1 /* 2 * Copyright (c) 2003, 2004 Jeffrey M. Hsu. All rights reserved. 3 * Copyright (c) 2003, 2004 The DragonFly Project. All rights reserved. 4 * 5 * This code is derived from software contributed to The DragonFly Project 6 * by Jeffrey M. Hsu. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of The DragonFly Project nor the names of its 17 * contributors may be used to endorse or promote products derived 18 * from this software without specific, prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 25 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING, 26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED 28 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 29 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 30 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 */ 33 34 /* 35 * Copyright (c) 1982, 1986, 1988, 1993 36 * The Regents of the University of California. All rights reserved. 37 * 38 * Redistribution and use in source and binary forms, with or without 39 * modification, are permitted provided that the following conditions 40 * are met: 41 * 1. Redistributions of source code must retain the above copyright 42 * notice, this list of conditions and the following disclaimer. 43 * 2. Redistributions in binary form must reproduce the above copyright 44 * notice, this list of conditions and the following disclaimer in the 45 * documentation and/or other materials provided with the distribution. 46 * 3. All advertising materials mentioning features or use of this software 47 * must display the following acknowledgement: 48 * This product includes software developed by the University of 49 * California, Berkeley and its contributors. 50 * 4. Neither the name of the University nor the names of its contributors 51 * may be used to endorse or promote products derived from this software 52 * without specific prior written permission. 53 * 54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 64 * SUCH DAMAGE. 65 * 66 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 67 * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $ 68 * $DragonFly: src/sys/netinet/ip_input.c,v 1.66 2007/04/04 06:13:26 dillon Exp $ 69 */ 70 71 #define _IP_VHL 72 73 #include "opt_bootp.h" 74 #include "opt_ipfw.h" 75 #include "opt_ipdn.h" 76 #include "opt_ipdivert.h" 77 #include "opt_ipfilter.h" 78 #include "opt_ipstealth.h" 79 #include "opt_ipsec.h" 80 81 #include <sys/param.h> 82 #include <sys/systm.h> 83 #include <sys/mbuf.h> 84 #include <sys/malloc.h> 85 #include <sys/mpipe.h> 86 #include <sys/domain.h> 87 #include <sys/protosw.h> 88 #include <sys/socket.h> 89 #include <sys/time.h> 90 #include <sys/globaldata.h> 91 #include <sys/thread.h> 92 #include <sys/kernel.h> 93 #include <sys/syslog.h> 94 #include <sys/sysctl.h> 95 #include <sys/in_cksum.h> 96 97 #include <sys/thread2.h> 98 #include <sys/msgport2.h> 99 100 #include <machine/stdarg.h> 101 102 #include <net/if.h> 103 #include <net/if_types.h> 104 #include <net/if_var.h> 105 #include <net/if_dl.h> 106 #include <net/pfil.h> 107 #include <net/route.h> 108 #include <net/netisr.h> 109 #include <net/intrq.h> 110 111 #include <netinet/in.h> 112 #include <netinet/in_systm.h> 113 #include <netinet/in_var.h> 114 #include <netinet/ip.h> 115 #include <netinet/in_pcb.h> 116 #include <netinet/ip_var.h> 117 #include <netinet/ip_icmp.h> 118 119 120 #include <sys/socketvar.h> 121 122 #include <net/ipfw/ip_fw.h> 123 #include <net/dummynet/ip_dummynet.h> 124 125 #ifdef IPSEC 126 #include <netinet6/ipsec.h> 127 #include <netproto/key/key.h> 128 #endif 129 130 #ifdef FAST_IPSEC 131 #include <netproto/ipsec/ipsec.h> 132 #include <netproto/ipsec/key.h> 133 #endif 134 135 int rsvp_on = 0; 136 static int ip_rsvp_on; 137 struct socket *ip_rsvpd; 138 139 int ipforwarding = 0; 140 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 141 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 142 143 static int ipsendredirects = 1; /* XXX */ 144 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 145 &ipsendredirects, 0, "Enable sending IP redirects"); 146 147 int ip_defttl = IPDEFTTL; 148 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 149 &ip_defttl, 0, "Maximum TTL on IP packets"); 150 151 static int ip_dosourceroute = 0; 152 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 153 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 154 155 static int ip_acceptsourceroute = 0; 156 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 157 CTLFLAG_RW, &ip_acceptsourceroute, 0, 158 "Enable accepting source routed IP packets"); 159 160 static int ip_keepfaith = 0; 161 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 162 &ip_keepfaith, 0, 163 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 164 165 static int nipq = 0; /* total # of reass queues */ 166 static int maxnipq; 167 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 168 &maxnipq, 0, 169 "Maximum number of IPv4 fragment reassembly queue entries"); 170 171 static int maxfragsperpacket; 172 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 173 &maxfragsperpacket, 0, 174 "Maximum number of IPv4 fragments allowed per packet"); 175 176 static int ip_sendsourcequench = 0; 177 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 178 &ip_sendsourcequench, 0, 179 "Enable the transmission of source quench packets"); 180 181 int ip_do_randomid = 0; 182 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 183 &ip_do_randomid, 0, 184 "Assign random ip_id values"); 185 /* 186 * XXX - Setting ip_checkinterface mostly implements the receive side of 187 * the Strong ES model described in RFC 1122, but since the routing table 188 * and transmit implementation do not implement the Strong ES model, 189 * setting this to 1 results in an odd hybrid. 190 * 191 * XXX - ip_checkinterface currently must be disabled if you use ipnat 192 * to translate the destination address to another local interface. 193 * 194 * XXX - ip_checkinterface must be disabled if you add IP aliases 195 * to the loopback interface instead of the interface where the 196 * packets for those addresses are received. 197 */ 198 static int ip_checkinterface = 0; 199 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 200 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 201 202 #ifdef DIAGNOSTIC 203 static int ipprintfs = 0; 204 #endif 205 206 static struct ifqueue ipintrq; 207 static int ipqmaxlen = IFQ_MAXLEN; 208 209 extern struct domain inetdomain; 210 extern struct protosw inetsw[]; 211 u_char ip_protox[IPPROTO_MAX]; 212 struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 213 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 214 u_long in_ifaddrhmask; /* mask for hash table */ 215 216 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 217 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 218 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 219 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 220 221 struct ip_stats ipstats_percpu[MAXCPU]; 222 #ifdef SMP 223 static int 224 sysctl_ipstats(SYSCTL_HANDLER_ARGS) 225 { 226 int cpu, error = 0; 227 228 for (cpu = 0; cpu < ncpus; ++cpu) { 229 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu], 230 sizeof(struct ip_stats)))) 231 break; 232 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu], 233 sizeof(struct ip_stats)))) 234 break; 235 } 236 237 return (error); 238 } 239 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW), 240 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics"); 241 #else 242 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 243 &ipstat, ip_stats, "IP statistics"); 244 #endif 245 246 /* Packet reassembly stuff */ 247 #define IPREASS_NHASH_LOG2 6 248 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 249 #define IPREASS_HMASK (IPREASS_NHASH - 1) 250 #define IPREASS_HASH(x,y) \ 251 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 252 253 static struct ipq ipq[IPREASS_NHASH]; 254 const int ipintrq_present = 1; 255 256 #ifdef IPCTL_DEFMTU 257 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 258 &ip_mtu, 0, "Default MTU"); 259 #endif 260 261 #ifdef IPSTEALTH 262 static int ipstealth = 0; 263 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, ""); 264 #else 265 static const int ipstealth = 0; 266 #endif 267 268 269 /* Firewall hooks */ 270 ip_fw_chk_t *ip_fw_chk_ptr; 271 int fw_enable = 1; 272 int fw_one_pass = 1; 273 274 /* Dummynet hooks */ 275 ip_dn_io_t *ip_dn_io_ptr; 276 277 struct pfil_head inet_pfil_hook; 278 279 /* 280 * XXX this is ugly -- the following two global variables are 281 * used to store packet state while it travels through the stack. 282 * Note that the code even makes assumptions on the size and 283 * alignment of fields inside struct ip_srcrt so e.g. adding some 284 * fields will break the code. This needs to be fixed. 285 * 286 * We need to save the IP options in case a protocol wants to respond 287 * to an incoming packet over the same route if the packet got here 288 * using IP source routing. This allows connection establishment and 289 * maintenance when the remote end is on a network that is not known 290 * to us. 291 */ 292 static int ip_nhops = 0; 293 294 static struct ip_srcrt { 295 struct in_addr dst; /* final destination */ 296 char nop; /* one NOP to align */ 297 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 298 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 299 } ip_srcrt; 300 301 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management"); 302 static struct malloc_pipe ipq_mpipe; 303 304 static void save_rte (u_char *, struct in_addr); 305 static int ip_dooptions (struct mbuf *m, int, 306 struct sockaddr_in *next_hop); 307 static void ip_forward (struct mbuf *m, boolean_t using_srcrt, 308 struct sockaddr_in *next_hop); 309 static void ip_freef (struct ipq *); 310 static int ip_input_handler (struct netmsg *); 311 static struct mbuf *ip_reass (struct mbuf *, struct ipq *, 312 struct ipq *, u_int32_t *); 313 314 /* 315 * IP initialization: fill in IP protocol switch table. 316 * All protocols not implemented in kernel go to raw IP protocol handler. 317 */ 318 void 319 ip_init(void) 320 { 321 struct protosw *pr; 322 int i; 323 #ifdef SMP 324 int cpu; 325 #endif 326 327 /* 328 * Make sure we can handle a reasonable number of fragments but 329 * cap it at 4000 (XXX). 330 */ 331 mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq), 332 IFQ_MAXLEN, 4000, 0, NULL); 333 TAILQ_INIT(&in_ifaddrhead); 334 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 335 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 336 if (pr == NULL) 337 panic("ip_init"); 338 for (i = 0; i < IPPROTO_MAX; i++) 339 ip_protox[i] = pr - inetsw; 340 for (pr = inetdomain.dom_protosw; 341 pr < inetdomain.dom_protoswNPROTOSW; pr++) 342 if (pr->pr_domain->dom_family == PF_INET && 343 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 344 ip_protox[pr->pr_protocol] = pr - inetsw; 345 346 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 347 inet_pfil_hook.ph_af = AF_INET; 348 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) { 349 kprintf("%s: WARNING: unable to register pfil hook, " 350 "error %d\n", __func__, i); 351 } 352 353 for (i = 0; i < IPREASS_NHASH; i++) 354 ipq[i].next = ipq[i].prev = &ipq[i]; 355 356 maxnipq = nmbclusters / 32; 357 maxfragsperpacket = 16; 358 359 ip_id = time_second & 0xffff; 360 ipintrq.ifq_maxlen = ipqmaxlen; 361 362 /* 363 * Initialize IP statistics counters for each CPU. 364 * 365 */ 366 #ifdef SMP 367 for (cpu = 0; cpu < ncpus; ++cpu) { 368 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats)); 369 } 370 #else 371 bzero(&ipstat, sizeof(struct ip_stats)); 372 #endif 373 374 netisr_register(NETISR_IP, ip_mport, ip_input_handler); 375 } 376 377 /* 378 * XXX watch out this one. It is perhaps used as a cache for 379 * the most recently used route ? it is cleared in in_addroute() 380 * when a new route is successfully created. 381 */ 382 struct route ipforward_rt[MAXCPU]; 383 384 /* Do transport protocol processing. */ 385 static void 386 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip, 387 struct sockaddr_in *nexthop) 388 { 389 /* 390 * Switch out to protocol's input routine. 391 */ 392 if (nexthop && ip->ip_p == IPPROTO_TCP) { 393 /* TCP needs IPFORWARD info if available */ 394 struct m_hdr tag; 395 396 tag.mh_type = MT_TAG; 397 tag.mh_flags = PACKET_TAG_IPFORWARD; 398 tag.mh_data = (caddr_t)nexthop; 399 tag.mh_next = m; 400 401 (*inetsw[ip_protox[ip->ip_p]].pr_input) 402 ((struct mbuf *)&tag, hlen, ip->ip_p); 403 } else { 404 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, ip->ip_p); 405 } 406 } 407 408 struct netmsg_transport_packet { 409 struct lwkt_msg nm_lmsg; 410 struct mbuf *nm_mbuf; 411 int nm_hlen; 412 boolean_t nm_hasnexthop; 413 struct sockaddr_in nm_nexthop; 414 }; 415 416 static int 417 transport_processing_handler(lwkt_msg_t lmsg) 418 { 419 struct netmsg_transport_packet *msg = (void *)lmsg; 420 struct sockaddr_in *nexthop; 421 struct ip *ip; 422 423 ip = mtod(msg->nm_mbuf, struct ip *); 424 nexthop = msg->nm_hasnexthop ? &msg->nm_nexthop : NULL; 425 transport_processing_oncpu(msg->nm_mbuf, msg->nm_hlen, ip, nexthop); 426 lwkt_replymsg(lmsg, 0); 427 return(EASYNC); 428 } 429 430 static int 431 ip_input_handler(struct netmsg *msg0) 432 { 433 struct mbuf *m = ((struct netmsg_packet *)msg0)->nm_packet; 434 435 ip_input(m); 436 /* msg0 was embedded in the mbuf, do not reply! */ 437 return(EASYNC); 438 } 439 440 /* 441 * IP input routine. Checksum and byte swap header. If fragmented 442 * try to reassemble. Process options. Pass to next level. 443 */ 444 void 445 ip_input(struct mbuf *m) 446 { 447 struct ip *ip; 448 struct ipq *fp; 449 struct in_ifaddr *ia = NULL; 450 struct ifaddr *ifa; 451 int i, hlen, checkif; 452 u_short sum; 453 struct in_addr pkt_dst; 454 u_int32_t divert_info = 0; /* packet divert/tee info */ 455 struct ip_fw_args args; 456 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */ 457 boolean_t needredispatch = FALSE; 458 struct in_addr odst; /* original dst address(NAT) */ 459 #if defined(FAST_IPSEC) || defined(IPDIVERT) 460 struct m_tag *mtag; 461 #endif 462 #ifdef FAST_IPSEC 463 struct tdb_ident *tdbi; 464 struct secpolicy *sp; 465 int error; 466 #endif 467 468 args.eh = NULL; 469 args.oif = NULL; 470 args.rule = NULL; 471 args.next_hop = NULL; 472 473 /* Grab info from MT_TAG mbufs prepended to the chain. */ 474 while (m != NULL && m->m_type == MT_TAG) { 475 switch(m->_m_tag_id) { 476 case PACKET_TAG_DUMMYNET: 477 args.rule = ((struct dn_pkt *)m)->rule; 478 break; 479 case PACKET_TAG_IPFORWARD: 480 args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data; 481 break; 482 default: 483 kprintf("ip_input: unrecognised MT_TAG tag %d\n", 484 m->_m_tag_id); 485 break; 486 } 487 m = m->m_next; 488 } 489 KASSERT(m != NULL && (m->m_flags & M_PKTHDR), ("ip_input: no HDR")); 490 491 if (args.rule != NULL) { /* dummynet already filtered us */ 492 ip = mtod(m, struct ip *); 493 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 494 goto iphack; 495 } 496 497 ipstat.ips_total++; 498 499 /* length checks already done in ip_demux() */ 500 KASSERT(m->m_len >= sizeof(ip), ("IP header not in one mbuf")); 501 502 ip = mtod(m, struct ip *); 503 504 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 505 ipstat.ips_badvers++; 506 goto bad; 507 } 508 509 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 510 /* length checks already done in ip_demux() */ 511 KASSERT(hlen >= sizeof(struct ip), ("IP header len too small")); 512 KASSERT(m->m_len >= hlen, ("packet shorter than IP header length")); 513 514 /* 127/8 must not appear on wire - RFC1122 */ 515 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 516 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 517 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) { 518 ipstat.ips_badaddr++; 519 goto bad; 520 } 521 } 522 523 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 524 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 525 } else { 526 if (hlen == sizeof(struct ip)) { 527 sum = in_cksum_hdr(ip); 528 } else { 529 sum = in_cksum(m, hlen); 530 } 531 } 532 if (sum != 0) { 533 ipstat.ips_badsum++; 534 goto bad; 535 } 536 537 #ifdef ALTQ 538 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 539 /* packet is dropped by traffic conditioner */ 540 return; 541 } 542 #endif 543 /* 544 * Convert fields to host representation. 545 */ 546 ip->ip_len = ntohs(ip->ip_len); 547 if (ip->ip_len < hlen) { 548 ipstat.ips_badlen++; 549 goto bad; 550 } 551 ip->ip_off = ntohs(ip->ip_off); 552 553 /* 554 * Check that the amount of data in the buffers 555 * is as at least much as the IP header would have us expect. 556 * Trim mbufs if longer than we expect. 557 * Drop packet if shorter than we expect. 558 */ 559 if (m->m_pkthdr.len < ip->ip_len) { 560 ipstat.ips_tooshort++; 561 goto bad; 562 } 563 if (m->m_pkthdr.len > ip->ip_len) { 564 if (m->m_len == m->m_pkthdr.len) { 565 m->m_len = ip->ip_len; 566 m->m_pkthdr.len = ip->ip_len; 567 } else 568 m_adj(m, ip->ip_len - m->m_pkthdr.len); 569 } 570 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF) 571 /* 572 * Bypass packet filtering for packets from a tunnel (gif). 573 */ 574 if (ipsec_gethist(m, NULL)) 575 goto pass; 576 #endif 577 578 /* 579 * IpHack's section. 580 * Right now when no processing on packet has done 581 * and it is still fresh out of network we do our black 582 * deals with it. 583 * - Firewall: deny/allow/divert 584 * - Xlate: translate packet's addr/port (NAT). 585 * - Pipe: pass pkt through dummynet. 586 * - Wrap: fake packet's addr/port <unimpl.> 587 * - Encapsulate: put it in another IP and send out. <unimp.> 588 */ 589 590 iphack: 591 592 /* 593 * Run through list of hooks for input packets. 594 * 595 * NB: Beware of the destination address changing (e.g. 596 * by NAT rewriting). When this happens, tell 597 * ip_forward to do the right thing. 598 */ 599 if (pfil_has_hooks(&inet_pfil_hook)) { 600 odst = ip->ip_dst; 601 if (pfil_run_hooks(&inet_pfil_hook, &m, 602 m->m_pkthdr.rcvif, PFIL_IN)) { 603 return; 604 } 605 if (m == NULL) /* consumed by filter */ 606 return; 607 ip = mtod(m, struct ip *); 608 using_srcrt = (odst.s_addr != ip->ip_dst.s_addr); 609 } 610 611 if (fw_enable && IPFW_LOADED) { 612 /* 613 * If we've been forwarded from the output side, then 614 * skip the firewall a second time 615 */ 616 if (args.next_hop != NULL) 617 goto ours; 618 619 args.m = m; 620 i = ip_fw_chk_ptr(&args); 621 m = args.m; 622 623 if ((i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */ 624 if (m != NULL) 625 m_freem(m); 626 return; 627 } 628 ip = mtod(m, struct ip *); /* just in case m changed */ 629 if (i == 0 && args.next_hop == NULL) /* common case */ 630 goto pass; 631 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG)) { 632 /* Send packet to the appropriate pipe */ 633 ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args); 634 return; 635 } 636 #ifdef IPDIVERT 637 if (i != 0 && !(i & IP_FW_PORT_DYNT_FLAG)) { 638 /* Divert or tee packet */ 639 divert_info = i; 640 goto ours; 641 } 642 #endif 643 if (i == 0 && args.next_hop != NULL) 644 goto pass; 645 /* 646 * if we get here, the packet must be dropped 647 */ 648 m_freem(m); 649 return; 650 } 651 pass: 652 653 /* 654 * Process options and, if not destined for us, 655 * ship it on. ip_dooptions returns 1 when an 656 * error was detected (causing an icmp message 657 * to be sent and the original packet to be freed). 658 */ 659 ip_nhops = 0; /* for source routed packets */ 660 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, args.next_hop)) 661 return; 662 663 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 664 * matter if it is destined to another node, or whether it is 665 * a multicast one, RSVP wants it! and prevents it from being forwarded 666 * anywhere else. Also checks if the rsvp daemon is running before 667 * grabbing the packet. 668 */ 669 if (rsvp_on && ip->ip_p == IPPROTO_RSVP) 670 goto ours; 671 672 /* 673 * Check our list of addresses, to see if the packet is for us. 674 * If we don't have any addresses, assume any unicast packet 675 * we receive might be for us (and let the upper layers deal 676 * with it). 677 */ 678 if (TAILQ_EMPTY(&in_ifaddrhead) && !(m->m_flags & (M_MCAST | M_BCAST))) 679 goto ours; 680 681 /* 682 * Cache the destination address of the packet; this may be 683 * changed by use of 'ipfw fwd'. 684 */ 685 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst; 686 687 /* 688 * Enable a consistency check between the destination address 689 * and the arrival interface for a unicast packet (the RFC 1122 690 * strong ES model) if IP forwarding is disabled and the packet 691 * is not locally generated and the packet is not subject to 692 * 'ipfw fwd'. 693 * 694 * XXX - Checking also should be disabled if the destination 695 * address is ipnat'ed to a different interface. 696 * 697 * XXX - Checking is incompatible with IP aliases added 698 * to the loopback interface instead of the interface where 699 * the packets are received. 700 */ 701 checkif = ip_checkinterface && 702 !ipforwarding && 703 m->m_pkthdr.rcvif != NULL && 704 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) && 705 (args.next_hop == NULL); 706 707 /* 708 * Check for exact addresses in the hash bucket. 709 */ 710 LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) { 711 /* 712 * If the address matches, verify that the packet 713 * arrived via the correct interface if checking is 714 * enabled. 715 */ 716 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 717 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 718 goto ours; 719 } 720 /* 721 * Check for broadcast addresses. 722 * 723 * Only accept broadcast packets that arrive via the matching 724 * interface. Reception of forwarded directed broadcasts would 725 * be handled via ip_forward() and ether_output() with the loopback 726 * into the stack for SIMPLEX interfaces handled by ether_output(). 727 */ 728 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 729 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 730 if (ifa->ifa_addr == NULL) /* shutdown/startup race */ 731 continue; 732 if (ifa->ifa_addr->sa_family != AF_INET) 733 continue; 734 ia = ifatoia(ifa); 735 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 736 pkt_dst.s_addr) 737 goto ours; 738 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 739 goto ours; 740 #ifdef BOOTP_COMPAT 741 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 742 goto ours; 743 #endif 744 } 745 } 746 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 747 struct in_multi *inm; 748 749 if (ip_mrouter != NULL) { 750 /* 751 * If we are acting as a multicast router, all 752 * incoming multicast packets are passed to the 753 * kernel-level multicast forwarding function. 754 * The packet is returned (relatively) intact; if 755 * ip_mforward() returns a non-zero value, the packet 756 * must be discarded, else it may be accepted below. 757 */ 758 if (ip_mforward != NULL && 759 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) { 760 ipstat.ips_cantforward++; 761 m_freem(m); 762 return; 763 } 764 765 /* 766 * The process-level routing daemon needs to receive 767 * all multicast IGMP packets, whether or not this 768 * host belongs to their destination groups. 769 */ 770 if (ip->ip_p == IPPROTO_IGMP) 771 goto ours; 772 ipstat.ips_forward++; 773 } 774 /* 775 * See if we belong to the destination multicast group on the 776 * arrival interface. 777 */ 778 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 779 if (inm == NULL) { 780 ipstat.ips_notmember++; 781 m_freem(m); 782 return; 783 } 784 goto ours; 785 } 786 if (ip->ip_dst.s_addr == INADDR_BROADCAST) 787 goto ours; 788 if (ip->ip_dst.s_addr == INADDR_ANY) 789 goto ours; 790 791 /* 792 * FAITH(Firewall Aided Internet Translator) 793 */ 794 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 795 if (ip_keepfaith) { 796 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 797 goto ours; 798 } 799 m_freem(m); 800 return; 801 } 802 803 /* 804 * Not for us; forward if possible and desirable. 805 */ 806 if (!ipforwarding) { 807 ipstat.ips_cantforward++; 808 m_freem(m); 809 } else { 810 #ifdef IPSEC 811 /* 812 * Enforce inbound IPsec SPD. 813 */ 814 if (ipsec4_in_reject(m, NULL)) { 815 ipsecstat.in_polvio++; 816 goto bad; 817 } 818 #endif 819 #ifdef FAST_IPSEC 820 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 821 crit_enter(); 822 if (mtag != NULL) { 823 tdbi = (struct tdb_ident *)m_tag_data(mtag); 824 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 825 } else { 826 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 827 IP_FORWARDING, &error); 828 } 829 if (sp == NULL) { /* NB: can happen if error */ 830 crit_exit(); 831 /*XXX error stat???*/ 832 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 833 goto bad; 834 } 835 836 /* 837 * Check security policy against packet attributes. 838 */ 839 error = ipsec_in_reject(sp, m); 840 KEY_FREESP(&sp); 841 crit_exit(); 842 if (error) { 843 ipstat.ips_cantforward++; 844 goto bad; 845 } 846 #endif 847 ip_forward(m, using_srcrt, args.next_hop); 848 } 849 return; 850 851 ours: 852 853 /* 854 * IPSTEALTH: Process non-routing options only 855 * if the packet is destined for us. 856 */ 857 if (ipstealth && 858 hlen > sizeof(struct ip) && 859 ip_dooptions(m, 1, args.next_hop)) 860 return; 861 862 /* Count the packet in the ip address stats */ 863 if (ia != NULL) { 864 ia->ia_ifa.if_ipackets++; 865 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 866 } 867 868 /* 869 * If offset or IP_MF are set, must reassemble. 870 * Otherwise, nothing need be done. 871 * (We could look in the reassembly queue to see 872 * if the packet was previously fragmented, 873 * but it's not worth the time; just let them time out.) 874 */ 875 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 876 877 /* If maxnipq is 0, never accept fragments. */ 878 if (maxnipq == 0) { 879 ipstat.ips_fragments++; 880 ipstat.ips_fragdropped++; 881 goto bad; 882 } 883 884 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 885 /* 886 * Look for queue of fragments 887 * of this datagram. 888 */ 889 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) 890 if (ip->ip_id == fp->ipq_id && 891 ip->ip_src.s_addr == fp->ipq_src.s_addr && 892 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 893 ip->ip_p == fp->ipq_p) 894 goto found; 895 896 fp = NULL; 897 898 /* 899 * Enforce upper bound on number of fragmented packets 900 * for which we attempt reassembly; 901 * If maxnipq is -1, accept all fragments without limitation. 902 */ 903 if ((nipq > maxnipq) && (maxnipq > 0)) { 904 /* 905 * drop something from the tail of the current queue 906 * before proceeding further 907 */ 908 if (ipq[sum].prev == &ipq[sum]) { /* gak */ 909 for (i = 0; i < IPREASS_NHASH; i++) { 910 if (ipq[i].prev != &ipq[i]) { 911 ipstat.ips_fragtimeout += 912 ipq[i].prev->ipq_nfrags; 913 ip_freef(ipq[i].prev); 914 break; 915 } 916 } 917 } else { 918 ipstat.ips_fragtimeout += 919 ipq[sum].prev->ipq_nfrags; 920 ip_freef(ipq[sum].prev); 921 } 922 } 923 found: 924 /* 925 * Adjust ip_len to not reflect header, 926 * convert offset of this to bytes. 927 */ 928 ip->ip_len -= hlen; 929 if (ip->ip_off & IP_MF) { 930 /* 931 * Make sure that fragments have a data length 932 * that's a non-zero multiple of 8 bytes. 933 */ 934 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 935 ipstat.ips_toosmall++; /* XXX */ 936 goto bad; 937 } 938 m->m_flags |= M_FRAG; 939 } else 940 m->m_flags &= ~M_FRAG; 941 ip->ip_off <<= 3; 942 943 /* 944 * Attempt reassembly; if it succeeds, proceed. 945 * ip_reass() will return a different mbuf, and update 946 * the divert info in divert_info. 947 */ 948 ipstat.ips_fragments++; 949 m->m_pkthdr.header = ip; 950 m = ip_reass(m, fp, &ipq[sum], &divert_info); 951 if (m == NULL) 952 return; 953 ipstat.ips_reassembled++; 954 needredispatch = TRUE; 955 ip = mtod(m, struct ip *); 956 /* Get the header length of the reassembled packet */ 957 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 958 #ifdef IPDIVERT 959 /* Restore original checksum before diverting packet */ 960 if (divert_info != 0) { 961 ip->ip_len += hlen; 962 ip->ip_len = htons(ip->ip_len); 963 ip->ip_off = htons(ip->ip_off); 964 ip->ip_sum = 0; 965 if (hlen == sizeof(struct ip)) 966 ip->ip_sum = in_cksum_hdr(ip); 967 else 968 ip->ip_sum = in_cksum(m, hlen); 969 ip->ip_off = ntohs(ip->ip_off); 970 ip->ip_len = ntohs(ip->ip_len); 971 ip->ip_len -= hlen; 972 } 973 #endif 974 } else { 975 ip->ip_len -= hlen; 976 } 977 978 #ifdef IPDIVERT 979 /* 980 * Divert or tee packet to the divert protocol if required. 981 */ 982 if (divert_info != 0) { 983 struct mbuf *clone = NULL; 984 985 /* Clone packet if we're doing a 'tee' */ 986 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 987 clone = m_dup(m, MB_DONTWAIT); 988 989 /* Restore packet header fields to original values */ 990 ip->ip_len += hlen; 991 ip->ip_len = htons(ip->ip_len); 992 ip->ip_off = htons(ip->ip_off); 993 994 /* Deliver packet to divert input routine */ 995 divert_packet(m, 1, divert_info & 0xffff); 996 ipstat.ips_delivered++; 997 998 /* If 'tee', continue with original packet */ 999 if (clone == NULL) 1000 return; 1001 m = clone; 1002 ip = mtod(m, struct ip *); 1003 ip->ip_len += hlen; 1004 /* 1005 * Jump backwards to complete processing of the 1006 * packet. But first clear divert_info to avoid 1007 * entering this block again. 1008 * We do not need to clear args.divert_rule 1009 * or args.next_hop as they will not be used. 1010 * 1011 * XXX Better safe than sorry, remove the DIVERT tag. 1012 */ 1013 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL); 1014 if (mtag != NULL) 1015 m_tag_delete(m, mtag); 1016 1017 divert_info = 0; 1018 goto pass; 1019 } 1020 #endif 1021 1022 #ifdef IPSEC 1023 /* 1024 * enforce IPsec policy checking if we are seeing last header. 1025 * note that we do not visit this with protocols with pcb layer 1026 * code - like udp/tcp/raw ip. 1027 */ 1028 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) && 1029 ipsec4_in_reject(m, NULL)) { 1030 ipsecstat.in_polvio++; 1031 goto bad; 1032 } 1033 #endif 1034 #if FAST_IPSEC 1035 /* 1036 * enforce IPsec policy checking if we are seeing last header. 1037 * note that we do not visit this with protocols with pcb layer 1038 * code - like udp/tcp/raw ip. 1039 */ 1040 if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) { 1041 /* 1042 * Check if the packet has already had IPsec processing 1043 * done. If so, then just pass it along. This tag gets 1044 * set during AH, ESP, etc. input handling, before the 1045 * packet is returned to the ip input queue for delivery. 1046 */ 1047 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 1048 crit_enter(); 1049 if (mtag != NULL) { 1050 tdbi = (struct tdb_ident *)m_tag_data(mtag); 1051 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 1052 } else { 1053 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 1054 IP_FORWARDING, &error); 1055 } 1056 if (sp != NULL) { 1057 /* 1058 * Check security policy against packet attributes. 1059 */ 1060 error = ipsec_in_reject(sp, m); 1061 KEY_FREESP(&sp); 1062 } else { 1063 /* XXX error stat??? */ 1064 error = EINVAL; 1065 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 1066 goto bad; 1067 } 1068 crit_exit(); 1069 if (error) 1070 goto bad; 1071 } 1072 #endif /* FAST_IPSEC */ 1073 1074 ipstat.ips_delivered++; 1075 if (needredispatch) { 1076 struct netmsg_transport_packet *msg; 1077 lwkt_port_t port; 1078 1079 ip->ip_off = htons(ip->ip_off); 1080 ip->ip_len = htons(ip->ip_len); 1081 port = ip_mport(&m); 1082 if (port == NULL) 1083 return; 1084 1085 msg = kmalloc(sizeof(struct netmsg_transport_packet), M_LWKTMSG, 1086 M_INTWAIT | M_NULLOK); 1087 if (msg == NULL) 1088 goto bad; 1089 1090 lwkt_initmsg(&msg->nm_lmsg, &netisr_afree_rport, 0, 1091 lwkt_cmd_func(transport_processing_handler), 1092 lwkt_cmd_op_none); 1093 msg->nm_hlen = hlen; 1094 msg->nm_hasnexthop = (args.next_hop != NULL); 1095 if (msg->nm_hasnexthop) 1096 msg->nm_nexthop = *args.next_hop; /* structure copy */ 1097 1098 msg->nm_mbuf = m; 1099 ip = mtod(m, struct ip *); 1100 ip->ip_len = ntohs(ip->ip_len); 1101 ip->ip_off = ntohs(ip->ip_off); 1102 lwkt_sendmsg(port, &msg->nm_lmsg); 1103 } else { 1104 transport_processing_oncpu(m, hlen, ip, args.next_hop); 1105 } 1106 return; 1107 1108 bad: 1109 m_freem(m); 1110 } 1111 1112 /* 1113 * Take incoming datagram fragment and try to reassemble it into 1114 * whole datagram. If a chain for reassembly of this datagram already 1115 * exists, then it is given as fp; otherwise have to make a chain. 1116 * 1117 * When IPDIVERT enabled, keep additional state with each packet that 1118 * tells us if we need to divert or tee the packet we're building. 1119 * In particular, *divinfo includes the port and TEE flag. 1120 */ 1121 1122 static struct mbuf * 1123 ip_reass(struct mbuf *m, struct ipq *fp, struct ipq *where, 1124 u_int32_t *divinfo) 1125 { 1126 struct ip *ip = mtod(m, struct ip *); 1127 struct mbuf *p = NULL, *q, *nq; 1128 struct mbuf *n; 1129 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 1130 int i, next; 1131 #ifdef IPDIVERT 1132 struct m_tag *mtag; 1133 #endif 1134 1135 /* 1136 * Presence of header sizes in mbufs 1137 * would confuse code below. 1138 */ 1139 m->m_data += hlen; 1140 m->m_len -= hlen; 1141 1142 /* 1143 * If first fragment to arrive, create a reassembly queue. 1144 */ 1145 if (fp == NULL) { 1146 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL) 1147 goto dropfrag; 1148 insque(fp, where); 1149 nipq++; 1150 fp->ipq_nfrags = 1; 1151 fp->ipq_ttl = IPFRAGTTL; 1152 fp->ipq_p = ip->ip_p; 1153 fp->ipq_id = ip->ip_id; 1154 fp->ipq_src = ip->ip_src; 1155 fp->ipq_dst = ip->ip_dst; 1156 fp->ipq_frags = m; 1157 m->m_nextpkt = NULL; 1158 #ifdef IPDIVERT 1159 fp->ipq_div_info = 0; 1160 #endif 1161 goto inserted; 1162 } else { 1163 fp->ipq_nfrags++; 1164 } 1165 1166 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1167 1168 /* 1169 * Find a segment which begins after this one does. 1170 */ 1171 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1172 if (GETIP(q)->ip_off > ip->ip_off) 1173 break; 1174 1175 /* 1176 * If there is a preceding segment, it may provide some of 1177 * our data already. If so, drop the data from the incoming 1178 * segment. If it provides all of our data, drop us, otherwise 1179 * stick new segment in the proper place. 1180 * 1181 * If some of the data is dropped from the the preceding 1182 * segment, then it's checksum is invalidated. 1183 */ 1184 if (p) { 1185 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1186 if (i > 0) { 1187 if (i >= ip->ip_len) 1188 goto dropfrag; 1189 m_adj(m, i); 1190 m->m_pkthdr.csum_flags = 0; 1191 ip->ip_off += i; 1192 ip->ip_len -= i; 1193 } 1194 m->m_nextpkt = p->m_nextpkt; 1195 p->m_nextpkt = m; 1196 } else { 1197 m->m_nextpkt = fp->ipq_frags; 1198 fp->ipq_frags = m; 1199 } 1200 1201 /* 1202 * While we overlap succeeding segments trim them or, 1203 * if they are completely covered, dequeue them. 1204 */ 1205 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1206 q = nq) { 1207 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1208 if (i < GETIP(q)->ip_len) { 1209 GETIP(q)->ip_len -= i; 1210 GETIP(q)->ip_off += i; 1211 m_adj(q, i); 1212 q->m_pkthdr.csum_flags = 0; 1213 break; 1214 } 1215 nq = q->m_nextpkt; 1216 m->m_nextpkt = nq; 1217 ipstat.ips_fragdropped++; 1218 fp->ipq_nfrags--; 1219 q->m_nextpkt = NULL; 1220 m_freem(q); 1221 } 1222 1223 inserted: 1224 1225 #ifdef IPDIVERT 1226 /* 1227 * Transfer firewall instructions to the fragment structure. 1228 * Only trust info in the fragment at offset 0. 1229 */ 1230 if (ip->ip_off == 0) { 1231 fp->ipq_div_info = *divinfo; 1232 } else { 1233 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL); 1234 if (mtag != NULL) 1235 m_tag_delete(m, mtag); 1236 } 1237 *divinfo = 0; 1238 #endif 1239 1240 /* 1241 * Check for complete reassembly and perform frag per packet 1242 * limiting. 1243 * 1244 * Frag limiting is performed here so that the nth frag has 1245 * a chance to complete the packet before we drop the packet. 1246 * As a result, n+1 frags are actually allowed per packet, but 1247 * only n will ever be stored. (n = maxfragsperpacket.) 1248 * 1249 */ 1250 next = 0; 1251 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1252 if (GETIP(q)->ip_off != next) { 1253 if (fp->ipq_nfrags > maxfragsperpacket) { 1254 ipstat.ips_fragdropped += fp->ipq_nfrags; 1255 ip_freef(fp); 1256 } 1257 return (NULL); 1258 } 1259 next += GETIP(q)->ip_len; 1260 } 1261 /* Make sure the last packet didn't have the IP_MF flag */ 1262 if (p->m_flags & M_FRAG) { 1263 if (fp->ipq_nfrags > maxfragsperpacket) { 1264 ipstat.ips_fragdropped += fp->ipq_nfrags; 1265 ip_freef(fp); 1266 } 1267 return (NULL); 1268 } 1269 1270 /* 1271 * Reassembly is complete. Make sure the packet is a sane size. 1272 */ 1273 q = fp->ipq_frags; 1274 ip = GETIP(q); 1275 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 1276 ipstat.ips_toolong++; 1277 ipstat.ips_fragdropped += fp->ipq_nfrags; 1278 ip_freef(fp); 1279 return (NULL); 1280 } 1281 1282 /* 1283 * Concatenate fragments. 1284 */ 1285 m = q; 1286 n = m->m_next; 1287 m->m_next = NULL; 1288 m_cat(m, n); 1289 nq = q->m_nextpkt; 1290 q->m_nextpkt = NULL; 1291 for (q = nq; q != NULL; q = nq) { 1292 nq = q->m_nextpkt; 1293 q->m_nextpkt = NULL; 1294 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1295 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1296 m_cat(m, q); 1297 } 1298 1299 #ifdef IPDIVERT 1300 /* 1301 * Extract firewall instructions from the fragment structure. 1302 */ 1303 *divinfo = fp->ipq_div_info; 1304 #endif 1305 1306 /* 1307 * Create header for new ip packet by 1308 * modifying header of first packet; 1309 * dequeue and discard fragment reassembly header. 1310 * Make header visible. 1311 */ 1312 ip->ip_len = next; 1313 ip->ip_src = fp->ipq_src; 1314 ip->ip_dst = fp->ipq_dst; 1315 remque(fp); 1316 nipq--; 1317 mpipe_free(&ipq_mpipe, fp); 1318 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 1319 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 1320 /* some debugging cruft by sklower, below, will go away soon */ 1321 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1322 int plen = 0; 1323 1324 for (n = m; n; n = n->m_next) 1325 plen += n->m_len; 1326 m->m_pkthdr.len = plen; 1327 } 1328 return (m); 1329 1330 dropfrag: 1331 #ifdef IPDIVERT 1332 *divinfo = 0; 1333 #endif 1334 ipstat.ips_fragdropped++; 1335 if (fp != NULL) 1336 fp->ipq_nfrags--; 1337 m_freem(m); 1338 return (NULL); 1339 1340 #undef GETIP 1341 } 1342 1343 /* 1344 * Free a fragment reassembly header and all 1345 * associated datagrams. 1346 */ 1347 static void 1348 ip_freef(struct ipq *fp) 1349 { 1350 struct mbuf *q; 1351 1352 while (fp->ipq_frags) { 1353 q = fp->ipq_frags; 1354 fp->ipq_frags = q->m_nextpkt; 1355 q->m_nextpkt = NULL; 1356 m_freem(q); 1357 } 1358 remque(fp); 1359 mpipe_free(&ipq_mpipe, fp); 1360 nipq--; 1361 } 1362 1363 /* 1364 * IP timer processing; 1365 * if a timer expires on a reassembly 1366 * queue, discard it. 1367 */ 1368 void 1369 ip_slowtimo(void) 1370 { 1371 struct ipq *fp; 1372 int i; 1373 1374 crit_enter(); 1375 for (i = 0; i < IPREASS_NHASH; i++) { 1376 fp = ipq[i].next; 1377 if (fp == NULL) 1378 continue; 1379 while (fp != &ipq[i]) { 1380 --fp->ipq_ttl; 1381 fp = fp->next; 1382 if (fp->prev->ipq_ttl == 0) { 1383 ipstat.ips_fragtimeout += fp->prev->ipq_nfrags; 1384 ip_freef(fp->prev); 1385 } 1386 } 1387 } 1388 /* 1389 * If we are over the maximum number of fragments 1390 * (due to the limit being lowered), drain off 1391 * enough to get down to the new limit. 1392 */ 1393 if (maxnipq >= 0 && nipq > maxnipq) { 1394 for (i = 0; i < IPREASS_NHASH; i++) { 1395 while (nipq > maxnipq && 1396 (ipq[i].next != &ipq[i])) { 1397 ipstat.ips_fragdropped += 1398 ipq[i].next->ipq_nfrags; 1399 ip_freef(ipq[i].next); 1400 } 1401 } 1402 } 1403 ipflow_slowtimo(); 1404 crit_exit(); 1405 } 1406 1407 /* 1408 * Drain off all datagram fragments. 1409 */ 1410 void 1411 ip_drain(void) 1412 { 1413 int i; 1414 1415 for (i = 0; i < IPREASS_NHASH; i++) { 1416 while (ipq[i].next != &ipq[i]) { 1417 ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags; 1418 ip_freef(ipq[i].next); 1419 } 1420 } 1421 in_rtqdrain(); 1422 } 1423 1424 /* 1425 * Do option processing on a datagram, 1426 * possibly discarding it if bad options are encountered, 1427 * or forwarding it if source-routed. 1428 * The pass argument is used when operating in the IPSTEALTH 1429 * mode to tell what options to process: 1430 * [LS]SRR (pass 0) or the others (pass 1). 1431 * The reason for as many as two passes is that when doing IPSTEALTH, 1432 * non-routing options should be processed only if the packet is for us. 1433 * Returns 1 if packet has been forwarded/freed, 1434 * 0 if the packet should be processed further. 1435 */ 1436 static int 1437 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop) 1438 { 1439 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET }; 1440 struct ip *ip = mtod(m, struct ip *); 1441 u_char *cp; 1442 struct in_ifaddr *ia; 1443 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB; 1444 boolean_t forward = FALSE; 1445 struct in_addr *sin, dst; 1446 n_time ntime; 1447 1448 dst = ip->ip_dst; 1449 cp = (u_char *)(ip + 1); 1450 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1451 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1452 opt = cp[IPOPT_OPTVAL]; 1453 if (opt == IPOPT_EOL) 1454 break; 1455 if (opt == IPOPT_NOP) 1456 optlen = 1; 1457 else { 1458 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1459 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1460 goto bad; 1461 } 1462 optlen = cp[IPOPT_OLEN]; 1463 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1464 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1465 goto bad; 1466 } 1467 } 1468 switch (opt) { 1469 1470 default: 1471 break; 1472 1473 /* 1474 * Source routing with record. 1475 * Find interface with current destination address. 1476 * If none on this machine then drop if strictly routed, 1477 * or do nothing if loosely routed. 1478 * Record interface address and bring up next address 1479 * component. If strictly routed make sure next 1480 * address is on directly accessible net. 1481 */ 1482 case IPOPT_LSRR: 1483 case IPOPT_SSRR: 1484 if (ipstealth && pass > 0) 1485 break; 1486 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1487 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1488 goto bad; 1489 } 1490 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1491 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1492 goto bad; 1493 } 1494 ipaddr.sin_addr = ip->ip_dst; 1495 ia = (struct in_ifaddr *) 1496 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1497 if (ia == NULL) { 1498 if (opt == IPOPT_SSRR) { 1499 type = ICMP_UNREACH; 1500 code = ICMP_UNREACH_SRCFAIL; 1501 goto bad; 1502 } 1503 if (!ip_dosourceroute) 1504 goto nosourcerouting; 1505 /* 1506 * Loose routing, and not at next destination 1507 * yet; nothing to do except forward. 1508 */ 1509 break; 1510 } 1511 off--; /* 0 origin */ 1512 if (off > optlen - (int)sizeof(struct in_addr)) { 1513 /* 1514 * End of source route. Should be for us. 1515 */ 1516 if (!ip_acceptsourceroute) 1517 goto nosourcerouting; 1518 save_rte(cp, ip->ip_src); 1519 break; 1520 } 1521 if (ipstealth) 1522 goto dropit; 1523 if (!ip_dosourceroute) { 1524 if (ipforwarding) { 1525 char buf[sizeof "aaa.bbb.ccc.ddd"]; 1526 1527 /* 1528 * Acting as a router, so generate ICMP 1529 */ 1530 nosourcerouting: 1531 strcpy(buf, inet_ntoa(ip->ip_dst)); 1532 log(LOG_WARNING, 1533 "attempted source route from %s to %s\n", 1534 inet_ntoa(ip->ip_src), buf); 1535 type = ICMP_UNREACH; 1536 code = ICMP_UNREACH_SRCFAIL; 1537 goto bad; 1538 } else { 1539 /* 1540 * Not acting as a router, 1541 * so silently drop. 1542 */ 1543 dropit: 1544 ipstat.ips_cantforward++; 1545 m_freem(m); 1546 return (1); 1547 } 1548 } 1549 1550 /* 1551 * locate outgoing interface 1552 */ 1553 memcpy(&ipaddr.sin_addr, cp + off, 1554 sizeof ipaddr.sin_addr); 1555 1556 if (opt == IPOPT_SSRR) { 1557 #define INA struct in_ifaddr * 1558 #define SA struct sockaddr * 1559 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) 1560 == NULL) 1561 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1562 } else 1563 ia = ip_rtaddr(ipaddr.sin_addr, 1564 &ipforward_rt[mycpuid]); 1565 if (ia == NULL) { 1566 type = ICMP_UNREACH; 1567 code = ICMP_UNREACH_SRCFAIL; 1568 goto bad; 1569 } 1570 ip->ip_dst = ipaddr.sin_addr; 1571 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1572 sizeof(struct in_addr)); 1573 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1574 /* 1575 * Let ip_intr's mcast routing check handle mcast pkts 1576 */ 1577 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1578 break; 1579 1580 case IPOPT_RR: 1581 if (ipstealth && pass == 0) 1582 break; 1583 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1584 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1585 goto bad; 1586 } 1587 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1588 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1589 goto bad; 1590 } 1591 /* 1592 * If no space remains, ignore. 1593 */ 1594 off--; /* 0 origin */ 1595 if (off > optlen - (int)sizeof(struct in_addr)) 1596 break; 1597 memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1598 sizeof ipaddr.sin_addr); 1599 /* 1600 * locate outgoing interface; if we're the destination, 1601 * use the incoming interface (should be same). 1602 */ 1603 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL && 1604 (ia = ip_rtaddr(ipaddr.sin_addr, 1605 &ipforward_rt[mycpuid])) 1606 == NULL) { 1607 type = ICMP_UNREACH; 1608 code = ICMP_UNREACH_HOST; 1609 goto bad; 1610 } 1611 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1612 sizeof(struct in_addr)); 1613 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1614 break; 1615 1616 case IPOPT_TS: 1617 if (ipstealth && pass == 0) 1618 break; 1619 code = cp - (u_char *)ip; 1620 if (optlen < 4 || optlen > 40) { 1621 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1622 goto bad; 1623 } 1624 if ((off = cp[IPOPT_OFFSET]) < 5) { 1625 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1626 goto bad; 1627 } 1628 if (off > optlen - (int)sizeof(int32_t)) { 1629 cp[IPOPT_OFFSET + 1] += (1 << 4); 1630 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1631 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1632 goto bad; 1633 } 1634 break; 1635 } 1636 off--; /* 0 origin */ 1637 sin = (struct in_addr *)(cp + off); 1638 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1639 1640 case IPOPT_TS_TSONLY: 1641 break; 1642 1643 case IPOPT_TS_TSANDADDR: 1644 if (off + sizeof(n_time) + 1645 sizeof(struct in_addr) > optlen) { 1646 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1647 goto bad; 1648 } 1649 ipaddr.sin_addr = dst; 1650 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1651 m->m_pkthdr.rcvif); 1652 if (ia == NULL) 1653 continue; 1654 memcpy(sin, &IA_SIN(ia)->sin_addr, 1655 sizeof(struct in_addr)); 1656 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1657 off += sizeof(struct in_addr); 1658 break; 1659 1660 case IPOPT_TS_PRESPEC: 1661 if (off + sizeof(n_time) + 1662 sizeof(struct in_addr) > optlen) { 1663 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1664 goto bad; 1665 } 1666 memcpy(&ipaddr.sin_addr, sin, 1667 sizeof(struct in_addr)); 1668 if (ifa_ifwithaddr((SA)&ipaddr) == NULL) 1669 continue; 1670 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1671 off += sizeof(struct in_addr); 1672 break; 1673 1674 default: 1675 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1676 goto bad; 1677 } 1678 ntime = iptime(); 1679 memcpy(cp + off, &ntime, sizeof(n_time)); 1680 cp[IPOPT_OFFSET] += sizeof(n_time); 1681 } 1682 } 1683 if (forward && ipforwarding) { 1684 ip_forward(m, TRUE, next_hop); 1685 return (1); 1686 } 1687 return (0); 1688 bad: 1689 icmp_error(m, type, code, 0, 0); 1690 ipstat.ips_badoptions++; 1691 return (1); 1692 } 1693 1694 /* 1695 * Given address of next destination (final or next hop), 1696 * return internet address info of interface to be used to get there. 1697 */ 1698 struct in_ifaddr * 1699 ip_rtaddr(struct in_addr dst, struct route *ro) 1700 { 1701 struct sockaddr_in *sin; 1702 1703 sin = (struct sockaddr_in *)&ro->ro_dst; 1704 1705 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) { 1706 if (ro->ro_rt != NULL) { 1707 RTFREE(ro->ro_rt); 1708 ro->ro_rt = NULL; 1709 } 1710 sin->sin_family = AF_INET; 1711 sin->sin_len = sizeof *sin; 1712 sin->sin_addr = dst; 1713 rtalloc_ign(ro, RTF_PRCLONING); 1714 } 1715 1716 if (ro->ro_rt == NULL) 1717 return (NULL); 1718 1719 return (ifatoia(ro->ro_rt->rt_ifa)); 1720 } 1721 1722 /* 1723 * Save incoming source route for use in replies, 1724 * to be picked up later by ip_srcroute if the receiver is interested. 1725 */ 1726 void 1727 save_rte(u_char *option, struct in_addr dst) 1728 { 1729 unsigned olen; 1730 1731 olen = option[IPOPT_OLEN]; 1732 #ifdef DIAGNOSTIC 1733 if (ipprintfs) 1734 kprintf("save_rte: olen %d\n", olen); 1735 #endif 1736 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1737 return; 1738 bcopy(option, ip_srcrt.srcopt, olen); 1739 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1740 ip_srcrt.dst = dst; 1741 } 1742 1743 /* 1744 * Retrieve incoming source route for use in replies, 1745 * in the same form used by setsockopt. 1746 * The first hop is placed before the options, will be removed later. 1747 */ 1748 struct mbuf * 1749 ip_srcroute(void) 1750 { 1751 struct in_addr *p, *q; 1752 struct mbuf *m; 1753 1754 if (ip_nhops == 0) 1755 return (NULL); 1756 m = m_get(MB_DONTWAIT, MT_HEADER); 1757 if (m == NULL) 1758 return (NULL); 1759 1760 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1761 1762 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1763 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1764 OPTSIZ; 1765 #ifdef DIAGNOSTIC 1766 if (ipprintfs) 1767 kprintf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1768 #endif 1769 1770 /* 1771 * First save first hop for return route 1772 */ 1773 p = &ip_srcrt.route[ip_nhops - 1]; 1774 *(mtod(m, struct in_addr *)) = *p--; 1775 #ifdef DIAGNOSTIC 1776 if (ipprintfs) 1777 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1778 #endif 1779 1780 /* 1781 * Copy option fields and padding (nop) to mbuf. 1782 */ 1783 ip_srcrt.nop = IPOPT_NOP; 1784 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1785 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &ip_srcrt.nop, 1786 OPTSIZ); 1787 q = (struct in_addr *)(mtod(m, caddr_t) + 1788 sizeof(struct in_addr) + OPTSIZ); 1789 #undef OPTSIZ 1790 /* 1791 * Record return path as an IP source route, 1792 * reversing the path (pointers are now aligned). 1793 */ 1794 while (p >= ip_srcrt.route) { 1795 #ifdef DIAGNOSTIC 1796 if (ipprintfs) 1797 kprintf(" %x", ntohl(q->s_addr)); 1798 #endif 1799 *q++ = *p--; 1800 } 1801 /* 1802 * Last hop goes to final destination. 1803 */ 1804 *q = ip_srcrt.dst; 1805 #ifdef DIAGNOSTIC 1806 if (ipprintfs) 1807 kprintf(" %x\n", ntohl(q->s_addr)); 1808 #endif 1809 return (m); 1810 } 1811 1812 /* 1813 * Strip out IP options. 1814 */ 1815 void 1816 ip_stripoptions(struct mbuf *m) 1817 { 1818 int datalen; 1819 struct ip *ip = mtod(m, struct ip *); 1820 caddr_t opts; 1821 int optlen; 1822 1823 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1824 opts = (caddr_t)(ip + 1); 1825 datalen = m->m_len - (sizeof(struct ip) + optlen); 1826 bcopy(opts + optlen, opts, datalen); 1827 m->m_len -= optlen; 1828 if (m->m_flags & M_PKTHDR) 1829 m->m_pkthdr.len -= optlen; 1830 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1831 } 1832 1833 u_char inetctlerrmap[PRC_NCMDS] = { 1834 0, 0, 0, 0, 1835 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1836 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1837 EMSGSIZE, EHOSTUNREACH, 0, 0, 1838 0, 0, 0, 0, 1839 ENOPROTOOPT, ECONNREFUSED 1840 }; 1841 1842 /* 1843 * Forward a packet. If some error occurs return the sender 1844 * an icmp packet. Note we can't always generate a meaningful 1845 * icmp message because icmp doesn't have a large enough repertoire 1846 * of codes and types. 1847 * 1848 * If not forwarding, just drop the packet. This could be confusing 1849 * if ipforwarding was zero but some routing protocol was advancing 1850 * us as a gateway to somewhere. However, we must let the routing 1851 * protocol deal with that. 1852 * 1853 * The using_srcrt parameter indicates whether the packet is being forwarded 1854 * via a source route. 1855 */ 1856 static void 1857 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop) 1858 { 1859 struct ip *ip = mtod(m, struct ip *); 1860 struct sockaddr_in *ipforward_rtaddr; 1861 struct rtentry *rt; 1862 int error, type = 0, code = 0, destmtu = 0; 1863 struct mbuf *mcopy; 1864 n_long dest; 1865 struct in_addr pkt_dst; 1866 struct m_hdr tag; 1867 struct route *cache_rt = &ipforward_rt[mycpuid]; 1868 1869 dest = INADDR_ANY; 1870 /* 1871 * Cache the destination address of the packet; this may be 1872 * changed by use of 'ipfw fwd'. 1873 */ 1874 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst; 1875 1876 #ifdef DIAGNOSTIC 1877 if (ipprintfs) 1878 kprintf("forward: src %x dst %x ttl %x\n", 1879 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl); 1880 #endif 1881 1882 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) { 1883 ipstat.ips_cantforward++; 1884 m_freem(m); 1885 return; 1886 } 1887 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) { 1888 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1889 return; 1890 } 1891 1892 ipforward_rtaddr = (struct sockaddr_in *) &cache_rt->ro_dst; 1893 if (cache_rt->ro_rt == NULL || 1894 ipforward_rtaddr->sin_addr.s_addr != pkt_dst.s_addr) { 1895 if (cache_rt->ro_rt != NULL) { 1896 RTFREE(cache_rt->ro_rt); 1897 cache_rt->ro_rt = NULL; 1898 } 1899 ipforward_rtaddr->sin_family = AF_INET; 1900 ipforward_rtaddr->sin_len = sizeof(struct sockaddr_in); 1901 ipforward_rtaddr->sin_addr = pkt_dst; 1902 rtalloc_ign(cache_rt, RTF_PRCLONING); 1903 if (cache_rt->ro_rt == NULL) { 1904 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1905 return; 1906 } 1907 } 1908 rt = cache_rt->ro_rt; 1909 1910 /* 1911 * Save the IP header and at most 8 bytes of the payload, 1912 * in case we need to generate an ICMP message to the src. 1913 * 1914 * XXX this can be optimized a lot by saving the data in a local 1915 * buffer on the stack (72 bytes at most), and only allocating the 1916 * mbuf if really necessary. The vast majority of the packets 1917 * are forwarded without having to send an ICMP back (either 1918 * because unnecessary, or because rate limited), so we are 1919 * really we are wasting a lot of work here. 1920 * 1921 * We don't use m_copy() because it might return a reference 1922 * to a shared cluster. Both this function and ip_output() 1923 * assume exclusive access to the IP header in `m', so any 1924 * data in a cluster may change before we reach icmp_error(). 1925 */ 1926 MGETHDR(mcopy, MB_DONTWAIT, m->m_type); 1927 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, MB_DONTWAIT)) { 1928 /* 1929 * It's probably ok if the pkthdr dup fails (because 1930 * the deep copy of the tag chain failed), but for now 1931 * be conservative and just discard the copy since 1932 * code below may some day want the tags. 1933 */ 1934 m_free(mcopy); 1935 mcopy = NULL; 1936 } 1937 if (mcopy != NULL) { 1938 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8, 1939 (int)ip->ip_len); 1940 mcopy->m_pkthdr.len = mcopy->m_len; 1941 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1942 } 1943 1944 if (!ipstealth) 1945 ip->ip_ttl -= IPTTLDEC; 1946 1947 /* 1948 * If forwarding packet using same interface that it came in on, 1949 * perhaps should send a redirect to sender to shortcut a hop. 1950 * Only send redirect if source is sending directly to us, 1951 * and if packet was not source routed (or has any options). 1952 * Also, don't send redirect if forwarding using a default route 1953 * or a route modified by a redirect. 1954 */ 1955 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1956 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) && 1957 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY && 1958 ipsendredirects && !using_srcrt && next_hop == NULL) { 1959 u_long src = ntohl(ip->ip_src.s_addr); 1960 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa; 1961 1962 if (rt_ifa != NULL && 1963 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) { 1964 if (rt->rt_flags & RTF_GATEWAY) 1965 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1966 else 1967 dest = pkt_dst.s_addr; 1968 /* 1969 * Router requirements says to only send 1970 * host redirects. 1971 */ 1972 type = ICMP_REDIRECT; 1973 code = ICMP_REDIRECT_HOST; 1974 #ifdef DIAGNOSTIC 1975 if (ipprintfs) 1976 kprintf("redirect (%d) to %x\n", code, dest); 1977 #endif 1978 } 1979 } 1980 1981 if (next_hop != NULL) { 1982 /* Pass IPFORWARD info if available */ 1983 tag.mh_type = MT_TAG; 1984 tag.mh_flags = PACKET_TAG_IPFORWARD; 1985 tag.mh_data = (caddr_t)next_hop; 1986 tag.mh_next = m; 1987 m = (struct mbuf *)&tag; 1988 } 1989 1990 error = ip_output(m, NULL, cache_rt, IP_FORWARDING, NULL, 1991 NULL); 1992 if (error == 0) { 1993 ipstat.ips_forward++; 1994 if (type == 0) { 1995 if (mcopy) { 1996 ipflow_create(cache_rt, mcopy); 1997 m_freem(mcopy); 1998 } 1999 return; /* most common case */ 2000 } else { 2001 ipstat.ips_redirectsent++; 2002 } 2003 } else { 2004 ipstat.ips_cantforward++; 2005 } 2006 2007 if (mcopy == NULL) 2008 return; 2009 2010 /* 2011 * Send ICMP message. 2012 */ 2013 2014 switch (error) { 2015 2016 case 0: /* forwarded, but need redirect */ 2017 /* type, code set above */ 2018 break; 2019 2020 case ENETUNREACH: /* shouldn't happen, checked above */ 2021 case EHOSTUNREACH: 2022 case ENETDOWN: 2023 case EHOSTDOWN: 2024 default: 2025 type = ICMP_UNREACH; 2026 code = ICMP_UNREACH_HOST; 2027 break; 2028 2029 case EMSGSIZE: 2030 type = ICMP_UNREACH; 2031 code = ICMP_UNREACH_NEEDFRAG; 2032 #ifdef IPSEC 2033 /* 2034 * If the packet is routed over IPsec tunnel, tell the 2035 * originator the tunnel MTU. 2036 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 2037 * XXX quickhack!!! 2038 */ 2039 if (cache_rt->ro_rt != NULL) { 2040 struct secpolicy *sp = NULL; 2041 int ipsecerror; 2042 int ipsechdr; 2043 struct route *ro; 2044 2045 sp = ipsec4_getpolicybyaddr(mcopy, 2046 IPSEC_DIR_OUTBOUND, 2047 IP_FORWARDING, 2048 &ipsecerror); 2049 2050 if (sp == NULL) 2051 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu; 2052 else { 2053 /* count IPsec header size */ 2054 ipsechdr = ipsec4_hdrsiz(mcopy, 2055 IPSEC_DIR_OUTBOUND, 2056 NULL); 2057 2058 /* 2059 * find the correct route for outer IPv4 2060 * header, compute tunnel MTU. 2061 * 2062 */ 2063 if (sp->req != NULL && sp->req->sav != NULL && 2064 sp->req->sav->sah != NULL) { 2065 ro = &sp->req->sav->sah->sa_route; 2066 if (ro->ro_rt != NULL && 2067 ro->ro_rt->rt_ifp != NULL) { 2068 destmtu = 2069 ro->ro_rt->rt_ifp->if_mtu; 2070 destmtu -= ipsechdr; 2071 } 2072 } 2073 2074 key_freesp(sp); 2075 } 2076 } 2077 #elif FAST_IPSEC 2078 /* 2079 * If the packet is routed over IPsec tunnel, tell the 2080 * originator the tunnel MTU. 2081 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 2082 * XXX quickhack!!! 2083 */ 2084 if (cache_rt->ro_rt != NULL) { 2085 struct secpolicy *sp = NULL; 2086 int ipsecerror; 2087 int ipsechdr; 2088 struct route *ro; 2089 2090 sp = ipsec_getpolicybyaddr(mcopy, 2091 IPSEC_DIR_OUTBOUND, 2092 IP_FORWARDING, 2093 &ipsecerror); 2094 2095 if (sp == NULL) 2096 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu; 2097 else { 2098 /* count IPsec header size */ 2099 ipsechdr = ipsec4_hdrsiz(mcopy, 2100 IPSEC_DIR_OUTBOUND, 2101 NULL); 2102 2103 /* 2104 * find the correct route for outer IPv4 2105 * header, compute tunnel MTU. 2106 */ 2107 2108 if (sp->req != NULL && 2109 sp->req->sav != NULL && 2110 sp->req->sav->sah != NULL) { 2111 ro = &sp->req->sav->sah->sa_route; 2112 if (ro->ro_rt != NULL && 2113 ro->ro_rt->rt_ifp != NULL) { 2114 destmtu = 2115 ro->ro_rt->rt_ifp->if_mtu; 2116 destmtu -= ipsechdr; 2117 } 2118 } 2119 2120 KEY_FREESP(&sp); 2121 } 2122 } 2123 #else /* !IPSEC && !FAST_IPSEC */ 2124 if (cache_rt->ro_rt != NULL) 2125 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu; 2126 #endif /*IPSEC*/ 2127 ipstat.ips_cantfrag++; 2128 break; 2129 2130 case ENOBUFS: 2131 /* 2132 * A router should not generate ICMP_SOURCEQUENCH as 2133 * required in RFC1812 Requirements for IP Version 4 Routers. 2134 * Source quench could be a big problem under DoS attacks, 2135 * or if the underlying interface is rate-limited. 2136 * Those who need source quench packets may re-enable them 2137 * via the net.inet.ip.sendsourcequench sysctl. 2138 */ 2139 if (!ip_sendsourcequench) { 2140 m_freem(mcopy); 2141 return; 2142 } else { 2143 type = ICMP_SOURCEQUENCH; 2144 code = 0; 2145 } 2146 break; 2147 2148 case EACCES: /* ipfw denied packet */ 2149 m_freem(mcopy); 2150 return; 2151 } 2152 icmp_error(mcopy, type, code, dest, destmtu); 2153 } 2154 2155 void 2156 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2157 struct mbuf *m) 2158 { 2159 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2160 struct timeval tv; 2161 2162 microtime(&tv); 2163 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2164 SCM_TIMESTAMP, SOL_SOCKET); 2165 if (*mp) 2166 mp = &(*mp)->m_next; 2167 } 2168 if (inp->inp_flags & INP_RECVDSTADDR) { 2169 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2170 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2171 if (*mp) 2172 mp = &(*mp)->m_next; 2173 } 2174 if (inp->inp_flags & INP_RECVTTL) { 2175 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 2176 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 2177 if (*mp) 2178 mp = &(*mp)->m_next; 2179 } 2180 #ifdef notyet 2181 /* XXX 2182 * Moving these out of udp_input() made them even more broken 2183 * than they already were. 2184 */ 2185 /* options were tossed already */ 2186 if (inp->inp_flags & INP_RECVOPTS) { 2187 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2188 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2189 if (*mp) 2190 mp = &(*mp)->m_next; 2191 } 2192 /* ip_srcroute doesn't do what we want here, need to fix */ 2193 if (inp->inp_flags & INP_RECVRETOPTS) { 2194 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 2195 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2196 if (*mp) 2197 mp = &(*mp)->m_next; 2198 } 2199 #endif 2200 if (inp->inp_flags & INP_RECVIF) { 2201 struct ifnet *ifp; 2202 struct sdlbuf { 2203 struct sockaddr_dl sdl; 2204 u_char pad[32]; 2205 } sdlbuf; 2206 struct sockaddr_dl *sdp; 2207 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2208 2209 if (((ifp = m->m_pkthdr.rcvif)) && 2210 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) { 2211 sdp = IF_LLSOCKADDR(ifp); 2212 /* 2213 * Change our mind and don't try copy. 2214 */ 2215 if ((sdp->sdl_family != AF_LINK) || 2216 (sdp->sdl_len > sizeof(sdlbuf))) { 2217 goto makedummy; 2218 } 2219 bcopy(sdp, sdl2, sdp->sdl_len); 2220 } else { 2221 makedummy: 2222 sdl2->sdl_len = 2223 offsetof(struct sockaddr_dl, sdl_data[0]); 2224 sdl2->sdl_family = AF_LINK; 2225 sdl2->sdl_index = 0; 2226 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2227 } 2228 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 2229 IP_RECVIF, IPPROTO_IP); 2230 if (*mp) 2231 mp = &(*mp)->m_next; 2232 } 2233 } 2234 2235 /* 2236 * XXX these routines are called from the upper part of the kernel. 2237 * 2238 * They could also be moved to ip_mroute.c, since all the RSVP 2239 * handling is done there already. 2240 */ 2241 int 2242 ip_rsvp_init(struct socket *so) 2243 { 2244 if (so->so_type != SOCK_RAW || 2245 so->so_proto->pr_protocol != IPPROTO_RSVP) 2246 return EOPNOTSUPP; 2247 2248 if (ip_rsvpd != NULL) 2249 return EADDRINUSE; 2250 2251 ip_rsvpd = so; 2252 /* 2253 * This may seem silly, but we need to be sure we don't over-increment 2254 * the RSVP counter, in case something slips up. 2255 */ 2256 if (!ip_rsvp_on) { 2257 ip_rsvp_on = 1; 2258 rsvp_on++; 2259 } 2260 2261 return 0; 2262 } 2263 2264 int 2265 ip_rsvp_done(void) 2266 { 2267 ip_rsvpd = NULL; 2268 /* 2269 * This may seem silly, but we need to be sure we don't over-decrement 2270 * the RSVP counter, in case something slips up. 2271 */ 2272 if (ip_rsvp_on) { 2273 ip_rsvp_on = 0; 2274 rsvp_on--; 2275 } 2276 return 0; 2277 } 2278 2279 void 2280 rsvp_input(struct mbuf *m, ...) /* XXX must fixup manually */ 2281 { 2282 int off, proto; 2283 __va_list ap; 2284 2285 __va_start(ap, m); 2286 off = __va_arg(ap, int); 2287 proto = __va_arg(ap, int); 2288 __va_end(ap); 2289 2290 if (rsvp_input_p) { /* call the real one if loaded */ 2291 rsvp_input_p(m, off, proto); 2292 return; 2293 } 2294 2295 /* Can still get packets with rsvp_on = 0 if there is a local member 2296 * of the group to which the RSVP packet is addressed. But in this 2297 * case we want to throw the packet away. 2298 */ 2299 2300 if (!rsvp_on) { 2301 m_freem(m); 2302 return; 2303 } 2304 2305 if (ip_rsvpd != NULL) { 2306 rip_input(m, off, proto); 2307 return; 2308 } 2309 /* Drop the packet */ 2310 m_freem(m); 2311 } 2312