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 1.84 2008/07/28 13:45:43 sephe 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 <machine/stdarg.h> 98 99 #include <net/if.h> 100 #include <net/if_types.h> 101 #include <net/if_var.h> 102 #include <net/if_dl.h> 103 #include <net/pfil.h> 104 #include <net/route.h> 105 #include <net/netisr.h> 106 107 #include <netinet/in.h> 108 #include <netinet/in_systm.h> 109 #include <netinet/in_var.h> 110 #include <netinet/ip.h> 111 #include <netinet/in_pcb.h> 112 #include <netinet/ip_var.h> 113 #include <netinet/ip_icmp.h> 114 115 #include <sys/thread2.h> 116 #include <sys/msgport2.h> 117 #include <net/netmsg2.h> 118 119 #include <sys/socketvar.h> 120 121 #include <net/ipfw/ip_fw.h> 122 #include <net/dummynet/ip_dummynet.h> 123 124 #ifdef IPSEC 125 #include <netinet6/ipsec.h> 126 #include <netproto/key/key.h> 127 #endif 128 129 #ifdef FAST_IPSEC 130 #include <netproto/ipsec/ipsec.h> 131 #include <netproto/ipsec/key.h> 132 #endif 133 134 int rsvp_on = 0; 135 static int ip_rsvp_on; 136 struct socket *ip_rsvpd; 137 138 int ipforwarding = 0; 139 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 140 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 141 142 static int ipsendredirects = 1; /* XXX */ 143 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 144 &ipsendredirects, 0, "Enable sending IP redirects"); 145 146 int ip_defttl = IPDEFTTL; 147 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 148 &ip_defttl, 0, "Maximum TTL on IP packets"); 149 150 static int ip_dosourceroute = 0; 151 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 152 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 153 154 static int ip_acceptsourceroute = 0; 155 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 156 CTLFLAG_RW, &ip_acceptsourceroute, 0, 157 "Enable accepting source routed IP packets"); 158 159 static int ip_keepfaith = 0; 160 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 161 &ip_keepfaith, 0, 162 "Enable packet capture for FAITH IPv4->IPv6 translator daemon"); 163 164 static int nipq = 0; /* total # of reass queues */ 165 static int maxnipq; 166 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 167 &maxnipq, 0, 168 "Maximum number of IPv4 fragment reassembly queue entries"); 169 170 static int maxfragsperpacket; 171 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 172 &maxfragsperpacket, 0, 173 "Maximum number of IPv4 fragments allowed per packet"); 174 175 static int ip_sendsourcequench = 0; 176 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 177 &ip_sendsourcequench, 0, 178 "Enable the transmission of source quench packets"); 179 180 int ip_do_randomid = 1; 181 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 182 &ip_do_randomid, 0, 183 "Assign random ip_id values"); 184 /* 185 * XXX - Setting ip_checkinterface mostly implements the receive side of 186 * the Strong ES model described in RFC 1122, but since the routing table 187 * and transmit implementation do not implement the Strong ES model, 188 * setting this to 1 results in an odd hybrid. 189 * 190 * XXX - ip_checkinterface currently must be disabled if you use ipnat 191 * to translate the destination address to another local interface. 192 * 193 * XXX - ip_checkinterface must be disabled if you add IP aliases 194 * to the loopback interface instead of the interface where the 195 * packets for those addresses are received. 196 */ 197 static int ip_checkinterface = 0; 198 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 199 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 200 201 #ifdef DIAGNOSTIC 202 static int ipprintfs = 0; 203 #endif 204 205 extern struct domain inetdomain; 206 extern struct protosw inetsw[]; 207 u_char ip_protox[IPPROTO_MAX]; 208 struct in_ifaddrhead in_ifaddrheads[MAXCPU]; /* first inet address */ 209 struct in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU]; 210 /* inet addr hash table */ 211 u_long in_ifaddrhmask; /* mask for hash table */ 212 213 struct ip_stats ipstats_percpu[MAXCPU]; 214 #ifdef SMP 215 static int 216 sysctl_ipstats(SYSCTL_HANDLER_ARGS) 217 { 218 int cpu, error = 0; 219 220 for (cpu = 0; cpu < ncpus; ++cpu) { 221 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu], 222 sizeof(struct ip_stats)))) 223 break; 224 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu], 225 sizeof(struct ip_stats)))) 226 break; 227 } 228 229 return (error); 230 } 231 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW), 232 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics"); 233 #else 234 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 235 &ipstat, ip_stats, "IP statistics"); 236 #endif 237 238 /* Packet reassembly stuff */ 239 #define IPREASS_NHASH_LOG2 6 240 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 241 #define IPREASS_HMASK (IPREASS_NHASH - 1) 242 #define IPREASS_HASH(x,y) \ 243 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 244 245 static struct ipq ipq[IPREASS_NHASH]; 246 247 #ifdef IPCTL_DEFMTU 248 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 249 &ip_mtu, 0, "Default MTU"); 250 #endif 251 252 #ifdef IPSTEALTH 253 static int ipstealth = 0; 254 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, ""); 255 #else 256 static const int ipstealth = 0; 257 #endif 258 259 260 /* Firewall hooks */ 261 ip_fw_chk_t *ip_fw_chk_ptr; 262 ip_fw_dn_io_t *ip_fw_dn_io_ptr; 263 int ip_fw_loaded; 264 int fw_enable = 1; 265 int fw_one_pass = 1; 266 267 struct pfil_head inet_pfil_hook; 268 269 /* 270 * XXX this is ugly -- the following two global variables are 271 * used to store packet state while it travels through the stack. 272 * Note that the code even makes assumptions on the size and 273 * alignment of fields inside struct ip_srcrt so e.g. adding some 274 * fields will break the code. This needs to be fixed. 275 * 276 * We need to save the IP options in case a protocol wants to respond 277 * to an incoming packet over the same route if the packet got here 278 * using IP source routing. This allows connection establishment and 279 * maintenance when the remote end is on a network that is not known 280 * to us. 281 */ 282 static int ip_nhops = 0; 283 284 static struct ip_srcrt { 285 struct in_addr dst; /* final destination */ 286 char nop; /* one NOP to align */ 287 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 288 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 289 } ip_srcrt; 290 291 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management"); 292 static struct malloc_pipe ipq_mpipe; 293 294 static void save_rte (u_char *, struct in_addr); 295 static int ip_dooptions (struct mbuf *m, int, 296 struct sockaddr_in *next_hop); 297 static void ip_freef (struct ipq *); 298 static void ip_input_handler (struct netmsg *); 299 static struct mbuf *ip_reass (struct mbuf *, struct ipq *, 300 struct ipq *, u_int32_t *); 301 302 /* 303 * IP initialization: fill in IP protocol switch table. 304 * All protocols not implemented in kernel go to raw IP protocol handler. 305 */ 306 void 307 ip_init(void) 308 { 309 struct protosw *pr; 310 int i; 311 #ifdef SMP 312 int cpu; 313 #endif 314 315 /* 316 * Make sure we can handle a reasonable number of fragments but 317 * cap it at 4000 (XXX). 318 */ 319 mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq), 320 IFQ_MAXLEN, 4000, 0, NULL); 321 for (i = 0; i < ncpus; ++i) { 322 TAILQ_INIT(&in_ifaddrheads[i]); 323 in_ifaddrhashtbls[i] = 324 hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 325 } 326 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 327 if (pr == NULL) 328 panic("ip_init"); 329 for (i = 0; i < IPPROTO_MAX; i++) 330 ip_protox[i] = pr - inetsw; 331 for (pr = inetdomain.dom_protosw; 332 pr < inetdomain.dom_protoswNPROTOSW; pr++) 333 if (pr->pr_domain->dom_family == PF_INET && 334 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 335 ip_protox[pr->pr_protocol] = pr - inetsw; 336 337 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 338 inet_pfil_hook.ph_af = AF_INET; 339 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) { 340 kprintf("%s: WARNING: unable to register pfil hook, " 341 "error %d\n", __func__, i); 342 } 343 344 for (i = 0; i < IPREASS_NHASH; i++) 345 ipq[i].next = ipq[i].prev = &ipq[i]; 346 347 maxnipq = nmbclusters / 32; 348 maxfragsperpacket = 16; 349 350 ip_id = time_second & 0xffff; 351 352 /* 353 * Initialize IP statistics counters for each CPU. 354 * 355 */ 356 #ifdef SMP 357 for (cpu = 0; cpu < ncpus; ++cpu) { 358 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats)); 359 } 360 #else 361 bzero(&ipstat, sizeof(struct ip_stats)); 362 #endif 363 364 netisr_register(NETISR_IP, ip_mport_in, ip_input_handler); 365 } 366 367 /* 368 * XXX watch out this one. It is perhaps used as a cache for 369 * the most recently used route ? it is cleared in in_addroute() 370 * when a new route is successfully created. 371 */ 372 struct route ipforward_rt[MAXCPU]; 373 374 /* Do transport protocol processing. */ 375 static void 376 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip, 377 struct sockaddr_in *nexthop) 378 { 379 /* 380 * Switch out to protocol's input routine. 381 */ 382 if (nexthop && ip->ip_p == IPPROTO_TCP) { 383 /* TCP needs IPFORWARD info if available */ 384 struct m_hdr tag; 385 386 tag.mh_type = MT_TAG; 387 tag.mh_flags = PACKET_TAG_IPFORWARD; 388 tag.mh_data = (caddr_t)nexthop; 389 tag.mh_next = m; 390 391 (*inetsw[ip_protox[ip->ip_p]].pr_input) 392 ((struct mbuf *)&tag, hlen, ip->ip_p); 393 } else { 394 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, ip->ip_p); 395 } 396 } 397 398 struct netmsg_transport_packet { 399 struct netmsg nm_netmsg; 400 struct mbuf *nm_mbuf; 401 int nm_hlen; 402 boolean_t nm_hasnexthop; 403 struct sockaddr_in nm_nexthop; 404 }; 405 406 static void 407 transport_processing_handler(netmsg_t netmsg) 408 { 409 struct netmsg_transport_packet *msg = (void *)netmsg; 410 struct sockaddr_in *nexthop; 411 struct ip *ip; 412 413 ip = mtod(msg->nm_mbuf, struct ip *); 414 nexthop = msg->nm_hasnexthop ? &msg->nm_nexthop : NULL; 415 transport_processing_oncpu(msg->nm_mbuf, msg->nm_hlen, ip, nexthop); 416 lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 0); 417 } 418 419 static void 420 ip_input_handler(struct netmsg *msg0) 421 { 422 struct mbuf *m = ((struct netmsg_packet *)msg0)->nm_packet; 423 424 ip_input(m); 425 /* msg0 was embedded in the mbuf, do not reply! */ 426 } 427 428 /* 429 * IP input routine. Checksum and byte swap header. If fragmented 430 * try to reassemble. Process options. Pass to next level. 431 */ 432 void 433 ip_input(struct mbuf *m) 434 { 435 struct ip *ip; 436 struct ipq *fp; 437 struct in_ifaddr *ia = NULL; 438 struct in_ifaddr_container *iac; 439 int i, hlen, checkif; 440 u_short sum; 441 struct in_addr pkt_dst; 442 u_int32_t divert_info = 0; /* packet divert/tee info */ 443 struct ip_fw_args args; 444 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */ 445 boolean_t needredispatch = FALSE; 446 struct in_addr odst; /* original dst address(NAT) */ 447 struct m_tag *mtag; 448 #ifdef FAST_IPSEC 449 struct tdb_ident *tdbi; 450 struct secpolicy *sp; 451 int error; 452 #endif 453 454 args.eh = NULL; 455 args.oif = NULL; 456 args.rule = NULL; 457 args.next_hop = NULL; 458 459 /* Grab info from MT_TAG mbufs prepended to the chain. */ 460 while (m != NULL && m->m_type == MT_TAG) { 461 switch(m->_m_tag_id) { 462 case PACKET_TAG_IPFORWARD: 463 args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data; 464 break; 465 default: 466 kprintf("ip_input: unrecognised MT_TAG tag %d\n", 467 m->_m_tag_id); 468 break; 469 } 470 m = m->m_next; 471 } 472 M_ASSERTPKTHDR(m); 473 474 /* Extract info from dummynet tag */ 475 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 476 if (mtag != NULL) { 477 args.rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 478 479 m_tag_delete(m, mtag); 480 mtag = NULL; 481 } 482 483 if (args.rule != NULL) { /* dummynet already filtered us */ 484 ip = mtod(m, struct ip *); 485 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 486 goto iphack; 487 } 488 489 ipstat.ips_total++; 490 491 /* length checks already done in ip_demux() */ 492 KASSERT(m->m_len >= sizeof(ip), ("IP header not in one mbuf")); 493 494 ip = mtod(m, struct ip *); 495 496 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 497 ipstat.ips_badvers++; 498 goto bad; 499 } 500 501 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 502 /* length checks already done in ip_demux() */ 503 KASSERT(hlen >= sizeof(struct ip), ("IP header len too small")); 504 KASSERT(m->m_len >= hlen, ("packet shorter than IP header length")); 505 506 /* 127/8 must not appear on wire - RFC1122 */ 507 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 508 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 509 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) { 510 ipstat.ips_badaddr++; 511 goto bad; 512 } 513 } 514 515 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 516 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 517 } else { 518 if (hlen == sizeof(struct ip)) { 519 sum = in_cksum_hdr(ip); 520 } else { 521 sum = in_cksum(m, hlen); 522 } 523 } 524 if (sum != 0) { 525 ipstat.ips_badsum++; 526 goto bad; 527 } 528 529 #ifdef ALTQ 530 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 531 /* packet is dropped by traffic conditioner */ 532 return; 533 } 534 #endif 535 /* 536 * Convert fields to host representation. 537 */ 538 ip->ip_len = ntohs(ip->ip_len); 539 if (ip->ip_len < hlen) { 540 ipstat.ips_badlen++; 541 goto bad; 542 } 543 ip->ip_off = ntohs(ip->ip_off); 544 545 /* 546 * Check that the amount of data in the buffers 547 * is as at least much as the IP header would have us expect. 548 * Trim mbufs if longer than we expect. 549 * Drop packet if shorter than we expect. 550 */ 551 if (m->m_pkthdr.len < ip->ip_len) { 552 ipstat.ips_tooshort++; 553 goto bad; 554 } 555 if (m->m_pkthdr.len > ip->ip_len) { 556 if (m->m_len == m->m_pkthdr.len) { 557 m->m_len = ip->ip_len; 558 m->m_pkthdr.len = ip->ip_len; 559 } else 560 m_adj(m, ip->ip_len - m->m_pkthdr.len); 561 } 562 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF) 563 /* 564 * Bypass packet filtering for packets from a tunnel (gif). 565 */ 566 if (ipsec_gethist(m, NULL)) 567 goto pass; 568 #endif 569 570 /* 571 * IpHack's section. 572 * Right now when no processing on packet has done 573 * and it is still fresh out of network we do our black 574 * deals with it. 575 * - Firewall: deny/allow/divert 576 * - Xlate: translate packet's addr/port (NAT). 577 * - Pipe: pass pkt through dummynet. 578 * - Wrap: fake packet's addr/port <unimpl.> 579 * - Encapsulate: put it in another IP and send out. <unimp.> 580 */ 581 582 iphack: 583 584 /* 585 * Run through list of hooks for input packets. 586 * 587 * NB: Beware of the destination address changing (e.g. 588 * by NAT rewriting). When this happens, tell 589 * ip_forward to do the right thing. 590 */ 591 if (pfil_has_hooks(&inet_pfil_hook)) { 592 odst = ip->ip_dst; 593 if (pfil_run_hooks(&inet_pfil_hook, &m, 594 m->m_pkthdr.rcvif, PFIL_IN)) { 595 return; 596 } 597 if (m == NULL) /* consumed by filter */ 598 return; 599 ip = mtod(m, struct ip *); 600 using_srcrt = (odst.s_addr != ip->ip_dst.s_addr); 601 } 602 603 if (fw_enable && IPFW_LOADED) { 604 /* 605 * If we've been forwarded from the output side, then 606 * skip the firewall a second time 607 */ 608 if (args.next_hop != NULL) 609 goto ours; 610 611 args.m = m; 612 i = ip_fw_chk_ptr(&args); 613 m = args.m; 614 615 if ((i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */ 616 if (m != NULL) 617 m_freem(m); 618 return; 619 } 620 ip = mtod(m, struct ip *); /* just in case m changed */ 621 if (i == 0 && args.next_hop == NULL) /* common case */ 622 goto pass; 623 if (i & IP_FW_PORT_DYNT_FLAG) { 624 /* Send packet to the appropriate pipe */ 625 ip_fw_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args); 626 return; 627 } 628 #ifdef IPDIVERT 629 if (i != 0 && !(i & IP_FW_PORT_DYNT_FLAG)) { 630 /* Divert or tee packet */ 631 divert_info = i; 632 goto ours; 633 } 634 #endif 635 if (i == 0 && args.next_hop != NULL) 636 goto pass; 637 /* 638 * if we get here, the packet must be dropped 639 */ 640 m_freem(m); 641 return; 642 } 643 pass: 644 645 /* 646 * Process options and, if not destined for us, 647 * ship it on. ip_dooptions returns 1 when an 648 * error was detected (causing an icmp message 649 * to be sent and the original packet to be freed). 650 */ 651 ip_nhops = 0; /* for source routed packets */ 652 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, args.next_hop)) 653 return; 654 655 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 656 * matter if it is destined to another node, or whether it is 657 * a multicast one, RSVP wants it! and prevents it from being forwarded 658 * anywhere else. Also checks if the rsvp daemon is running before 659 * grabbing the packet. 660 */ 661 if (rsvp_on && ip->ip_p == IPPROTO_RSVP) 662 goto ours; 663 664 /* 665 * Check our list of addresses, to see if the packet is for us. 666 * If we don't have any addresses, assume any unicast packet 667 * we receive might be for us (and let the upper layers deal 668 * with it). 669 */ 670 if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) && 671 !(m->m_flags & (M_MCAST | M_BCAST))) 672 goto ours; 673 674 /* 675 * Cache the destination address of the packet; this may be 676 * changed by use of 'ipfw fwd'. 677 */ 678 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst; 679 680 /* 681 * Enable a consistency check between the destination address 682 * and the arrival interface for a unicast packet (the RFC 1122 683 * strong ES model) if IP forwarding is disabled and the packet 684 * is not locally generated and the packet is not subject to 685 * 'ipfw fwd'. 686 * 687 * XXX - Checking also should be disabled if the destination 688 * address is ipnat'ed to a different interface. 689 * 690 * XXX - Checking is incompatible with IP aliases added 691 * to the loopback interface instead of the interface where 692 * the packets are received. 693 */ 694 checkif = ip_checkinterface && 695 !ipforwarding && 696 m->m_pkthdr.rcvif != NULL && 697 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) && 698 (args.next_hop == NULL); 699 700 /* 701 * Check for exact addresses in the hash bucket. 702 */ 703 LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) { 704 ia = iac->ia; 705 706 /* 707 * If the address matches, verify that the packet 708 * arrived via the correct interface if checking is 709 * enabled. 710 */ 711 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 712 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 713 goto ours; 714 } 715 ia = NULL; 716 717 /* 718 * Check for broadcast addresses. 719 * 720 * Only accept broadcast packets that arrive via the matching 721 * interface. Reception of forwarded directed broadcasts would 722 * be handled via ip_forward() and ether_output() with the loopback 723 * into the stack for SIMPLEX interfaces handled by ether_output(). 724 */ 725 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 726 struct ifaddr_container *ifac; 727 728 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid], 729 ifa_link) { 730 struct ifaddr *ifa = ifac->ifa; 731 732 if (ifa->ifa_addr == NULL) /* shutdown/startup race */ 733 continue; 734 if (ifa->ifa_addr->sa_family != AF_INET) 735 continue; 736 ia = ifatoia(ifa); 737 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 738 pkt_dst.s_addr) 739 goto ours; 740 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 741 goto ours; 742 #ifdef BOOTP_COMPAT 743 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 744 goto ours; 745 #endif 746 } 747 } 748 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 749 struct in_multi *inm; 750 751 if (ip_mrouter != NULL) { 752 /* 753 * If we are acting as a multicast router, all 754 * incoming multicast packets are passed to the 755 * kernel-level multicast forwarding function. 756 * The packet is returned (relatively) intact; if 757 * ip_mforward() returns a non-zero value, the packet 758 * must be discarded, else it may be accepted below. 759 */ 760 if (ip_mforward != NULL && 761 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) { 762 ipstat.ips_cantforward++; 763 m_freem(m); 764 return; 765 } 766 767 /* 768 * The process-level routing daemon needs to receive 769 * all multicast IGMP packets, whether or not this 770 * host belongs to their destination groups. 771 */ 772 if (ip->ip_p == IPPROTO_IGMP) 773 goto ours; 774 ipstat.ips_forward++; 775 } 776 /* 777 * See if we belong to the destination multicast group on the 778 * arrival interface. 779 */ 780 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 781 if (inm == NULL) { 782 ipstat.ips_notmember++; 783 m_freem(m); 784 return; 785 } 786 goto ours; 787 } 788 if (ip->ip_dst.s_addr == INADDR_BROADCAST) 789 goto ours; 790 if (ip->ip_dst.s_addr == INADDR_ANY) 791 goto ours; 792 793 /* 794 * FAITH(Firewall Aided Internet Translator) 795 */ 796 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 797 if (ip_keepfaith) { 798 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 799 goto ours; 800 } 801 m_freem(m); 802 return; 803 } 804 805 /* 806 * Not for us; forward if possible and desirable. 807 */ 808 if (!ipforwarding) { 809 ipstat.ips_cantforward++; 810 m_freem(m); 811 } else { 812 #ifdef IPSEC 813 /* 814 * Enforce inbound IPsec SPD. 815 */ 816 if (ipsec4_in_reject(m, NULL)) { 817 ipsecstat.in_polvio++; 818 goto bad; 819 } 820 #endif 821 #ifdef FAST_IPSEC 822 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 823 crit_enter(); 824 if (mtag != NULL) { 825 tdbi = (struct tdb_ident *)m_tag_data(mtag); 826 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 827 } else { 828 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 829 IP_FORWARDING, &error); 830 } 831 if (sp == NULL) { /* NB: can happen if error */ 832 crit_exit(); 833 /*XXX error stat???*/ 834 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 835 goto bad; 836 } 837 838 /* 839 * Check security policy against packet attributes. 840 */ 841 error = ipsec_in_reject(sp, m); 842 KEY_FREESP(&sp); 843 crit_exit(); 844 if (error) { 845 ipstat.ips_cantforward++; 846 goto bad; 847 } 848 #endif 849 ip_forward(m, using_srcrt, args.next_hop); 850 } 851 return; 852 853 ours: 854 855 /* 856 * IPSTEALTH: Process non-routing options only 857 * if the packet is destined for us. 858 */ 859 if (ipstealth && 860 hlen > sizeof(struct ip) && 861 ip_dooptions(m, 1, args.next_hop)) 862 return; 863 864 /* Count the packet in the ip address stats */ 865 if (ia != NULL) { 866 ia->ia_ifa.if_ipackets++; 867 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 868 } 869 870 /* 871 * If offset or IP_MF are set, must reassemble. 872 * Otherwise, nothing need be done. 873 * (We could look in the reassembly queue to see 874 * if the packet was previously fragmented, 875 * but it's not worth the time; just let them time out.) 876 */ 877 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 878 879 /* If maxnipq is 0, never accept fragments. */ 880 if (maxnipq == 0) { 881 ipstat.ips_fragments++; 882 ipstat.ips_fragdropped++; 883 goto bad; 884 } 885 886 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 887 /* 888 * Look for queue of fragments 889 * of this datagram. 890 */ 891 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) 892 if (ip->ip_id == fp->ipq_id && 893 ip->ip_src.s_addr == fp->ipq_src.s_addr && 894 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 895 ip->ip_p == fp->ipq_p) 896 goto found; 897 898 fp = NULL; 899 900 /* 901 * Enforce upper bound on number of fragmented packets 902 * for which we attempt reassembly; 903 * If maxnipq is -1, accept all fragments without limitation. 904 */ 905 if ((nipq > maxnipq) && (maxnipq > 0)) { 906 /* 907 * drop something from the tail of the current queue 908 * before proceeding further 909 */ 910 if (ipq[sum].prev == &ipq[sum]) { /* gak */ 911 for (i = 0; i < IPREASS_NHASH; i++) { 912 if (ipq[i].prev != &ipq[i]) { 913 ipstat.ips_fragtimeout += 914 ipq[i].prev->ipq_nfrags; 915 ip_freef(ipq[i].prev); 916 break; 917 } 918 } 919 } else { 920 ipstat.ips_fragtimeout += 921 ipq[sum].prev->ipq_nfrags; 922 ip_freef(ipq[sum].prev); 923 } 924 } 925 found: 926 /* 927 * Adjust ip_len to not reflect header, 928 * convert offset of this to bytes. 929 */ 930 ip->ip_len -= hlen; 931 if (ip->ip_off & IP_MF) { 932 /* 933 * Make sure that fragments have a data length 934 * that's a non-zero multiple of 8 bytes. 935 */ 936 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 937 ipstat.ips_toosmall++; /* XXX */ 938 goto bad; 939 } 940 m->m_flags |= M_FRAG; 941 } else 942 m->m_flags &= ~M_FRAG; 943 ip->ip_off <<= 3; 944 945 /* 946 * Attempt reassembly; if it succeeds, proceed. 947 * ip_reass() will return a different mbuf, and update 948 * the divert info in divert_info. 949 */ 950 ipstat.ips_fragments++; 951 m->m_pkthdr.header = ip; 952 m = ip_reass(m, fp, &ipq[sum], &divert_info); 953 if (m == NULL) 954 return; 955 ipstat.ips_reassembled++; 956 needredispatch = TRUE; 957 ip = mtod(m, struct ip *); 958 /* Get the header length of the reassembled packet */ 959 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 960 #ifdef IPDIVERT 961 /* Restore original checksum before diverting packet */ 962 if (divert_info != 0) { 963 ip->ip_len += hlen; 964 ip->ip_len = htons(ip->ip_len); 965 ip->ip_off = htons(ip->ip_off); 966 ip->ip_sum = 0; 967 if (hlen == sizeof(struct ip)) 968 ip->ip_sum = in_cksum_hdr(ip); 969 else 970 ip->ip_sum = in_cksum(m, hlen); 971 ip->ip_off = ntohs(ip->ip_off); 972 ip->ip_len = ntohs(ip->ip_len); 973 ip->ip_len -= hlen; 974 } 975 #endif 976 } else { 977 ip->ip_len -= hlen; 978 } 979 980 #ifdef IPDIVERT 981 /* 982 * Divert or tee packet to the divert protocol if required. 983 */ 984 if (divert_info != 0) { 985 struct mbuf *clone = NULL; 986 987 /* Clone packet if we're doing a 'tee' */ 988 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 989 clone = m_dup(m, MB_DONTWAIT); 990 991 /* Restore packet header fields to original values */ 992 ip->ip_len += hlen; 993 ip->ip_len = htons(ip->ip_len); 994 ip->ip_off = htons(ip->ip_off); 995 996 /* Deliver packet to divert input routine */ 997 divert_packet(m, 1, divert_info & 0xffff); 998 ipstat.ips_delivered++; 999 1000 /* If 'tee', continue with original packet */ 1001 if (clone == NULL) 1002 return; 1003 m = clone; 1004 ip = mtod(m, struct ip *); 1005 ip->ip_len += hlen; 1006 /* 1007 * Jump backwards to complete processing of the 1008 * packet. But first clear divert_info to avoid 1009 * entering this block again. 1010 * We do not need to clear args.divert_rule 1011 * or args.next_hop as they will not be used. 1012 * 1013 * XXX Better safe than sorry, remove the DIVERT tag. 1014 */ 1015 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL); 1016 if (mtag != NULL) 1017 m_tag_delete(m, mtag); 1018 1019 divert_info = 0; 1020 goto pass; 1021 } 1022 #endif 1023 1024 #ifdef IPSEC 1025 /* 1026 * enforce IPsec policy checking if we are seeing last header. 1027 * note that we do not visit this with protocols with pcb layer 1028 * code - like udp/tcp/raw ip. 1029 */ 1030 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) && 1031 ipsec4_in_reject(m, NULL)) { 1032 ipsecstat.in_polvio++; 1033 goto bad; 1034 } 1035 #endif 1036 #if FAST_IPSEC 1037 /* 1038 * enforce IPsec policy checking if we are seeing last header. 1039 * note that we do not visit this with protocols with pcb layer 1040 * code - like udp/tcp/raw ip. 1041 */ 1042 if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) { 1043 /* 1044 * Check if the packet has already had IPsec processing 1045 * done. If so, then just pass it along. This tag gets 1046 * set during AH, ESP, etc. input handling, before the 1047 * packet is returned to the ip input queue for delivery. 1048 */ 1049 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 1050 crit_enter(); 1051 if (mtag != NULL) { 1052 tdbi = (struct tdb_ident *)m_tag_data(mtag); 1053 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 1054 } else { 1055 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 1056 IP_FORWARDING, &error); 1057 } 1058 if (sp != NULL) { 1059 /* 1060 * Check security policy against packet attributes. 1061 */ 1062 error = ipsec_in_reject(sp, m); 1063 KEY_FREESP(&sp); 1064 } else { 1065 /* XXX error stat??? */ 1066 error = EINVAL; 1067 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 1068 goto bad; 1069 } 1070 crit_exit(); 1071 if (error) 1072 goto bad; 1073 } 1074 #endif /* FAST_IPSEC */ 1075 1076 ipstat.ips_delivered++; 1077 if (needredispatch) { 1078 struct netmsg_transport_packet *msg; 1079 lwkt_port_t port; 1080 1081 ip->ip_off = htons(ip->ip_off); 1082 ip->ip_len = htons(ip->ip_len); 1083 port = ip_mport_in(&m); 1084 if (port == NULL) 1085 return; 1086 1087 msg = kmalloc(sizeof(struct netmsg_transport_packet), M_LWKTMSG, 1088 M_INTWAIT | M_NULLOK); 1089 if (msg == NULL) 1090 goto bad; 1091 1092 netmsg_init(&msg->nm_netmsg, &netisr_afree_rport, 0, 1093 transport_processing_handler); 1094 msg->nm_hlen = hlen; 1095 msg->nm_hasnexthop = (args.next_hop != NULL); 1096 if (msg->nm_hasnexthop) 1097 msg->nm_nexthop = *args.next_hop; /* structure copy */ 1098 1099 msg->nm_mbuf = m; 1100 ip = mtod(m, struct ip *); 1101 ip->ip_len = ntohs(ip->ip_len); 1102 ip->ip_off = ntohs(ip->ip_off); 1103 lwkt_sendmsg(port, &msg->nm_netmsg.nm_lmsg); 1104 } else { 1105 transport_processing_oncpu(m, hlen, ip, args.next_hop); 1106 } 1107 return; 1108 1109 bad: 1110 m_freem(m); 1111 } 1112 1113 /* 1114 * Take incoming datagram fragment and try to reassemble it into 1115 * whole datagram. If a chain for reassembly of this datagram already 1116 * exists, then it is given as fp; otherwise have to make a chain. 1117 * 1118 * When IPDIVERT enabled, keep additional state with each packet that 1119 * tells us if we need to divert or tee the packet we're building. 1120 * In particular, *divinfo includes the port and TEE flag. 1121 */ 1122 1123 static struct mbuf * 1124 ip_reass(struct mbuf *m, struct ipq *fp, struct ipq *where, 1125 u_int32_t *divinfo) 1126 { 1127 struct ip *ip = mtod(m, struct ip *); 1128 struct mbuf *p = NULL, *q, *nq; 1129 struct mbuf *n; 1130 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 1131 int i, next; 1132 #ifdef IPDIVERT 1133 struct m_tag *mtag; 1134 #endif 1135 1136 /* 1137 * If the hardware has not done csum over this fragment 1138 * then csum_data is not valid at all. 1139 */ 1140 if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) 1141 == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) { 1142 m->m_pkthdr.csum_data = 0; 1143 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 1144 } 1145 1146 /* 1147 * Presence of header sizes in mbufs 1148 * would confuse code below. 1149 */ 1150 m->m_data += hlen; 1151 m->m_len -= hlen; 1152 1153 /* 1154 * If first fragment to arrive, create a reassembly queue. 1155 */ 1156 if (fp == NULL) { 1157 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL) 1158 goto dropfrag; 1159 insque(fp, where); 1160 nipq++; 1161 fp->ipq_nfrags = 1; 1162 fp->ipq_ttl = IPFRAGTTL; 1163 fp->ipq_p = ip->ip_p; 1164 fp->ipq_id = ip->ip_id; 1165 fp->ipq_src = ip->ip_src; 1166 fp->ipq_dst = ip->ip_dst; 1167 fp->ipq_frags = m; 1168 m->m_nextpkt = NULL; 1169 #ifdef IPDIVERT 1170 fp->ipq_div_info = 0; 1171 #endif 1172 goto inserted; 1173 } else { 1174 fp->ipq_nfrags++; 1175 } 1176 1177 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1178 1179 /* 1180 * Find a segment which begins after this one does. 1181 */ 1182 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1183 if (GETIP(q)->ip_off > ip->ip_off) 1184 break; 1185 1186 /* 1187 * If there is a preceding segment, it may provide some of 1188 * our data already. If so, drop the data from the incoming 1189 * segment. If it provides all of our data, drop us, otherwise 1190 * stick new segment in the proper place. 1191 * 1192 * If some of the data is dropped from the the preceding 1193 * segment, then it's checksum is invalidated. 1194 */ 1195 if (p) { 1196 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1197 if (i > 0) { 1198 if (i >= ip->ip_len) 1199 goto dropfrag; 1200 m_adj(m, i); 1201 m->m_pkthdr.csum_flags = 0; 1202 ip->ip_off += i; 1203 ip->ip_len -= i; 1204 } 1205 m->m_nextpkt = p->m_nextpkt; 1206 p->m_nextpkt = m; 1207 } else { 1208 m->m_nextpkt = fp->ipq_frags; 1209 fp->ipq_frags = m; 1210 } 1211 1212 /* 1213 * While we overlap succeeding segments trim them or, 1214 * if they are completely covered, dequeue them. 1215 */ 1216 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1217 q = nq) { 1218 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1219 if (i < GETIP(q)->ip_len) { 1220 GETIP(q)->ip_len -= i; 1221 GETIP(q)->ip_off += i; 1222 m_adj(q, i); 1223 q->m_pkthdr.csum_flags = 0; 1224 break; 1225 } 1226 nq = q->m_nextpkt; 1227 m->m_nextpkt = nq; 1228 ipstat.ips_fragdropped++; 1229 fp->ipq_nfrags--; 1230 q->m_nextpkt = NULL; 1231 m_freem(q); 1232 } 1233 1234 inserted: 1235 1236 #ifdef IPDIVERT 1237 /* 1238 * Transfer firewall instructions to the fragment structure. 1239 * Only trust info in the fragment at offset 0. 1240 */ 1241 if (ip->ip_off == 0) { 1242 fp->ipq_div_info = *divinfo; 1243 } else { 1244 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL); 1245 if (mtag != NULL) 1246 m_tag_delete(m, mtag); 1247 } 1248 *divinfo = 0; 1249 #endif 1250 1251 /* 1252 * Check for complete reassembly and perform frag per packet 1253 * limiting. 1254 * 1255 * Frag limiting is performed here so that the nth frag has 1256 * a chance to complete the packet before we drop the packet. 1257 * As a result, n+1 frags are actually allowed per packet, but 1258 * only n will ever be stored. (n = maxfragsperpacket.) 1259 * 1260 */ 1261 next = 0; 1262 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1263 if (GETIP(q)->ip_off != next) { 1264 if (fp->ipq_nfrags > maxfragsperpacket) { 1265 ipstat.ips_fragdropped += fp->ipq_nfrags; 1266 ip_freef(fp); 1267 } 1268 return (NULL); 1269 } 1270 next += GETIP(q)->ip_len; 1271 } 1272 /* Make sure the last packet didn't have the IP_MF flag */ 1273 if (p->m_flags & M_FRAG) { 1274 if (fp->ipq_nfrags > maxfragsperpacket) { 1275 ipstat.ips_fragdropped += fp->ipq_nfrags; 1276 ip_freef(fp); 1277 } 1278 return (NULL); 1279 } 1280 1281 /* 1282 * Reassembly is complete. Make sure the packet is a sane size. 1283 */ 1284 q = fp->ipq_frags; 1285 ip = GETIP(q); 1286 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 1287 ipstat.ips_toolong++; 1288 ipstat.ips_fragdropped += fp->ipq_nfrags; 1289 ip_freef(fp); 1290 return (NULL); 1291 } 1292 1293 /* 1294 * Concatenate fragments. 1295 */ 1296 m = q; 1297 n = m->m_next; 1298 m->m_next = NULL; 1299 m_cat(m, n); 1300 nq = q->m_nextpkt; 1301 q->m_nextpkt = NULL; 1302 for (q = nq; q != NULL; q = nq) { 1303 nq = q->m_nextpkt; 1304 q->m_nextpkt = NULL; 1305 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1306 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1307 m_cat(m, q); 1308 } 1309 1310 /* 1311 * Clean up the 1's complement checksum. Carry over 16 bits must 1312 * be added back. This assumes no more then 65535 packet fragments 1313 * were reassembled. A second carry can also occur (but not a third). 1314 */ 1315 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) + 1316 (m->m_pkthdr.csum_data >> 16); 1317 if (m->m_pkthdr.csum_data > 0xFFFF) 1318 m->m_pkthdr.csum_data -= 0xFFFF; 1319 1320 1321 #ifdef IPDIVERT 1322 /* 1323 * Extract firewall instructions from the fragment structure. 1324 */ 1325 *divinfo = fp->ipq_div_info; 1326 #endif 1327 1328 /* 1329 * Create header for new ip packet by 1330 * modifying header of first packet; 1331 * dequeue and discard fragment reassembly header. 1332 * Make header visible. 1333 */ 1334 ip->ip_len = next; 1335 ip->ip_src = fp->ipq_src; 1336 ip->ip_dst = fp->ipq_dst; 1337 remque(fp); 1338 nipq--; 1339 mpipe_free(&ipq_mpipe, fp); 1340 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 1341 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 1342 /* some debugging cruft by sklower, below, will go away soon */ 1343 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1344 int plen = 0; 1345 1346 for (n = m; n; n = n->m_next) 1347 plen += n->m_len; 1348 m->m_pkthdr.len = plen; 1349 } 1350 return (m); 1351 1352 dropfrag: 1353 #ifdef IPDIVERT 1354 *divinfo = 0; 1355 #endif 1356 ipstat.ips_fragdropped++; 1357 if (fp != NULL) 1358 fp->ipq_nfrags--; 1359 m_freem(m); 1360 return (NULL); 1361 1362 #undef GETIP 1363 } 1364 1365 /* 1366 * Free a fragment reassembly header and all 1367 * associated datagrams. 1368 */ 1369 static void 1370 ip_freef(struct ipq *fp) 1371 { 1372 struct mbuf *q; 1373 1374 while (fp->ipq_frags) { 1375 q = fp->ipq_frags; 1376 fp->ipq_frags = q->m_nextpkt; 1377 q->m_nextpkt = NULL; 1378 m_freem(q); 1379 } 1380 remque(fp); 1381 mpipe_free(&ipq_mpipe, fp); 1382 nipq--; 1383 } 1384 1385 /* 1386 * IP timer processing; 1387 * if a timer expires on a reassembly 1388 * queue, discard it. 1389 */ 1390 void 1391 ip_slowtimo(void) 1392 { 1393 struct ipq *fp; 1394 int i; 1395 1396 crit_enter(); 1397 for (i = 0; i < IPREASS_NHASH; i++) { 1398 fp = ipq[i].next; 1399 if (fp == NULL) 1400 continue; 1401 while (fp != &ipq[i]) { 1402 --fp->ipq_ttl; 1403 fp = fp->next; 1404 if (fp->prev->ipq_ttl == 0) { 1405 ipstat.ips_fragtimeout += fp->prev->ipq_nfrags; 1406 ip_freef(fp->prev); 1407 } 1408 } 1409 } 1410 /* 1411 * If we are over the maximum number of fragments 1412 * (due to the limit being lowered), drain off 1413 * enough to get down to the new limit. 1414 */ 1415 if (maxnipq >= 0 && nipq > maxnipq) { 1416 for (i = 0; i < IPREASS_NHASH; i++) { 1417 while (nipq > maxnipq && 1418 (ipq[i].next != &ipq[i])) { 1419 ipstat.ips_fragdropped += 1420 ipq[i].next->ipq_nfrags; 1421 ip_freef(ipq[i].next); 1422 } 1423 } 1424 } 1425 ipflow_slowtimo(); 1426 crit_exit(); 1427 } 1428 1429 /* 1430 * Drain off all datagram fragments. 1431 */ 1432 void 1433 ip_drain(void) 1434 { 1435 int i; 1436 1437 for (i = 0; i < IPREASS_NHASH; i++) { 1438 while (ipq[i].next != &ipq[i]) { 1439 ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags; 1440 ip_freef(ipq[i].next); 1441 } 1442 } 1443 in_rtqdrain(); 1444 } 1445 1446 /* 1447 * Do option processing on a datagram, 1448 * possibly discarding it if bad options are encountered, 1449 * or forwarding it if source-routed. 1450 * The pass argument is used when operating in the IPSTEALTH 1451 * mode to tell what options to process: 1452 * [LS]SRR (pass 0) or the others (pass 1). 1453 * The reason for as many as two passes is that when doing IPSTEALTH, 1454 * non-routing options should be processed only if the packet is for us. 1455 * Returns 1 if packet has been forwarded/freed, 1456 * 0 if the packet should be processed further. 1457 */ 1458 static int 1459 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop) 1460 { 1461 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET }; 1462 struct ip *ip = mtod(m, struct ip *); 1463 u_char *cp; 1464 struct in_ifaddr *ia; 1465 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB; 1466 boolean_t forward = FALSE; 1467 struct in_addr *sin, dst; 1468 n_time ntime; 1469 1470 dst = ip->ip_dst; 1471 cp = (u_char *)(ip + 1); 1472 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1473 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1474 opt = cp[IPOPT_OPTVAL]; 1475 if (opt == IPOPT_EOL) 1476 break; 1477 if (opt == IPOPT_NOP) 1478 optlen = 1; 1479 else { 1480 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1481 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1482 goto bad; 1483 } 1484 optlen = cp[IPOPT_OLEN]; 1485 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1486 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1487 goto bad; 1488 } 1489 } 1490 switch (opt) { 1491 1492 default: 1493 break; 1494 1495 /* 1496 * Source routing with record. 1497 * Find interface with current destination address. 1498 * If none on this machine then drop if strictly routed, 1499 * or do nothing if loosely routed. 1500 * Record interface address and bring up next address 1501 * component. If strictly routed make sure next 1502 * address is on directly accessible net. 1503 */ 1504 case IPOPT_LSRR: 1505 case IPOPT_SSRR: 1506 if (ipstealth && pass > 0) 1507 break; 1508 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1509 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1510 goto bad; 1511 } 1512 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1513 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1514 goto bad; 1515 } 1516 ipaddr.sin_addr = ip->ip_dst; 1517 ia = (struct in_ifaddr *) 1518 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1519 if (ia == NULL) { 1520 if (opt == IPOPT_SSRR) { 1521 type = ICMP_UNREACH; 1522 code = ICMP_UNREACH_SRCFAIL; 1523 goto bad; 1524 } 1525 if (!ip_dosourceroute) 1526 goto nosourcerouting; 1527 /* 1528 * Loose routing, and not at next destination 1529 * yet; nothing to do except forward. 1530 */ 1531 break; 1532 } 1533 off--; /* 0 origin */ 1534 if (off > optlen - (int)sizeof(struct in_addr)) { 1535 /* 1536 * End of source route. Should be for us. 1537 */ 1538 if (!ip_acceptsourceroute) 1539 goto nosourcerouting; 1540 save_rte(cp, ip->ip_src); 1541 break; 1542 } 1543 if (ipstealth) 1544 goto dropit; 1545 if (!ip_dosourceroute) { 1546 if (ipforwarding) { 1547 char buf[sizeof "aaa.bbb.ccc.ddd"]; 1548 1549 /* 1550 * Acting as a router, so generate ICMP 1551 */ 1552 nosourcerouting: 1553 strcpy(buf, inet_ntoa(ip->ip_dst)); 1554 log(LOG_WARNING, 1555 "attempted source route from %s to %s\n", 1556 inet_ntoa(ip->ip_src), buf); 1557 type = ICMP_UNREACH; 1558 code = ICMP_UNREACH_SRCFAIL; 1559 goto bad; 1560 } else { 1561 /* 1562 * Not acting as a router, 1563 * so silently drop. 1564 */ 1565 dropit: 1566 ipstat.ips_cantforward++; 1567 m_freem(m); 1568 return (1); 1569 } 1570 } 1571 1572 /* 1573 * locate outgoing interface 1574 */ 1575 memcpy(&ipaddr.sin_addr, cp + off, 1576 sizeof ipaddr.sin_addr); 1577 1578 if (opt == IPOPT_SSRR) { 1579 #define INA struct in_ifaddr * 1580 #define SA struct sockaddr * 1581 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) 1582 == NULL) 1583 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1584 } else 1585 ia = ip_rtaddr(ipaddr.sin_addr, 1586 &ipforward_rt[mycpuid]); 1587 if (ia == NULL) { 1588 type = ICMP_UNREACH; 1589 code = ICMP_UNREACH_SRCFAIL; 1590 goto bad; 1591 } 1592 ip->ip_dst = ipaddr.sin_addr; 1593 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1594 sizeof(struct in_addr)); 1595 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1596 /* 1597 * Let ip_intr's mcast routing check handle mcast pkts 1598 */ 1599 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1600 break; 1601 1602 case IPOPT_RR: 1603 if (ipstealth && pass == 0) 1604 break; 1605 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1606 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1607 goto bad; 1608 } 1609 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1610 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1611 goto bad; 1612 } 1613 /* 1614 * If no space remains, ignore. 1615 */ 1616 off--; /* 0 origin */ 1617 if (off > optlen - (int)sizeof(struct in_addr)) 1618 break; 1619 memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1620 sizeof ipaddr.sin_addr); 1621 /* 1622 * locate outgoing interface; if we're the destination, 1623 * use the incoming interface (should be same). 1624 */ 1625 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL && 1626 (ia = ip_rtaddr(ipaddr.sin_addr, 1627 &ipforward_rt[mycpuid])) 1628 == NULL) { 1629 type = ICMP_UNREACH; 1630 code = ICMP_UNREACH_HOST; 1631 goto bad; 1632 } 1633 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1634 sizeof(struct in_addr)); 1635 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1636 break; 1637 1638 case IPOPT_TS: 1639 if (ipstealth && pass == 0) 1640 break; 1641 code = cp - (u_char *)ip; 1642 if (optlen < 4 || optlen > 40) { 1643 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1644 goto bad; 1645 } 1646 if ((off = cp[IPOPT_OFFSET]) < 5) { 1647 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1648 goto bad; 1649 } 1650 if (off > optlen - (int)sizeof(int32_t)) { 1651 cp[IPOPT_OFFSET + 1] += (1 << 4); 1652 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1653 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1654 goto bad; 1655 } 1656 break; 1657 } 1658 off--; /* 0 origin */ 1659 sin = (struct in_addr *)(cp + off); 1660 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1661 1662 case IPOPT_TS_TSONLY: 1663 break; 1664 1665 case IPOPT_TS_TSANDADDR: 1666 if (off + sizeof(n_time) + 1667 sizeof(struct in_addr) > optlen) { 1668 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1669 goto bad; 1670 } 1671 ipaddr.sin_addr = dst; 1672 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1673 m->m_pkthdr.rcvif); 1674 if (ia == NULL) 1675 continue; 1676 memcpy(sin, &IA_SIN(ia)->sin_addr, 1677 sizeof(struct in_addr)); 1678 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1679 off += sizeof(struct in_addr); 1680 break; 1681 1682 case IPOPT_TS_PRESPEC: 1683 if (off + sizeof(n_time) + 1684 sizeof(struct in_addr) > optlen) { 1685 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1686 goto bad; 1687 } 1688 memcpy(&ipaddr.sin_addr, sin, 1689 sizeof(struct in_addr)); 1690 if (ifa_ifwithaddr((SA)&ipaddr) == NULL) 1691 continue; 1692 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1693 off += sizeof(struct in_addr); 1694 break; 1695 1696 default: 1697 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1698 goto bad; 1699 } 1700 ntime = iptime(); 1701 memcpy(cp + off, &ntime, sizeof(n_time)); 1702 cp[IPOPT_OFFSET] += sizeof(n_time); 1703 } 1704 } 1705 if (forward && ipforwarding) { 1706 ip_forward(m, TRUE, next_hop); 1707 return (1); 1708 } 1709 return (0); 1710 bad: 1711 icmp_error(m, type, code, 0, 0); 1712 ipstat.ips_badoptions++; 1713 return (1); 1714 } 1715 1716 /* 1717 * Given address of next destination (final or next hop), 1718 * return internet address info of interface to be used to get there. 1719 */ 1720 struct in_ifaddr * 1721 ip_rtaddr(struct in_addr dst, struct route *ro) 1722 { 1723 struct sockaddr_in *sin; 1724 1725 sin = (struct sockaddr_in *)&ro->ro_dst; 1726 1727 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) { 1728 if (ro->ro_rt != NULL) { 1729 RTFREE(ro->ro_rt); 1730 ro->ro_rt = NULL; 1731 } 1732 sin->sin_family = AF_INET; 1733 sin->sin_len = sizeof *sin; 1734 sin->sin_addr = dst; 1735 rtalloc_ign(ro, RTF_PRCLONING); 1736 } 1737 1738 if (ro->ro_rt == NULL) 1739 return (NULL); 1740 1741 return (ifatoia(ro->ro_rt->rt_ifa)); 1742 } 1743 1744 /* 1745 * Save incoming source route for use in replies, 1746 * to be picked up later by ip_srcroute if the receiver is interested. 1747 */ 1748 void 1749 save_rte(u_char *option, struct in_addr dst) 1750 { 1751 unsigned olen; 1752 1753 olen = option[IPOPT_OLEN]; 1754 #ifdef DIAGNOSTIC 1755 if (ipprintfs) 1756 kprintf("save_rte: olen %d\n", olen); 1757 #endif 1758 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1759 return; 1760 bcopy(option, ip_srcrt.srcopt, olen); 1761 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1762 ip_srcrt.dst = dst; 1763 } 1764 1765 /* 1766 * Retrieve incoming source route for use in replies, 1767 * in the same form used by setsockopt. 1768 * The first hop is placed before the options, will be removed later. 1769 */ 1770 struct mbuf * 1771 ip_srcroute(void) 1772 { 1773 struct in_addr *p, *q; 1774 struct mbuf *m; 1775 1776 if (ip_nhops == 0) 1777 return (NULL); 1778 m = m_get(MB_DONTWAIT, MT_HEADER); 1779 if (m == NULL) 1780 return (NULL); 1781 1782 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1783 1784 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1785 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1786 OPTSIZ; 1787 #ifdef DIAGNOSTIC 1788 if (ipprintfs) 1789 kprintf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1790 #endif 1791 1792 /* 1793 * First save first hop for return route 1794 */ 1795 p = &ip_srcrt.route[ip_nhops - 1]; 1796 *(mtod(m, struct in_addr *)) = *p--; 1797 #ifdef DIAGNOSTIC 1798 if (ipprintfs) 1799 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1800 #endif 1801 1802 /* 1803 * Copy option fields and padding (nop) to mbuf. 1804 */ 1805 ip_srcrt.nop = IPOPT_NOP; 1806 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1807 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &ip_srcrt.nop, 1808 OPTSIZ); 1809 q = (struct in_addr *)(mtod(m, caddr_t) + 1810 sizeof(struct in_addr) + OPTSIZ); 1811 #undef OPTSIZ 1812 /* 1813 * Record return path as an IP source route, 1814 * reversing the path (pointers are now aligned). 1815 */ 1816 while (p >= ip_srcrt.route) { 1817 #ifdef DIAGNOSTIC 1818 if (ipprintfs) 1819 kprintf(" %x", ntohl(q->s_addr)); 1820 #endif 1821 *q++ = *p--; 1822 } 1823 /* 1824 * Last hop goes to final destination. 1825 */ 1826 *q = ip_srcrt.dst; 1827 #ifdef DIAGNOSTIC 1828 if (ipprintfs) 1829 kprintf(" %x\n", ntohl(q->s_addr)); 1830 #endif 1831 return (m); 1832 } 1833 1834 /* 1835 * Strip out IP options. 1836 */ 1837 void 1838 ip_stripoptions(struct mbuf *m) 1839 { 1840 int datalen; 1841 struct ip *ip = mtod(m, struct ip *); 1842 caddr_t opts; 1843 int optlen; 1844 1845 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1846 opts = (caddr_t)(ip + 1); 1847 datalen = m->m_len - (sizeof(struct ip) + optlen); 1848 bcopy(opts + optlen, opts, datalen); 1849 m->m_len -= optlen; 1850 if (m->m_flags & M_PKTHDR) 1851 m->m_pkthdr.len -= optlen; 1852 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1853 } 1854 1855 u_char inetctlerrmap[PRC_NCMDS] = { 1856 0, 0, 0, 0, 1857 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1858 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1859 EMSGSIZE, EHOSTUNREACH, 0, 0, 1860 0, 0, 0, 0, 1861 ENOPROTOOPT, ECONNREFUSED 1862 }; 1863 1864 /* 1865 * Forward a packet. If some error occurs return the sender 1866 * an icmp packet. Note we can't always generate a meaningful 1867 * icmp message because icmp doesn't have a large enough repertoire 1868 * of codes and types. 1869 * 1870 * If not forwarding, just drop the packet. This could be confusing 1871 * if ipforwarding was zero but some routing protocol was advancing 1872 * us as a gateway to somewhere. However, we must let the routing 1873 * protocol deal with that. 1874 * 1875 * The using_srcrt parameter indicates whether the packet is being forwarded 1876 * via a source route. 1877 */ 1878 void 1879 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop) 1880 { 1881 struct ip *ip = mtod(m, struct ip *); 1882 struct sockaddr_in *ipforward_rtaddr; 1883 struct rtentry *rt; 1884 int error, type = 0, code = 0, destmtu = 0; 1885 struct mbuf *mcopy; 1886 n_long dest; 1887 struct in_addr pkt_dst; 1888 struct m_hdr tag; 1889 struct route *cache_rt = &ipforward_rt[mycpuid]; 1890 1891 dest = INADDR_ANY; 1892 /* 1893 * Cache the destination address of the packet; this may be 1894 * changed by use of 'ipfw fwd'. 1895 */ 1896 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst; 1897 1898 #ifdef DIAGNOSTIC 1899 if (ipprintfs) 1900 kprintf("forward: src %x dst %x ttl %x\n", 1901 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl); 1902 #endif 1903 1904 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) { 1905 ipstat.ips_cantforward++; 1906 m_freem(m); 1907 return; 1908 } 1909 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) { 1910 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1911 return; 1912 } 1913 1914 ipforward_rtaddr = (struct sockaddr_in *) &cache_rt->ro_dst; 1915 if (cache_rt->ro_rt == NULL || 1916 ipforward_rtaddr->sin_addr.s_addr != pkt_dst.s_addr) { 1917 if (cache_rt->ro_rt != NULL) { 1918 RTFREE(cache_rt->ro_rt); 1919 cache_rt->ro_rt = NULL; 1920 } 1921 ipforward_rtaddr->sin_family = AF_INET; 1922 ipforward_rtaddr->sin_len = sizeof(struct sockaddr_in); 1923 ipforward_rtaddr->sin_addr = pkt_dst; 1924 rtalloc_ign(cache_rt, RTF_PRCLONING); 1925 if (cache_rt->ro_rt == NULL) { 1926 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1927 return; 1928 } 1929 } 1930 rt = cache_rt->ro_rt; 1931 1932 /* 1933 * Save the IP header and at most 8 bytes of the payload, 1934 * in case we need to generate an ICMP message to the src. 1935 * 1936 * XXX this can be optimized a lot by saving the data in a local 1937 * buffer on the stack (72 bytes at most), and only allocating the 1938 * mbuf if really necessary. The vast majority of the packets 1939 * are forwarded without having to send an ICMP back (either 1940 * because unnecessary, or because rate limited), so we are 1941 * really we are wasting a lot of work here. 1942 * 1943 * We don't use m_copy() because it might return a reference 1944 * to a shared cluster. Both this function and ip_output() 1945 * assume exclusive access to the IP header in `m', so any 1946 * data in a cluster may change before we reach icmp_error(). 1947 */ 1948 MGETHDR(mcopy, MB_DONTWAIT, m->m_type); 1949 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, MB_DONTWAIT)) { 1950 /* 1951 * It's probably ok if the pkthdr dup fails (because 1952 * the deep copy of the tag chain failed), but for now 1953 * be conservative and just discard the copy since 1954 * code below may some day want the tags. 1955 */ 1956 m_free(mcopy); 1957 mcopy = NULL; 1958 } 1959 if (mcopy != NULL) { 1960 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8, 1961 (int)ip->ip_len); 1962 mcopy->m_pkthdr.len = mcopy->m_len; 1963 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1964 } 1965 1966 if (!ipstealth) 1967 ip->ip_ttl -= IPTTLDEC; 1968 1969 /* 1970 * If forwarding packet using same interface that it came in on, 1971 * perhaps should send a redirect to sender to shortcut a hop. 1972 * Only send redirect if source is sending directly to us, 1973 * and if packet was not source routed (or has any options). 1974 * Also, don't send redirect if forwarding using a default route 1975 * or a route modified by a redirect. 1976 */ 1977 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1978 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) && 1979 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY && 1980 ipsendredirects && !using_srcrt && next_hop == NULL) { 1981 u_long src = ntohl(ip->ip_src.s_addr); 1982 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa; 1983 1984 if (rt_ifa != NULL && 1985 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) { 1986 if (rt->rt_flags & RTF_GATEWAY) 1987 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1988 else 1989 dest = pkt_dst.s_addr; 1990 /* 1991 * Router requirements says to only send 1992 * host redirects. 1993 */ 1994 type = ICMP_REDIRECT; 1995 code = ICMP_REDIRECT_HOST; 1996 #ifdef DIAGNOSTIC 1997 if (ipprintfs) 1998 kprintf("redirect (%d) to %x\n", code, dest); 1999 #endif 2000 } 2001 } 2002 2003 if (next_hop != NULL) { 2004 /* Pass IPFORWARD info if available */ 2005 tag.mh_type = MT_TAG; 2006 tag.mh_flags = PACKET_TAG_IPFORWARD; 2007 tag.mh_data = (caddr_t)next_hop; 2008 tag.mh_next = m; 2009 m = (struct mbuf *)&tag; 2010 } 2011 2012 error = ip_output(m, NULL, cache_rt, IP_FORWARDING, NULL, 2013 NULL); 2014 if (error == 0) { 2015 ipstat.ips_forward++; 2016 if (type == 0) { 2017 if (mcopy) { 2018 ipflow_create(cache_rt, mcopy); 2019 m_freem(mcopy); 2020 } 2021 return; /* most common case */ 2022 } else { 2023 ipstat.ips_redirectsent++; 2024 } 2025 } else { 2026 ipstat.ips_cantforward++; 2027 } 2028 2029 if (mcopy == NULL) 2030 return; 2031 2032 /* 2033 * Send ICMP message. 2034 */ 2035 2036 switch (error) { 2037 2038 case 0: /* forwarded, but need redirect */ 2039 /* type, code set above */ 2040 break; 2041 2042 case ENETUNREACH: /* shouldn't happen, checked above */ 2043 case EHOSTUNREACH: 2044 case ENETDOWN: 2045 case EHOSTDOWN: 2046 default: 2047 type = ICMP_UNREACH; 2048 code = ICMP_UNREACH_HOST; 2049 break; 2050 2051 case EMSGSIZE: 2052 type = ICMP_UNREACH; 2053 code = ICMP_UNREACH_NEEDFRAG; 2054 #ifdef IPSEC 2055 /* 2056 * If the packet is routed over IPsec tunnel, tell the 2057 * originator the tunnel MTU. 2058 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 2059 * XXX quickhack!!! 2060 */ 2061 if (cache_rt->ro_rt != NULL) { 2062 struct secpolicy *sp = NULL; 2063 int ipsecerror; 2064 int ipsechdr; 2065 struct route *ro; 2066 2067 sp = ipsec4_getpolicybyaddr(mcopy, 2068 IPSEC_DIR_OUTBOUND, 2069 IP_FORWARDING, 2070 &ipsecerror); 2071 2072 if (sp == NULL) 2073 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu; 2074 else { 2075 /* count IPsec header size */ 2076 ipsechdr = ipsec4_hdrsiz(mcopy, 2077 IPSEC_DIR_OUTBOUND, 2078 NULL); 2079 2080 /* 2081 * find the correct route for outer IPv4 2082 * header, compute tunnel MTU. 2083 * 2084 */ 2085 if (sp->req != NULL && sp->req->sav != NULL && 2086 sp->req->sav->sah != NULL) { 2087 ro = &sp->req->sav->sah->sa_route; 2088 if (ro->ro_rt != NULL && 2089 ro->ro_rt->rt_ifp != NULL) { 2090 destmtu = 2091 ro->ro_rt->rt_ifp->if_mtu; 2092 destmtu -= ipsechdr; 2093 } 2094 } 2095 2096 key_freesp(sp); 2097 } 2098 } 2099 #elif FAST_IPSEC 2100 /* 2101 * If the packet is routed over IPsec tunnel, tell the 2102 * originator the tunnel MTU. 2103 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 2104 * XXX quickhack!!! 2105 */ 2106 if (cache_rt->ro_rt != NULL) { 2107 struct secpolicy *sp = NULL; 2108 int ipsecerror; 2109 int ipsechdr; 2110 struct route *ro; 2111 2112 sp = ipsec_getpolicybyaddr(mcopy, 2113 IPSEC_DIR_OUTBOUND, 2114 IP_FORWARDING, 2115 &ipsecerror); 2116 2117 if (sp == NULL) 2118 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu; 2119 else { 2120 /* count IPsec header size */ 2121 ipsechdr = ipsec4_hdrsiz(mcopy, 2122 IPSEC_DIR_OUTBOUND, 2123 NULL); 2124 2125 /* 2126 * find the correct route for outer IPv4 2127 * header, compute tunnel MTU. 2128 */ 2129 2130 if (sp->req != NULL && 2131 sp->req->sav != NULL && 2132 sp->req->sav->sah != NULL) { 2133 ro = &sp->req->sav->sah->sa_route; 2134 if (ro->ro_rt != NULL && 2135 ro->ro_rt->rt_ifp != NULL) { 2136 destmtu = 2137 ro->ro_rt->rt_ifp->if_mtu; 2138 destmtu -= ipsechdr; 2139 } 2140 } 2141 2142 KEY_FREESP(&sp); 2143 } 2144 } 2145 #else /* !IPSEC && !FAST_IPSEC */ 2146 if (cache_rt->ro_rt != NULL) 2147 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu; 2148 #endif /*IPSEC*/ 2149 ipstat.ips_cantfrag++; 2150 break; 2151 2152 case ENOBUFS: 2153 /* 2154 * A router should not generate ICMP_SOURCEQUENCH as 2155 * required in RFC1812 Requirements for IP Version 4 Routers. 2156 * Source quench could be a big problem under DoS attacks, 2157 * or if the underlying interface is rate-limited. 2158 * Those who need source quench packets may re-enable them 2159 * via the net.inet.ip.sendsourcequench sysctl. 2160 */ 2161 if (!ip_sendsourcequench) { 2162 m_freem(mcopy); 2163 return; 2164 } else { 2165 type = ICMP_SOURCEQUENCH; 2166 code = 0; 2167 } 2168 break; 2169 2170 case EACCES: /* ipfw denied packet */ 2171 m_freem(mcopy); 2172 return; 2173 } 2174 icmp_error(mcopy, type, code, dest, destmtu); 2175 } 2176 2177 void 2178 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2179 struct mbuf *m) 2180 { 2181 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2182 struct timeval tv; 2183 2184 microtime(&tv); 2185 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2186 SCM_TIMESTAMP, SOL_SOCKET); 2187 if (*mp) 2188 mp = &(*mp)->m_next; 2189 } 2190 if (inp->inp_flags & INP_RECVDSTADDR) { 2191 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2192 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2193 if (*mp) 2194 mp = &(*mp)->m_next; 2195 } 2196 if (inp->inp_flags & INP_RECVTTL) { 2197 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 2198 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 2199 if (*mp) 2200 mp = &(*mp)->m_next; 2201 } 2202 #ifdef notyet 2203 /* XXX 2204 * Moving these out of udp_input() made them even more broken 2205 * than they already were. 2206 */ 2207 /* options were tossed already */ 2208 if (inp->inp_flags & INP_RECVOPTS) { 2209 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2210 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2211 if (*mp) 2212 mp = &(*mp)->m_next; 2213 } 2214 /* ip_srcroute doesn't do what we want here, need to fix */ 2215 if (inp->inp_flags & INP_RECVRETOPTS) { 2216 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 2217 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2218 if (*mp) 2219 mp = &(*mp)->m_next; 2220 } 2221 #endif 2222 if (inp->inp_flags & INP_RECVIF) { 2223 struct ifnet *ifp; 2224 struct sdlbuf { 2225 struct sockaddr_dl sdl; 2226 u_char pad[32]; 2227 } sdlbuf; 2228 struct sockaddr_dl *sdp; 2229 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2230 2231 if (((ifp = m->m_pkthdr.rcvif)) && 2232 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) { 2233 sdp = IF_LLSOCKADDR(ifp); 2234 /* 2235 * Change our mind and don't try copy. 2236 */ 2237 if ((sdp->sdl_family != AF_LINK) || 2238 (sdp->sdl_len > sizeof(sdlbuf))) { 2239 goto makedummy; 2240 } 2241 bcopy(sdp, sdl2, sdp->sdl_len); 2242 } else { 2243 makedummy: 2244 sdl2->sdl_len = 2245 offsetof(struct sockaddr_dl, sdl_data[0]); 2246 sdl2->sdl_family = AF_LINK; 2247 sdl2->sdl_index = 0; 2248 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2249 } 2250 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 2251 IP_RECVIF, IPPROTO_IP); 2252 if (*mp) 2253 mp = &(*mp)->m_next; 2254 } 2255 } 2256 2257 /* 2258 * XXX these routines are called from the upper part of the kernel. 2259 * 2260 * They could also be moved to ip_mroute.c, since all the RSVP 2261 * handling is done there already. 2262 */ 2263 int 2264 ip_rsvp_init(struct socket *so) 2265 { 2266 if (so->so_type != SOCK_RAW || 2267 so->so_proto->pr_protocol != IPPROTO_RSVP) 2268 return EOPNOTSUPP; 2269 2270 if (ip_rsvpd != NULL) 2271 return EADDRINUSE; 2272 2273 ip_rsvpd = so; 2274 /* 2275 * This may seem silly, but we need to be sure we don't over-increment 2276 * the RSVP counter, in case something slips up. 2277 */ 2278 if (!ip_rsvp_on) { 2279 ip_rsvp_on = 1; 2280 rsvp_on++; 2281 } 2282 2283 return 0; 2284 } 2285 2286 int 2287 ip_rsvp_done(void) 2288 { 2289 ip_rsvpd = NULL; 2290 /* 2291 * This may seem silly, but we need to be sure we don't over-decrement 2292 * the RSVP counter, in case something slips up. 2293 */ 2294 if (ip_rsvp_on) { 2295 ip_rsvp_on = 0; 2296 rsvp_on--; 2297 } 2298 return 0; 2299 } 2300 2301 void 2302 rsvp_input(struct mbuf *m, ...) /* XXX must fixup manually */ 2303 { 2304 int off, proto; 2305 __va_list ap; 2306 2307 __va_start(ap, m); 2308 off = __va_arg(ap, int); 2309 proto = __va_arg(ap, int); 2310 __va_end(ap); 2311 2312 if (rsvp_input_p) { /* call the real one if loaded */ 2313 rsvp_input_p(m, off, proto); 2314 return; 2315 } 2316 2317 /* Can still get packets with rsvp_on = 0 if there is a local member 2318 * of the group to which the RSVP packet is addressed. But in this 2319 * case we want to throw the packet away. 2320 */ 2321 2322 if (!rsvp_on) { 2323 m_freem(m); 2324 return; 2325 } 2326 2327 if (ip_rsvpd != NULL) { 2328 rip_input(m, off, proto); 2329 return; 2330 } 2331 /* Drop the packet */ 2332 m_freem(m); 2333 } 2334