ip_reass.c revision 77665
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 77665 2001-06-03 23:33:23Z jesper $ 35 */ 36 37#define _IP_VHL 38 39#include "opt_bootp.h" 40#include "opt_ipfw.h" 41#include "opt_ipdn.h" 42#include "opt_ipdivert.h" 43#include "opt_ipfilter.h" 44#include "opt_ipstealth.h" 45#include "opt_ipsec.h" 46#include "opt_pfil_hooks.h" 47#include "opt_random_ip_id.h" 48 49#include <sys/param.h> 50#include <sys/systm.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_var.h> 64#include <net/if_dl.h> 65#include <net/route.h> 66#include <net/netisr.h> 67#include <net/intrq.h> 68 69#include <netinet/in.h> 70#include <netinet/in_systm.h> 71#include <netinet/in_var.h> 72#include <netinet/ip.h> 73#include <netinet/in_pcb.h> 74#include <netinet/ip_var.h> 75#include <netinet/ip_icmp.h> 76#include <machine/in_cksum.h> 77 78#include <netinet/ipprotosw.h> 79 80#include <sys/socketvar.h> 81 82#include <netinet/ip_fw.h> 83 84#ifdef IPSEC 85#include <netinet6/ipsec.h> 86#include <netkey/key.h> 87#endif 88 89#include "faith.h" 90#if defined(NFAITH) && NFAITH > 0 91#include <net/if_types.h> 92#endif 93 94#ifdef DUMMYNET 95#include <netinet/ip_dummynet.h> 96#endif 97 98int rsvp_on = 0; 99static int ip_rsvp_on; 100struct socket *ip_rsvpd; 101 102int ipforwarding = 0; 103SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 104 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 105 106static int ipsendredirects = 1; /* XXX */ 107SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 108 &ipsendredirects, 0, "Enable sending IP redirects"); 109 110int ip_defttl = IPDEFTTL; 111SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 112 &ip_defttl, 0, "Maximum TTL on IP packets"); 113 114static int ip_dosourceroute = 0; 115SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 116 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 117 118static int ip_acceptsourceroute = 0; 119SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 120 CTLFLAG_RW, &ip_acceptsourceroute, 0, 121 "Enable accepting source routed IP packets"); 122 123static int ip_keepfaith = 0; 124SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 125 &ip_keepfaith, 0, 126 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 127 128static int ip_nfragpackets = 0; 129static int ip_maxfragpackets = 200; 130SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 131 &ip_maxfragpackets, 0, 132 "Maximum number of IPv4 fragment reassembly queue entries"); 133 134/* 135 * XXX - Setting ip_checkinterface mostly implements the receive side of 136 * the Strong ES model described in RFC 1122, but since the routing table 137 * and transmit implementation do not implement the Strong ES model, 138 * setting this to 1 results in an odd hybrid. 139 * 140 * XXX - ip_checkinterface currently must be disabled if you use ipnat 141 * to translate the destination address to another local interface. 142 * 143 * XXX - ip_checkinterface must be disabled if you add IP aliases 144 * to the loopback interface instead of the interface where the 145 * packets for those addresses are received. 146 */ 147static int ip_checkinterface = 1; 148SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 149 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 150 151#ifdef DIAGNOSTIC 152static int ipprintfs = 0; 153#endif 154 155extern struct domain inetdomain; 156extern struct ipprotosw inetsw[]; 157u_char ip_protox[IPPROTO_MAX]; 158static int ipqmaxlen = IFQ_MAXLEN; 159struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 160SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 161 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 162SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 163 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 164 165struct ipstat ipstat; 166SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD, 167 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 168 169/* Packet reassembly stuff */ 170#define IPREASS_NHASH_LOG2 6 171#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 172#define IPREASS_HMASK (IPREASS_NHASH - 1) 173#define IPREASS_HASH(x,y) \ 174 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 175 176static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 177static int nipq = 0; /* total # of reass queues */ 178static int maxnipq; 179const int ipintrq_present = 1; 180 181#ifdef IPCTL_DEFMTU 182SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 183 &ip_mtu, 0, "Default MTU"); 184#endif 185 186#ifdef IPSTEALTH 187static int ipstealth = 0; 188SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 189 &ipstealth, 0, ""); 190#endif 191 192 193/* Firewall hooks */ 194ip_fw_chk_t *ip_fw_chk_ptr; 195ip_fw_ctl_t *ip_fw_ctl_ptr; 196int fw_enable = 1 ; 197 198#ifdef DUMMYNET 199ip_dn_ctl_t *ip_dn_ctl_ptr; 200#endif 201 202 203/* 204 * We need to save the IP options in case a protocol wants to respond 205 * to an incoming packet over the same route if the packet got here 206 * using IP source routing. This allows connection establishment and 207 * maintenance when the remote end is on a network that is not known 208 * to us. 209 */ 210static int ip_nhops = 0; 211static struct ip_srcrt { 212 struct in_addr dst; /* final destination */ 213 char nop; /* one NOP to align */ 214 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 215 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 216} ip_srcrt; 217 218struct sockaddr_in *ip_fw_fwd_addr; 219 220static void save_rte __P((u_char *, struct in_addr)); 221static int ip_dooptions __P((struct mbuf *)); 222static void ip_forward __P((struct mbuf *, int)); 223static void ip_freef __P((struct ipqhead *, struct ipq *)); 224#ifdef IPDIVERT 225static struct mbuf *ip_reass __P((struct mbuf *, struct ipqhead *, struct ipq *, u_int32_t *, u_int16_t *)); 226#else 227static struct mbuf *ip_reass __P((struct mbuf *, struct ipqhead *, struct ipq *)); 228#endif 229static struct in_ifaddr *ip_rtaddr __P((struct in_addr)); 230static void ipintr __P((void)); 231 232/* 233 * IP initialization: fill in IP protocol switch table. 234 * All protocols not implemented in kernel go to raw IP protocol handler. 235 */ 236void 237ip_init() 238{ 239 register struct ipprotosw *pr; 240 register int i; 241 242 TAILQ_INIT(&in_ifaddrhead); 243 pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 244 if (pr == 0) 245 panic("ip_init"); 246 for (i = 0; i < IPPROTO_MAX; i++) 247 ip_protox[i] = pr - inetsw; 248 for (pr = (struct ipprotosw *)inetdomain.dom_protosw; 249 pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++) 250 if (pr->pr_domain->dom_family == PF_INET && 251 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 252 ip_protox[pr->pr_protocol] = pr - inetsw; 253 254 for (i = 0; i < IPREASS_NHASH; i++) 255 TAILQ_INIT(&ipq[i]); 256 257 maxnipq = nmbclusters/4; 258 259#ifndef RANDOM_IP_ID 260 ip_id = time_second & 0xffff; 261#endif 262 ipintrq.ifq_maxlen = ipqmaxlen; 263 mtx_init(&ipintrq.ifq_mtx, "ip_inq", MTX_DEF); 264 265 register_netisr(NETISR_IP, ipintr); 266} 267 268static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 269struct route ipforward_rt; 270 271/* 272 * Ip input routine. Checksum and byte swap header. If fragmented 273 * try to reassemble. Process options. Pass to next level. 274 */ 275void 276ip_input(struct mbuf *m) 277{ 278 struct ip *ip; 279 struct ipq *fp; 280 struct in_ifaddr *ia = NULL; 281 int i, hlen, checkif; 282 u_short sum; 283 u_int16_t divert_cookie; /* firewall cookie */ 284 struct in_addr pkt_dst; 285#ifdef IPDIVERT 286 u_int32_t divert_info = 0; /* packet divert/tee info */ 287#endif 288 struct ip_fw_chain *rule = NULL; 289#ifdef PFIL_HOOKS 290 struct packet_filter_hook *pfh; 291 struct mbuf *m0; 292 int rv; 293#endif /* PFIL_HOOKS */ 294 295#ifdef IPDIVERT 296 /* Get and reset firewall cookie */ 297 divert_cookie = ip_divert_cookie; 298 ip_divert_cookie = 0; 299#else 300 divert_cookie = 0; 301#endif 302 303#if defined(IPFIREWALL) && defined(DUMMYNET) 304 /* 305 * dummynet packet are prepended a vestigial mbuf with 306 * m_type = MT_DUMMYNET and m_data pointing to the matching 307 * rule. 308 */ 309 if (m->m_type == MT_DUMMYNET) { 310 rule = (struct ip_fw_chain *)(m->m_data) ; 311 m = m->m_next ; 312 ip = mtod(m, struct ip *); 313 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 314 goto iphack ; 315 } else 316 rule = NULL ; 317#endif 318 319#ifdef DIAGNOSTIC 320 if (m == NULL || (m->m_flags & M_PKTHDR) == 0) 321 panic("ip_input no HDR"); 322#endif 323 ipstat.ips_total++; 324 325 if (m->m_pkthdr.len < sizeof(struct ip)) 326 goto tooshort; 327 328 if (m->m_len < sizeof (struct ip) && 329 (m = m_pullup(m, sizeof (struct ip))) == 0) { 330 ipstat.ips_toosmall++; 331 return; 332 } 333 ip = mtod(m, struct ip *); 334 335 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 336 ipstat.ips_badvers++; 337 goto bad; 338 } 339 340 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 341 if (hlen < sizeof(struct ip)) { /* minimum header length */ 342 ipstat.ips_badhlen++; 343 goto bad; 344 } 345 if (hlen > m->m_len) { 346 if ((m = m_pullup(m, hlen)) == 0) { 347 ipstat.ips_badhlen++; 348 return; 349 } 350 ip = mtod(m, struct ip *); 351 } 352 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 353 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 354 } else { 355 if (hlen == sizeof(struct ip)) { 356 sum = in_cksum_hdr(ip); 357 } else { 358 sum = in_cksum(m, hlen); 359 } 360 } 361 if (sum) { 362 ipstat.ips_badsum++; 363 goto bad; 364 } 365 366 /* 367 * Convert fields to host representation. 368 */ 369 NTOHS(ip->ip_len); 370 if (ip->ip_len < hlen) { 371 ipstat.ips_badlen++; 372 goto bad; 373 } 374 NTOHS(ip->ip_off); 375 376 /* 377 * Check that the amount of data in the buffers 378 * is as at least much as the IP header would have us expect. 379 * Trim mbufs if longer than we expect. 380 * Drop packet if shorter than we expect. 381 */ 382 if (m->m_pkthdr.len < ip->ip_len) { 383tooshort: 384 ipstat.ips_tooshort++; 385 goto bad; 386 } 387 if (m->m_pkthdr.len > ip->ip_len) { 388 if (m->m_len == m->m_pkthdr.len) { 389 m->m_len = ip->ip_len; 390 m->m_pkthdr.len = ip->ip_len; 391 } else 392 m_adj(m, ip->ip_len - m->m_pkthdr.len); 393 } 394 395 /* 396 * Don't accept packets with a loopback destination address 397 * unless they arrived via the loopback interface. 398 */ 399 if ((ntohl(ip->ip_dst.s_addr) & IN_CLASSA_NET) == 400 (IN_LOOPBACKNET << IN_CLASSA_NSHIFT) && 401 (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 402 goto bad; 403 } 404 405 /* 406 * IpHack's section. 407 * Right now when no processing on packet has done 408 * and it is still fresh out of network we do our black 409 * deals with it. 410 * - Firewall: deny/allow/divert 411 * - Xlate: translate packet's addr/port (NAT). 412 * - Pipe: pass pkt through dummynet. 413 * - Wrap: fake packet's addr/port <unimpl.> 414 * - Encapsulate: put it in another IP and send out. <unimp.> 415 */ 416 417#if defined(IPFIREWALL) && defined(DUMMYNET) 418iphack: 419#endif 420 421#ifdef PFIL_HOOKS 422 /* 423 * Run through list of hooks for input packets. If there are any 424 * filters which require that additional packets in the flow are 425 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 426 * Note that filters must _never_ set this flag, as another filter 427 * in the list may have previously cleared it. 428 */ 429 m0 = m; 430 pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 431 for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link)) 432 if (pfh->pfil_func) { 433 rv = pfh->pfil_func(ip, hlen, 434 m->m_pkthdr.rcvif, 0, &m0); 435 if (rv) 436 return; 437 m = m0; 438 if (m == NULL) 439 return; 440 ip = mtod(m, struct ip *); 441 } 442#endif /* PFIL_HOOKS */ 443 444 if (fw_enable && ip_fw_chk_ptr) { 445#ifdef IPFIREWALL_FORWARD 446 /* 447 * If we've been forwarded from the output side, then 448 * skip the firewall a second time 449 */ 450 if (ip_fw_fwd_addr) 451 goto ours; 452#endif /* IPFIREWALL_FORWARD */ 453 /* 454 * See the comment in ip_output for the return values 455 * produced by the firewall. 456 */ 457 i = (*ip_fw_chk_ptr)(&ip, 458 hlen, NULL, &divert_cookie, &m, &rule, &ip_fw_fwd_addr); 459 if (i & IP_FW_PORT_DENY_FLAG) { /* XXX new interface-denied */ 460 if (m) 461 m_freem(m); 462 return ; 463 } 464 if (m == NULL) { /* Packet discarded by firewall */ 465 static int __debug=10; 466 if (__debug >0) { 467 printf("firewall returns NULL, please update!\n"); 468 __debug-- ; 469 } 470 return; 471 } 472 if (i == 0 && ip_fw_fwd_addr == NULL) /* common case */ 473 goto pass; 474#ifdef DUMMYNET 475 if ((i & IP_FW_PORT_DYNT_FLAG) != 0) { 476 /* Send packet to the appropriate pipe */ 477 dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule, 478 0); 479 return; 480 } 481#endif 482#ifdef IPDIVERT 483 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 484 /* Divert or tee packet */ 485 divert_info = i; 486 goto ours; 487 } 488#endif 489#ifdef IPFIREWALL_FORWARD 490 if (i == 0 && ip_fw_fwd_addr != NULL) 491 goto pass; 492#endif 493 /* 494 * if we get here, the packet must be dropped 495 */ 496 m_freem(m); 497 return; 498 } 499pass: 500 501 /* 502 * Process options and, if not destined for us, 503 * ship it on. ip_dooptions returns 1 when an 504 * error was detected (causing an icmp message 505 * to be sent and the original packet to be freed). 506 */ 507 ip_nhops = 0; /* for source routed packets */ 508 if (hlen > sizeof (struct ip) && ip_dooptions(m)) { 509#ifdef IPFIREWALL_FORWARD 510 ip_fw_fwd_addr = NULL; 511#endif 512 return; 513 } 514 515 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 516 * matter if it is destined to another node, or whether it is 517 * a multicast one, RSVP wants it! and prevents it from being forwarded 518 * anywhere else. Also checks if the rsvp daemon is running before 519 * grabbing the packet. 520 */ 521 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 522 goto ours; 523 524 /* 525 * Check our list of addresses, to see if the packet is for us. 526 * If we don't have any addresses, assume any unicast packet 527 * we receive might be for us (and let the upper layers deal 528 * with it). 529 */ 530 if (TAILQ_EMPTY(&in_ifaddrhead) && 531 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 532 goto ours; 533 534 /* 535 * Cache the destination address of the packet; this may be 536 * changed by use of 'ipfw fwd'. 537 */ 538 pkt_dst = ip_fw_fwd_addr == NULL ? 539 ip->ip_dst : ip_fw_fwd_addr->sin_addr; 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->if_flags & IFF_LOOPBACK) == 0) && 557 (ip_fw_fwd_addr == NULL); 558 559 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) { 560#define satosin(sa) ((struct sockaddr_in *)(sa)) 561 562#ifdef BOOTP_COMPAT 563 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 564 goto ours; 565#endif 566 /* 567 * If the address matches, verify that the packet 568 * arrived via the correct interface if checking is 569 * enabled. 570 */ 571 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 572 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 573 goto ours; 574 /* 575 * Only accept broadcast packets that arrive via the 576 * matching interface. Reception of forwarded directed 577 * broadcasts would be handled via ip_forward() and 578 * ether_output() with the loopback into the stack for 579 * SIMPLEX interfaces handled by ether_output(). 580 */ 581 if (ia->ia_ifp == m->m_pkthdr.rcvif && 582 ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) { 583 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 584 pkt_dst.s_addr) 585 goto ours; 586 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 587 goto ours; 588 } 589 } 590 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 591 struct in_multi *inm; 592 if (ip_mrouter) { 593 /* 594 * If we are acting as a multicast router, all 595 * incoming multicast packets are passed to the 596 * kernel-level multicast forwarding function. 597 * The packet is returned (relatively) intact; if 598 * ip_mforward() returns a non-zero value, the packet 599 * must be discarded, else it may be accepted below. 600 */ 601 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 602 ipstat.ips_cantforward++; 603 m_freem(m); 604 return; 605 } 606 607 /* 608 * The process-level routing demon needs to receive 609 * all multicast IGMP packets, whether or not this 610 * host belongs to their destination groups. 611 */ 612 if (ip->ip_p == IPPROTO_IGMP) 613 goto ours; 614 ipstat.ips_forward++; 615 } 616 /* 617 * See if we belong to the destination multicast group on the 618 * arrival interface. 619 */ 620 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 621 if (inm == NULL) { 622 ipstat.ips_notmember++; 623 m_freem(m); 624 return; 625 } 626 goto ours; 627 } 628 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 629 goto ours; 630 if (ip->ip_dst.s_addr == INADDR_ANY) 631 goto ours; 632 633#if defined(NFAITH) && 0 < NFAITH 634 /* 635 * FAITH(Firewall Aided Internet Translator) 636 */ 637 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 638 if (ip_keepfaith) { 639 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 640 goto ours; 641 } 642 m_freem(m); 643 return; 644 } 645#endif 646 /* 647 * Not for us; forward if possible and desirable. 648 */ 649 if (ipforwarding == 0) { 650 ipstat.ips_cantforward++; 651 m_freem(m); 652 } else 653 ip_forward(m, 0); 654#ifdef IPFIREWALL_FORWARD 655 ip_fw_fwd_addr = NULL; 656#endif 657 return; 658 659ours: 660 /* Count the packet in the ip address stats */ 661 if (ia != NULL) { 662 ia->ia_ifa.if_ipackets++; 663 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 664 } 665 666 /* 667 * If offset or IP_MF are set, must reassemble. 668 * Otherwise, nothing need be done. 669 * (We could look in the reassembly queue to see 670 * if the packet was previously fragmented, 671 * but it's not worth the time; just let them time out.) 672 */ 673 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 674 675 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 676 /* 677 * Look for queue of fragments 678 * of this datagram. 679 */ 680 TAILQ_FOREACH(fp, &ipq[sum], ipq_list) 681 if (ip->ip_id == fp->ipq_id && 682 ip->ip_src.s_addr == fp->ipq_src.s_addr && 683 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 684 ip->ip_p == fp->ipq_p) 685 goto found; 686 687 fp = 0; 688 689 /* check if there's a place for the new queue */ 690 if (nipq > maxnipq) { 691 /* 692 * drop something from the tail of the current queue 693 * before proceeding further 694 */ 695 struct ipq *q = TAILQ_LAST(&ipq[sum], ipqhead); 696 if (q == NULL) { /* gak */ 697 for (i = 0; i < IPREASS_NHASH; i++) { 698 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 699 if (r) { 700 ip_freef(&ipq[i], r); 701 break; 702 } 703 } 704 } else 705 ip_freef(&ipq[sum], q); 706 } 707found: 708 /* 709 * Adjust ip_len to not reflect header, 710 * convert offset of this to bytes. 711 */ 712 ip->ip_len -= hlen; 713 if (ip->ip_off & IP_MF) { 714 /* 715 * Make sure that fragments have a data length 716 * that's a non-zero multiple of 8 bytes. 717 */ 718 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 719 ipstat.ips_toosmall++; /* XXX */ 720 goto bad; 721 } 722 m->m_flags |= M_FRAG; 723 } 724 ip->ip_off <<= 3; 725 726 /* 727 * Attempt reassembly; if it succeeds, proceed. 728 */ 729 ipstat.ips_fragments++; 730 m->m_pkthdr.header = ip; 731#ifdef IPDIVERT 732 m = ip_reass(m, 733 &ipq[sum], fp, &divert_info, &divert_cookie); 734#else 735 m = ip_reass(m, &ipq[sum], fp); 736#endif 737 if (m == 0) { 738#ifdef IPFIREWALL_FORWARD 739 ip_fw_fwd_addr = NULL; 740#endif 741 return; 742 } 743 ipstat.ips_reassembled++; 744 ip = mtod(m, struct ip *); 745 /* Get the header length of the reassembled packet */ 746 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 747#ifdef IPDIVERT 748 /* Restore original checksum before diverting packet */ 749 if (divert_info != 0) { 750 ip->ip_len += hlen; 751 HTONS(ip->ip_len); 752 HTONS(ip->ip_off); 753 ip->ip_sum = 0; 754 if (hlen == sizeof(struct ip)) 755 ip->ip_sum = in_cksum_hdr(ip); 756 else 757 ip->ip_sum = in_cksum(m, hlen); 758 NTOHS(ip->ip_off); 759 NTOHS(ip->ip_len); 760 ip->ip_len -= hlen; 761 } 762#endif 763 } else 764 ip->ip_len -= hlen; 765 766#ifdef IPDIVERT 767 /* 768 * Divert or tee packet to the divert protocol if required. 769 * 770 * If divert_info is zero then cookie should be too, so we shouldn't 771 * need to clear them here. Assume divert_packet() does so also. 772 */ 773 if (divert_info != 0) { 774 struct mbuf *clone = NULL; 775 776 /* Clone packet if we're doing a 'tee' */ 777 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 778 clone = m_dup(m, M_DONTWAIT); 779 780 /* Restore packet header fields to original values */ 781 ip->ip_len += hlen; 782 HTONS(ip->ip_len); 783 HTONS(ip->ip_off); 784 785 /* Deliver packet to divert input routine */ 786 ip_divert_cookie = divert_cookie; 787 divert_packet(m, 1, divert_info & 0xffff); 788 ipstat.ips_delivered++; 789 790 /* If 'tee', continue with original packet */ 791 if (clone == NULL) 792 return; 793 m = clone; 794 ip = mtod(m, struct ip *); 795 } 796#endif 797 798 /* 799 * Switch out to protocol's input routine. 800 */ 801 ipstat.ips_delivered++; 802 { 803 int off = hlen, nh = ip->ip_p; 804 805 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off, nh); 806#ifdef IPFIREWALL_FORWARD 807 ip_fw_fwd_addr = NULL; /* tcp needed it */ 808#endif 809 return; 810 } 811bad: 812#ifdef IPFIREWALL_FORWARD 813 ip_fw_fwd_addr = NULL; 814#endif 815 m_freem(m); 816} 817 818/* 819 * IP software interrupt routine - to go away sometime soon 820 */ 821static void 822ipintr(void) 823{ 824 struct mbuf *m; 825 826 while (1) { 827 IF_DEQUEUE(&ipintrq, m); 828 if (m == 0) 829 return; 830 ip_input(m); 831 } 832} 833 834/* 835 * Take incoming datagram fragment and try to reassemble it into 836 * whole datagram. If a chain for reassembly of this datagram already 837 * exists, then it is given as fp; otherwise have to make a chain. 838 * 839 * When IPDIVERT enabled, keep additional state with each packet that 840 * tells us if we need to divert or tee the packet we're building. 841 */ 842 843static struct mbuf * 844#ifdef IPDIVERT 845ip_reass(m, head, fp, divinfo, divcookie) 846#else 847ip_reass(m, head, fp) 848#endif 849 struct mbuf *m; 850 struct ipqhead *head; 851 struct ipq *fp; 852#ifdef IPDIVERT 853 u_int32_t *divinfo; 854 u_int16_t *divcookie; 855#endif 856{ 857 struct ip *ip = mtod(m, struct ip *); 858 register struct mbuf *p, *q, *nq; 859 struct mbuf *t; 860 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 861 int i, next; 862 863 /* 864 * Presence of header sizes in mbufs 865 * would confuse code below. 866 */ 867 m->m_data += hlen; 868 m->m_len -= hlen; 869 870 /* 871 * If first fragment to arrive, create a reassembly queue. 872 */ 873 if (fp == 0) { 874 /* 875 * Enforce upper bound on number of fragmented packets 876 * for which we attempt reassembly; 877 * If maxfrag is 0, never accept fragments. 878 * If maxfrag is -1, accept all fragments without limitation. 879 */ 880 if ((ip_maxfragpackets >= 0) && (ip_nfragpackets >= ip_maxfragpackets)) 881 goto dropfrag; 882 ip_nfragpackets++; 883 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 884 goto dropfrag; 885 fp = mtod(t, struct ipq *); 886 TAILQ_INSERT_HEAD(head, fp, ipq_list); 887 nipq++; 888 fp->ipq_ttl = IPFRAGTTL; 889 fp->ipq_p = ip->ip_p; 890 fp->ipq_id = ip->ip_id; 891 fp->ipq_src = ip->ip_src; 892 fp->ipq_dst = ip->ip_dst; 893 fp->ipq_frags = m; 894 m->m_nextpkt = NULL; 895#ifdef IPDIVERT 896 fp->ipq_div_info = 0; 897 fp->ipq_div_cookie = 0; 898#endif 899 goto inserted; 900 } 901 902#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 903 904 /* 905 * Find a segment which begins after this one does. 906 */ 907 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 908 if (GETIP(q)->ip_off > ip->ip_off) 909 break; 910 911 /* 912 * If there is a preceding segment, it may provide some of 913 * our data already. If so, drop the data from the incoming 914 * segment. If it provides all of our data, drop us, otherwise 915 * stick new segment in the proper place. 916 * 917 * If some of the data is dropped from the the preceding 918 * segment, then it's checksum is invalidated. 919 */ 920 if (p) { 921 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 922 if (i > 0) { 923 if (i >= ip->ip_len) 924 goto dropfrag; 925 m_adj(m, i); 926 m->m_pkthdr.csum_flags = 0; 927 ip->ip_off += i; 928 ip->ip_len -= i; 929 } 930 m->m_nextpkt = p->m_nextpkt; 931 p->m_nextpkt = m; 932 } else { 933 m->m_nextpkt = fp->ipq_frags; 934 fp->ipq_frags = m; 935 } 936 937 /* 938 * While we overlap succeeding segments trim them or, 939 * if they are completely covered, dequeue them. 940 */ 941 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 942 q = nq) { 943 i = (ip->ip_off + ip->ip_len) - 944 GETIP(q)->ip_off; 945 if (i < GETIP(q)->ip_len) { 946 GETIP(q)->ip_len -= i; 947 GETIP(q)->ip_off += i; 948 m_adj(q, i); 949 q->m_pkthdr.csum_flags = 0; 950 break; 951 } 952 nq = q->m_nextpkt; 953 m->m_nextpkt = nq; 954 m_freem(q); 955 } 956 957inserted: 958 959#ifdef IPDIVERT 960 /* 961 * Transfer firewall instructions to the fragment structure. 962 * Any fragment diverting causes the whole packet to divert. 963 */ 964 fp->ipq_div_info = *divinfo; 965 fp->ipq_div_cookie = *divcookie; 966 *divinfo = 0; 967 *divcookie = 0; 968#endif 969 970 /* 971 * Check for complete reassembly. 972 */ 973 next = 0; 974 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 975 if (GETIP(q)->ip_off != next) 976 return (0); 977 next += GETIP(q)->ip_len; 978 } 979 /* Make sure the last packet didn't have the IP_MF flag */ 980 if (p->m_flags & M_FRAG) 981 return (0); 982 983 /* 984 * Reassembly is complete. Make sure the packet is a sane size. 985 */ 986 q = fp->ipq_frags; 987 ip = GETIP(q); 988 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 989 ipstat.ips_toolong++; 990 ip_freef(head, fp); 991 return (0); 992 } 993 994 /* 995 * Concatenate fragments. 996 */ 997 m = q; 998 t = m->m_next; 999 m->m_next = 0; 1000 m_cat(m, t); 1001 nq = q->m_nextpkt; 1002 q->m_nextpkt = 0; 1003 for (q = nq; q != NULL; q = nq) { 1004 nq = q->m_nextpkt; 1005 q->m_nextpkt = NULL; 1006 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1007 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1008 m_cat(m, q); 1009 } 1010 1011#ifdef IPDIVERT 1012 /* 1013 * Extract firewall instructions from the fragment structure. 1014 */ 1015 *divinfo = fp->ipq_div_info; 1016 *divcookie = fp->ipq_div_cookie; 1017#endif 1018 1019 /* 1020 * Create header for new ip packet by 1021 * modifying header of first packet; 1022 * dequeue and discard fragment reassembly header. 1023 * Make header visible. 1024 */ 1025 ip->ip_len = next; 1026 ip->ip_src = fp->ipq_src; 1027 ip->ip_dst = fp->ipq_dst; 1028 TAILQ_REMOVE(head, fp, ipq_list); 1029 nipq--; 1030 (void) m_free(dtom(fp)); 1031 ip_nfragpackets--; 1032 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 1033 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 1034 /* some debugging cruft by sklower, below, will go away soon */ 1035 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1036 register int plen = 0; 1037 for (t = m; t; t = t->m_next) 1038 plen += t->m_len; 1039 m->m_pkthdr.len = plen; 1040 } 1041 return (m); 1042 1043dropfrag: 1044#ifdef IPDIVERT 1045 *divinfo = 0; 1046 *divcookie = 0; 1047#endif 1048 ipstat.ips_fragdropped++; 1049 m_freem(m); 1050 return (0); 1051 1052#undef GETIP 1053} 1054 1055/* 1056 * Free a fragment reassembly header and all 1057 * associated datagrams. 1058 */ 1059static void 1060ip_freef(fhp, fp) 1061 struct ipqhead *fhp; 1062 struct ipq *fp; 1063{ 1064 register struct mbuf *q; 1065 1066 while (fp->ipq_frags) { 1067 q = fp->ipq_frags; 1068 fp->ipq_frags = q->m_nextpkt; 1069 m_freem(q); 1070 } 1071 TAILQ_REMOVE(fhp, fp, ipq_list); 1072 (void) m_free(dtom(fp)); 1073 ip_nfragpackets--; 1074 nipq--; 1075} 1076 1077/* 1078 * IP timer processing; 1079 * if a timer expires on a reassembly 1080 * queue, discard it. 1081 */ 1082void 1083ip_slowtimo() 1084{ 1085 register struct ipq *fp; 1086 int s = splnet(); 1087 int i; 1088 1089 for (i = 0; i < IPREASS_NHASH; i++) { 1090 for(fp = TAILQ_FIRST(&ipq[i]); fp;) { 1091 struct ipq *fpp; 1092 1093 fpp = fp; 1094 fp = TAILQ_NEXT(fp, ipq_list); 1095 if(--fpp->ipq_ttl == 0) { 1096 ipstat.ips_fragtimeout++; 1097 ip_freef(&ipq[i], fpp); 1098 } 1099 } 1100 } 1101 /* 1102 * If we are over the maximum number of fragments 1103 * (due to the limit being lowered), drain off 1104 * enough to get down to the new limit. 1105 */ 1106 for (i = 0; i < IPREASS_NHASH; i++) { 1107 if (ip_maxfragpackets >= 0) { 1108 while (ip_nfragpackets > ip_maxfragpackets && 1109 !TAILQ_EMPTY(&ipq[i])) { 1110 ipstat.ips_fragdropped++; 1111 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1112 } 1113 } 1114 } 1115 ipflow_slowtimo(); 1116 splx(s); 1117} 1118 1119/* 1120 * Drain off all datagram fragments. 1121 */ 1122void 1123ip_drain() 1124{ 1125 int i; 1126 1127 for (i = 0; i < IPREASS_NHASH; i++) { 1128 while(!TAILQ_EMPTY(&ipq[i])) { 1129 ipstat.ips_fragdropped++; 1130 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1131 } 1132 } 1133 in_rtqdrain(); 1134} 1135 1136/* 1137 * Do option processing on a datagram, 1138 * possibly discarding it if bad options are encountered, 1139 * or forwarding it if source-routed. 1140 * Returns 1 if packet has been forwarded/freed, 1141 * 0 if the packet should be processed further. 1142 */ 1143static int 1144ip_dooptions(m) 1145 struct mbuf *m; 1146{ 1147 register struct ip *ip = mtod(m, struct ip *); 1148 register u_char *cp; 1149 register struct ip_timestamp *ipt; 1150 register struct in_ifaddr *ia; 1151 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1152 struct in_addr *sin, dst; 1153 n_time ntime; 1154 1155 dst = ip->ip_dst; 1156 cp = (u_char *)(ip + 1); 1157 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1158 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1159 opt = cp[IPOPT_OPTVAL]; 1160 if (opt == IPOPT_EOL) 1161 break; 1162 if (opt == IPOPT_NOP) 1163 optlen = 1; 1164 else { 1165 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1166 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1167 goto bad; 1168 } 1169 optlen = cp[IPOPT_OLEN]; 1170 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1171 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1172 goto bad; 1173 } 1174 } 1175 switch (opt) { 1176 1177 default: 1178 break; 1179 1180 /* 1181 * Source routing with record. 1182 * Find interface with current destination address. 1183 * If none on this machine then drop if strictly routed, 1184 * or do nothing if loosely routed. 1185 * Record interface address and bring up next address 1186 * component. If strictly routed make sure next 1187 * address is on directly accessible net. 1188 */ 1189 case IPOPT_LSRR: 1190 case IPOPT_SSRR: 1191 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1192 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1193 goto bad; 1194 } 1195 ipaddr.sin_addr = ip->ip_dst; 1196 ia = (struct in_ifaddr *) 1197 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1198 if (ia == 0) { 1199 if (opt == IPOPT_SSRR) { 1200 type = ICMP_UNREACH; 1201 code = ICMP_UNREACH_SRCFAIL; 1202 goto bad; 1203 } 1204 if (!ip_dosourceroute) 1205 goto nosourcerouting; 1206 /* 1207 * Loose routing, and not at next destination 1208 * yet; nothing to do except forward. 1209 */ 1210 break; 1211 } 1212 off--; /* 0 origin */ 1213 if (off > optlen - (int)sizeof(struct in_addr)) { 1214 /* 1215 * End of source route. Should be for us. 1216 */ 1217 if (!ip_acceptsourceroute) 1218 goto nosourcerouting; 1219 save_rte(cp, ip->ip_src); 1220 break; 1221 } 1222 1223 if (!ip_dosourceroute) { 1224 if (ipforwarding) { 1225 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1226 /* 1227 * Acting as a router, so generate ICMP 1228 */ 1229nosourcerouting: 1230 strcpy(buf, inet_ntoa(ip->ip_dst)); 1231 log(LOG_WARNING, 1232 "attempted source route from %s to %s\n", 1233 inet_ntoa(ip->ip_src), buf); 1234 type = ICMP_UNREACH; 1235 code = ICMP_UNREACH_SRCFAIL; 1236 goto bad; 1237 } else { 1238 /* 1239 * Not acting as a router, so silently drop. 1240 */ 1241 ipstat.ips_cantforward++; 1242 m_freem(m); 1243 return (1); 1244 } 1245 } 1246 1247 /* 1248 * locate outgoing interface 1249 */ 1250 (void)memcpy(&ipaddr.sin_addr, cp + off, 1251 sizeof(ipaddr.sin_addr)); 1252 1253 if (opt == IPOPT_SSRR) { 1254#define INA struct in_ifaddr * 1255#define SA struct sockaddr * 1256 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1257 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1258 } else 1259 ia = ip_rtaddr(ipaddr.sin_addr); 1260 if (ia == 0) { 1261 type = ICMP_UNREACH; 1262 code = ICMP_UNREACH_SRCFAIL; 1263 goto bad; 1264 } 1265 ip->ip_dst = ipaddr.sin_addr; 1266 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1267 sizeof(struct in_addr)); 1268 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1269 /* 1270 * Let ip_intr's mcast routing check handle mcast pkts 1271 */ 1272 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1273 break; 1274 1275 case IPOPT_RR: 1276 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1277 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1278 goto bad; 1279 } 1280 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1281 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1282 goto bad; 1283 } 1284 /* 1285 * If no space remains, ignore. 1286 */ 1287 off--; /* 0 origin */ 1288 if (off > optlen - (int)sizeof(struct in_addr)) 1289 break; 1290 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1291 sizeof(ipaddr.sin_addr)); 1292 /* 1293 * locate outgoing interface; if we're the destination, 1294 * use the incoming interface (should be same). 1295 */ 1296 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1297 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1298 type = ICMP_UNREACH; 1299 code = ICMP_UNREACH_HOST; 1300 goto bad; 1301 } 1302 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1303 sizeof(struct in_addr)); 1304 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1305 break; 1306 1307 case IPOPT_TS: 1308 code = cp - (u_char *)ip; 1309 ipt = (struct ip_timestamp *)cp; 1310 if (ipt->ipt_len < 5) 1311 goto bad; 1312 if (ipt->ipt_ptr > 1313 ipt->ipt_len - (int)sizeof(int32_t)) { 1314 if (++ipt->ipt_oflw == 0) 1315 goto bad; 1316 break; 1317 } 1318 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1319 switch (ipt->ipt_flg) { 1320 1321 case IPOPT_TS_TSONLY: 1322 break; 1323 1324 case IPOPT_TS_TSANDADDR: 1325 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1326 sizeof(struct in_addr) > ipt->ipt_len) 1327 goto bad; 1328 ipaddr.sin_addr = dst; 1329 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1330 m->m_pkthdr.rcvif); 1331 if (ia == 0) 1332 continue; 1333 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1334 sizeof(struct in_addr)); 1335 ipt->ipt_ptr += sizeof(struct in_addr); 1336 break; 1337 1338 case IPOPT_TS_PRESPEC: 1339 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1340 sizeof(struct in_addr) > ipt->ipt_len) 1341 goto bad; 1342 (void)memcpy(&ipaddr.sin_addr, sin, 1343 sizeof(struct in_addr)); 1344 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1345 continue; 1346 ipt->ipt_ptr += sizeof(struct in_addr); 1347 break; 1348 1349 default: 1350 goto bad; 1351 } 1352 ntime = iptime(); 1353 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 1354 sizeof(n_time)); 1355 ipt->ipt_ptr += sizeof(n_time); 1356 } 1357 } 1358 if (forward && ipforwarding) { 1359 ip_forward(m, 1); 1360 return (1); 1361 } 1362 return (0); 1363bad: 1364 icmp_error(m, type, code, 0, 0); 1365 ipstat.ips_badoptions++; 1366 return (1); 1367} 1368 1369/* 1370 * Given address of next destination (final or next hop), 1371 * return internet address info of interface to be used to get there. 1372 */ 1373static struct in_ifaddr * 1374ip_rtaddr(dst) 1375 struct in_addr dst; 1376{ 1377 register struct sockaddr_in *sin; 1378 1379 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 1380 1381 if (ipforward_rt.ro_rt == 0 || 1382 !(ipforward_rt.ro_rt->rt_flags & RTF_UP) || 1383 dst.s_addr != sin->sin_addr.s_addr) { 1384 if (ipforward_rt.ro_rt) { 1385 RTFREE(ipforward_rt.ro_rt); 1386 ipforward_rt.ro_rt = 0; 1387 } 1388 sin->sin_family = AF_INET; 1389 sin->sin_len = sizeof(*sin); 1390 sin->sin_addr = dst; 1391 1392 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1393 } 1394 if (ipforward_rt.ro_rt == 0) 1395 return ((struct in_ifaddr *)0); 1396 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 1397} 1398 1399/* 1400 * Save incoming source route for use in replies, 1401 * to be picked up later by ip_srcroute if the receiver is interested. 1402 */ 1403void 1404save_rte(option, dst) 1405 u_char *option; 1406 struct in_addr dst; 1407{ 1408 unsigned olen; 1409 1410 olen = option[IPOPT_OLEN]; 1411#ifdef DIAGNOSTIC 1412 if (ipprintfs) 1413 printf("save_rte: olen %d\n", olen); 1414#endif 1415 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1416 return; 1417 bcopy(option, ip_srcrt.srcopt, olen); 1418 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1419 ip_srcrt.dst = dst; 1420} 1421 1422/* 1423 * Retrieve incoming source route for use in replies, 1424 * in the same form used by setsockopt. 1425 * The first hop is placed before the options, will be removed later. 1426 */ 1427struct mbuf * 1428ip_srcroute() 1429{ 1430 register struct in_addr *p, *q; 1431 register struct mbuf *m; 1432 1433 if (ip_nhops == 0) 1434 return ((struct mbuf *)0); 1435 m = m_get(M_DONTWAIT, MT_HEADER); 1436 if (m == 0) 1437 return ((struct mbuf *)0); 1438 1439#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1440 1441 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1442 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1443 OPTSIZ; 1444#ifdef DIAGNOSTIC 1445 if (ipprintfs) 1446 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1447#endif 1448 1449 /* 1450 * First save first hop for return route 1451 */ 1452 p = &ip_srcrt.route[ip_nhops - 1]; 1453 *(mtod(m, struct in_addr *)) = *p--; 1454#ifdef DIAGNOSTIC 1455 if (ipprintfs) 1456 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1457#endif 1458 1459 /* 1460 * Copy option fields and padding (nop) to mbuf. 1461 */ 1462 ip_srcrt.nop = IPOPT_NOP; 1463 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1464 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1465 &ip_srcrt.nop, OPTSIZ); 1466 q = (struct in_addr *)(mtod(m, caddr_t) + 1467 sizeof(struct in_addr) + OPTSIZ); 1468#undef OPTSIZ 1469 /* 1470 * Record return path as an IP source route, 1471 * reversing the path (pointers are now aligned). 1472 */ 1473 while (p >= ip_srcrt.route) { 1474#ifdef DIAGNOSTIC 1475 if (ipprintfs) 1476 printf(" %lx", (u_long)ntohl(q->s_addr)); 1477#endif 1478 *q++ = *p--; 1479 } 1480 /* 1481 * Last hop goes to final destination. 1482 */ 1483 *q = ip_srcrt.dst; 1484#ifdef DIAGNOSTIC 1485 if (ipprintfs) 1486 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1487#endif 1488 return (m); 1489} 1490 1491/* 1492 * Strip out IP options, at higher 1493 * level protocol in the kernel. 1494 * Second argument is buffer to which options 1495 * will be moved, and return value is their length. 1496 * XXX should be deleted; last arg currently ignored. 1497 */ 1498void 1499ip_stripoptions(m, mopt) 1500 register struct mbuf *m; 1501 struct mbuf *mopt; 1502{ 1503 register int i; 1504 struct ip *ip = mtod(m, struct ip *); 1505 register caddr_t opts; 1506 int olen; 1507 1508 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1509 opts = (caddr_t)(ip + 1); 1510 i = m->m_len - (sizeof (struct ip) + olen); 1511 bcopy(opts + olen, opts, (unsigned)i); 1512 m->m_len -= olen; 1513 if (m->m_flags & M_PKTHDR) 1514 m->m_pkthdr.len -= olen; 1515 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1516} 1517 1518u_char inetctlerrmap[PRC_NCMDS] = { 1519 0, 0, 0, 0, 1520 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1521 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1522 EMSGSIZE, EHOSTUNREACH, 0, 0, 1523 0, 0, 0, 0, 1524 ENOPROTOOPT, ENETRESET 1525}; 1526 1527/* 1528 * Forward a packet. If some error occurs return the sender 1529 * an icmp packet. Note we can't always generate a meaningful 1530 * icmp message because icmp doesn't have a large enough repertoire 1531 * of codes and types. 1532 * 1533 * If not forwarding, just drop the packet. This could be confusing 1534 * if ipforwarding was zero but some routing protocol was advancing 1535 * us as a gateway to somewhere. However, we must let the routing 1536 * protocol deal with that. 1537 * 1538 * The srcrt parameter indicates whether the packet is being forwarded 1539 * via a source route. 1540 */ 1541static void 1542ip_forward(m, srcrt) 1543 struct mbuf *m; 1544 int srcrt; 1545{ 1546 register struct ip *ip = mtod(m, struct ip *); 1547 register struct rtentry *rt; 1548 int error, type = 0, code = 0; 1549 struct mbuf *mcopy; 1550 n_long dest; 1551 struct ifnet *destifp; 1552#ifdef IPSEC 1553 struct ifnet dummyifp; 1554#endif 1555 1556 dest = 0; 1557#ifdef DIAGNOSTIC 1558 if (ipprintfs) 1559 printf("forward: src %lx dst %lx ttl %x\n", 1560 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr, 1561 ip->ip_ttl); 1562#endif 1563 1564 1565 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1566 ipstat.ips_cantforward++; 1567 m_freem(m); 1568 return; 1569 } 1570#ifdef IPSTEALTH 1571 if (!ipstealth) { 1572#endif 1573 if (ip->ip_ttl <= IPTTLDEC) { 1574 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1575 dest, 0); 1576 return; 1577 } 1578#ifdef IPSTEALTH 1579 } 1580#endif 1581 1582 if (ip_rtaddr(ip->ip_dst) == 0) { 1583 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1584 return; 1585 } else 1586 rt = ipforward_rt.ro_rt; 1587 1588 /* 1589 * Save the IP header and at most 8 bytes of the payload, 1590 * in case we need to generate an ICMP message to the src. 1591 * 1592 * We don't use m_copy() because it might return a reference 1593 * to a shared cluster. Both this function and ip_output() 1594 * assume exclusive access to the IP header in `m', so any 1595 * data in a cluster may change before we reach icmp_error(). 1596 */ 1597 MGET(mcopy, M_DONTWAIT, m->m_type); 1598 if (mcopy != NULL) { 1599 M_COPY_PKTHDR(mcopy, m); 1600 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8, 1601 (int)ip->ip_len); 1602 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1603 } 1604 1605#ifdef IPSTEALTH 1606 if (!ipstealth) { 1607#endif 1608 ip->ip_ttl -= IPTTLDEC; 1609#ifdef IPSTEALTH 1610 } 1611#endif 1612 1613 /* 1614 * If forwarding packet using same interface that it came in on, 1615 * perhaps should send a redirect to sender to shortcut a hop. 1616 * Only send redirect if source is sending directly to us, 1617 * and if packet was not source routed (or has any options). 1618 * Also, don't send redirect if forwarding using a default route 1619 * or a route modified by a redirect. 1620 */ 1621#define satosin(sa) ((struct sockaddr_in *)(sa)) 1622 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1623 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1624 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1625 ipsendredirects && !srcrt) { 1626#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1627 u_long src = ntohl(ip->ip_src.s_addr); 1628 1629 if (RTA(rt) && 1630 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1631 if (rt->rt_flags & RTF_GATEWAY) 1632 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1633 else 1634 dest = ip->ip_dst.s_addr; 1635 /* Router requirements says to only send host redirects */ 1636 type = ICMP_REDIRECT; 1637 code = ICMP_REDIRECT_HOST; 1638#ifdef DIAGNOSTIC 1639 if (ipprintfs) 1640 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1641#endif 1642 } 1643 } 1644 1645 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1646 IP_FORWARDING, 0); 1647 if (error) 1648 ipstat.ips_cantforward++; 1649 else { 1650 ipstat.ips_forward++; 1651 if (type) 1652 ipstat.ips_redirectsent++; 1653 else { 1654 if (mcopy) { 1655 ipflow_create(&ipforward_rt, mcopy); 1656 m_freem(mcopy); 1657 } 1658 return; 1659 } 1660 } 1661 if (mcopy == NULL) 1662 return; 1663 destifp = NULL; 1664 1665 switch (error) { 1666 1667 case 0: /* forwarded, but need redirect */ 1668 /* type, code set above */ 1669 break; 1670 1671 case ENETUNREACH: /* shouldn't happen, checked above */ 1672 case EHOSTUNREACH: 1673 case ENETDOWN: 1674 case EHOSTDOWN: 1675 default: 1676 type = ICMP_UNREACH; 1677 code = ICMP_UNREACH_HOST; 1678 break; 1679 1680 case EMSGSIZE: 1681 type = ICMP_UNREACH; 1682 code = ICMP_UNREACH_NEEDFRAG; 1683#ifndef IPSEC 1684 if (ipforward_rt.ro_rt) 1685 destifp = ipforward_rt.ro_rt->rt_ifp; 1686#else 1687 /* 1688 * If the packet is routed over IPsec tunnel, tell the 1689 * originator the tunnel MTU. 1690 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1691 * XXX quickhack!!! 1692 */ 1693 if (ipforward_rt.ro_rt) { 1694 struct secpolicy *sp = NULL; 1695 int ipsecerror; 1696 int ipsechdr; 1697 struct route *ro; 1698 1699 sp = ipsec4_getpolicybyaddr(mcopy, 1700 IPSEC_DIR_OUTBOUND, 1701 IP_FORWARDING, 1702 &ipsecerror); 1703 1704 if (sp == NULL) 1705 destifp = ipforward_rt.ro_rt->rt_ifp; 1706 else { 1707 /* count IPsec header size */ 1708 ipsechdr = ipsec4_hdrsiz(mcopy, 1709 IPSEC_DIR_OUTBOUND, 1710 NULL); 1711 1712 /* 1713 * find the correct route for outer IPv4 1714 * header, compute tunnel MTU. 1715 * 1716 * XXX BUG ALERT 1717 * The "dummyifp" code relies upon the fact 1718 * that icmp_error() touches only ifp->if_mtu. 1719 */ 1720 /*XXX*/ 1721 destifp = NULL; 1722 if (sp->req != NULL 1723 && sp->req->sav != NULL 1724 && sp->req->sav->sah != NULL) { 1725 ro = &sp->req->sav->sah->sa_route; 1726 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1727 dummyifp.if_mtu = 1728 ro->ro_rt->rt_ifp->if_mtu; 1729 dummyifp.if_mtu -= ipsechdr; 1730 destifp = &dummyifp; 1731 } 1732 } 1733 1734 key_freesp(sp); 1735 } 1736 } 1737#endif /*IPSEC*/ 1738 ipstat.ips_cantfrag++; 1739 break; 1740 1741 case ENOBUFS: 1742 type = ICMP_SOURCEQUENCH; 1743 code = 0; 1744 break; 1745 1746 case EACCES: /* ipfw denied packet */ 1747 m_freem(mcopy); 1748 return; 1749 } 1750 icmp_error(mcopy, type, code, dest, destifp); 1751} 1752 1753void 1754ip_savecontrol(inp, mp, ip, m) 1755 register struct inpcb *inp; 1756 register struct mbuf **mp; 1757 register struct ip *ip; 1758 register struct mbuf *m; 1759{ 1760 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1761 struct timeval tv; 1762 1763 microtime(&tv); 1764 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1765 SCM_TIMESTAMP, SOL_SOCKET); 1766 if (*mp) 1767 mp = &(*mp)->m_next; 1768 } 1769 if (inp->inp_flags & INP_RECVDSTADDR) { 1770 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1771 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1772 if (*mp) 1773 mp = &(*mp)->m_next; 1774 } 1775#ifdef notyet 1776 /* XXX 1777 * Moving these out of udp_input() made them even more broken 1778 * than they already were. 1779 */ 1780 /* options were tossed already */ 1781 if (inp->inp_flags & INP_RECVOPTS) { 1782 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1783 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1784 if (*mp) 1785 mp = &(*mp)->m_next; 1786 } 1787 /* ip_srcroute doesn't do what we want here, need to fix */ 1788 if (inp->inp_flags & INP_RECVRETOPTS) { 1789 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1790 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1791 if (*mp) 1792 mp = &(*mp)->m_next; 1793 } 1794#endif 1795 if (inp->inp_flags & INP_RECVIF) { 1796 struct ifnet *ifp; 1797 struct sdlbuf { 1798 struct sockaddr_dl sdl; 1799 u_char pad[32]; 1800 } sdlbuf; 1801 struct sockaddr_dl *sdp; 1802 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1803 1804 if (((ifp = m->m_pkthdr.rcvif)) 1805 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1806 sdp = (struct sockaddr_dl *)(ifnet_addrs 1807 [ifp->if_index - 1]->ifa_addr); 1808 /* 1809 * Change our mind and don't try copy. 1810 */ 1811 if ((sdp->sdl_family != AF_LINK) 1812 || (sdp->sdl_len > sizeof(sdlbuf))) { 1813 goto makedummy; 1814 } 1815 bcopy(sdp, sdl2, sdp->sdl_len); 1816 } else { 1817makedummy: 1818 sdl2->sdl_len 1819 = offsetof(struct sockaddr_dl, sdl_data[0]); 1820 sdl2->sdl_family = AF_LINK; 1821 sdl2->sdl_index = 0; 1822 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1823 } 1824 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1825 IP_RECVIF, IPPROTO_IP); 1826 if (*mp) 1827 mp = &(*mp)->m_next; 1828 } 1829} 1830 1831int 1832ip_rsvp_init(struct socket *so) 1833{ 1834 if (so->so_type != SOCK_RAW || 1835 so->so_proto->pr_protocol != IPPROTO_RSVP) 1836 return EOPNOTSUPP; 1837 1838 if (ip_rsvpd != NULL) 1839 return EADDRINUSE; 1840 1841 ip_rsvpd = so; 1842 /* 1843 * This may seem silly, but we need to be sure we don't over-increment 1844 * the RSVP counter, in case something slips up. 1845 */ 1846 if (!ip_rsvp_on) { 1847 ip_rsvp_on = 1; 1848 rsvp_on++; 1849 } 1850 1851 return 0; 1852} 1853 1854int 1855ip_rsvp_done(void) 1856{ 1857 ip_rsvpd = NULL; 1858 /* 1859 * This may seem silly, but we need to be sure we don't over-decrement 1860 * the RSVP counter, in case something slips up. 1861 */ 1862 if (ip_rsvp_on) { 1863 ip_rsvp_on = 0; 1864 rsvp_on--; 1865 } 1866 return 0; 1867} 1868