ip_reass.c revision 107081
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: head/sys/netinet/ip_input.c 107081 2002-11-19 17:06:06Z silby $ 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 94int rsvp_on = 0; 95 96int ipforwarding = 0; 97SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 98 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 99 100static int ipsendredirects = 1; /* XXX */ 101SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 102 &ipsendredirects, 0, "Enable sending IP redirects"); 103 104int ip_defttl = IPDEFTTL; 105SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 106 &ip_defttl, 0, "Maximum TTL on IP packets"); 107 108static int ip_dosourceroute = 0; 109SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 110 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 111 112static int ip_acceptsourceroute = 0; 113SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 114 CTLFLAG_RW, &ip_acceptsourceroute, 0, 115 "Enable accepting source routed IP packets"); 116 117static int ip_keepfaith = 0; 118SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 119 &ip_keepfaith, 0, 120 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 121 122static int ip_nfragpackets = 0; 123static int ip_maxfragpackets; /* initialized in ip_init() */ 124SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 125 &ip_maxfragpackets, 0, 126 "Maximum number of IPv4 fragment reassembly queue entries"); 127 128static int ip_sendsourcequench = 0; 129SYSCTL_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 */ 146static int ip_checkinterface = 1; 147SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 148 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 149 150#ifdef DIAGNOSTIC 151static int ipprintfs = 0; 152#endif 153 154static int ipqmaxlen = IFQ_MAXLEN; 155 156extern struct domain inetdomain; 157extern struct protosw inetsw[]; 158u_char ip_protox[IPPROTO_MAX]; 159struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 160struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 161u_long in_ifaddrhmask; /* mask for hash table */ 162 163SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 164 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 165SYSCTL_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 168struct ipstat ipstat; 169SYSCTL_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 179static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 180static int nipq = 0; /* total # of reass queues */ 181static int maxnipq; 182 183#ifdef IPCTL_DEFMTU 184SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 185 &ip_mtu, 0, "Default MTU"); 186#endif 187 188#ifdef IPSTEALTH 189static int ipstealth = 0; 190SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 191 &ipstealth, 0, ""); 192#endif 193 194 195/* Firewall hooks */ 196ip_fw_chk_t *ip_fw_chk_ptr; 197int fw_enable = 1 ; 198 199/* Dummynet hooks */ 200ip_dn_io_t *ip_dn_io_ptr; 201 202 203/* 204 * XXX this is ugly -- the following two global variables are 205 * used to store packet state while it travels through the stack. 206 * Note that the code even makes assumptions on the size and 207 * alignment of fields inside struct ip_srcrt so e.g. adding some 208 * fields will break the code. This needs to be fixed. 209 * 210 * We need to save the IP options in case a protocol wants to respond 211 * to an incoming packet over the same route if the packet got here 212 * using IP source routing. This allows connection establishment and 213 * maintenance when the remote end is on a network that is not known 214 * to us. 215 */ 216static int ip_nhops = 0; 217static struct ip_srcrt { 218 struct in_addr dst; /* final destination */ 219 char nop; /* one NOP to align */ 220 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 221 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 222} ip_srcrt; 223 224static void save_rte(u_char *, struct in_addr); 225static int ip_dooptions(struct mbuf *m, int, 226 struct sockaddr_in *next_hop); 227static void ip_forward(struct mbuf *m, int srcrt, 228 struct sockaddr_in *next_hop); 229static void ip_freef(struct ipqhead *, struct ipq *); 230static struct mbuf *ip_reass(struct mbuf *, struct ipqhead *, 231 struct ipq *, u_int32_t *, u_int16_t *); 232static void ipintr(void); 233 234/* 235 * IP initialization: fill in IP protocol switch table. 236 * All protocols not implemented in kernel go to raw IP protocol handler. 237 */ 238void 239ip_init() 240{ 241 register struct protosw *pr; 242 register int i; 243 244 TAILQ_INIT(&in_ifaddrhead); 245 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 246 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 247 if (pr == 0) 248 panic("ip_init"); 249 for (i = 0; i < IPPROTO_MAX; i++) 250 ip_protox[i] = pr - inetsw; 251 for (pr = inetdomain.dom_protosw; 252 pr < inetdomain.dom_protoswNPROTOSW; pr++) 253 if (pr->pr_domain->dom_family == PF_INET && 254 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 255 ip_protox[pr->pr_protocol] = pr - inetsw; 256 257 for (i = 0; i < IPREASS_NHASH; i++) 258 TAILQ_INIT(&ipq[i]); 259 260 maxnipq = nmbclusters / 4; 261 ip_maxfragpackets = nmbclusters / 4; 262 263#ifndef RANDOM_IP_ID 264 ip_id = time_second & 0xffff; 265#endif 266 ipintrq.ifq_maxlen = ipqmaxlen; 267 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF); 268 ipintrq_present = 1; 269 270 register_netisr(NETISR_IP, ipintr); 271} 272 273/* 274 * XXX watch out this one. It is perhaps used as a cache for 275 * the most recently used route ? it is cleared in in_addroute() 276 * when a new route is successfully created. 277 */ 278struct route ipforward_rt; 279 280/* 281 * Ip input routine. Checksum and byte swap header. If fragmented 282 * try to reassemble. Process options. Pass to next level. 283 */ 284void 285ip_input(struct mbuf *m) 286{ 287 struct ip *ip; 288 struct ipq *fp; 289 struct in_ifaddr *ia = NULL; 290 struct ifaddr *ifa; 291 int i, hlen, checkif; 292 u_short sum; 293 struct in_addr pkt_dst; 294 u_int32_t divert_info = 0; /* packet divert/tee info */ 295 struct ip_fw_args args; 296#ifdef PFIL_HOOKS 297 struct packet_filter_hook *pfh; 298 struct mbuf *m0; 299 int rv; 300#endif /* PFIL_HOOKS */ 301#ifdef FAST_IPSEC 302 struct m_tag *mtag; 303 struct tdb_ident *tdbi; 304 struct secpolicy *sp; 305 int s, error; 306#endif /* FAST_IPSEC */ 307 308 args.eh = NULL; 309 args.oif = NULL; 310 args.rule = NULL; 311 args.divert_rule = 0; /* divert cookie */ 312 args.next_hop = NULL; 313 314 /* Grab info from MT_TAG mbufs prepended to the chain. */ 315 for (; m && m->m_type == MT_TAG; m = m->m_next) { 316 switch(m->_m_tag_id) { 317 default: 318 printf("ip_input: unrecognised MT_TAG tag %d\n", 319 m->_m_tag_id); 320 break; 321 322 case PACKET_TAG_DUMMYNET: 323 args.rule = ((struct dn_pkt *)m)->rule; 324 break; 325 326 case PACKET_TAG_DIVERT: 327 args.divert_rule = (intptr_t)m->m_hdr.mh_data & 0xffff; 328 break; 329 330 case PACKET_TAG_IPFORWARD: 331 args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data; 332 break; 333 } 334 } 335 336 KASSERT(m != NULL && (m->m_flags & M_PKTHDR) != 0, 337 ("ip_input: no HDR")); 338 339 if (args.rule) { /* dummynet already filtered us */ 340 ip = mtod(m, struct ip *); 341 hlen = ip->ip_hl << 2; 342 goto iphack ; 343 } 344 345 ipstat.ips_total++; 346 347 if (m->m_pkthdr.len < sizeof(struct ip)) 348 goto tooshort; 349 350 if (m->m_len < sizeof (struct ip) && 351 (m = m_pullup(m, sizeof (struct ip))) == 0) { 352 ipstat.ips_toosmall++; 353 return; 354 } 355 ip = mtod(m, struct ip *); 356 357 if (ip->ip_v != IPVERSION) { 358 ipstat.ips_badvers++; 359 goto bad; 360 } 361 362 hlen = ip->ip_hl << 2; 363 if (hlen < sizeof(struct ip)) { /* minimum header length */ 364 ipstat.ips_badhlen++; 365 goto bad; 366 } 367 if (hlen > m->m_len) { 368 if ((m = m_pullup(m, hlen)) == 0) { 369 ipstat.ips_badhlen++; 370 return; 371 } 372 ip = mtod(m, struct ip *); 373 } 374 375 /* 127/8 must not appear on wire - RFC1122 */ 376 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 377 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 378 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 379 ipstat.ips_badaddr++; 380 goto bad; 381 } 382 } 383 384 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 385 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 386 } else { 387 if (hlen == sizeof(struct ip)) { 388 sum = in_cksum_hdr(ip); 389 } else { 390 sum = in_cksum(m, hlen); 391 } 392 } 393 if (sum) { 394 ipstat.ips_badsum++; 395 goto bad; 396 } 397 398 /* 399 * Convert fields to host representation. 400 */ 401 ip->ip_len = ntohs(ip->ip_len); 402 if (ip->ip_len < hlen) { 403 ipstat.ips_badlen++; 404 goto bad; 405 } 406 ip->ip_off = ntohs(ip->ip_off); 407 408 /* 409 * Check that the amount of data in the buffers 410 * is as at least much as the IP header would have us expect. 411 * Trim mbufs if longer than we expect. 412 * Drop packet if shorter than we expect. 413 */ 414 if (m->m_pkthdr.len < ip->ip_len) { 415tooshort: 416 ipstat.ips_tooshort++; 417 goto bad; 418 } 419 if (m->m_pkthdr.len > ip->ip_len) { 420 if (m->m_len == m->m_pkthdr.len) { 421 m->m_len = ip->ip_len; 422 m->m_pkthdr.len = ip->ip_len; 423 } else 424 m_adj(m, ip->ip_len - m->m_pkthdr.len); 425 } 426 427 /* 428 * IpHack's section. 429 * Right now when no processing on packet has done 430 * and it is still fresh out of network we do our black 431 * deals with it. 432 * - Firewall: deny/allow/divert 433 * - Xlate: translate packet's addr/port (NAT). 434 * - Pipe: pass pkt through dummynet. 435 * - Wrap: fake packet's addr/port <unimpl.> 436 * - Encapsulate: put it in another IP and send out. <unimp.> 437 */ 438 439iphack: 440 441#ifdef PFIL_HOOKS 442 /* 443 * Run through list of hooks for input packets. If there are any 444 * filters which require that additional packets in the flow are 445 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 446 * Note that filters must _never_ set this flag, as another filter 447 * in the list may have previously cleared it. 448 */ 449 m0 = m; 450 pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 451 for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link)) 452 if (pfh->pfil_func) { 453 rv = pfh->pfil_func(ip, hlen, 454 m->m_pkthdr.rcvif, 0, &m0); 455 if (rv) 456 return; 457 m = m0; 458 if (m == NULL) 459 return; 460 ip = mtod(m, struct ip *); 461 } 462#endif /* PFIL_HOOKS */ 463 464 if (fw_enable && IPFW_LOADED) { 465 /* 466 * If we've been forwarded from the output side, then 467 * skip the firewall a second time 468 */ 469 if (args.next_hop) 470 goto ours; 471 472 args.m = m; 473 i = ip_fw_chk_ptr(&args); 474 m = args.m; 475 476 if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */ 477 if (m) 478 m_freem(m); 479 return; 480 } 481 ip = mtod(m, struct ip *); /* just in case m changed */ 482 if (i == 0 && args.next_hop == NULL) /* common case */ 483 goto pass; 484 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) { 485 /* Send packet to the appropriate pipe */ 486 ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args); 487 return; 488 } 489#ifdef IPDIVERT 490 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 491 /* Divert or tee packet */ 492 divert_info = i; 493 goto ours; 494 } 495#endif 496 if (i == 0 && args.next_hop != NULL) 497 goto pass; 498 /* 499 * if we get here, the packet must be dropped 500 */ 501 m_freem(m); 502 return; 503 } 504pass: 505 506 /* 507 * Process options and, if not destined for us, 508 * ship it on. ip_dooptions returns 1 when an 509 * error was detected (causing an icmp message 510 * to be sent and the original packet to be freed). 511 */ 512 ip_nhops = 0; /* for source routed packets */ 513 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.next_hop)) 514 return; 515 516 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 517 * matter if it is destined to another node, or whether it is 518 * a multicast one, RSVP wants it! and prevents it from being forwarded 519 * anywhere else. Also checks if the rsvp daemon is running before 520 * grabbing the packet. 521 */ 522 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 523 goto ours; 524 525 /* 526 * Check our list of addresses, to see if the packet is for us. 527 * If we don't have any addresses, assume any unicast packet 528 * we receive might be for us (and let the upper layers deal 529 * with it). 530 */ 531 if (TAILQ_EMPTY(&in_ifaddrhead) && 532 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 533 goto ours; 534 535 /* 536 * Cache the destination address of the packet; this may be 537 * changed by use of 'ipfw fwd'. 538 */ 539 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst; 540 541 /* 542 * Enable a consistency check between the destination address 543 * and the arrival interface for a unicast packet (the RFC 1122 544 * strong ES model) if IP forwarding is disabled and the packet 545 * is not locally generated and the packet is not subject to 546 * 'ipfw fwd'. 547 * 548 * XXX - Checking also should be disabled if the destination 549 * address is ipnat'ed to a different interface. 550 * 551 * XXX - Checking is incompatible with IP aliases added 552 * to the loopback interface instead of the interface where 553 * the packets are received. 554 */ 555 checkif = ip_checkinterface && (ipforwarding == 0) && 556 m->m_pkthdr.rcvif != NULL && 557 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 558 (args.next_hop == NULL); 559 560 /* 561 * Check for exact addresses in the hash bucket. 562 */ 563 LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) { 564 /* 565 * If the address matches, verify that the packet 566 * arrived via the correct interface if checking is 567 * enabled. 568 */ 569 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 570 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 571 goto ours; 572 } 573 /* 574 * Check for broadcast addresses. 575 * 576 * Only accept broadcast packets that arrive via the matching 577 * interface. Reception of forwarded directed broadcasts would 578 * be handled via ip_forward() and ether_output() with the loopback 579 * into the stack for SIMPLEX interfaces handled by ether_output(). 580 */ 581 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 582 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 583 if (ifa->ifa_addr->sa_family != AF_INET) 584 continue; 585 ia = ifatoia(ifa); 586 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 587 pkt_dst.s_addr) 588 goto ours; 589 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 590 goto ours; 591#ifdef BOOTP_COMPAT 592 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 593 goto ours; 594#endif 595 } 596 } 597 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 598 struct in_multi *inm; 599 if (ip_mrouter) { 600 /* 601 * If we are acting as a multicast router, all 602 * incoming multicast packets are passed to the 603 * kernel-level multicast forwarding function. 604 * The packet is returned (relatively) intact; if 605 * ip_mforward() returns a non-zero value, the packet 606 * must be discarded, else it may be accepted below. 607 */ 608 if (ip_mforward && 609 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 610 ipstat.ips_cantforward++; 611 m_freem(m); 612 return; 613 } 614 615 /* 616 * The process-level routing daemon needs to receive 617 * all multicast IGMP packets, whether or not this 618 * host belongs to their destination groups. 619 */ 620 if (ip->ip_p == IPPROTO_IGMP) 621 goto ours; 622 ipstat.ips_forward++; 623 } 624 /* 625 * See if we belong to the destination multicast group on the 626 * arrival interface. 627 */ 628 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 629 if (inm == NULL) { 630 ipstat.ips_notmember++; 631 m_freem(m); 632 return; 633 } 634 goto ours; 635 } 636 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 637 goto ours; 638 if (ip->ip_dst.s_addr == INADDR_ANY) 639 goto ours; 640 641 /* 642 * FAITH(Firewall Aided Internet Translator) 643 */ 644 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 645 if (ip_keepfaith) { 646 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 647 goto ours; 648 } 649 m_freem(m); 650 return; 651 } 652 653 /* 654 * Not for us; forward if possible and desirable. 655 */ 656 if (ipforwarding == 0) { 657 ipstat.ips_cantforward++; 658 m_freem(m); 659 } else { 660#ifdef IPSEC 661 /* 662 * Enforce inbound IPsec SPD. 663 */ 664 if (ipsec4_in_reject(m, NULL)) { 665 ipsecstat.in_polvio++; 666 goto bad; 667 } 668#endif /* IPSEC */ 669#ifdef FAST_IPSEC 670 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 671 s = splnet(); 672 if (mtag != NULL) { 673 tdbi = (struct tdb_ident *)(mtag + 1); 674 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 675 } else { 676 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 677 IP_FORWARDING, &error); 678 } 679 if (sp == NULL) { /* NB: can happen if error */ 680 splx(s); 681 /*XXX error stat???*/ 682 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 683 goto bad; 684 } 685 686 /* 687 * Check security policy against packet attributes. 688 */ 689 error = ipsec_in_reject(sp, m); 690 KEY_FREESP(&sp); 691 splx(s); 692 if (error) { 693 ipstat.ips_cantforward++; 694 goto bad; 695 } 696#endif /* FAST_IPSEC */ 697 ip_forward(m, 0, args.next_hop); 698 } 699 return; 700 701ours: 702#ifdef IPSTEALTH 703 /* 704 * IPSTEALTH: Process non-routing options only 705 * if the packet is destined for us. 706 */ 707 if (ipstealth && hlen > sizeof (struct ip) && 708 ip_dooptions(m, 1, args.next_hop)) 709 return; 710#endif /* IPSTEALTH */ 711 712 /* Count the packet in the ip address stats */ 713 if (ia != NULL) { 714 ia->ia_ifa.if_ipackets++; 715 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 716 } 717 718 /* 719 * If offset or IP_MF are set, must reassemble. 720 * Otherwise, nothing need be done. 721 * (We could look in the reassembly queue to see 722 * if the packet was previously fragmented, 723 * but it's not worth the time; just let them time out.) 724 */ 725 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 726 727 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 728 /* 729 * Look for queue of fragments 730 * of this datagram. 731 */ 732 TAILQ_FOREACH(fp, &ipq[sum], ipq_list) 733 if (ip->ip_id == fp->ipq_id && 734 ip->ip_src.s_addr == fp->ipq_src.s_addr && 735 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 736#ifdef MAC 737 mac_fragment_match(m, fp) && 738#endif 739 ip->ip_p == fp->ipq_p) 740 goto found; 741 742 fp = 0; 743 744 /* check if there's a place for the new queue */ 745 if (nipq > maxnipq) { 746 /* 747 * drop something from the tail of the current queue 748 * before proceeding further 749 */ 750 struct ipq *q = TAILQ_LAST(&ipq[sum], ipqhead); 751 if (q == NULL) { /* gak */ 752 for (i = 0; i < IPREASS_NHASH; i++) { 753 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 754 if (r) { 755 ip_freef(&ipq[i], r); 756 break; 757 } 758 } 759 } else 760 ip_freef(&ipq[sum], q); 761 } 762found: 763 /* 764 * Adjust ip_len to not reflect header, 765 * convert offset of this to bytes. 766 */ 767 ip->ip_len -= hlen; 768 if (ip->ip_off & IP_MF) { 769 /* 770 * Make sure that fragments have a data length 771 * that's a non-zero multiple of 8 bytes. 772 */ 773 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 774 ipstat.ips_toosmall++; /* XXX */ 775 goto bad; 776 } 777 m->m_flags |= M_FRAG; 778 } else 779 m->m_flags &= ~M_FRAG; 780 ip->ip_off <<= 3; 781 782 /* 783 * Attempt reassembly; if it succeeds, proceed. 784 * ip_reass() will return a different mbuf, and update 785 * the divert info in divert_info and args.divert_rule. 786 */ 787 ipstat.ips_fragments++; 788 m->m_pkthdr.header = ip; 789 m = ip_reass(m, 790 &ipq[sum], fp, &divert_info, &args.divert_rule); 791 if (m == 0) 792 return; 793 ipstat.ips_reassembled++; 794 ip = mtod(m, struct ip *); 795 /* Get the header length of the reassembled packet */ 796 hlen = ip->ip_hl << 2; 797#ifdef IPDIVERT 798 /* Restore original checksum before diverting packet */ 799 if (divert_info != 0) { 800 ip->ip_len += hlen; 801 ip->ip_len = htons(ip->ip_len); 802 ip->ip_off = htons(ip->ip_off); 803 ip->ip_sum = 0; 804 if (hlen == sizeof(struct ip)) 805 ip->ip_sum = in_cksum_hdr(ip); 806 else 807 ip->ip_sum = in_cksum(m, hlen); 808 ip->ip_off = ntohs(ip->ip_off); 809 ip->ip_len = ntohs(ip->ip_len); 810 ip->ip_len -= hlen; 811 } 812#endif 813 } else 814 ip->ip_len -= hlen; 815 816#ifdef IPDIVERT 817 /* 818 * Divert or tee packet to the divert protocol if required. 819 */ 820 if (divert_info != 0) { 821 struct mbuf *clone = NULL; 822 823 /* Clone packet if we're doing a 'tee' */ 824 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 825 clone = m_dup(m, M_DONTWAIT); 826 827 /* Restore packet header fields to original values */ 828 ip->ip_len += hlen; 829 ip->ip_len = htons(ip->ip_len); 830 ip->ip_off = htons(ip->ip_off); 831 832 /* Deliver packet to divert input routine */ 833 divert_packet(m, 1, divert_info & 0xffff, args.divert_rule); 834 ipstat.ips_delivered++; 835 836 /* If 'tee', continue with original packet */ 837 if (clone == NULL) 838 return; 839 m = clone; 840 ip = mtod(m, struct ip *); 841 ip->ip_len += hlen; 842 /* 843 * Jump backwards to complete processing of the 844 * packet. But first clear divert_info to avoid 845 * entering this block again. 846 * We do not need to clear args.divert_rule 847 * or args.next_hop as they will not be used. 848 */ 849 divert_info = 0; 850 goto pass; 851 } 852#endif 853 854#ifdef IPSEC 855 /* 856 * enforce IPsec policy checking if we are seeing last header. 857 * note that we do not visit this with protocols with pcb layer 858 * code - like udp/tcp/raw ip. 859 */ 860 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 861 ipsec4_in_reject(m, NULL)) { 862 ipsecstat.in_polvio++; 863 goto bad; 864 } 865#endif 866#if FAST_IPSEC 867 /* 868 * enforce IPsec policy checking if we are seeing last header. 869 * note that we do not visit this with protocols with pcb layer 870 * code - like udp/tcp/raw ip. 871 */ 872 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 873 /* 874 * Check if the packet has already had IPsec processing 875 * done. If so, then just pass it along. This tag gets 876 * set during AH, ESP, etc. input handling, before the 877 * packet is returned to the ip input queue for delivery. 878 */ 879 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 880 s = splnet(); 881 if (mtag != NULL) { 882 tdbi = (struct tdb_ident *)(mtag + 1); 883 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 884 } else { 885 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 886 IP_FORWARDING, &error); 887 } 888 if (sp != NULL) { 889 /* 890 * Check security policy against packet attributes. 891 */ 892 error = ipsec_in_reject(sp, m); 893 KEY_FREESP(&sp); 894 } else { 895 /* XXX error stat??? */ 896 error = EINVAL; 897DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 898 goto bad; 899 } 900 splx(s); 901 if (error) 902 goto bad; 903 } 904#endif /* FAST_IPSEC */ 905 906 /* 907 * Switch out to protocol's input routine. 908 */ 909 ipstat.ips_delivered++; 910 if (args.next_hop && ip->ip_p == IPPROTO_TCP) { 911 /* TCP needs IPFORWARD info if available */ 912 struct m_hdr tag; 913 914 tag.mh_type = MT_TAG; 915 tag.mh_flags = PACKET_TAG_IPFORWARD; 916 tag.mh_data = (caddr_t)args.next_hop; 917 tag.mh_next = m; 918 919 (*inetsw[ip_protox[ip->ip_p]].pr_input)( 920 (struct mbuf *)&tag, hlen); 921 } else 922 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 923 return; 924bad: 925 m_freem(m); 926} 927 928/* 929 * IP software interrupt routine - to go away sometime soon 930 */ 931static void 932ipintr(void) 933{ 934 struct mbuf *m; 935 936 while (1) { 937 IF_DEQUEUE(&ipintrq, m); 938 if (m == 0) 939 return; 940 ip_input(m); 941 } 942} 943 944/* 945 * Take incoming datagram fragment and try to reassemble it into 946 * whole datagram. If a chain for reassembly of this datagram already 947 * exists, then it is given as fp; otherwise have to make a chain. 948 * 949 * When IPDIVERT enabled, keep additional state with each packet that 950 * tells us if we need to divert or tee the packet we're building. 951 * In particular, *divinfo includes the port and TEE flag, 952 * *divert_rule is the number of the matching rule. 953 */ 954 955static struct mbuf * 956ip_reass(struct mbuf *m, struct ipqhead *head, struct ipq *fp, 957 u_int32_t *divinfo, u_int16_t *divert_rule) 958{ 959 struct ip *ip = mtod(m, struct ip *); 960 register struct mbuf *p, *q, *nq; 961 struct mbuf *t; 962 int hlen = ip->ip_hl << 2; 963 int i, next; 964 965 /* 966 * Presence of header sizes in mbufs 967 * would confuse code below. 968 */ 969 m->m_data += hlen; 970 m->m_len -= hlen; 971 972 /* 973 * If first fragment to arrive, create a reassembly queue. 974 */ 975 if (fp == 0) { 976 /* 977 * Enforce upper bound on number of fragmented packets 978 * for which we attempt reassembly; 979 * If maxfrag is 0, never accept fragments. 980 * If maxfrag is -1, accept all fragments without limitation. 981 */ 982 if ((ip_maxfragpackets >= 0) && (ip_nfragpackets >= ip_maxfragpackets)) 983 goto dropfrag; 984 ip_nfragpackets++; 985 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 986 goto dropfrag; 987 fp = mtod(t, struct ipq *); 988#ifdef MAC 989 mac_init_ipq(fp); 990 mac_create_ipq(m, fp); 991#endif 992 TAILQ_INSERT_HEAD(head, fp, ipq_list); 993 nipq++; 994 fp->ipq_ttl = IPFRAGTTL; 995 fp->ipq_p = ip->ip_p; 996 fp->ipq_id = ip->ip_id; 997 fp->ipq_src = ip->ip_src; 998 fp->ipq_dst = ip->ip_dst; 999 fp->ipq_frags = m; 1000 m->m_nextpkt = NULL; 1001#ifdef IPDIVERT 1002 fp->ipq_div_info = 0; 1003 fp->ipq_div_cookie = 0; 1004#endif 1005 goto inserted; 1006 } else { 1007#ifdef MAC 1008 mac_update_ipq(m, fp); 1009#endif 1010 } 1011 1012#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1013 1014 /* 1015 * Find a segment which begins after this one does. 1016 */ 1017 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1018 if (GETIP(q)->ip_off > ip->ip_off) 1019 break; 1020 1021 /* 1022 * If there is a preceding segment, it may provide some of 1023 * our data already. If so, drop the data from the incoming 1024 * segment. If it provides all of our data, drop us, otherwise 1025 * stick new segment in the proper place. 1026 * 1027 * If some of the data is dropped from the the preceding 1028 * segment, then it's checksum is invalidated. 1029 */ 1030 if (p) { 1031 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1032 if (i > 0) { 1033 if (i >= ip->ip_len) 1034 goto dropfrag; 1035 m_adj(m, i); 1036 m->m_pkthdr.csum_flags = 0; 1037 ip->ip_off += i; 1038 ip->ip_len -= i; 1039 } 1040 m->m_nextpkt = p->m_nextpkt; 1041 p->m_nextpkt = m; 1042 } else { 1043 m->m_nextpkt = fp->ipq_frags; 1044 fp->ipq_frags = m; 1045 } 1046 1047 /* 1048 * While we overlap succeeding segments trim them or, 1049 * if they are completely covered, dequeue them. 1050 */ 1051 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1052 q = nq) { 1053 i = (ip->ip_off + ip->ip_len) - 1054 GETIP(q)->ip_off; 1055 if (i < GETIP(q)->ip_len) { 1056 GETIP(q)->ip_len -= i; 1057 GETIP(q)->ip_off += i; 1058 m_adj(q, i); 1059 q->m_pkthdr.csum_flags = 0; 1060 break; 1061 } 1062 nq = q->m_nextpkt; 1063 m->m_nextpkt = nq; 1064 m_freem(q); 1065 } 1066 1067inserted: 1068 1069#ifdef IPDIVERT 1070 /* 1071 * Transfer firewall instructions to the fragment structure. 1072 * Only trust info in the fragment at offset 0. 1073 */ 1074 if (ip->ip_off == 0) { 1075 fp->ipq_div_info = *divinfo; 1076 fp->ipq_div_cookie = *divert_rule; 1077 } 1078 *divinfo = 0; 1079 *divert_rule = 0; 1080#endif 1081 1082 /* 1083 * Check for complete reassembly. 1084 */ 1085 next = 0; 1086 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1087 if (GETIP(q)->ip_off != next) 1088 return (0); 1089 next += GETIP(q)->ip_len; 1090 } 1091 /* Make sure the last packet didn't have the IP_MF flag */ 1092 if (p->m_flags & M_FRAG) 1093 return (0); 1094 1095 /* 1096 * Reassembly is complete. Make sure the packet is a sane size. 1097 */ 1098 q = fp->ipq_frags; 1099 ip = GETIP(q); 1100 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1101 ipstat.ips_toolong++; 1102 ip_freef(head, fp); 1103 return (0); 1104 } 1105 1106 /* 1107 * Concatenate fragments. 1108 */ 1109 m = q; 1110 t = m->m_next; 1111 m->m_next = 0; 1112 m_cat(m, t); 1113 nq = q->m_nextpkt; 1114 q->m_nextpkt = 0; 1115 for (q = nq; q != NULL; q = nq) { 1116 nq = q->m_nextpkt; 1117 q->m_nextpkt = NULL; 1118 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1119 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1120 m_cat(m, q); 1121 } 1122#ifdef MAC 1123 mac_create_datagram_from_ipq(fp, m); 1124 mac_destroy_ipq(fp); 1125#endif 1126 1127#ifdef IPDIVERT 1128 /* 1129 * Extract firewall instructions from the fragment structure. 1130 */ 1131 *divinfo = fp->ipq_div_info; 1132 *divert_rule = fp->ipq_div_cookie; 1133#endif 1134 1135 /* 1136 * Create header for new ip packet by 1137 * modifying header of first packet; 1138 * dequeue and discard fragment reassembly header. 1139 * Make header visible. 1140 */ 1141 ip->ip_len = next; 1142 ip->ip_src = fp->ipq_src; 1143 ip->ip_dst = fp->ipq_dst; 1144 TAILQ_REMOVE(head, fp, ipq_list); 1145 nipq--; 1146 (void) m_free(dtom(fp)); 1147 ip_nfragpackets--; 1148 m->m_len += (ip->ip_hl << 2); 1149 m->m_data -= (ip->ip_hl << 2); 1150 /* some debugging cruft by sklower, below, will go away soon */ 1151 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1152 m_fixhdr(m); 1153 return (m); 1154 1155dropfrag: 1156#ifdef IPDIVERT 1157 *divinfo = 0; 1158 *divert_rule = 0; 1159#endif 1160 ipstat.ips_fragdropped++; 1161 m_freem(m); 1162 return (0); 1163 1164#undef GETIP 1165} 1166 1167/* 1168 * Free a fragment reassembly header and all 1169 * associated datagrams. 1170 */ 1171static void 1172ip_freef(fhp, fp) 1173 struct ipqhead *fhp; 1174 struct ipq *fp; 1175{ 1176 register struct mbuf *q; 1177 1178 while (fp->ipq_frags) { 1179 q = fp->ipq_frags; 1180 fp->ipq_frags = q->m_nextpkt; 1181 m_freem(q); 1182 } 1183 TAILQ_REMOVE(fhp, fp, ipq_list); 1184 (void) m_free(dtom(fp)); 1185 ip_nfragpackets--; 1186 nipq--; 1187} 1188 1189/* 1190 * IP timer processing; 1191 * if a timer expires on a reassembly 1192 * queue, discard it. 1193 */ 1194void 1195ip_slowtimo() 1196{ 1197 register struct ipq *fp; 1198 int s = splnet(); 1199 int i; 1200 1201 for (i = 0; i < IPREASS_NHASH; i++) { 1202 for(fp = TAILQ_FIRST(&ipq[i]); fp;) { 1203 struct ipq *fpp; 1204 1205 fpp = fp; 1206 fp = TAILQ_NEXT(fp, ipq_list); 1207 if(--fpp->ipq_ttl == 0) { 1208 ipstat.ips_fragtimeout++; 1209 ip_freef(&ipq[i], fpp); 1210 } 1211 } 1212 } 1213 /* 1214 * If we are over the maximum number of fragments 1215 * (due to the limit being lowered), drain off 1216 * enough to get down to the new limit. 1217 */ 1218 for (i = 0; i < IPREASS_NHASH; i++) { 1219 if (ip_maxfragpackets >= 0) { 1220 while (ip_nfragpackets > ip_maxfragpackets && 1221 !TAILQ_EMPTY(&ipq[i])) { 1222 ipstat.ips_fragdropped++; 1223 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1224 } 1225 } 1226 } 1227 ipflow_slowtimo(); 1228 splx(s); 1229} 1230 1231/* 1232 * Drain off all datagram fragments. 1233 */ 1234void 1235ip_drain() 1236{ 1237 int i; 1238 1239 for (i = 0; i < IPREASS_NHASH; i++) { 1240 while(!TAILQ_EMPTY(&ipq[i])) { 1241 ipstat.ips_fragdropped++; 1242 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1243 } 1244 } 1245 in_rtqdrain(); 1246} 1247 1248/* 1249 * Do option processing on a datagram, 1250 * possibly discarding it if bad options are encountered, 1251 * or forwarding it if source-routed. 1252 * The pass argument is used when operating in the IPSTEALTH 1253 * mode to tell what options to process: 1254 * [LS]SRR (pass 0) or the others (pass 1). 1255 * The reason for as many as two passes is that when doing IPSTEALTH, 1256 * non-routing options should be processed only if the packet is for us. 1257 * Returns 1 if packet has been forwarded/freed, 1258 * 0 if the packet should be processed further. 1259 */ 1260static int 1261ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop) 1262{ 1263 struct ip *ip = mtod(m, struct ip *); 1264 u_char *cp; 1265 struct in_ifaddr *ia; 1266 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1267 struct in_addr *sin, dst; 1268 n_time ntime; 1269 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 1270 1271 dst = ip->ip_dst; 1272 cp = (u_char *)(ip + 1); 1273 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1274 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1275 opt = cp[IPOPT_OPTVAL]; 1276 if (opt == IPOPT_EOL) 1277 break; 1278 if (opt == IPOPT_NOP) 1279 optlen = 1; 1280 else { 1281 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1282 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1283 goto bad; 1284 } 1285 optlen = cp[IPOPT_OLEN]; 1286 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1287 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1288 goto bad; 1289 } 1290 } 1291 switch (opt) { 1292 1293 default: 1294 break; 1295 1296 /* 1297 * Source routing with record. 1298 * Find interface with current destination address. 1299 * If none on this machine then drop if strictly routed, 1300 * or do nothing if loosely routed. 1301 * Record interface address and bring up next address 1302 * component. If strictly routed make sure next 1303 * address is on directly accessible net. 1304 */ 1305 case IPOPT_LSRR: 1306 case IPOPT_SSRR: 1307#ifdef IPSTEALTH 1308 if (ipstealth && pass > 0) 1309 break; 1310#endif 1311 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1312 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1313 goto bad; 1314 } 1315 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1316 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1317 goto bad; 1318 } 1319 ipaddr.sin_addr = ip->ip_dst; 1320 ia = (struct in_ifaddr *) 1321 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1322 if (ia == 0) { 1323 if (opt == IPOPT_SSRR) { 1324 type = ICMP_UNREACH; 1325 code = ICMP_UNREACH_SRCFAIL; 1326 goto bad; 1327 } 1328 if (!ip_dosourceroute) 1329 goto nosourcerouting; 1330 /* 1331 * Loose routing, and not at next destination 1332 * yet; nothing to do except forward. 1333 */ 1334 break; 1335 } 1336 off--; /* 0 origin */ 1337 if (off > optlen - (int)sizeof(struct in_addr)) { 1338 /* 1339 * End of source route. Should be for us. 1340 */ 1341 if (!ip_acceptsourceroute) 1342 goto nosourcerouting; 1343 save_rte(cp, ip->ip_src); 1344 break; 1345 } 1346#ifdef IPSTEALTH 1347 if (ipstealth) 1348 goto dropit; 1349#endif 1350 if (!ip_dosourceroute) { 1351 if (ipforwarding) { 1352 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1353 /* 1354 * Acting as a router, so generate ICMP 1355 */ 1356nosourcerouting: 1357 strcpy(buf, inet_ntoa(ip->ip_dst)); 1358 log(LOG_WARNING, 1359 "attempted source route from %s to %s\n", 1360 inet_ntoa(ip->ip_src), buf); 1361 type = ICMP_UNREACH; 1362 code = ICMP_UNREACH_SRCFAIL; 1363 goto bad; 1364 } else { 1365 /* 1366 * Not acting as a router, so silently drop. 1367 */ 1368#ifdef IPSTEALTH 1369dropit: 1370#endif 1371 ipstat.ips_cantforward++; 1372 m_freem(m); 1373 return (1); 1374 } 1375 } 1376 1377 /* 1378 * locate outgoing interface 1379 */ 1380 (void)memcpy(&ipaddr.sin_addr, cp + off, 1381 sizeof(ipaddr.sin_addr)); 1382 1383 if (opt == IPOPT_SSRR) { 1384#define INA struct in_ifaddr * 1385#define SA struct sockaddr * 1386 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1387 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1388 } else 1389 ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt); 1390 if (ia == 0) { 1391 type = ICMP_UNREACH; 1392 code = ICMP_UNREACH_SRCFAIL; 1393 goto bad; 1394 } 1395 ip->ip_dst = ipaddr.sin_addr; 1396 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1397 sizeof(struct in_addr)); 1398 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1399 /* 1400 * Let ip_intr's mcast routing check handle mcast pkts 1401 */ 1402 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1403 break; 1404 1405 case IPOPT_RR: 1406#ifdef IPSTEALTH 1407 if (ipstealth && pass == 0) 1408 break; 1409#endif 1410 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1411 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1412 goto bad; 1413 } 1414 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1415 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1416 goto bad; 1417 } 1418 /* 1419 * If no space remains, ignore. 1420 */ 1421 off--; /* 0 origin */ 1422 if (off > optlen - (int)sizeof(struct in_addr)) 1423 break; 1424 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1425 sizeof(ipaddr.sin_addr)); 1426 /* 1427 * locate outgoing interface; if we're the destination, 1428 * use the incoming interface (should be same). 1429 */ 1430 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1431 (ia = ip_rtaddr(ipaddr.sin_addr, 1432 &ipforward_rt)) == 0) { 1433 type = ICMP_UNREACH; 1434 code = ICMP_UNREACH_HOST; 1435 goto bad; 1436 } 1437 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1438 sizeof(struct in_addr)); 1439 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1440 break; 1441 1442 case IPOPT_TS: 1443#ifdef IPSTEALTH 1444 if (ipstealth && pass == 0) 1445 break; 1446#endif 1447 code = cp - (u_char *)ip; 1448 if (optlen < 4 || optlen > 40) { 1449 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1450 goto bad; 1451 } 1452 if ((off = cp[IPOPT_OFFSET]) < 5) { 1453 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1454 goto bad; 1455 } 1456 if (off > optlen - (int)sizeof(int32_t)) { 1457 cp[IPOPT_OFFSET + 1] += (1 << 4); 1458 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1459 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1460 goto bad; 1461 } 1462 break; 1463 } 1464 off--; /* 0 origin */ 1465 sin = (struct in_addr *)(cp + off); 1466 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1467 1468 case IPOPT_TS_TSONLY: 1469 break; 1470 1471 case IPOPT_TS_TSANDADDR: 1472 if (off + sizeof(n_time) + 1473 sizeof(struct in_addr) > optlen) { 1474 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1475 goto bad; 1476 } 1477 ipaddr.sin_addr = dst; 1478 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1479 m->m_pkthdr.rcvif); 1480 if (ia == 0) 1481 continue; 1482 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1483 sizeof(struct in_addr)); 1484 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1485 off += sizeof(struct in_addr); 1486 break; 1487 1488 case IPOPT_TS_PRESPEC: 1489 if (off + sizeof(n_time) + 1490 sizeof(struct in_addr) > optlen) { 1491 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1492 goto bad; 1493 } 1494 (void)memcpy(&ipaddr.sin_addr, sin, 1495 sizeof(struct in_addr)); 1496 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1497 continue; 1498 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1499 off += sizeof(struct in_addr); 1500 break; 1501 1502 default: 1503 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1504 goto bad; 1505 } 1506 ntime = iptime(); 1507 (void)memcpy(cp + off, &ntime, sizeof(n_time)); 1508 cp[IPOPT_OFFSET] += sizeof(n_time); 1509 } 1510 } 1511 if (forward && ipforwarding) { 1512 ip_forward(m, 1, next_hop); 1513 return (1); 1514 } 1515 return (0); 1516bad: 1517 icmp_error(m, type, code, 0, 0); 1518 ipstat.ips_badoptions++; 1519 return (1); 1520} 1521 1522/* 1523 * Given address of next destination (final or next hop), 1524 * return internet address info of interface to be used to get there. 1525 */ 1526struct in_ifaddr * 1527ip_rtaddr(dst, rt) 1528 struct in_addr dst; 1529 struct route *rt; 1530{ 1531 register struct sockaddr_in *sin; 1532 1533 sin = (struct sockaddr_in *)&rt->ro_dst; 1534 1535 if (rt->ro_rt == 0 || 1536 !(rt->ro_rt->rt_flags & RTF_UP) || 1537 dst.s_addr != sin->sin_addr.s_addr) { 1538 if (rt->ro_rt) { 1539 RTFREE(rt->ro_rt); 1540 rt->ro_rt = 0; 1541 } 1542 sin->sin_family = AF_INET; 1543 sin->sin_len = sizeof(*sin); 1544 sin->sin_addr = dst; 1545 1546 rtalloc_ign(rt, RTF_PRCLONING); 1547 } 1548 if (rt->ro_rt == 0) 1549 return ((struct in_ifaddr *)0); 1550 return (ifatoia(rt->ro_rt->rt_ifa)); 1551} 1552 1553/* 1554 * Save incoming source route for use in replies, 1555 * to be picked up later by ip_srcroute if the receiver is interested. 1556 */ 1557static void 1558save_rte(option, dst) 1559 u_char *option; 1560 struct in_addr dst; 1561{ 1562 unsigned olen; 1563 1564 olen = option[IPOPT_OLEN]; 1565#ifdef DIAGNOSTIC 1566 if (ipprintfs) 1567 printf("save_rte: olen %d\n", olen); 1568#endif 1569 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1570 return; 1571 bcopy(option, ip_srcrt.srcopt, olen); 1572 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1573 ip_srcrt.dst = dst; 1574} 1575 1576/* 1577 * Retrieve incoming source route for use in replies, 1578 * in the same form used by setsockopt. 1579 * The first hop is placed before the options, will be removed later. 1580 */ 1581struct mbuf * 1582ip_srcroute() 1583{ 1584 register struct in_addr *p, *q; 1585 register struct mbuf *m; 1586 1587 if (ip_nhops == 0) 1588 return ((struct mbuf *)0); 1589 m = m_get(M_DONTWAIT, MT_HEADER); 1590 if (m == 0) 1591 return ((struct mbuf *)0); 1592 1593#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1594 1595 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1596 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1597 OPTSIZ; 1598#ifdef DIAGNOSTIC 1599 if (ipprintfs) 1600 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1601#endif 1602 1603 /* 1604 * First save first hop for return route 1605 */ 1606 p = &ip_srcrt.route[ip_nhops - 1]; 1607 *(mtod(m, struct in_addr *)) = *p--; 1608#ifdef DIAGNOSTIC 1609 if (ipprintfs) 1610 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1611#endif 1612 1613 /* 1614 * Copy option fields and padding (nop) to mbuf. 1615 */ 1616 ip_srcrt.nop = IPOPT_NOP; 1617 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1618 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1619 &ip_srcrt.nop, OPTSIZ); 1620 q = (struct in_addr *)(mtod(m, caddr_t) + 1621 sizeof(struct in_addr) + OPTSIZ); 1622#undef OPTSIZ 1623 /* 1624 * Record return path as an IP source route, 1625 * reversing the path (pointers are now aligned). 1626 */ 1627 while (p >= ip_srcrt.route) { 1628#ifdef DIAGNOSTIC 1629 if (ipprintfs) 1630 printf(" %lx", (u_long)ntohl(q->s_addr)); 1631#endif 1632 *q++ = *p--; 1633 } 1634 /* 1635 * Last hop goes to final destination. 1636 */ 1637 *q = ip_srcrt.dst; 1638#ifdef DIAGNOSTIC 1639 if (ipprintfs) 1640 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1641#endif 1642 return (m); 1643} 1644 1645/* 1646 * Strip out IP options, at higher 1647 * level protocol in the kernel. 1648 * Second argument is buffer to which options 1649 * will be moved, and return value is their length. 1650 * XXX should be deleted; last arg currently ignored. 1651 */ 1652void 1653ip_stripoptions(m, mopt) 1654 register struct mbuf *m; 1655 struct mbuf *mopt; 1656{ 1657 register int i; 1658 struct ip *ip = mtod(m, struct ip *); 1659 register caddr_t opts; 1660 int olen; 1661 1662 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1663 opts = (caddr_t)(ip + 1); 1664 i = m->m_len - (sizeof (struct ip) + olen); 1665 bcopy(opts + olen, opts, (unsigned)i); 1666 m->m_len -= olen; 1667 if (m->m_flags & M_PKTHDR) 1668 m->m_pkthdr.len -= olen; 1669 ip->ip_v = IPVERSION; 1670 ip->ip_hl = sizeof(struct ip) >> 2; 1671} 1672 1673u_char inetctlerrmap[PRC_NCMDS] = { 1674 0, 0, 0, 0, 1675 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1676 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1677 EMSGSIZE, EHOSTUNREACH, 0, 0, 1678 0, 0, 0, 0, 1679 ENOPROTOOPT, ECONNREFUSED 1680}; 1681 1682/* 1683 * Forward a packet. If some error occurs return the sender 1684 * an icmp packet. Note we can't always generate a meaningful 1685 * icmp message because icmp doesn't have a large enough repertoire 1686 * of codes and types. 1687 * 1688 * If not forwarding, just drop the packet. This could be confusing 1689 * if ipforwarding was zero but some routing protocol was advancing 1690 * us as a gateway to somewhere. However, we must let the routing 1691 * protocol deal with that. 1692 * 1693 * The srcrt parameter indicates whether the packet is being forwarded 1694 * via a source route. 1695 */ 1696static void 1697ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop) 1698{ 1699 struct ip *ip = mtod(m, struct ip *); 1700 struct rtentry *rt; 1701 int error, type = 0, code = 0; 1702 struct mbuf *mcopy; 1703 n_long dest; 1704 struct in_addr pkt_dst; 1705 struct ifnet *destifp; 1706#if defined(IPSEC) || defined(FAST_IPSEC) 1707 struct ifnet dummyifp; 1708#endif 1709 1710 dest = 0; 1711 /* 1712 * Cache the destination address of the packet; this may be 1713 * changed by use of 'ipfw fwd'. 1714 */ 1715 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst; 1716 1717#ifdef DIAGNOSTIC 1718 if (ipprintfs) 1719 printf("forward: src %lx dst %lx ttl %x\n", 1720 (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr, 1721 ip->ip_ttl); 1722#endif 1723 1724 1725 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) { 1726 ipstat.ips_cantforward++; 1727 m_freem(m); 1728 return; 1729 } 1730#ifdef IPSTEALTH 1731 if (!ipstealth) { 1732#endif 1733 if (ip->ip_ttl <= IPTTLDEC) { 1734 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1735 dest, 0); 1736 return; 1737 } 1738#ifdef IPSTEALTH 1739 } 1740#endif 1741 1742 if (ip_rtaddr(pkt_dst, &ipforward_rt) == 0) { 1743 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1744 return; 1745 } else 1746 rt = ipforward_rt.ro_rt; 1747 1748 /* 1749 * Save the IP header and at most 8 bytes of the payload, 1750 * in case we need to generate an ICMP message to the src. 1751 * 1752 * XXX this can be optimized a lot by saving the data in a local 1753 * buffer on the stack (72 bytes at most), and only allocating the 1754 * mbuf if really necessary. The vast majority of the packets 1755 * are forwarded without having to send an ICMP back (either 1756 * because unnecessary, or because rate limited), so we are 1757 * really we are wasting a lot of work here. 1758 * 1759 * We don't use m_copy() because it might return a reference 1760 * to a shared cluster. Both this function and ip_output() 1761 * assume exclusive access to the IP header in `m', so any 1762 * data in a cluster may change before we reach icmp_error(). 1763 */ 1764 MGET(mcopy, M_DONTWAIT, m->m_type); 1765 if (mcopy != NULL) { 1766 M_COPY_PKTHDR(mcopy, m); 1767 mcopy->m_len = imin((ip->ip_hl << 2) + 8, 1768 (int)ip->ip_len); 1769 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1770#ifdef MAC 1771 /* 1772 * XXXMAC: This will eventually become an explicit 1773 * labeling point. 1774 */ 1775 mac_create_mbuf_from_mbuf(m, mcopy); 1776#endif 1777 } 1778 1779#ifdef IPSTEALTH 1780 if (!ipstealth) { 1781#endif 1782 ip->ip_ttl -= IPTTLDEC; 1783#ifdef IPSTEALTH 1784 } 1785#endif 1786 1787 /* 1788 * If forwarding packet using same interface that it came in on, 1789 * perhaps should send a redirect to sender to shortcut a hop. 1790 * Only send redirect if source is sending directly to us, 1791 * and if packet was not source routed (or has any options). 1792 * Also, don't send redirect if forwarding using a default route 1793 * or a route modified by a redirect. 1794 */ 1795 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1796 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1797 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1798 ipsendredirects && !srcrt && !next_hop) { 1799#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1800 u_long src = ntohl(ip->ip_src.s_addr); 1801 1802 if (RTA(rt) && 1803 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1804 if (rt->rt_flags & RTF_GATEWAY) 1805 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1806 else 1807 dest = pkt_dst.s_addr; 1808 /* Router requirements says to only send host redirects */ 1809 type = ICMP_REDIRECT; 1810 code = ICMP_REDIRECT_HOST; 1811#ifdef DIAGNOSTIC 1812 if (ipprintfs) 1813 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1814#endif 1815 } 1816 } 1817 1818 { 1819 struct m_hdr tag; 1820 1821 if (next_hop) { 1822 /* Pass IPFORWARD info if available */ 1823 1824 tag.mh_type = MT_TAG; 1825 tag.mh_flags = PACKET_TAG_IPFORWARD; 1826 tag.mh_data = (caddr_t)next_hop; 1827 tag.mh_next = m; 1828 m = (struct mbuf *)&tag; 1829 } 1830 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1831 IP_FORWARDING, 0, NULL); 1832 } 1833 if (error) 1834 ipstat.ips_cantforward++; 1835 else { 1836 ipstat.ips_forward++; 1837 if (type) 1838 ipstat.ips_redirectsent++; 1839 else { 1840 if (mcopy) { 1841 ipflow_create(&ipforward_rt, mcopy); 1842 m_freem(mcopy); 1843 } 1844 return; 1845 } 1846 } 1847 if (mcopy == NULL) 1848 return; 1849 destifp = NULL; 1850 1851 switch (error) { 1852 1853 case 0: /* forwarded, but need redirect */ 1854 /* type, code set above */ 1855 break; 1856 1857 case ENETUNREACH: /* shouldn't happen, checked above */ 1858 case EHOSTUNREACH: 1859 case ENETDOWN: 1860 case EHOSTDOWN: 1861 default: 1862 type = ICMP_UNREACH; 1863 code = ICMP_UNREACH_HOST; 1864 break; 1865 1866 case EMSGSIZE: 1867 type = ICMP_UNREACH; 1868 code = ICMP_UNREACH_NEEDFRAG; 1869#ifdef IPSEC 1870 /* 1871 * If the packet is routed over IPsec tunnel, tell the 1872 * originator the tunnel MTU. 1873 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1874 * XXX quickhack!!! 1875 */ 1876 if (ipforward_rt.ro_rt) { 1877 struct secpolicy *sp = NULL; 1878 int ipsecerror; 1879 int ipsechdr; 1880 struct route *ro; 1881 1882 sp = ipsec4_getpolicybyaddr(mcopy, 1883 IPSEC_DIR_OUTBOUND, 1884 IP_FORWARDING, 1885 &ipsecerror); 1886 1887 if (sp == NULL) 1888 destifp = ipforward_rt.ro_rt->rt_ifp; 1889 else { 1890 /* count IPsec header size */ 1891 ipsechdr = ipsec4_hdrsiz(mcopy, 1892 IPSEC_DIR_OUTBOUND, 1893 NULL); 1894 1895 /* 1896 * find the correct route for outer IPv4 1897 * header, compute tunnel MTU. 1898 * 1899 * XXX BUG ALERT 1900 * The "dummyifp" code relies upon the fact 1901 * that icmp_error() touches only ifp->if_mtu. 1902 */ 1903 /*XXX*/ 1904 destifp = NULL; 1905 if (sp->req != NULL 1906 && sp->req->sav != NULL 1907 && sp->req->sav->sah != NULL) { 1908 ro = &sp->req->sav->sah->sa_route; 1909 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1910 dummyifp.if_mtu = 1911 ro->ro_rt->rt_ifp->if_mtu; 1912 dummyifp.if_mtu -= ipsechdr; 1913 destifp = &dummyifp; 1914 } 1915 } 1916 1917 key_freesp(sp); 1918 } 1919 } 1920#elif FAST_IPSEC 1921 /* 1922 * If the packet is routed over IPsec tunnel, tell the 1923 * originator the tunnel MTU. 1924 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1925 * XXX quickhack!!! 1926 */ 1927 if (ipforward_rt.ro_rt) { 1928 struct secpolicy *sp = NULL; 1929 int ipsecerror; 1930 int ipsechdr; 1931 struct route *ro; 1932 1933 sp = ipsec_getpolicybyaddr(mcopy, 1934 IPSEC_DIR_OUTBOUND, 1935 IP_FORWARDING, 1936 &ipsecerror); 1937 1938 if (sp == NULL) 1939 destifp = ipforward_rt.ro_rt->rt_ifp; 1940 else { 1941 /* count IPsec header size */ 1942 ipsechdr = ipsec4_hdrsiz(mcopy, 1943 IPSEC_DIR_OUTBOUND, 1944 NULL); 1945 1946 /* 1947 * find the correct route for outer IPv4 1948 * header, compute tunnel MTU. 1949 * 1950 * XXX BUG ALERT 1951 * The "dummyifp" code relies upon the fact 1952 * that icmp_error() touches only ifp->if_mtu. 1953 */ 1954 /*XXX*/ 1955 destifp = NULL; 1956 if (sp->req != NULL 1957 && sp->req->sav != NULL 1958 && sp->req->sav->sah != NULL) { 1959 ro = &sp->req->sav->sah->sa_route; 1960 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1961 dummyifp.if_mtu = 1962 ro->ro_rt->rt_ifp->if_mtu; 1963 dummyifp.if_mtu -= ipsechdr; 1964 destifp = &dummyifp; 1965 } 1966 } 1967 1968 KEY_FREESP(&sp); 1969 } 1970 } 1971#else /* !IPSEC && !FAST_IPSEC */ 1972 if (ipforward_rt.ro_rt) 1973 destifp = ipforward_rt.ro_rt->rt_ifp; 1974#endif /*IPSEC*/ 1975 ipstat.ips_cantfrag++; 1976 break; 1977 1978 case ENOBUFS: 1979 /* 1980 * A router should not generate ICMP_SOURCEQUENCH as 1981 * required in RFC1812 Requirements for IP Version 4 Routers. 1982 * Source quench could be a big problem under DoS attacks, 1983 * or if the underlying interface is rate-limited. 1984 * Those who need source quench packets may re-enable them 1985 * via the net.inet.ip.sendsourcequench sysctl. 1986 */ 1987 if (ip_sendsourcequench == 0) { 1988 m_freem(mcopy); 1989 return; 1990 } else { 1991 type = ICMP_SOURCEQUENCH; 1992 code = 0; 1993 } 1994 break; 1995 1996 case EACCES: /* ipfw denied packet */ 1997 m_freem(mcopy); 1998 return; 1999 } 2000 icmp_error(mcopy, type, code, dest, destifp); 2001} 2002 2003void 2004ip_savecontrol(inp, mp, ip, m) 2005 register struct inpcb *inp; 2006 register struct mbuf **mp; 2007 register struct ip *ip; 2008 register struct mbuf *m; 2009{ 2010 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2011 struct timeval tv; 2012 2013 microtime(&tv); 2014 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2015 SCM_TIMESTAMP, SOL_SOCKET); 2016 if (*mp) 2017 mp = &(*mp)->m_next; 2018 } 2019 if (inp->inp_flags & INP_RECVDSTADDR) { 2020 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2021 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2022 if (*mp) 2023 mp = &(*mp)->m_next; 2024 } 2025#ifdef notyet 2026 /* XXX 2027 * Moving these out of udp_input() made them even more broken 2028 * than they already were. 2029 */ 2030 /* options were tossed already */ 2031 if (inp->inp_flags & INP_RECVOPTS) { 2032 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2033 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2034 if (*mp) 2035 mp = &(*mp)->m_next; 2036 } 2037 /* ip_srcroute doesn't do what we want here, need to fix */ 2038 if (inp->inp_flags & INP_RECVRETOPTS) { 2039 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 2040 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2041 if (*mp) 2042 mp = &(*mp)->m_next; 2043 } 2044#endif 2045 if (inp->inp_flags & INP_RECVIF) { 2046 struct ifnet *ifp; 2047 struct sdlbuf { 2048 struct sockaddr_dl sdl; 2049 u_char pad[32]; 2050 } sdlbuf; 2051 struct sockaddr_dl *sdp; 2052 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2053 2054 if (((ifp = m->m_pkthdr.rcvif)) 2055 && ( ifp->if_index && (ifp->if_index <= if_index))) { 2056 sdp = (struct sockaddr_dl *) 2057 (ifaddr_byindex(ifp->if_index)->ifa_addr); 2058 /* 2059 * Change our mind and don't try copy. 2060 */ 2061 if ((sdp->sdl_family != AF_LINK) 2062 || (sdp->sdl_len > sizeof(sdlbuf))) { 2063 goto makedummy; 2064 } 2065 bcopy(sdp, sdl2, sdp->sdl_len); 2066 } else { 2067makedummy: 2068 sdl2->sdl_len 2069 = offsetof(struct sockaddr_dl, sdl_data[0]); 2070 sdl2->sdl_family = AF_LINK; 2071 sdl2->sdl_index = 0; 2072 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2073 } 2074 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 2075 IP_RECVIF, IPPROTO_IP); 2076 if (*mp) 2077 mp = &(*mp)->m_next; 2078 } 2079} 2080 2081/* 2082 * XXX these routines are called from the upper part of the kernel. 2083 * They need to be locked when we remove Giant. 2084 * 2085 * They could also be moved to ip_mroute.c, since all the RSVP 2086 * handling is done there already. 2087 */ 2088static int ip_rsvp_on; 2089struct socket *ip_rsvpd; 2090int 2091ip_rsvp_init(struct socket *so) 2092{ 2093 if (so->so_type != SOCK_RAW || 2094 so->so_proto->pr_protocol != IPPROTO_RSVP) 2095 return EOPNOTSUPP; 2096 2097 if (ip_rsvpd != NULL) 2098 return EADDRINUSE; 2099 2100 ip_rsvpd = so; 2101 /* 2102 * This may seem silly, but we need to be sure we don't over-increment 2103 * the RSVP counter, in case something slips up. 2104 */ 2105 if (!ip_rsvp_on) { 2106 ip_rsvp_on = 1; 2107 rsvp_on++; 2108 } 2109 2110 return 0; 2111} 2112 2113int 2114ip_rsvp_done(void) 2115{ 2116 ip_rsvpd = NULL; 2117 /* 2118 * This may seem silly, but we need to be sure we don't over-decrement 2119 * the RSVP counter, in case something slips up. 2120 */ 2121 if (ip_rsvp_on) { 2122 ip_rsvp_on = 0; 2123 rsvp_on--; 2124 } 2125 return 0; 2126} 2127 2128void 2129rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 2130{ 2131 if (rsvp_input_p) { /* call the real one if loaded */ 2132 rsvp_input_p(m, off); 2133 return; 2134 } 2135 2136 /* Can still get packets with rsvp_on = 0 if there is a local member 2137 * of the group to which the RSVP packet is addressed. But in this 2138 * case we want to throw the packet away. 2139 */ 2140 2141 if (!rsvp_on) { 2142 m_freem(m); 2143 return; 2144 } 2145 2146 if (ip_rsvpd != NULL) { 2147 rip_input(m, off); 2148 return; 2149 } 2150 /* Drop the packet */ 2151 m_freem(m); 2152} 2153