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 1.73 2007/11/16 05:07:36 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 translater 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 = 0; 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_ifaddrhead; /* first inet address */ 209 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 210 u_long in_ifaddrhmask; /* mask for hash table */ 211 212 struct ip_stats ipstats_percpu[MAXCPU]; 213 #ifdef SMP 214 static int 215 sysctl_ipstats(SYSCTL_HANDLER_ARGS) 216 { 217 int cpu, error = 0; 218 219 for (cpu = 0; cpu < ncpus; ++cpu) { 220 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu], 221 sizeof(struct ip_stats)))) 222 break; 223 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu], 224 sizeof(struct ip_stats)))) 225 break; 226 } 227 228 return (error); 229 } 230 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW), 231 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics"); 232 #else 233 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 234 &ipstat, ip_stats, "IP statistics"); 235 #endif 236 237 /* Packet reassembly stuff */ 238 #define IPREASS_NHASH_LOG2 6 239 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 240 #define IPREASS_HMASK (IPREASS_NHASH - 1) 241 #define IPREASS_HASH(x,y) \ 242 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 243 244 static struct ipq ipq[IPREASS_NHASH]; 245 246 #ifdef IPCTL_DEFMTU 247 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 248 &ip_mtu, 0, "Default MTU"); 249 #endif 250 251 #ifdef IPSTEALTH 252 static int ipstealth = 0; 253 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, ""); 254 #else 255 static const int ipstealth = 0; 256 #endif 257 258 259 /* Firewall hooks */ 260 ip_fw_chk_t *ip_fw_chk_ptr; 261 ip_fw_dn_io_t *ip_fw_dn_io_ptr; 262 int fw_enable = 1; 263 int fw_one_pass = 1; 264 265 struct pfil_head inet_pfil_hook; 266 267 /* 268 * XXX this is ugly -- the following two global variables are 269 * used to store packet state while it travels through the stack. 270 * Note that the code even makes assumptions on the size and 271 * alignment of fields inside struct ip_srcrt so e.g. adding some 272 * fields will break the code. This needs to be fixed. 273 * 274 * We need to save the IP options in case a protocol wants to respond 275 * to an incoming packet over the same route if the packet got here 276 * using IP source routing. This allows connection establishment and 277 * maintenance when the remote end is on a network that is not known 278 * to us. 279 */ 280 static int ip_nhops = 0; 281 282 static struct ip_srcrt { 283 struct in_addr dst; /* final destination */ 284 char nop; /* one NOP to align */ 285 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 286 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 287 } ip_srcrt; 288 289 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management"); 290 static struct malloc_pipe ipq_mpipe; 291 292 static void save_rte (u_char *, struct in_addr); 293 static int ip_dooptions (struct mbuf *m, int, 294 struct sockaddr_in *next_hop); 295 static void ip_forward (struct mbuf *m, boolean_t using_srcrt, 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 TAILQ_INIT(&in_ifaddrhead); 322 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 323 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 324 if (pr == NULL) 325 panic("ip_init"); 326 for (i = 0; i < IPPROTO_MAX; i++) 327 ip_protox[i] = pr - inetsw; 328 for (pr = inetdomain.dom_protosw; 329 pr < inetdomain.dom_protoswNPROTOSW; pr++) 330 if (pr->pr_domain->dom_family == PF_INET && 331 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 332 ip_protox[pr->pr_protocol] = pr - inetsw; 333 334 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 335 inet_pfil_hook.ph_af = AF_INET; 336 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) { 337 kprintf("%s: WARNING: unable to register pfil hook, " 338 "error %d\n", __func__, i); 339 } 340 341 for (i = 0; i < IPREASS_NHASH; i++) 342 ipq[i].next = ipq[i].prev = &ipq[i]; 343 344 maxnipq = nmbclusters / 32; 345 maxfragsperpacket = 16; 346 347 ip_id = time_second & 0xffff; 348 349 /* 350 * Initialize IP statistics counters for each CPU. 351 * 352 */ 353 #ifdef SMP 354 for (cpu = 0; cpu < ncpus; ++cpu) { 355 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats)); 356 } 357 #else 358 bzero(&ipstat, sizeof(struct ip_stats)); 359 #endif 360 361 netisr_register(NETISR_IP, ip_mport, ip_input_handler); 362 } 363 364 /* 365 * XXX watch out this one. It is perhaps used as a cache for 366 * the most recently used route ? it is cleared in in_addroute() 367 * when a new route is successfully created. 368 */ 369 struct route ipforward_rt[MAXCPU]; 370 371 /* Do transport protocol processing. */ 372 static void 373 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip, 374 struct sockaddr_in *nexthop) 375 { 376 /* 377 * Switch out to protocol's input routine. 378 */ 379 if (nexthop && ip->ip_p == IPPROTO_TCP) { 380 /* TCP needs IPFORWARD info if available */ 381 struct m_hdr tag; 382 383 tag.mh_type = MT_TAG; 384 tag.mh_flags = PACKET_TAG_IPFORWARD; 385 tag.mh_data = (caddr_t)nexthop; 386 tag.mh_next = m; 387 388 (*inetsw[ip_protox[ip->ip_p]].pr_input) 389 ((struct mbuf *)&tag, hlen, ip->ip_p); 390 } else { 391 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, ip->ip_p); 392 } 393 } 394 395 struct netmsg_transport_packet { 396 struct netmsg nm_netmsg; 397 struct mbuf *nm_mbuf; 398 int nm_hlen; 399 boolean_t nm_hasnexthop; 400 struct sockaddr_in nm_nexthop; 401 }; 402 403 static void 404 transport_processing_handler(netmsg_t netmsg) 405 { 406 struct netmsg_transport_packet *msg = (void *)netmsg; 407 struct sockaddr_in *nexthop; 408 struct ip *ip; 409 410 ip = mtod(msg->nm_mbuf, struct ip *); 411 nexthop = msg->nm_hasnexthop ? &msg->nm_nexthop : NULL; 412 transport_processing_oncpu(msg->nm_mbuf, msg->nm_hlen, ip, nexthop); 413 lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 0); 414 } 415 416 static void 417 ip_input_handler(struct netmsg *msg0) 418 { 419 struct mbuf *m = ((struct netmsg_packet *)msg0)->nm_packet; 420 421 ip_input(m); 422 /* msg0 was embedded in the mbuf, do not reply! */ 423 } 424 425 /* 426 * IP input routine. Checksum and byte swap header. If fragmented 427 * try to reassemble. Process options. Pass to next level. 428 */ 429 void 430 ip_input(struct mbuf *m) 431 { 432 struct ip *ip; 433 struct ipq *fp; 434 struct in_ifaddr *ia = NULL; 435 struct ifaddr *ifa; 436 int i, hlen, checkif; 437 u_short sum; 438 struct in_addr pkt_dst; 439 u_int32_t divert_info = 0; /* packet divert/tee info */ 440 struct ip_fw_args args; 441 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */ 442 boolean_t needredispatch = FALSE; 443 struct in_addr odst; /* original dst address(NAT) */ 444 struct m_tag *mtag; 445 #ifdef FAST_IPSEC 446 struct tdb_ident *tdbi; 447 struct secpolicy *sp; 448 int error; 449 #endif 450 451 args.eh = NULL; 452 args.oif = NULL; 453 args.rule = NULL; 454 args.next_hop = NULL; 455 456 /* Grab info from MT_TAG mbufs prepended to the chain. */ 457 while (m != NULL && m->m_type == MT_TAG) { 458 switch(m->_m_tag_id) { 459 case PACKET_TAG_IPFORWARD: 460 args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data; 461 break; 462 default: 463 kprintf("ip_input: unrecognised MT_TAG tag %d\n", 464 m->_m_tag_id); 465 break; 466 } 467 m = m->m_next; 468 } 469 KASSERT(m != NULL && (m->m_flags & M_PKTHDR), ("ip_input: no HDR")); 470 471 /* Extract info from dummynet tag */ 472 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 473 if (mtag != NULL) { 474 args.rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 475 476 m_tag_delete(m, mtag); 477 mtag = NULL; 478 } 479 480 if (args.rule != NULL) { /* dummynet already filtered us */ 481 ip = mtod(m, struct ip *); 482 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 483 goto iphack; 484 } 485 486 ipstat.ips_total++; 487 488 /* length checks already done in ip_demux() */ 489 KASSERT(m->m_len >= sizeof(ip), ("IP header not in one mbuf")); 490 491 ip = mtod(m, struct ip *); 492 493 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 494 ipstat.ips_badvers++; 495 goto bad; 496 } 497 498 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 499 /* length checks already done in ip_demux() */ 500 KASSERT(hlen >= sizeof(struct ip), ("IP header len too small")); 501 KASSERT(m->m_len >= hlen, ("packet shorter than IP header length")); 502 503 /* 127/8 must not appear on wire - RFC1122 */ 504 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 505 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 506 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) { 507 ipstat.ips_badaddr++; 508 goto bad; 509 } 510 } 511 512 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 513 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 514 } else { 515 if (hlen == sizeof(struct ip)) { 516 sum = in_cksum_hdr(ip); 517 } else { 518 sum = in_cksum(m, hlen); 519 } 520 } 521 if (sum != 0) { 522 ipstat.ips_badsum++; 523 goto bad; 524 } 525 526 #ifdef ALTQ 527 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 528 /* packet is dropped by traffic conditioner */ 529 return; 530 } 531 #endif 532 /* 533 * Convert fields to host representation. 534 */ 535 ip->ip_len = ntohs(ip->ip_len); 536 if (ip->ip_len < hlen) { 537 ipstat.ips_badlen++; 538 goto bad; 539 } 540 ip->ip_off = ntohs(ip->ip_off); 541 542 /* 543 * Check that the amount of data in the buffers 544 * is as at least much as the IP header would have us expect. 545 * Trim mbufs if longer than we expect. 546 * Drop packet if shorter than we expect. 547 */ 548 if (m->m_pkthdr.len < ip->ip_len) { 549 ipstat.ips_tooshort++; 550 goto bad; 551 } 552 if (m->m_pkthdr.len > ip->ip_len) { 553 if (m->m_len == m->m_pkthdr.len) { 554 m->m_len = ip->ip_len; 555 m->m_pkthdr.len = ip->ip_len; 556 } else 557 m_adj(m, ip->ip_len - m->m_pkthdr.len); 558 } 559 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF) 560 /* 561 * Bypass packet filtering for packets from a tunnel (gif). 562 */ 563 if (ipsec_gethist(m, NULL)) 564 goto pass; 565 #endif 566 567 /* 568 * IpHack's section. 569 * Right now when no processing on packet has done 570 * and it is still fresh out of network we do our black 571 * deals with it. 572 * - Firewall: deny/allow/divert 573 * - Xlate: translate packet's addr/port (NAT). 574 * - Pipe: pass pkt through dummynet. 575 * - Wrap: fake packet's addr/port <unimpl.> 576 * - Encapsulate: put it in another IP and send out. <unimp.> 577 */ 578 579 iphack: 580 581 /* 582 * Run through list of hooks for input packets. 583 * 584 * NB: Beware of the destination address changing (e.g. 585 * by NAT rewriting). When this happens, tell 586 * ip_forward to do the right thing. 587 */ 588 if (pfil_has_hooks(&inet_pfil_hook)) { 589 odst = ip->ip_dst; 590 if (pfil_run_hooks(&inet_pfil_hook, &m, 591 m->m_pkthdr.rcvif, PFIL_IN)) { 592 return; 593 } 594 if (m == NULL) /* consumed by filter */ 595 return; 596 ip = mtod(m, struct ip *); 597 using_srcrt = (odst.s_addr != ip->ip_dst.s_addr); 598 } 599 600 if (fw_enable && IPFW_LOADED) { 601 /* 602 * If we've been forwarded from the output side, then 603 * skip the firewall a second time 604 */ 605 if (args.next_hop != NULL) 606 goto ours; 607 608 args.m = m; 609 i = ip_fw_chk_ptr(&args); 610 m = args.m; 611 612 if ((i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */ 613 if (m != NULL) 614 m_freem(m); 615 return; 616 } 617 ip = mtod(m, struct ip *); /* just in case m changed */ 618 if (i == 0 && args.next_hop == NULL) /* common case */ 619 goto pass; 620 if (i & IP_FW_PORT_DYNT_FLAG) { 621 /* Send packet to the appropriate pipe */ 622 ip_fw_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args); 623 return; 624 } 625 #ifdef IPDIVERT 626 if (i != 0 && !(i & IP_FW_PORT_DYNT_FLAG)) { 627 /* Divert or tee packet */ 628 divert_info = i; 629 goto ours; 630 } 631 #endif 632 if (i == 0 && args.next_hop != NULL) 633 goto pass; 634 /* 635 * if we get here, the packet must be dropped 636 */ 637 m_freem(m); 638 return; 639 } 640 pass: 641 642 /* 643 * Process options and, if not destined for us, 644 * ship it on. ip_dooptions returns 1 when an 645 * error was detected (causing an icmp message 646 * to be sent and the original packet to be freed). 647 */ 648 ip_nhops = 0; /* for source routed packets */ 649 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, args.next_hop)) 650 return; 651 652 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 653 * matter if it is destined to another node, or whether it is 654 * a multicast one, RSVP wants it! and prevents it from being forwarded 655 * anywhere else. Also checks if the rsvp daemon is running before 656 * grabbing the packet. 657 */ 658 if (rsvp_on && ip->ip_p == IPPROTO_RSVP) 659 goto ours; 660 661 /* 662 * Check our list of addresses, to see if the packet is for us. 663 * If we don't have any addresses, assume any unicast packet 664 * we receive might be for us (and let the upper layers deal 665 * with it). 666 */ 667 if (TAILQ_EMPTY(&in_ifaddrhead) && !(m->m_flags & (M_MCAST | M_BCAST))) 668 goto ours; 669 670 /* 671 * Cache the destination address of the packet; this may be 672 * changed by use of 'ipfw fwd'. 673 */ 674 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst; 675 676 /* 677 * Enable a consistency check between the destination address 678 * and the arrival interface for a unicast packet (the RFC 1122 679 * strong ES model) if IP forwarding is disabled and the packet 680 * is not locally generated and the packet is not subject to 681 * 'ipfw fwd'. 682 * 683 * XXX - Checking also should be disabled if the destination 684 * address is ipnat'ed to a different interface. 685 * 686 * XXX - Checking is incompatible with IP aliases added 687 * to the loopback interface instead of the interface where 688 * the packets are received. 689 */ 690 checkif = ip_checkinterface && 691 !ipforwarding && 692 m->m_pkthdr.rcvif != NULL && 693 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) && 694 (args.next_hop == NULL); 695 696 /* 697 * Check for exact addresses in the hash bucket. 698 */ 699 LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) { 700 /* 701 * If the address matches, verify that the packet 702 * arrived via the correct interface if checking is 703 * enabled. 704 */ 705 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 706 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 707 goto ours; 708 } 709 /* 710 * Check for broadcast addresses. 711 * 712 * Only accept broadcast packets that arrive via the matching 713 * interface. Reception of forwarded directed broadcasts would 714 * be handled via ip_forward() and ether_output() with the loopback 715 * into the stack for SIMPLEX interfaces handled by ether_output(). 716 */ 717 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 718 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 719 if (ifa->ifa_addr == NULL) /* shutdown/startup race */ 720 continue; 721 if (ifa->ifa_addr->sa_family != AF_INET) 722 continue; 723 ia = ifatoia(ifa); 724 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 725 pkt_dst.s_addr) 726 goto ours; 727 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 728 goto ours; 729 #ifdef BOOTP_COMPAT 730 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 731 goto ours; 732 #endif 733 } 734 } 735 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 736 struct in_multi *inm; 737 738 if (ip_mrouter != NULL) { 739 /* 740 * If we are acting as a multicast router, all 741 * incoming multicast packets are passed to the 742 * kernel-level multicast forwarding function. 743 * The packet is returned (relatively) intact; if 744 * ip_mforward() returns a non-zero value, the packet 745 * must be discarded, else it may be accepted below. 746 */ 747 if (ip_mforward != NULL && 748 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) { 749 ipstat.ips_cantforward++; 750 m_freem(m); 751 return; 752 } 753 754 /* 755 * The process-level routing daemon needs to receive 756 * all multicast IGMP packets, whether or not this 757 * host belongs to their destination groups. 758 */ 759 if (ip->ip_p == IPPROTO_IGMP) 760 goto ours; 761 ipstat.ips_forward++; 762 } 763 /* 764 * See if we belong to the destination multicast group on the 765 * arrival interface. 766 */ 767 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 768 if (inm == NULL) { 769 ipstat.ips_notmember++; 770 m_freem(m); 771 return; 772 } 773 goto ours; 774 } 775 if (ip->ip_dst.s_addr == INADDR_BROADCAST) 776 goto ours; 777 if (ip->ip_dst.s_addr == INADDR_ANY) 778 goto ours; 779 780 /* 781 * FAITH(Firewall Aided Internet Translator) 782 */ 783 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 784 if (ip_keepfaith) { 785 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 786 goto ours; 787 } 788 m_freem(m); 789 return; 790 } 791 792 /* 793 * Not for us; forward if possible and desirable. 794 */ 795 if (!ipforwarding) { 796 ipstat.ips_cantforward++; 797 m_freem(m); 798 } else { 799 #ifdef IPSEC 800 /* 801 * Enforce inbound IPsec SPD. 802 */ 803 if (ipsec4_in_reject(m, NULL)) { 804 ipsecstat.in_polvio++; 805 goto bad; 806 } 807 #endif 808 #ifdef FAST_IPSEC 809 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 810 crit_enter(); 811 if (mtag != NULL) { 812 tdbi = (struct tdb_ident *)m_tag_data(mtag); 813 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 814 } else { 815 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 816 IP_FORWARDING, &error); 817 } 818 if (sp == NULL) { /* NB: can happen if error */ 819 crit_exit(); 820 /*XXX error stat???*/ 821 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 822 goto bad; 823 } 824 825 /* 826 * Check security policy against packet attributes. 827 */ 828 error = ipsec_in_reject(sp, m); 829 KEY_FREESP(&sp); 830 crit_exit(); 831 if (error) { 832 ipstat.ips_cantforward++; 833 goto bad; 834 } 835 #endif 836 ip_forward(m, using_srcrt, args.next_hop); 837 } 838 return; 839 840 ours: 841 842 /* 843 * IPSTEALTH: Process non-routing options only 844 * if the packet is destined for us. 845 */ 846 if (ipstealth && 847 hlen > sizeof(struct ip) && 848 ip_dooptions(m, 1, args.next_hop)) 849 return; 850 851 /* Count the packet in the ip address stats */ 852 if (ia != NULL) { 853 ia->ia_ifa.if_ipackets++; 854 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 855 } 856 857 /* 858 * If offset or IP_MF are set, must reassemble. 859 * Otherwise, nothing need be done. 860 * (We could look in the reassembly queue to see 861 * if the packet was previously fragmented, 862 * but it's not worth the time; just let them time out.) 863 */ 864 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 865 866 /* If maxnipq is 0, never accept fragments. */ 867 if (maxnipq == 0) { 868 ipstat.ips_fragments++; 869 ipstat.ips_fragdropped++; 870 goto bad; 871 } 872 873 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 874 /* 875 * Look for queue of fragments 876 * of this datagram. 877 */ 878 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) 879 if (ip->ip_id == fp->ipq_id && 880 ip->ip_src.s_addr == fp->ipq_src.s_addr && 881 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 882 ip->ip_p == fp->ipq_p) 883 goto found; 884 885 fp = NULL; 886 887 /* 888 * Enforce upper bound on number of fragmented packets 889 * for which we attempt reassembly; 890 * If maxnipq is -1, accept all fragments without limitation. 891 */ 892 if ((nipq > maxnipq) && (maxnipq > 0)) { 893 /* 894 * drop something from the tail of the current queue 895 * before proceeding further 896 */ 897 if (ipq[sum].prev == &ipq[sum]) { /* gak */ 898 for (i = 0; i < IPREASS_NHASH; i++) { 899 if (ipq[i].prev != &ipq[i]) { 900 ipstat.ips_fragtimeout += 901 ipq[i].prev->ipq_nfrags; 902 ip_freef(ipq[i].prev); 903 break; 904 } 905 } 906 } else { 907 ipstat.ips_fragtimeout += 908 ipq[sum].prev->ipq_nfrags; 909 ip_freef(ipq[sum].prev); 910 } 911 } 912 found: 913 /* 914 * Adjust ip_len to not reflect header, 915 * convert offset of this to bytes. 916 */ 917 ip->ip_len -= hlen; 918 if (ip->ip_off & IP_MF) { 919 /* 920 * Make sure that fragments have a data length 921 * that's a non-zero multiple of 8 bytes. 922 */ 923 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 924 ipstat.ips_toosmall++; /* XXX */ 925 goto bad; 926 } 927 m->m_flags |= M_FRAG; 928 } else 929 m->m_flags &= ~M_FRAG; 930 ip->ip_off <<= 3; 931 932 /* 933 * Attempt reassembly; if it succeeds, proceed. 934 * ip_reass() will return a different mbuf, and update 935 * the divert info in divert_info. 936 */ 937 ipstat.ips_fragments++; 938 m->m_pkthdr.header = ip; 939 m = ip_reass(m, fp, &ipq[sum], &divert_info); 940 if (m == NULL) 941 return; 942 ipstat.ips_reassembled++; 943 needredispatch = TRUE; 944 ip = mtod(m, struct ip *); 945 /* Get the header length of the reassembled packet */ 946 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 947 #ifdef IPDIVERT 948 /* Restore original checksum before diverting packet */ 949 if (divert_info != 0) { 950 ip->ip_len += hlen; 951 ip->ip_len = htons(ip->ip_len); 952 ip->ip_off = htons(ip->ip_off); 953 ip->ip_sum = 0; 954 if (hlen == sizeof(struct ip)) 955 ip->ip_sum = in_cksum_hdr(ip); 956 else 957 ip->ip_sum = in_cksum(m, hlen); 958 ip->ip_off = ntohs(ip->ip_off); 959 ip->ip_len = ntohs(ip->ip_len); 960 ip->ip_len -= hlen; 961 } 962 #endif 963 } else { 964 ip->ip_len -= hlen; 965 } 966 967 #ifdef IPDIVERT 968 /* 969 * Divert or tee packet to the divert protocol if required. 970 */ 971 if (divert_info != 0) { 972 struct mbuf *clone = NULL; 973 974 /* Clone packet if we're doing a 'tee' */ 975 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 976 clone = m_dup(m, MB_DONTWAIT); 977 978 /* Restore packet header fields to original values */ 979 ip->ip_len += hlen; 980 ip->ip_len = htons(ip->ip_len); 981 ip->ip_off = htons(ip->ip_off); 982 983 /* Deliver packet to divert input routine */ 984 divert_packet(m, 1, divert_info & 0xffff); 985 ipstat.ips_delivered++; 986 987 /* If 'tee', continue with original packet */ 988 if (clone == NULL) 989 return; 990 m = clone; 991 ip = mtod(m, struct ip *); 992 ip->ip_len += hlen; 993 /* 994 * Jump backwards to complete processing of the 995 * packet. But first clear divert_info to avoid 996 * entering this block again. 997 * We do not need to clear args.divert_rule 998 * or args.next_hop as they will not be used. 999 * 1000 * XXX Better safe than sorry, remove the DIVERT tag. 1001 */ 1002 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL); 1003 if (mtag != NULL) 1004 m_tag_delete(m, mtag); 1005 1006 divert_info = 0; 1007 goto pass; 1008 } 1009 #endif 1010 1011 #ifdef IPSEC 1012 /* 1013 * enforce IPsec policy checking if we are seeing last header. 1014 * note that we do not visit this with protocols with pcb layer 1015 * code - like udp/tcp/raw ip. 1016 */ 1017 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) && 1018 ipsec4_in_reject(m, NULL)) { 1019 ipsecstat.in_polvio++; 1020 goto bad; 1021 } 1022 #endif 1023 #if FAST_IPSEC 1024 /* 1025 * enforce IPsec policy checking if we are seeing last header. 1026 * note that we do not visit this with protocols with pcb layer 1027 * code - like udp/tcp/raw ip. 1028 */ 1029 if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) { 1030 /* 1031 * Check if the packet has already had IPsec processing 1032 * done. If so, then just pass it along. This tag gets 1033 * set during AH, ESP, etc. input handling, before the 1034 * packet is returned to the ip input queue for delivery. 1035 */ 1036 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 1037 crit_enter(); 1038 if (mtag != NULL) { 1039 tdbi = (struct tdb_ident *)m_tag_data(mtag); 1040 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 1041 } else { 1042 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 1043 IP_FORWARDING, &error); 1044 } 1045 if (sp != NULL) { 1046 /* 1047 * Check security policy against packet attributes. 1048 */ 1049 error = ipsec_in_reject(sp, m); 1050 KEY_FREESP(&sp); 1051 } else { 1052 /* XXX error stat??? */ 1053 error = EINVAL; 1054 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 1055 goto bad; 1056 } 1057 crit_exit(); 1058 if (error) 1059 goto bad; 1060 } 1061 #endif /* FAST_IPSEC */ 1062 1063 ipstat.ips_delivered++; 1064 if (needredispatch) { 1065 struct netmsg_transport_packet *msg; 1066 lwkt_port_t port; 1067 1068 ip->ip_off = htons(ip->ip_off); 1069 ip->ip_len = htons(ip->ip_len); 1070 port = ip_mport(&m); 1071 if (port == NULL) 1072 return; 1073 1074 msg = kmalloc(sizeof(struct netmsg_transport_packet), M_LWKTMSG, 1075 M_INTWAIT | M_NULLOK); 1076 if (msg == NULL) 1077 goto bad; 1078 1079 netmsg_init(&msg->nm_netmsg, &netisr_afree_rport, 0, 1080 transport_processing_handler); 1081 msg->nm_hlen = hlen; 1082 msg->nm_hasnexthop = (args.next_hop != NULL); 1083 if (msg->nm_hasnexthop) 1084 msg->nm_nexthop = *args.next_hop; /* structure copy */ 1085 1086 msg->nm_mbuf = m; 1087 ip = mtod(m, struct ip *); 1088 ip->ip_len = ntohs(ip->ip_len); 1089 ip->ip_off = ntohs(ip->ip_off); 1090 lwkt_sendmsg(port, &msg->nm_netmsg.nm_lmsg); 1091 } else { 1092 transport_processing_oncpu(m, hlen, ip, args.next_hop); 1093 } 1094 return; 1095 1096 bad: 1097 m_freem(m); 1098 } 1099 1100 /* 1101 * Take incoming datagram fragment and try to reassemble it into 1102 * whole datagram. If a chain for reassembly of this datagram already 1103 * exists, then it is given as fp; otherwise have to make a chain. 1104 * 1105 * When IPDIVERT enabled, keep additional state with each packet that 1106 * tells us if we need to divert or tee the packet we're building. 1107 * In particular, *divinfo includes the port and TEE flag. 1108 */ 1109 1110 static struct mbuf * 1111 ip_reass(struct mbuf *m, struct ipq *fp, struct ipq *where, 1112 u_int32_t *divinfo) 1113 { 1114 struct ip *ip = mtod(m, struct ip *); 1115 struct mbuf *p = NULL, *q, *nq; 1116 struct mbuf *n; 1117 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 1118 int i, next; 1119 #ifdef IPDIVERT 1120 struct m_tag *mtag; 1121 #endif 1122 1123 /* 1124 * If the hardware has not done csum over this fragment 1125 * then csum_data is not valid at all. 1126 */ 1127 if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) 1128 == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) { 1129 m->m_pkthdr.csum_data = 0; 1130 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 1131 } 1132 1133 /* 1134 * Presence of header sizes in mbufs 1135 * would confuse code below. 1136 */ 1137 m->m_data += hlen; 1138 m->m_len -= hlen; 1139 1140 /* 1141 * If first fragment to arrive, create a reassembly queue. 1142 */ 1143 if (fp == NULL) { 1144 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL) 1145 goto dropfrag; 1146 insque(fp, where); 1147 nipq++; 1148 fp->ipq_nfrags = 1; 1149 fp->ipq_ttl = IPFRAGTTL; 1150 fp->ipq_p = ip->ip_p; 1151 fp->ipq_id = ip->ip_id; 1152 fp->ipq_src = ip->ip_src; 1153 fp->ipq_dst = ip->ip_dst; 1154 fp->ipq_frags = m; 1155 m->m_nextpkt = NULL; 1156 #ifdef IPDIVERT 1157 fp->ipq_div_info = 0; 1158 #endif 1159 goto inserted; 1160 } else { 1161 fp->ipq_nfrags++; 1162 } 1163 1164 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1165 1166 /* 1167 * Find a segment which begins after this one does. 1168 */ 1169 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1170 if (GETIP(q)->ip_off > ip->ip_off) 1171 break; 1172 1173 /* 1174 * If there is a preceding segment, it may provide some of 1175 * our data already. If so, drop the data from the incoming 1176 * segment. If it provides all of our data, drop us, otherwise 1177 * stick new segment in the proper place. 1178 * 1179 * If some of the data is dropped from the the preceding 1180 * segment, then it's checksum is invalidated. 1181 */ 1182 if (p) { 1183 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1184 if (i > 0) { 1185 if (i >= ip->ip_len) 1186 goto dropfrag; 1187 m_adj(m, i); 1188 m->m_pkthdr.csum_flags = 0; 1189 ip->ip_off += i; 1190 ip->ip_len -= i; 1191 } 1192 m->m_nextpkt = p->m_nextpkt; 1193 p->m_nextpkt = m; 1194 } else { 1195 m->m_nextpkt = fp->ipq_frags; 1196 fp->ipq_frags = m; 1197 } 1198 1199 /* 1200 * While we overlap succeeding segments trim them or, 1201 * if they are completely covered, dequeue them. 1202 */ 1203 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1204 q = nq) { 1205 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1206 if (i < GETIP(q)->ip_len) { 1207 GETIP(q)->ip_len -= i; 1208 GETIP(q)->ip_off += i; 1209 m_adj(q, i); 1210 q->m_pkthdr.csum_flags = 0; 1211 break; 1212 } 1213 nq = q->m_nextpkt; 1214 m->m_nextpkt = nq; 1215 ipstat.ips_fragdropped++; 1216 fp->ipq_nfrags--; 1217 q->m_nextpkt = NULL; 1218 m_freem(q); 1219 } 1220 1221 inserted: 1222 1223 #ifdef IPDIVERT 1224 /* 1225 * Transfer firewall instructions to the fragment structure. 1226 * Only trust info in the fragment at offset 0. 1227 */ 1228 if (ip->ip_off == 0) { 1229 fp->ipq_div_info = *divinfo; 1230 } else { 1231 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL); 1232 if (mtag != NULL) 1233 m_tag_delete(m, mtag); 1234 } 1235 *divinfo = 0; 1236 #endif 1237 1238 /* 1239 * Check for complete reassembly and perform frag per packet 1240 * limiting. 1241 * 1242 * Frag limiting is performed here so that the nth frag has 1243 * a chance to complete the packet before we drop the packet. 1244 * As a result, n+1 frags are actually allowed per packet, but 1245 * only n will ever be stored. (n = maxfragsperpacket.) 1246 * 1247 */ 1248 next = 0; 1249 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1250 if (GETIP(q)->ip_off != next) { 1251 if (fp->ipq_nfrags > maxfragsperpacket) { 1252 ipstat.ips_fragdropped += fp->ipq_nfrags; 1253 ip_freef(fp); 1254 } 1255 return (NULL); 1256 } 1257 next += GETIP(q)->ip_len; 1258 } 1259 /* Make sure the last packet didn't have the IP_MF flag */ 1260 if (p->m_flags & M_FRAG) { 1261 if (fp->ipq_nfrags > maxfragsperpacket) { 1262 ipstat.ips_fragdropped += fp->ipq_nfrags; 1263 ip_freef(fp); 1264 } 1265 return (NULL); 1266 } 1267 1268 /* 1269 * Reassembly is complete. Make sure the packet is a sane size. 1270 */ 1271 q = fp->ipq_frags; 1272 ip = GETIP(q); 1273 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 1274 ipstat.ips_toolong++; 1275 ipstat.ips_fragdropped += fp->ipq_nfrags; 1276 ip_freef(fp); 1277 return (NULL); 1278 } 1279 1280 /* 1281 * Concatenate fragments. 1282 */ 1283 m = q; 1284 n = m->m_next; 1285 m->m_next = NULL; 1286 m_cat(m, n); 1287 nq = q->m_nextpkt; 1288 q->m_nextpkt = NULL; 1289 for (q = nq; q != NULL; q = nq) { 1290 nq = q->m_nextpkt; 1291 q->m_nextpkt = NULL; 1292 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1293 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1294 m_cat(m, q); 1295 } 1296 1297 /* 1298 * Clean up the 1's complement checksum. Carry over 16 bits must 1299 * be added back. This assumes no more then 65535 packet fragments 1300 * were reassembled. A second carry can also occur (but not a third). 1301 */ 1302 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) + 1303 (m->m_pkthdr.csum_data >> 16); 1304 if (m->m_pkthdr.csum_data > 0xFFFF) 1305 m->m_pkthdr.csum_data -= 0xFFFF; 1306 1307 1308 #ifdef IPDIVERT 1309 /* 1310 * Extract firewall instructions from the fragment structure. 1311 */ 1312 *divinfo = fp->ipq_div_info; 1313 #endif 1314 1315 /* 1316 * Create header for new ip packet by 1317 * modifying header of first packet; 1318 * dequeue and discard fragment reassembly header. 1319 * Make header visible. 1320 */ 1321 ip->ip_len = next; 1322 ip->ip_src = fp->ipq_src; 1323 ip->ip_dst = fp->ipq_dst; 1324 remque(fp); 1325 nipq--; 1326 mpipe_free(&ipq_mpipe, fp); 1327 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 1328 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 1329 /* some debugging cruft by sklower, below, will go away soon */ 1330 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1331 int plen = 0; 1332 1333 for (n = m; n; n = n->m_next) 1334 plen += n->m_len; 1335 m->m_pkthdr.len = plen; 1336 } 1337 return (m); 1338 1339 dropfrag: 1340 #ifdef IPDIVERT 1341 *divinfo = 0; 1342 #endif 1343 ipstat.ips_fragdropped++; 1344 if (fp != NULL) 1345 fp->ipq_nfrags--; 1346 m_freem(m); 1347 return (NULL); 1348 1349 #undef GETIP 1350 } 1351 1352 /* 1353 * Free a fragment reassembly header and all 1354 * associated datagrams. 1355 */ 1356 static void 1357 ip_freef(struct ipq *fp) 1358 { 1359 struct mbuf *q; 1360 1361 while (fp->ipq_frags) { 1362 q = fp->ipq_frags; 1363 fp->ipq_frags = q->m_nextpkt; 1364 q->m_nextpkt = NULL; 1365 m_freem(q); 1366 } 1367 remque(fp); 1368 mpipe_free(&ipq_mpipe, fp); 1369 nipq--; 1370 } 1371 1372 /* 1373 * IP timer processing; 1374 * if a timer expires on a reassembly 1375 * queue, discard it. 1376 */ 1377 void 1378 ip_slowtimo(void) 1379 { 1380 struct ipq *fp; 1381 int i; 1382 1383 crit_enter(); 1384 for (i = 0; i < IPREASS_NHASH; i++) { 1385 fp = ipq[i].next; 1386 if (fp == NULL) 1387 continue; 1388 while (fp != &ipq[i]) { 1389 --fp->ipq_ttl; 1390 fp = fp->next; 1391 if (fp->prev->ipq_ttl == 0) { 1392 ipstat.ips_fragtimeout += fp->prev->ipq_nfrags; 1393 ip_freef(fp->prev); 1394 } 1395 } 1396 } 1397 /* 1398 * If we are over the maximum number of fragments 1399 * (due to the limit being lowered), drain off 1400 * enough to get down to the new limit. 1401 */ 1402 if (maxnipq >= 0 && nipq > maxnipq) { 1403 for (i = 0; i < IPREASS_NHASH; i++) { 1404 while (nipq > maxnipq && 1405 (ipq[i].next != &ipq[i])) { 1406 ipstat.ips_fragdropped += 1407 ipq[i].next->ipq_nfrags; 1408 ip_freef(ipq[i].next); 1409 } 1410 } 1411 } 1412 ipflow_slowtimo(); 1413 crit_exit(); 1414 } 1415 1416 /* 1417 * Drain off all datagram fragments. 1418 */ 1419 void 1420 ip_drain(void) 1421 { 1422 int i; 1423 1424 for (i = 0; i < IPREASS_NHASH; i++) { 1425 while (ipq[i].next != &ipq[i]) { 1426 ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags; 1427 ip_freef(ipq[i].next); 1428 } 1429 } 1430 in_rtqdrain(); 1431 } 1432 1433 /* 1434 * Do option processing on a datagram, 1435 * possibly discarding it if bad options are encountered, 1436 * or forwarding it if source-routed. 1437 * The pass argument is used when operating in the IPSTEALTH 1438 * mode to tell what options to process: 1439 * [LS]SRR (pass 0) or the others (pass 1). 1440 * The reason for as many as two passes is that when doing IPSTEALTH, 1441 * non-routing options should be processed only if the packet is for us. 1442 * Returns 1 if packet has been forwarded/freed, 1443 * 0 if the packet should be processed further. 1444 */ 1445 static int 1446 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop) 1447 { 1448 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET }; 1449 struct ip *ip = mtod(m, struct ip *); 1450 u_char *cp; 1451 struct in_ifaddr *ia; 1452 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB; 1453 boolean_t forward = FALSE; 1454 struct in_addr *sin, dst; 1455 n_time ntime; 1456 1457 dst = ip->ip_dst; 1458 cp = (u_char *)(ip + 1); 1459 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1460 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1461 opt = cp[IPOPT_OPTVAL]; 1462 if (opt == IPOPT_EOL) 1463 break; 1464 if (opt == IPOPT_NOP) 1465 optlen = 1; 1466 else { 1467 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1468 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1469 goto bad; 1470 } 1471 optlen = cp[IPOPT_OLEN]; 1472 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1473 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1474 goto bad; 1475 } 1476 } 1477 switch (opt) { 1478 1479 default: 1480 break; 1481 1482 /* 1483 * Source routing with record. 1484 * Find interface with current destination address. 1485 * If none on this machine then drop if strictly routed, 1486 * or do nothing if loosely routed. 1487 * Record interface address and bring up next address 1488 * component. If strictly routed make sure next 1489 * address is on directly accessible net. 1490 */ 1491 case IPOPT_LSRR: 1492 case IPOPT_SSRR: 1493 if (ipstealth && pass > 0) 1494 break; 1495 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1496 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1497 goto bad; 1498 } 1499 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1500 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1501 goto bad; 1502 } 1503 ipaddr.sin_addr = ip->ip_dst; 1504 ia = (struct in_ifaddr *) 1505 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1506 if (ia == NULL) { 1507 if (opt == IPOPT_SSRR) { 1508 type = ICMP_UNREACH; 1509 code = ICMP_UNREACH_SRCFAIL; 1510 goto bad; 1511 } 1512 if (!ip_dosourceroute) 1513 goto nosourcerouting; 1514 /* 1515 * Loose routing, and not at next destination 1516 * yet; nothing to do except forward. 1517 */ 1518 break; 1519 } 1520 off--; /* 0 origin */ 1521 if (off > optlen - (int)sizeof(struct in_addr)) { 1522 /* 1523 * End of source route. Should be for us. 1524 */ 1525 if (!ip_acceptsourceroute) 1526 goto nosourcerouting; 1527 save_rte(cp, ip->ip_src); 1528 break; 1529 } 1530 if (ipstealth) 1531 goto dropit; 1532 if (!ip_dosourceroute) { 1533 if (ipforwarding) { 1534 char buf[sizeof "aaa.bbb.ccc.ddd"]; 1535 1536 /* 1537 * Acting as a router, so generate ICMP 1538 */ 1539 nosourcerouting: 1540 strcpy(buf, inet_ntoa(ip->ip_dst)); 1541 log(LOG_WARNING, 1542 "attempted source route from %s to %s\n", 1543 inet_ntoa(ip->ip_src), buf); 1544 type = ICMP_UNREACH; 1545 code = ICMP_UNREACH_SRCFAIL; 1546 goto bad; 1547 } else { 1548 /* 1549 * Not acting as a router, 1550 * so silently drop. 1551 */ 1552 dropit: 1553 ipstat.ips_cantforward++; 1554 m_freem(m); 1555 return (1); 1556 } 1557 } 1558 1559 /* 1560 * locate outgoing interface 1561 */ 1562 memcpy(&ipaddr.sin_addr, cp + off, 1563 sizeof ipaddr.sin_addr); 1564 1565 if (opt == IPOPT_SSRR) { 1566 #define INA struct in_ifaddr * 1567 #define SA struct sockaddr * 1568 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) 1569 == NULL) 1570 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1571 } else 1572 ia = ip_rtaddr(ipaddr.sin_addr, 1573 &ipforward_rt[mycpuid]); 1574 if (ia == NULL) { 1575 type = ICMP_UNREACH; 1576 code = ICMP_UNREACH_SRCFAIL; 1577 goto bad; 1578 } 1579 ip->ip_dst = ipaddr.sin_addr; 1580 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1581 sizeof(struct in_addr)); 1582 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1583 /* 1584 * Let ip_intr's mcast routing check handle mcast pkts 1585 */ 1586 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1587 break; 1588 1589 case IPOPT_RR: 1590 if (ipstealth && pass == 0) 1591 break; 1592 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1593 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1594 goto bad; 1595 } 1596 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1597 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1598 goto bad; 1599 } 1600 /* 1601 * If no space remains, ignore. 1602 */ 1603 off--; /* 0 origin */ 1604 if (off > optlen - (int)sizeof(struct in_addr)) 1605 break; 1606 memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1607 sizeof ipaddr.sin_addr); 1608 /* 1609 * locate outgoing interface; if we're the destination, 1610 * use the incoming interface (should be same). 1611 */ 1612 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL && 1613 (ia = ip_rtaddr(ipaddr.sin_addr, 1614 &ipforward_rt[mycpuid])) 1615 == NULL) { 1616 type = ICMP_UNREACH; 1617 code = ICMP_UNREACH_HOST; 1618 goto bad; 1619 } 1620 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1621 sizeof(struct in_addr)); 1622 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1623 break; 1624 1625 case IPOPT_TS: 1626 if (ipstealth && pass == 0) 1627 break; 1628 code = cp - (u_char *)ip; 1629 if (optlen < 4 || optlen > 40) { 1630 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1631 goto bad; 1632 } 1633 if ((off = cp[IPOPT_OFFSET]) < 5) { 1634 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1635 goto bad; 1636 } 1637 if (off > optlen - (int)sizeof(int32_t)) { 1638 cp[IPOPT_OFFSET + 1] += (1 << 4); 1639 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1640 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1641 goto bad; 1642 } 1643 break; 1644 } 1645 off--; /* 0 origin */ 1646 sin = (struct in_addr *)(cp + off); 1647 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1648 1649 case IPOPT_TS_TSONLY: 1650 break; 1651 1652 case IPOPT_TS_TSANDADDR: 1653 if (off + sizeof(n_time) + 1654 sizeof(struct in_addr) > optlen) { 1655 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1656 goto bad; 1657 } 1658 ipaddr.sin_addr = dst; 1659 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1660 m->m_pkthdr.rcvif); 1661 if (ia == NULL) 1662 continue; 1663 memcpy(sin, &IA_SIN(ia)->sin_addr, 1664 sizeof(struct in_addr)); 1665 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1666 off += sizeof(struct in_addr); 1667 break; 1668 1669 case IPOPT_TS_PRESPEC: 1670 if (off + sizeof(n_time) + 1671 sizeof(struct in_addr) > optlen) { 1672 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1673 goto bad; 1674 } 1675 memcpy(&ipaddr.sin_addr, sin, 1676 sizeof(struct in_addr)); 1677 if (ifa_ifwithaddr((SA)&ipaddr) == NULL) 1678 continue; 1679 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1680 off += sizeof(struct in_addr); 1681 break; 1682 1683 default: 1684 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1685 goto bad; 1686 } 1687 ntime = iptime(); 1688 memcpy(cp + off, &ntime, sizeof(n_time)); 1689 cp[IPOPT_OFFSET] += sizeof(n_time); 1690 } 1691 } 1692 if (forward && ipforwarding) { 1693 ip_forward(m, TRUE, next_hop); 1694 return (1); 1695 } 1696 return (0); 1697 bad: 1698 icmp_error(m, type, code, 0, 0); 1699 ipstat.ips_badoptions++; 1700 return (1); 1701 } 1702 1703 /* 1704 * Given address of next destination (final or next hop), 1705 * return internet address info of interface to be used to get there. 1706 */ 1707 struct in_ifaddr * 1708 ip_rtaddr(struct in_addr dst, struct route *ro) 1709 { 1710 struct sockaddr_in *sin; 1711 1712 sin = (struct sockaddr_in *)&ro->ro_dst; 1713 1714 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) { 1715 if (ro->ro_rt != NULL) { 1716 RTFREE(ro->ro_rt); 1717 ro->ro_rt = NULL; 1718 } 1719 sin->sin_family = AF_INET; 1720 sin->sin_len = sizeof *sin; 1721 sin->sin_addr = dst; 1722 rtalloc_ign(ro, RTF_PRCLONING); 1723 } 1724 1725 if (ro->ro_rt == NULL) 1726 return (NULL); 1727 1728 return (ifatoia(ro->ro_rt->rt_ifa)); 1729 } 1730 1731 /* 1732 * Save incoming source route for use in replies, 1733 * to be picked up later by ip_srcroute if the receiver is interested. 1734 */ 1735 void 1736 save_rte(u_char *option, struct in_addr dst) 1737 { 1738 unsigned olen; 1739 1740 olen = option[IPOPT_OLEN]; 1741 #ifdef DIAGNOSTIC 1742 if (ipprintfs) 1743 kprintf("save_rte: olen %d\n", olen); 1744 #endif 1745 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1746 return; 1747 bcopy(option, ip_srcrt.srcopt, olen); 1748 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1749 ip_srcrt.dst = dst; 1750 } 1751 1752 /* 1753 * Retrieve incoming source route for use in replies, 1754 * in the same form used by setsockopt. 1755 * The first hop is placed before the options, will be removed later. 1756 */ 1757 struct mbuf * 1758 ip_srcroute(void) 1759 { 1760 struct in_addr *p, *q; 1761 struct mbuf *m; 1762 1763 if (ip_nhops == 0) 1764 return (NULL); 1765 m = m_get(MB_DONTWAIT, MT_HEADER); 1766 if (m == NULL) 1767 return (NULL); 1768 1769 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1770 1771 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1772 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1773 OPTSIZ; 1774 #ifdef DIAGNOSTIC 1775 if (ipprintfs) 1776 kprintf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1777 #endif 1778 1779 /* 1780 * First save first hop for return route 1781 */ 1782 p = &ip_srcrt.route[ip_nhops - 1]; 1783 *(mtod(m, struct in_addr *)) = *p--; 1784 #ifdef DIAGNOSTIC 1785 if (ipprintfs) 1786 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1787 #endif 1788 1789 /* 1790 * Copy option fields and padding (nop) to mbuf. 1791 */ 1792 ip_srcrt.nop = IPOPT_NOP; 1793 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1794 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &ip_srcrt.nop, 1795 OPTSIZ); 1796 q = (struct in_addr *)(mtod(m, caddr_t) + 1797 sizeof(struct in_addr) + OPTSIZ); 1798 #undef OPTSIZ 1799 /* 1800 * Record return path as an IP source route, 1801 * reversing the path (pointers are now aligned). 1802 */ 1803 while (p >= ip_srcrt.route) { 1804 #ifdef DIAGNOSTIC 1805 if (ipprintfs) 1806 kprintf(" %x", ntohl(q->s_addr)); 1807 #endif 1808 *q++ = *p--; 1809 } 1810 /* 1811 * Last hop goes to final destination. 1812 */ 1813 *q = ip_srcrt.dst; 1814 #ifdef DIAGNOSTIC 1815 if (ipprintfs) 1816 kprintf(" %x\n", ntohl(q->s_addr)); 1817 #endif 1818 return (m); 1819 } 1820 1821 /* 1822 * Strip out IP options. 1823 */ 1824 void 1825 ip_stripoptions(struct mbuf *m) 1826 { 1827 int datalen; 1828 struct ip *ip = mtod(m, struct ip *); 1829 caddr_t opts; 1830 int optlen; 1831 1832 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1833 opts = (caddr_t)(ip + 1); 1834 datalen = m->m_len - (sizeof(struct ip) + optlen); 1835 bcopy(opts + optlen, opts, datalen); 1836 m->m_len -= optlen; 1837 if (m->m_flags & M_PKTHDR) 1838 m->m_pkthdr.len -= optlen; 1839 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1840 } 1841 1842 u_char inetctlerrmap[PRC_NCMDS] = { 1843 0, 0, 0, 0, 1844 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1845 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1846 EMSGSIZE, EHOSTUNREACH, 0, 0, 1847 0, 0, 0, 0, 1848 ENOPROTOOPT, ECONNREFUSED 1849 }; 1850 1851 /* 1852 * Forward a packet. If some error occurs return the sender 1853 * an icmp packet. Note we can't always generate a meaningful 1854 * icmp message because icmp doesn't have a large enough repertoire 1855 * of codes and types. 1856 * 1857 * If not forwarding, just drop the packet. This could be confusing 1858 * if ipforwarding was zero but some routing protocol was advancing 1859 * us as a gateway to somewhere. However, we must let the routing 1860 * protocol deal with that. 1861 * 1862 * The using_srcrt parameter indicates whether the packet is being forwarded 1863 * via a source route. 1864 */ 1865 static void 1866 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop) 1867 { 1868 struct ip *ip = mtod(m, struct ip *); 1869 struct sockaddr_in *ipforward_rtaddr; 1870 struct rtentry *rt; 1871 int error, type = 0, code = 0, destmtu = 0; 1872 struct mbuf *mcopy; 1873 n_long dest; 1874 struct in_addr pkt_dst; 1875 struct m_hdr tag; 1876 struct route *cache_rt = &ipforward_rt[mycpuid]; 1877 1878 dest = INADDR_ANY; 1879 /* 1880 * Cache the destination address of the packet; this may be 1881 * changed by use of 'ipfw fwd'. 1882 */ 1883 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst; 1884 1885 #ifdef DIAGNOSTIC 1886 if (ipprintfs) 1887 kprintf("forward: src %x dst %x ttl %x\n", 1888 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl); 1889 #endif 1890 1891 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) { 1892 ipstat.ips_cantforward++; 1893 m_freem(m); 1894 return; 1895 } 1896 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) { 1897 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1898 return; 1899 } 1900 1901 ipforward_rtaddr = (struct sockaddr_in *) &cache_rt->ro_dst; 1902 if (cache_rt->ro_rt == NULL || 1903 ipforward_rtaddr->sin_addr.s_addr != pkt_dst.s_addr) { 1904 if (cache_rt->ro_rt != NULL) { 1905 RTFREE(cache_rt->ro_rt); 1906 cache_rt->ro_rt = NULL; 1907 } 1908 ipforward_rtaddr->sin_family = AF_INET; 1909 ipforward_rtaddr->sin_len = sizeof(struct sockaddr_in); 1910 ipforward_rtaddr->sin_addr = pkt_dst; 1911 rtalloc_ign(cache_rt, RTF_PRCLONING); 1912 if (cache_rt->ro_rt == NULL) { 1913 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1914 return; 1915 } 1916 } 1917 rt = cache_rt->ro_rt; 1918 1919 /* 1920 * Save the IP header and at most 8 bytes of the payload, 1921 * in case we need to generate an ICMP message to the src. 1922 * 1923 * XXX this can be optimized a lot by saving the data in a local 1924 * buffer on the stack (72 bytes at most), and only allocating the 1925 * mbuf if really necessary. The vast majority of the packets 1926 * are forwarded without having to send an ICMP back (either 1927 * because unnecessary, or because rate limited), so we are 1928 * really we are wasting a lot of work here. 1929 * 1930 * We don't use m_copy() because it might return a reference 1931 * to a shared cluster. Both this function and ip_output() 1932 * assume exclusive access to the IP header in `m', so any 1933 * data in a cluster may change before we reach icmp_error(). 1934 */ 1935 MGETHDR(mcopy, MB_DONTWAIT, m->m_type); 1936 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, MB_DONTWAIT)) { 1937 /* 1938 * It's probably ok if the pkthdr dup fails (because 1939 * the deep copy of the tag chain failed), but for now 1940 * be conservative and just discard the copy since 1941 * code below may some day want the tags. 1942 */ 1943 m_free(mcopy); 1944 mcopy = NULL; 1945 } 1946 if (mcopy != NULL) { 1947 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8, 1948 (int)ip->ip_len); 1949 mcopy->m_pkthdr.len = mcopy->m_len; 1950 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1951 } 1952 1953 if (!ipstealth) 1954 ip->ip_ttl -= IPTTLDEC; 1955 1956 /* 1957 * If forwarding packet using same interface that it came in on, 1958 * perhaps should send a redirect to sender to shortcut a hop. 1959 * Only send redirect if source is sending directly to us, 1960 * and if packet was not source routed (or has any options). 1961 * Also, don't send redirect if forwarding using a default route 1962 * or a route modified by a redirect. 1963 */ 1964 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1965 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) && 1966 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY && 1967 ipsendredirects && !using_srcrt && next_hop == NULL) { 1968 u_long src = ntohl(ip->ip_src.s_addr); 1969 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa; 1970 1971 if (rt_ifa != NULL && 1972 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) { 1973 if (rt->rt_flags & RTF_GATEWAY) 1974 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1975 else 1976 dest = pkt_dst.s_addr; 1977 /* 1978 * Router requirements says to only send 1979 * host redirects. 1980 */ 1981 type = ICMP_REDIRECT; 1982 code = ICMP_REDIRECT_HOST; 1983 #ifdef DIAGNOSTIC 1984 if (ipprintfs) 1985 kprintf("redirect (%d) to %x\n", code, dest); 1986 #endif 1987 } 1988 } 1989 1990 if (next_hop != NULL) { 1991 /* Pass IPFORWARD info if available */ 1992 tag.mh_type = MT_TAG; 1993 tag.mh_flags = PACKET_TAG_IPFORWARD; 1994 tag.mh_data = (caddr_t)next_hop; 1995 tag.mh_next = m; 1996 m = (struct mbuf *)&tag; 1997 } 1998 1999 error = ip_output(m, NULL, cache_rt, IP_FORWARDING, NULL, 2000 NULL); 2001 if (error == 0) { 2002 ipstat.ips_forward++; 2003 if (type == 0) { 2004 if (mcopy) { 2005 ipflow_create(cache_rt, mcopy); 2006 m_freem(mcopy); 2007 } 2008 return; /* most common case */ 2009 } else { 2010 ipstat.ips_redirectsent++; 2011 } 2012 } else { 2013 ipstat.ips_cantforward++; 2014 } 2015 2016 if (mcopy == NULL) 2017 return; 2018 2019 /* 2020 * Send ICMP message. 2021 */ 2022 2023 switch (error) { 2024 2025 case 0: /* forwarded, but need redirect */ 2026 /* type, code set above */ 2027 break; 2028 2029 case ENETUNREACH: /* shouldn't happen, checked above */ 2030 case EHOSTUNREACH: 2031 case ENETDOWN: 2032 case EHOSTDOWN: 2033 default: 2034 type = ICMP_UNREACH; 2035 code = ICMP_UNREACH_HOST; 2036 break; 2037 2038 case EMSGSIZE: 2039 type = ICMP_UNREACH; 2040 code = ICMP_UNREACH_NEEDFRAG; 2041 #ifdef IPSEC 2042 /* 2043 * If the packet is routed over IPsec tunnel, tell the 2044 * originator the tunnel MTU. 2045 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 2046 * XXX quickhack!!! 2047 */ 2048 if (cache_rt->ro_rt != NULL) { 2049 struct secpolicy *sp = NULL; 2050 int ipsecerror; 2051 int ipsechdr; 2052 struct route *ro; 2053 2054 sp = ipsec4_getpolicybyaddr(mcopy, 2055 IPSEC_DIR_OUTBOUND, 2056 IP_FORWARDING, 2057 &ipsecerror); 2058 2059 if (sp == NULL) 2060 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu; 2061 else { 2062 /* count IPsec header size */ 2063 ipsechdr = ipsec4_hdrsiz(mcopy, 2064 IPSEC_DIR_OUTBOUND, 2065 NULL); 2066 2067 /* 2068 * find the correct route for outer IPv4 2069 * header, compute tunnel MTU. 2070 * 2071 */ 2072 if (sp->req != NULL && sp->req->sav != NULL && 2073 sp->req->sav->sah != NULL) { 2074 ro = &sp->req->sav->sah->sa_route; 2075 if (ro->ro_rt != NULL && 2076 ro->ro_rt->rt_ifp != NULL) { 2077 destmtu = 2078 ro->ro_rt->rt_ifp->if_mtu; 2079 destmtu -= ipsechdr; 2080 } 2081 } 2082 2083 key_freesp(sp); 2084 } 2085 } 2086 #elif FAST_IPSEC 2087 /* 2088 * If the packet is routed over IPsec tunnel, tell the 2089 * originator the tunnel MTU. 2090 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 2091 * XXX quickhack!!! 2092 */ 2093 if (cache_rt->ro_rt != NULL) { 2094 struct secpolicy *sp = NULL; 2095 int ipsecerror; 2096 int ipsechdr; 2097 struct route *ro; 2098 2099 sp = ipsec_getpolicybyaddr(mcopy, 2100 IPSEC_DIR_OUTBOUND, 2101 IP_FORWARDING, 2102 &ipsecerror); 2103 2104 if (sp == NULL) 2105 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu; 2106 else { 2107 /* count IPsec header size */ 2108 ipsechdr = ipsec4_hdrsiz(mcopy, 2109 IPSEC_DIR_OUTBOUND, 2110 NULL); 2111 2112 /* 2113 * find the correct route for outer IPv4 2114 * header, compute tunnel MTU. 2115 */ 2116 2117 if (sp->req != NULL && 2118 sp->req->sav != NULL && 2119 sp->req->sav->sah != NULL) { 2120 ro = &sp->req->sav->sah->sa_route; 2121 if (ro->ro_rt != NULL && 2122 ro->ro_rt->rt_ifp != NULL) { 2123 destmtu = 2124 ro->ro_rt->rt_ifp->if_mtu; 2125 destmtu -= ipsechdr; 2126 } 2127 } 2128 2129 KEY_FREESP(&sp); 2130 } 2131 } 2132 #else /* !IPSEC && !FAST_IPSEC */ 2133 if (cache_rt->ro_rt != NULL) 2134 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu; 2135 #endif /*IPSEC*/ 2136 ipstat.ips_cantfrag++; 2137 break; 2138 2139 case ENOBUFS: 2140 /* 2141 * A router should not generate ICMP_SOURCEQUENCH as 2142 * required in RFC1812 Requirements for IP Version 4 Routers. 2143 * Source quench could be a big problem under DoS attacks, 2144 * or if the underlying interface is rate-limited. 2145 * Those who need source quench packets may re-enable them 2146 * via the net.inet.ip.sendsourcequench sysctl. 2147 */ 2148 if (!ip_sendsourcequench) { 2149 m_freem(mcopy); 2150 return; 2151 } else { 2152 type = ICMP_SOURCEQUENCH; 2153 code = 0; 2154 } 2155 break; 2156 2157 case EACCES: /* ipfw denied packet */ 2158 m_freem(mcopy); 2159 return; 2160 } 2161 icmp_error(mcopy, type, code, dest, destmtu); 2162 } 2163 2164 void 2165 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2166 struct mbuf *m) 2167 { 2168 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2169 struct timeval tv; 2170 2171 microtime(&tv); 2172 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2173 SCM_TIMESTAMP, SOL_SOCKET); 2174 if (*mp) 2175 mp = &(*mp)->m_next; 2176 } 2177 if (inp->inp_flags & INP_RECVDSTADDR) { 2178 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2179 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2180 if (*mp) 2181 mp = &(*mp)->m_next; 2182 } 2183 if (inp->inp_flags & INP_RECVTTL) { 2184 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 2185 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 2186 if (*mp) 2187 mp = &(*mp)->m_next; 2188 } 2189 #ifdef notyet 2190 /* XXX 2191 * Moving these out of udp_input() made them even more broken 2192 * than they already were. 2193 */ 2194 /* options were tossed already */ 2195 if (inp->inp_flags & INP_RECVOPTS) { 2196 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2197 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2198 if (*mp) 2199 mp = &(*mp)->m_next; 2200 } 2201 /* ip_srcroute doesn't do what we want here, need to fix */ 2202 if (inp->inp_flags & INP_RECVRETOPTS) { 2203 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 2204 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2205 if (*mp) 2206 mp = &(*mp)->m_next; 2207 } 2208 #endif 2209 if (inp->inp_flags & INP_RECVIF) { 2210 struct ifnet *ifp; 2211 struct sdlbuf { 2212 struct sockaddr_dl sdl; 2213 u_char pad[32]; 2214 } sdlbuf; 2215 struct sockaddr_dl *sdp; 2216 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2217 2218 if (((ifp = m->m_pkthdr.rcvif)) && 2219 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) { 2220 sdp = IF_LLSOCKADDR(ifp); 2221 /* 2222 * Change our mind and don't try copy. 2223 */ 2224 if ((sdp->sdl_family != AF_LINK) || 2225 (sdp->sdl_len > sizeof(sdlbuf))) { 2226 goto makedummy; 2227 } 2228 bcopy(sdp, sdl2, sdp->sdl_len); 2229 } else { 2230 makedummy: 2231 sdl2->sdl_len = 2232 offsetof(struct sockaddr_dl, sdl_data[0]); 2233 sdl2->sdl_family = AF_LINK; 2234 sdl2->sdl_index = 0; 2235 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2236 } 2237 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 2238 IP_RECVIF, IPPROTO_IP); 2239 if (*mp) 2240 mp = &(*mp)->m_next; 2241 } 2242 } 2243 2244 /* 2245 * XXX these routines are called from the upper part of the kernel. 2246 * 2247 * They could also be moved to ip_mroute.c, since all the RSVP 2248 * handling is done there already. 2249 */ 2250 int 2251 ip_rsvp_init(struct socket *so) 2252 { 2253 if (so->so_type != SOCK_RAW || 2254 so->so_proto->pr_protocol != IPPROTO_RSVP) 2255 return EOPNOTSUPP; 2256 2257 if (ip_rsvpd != NULL) 2258 return EADDRINUSE; 2259 2260 ip_rsvpd = so; 2261 /* 2262 * This may seem silly, but we need to be sure we don't over-increment 2263 * the RSVP counter, in case something slips up. 2264 */ 2265 if (!ip_rsvp_on) { 2266 ip_rsvp_on = 1; 2267 rsvp_on++; 2268 } 2269 2270 return 0; 2271 } 2272 2273 int 2274 ip_rsvp_done(void) 2275 { 2276 ip_rsvpd = NULL; 2277 /* 2278 * This may seem silly, but we need to be sure we don't over-decrement 2279 * the RSVP counter, in case something slips up. 2280 */ 2281 if (ip_rsvp_on) { 2282 ip_rsvp_on = 0; 2283 rsvp_on--; 2284 } 2285 return 0; 2286 } 2287 2288 void 2289 rsvp_input(struct mbuf *m, ...) /* XXX must fixup manually */ 2290 { 2291 int off, proto; 2292 __va_list ap; 2293 2294 __va_start(ap, m); 2295 off = __va_arg(ap, int); 2296 proto = __va_arg(ap, int); 2297 __va_end(ap); 2298 2299 if (rsvp_input_p) { /* call the real one if loaded */ 2300 rsvp_input_p(m, off, proto); 2301 return; 2302 } 2303 2304 /* Can still get packets with rsvp_on = 0 if there is a local member 2305 * of the group to which the RSVP packet is addressed. But in this 2306 * case we want to throw the packet away. 2307 */ 2308 2309 if (!rsvp_on) { 2310 m_freem(m); 2311 return; 2312 } 2313 2314 if (ip_rsvpd != NULL) { 2315 rip_input(m, off, proto); 2316 return; 2317 } 2318 /* Drop the packet */ 2319 m_freem(m); 2320 } 2321