, 791 &ipq[sum], fp, &divert_info, &args.divert_rule); 792 if (m == 0) 793 return; 794 ipstat.ips_reassembled++; 795 ip = mtod(m, struct ip *); 796 /* Get the header length of the reassembled packet */ 797 hlen = ip->ip_hl << 2; 798 #ifdef IPDIVERT 799 /* Restore original checksum before diverting packet */ 800 if (divert_info != 0) { 801 ip->ip_len += hlen; 802 ip->ip_len = htons(ip->ip_len); 803 ip->ip_off = htons(ip->ip_off); 804 ip->ip_sum = 0; 805 if (hlen == sizeof(struct ip)) 806 ip->ip_sum = in_cksum_hdr(ip); 807 else 808 ip->ip_sum = in_cksum(m, hlen); 809 ip->ip_off = ntohs(ip->ip_off); 810 ip->ip_len = ntohs(ip->ip_len); 811 ip->ip_len -= hlen; 812 } 813 #endif 814 } else 815 ip->ip_len -= hlen; 816 817 #ifdef IPDIVERT 818 /* 819 * Divert or tee packet to the divert protocol if required. 820 */ 821 if (divert_info != 0) { 822 struct mbuf *clone = NULL; 823 824 /* Clone packet if we're doing a 'tee' */ 825 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 826 clone = m_dup(m, M_DONTWAIT); 827 828 /* Restore packet header fields to original values */ 829 ip->ip_len += hlen; 830 ip->ip_len = htons(ip->ip_len); 831 ip->ip_off = htons(ip->ip_off); 832 833 /* Deliver packet to divert input routine */ 834 divert_packet(m, 1, divert_info & 0xffff, args.divert_rule); 835 ipstat.ips_delivered++; 836 837 /* If 'tee', continue with original packet */ 838 if (clone == NULL) 839 return; 840 m = clone; 841 ip = mtod(m, struct ip *); 842 ip->ip_len += hlen; 843 /* 844 * Jump backwards to complete processing of the 845 * packet. But first clear divert_info to avoid 846 * entering this block again. 847 * We do not need to clear args.divert_rule 848 * or args.next_hop as they will not be used. 849 */ 850 divert_info = 0; 851 goto pass; 852 } 853 #endif 854 855 #ifdef IPSEC 856 /* 857 * enforce IPsec policy checking if we are seeing last header. 858 * note that we do not visit this with protocols with pcb layer 859 * code - like udp/tcp/raw ip. 860 */ 861 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 862 ipsec4_in_reject(m, NULL)) { 863 ipsecstat.in_polvio++; 864 goto bad; 865 } 866 #endif 867 #if FAST_IPSEC 868 /* 869 * enforce IPsec policy checking if we are seeing last header. 870 * note that we do not visit this with protocols with pcb layer 871 * code - like udp/tcp/raw ip. 872 */ 873 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 874 /* 875 * Check if the packet has already had IPsec processing 876 * done. If so, then just pass it along. This tag gets 877 * set during AH, ESP, etc. input handling, before the 878 * packet is returned to the ip input queue for delivery. 879 */ 880 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 881 s = splnet(); 882 if (mtag != NULL) { 883 tdbi = (struct tdb_ident *)(mtag + 1); 884 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 885 } else { 886 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 887 IP_FORWARDING, &error); 888 } 889 if (sp != NULL) { 890 /* 891 * Check security policy against packet attributes. 892 */ 893 error = ipsec_in_reject(sp, m); 894 KEY_FREESP(&sp); 895 } else { 896 /* XXX error stat??? */ 897 error = EINVAL; 898 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 899 goto bad; 900 } 901 splx(s); 902 if (error) 903 goto bad; 904 } 905 #endif /* FAST_IPSEC */ 906 907 /* 908 * Switch out to protocol's input routine. 909 */ 910 ipstat.ips_delivered++; 911 if (args.next_hop && ip->ip_p == IPPROTO_TCP) { 912 /* TCP needs IPFORWARD info if available */ 913 struct m_hdr tag; 914 915 tag.mh_type = MT_TAG; 916 tag.mh_flags = PACKET_TAG_IPFORWARD; 917 tag.mh_data = (caddr_t)args.next_hop; 918 tag.mh_next = m; 919 920 (*inetsw[ip_protox[ip->ip_p]].pr_input)( 921 (struct mbuf *)&tag, hlen); 922 } else 923 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 924 return; 925 bad: 926 m_freem(m); 927 } 928 929 /* 930 * IP software interrupt routine - to go away sometime soon 931 */ 932 static void 933 ipintr(void) 934 { 935 struct mbuf *m; 936 937 while (1) { 938 IF_DEQUEUE(&ipintrq, m); 939 if (m == 0) 940 return; 941 ip_input(m); 942 } 943 } 944 945 /* 946 * Take incoming datagram fragment and try to reassemble it into 947 * whole datagram. If a chain for reassembly of this datagram already 948 * exists, then it is given as fp; otherwise have to make a chain. 949 * 950 * When IPDIVERT enabled, keep additional state with each packet that 951 * tells us if we need to divert or tee the packet we're building. 952 * In particular, *divinfo includes the port and TEE flag, 953 * *divert_rule is the number of the matching rule. 954 */ 955 956 static struct mbuf * 957 ip_reass(struct mbuf *m, struct ipqhead *head, struct ipq *fp, 958 u_int32_t *divinfo, u_int16_t *divert_rule) 959 { 960 struct ip *ip = mtod(m, struct ip *); 961 register struct mbuf *p, *q, *nq; 962 struct mbuf *t; 963 int hlen = ip->ip_hl << 2; 964 int i, next; 965 966 /* 967 * Presence of header sizes in mbufs 968 * would confuse code below. 969 */ 970 m->m_data += hlen; 971 m->m_len -= hlen; 972 973 /* 974 * If first fragment to arrive, create a reassembly queue. 975 */ 976 if (fp == 0) { 977 /* 978 * Enforce upper bound on number of fragmented packets 979 * for which we attempt reassembly; 980 * If maxfrag is 0, never accept fragments. 981 * If maxfrag is -1, accept all fragments without limitation. 982 */ 983 if ((ip_maxfragpackets >= 0) && (ip_nfragpackets >= ip_maxfragpackets)) 984 goto dropfrag; 985 ip_nfragpackets++; 986 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 987 goto dropfrag; 988 fp = mtod(t, struct ipq *); 989 #ifdef MAC 990 mac_init_ipq(fp); 991 mac_create_ipq(m, fp); 992 #endif 993 TAILQ_INSERT_HEAD(head, fp, ipq_list); 994 nipq++; 995 fp->ipq_ttl = IPFRAGTTL; 996 fp->ipq_p = ip->ip_p; 997 fp->ipq_id = ip->ip_id; 998 fp->ipq_src = ip->ip_src; 999 fp->ipq_dst = ip->ip_dst; 1000 fp->ipq_frags = m; 1001 m->m_nextpkt = NULL; 1002 #ifdef IPDIVERT 1003 fp->ipq_div_info = 0; 1004 fp->ipq_div_cookie = 0; 1005 #endif 1006 goto inserted; 1007 } else { 1008 #ifdef MAC 1009 mac_update_ipq(m, fp); 1010 #endif 1011 } 1012 1013 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1014 1015 /* 1016 * Find a segment which begins after this one does. 1017 */ 1018 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1019 if (GETIP(q)->ip_off > ip->ip_off) 1020 break; 1021 1022 /* 1023 * If there is a preceding segment, it may provide some of 1024 * our data already. If so, drop the data from the incoming 1025 * segment. If it provides all of our data, drop us, otherwise 1026 * stick new segment in the proper place. 1027 * 1028 * If some of the data is dropped from the the preceding 1029 * segment, then it's checksum is invalidated. 1030 */ 1031 if (p) { 1032 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1033 if (i > 0) { 1034 if (i >= ip->ip_len) 1035 goto dropfrag; 1036 m_adj(m, i); 1037 m->m_pkthdr.csum_flags = 0; 1038 ip->ip_off += i; 1039 ip->ip_len -= i; 1040 } 1041 m->m_nextpkt = p->m_nextpkt; 1042 p->m_nextpkt = m; 1043 } else { 1044 m->m_nextpkt = fp->ipq_frags; 1045 fp->ipq_frags = m; 1046 } 1047 1048 /* 1049 * While we overlap succeeding segments trim them or, 1050 * if they are completely covered, dequeue them. 1051 */ 1052 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1053 q = nq) { 1054 i = (ip->ip_off + ip->ip_len) - 1055 GETIP(q)->ip_off; 1056 if (i < GETIP(q)->ip_len) { 1057 GETIP(q)->ip_len -= i; 1058 GETIP(q)->ip_off += i; 1059 m_adj(q, i); 1060 q->m_pkthdr.csum_flags = 0; 1061 break; 1062 } 1063 nq = q->m_nextpkt; 1064 m->m_nextpkt = nq; 1065 m_freem(q); 1066 } 1067 1068 inserted: 1069 1070 #ifdef IPDIVERT 1071 /* 1072 * Transfer firewall instructions to the fragment structure. 1073 * Only trust info in the fragment at offset 0. 1074 */ 1075 if (ip->ip_off == 0) { 1076 fp->ipq_div_info = *divinfo; 1077 fp->ipq_div_cookie = *divert_rule; 1078 } 1079 *divinfo = 0; 1080 *divert_rule = 0; 1081 #endif 1082 1083 /* 1084 * Check for complete reassembly. 1085 */ 1086 next = 0; 1087 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1088 if (GETIP(q)->ip_off != next) 1089 return (0); 1090 next += GETIP(q)->ip_len; 1091 } 1092 /* Make sure the last packet didn't have the IP_MF flag */ 1093 if (p->m_flags & M_FRAG) 1094 return (0); 1095 1096 /* 1097 * Reassembly is complete. Make sure the packet is a sane size. 1098 */ 1099 q = fp->ipq_frags; 1100 ip = GETIP(q); 1101 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1102 ipstat.ips_toolong++; 1103 ip_freef(head, fp); 1104 return (0); 1105 } 1106 1107 /* 1108 * Concatenate fragments. 1109 */ 1110 m = q; 1111 t = m->m_next; 1112 m->m_next = 0; 1113 m_cat(m, t); 1114 nq = q->m_nextpkt; 1115 q->m_nextpkt = 0; 1116 for (q = nq; q != NULL; q = nq) { 1117 nq = q->m_nextpkt; 1118 q->m_nextpkt = NULL; 1119 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1120 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1121 m_cat(m, q); 1122 } 1123 #ifdef MAC 1124 mac_create_datagram_from_ipq(fp, m); 1125 mac_destroy_ipq(fp); 1126 #endif 1127 1128 #ifdef IPDIVERT 1129 /* 1130 * Extract firewall instructions from the fragment structure. 1131 */ 1132 *divinfo = fp->ipq_div_info; 1133 *divert_rule = fp->ipq_div_cookie; 1134 #endif 1135 1136 /* 1137 * Create header for new ip packet by 1138 * modifying header of first packet; 1139 * dequeue and discard fragment reassembly header. 1140 * Make header visible. 1141 */ 1142 ip->ip_len = next; 1143 ip->ip_src = fp->ipq_src; 1144 ip->ip_dst = fp->ipq_dst; 1145 TAILQ_REMOVE(head, fp, ipq_list); 1146 nipq--; 1147 (void) m_free(dtom(fp)); 1148 ip_nfragpackets--; 1149 m->m_len += (ip->ip_hl << 2); 1150 m->m_data -= (ip->ip_hl << 2); 1151 /* some debugging cruft by sklower, below, will go away soon */ 1152 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1153 m_fixhdr(m); 1154 return (m); 1155 1156 dropfrag: 1157 #ifdef IPDIVERT 1158 *divinfo = 0; 1159 *divert_rule = 0; 1160 #endif 1161 ipstat.ips_fragdropped++; 1162 m_freem(m); 1163 return (0); 1164 1165 #undef GETIP 1166 } 1167 1168 /* 1169 * Free a fragment reassembly header and all 1170 * associated datagrams. 1171 */ 1172 static void 1173 ip_freef(fhp, fp) 1174 struct ipqhead *fhp; 1175 struct ipq *fp; 1176 { 1177 register struct mbuf *q; 1178 1179 while (fp->ipq_frags) { 1180 q = fp->ipq_frags; 1181 fp->ipq_frags = q->m_nextpkt; 1182 m_freem(q); 1183 } 1184 TAILQ_REMOVE(fhp, fp, ipq_list); 1185 (void) m_free(dtom(fp)); 1186 ip_nfragpackets--; 1187 nipq--; 1188 } 1189 1190 /* 1191 * IP timer processing; 1192 * if a timer expires on a reassembly 1193 * queue, discard it. 1194 */ 1195 void 1196 ip_slowtimo() 1197 { 1198 register struct ipq *fp; 1199 int s = splnet(); 1200 int i; 1201 1202 for (i = 0; i < IPREASS_NHASH; i++) { 1203 for(fp = TAILQ_FIRST(&ipq[i]); fp;) { 1204 struct ipq *fpp; 1205 1206 fpp = fp; 1207 fp = TAILQ_NEXT(fp, ipq_list); 1208 if(--fpp->ipq_ttl == 0) { 1209 ipstat.ips_fragtimeout++; 1210 ip_freef(&ipq[i], fpp); 1211 } 1212 } 1213 } 1214 /* 1215 * If we are over the maximum number of fragments 1216 * (due to the limit being lowered), drain off 1217 * enough to get down to the new limit. 1218 */ 1219 for (i = 0; i < IPREASS_NHASH; i++) { 1220 if (ip_maxfragpackets >= 0) { 1221 while (ip_nfragpackets > ip_maxfragpackets && 1222 !TAILQ_EMPTY(&ipq[i])) { 1223 ipstat.ips_fragdropped++; 1224 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1225 } 1226 } 1227 } 1228 ipflow_slowtimo(); 1229 splx(s); 1230 } 1231 1232 /* 1233 * Drain off all datagram fragments. 1234 */ 1235 void 1236 ip_drain() 1237 { 1238 int i; 1239 1240 for (i = 0; i < IPREASS_NHASH; i++) { 1241 while(!TAILQ_EMPTY(&ipq[i])) { 1242 ipstat.ips_fragdropped++; 1243 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1244 } 1245 } 1246 in_rtqdrain(); 1247 } 1248 1249 /* 1250 * Do option processing on a datagram, 1251 * possibly discarding it if bad options are encountered, 1252 * or forwarding it if source-routed. 1253 * The pass argument is used when operating in the IPSTEALTH 1254 * mode to tell what options to process: 1255 * [LS]SRR (pass 0) or the others (pass 1). 1256 * The reason for as many as two passes is that when doing IPSTEALTH, 1257 * non-routing options should be processed only if the packet is for us. 1258 * Returns 1 if packet has been forwarded/freed, 1259 * 0 if the packet should be processed further. 1260 */ 1261 static int 1262 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop) 1263 { 1264 struct ip *ip = mtod(m, struct ip *); 1265 u_char *cp; 1266 struct in_ifaddr *ia; 1267 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1268 struct in_addr *sin, dst; 1269 n_time ntime; 1270 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 1271 1272 dst = ip->ip_dst; 1273 cp = (u_char *)(ip + 1); 1274 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1275 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1276 opt = cp[IPOPT_OPTVAL]; 1277 if (opt == IPOPT_EOL) 1278 break; 1279 if (opt == IPOPT_NOP) 1280 optlen = 1; 1281 else { 1282 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1283 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1284 goto bad; 1285 } 1286 optlen = cp[IPOPT_OLEN]; 1287 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1288 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1289 goto bad; 1290 } 1291 } 1292 switch (opt) { 1293 1294 default: 1295 break; 1296 1297 /* 1298 * Source routing with record. 1299 * Find interface with current destination address. 1300 * If none on this machine then drop if strictly routed, 1301 * or do nothing if loosely routed. 1302 * Record interface address and bring up next address 1303 * component. If strictly routed make sure next 1304 * address is on directly accessible net. 1305 */ 1306 case IPOPT_LSRR: 1307 case IPOPT_SSRR: 1308 #ifdef IPSTEALTH 1309 if (ipstealth && pass > 0) 1310 break; 1311 #endif 1312 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1313 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1314 goto bad; 1315 } 1316 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1317 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1318 goto bad; 1319 } 1320 ipaddr.sin_addr = ip->ip_dst; 1321 ia = (struct in_ifaddr *) 1322 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1323 if (ia == 0) { 1324 if (opt == IPOPT_SSRR) { 1325 type = ICMP_UNREACH; 1326 code = ICMP_UNREACH_SRCFAIL; 1327 goto bad; 1328 } 1329 if (!ip_dosourceroute) 1330 goto nosourcerouting; 1331 /* 1332 * Loose routing, and not at next destination 1333 * yet; nothing to do except forward. 1334 */ 1335 break; 1336 } 1337 off--; /* 0 origin */ 1338 if (off > optlen - (int)sizeof(struct in_addr)) { 1339 /* 1340 * End of source route. Should be for us. 1341 */ 1342 if (!ip_acceptsourceroute) 1343 goto nosourcerouting; 1344 save_rte(cp, ip->ip_src); 1345 break; 1346 } 1347 #ifdef IPSTEALTH 1348 if (ipstealth) 1349 goto dropit; 1350 #endif 1351 if (!ip_dosourceroute) { 1352 if (ipforwarding) { 1353 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1354 /* 1355 * Acting as a router, so generate ICMP 1356 */ 1357 nosourcerouting: 1358 strcpy(buf, inet_ntoa(ip->ip_dst)); 1359 log(LOG_WARNING, 1360 "attempted source route from %s to %s\n", 1361 inet_ntoa(ip->ip_src), buf); 1362 type = ICMP_UNREACH; 1363 code = ICMP_UNREACH_SRCFAIL; 1364 goto bad; 1365 } else { 1366 /* 1367 * Not acting as a router, so silently drop. 1368 */ 1369 #ifdef IPSTEALTH 1370 dropit: 1371 #endif 1372 ipstat.ips_cantforward++; 1373 m_freem(m); 1374 return (1); 1375 } 1376 } 1377 1378 /* 1379 * locate outgoing interface 1380 */ 1381 (void)memcpy(&ipaddr.sin_addr, cp + off, 1382 sizeof(ipaddr.sin_addr)); 1383 1384 if (opt == IPOPT_SSRR) { 1385 #define INA struct in_ifaddr * 1386 #define SA struct sockaddr * 1387 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1388 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1389 } else 1390 ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt); 1391 if (ia == 0) { 1392 type = ICMP_UNREACH; 1393 code = ICMP_UNREACH_SRCFAIL; 1394 goto bad; 1395 } 1396 ip->ip_dst = ipaddr.sin_addr; 1397 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1398 sizeof(struct in_addr)); 1399 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1400 /* 1401 * Let ip_intr's mcast routing check handle mcast pkts 1402 */ 1403 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1404 break; 1405 1406 case IPOPT_RR: 1407 #ifdef IPSTEALTH 1408 if (ipstealth && pass == 0) 1409 break; 1410 #endif 1411 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1412 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1413 goto bad; 1414 } 1415 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1416 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1417 goto bad; 1418 } 1419 /* 1420 * If no space remains, ignore. 1421 */ 1422 off--; /* 0 origin */ 1423 if (off > optlen - (int)sizeof(struct in_addr)) 1424 break; 1425 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1426 sizeof(ipaddr.sin_addr)); 1427 /* 1428 * locate outgoing interface; if we're the destination, 1429 * use the incoming interface (should be same). 1430 */ 1431 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1432 (ia = ip_rtaddr(ipaddr.sin_addr, 1433 &ipforward_rt)) == 0) { 1434 type = ICMP_UNREACH; 1435 code = ICMP_UNREACH_HOST; 1436 goto bad; 1437 } 1438 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1439 sizeof(struct in_addr)); 1440 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1441 break; 1442 1443 case IPOPT_TS: 1444 #ifdef IPSTEALTH 1445 if (ipstealth && pass == 0) 1446 break; 1447 #endif 1448 code = cp - (u_char *)ip; 1449 if (optlen < 4 || optlen > 40) { 1450 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1451 goto bad; 1452 } 1453 if ((off = cp[IPOPT_OFFSET]) < 5) { 1454 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1455 goto bad; 1456 } 1457 if (off > optlen - (int)sizeof(int32_t)) { 1458 cp[IPOPT_OFFSET + 1] += (1 << 4); 1459 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1460 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1461 goto bad; 1462 } 1463 break; 1464 } 1465 off--; /* 0 origin */ 1466 sin = (struct in_addr *)(cp + off); 1467 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1468 1469 case IPOPT_TS_TSONLY: 1470 break; 1471 1472 case IPOPT_TS_TSANDADDR: 1473 if (off + sizeof(n_time) + 1474 sizeof(struct in_addr) > optlen) { 1475 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1476 goto bad; 1477 } 1478 ipaddr.sin_addr = dst; 1479 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1480 m->m_pkthdr.rcvif); 1481 if (ia == 0) 1482 continue; 1483 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1484 sizeof(struct in_addr)); 1485 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1486 off += sizeof(struct in_addr); 1487 break; 1488 1489 case IPOPT_TS_PRESPEC: 1490 if (off + sizeof(n_time) + 1491 sizeof(struct in_addr) > optlen) { 1492 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1493 goto bad; 1494 } 1495 (void)memcpy(&ipaddr.sin_addr, sin, 1496 sizeof(struct in_addr)); 1497 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1498 continue; 1499 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1500 off += sizeof(struct in_addr); 1501 break; 1502 1503 default: 1504 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1505 goto bad; 1506 } 1507 ntime = iptime(); 1508 (void)memcpy(cp + off, &ntime, sizeof(n_time)); 1509 cp[IPOPT_OFFSET] += sizeof(n_time); 1510 } 1511 } 1512 if (forward && ipforwarding) { 1513 ip_forward(m, 1, next_hop); 1514 return (1); 1515 } 1516 return (0); 1517 bad: 1518 icmp_error(m, type, code, 0, 0); 1519 ipstat.ips_badoptions++; 1520 return (1); 1521 } 1522 1523 /* 1524 * Given address of next destination (final or next hop), 1525 * return internet address info of interface to be used to get there. 1526 */ 1527 struct in_ifaddr * 1528 ip_rtaddr(dst, rt) 1529 struct in_addr dst; 1530 struct route *rt; 1531 { 1532 register struct sockaddr_in *sin; 1533 1534 sin = (struct sockaddr_in *)&rt->ro_dst; 1535 1536 if (rt->ro_rt == 0 || 1537 !(rt->ro_rt->rt_flags & RTF_UP) || 1538 dst.s_addr != sin->sin_addr.s_addr) { 1539 if (rt->ro_rt) { 1540 RTFREE(rt->ro_rt); 1541 rt->ro_rt = 0; 1542 } 1543 sin->sin_family = AF_INET; 1544 sin->sin_len = sizeof(*sin); 1545 sin->sin_addr = dst; 1546 1547 rtalloc_ign(rt, RTF_PRCLONING); 1548 } 1549 if (rt->ro_rt == 0) 1550 return ((struct in_ifaddr *)0); 1551 return (ifatoia(rt->ro_rt->rt_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(option, dst) 1560 u_char *option; 1561 struct in_addr dst; 1562 { 1563 unsigned olen; 1564 1565 olen = option[IPOPT_OLEN]; 1566 #ifdef DIAGNOSTIC 1567 if (ipprintfs) 1568 printf("save_rte: olen %d\n", olen); 1569 #endif 1570 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1571 return; 1572 bcopy(option, ip_srcrt.srcopt, olen); 1573 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1574 ip_srcrt.dst = dst; 1575 } 1576 1577 /* 1578 * Retrieve incoming source route for use in replies, 1579 * in the same form used by setsockopt. 1580 * The first hop is placed before the options, will be removed later. 1581 */ 1582 struct mbuf * 1583 ip_srcroute() 1584 { 1585 register struct in_addr *p, *q; 1586 register struct mbuf *m; 1587 1588 if (ip_nhops == 0) 1589 return ((struct mbuf *)0); 1590 m = m_get(M_DONTWAIT, MT_HEADER); 1591 if (m == 0) 1592 return ((struct mbuf *)0); 1593 1594 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1595 1596 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1597 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1598 OPTSIZ; 1599 #ifdef DIAGNOSTIC 1600 if (ipprintfs) 1601 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1602 #endif 1603 1604 /* 1605 * First save first hop for return route 1606 */ 1607 p = &ip_srcrt.route[ip_nhops - 1]; 1608 *(mtod(m, struct in_addr *)) = *p--; 1609 #ifdef DIAGNOSTIC 1610 if (ipprintfs) 1611 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1612 #endif 1613 1614 /* 1615 * Copy option fields and padding (nop) to mbuf. 1616 */ 1617 ip_srcrt.nop = IPOPT_NOP; 1618 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1619 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1620 &ip_srcrt.nop, OPTSIZ); 1621 q = (struct in_addr *)(mtod(m, caddr_t) + 1622 sizeof(struct in_addr) + OPTSIZ); 1623 #undef OPTSIZ 1624 /* 1625 * Record return path as an IP source route, 1626 * reversing the path (pointers are now aligned). 1627 */ 1628 while (p >= ip_srcrt.route) { 1629 #ifdef DIAGNOSTIC 1630 if (ipprintfs) 1631 printf(" %lx", (u_long)ntohl(q->s_addr)); 1632 #endif 1633 *q++ = *p--; 1634 } 1635 /* 1636 * Last hop goes to final destination. 1637 */ 1638 *q = ip_srcrt.dst; 1639 #ifdef DIAGNOSTIC 1640 if (ipprintfs) 1641 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1642 #endif 1643 return (m); 1644 } 1645 1646 /* 1647 * Strip out IP options, at higher 1648 * level protocol in the kernel. 1649 * Second argument is buffer to which options 1650 * will be moved, and return value is their length. 1651 * XXX should be deleted; last arg currently ignored. 1652 */ 1653 void 1654 ip_stripoptions(m, mopt) 1655 register struct mbuf *m; 1656 struct mbuf *mopt; 1657 { 1658 register int i; 1659 struct ip *ip = mtod(m, struct ip *); 1660 register caddr_t opts; 1661 int olen; 1662 1663 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1664 opts = (caddr_t)(ip + 1); 1665 i = m->m_len - (sizeof (struct ip) + olen); 1666 bcopy(opts + olen, opts, (unsigned)i); 1667 m->m_len -= olen; 1668 if (m->m_flags & M_PKTHDR) 1669 m->m_pkthdr.len -= olen; 1670 ip->ip_v = IPVERSION; 1671 ip->ip_hl = sizeof(struct ip) >> 2; 1672 } 1673 1674 u_char inetctlerrmap[PRC_NCMDS] = { 1675 0, 0, 0, 0, 1676 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1677 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1678 EMSGSIZE, EHOSTUNREACH, 0, 0, 1679 0, 0, 0, 0, 1680 ENOPROTOOPT, ECONNREFUSED 1681 }; 1682 1683 /* 1684 * Forward a packet. If some error occurs return the sender 1685 * an icmp packet. Note we can't always generate a meaningful 1686 * icmp message because icmp doesn't have a large enough repertoire 1687 * of codes and types. 1688 * 1689 * If not forwarding, just drop the packet. This could be confusing 1690 * if ipforwarding was zero but some routing protocol was advancing 1691 * us as a gateway to somewhere. However, we must let the routing 1692 * protocol deal with that. 1693 * 1694 * The srcrt parameter indicates whether the packet is being forwarded 1695 * via a source route. 1696 */ 1697 static void 1698 ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop) 1699 { 1700 struct ip *ip = mtod(m, struct ip *); 1701 struct rtentry *rt; 1702 int error, type = 0, code = 0; 1703 struct mbuf *mcopy; 1704 n_long dest; 1705 struct in_addr pkt_dst; 1706 struct ifnet *destifp; 1707 #if defined(IPSEC) || defined(FAST_IPSEC) 1708 struct ifnet dummyifp; 1709 #endif 1710 1711 dest = 0; 1712 /* 1713 * Cache the destination address of the packet; this may be 1714 * changed by use of 'ipfw fwd'. 1715 */ 1716 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst; 1717 1718 #ifdef DIAGNOSTIC 1719 if (ipprintfs) 1720 printf("forward: src %lx dst %lx ttl %x\n", 1721 (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr, 1722 ip->ip_ttl); 1723 #endif 1724 1725 1726 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) { 1727 ipstat.ips_cantforward++; 1728 m_freem(m); 1729 return; 1730 } 1731 #ifdef IPSTEALTH 1732 if (!ipstealth) { 1733 #endif 1734 if (ip->ip_ttl <= IPTTLDEC) { 1735 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1736 dest, 0); 1737 return; 1738 } 1739 #ifdef IPSTEALTH 1740 } 1741 #endif 1742 1743 if (ip_rtaddr(pkt_dst, &ipforward_rt) == 0) { 1744 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1745 return; 1746 } else 1747 rt = ipforward_rt.ro_rt; 1748 1749 /* 1750 * Save the IP header and at most 8 bytes of the payload, 1751 * in case we need to generate an ICMP message to the src. 1752 * 1753 * XXX this can be optimized a lot by saving the data in a local 1754 * buffer on the stack (72 bytes at most), and only allocating the 1755 * mbuf if really necessary. The vast majority of the packets 1756 * are forwarded without having to send an ICMP back (either 1757 * because unnecessary, or because rate limited), so we are 1758 * really we are wasting a lot of work here. 1759 * 1760 * We don't use m_copy() because it might return a reference 1761 * to a shared cluster. Both this function and ip_output() 1762 * assume exclusive access to the IP header in `m', so any 1763 * data in a cluster may change before we reach icmp_error(). 1764 */ 1765 MGET(mcopy, M_DONTWAIT, m->m_type); 1766 if (mcopy != NULL) { 1767 M_COPY_PKTHDR(mcopy, m); 1768 mcopy->m_len = imin((ip->ip_hl << 2) + 8, 1769 (int)ip->ip_len); 1770 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1771 #ifdef MAC 1772 /* 1773 * XXXMAC: This will eventually become an explicit 1774 * labeling point. 1775 */ 1776 mac_create_mbuf_from_mbuf(m, mcopy); 1777 #endif 1778 } 1779 1780 #ifdef IPSTEALTH 1781 if (!ipstealth) { 1782 #endif 1783 ip->ip_ttl -= IPTTLDEC; 1784 #ifdef IPSTEALTH 1785 } 1786 #endif 1787 1788 /* 1789 * If forwarding packet using same interface that it came in on, 1790 * perhaps should send a redirect to sender to shortcut a hop. 1791 * Only send redirect if source is sending directly to us, 1792 * and if packet was not source routed (or has any options). 1793 * Also, don't send redirect if forwarding using a default route 1794 * or a route modified by a redirect. 1795 */ 1796 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1797 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1798 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1799 ipsendredirects && !srcrt && !next_hop) { 1800 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1801 u_long src = ntohl(ip->ip_src.s_addr); 1802 1803 if (RTA(rt) && 1804 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1805 if (rt->rt_flags & RTF_GATEWAY) 1806 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1807 else 1808 dest = pkt_dst.s_addr; 1809 /* Router requirements says to only send host redirects */ 1810 type = ICMP_REDIRECT; 1811 code = ICMP_REDIRECT_HOST; 1812 #ifdef DIAGNOSTIC 1813 if (ipprintfs) 1814 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1815 #endif 1816 } 1817 } 1818 1819 { 1820 struct m_hdr tag; 1821 1822 if (next_hop) { 1823 /* Pass IPFORWARD info if available */ 1824 1825 tag.mh_type = MT_TAG; 1826 tag.mh_flags = PACKET_TAG_IPFORWARD; 1827 tag.mh_data = (caddr_t)next_hop; 1828 tag.mh_next = m; 1829 m = (struct mbuf *)&tag; 1830 } 1831 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1832 IP_FORWARDING, 0, NULL); 1833 } 1834 if (error) 1835 ipstat.ips_cantforward++; 1836 else { 1837 ipstat.ips_forward++; 1838 if (type) 1839 ipstat.ips_redirectsent++; 1840 else { 1841 if (mcopy) { 1842 ipflow_create(&ipforward_rt, mcopy); 1843 m_freem(mcopy); 1844 } 1845 return; 1846 } 1847 } 1848 if (mcopy == NULL) 1849 return; 1850 destifp = NULL; 1851 1852 switch (error) { 1853 1854 case 0: /* forwarded, but need redirect */ 1855 /* type, code set above */ 1856 break; 1857 1858 case ENETUNREACH: /* shouldn't happen, checked above */ 1859 case EHOSTUNREACH: 1860 case ENETDOWN: 1861 case EHOSTDOWN: 1862 default: 1863 type = ICMP_UNREACH; 1864 code = ICMP_UNREACH_HOST; 1865 break; 1866 1867 case EMSGSIZE: 1868 type = ICMP_UNREACH; 1869 code = ICMP_UNREACH_NEEDFRAG; 1870 #ifdef IPSEC 1871 /* 1872 * If the packet is routed over IPsec tunnel, tell the 1873 * originator the tunnel MTU. 1874 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1875 * XXX quickhack!!! 1876 */ 1877 if (ipforward_rt.ro_rt) { 1878 struct secpolicy *sp = NULL; 1879 int ipsecerror; 1880 int ipsechdr; 1881 struct route *ro; 1882 1883 sp = ipsec4_getpolicybyaddr(mcopy, 1884 IPSEC_DIR_OUTBOUND, 1885 IP_FORWARDING, 1886 &ipsecerror); 1887 1888 if (sp == NULL) 1889 destifp = ipforward_rt.ro_rt->rt_ifp; 1890 else { 1891 /* count IPsec header size */ 1892 ipsechdr = ipsec4_hdrsiz(mcopy, 1893 IPSEC_DIR_OUTBOUND, 1894 NULL); 1895 1896 /* 1897 * find the correct route for outer IPv4 1898 * header, compute tunnel MTU. 1899 * 1900 * XXX BUG ALERT 1901 * The "dummyifp" code relies upon the fact 1902 * that icmp_error() touches only ifp->if_mtu. 1903 */ 1904 /*XXX*/ 1905 destifp = NULL; 1906 if (sp->req != NULL 1907 && sp->req->sav != NULL 1908 && sp->req->sav->sah != NULL) { 1909 ro = &sp->req->sav->sah->sa_route; 1910 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1911 dummyifp.if_mtu = 1912 ro->ro_rt->rt_ifp->if_mtu; 1913 dummyifp.if_mtu -= ipsechdr; 1914 destifp = &dummyifp; 1915 } 1916 } 1917 1918 key_freesp(sp); 1919 } 1920 } 1921 #elif FAST_IPSEC 1922 /* 1923 * If the packet is routed over IPsec tunnel, tell the 1924 * originator the tunnel MTU. 1925 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1926 * XXX quickhack!!! 1927 */ 1928 if (ipforward_rt.ro_rt) { 1929 struct secpolicy *sp = NULL; 1930 int ipsecerror; 1931 int ipsechdr; 1932 struct route *ro; 1933 1934 sp = ipsec_getpolicybyaddr(mcopy, 1935 IPSEC_DIR_OUTBOUND, 1936 IP_FORWARDING, 1937 &ipsecerror); 1938 1939 if (sp == NULL) 1940 destifp = ipforward_rt.ro_rt->rt_ifp; 1941 else { 1942 /* count IPsec header size */ 1943 ipsechdr = ipsec4_hdrsiz(mcopy, 1944 IPSEC_DIR_OUTBOUND, 1945 NULL); 1946 1947 /* 1948 * find the correct route for outer IPv4 1949 * header, compute tunnel MTU. 1950 * 1951 * XXX BUG ALERT 1952 * The "dummyifp" code relies upon the fact 1953 * that icmp_error() touches only ifp->if_mtu. 1954 */ 1955 /*XXX*/ 1956 destifp = NULL; 1957 if (sp->req != NULL 1958 && sp->req->sav != NULL 1959 && sp->req->sav->sah != NULL) { 1960 ro = &sp->req->sav->sah->sa_route; 1961 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1962 dummyifp.if_mtu = 1963 ro->ro_rt->rt_ifp->if_mtu; 1964 dummyifp.if_mtu -= ipsechdr; 1965 destifp = &dummyifp; 1966 } 1967 } 1968 1969 KEY_FREESP(&sp); 1970 } 1971 } 1972 #else /* !IPSEC && !FAST_IPSEC */ 1973 if (ipforward_rt.ro_rt) 1974 destifp = ipforward_rt.ro_rt->rt_ifp; 1975 #endif /*IPSEC*/ 1976 ipstat.ips_cantfrag++; 1977 break; 1978 1979 case ENOBUFS: 1980 /* 1981 * A router should not generate ICMP_SOURCEQUENCH as 1982 * required in RFC1812 Requirements for IP Version 4 Routers. 1983 * Source quench could be a big problem under DoS attacks, 1984 * or if the underlying interface is rate-limited. 1985 * Those who need source quench packets may re-enable them 1986 * via the net.inet.ip.sendsourcequench sysctl. 1987 */ 1988 if (ip_sendsourcequench == 0) { 1989 m_freem(mcopy); 1990 return; 1991 } else { 1992 type = ICMP_SOURCEQUENCH; 1993 code = 0; 1994 } 1995 break; 1996 1997 case EACCES: /* ipfw denied packet */ 1998 m_freem(mcopy); 1999 return; 2000 } 2001 icmp_error(mcopy, type, code, dest, destifp); 2002 } 2003 2004 void 2005 ip_savecontrol(inp, mp, ip, m) 2006 register struct inpcb *inp; 2007 register struct mbuf **mp; 2008 register struct ip *ip; 2009 register struct mbuf *m; 2010 { 2011 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2012 struct timeval tv; 2013 2014 microtime(&tv); 2015 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2016 SCM_TIMESTAMP, SOL_SOCKET); 2017 if (*mp) 2018 mp = &(*mp)->m_next; 2019 } 2020 if (inp->inp_flags & INP_RECVDSTADDR) { 2021 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2022 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2023 if (*mp) 2024 mp = &(*mp)->m_next; 2025 } 2026 #ifdef notyet 2027 /* XXX 2028 * Moving these out of udp_input() made them even more broken 2029 * than they already were. 2030 */ 2031 /* options were tossed already */ 2032 if (inp->inp_flags & INP_RECVOPTS) { 2033 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2034 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2035 if (*mp) 2036 mp = &(*mp)->m_next; 2037 } 2038 /* ip_srcroute doesn't do what we want here, need to fix */ 2039 if (inp->inp_flags & INP_RECVRETOPTS) { 2040 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 2041 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2042 if (*mp) 2043 mp = &(*mp)->m_next; 2044 } 2045 #endif 2046 if (inp->inp_flags & INP_RECVIF) { 2047 struct ifnet *ifp; 2048 struct sdlbuf { 2049 struct sockaddr_dl sdl; 2050 u_char pad[32]; 2051 } sdlbuf; 2052 struct sockaddr_dl *sdp; 2053 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2054 2055 if (((ifp = m->m_pkthdr.rcvif)) 2056 && ( ifp->if_index && (ifp->if_index <= if_index))) { 2057 sdp = (struct sockaddr_dl *) 2058 (ifaddr_byindex(ifp->if_index)->ifa_addr); 2059 /* 2060 * Change our mind and don't try copy. 2061 */ 2062 if ((sdp->sdl_family != AF_LINK) 2063 || (sdp->sdl_len > sizeof(sdlbuf))) { 2064 goto makedummy; 2065 } 2066 bcopy(sdp, sdl2, sdp->sdl_len); 2067 } else { 2068 makedummy: 2069 sdl2->sdl_len 2070 = offsetof(struct sockaddr_dl, sdl_data[0]); 2071 sdl2->sdl_family = AF_LINK; 2072 sdl2->sdl_index = 0; 2073 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2074 } 2075 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 2076 IP_RECVIF, IPPROTO_IP); 2077 if (*mp) 2078 mp = &(*mp)->m_next; 2079 } 2080 } 2081 2082 /* 2083 * XXX these routines are called from the upper part of the kernel. 2084 * They need to be locked when we remove Giant. 2085 * 2086 * They could also be moved to ip_mroute.c, since all the RSVP 2087 * handling is done there already. 2088 */ 2089 static int ip_rsvp_on; 2090 struct socket *ip_rsvpd; 2091 int 2092 ip_rsvp_init(struct socket *so) 2093 { 2094 if (so->so_type != SOCK_RAW || 2095 so->so_proto->pr_protocol != IPPROTO_RSVP) 2096 return EOPNOTSUPP; 2097 2098 if (ip_rsvpd != NULL) 2099 return EADDRINUSE; 2100 2101 ip_rsvpd = so; 2102 /* 2103 * This may seem silly, but we need to be sure we don't over-increment 2104 * the RSVP counter, in case something slips up. 2105 */ 2106 if (!ip_rsvp_on) { 2107 ip_rsvp_on = 1; 2108 rsvp_on++; 2109 } 2110 2111 return 0; 2112 } 2113 2114 int 2115 ip_rsvp_done(void) 2116 { 2117 ip_rsvpd = NULL; 2118 /* 2119 * This may seem silly, but we need to be sure we don't over-decrement 2120 * the RSVP counter, in case something slips up. 2121 */ 2122 if (ip_rsvp_on) { 2123 ip_rsvp_on = 0; 2124 rsvp_on--; 2125 } 2126 return 0; 2127 } 2128 2129 void 2130 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 2131 { 2132 if (rsvp_input_p) { /* call the real one if loaded */ 2133 rsvp_input_p(m, off); 2134 return; 2135 } 2136 2137 /* Can still get packets with rsvp_on = 0 if there is a local member 2138 * of the group to which the RSVP packet is addressed. But in this 2139 * case we want to throw the packet away. 2140 */ 2141 2142 if (!rsvp_on) { 2143 m_freem(m); 2144 return; 2145 } 2146 2147 if (ip_rsvpd != NULL) { 2148 rip_input(m, off); 2149 return; 2150 } 2151 /* Drop the packet */ 2152 m_freem(m); 2153 } 2154