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. Neither the name of the University nor the names of its contributors 47 * may be used to endorse or promote products derived from this software 48 * without specific prior written permission. 49 * 50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 60 * SUCH DAMAGE. 61 * 62 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 63 * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $ 64 */ 65 66 #define _IP_VHL 67 68 #include "opt_bootp.h" 69 #include "opt_ipdn.h" 70 #include "opt_ipdivert.h" 71 #include "opt_ipstealth.h" 72 #include "opt_ipsec.h" 73 #include "opt_rss.h" 74 75 #include <sys/param.h> 76 #include <sys/systm.h> 77 #include <sys/mbuf.h> 78 #include <sys/malloc.h> 79 #include <sys/mpipe.h> 80 #include <sys/domain.h> 81 #include <sys/protosw.h> 82 #include <sys/socket.h> 83 #include <sys/time.h> 84 #include <sys/globaldata.h> 85 #include <sys/thread.h> 86 #include <sys/kernel.h> 87 #include <sys/syslog.h> 88 #include <sys/sysctl.h> 89 #include <sys/in_cksum.h> 90 #include <sys/lock.h> 91 92 #include <sys/mplock2.h> 93 94 #include <machine/stdarg.h> 95 96 #include <net/if.h> 97 #include <net/if_types.h> 98 #include <net/if_var.h> 99 #include <net/if_dl.h> 100 #include <net/pfil.h> 101 #include <net/route.h> 102 #include <net/netisr2.h> 103 104 #include <netinet/in.h> 105 #include <netinet/in_systm.h> 106 #include <netinet/in_var.h> 107 #include <netinet/ip.h> 108 #include <netinet/in_pcb.h> 109 #include <netinet/ip_var.h> 110 #include <netinet/ip_icmp.h> 111 #include <netinet/ip_divert.h> 112 #include <netinet/ip_flow.h> 113 114 #include <sys/thread2.h> 115 #include <sys/msgport2.h> 116 #include <net/netmsg2.h> 117 118 #include <sys/socketvar.h> 119 120 #include <net/ipfw/ip_fw.h> 121 #include <net/dummynet/ip_dummynet.h> 122 123 #ifdef IPSEC 124 #include <netinet6/ipsec.h> 125 #include <netproto/key/key.h> 126 #endif 127 128 #ifdef FAST_IPSEC 129 #include <netproto/ipsec/ipsec.h> 130 #include <netproto/ipsec/key.h> 131 #endif 132 133 int rsvp_on = 0; 134 static int ip_rsvp_on; 135 struct socket *ip_rsvpd; 136 137 int ipforwarding = 0; 138 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 139 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 140 141 static int ipsendredirects = 1; /* XXX */ 142 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 143 &ipsendredirects, 0, "Enable sending IP redirects"); 144 145 int ip_defttl = IPDEFTTL; 146 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 147 &ip_defttl, 0, "Maximum TTL on IP packets"); 148 149 static int ip_dosourceroute = 0; 150 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 151 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 152 153 static int ip_acceptsourceroute = 0; 154 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 155 CTLFLAG_RW, &ip_acceptsourceroute, 0, 156 "Enable accepting source routed IP packets"); 157 158 static int ip_keepfaith = 0; 159 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 160 &ip_keepfaith, 0, 161 "Enable packet capture for FAITH IPv4->IPv6 translator daemon"); 162 163 static int nipq = 0; /* total # of reass queues */ 164 static int maxnipq; 165 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 166 &maxnipq, 0, 167 "Maximum number of IPv4 fragment reassembly queue entries"); 168 169 static int maxfragsperpacket; 170 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 171 &maxfragsperpacket, 0, 172 "Maximum number of IPv4 fragments allowed per packet"); 173 174 static int ip_sendsourcequench = 0; 175 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 176 &ip_sendsourcequench, 0, 177 "Enable the transmission of source quench packets"); 178 179 int ip_do_randomid = 1; 180 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 181 &ip_do_randomid, 0, 182 "Assign random ip_id values"); 183 /* 184 * XXX - Setting ip_checkinterface mostly implements the receive side of 185 * the Strong ES model described in RFC 1122, but since the routing table 186 * and transmit implementation do not implement the Strong ES model, 187 * setting this to 1 results in an odd hybrid. 188 * 189 * XXX - ip_checkinterface currently must be disabled if you use ipnat 190 * to translate the destination address to another local interface. 191 * 192 * XXX - ip_checkinterface must be disabled if you add IP aliases 193 * to the loopback interface instead of the interface where the 194 * packets for those addresses are received. 195 */ 196 static int ip_checkinterface = 0; 197 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 198 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 199 200 static u_long ip_hash_count = 0; 201 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, hash_count, CTLFLAG_RD, 202 &ip_hash_count, 0, "Number of packets hashed by IP"); 203 204 #ifdef RSS_DEBUG 205 static u_long ip_rehash_count = 0; 206 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, rehash_count, CTLFLAG_RD, 207 &ip_rehash_count, 0, "Number of packets rehashed by IP"); 208 209 static u_long ip_dispatch_fast = 0; 210 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_fast_count, CTLFLAG_RD, 211 &ip_dispatch_fast, 0, "Number of packets handled on current CPU"); 212 213 static u_long ip_dispatch_slow = 0; 214 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_slow_count, CTLFLAG_RD, 215 &ip_dispatch_slow, 0, "Number of packets messaged to another CPU"); 216 #endif 217 218 static struct lwkt_token ipq_token = LWKT_TOKEN_INITIALIZER(ipq_token); 219 220 #ifdef DIAGNOSTIC 221 static int ipprintfs = 0; 222 #endif 223 224 extern struct domain inetdomain; 225 extern struct protosw inetsw[]; 226 u_char ip_protox[IPPROTO_MAX]; 227 struct in_ifaddrhead in_ifaddrheads[MAXCPU]; /* first inet address */ 228 struct in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU]; 229 /* inet addr hash table */ 230 u_long in_ifaddrhmask; /* mask for hash table */ 231 232 static struct mbuf *ipforward_mtemp[MAXCPU]; 233 234 struct ip_stats ipstats_percpu[MAXCPU] __cachealign; 235 236 static int 237 sysctl_ipstats(SYSCTL_HANDLER_ARGS) 238 { 239 int cpu, error = 0; 240 241 for (cpu = 0; cpu < ncpus; ++cpu) { 242 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu], 243 sizeof(struct ip_stats)))) 244 break; 245 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu], 246 sizeof(struct ip_stats)))) 247 break; 248 } 249 250 return (error); 251 } 252 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW), 253 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics"); 254 255 /* Packet reassembly stuff */ 256 #define IPREASS_NHASH_LOG2 6 257 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 258 #define IPREASS_HMASK (IPREASS_NHASH - 1) 259 #define IPREASS_HASH(x,y) \ 260 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 261 262 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 263 264 #ifdef IPCTL_DEFMTU 265 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 266 &ip_mtu, 0, "Default MTU"); 267 #endif 268 269 #ifdef IPSTEALTH 270 static int ipstealth = 0; 271 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, ""); 272 #else 273 static const int ipstealth = 0; 274 #endif 275 276 struct mbuf *(*ip_divert_p)(struct mbuf *, int, int); 277 278 struct pfil_head inet_pfil_hook; 279 280 /* 281 * struct ip_srcrt_opt is used to store packet state while it travels 282 * through the stack. 283 * 284 * XXX Note that the code even makes assumptions on the size and 285 * alignment of fields inside struct ip_srcrt so e.g. adding some 286 * fields will break the code. This needs to be fixed. 287 * 288 * We need to save the IP options in case a protocol wants to respond 289 * to an incoming packet over the same route if the packet got here 290 * using IP source routing. This allows connection establishment and 291 * maintenance when the remote end is on a network that is not known 292 * to us. 293 */ 294 struct ip_srcrt { 295 struct in_addr dst; /* final destination */ 296 char nop; /* one NOP to align */ 297 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 298 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 299 }; 300 301 struct ip_srcrt_opt { 302 int ip_nhops; 303 struct ip_srcrt ip_srcrt; 304 }; 305 306 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management"); 307 static struct malloc_pipe ipq_mpipe; 308 309 static void save_rte(struct mbuf *, u_char *, struct in_addr); 310 static int ip_dooptions(struct mbuf *m, int, struct sockaddr_in *); 311 static void ip_freef(struct ipqhead *, struct ipq *); 312 static void ip_input_handler(netmsg_t); 313 314 /* 315 * IP initialization: fill in IP protocol switch table. 316 * All protocols not implemented in kernel go to raw IP protocol handler. 317 */ 318 void 319 ip_init(void) 320 { 321 struct protosw *pr; 322 int i; 323 int cpu; 324 325 /* 326 * Make sure we can handle a reasonable number of fragments but 327 * cap it at 4000 (XXX). 328 */ 329 mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq), 330 IFQ_MAXLEN, 4000, 0, NULL, NULL, NULL); 331 for (i = 0; i < ncpus; ++i) { 332 TAILQ_INIT(&in_ifaddrheads[i]); 333 in_ifaddrhashtbls[i] = 334 hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 335 } 336 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 337 if (pr == NULL) 338 panic("ip_init"); 339 for (i = 0; i < IPPROTO_MAX; i++) 340 ip_protox[i] = pr - inetsw; 341 for (pr = inetdomain.dom_protosw; 342 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 343 if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol) { 344 if (pr->pr_protocol != IPPROTO_RAW) 345 ip_protox[pr->pr_protocol] = pr - inetsw; 346 } 347 } 348 349 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 350 inet_pfil_hook.ph_af = AF_INET; 351 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) { 352 kprintf("%s: WARNING: unable to register pfil hook, " 353 "error %d\n", __func__, i); 354 } 355 356 for (i = 0; i < IPREASS_NHASH; i++) 357 TAILQ_INIT(&ipq[i]); 358 359 maxnipq = nmbclusters / 32; 360 maxfragsperpacket = 16; 361 362 ip_id = time_second & 0xffff; /* time_second survives reboots */ 363 364 for (cpu = 0; cpu < ncpus; ++cpu) { 365 /* 366 * Initialize IP statistics counters for each CPU. 367 */ 368 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats)); 369 370 /* 371 * Preallocate mbuf template for forwarding 372 */ 373 MGETHDR(ipforward_mtemp[cpu], MB_WAIT, MT_DATA); 374 } 375 376 netisr_register(NETISR_IP, ip_input_handler, ip_hashfn_in); 377 netisr_register_hashcheck(NETISR_IP, ip_hashcheck); 378 } 379 380 /* Do transport protocol processing. */ 381 static void 382 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip) 383 { 384 const struct protosw *pr = &inetsw[ip_protox[ip->ip_p]]; 385 386 /* 387 * Switch out to protocol's input routine. 388 */ 389 PR_GET_MPLOCK(pr); 390 pr->pr_input(&m, &hlen, ip->ip_p); 391 PR_REL_MPLOCK(pr); 392 } 393 394 static void 395 transport_processing_handler(netmsg_t msg) 396 { 397 struct netmsg_packet *pmsg = &msg->packet; 398 struct ip *ip; 399 int hlen; 400 401 ip = mtod(pmsg->nm_packet, struct ip *); 402 hlen = pmsg->base.lmsg.u.ms_result; 403 404 transport_processing_oncpu(pmsg->nm_packet, hlen, ip); 405 /* msg was embedded in the mbuf, do not reply! */ 406 } 407 408 static void 409 ip_input_handler(netmsg_t msg) 410 { 411 ip_input(msg->packet.nm_packet); 412 /* msg was embedded in the mbuf, do not reply! */ 413 } 414 415 /* 416 * IP input routine. Checksum and byte swap header. If fragmented 417 * try to reassemble. Process options. Pass to next level. 418 */ 419 void 420 ip_input(struct mbuf *m) 421 { 422 struct ip *ip; 423 struct in_ifaddr *ia = NULL; 424 struct in_ifaddr_container *iac; 425 int hlen, checkif; 426 u_short sum; 427 struct in_addr pkt_dst; 428 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */ 429 struct in_addr odst; /* original dst address(NAT) */ 430 struct m_tag *mtag; 431 struct sockaddr_in *next_hop = NULL; 432 lwkt_port_t port; 433 #ifdef FAST_IPSEC 434 struct tdb_ident *tdbi; 435 struct secpolicy *sp; 436 int error; 437 #endif 438 439 M_ASSERTPKTHDR(m); 440 441 /* 442 * This routine is called from numerous places which may not have 443 * characterized the packet. 444 */ 445 if ((m->m_flags & M_HASH) == 0) { 446 atomic_add_long(&ip_hash_count, 1); 447 ip_hashfn(&m, 0, IP_MPORT_IN); 448 if (m == NULL) 449 return; 450 KKASSERT(m->m_flags & M_HASH); 451 } 452 ip = mtod(m, struct ip *); 453 454 /* 455 * Pull out certain tags 456 */ 457 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 458 /* Next hop */ 459 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 460 KKASSERT(mtag != NULL); 461 next_hop = m_tag_data(mtag); 462 } 463 464 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 465 /* dummynet already filtered us */ 466 ip = mtod(m, struct ip *); 467 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 468 goto iphack; 469 } 470 471 ipstat.ips_total++; 472 473 /* length checks already done in ip_hashfn() */ 474 KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf")); 475 476 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 477 ipstat.ips_badvers++; 478 goto bad; 479 } 480 481 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 482 /* length checks already done in ip_hashfn() */ 483 KASSERT(hlen >= sizeof(struct ip), ("IP header len too small")); 484 KASSERT(m->m_len >= hlen, ("complete IP header not in one mbuf")); 485 486 /* 127/8 must not appear on wire - RFC1122 */ 487 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 488 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 489 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) { 490 ipstat.ips_badaddr++; 491 goto bad; 492 } 493 } 494 495 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 496 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 497 } else { 498 if (hlen == sizeof(struct ip)) 499 sum = in_cksum_hdr(ip); 500 else 501 sum = in_cksum(m, hlen); 502 } 503 if (sum != 0) { 504 ipstat.ips_badsum++; 505 goto bad; 506 } 507 508 #ifdef ALTQ 509 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 510 /* packet is dropped by traffic conditioner */ 511 return; 512 } 513 #endif 514 /* 515 * Convert fields to host representation. 516 */ 517 ip->ip_len = ntohs(ip->ip_len); 518 ip->ip_off = ntohs(ip->ip_off); 519 520 /* length checks already done in ip_hashfn() */ 521 KASSERT(ip->ip_len >= hlen, ("total length less then header length")); 522 KASSERT(m->m_pkthdr.len >= ip->ip_len, ("mbuf too short")); 523 524 /* 525 * Trim mbufs if longer than the IP header would have us expect. 526 */ 527 if (m->m_pkthdr.len > ip->ip_len) { 528 if (m->m_len == m->m_pkthdr.len) { 529 m->m_len = ip->ip_len; 530 m->m_pkthdr.len = ip->ip_len; 531 } else { 532 m_adj(m, ip->ip_len - m->m_pkthdr.len); 533 } 534 } 535 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF) 536 /* 537 * Bypass packet filtering for packets from a tunnel (gif). 538 */ 539 if (ipsec_gethist(m, NULL)) 540 goto pass; 541 #endif 542 543 /* 544 * IpHack's section. 545 * Right now when no processing on packet has done 546 * and it is still fresh out of network we do our black 547 * deals with it. 548 * - Firewall: deny/allow/divert 549 * - Xlate: translate packet's addr/port (NAT). 550 * - Pipe: pass pkt through dummynet. 551 * - Wrap: fake packet's addr/port <unimpl.> 552 * - Encapsulate: put it in another IP and send out. <unimp.> 553 */ 554 555 iphack: 556 /* 557 * If we've been forwarded from the output side, then 558 * skip the firewall a second time 559 */ 560 if (next_hop != NULL) 561 goto ours; 562 563 /* No pfil hooks */ 564 if (!pfil_has_hooks(&inet_pfil_hook)) { 565 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 566 /* 567 * Strip dummynet tags from stranded packets 568 */ 569 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 570 KKASSERT(mtag != NULL); 571 m_tag_delete(m, mtag); 572 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 573 } 574 goto pass; 575 } 576 577 /* 578 * Run through list of hooks for input packets. 579 * 580 * NOTE! If the packet is rewritten pf/ipfw/whoever must 581 * clear M_HASH. 582 */ 583 odst = ip->ip_dst; 584 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN)) 585 return; 586 if (m == NULL) /* consumed by filter */ 587 return; 588 ip = mtod(m, struct ip *); 589 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 590 using_srcrt = (odst.s_addr != ip->ip_dst.s_addr); 591 592 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 593 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 594 KKASSERT(mtag != NULL); 595 next_hop = m_tag_data(mtag); 596 } 597 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 598 ip_dn_queue(m); 599 return; 600 } 601 if (m->m_pkthdr.fw_flags & FW_MBUF_REDISPATCH) { 602 m->m_pkthdr.fw_flags &= ~FW_MBUF_REDISPATCH; 603 } 604 pass: 605 /* 606 * Process options and, if not destined for us, 607 * ship it on. ip_dooptions returns 1 when an 608 * error was detected (causing an icmp message 609 * to be sent and the original packet to be freed). 610 */ 611 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, next_hop)) 612 return; 613 614 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 615 * matter if it is destined to another node, or whether it is 616 * a multicast one, RSVP wants it! and prevents it from being forwarded 617 * anywhere else. Also checks if the rsvp daemon is running before 618 * grabbing the packet. 619 */ 620 if (rsvp_on && ip->ip_p == IPPROTO_RSVP) 621 goto ours; 622 623 /* 624 * Check our list of addresses, to see if the packet is for us. 625 * If we don't have any addresses, assume any unicast packet 626 * we receive might be for us (and let the upper layers deal 627 * with it). 628 */ 629 if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) && 630 !(m->m_flags & (M_MCAST | M_BCAST))) 631 goto ours; 632 633 /* 634 * Cache the destination address of the packet; this may be 635 * changed by use of 'ipfw fwd'. 636 */ 637 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst; 638 639 /* 640 * Enable a consistency check between the destination address 641 * and the arrival interface for a unicast packet (the RFC 1122 642 * strong ES model) if IP forwarding is disabled and the packet 643 * is not locally generated and the packet is not subject to 644 * 'ipfw fwd'. 645 * 646 * XXX - Checking also should be disabled if the destination 647 * address is ipnat'ed to a different interface. 648 * 649 * XXX - Checking is incompatible with IP aliases added 650 * to the loopback interface instead of the interface where 651 * the packets are received. 652 */ 653 checkif = ip_checkinterface && 654 !ipforwarding && 655 m->m_pkthdr.rcvif != NULL && 656 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) && 657 next_hop == NULL; 658 659 /* 660 * Check for exact addresses in the hash bucket. 661 */ 662 LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) { 663 ia = iac->ia; 664 665 /* 666 * If the address matches, verify that the packet 667 * arrived via the correct interface if checking is 668 * enabled. 669 */ 670 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 671 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 672 goto ours; 673 } 674 ia = NULL; 675 676 /* 677 * Check for broadcast addresses. 678 * 679 * Only accept broadcast packets that arrive via the matching 680 * interface. Reception of forwarded directed broadcasts would 681 * be handled via ip_forward() and ether_output() with the loopback 682 * into the stack for SIMPLEX interfaces handled by ether_output(). 683 */ 684 if (m->m_pkthdr.rcvif != NULL && 685 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 686 struct ifaddr_container *ifac; 687 688 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid], 689 ifa_link) { 690 struct ifaddr *ifa = ifac->ifa; 691 692 if (ifa->ifa_addr == NULL) /* shutdown/startup race */ 693 continue; 694 if (ifa->ifa_addr->sa_family != AF_INET) 695 continue; 696 ia = ifatoia(ifa); 697 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 698 pkt_dst.s_addr) 699 goto ours; 700 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 701 goto ours; 702 #ifdef BOOTP_COMPAT 703 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 704 goto ours; 705 #endif 706 } 707 } 708 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 709 struct in_multi *inm; 710 711 /* XXX Multicast is not MPSAFE yet */ 712 get_mplock(); 713 714 if (ip_mrouter != NULL) { 715 /* 716 * If we are acting as a multicast router, all 717 * incoming multicast packets are passed to the 718 * kernel-level multicast forwarding function. 719 * The packet is returned (relatively) intact; if 720 * ip_mforward() returns a non-zero value, the packet 721 * must be discarded, else it may be accepted below. 722 */ 723 if (ip_mforward != NULL && 724 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) { 725 rel_mplock(); 726 ipstat.ips_cantforward++; 727 m_freem(m); 728 return; 729 } 730 731 /* 732 * The process-level routing daemon needs to receive 733 * all multicast IGMP packets, whether or not this 734 * host belongs to their destination groups. 735 */ 736 if (ip->ip_p == IPPROTO_IGMP) { 737 rel_mplock(); 738 goto ours; 739 } 740 ipstat.ips_forward++; 741 } 742 /* 743 * See if we belong to the destination multicast group on the 744 * arrival interface. 745 */ 746 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 747 if (inm == NULL) { 748 rel_mplock(); 749 ipstat.ips_notmember++; 750 m_freem(m); 751 return; 752 } 753 754 rel_mplock(); 755 goto ours; 756 } 757 if (ip->ip_dst.s_addr == INADDR_BROADCAST) 758 goto ours; 759 if (ip->ip_dst.s_addr == INADDR_ANY) 760 goto ours; 761 762 /* 763 * FAITH(Firewall Aided Internet Translator) 764 */ 765 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 766 if (ip_keepfaith) { 767 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 768 goto ours; 769 } 770 m_freem(m); 771 return; 772 } 773 774 /* 775 * Not for us; forward if possible and desirable. 776 */ 777 if (!ipforwarding) { 778 ipstat.ips_cantforward++; 779 m_freem(m); 780 } else { 781 #ifdef IPSEC 782 /* 783 * Enforce inbound IPsec SPD. 784 */ 785 if (ipsec4_in_reject(m, NULL)) { 786 ipsecstat.in_polvio++; 787 goto bad; 788 } 789 #endif 790 #ifdef FAST_IPSEC 791 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 792 crit_enter(); 793 if (mtag != NULL) { 794 tdbi = (struct tdb_ident *)m_tag_data(mtag); 795 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 796 } else { 797 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 798 IP_FORWARDING, &error); 799 } 800 if (sp == NULL) { /* NB: can happen if error */ 801 crit_exit(); 802 /*XXX error stat???*/ 803 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 804 goto bad; 805 } 806 807 /* 808 * Check security policy against packet attributes. 809 */ 810 error = ipsec_in_reject(sp, m); 811 KEY_FREESP(&sp); 812 crit_exit(); 813 if (error) { 814 ipstat.ips_cantforward++; 815 goto bad; 816 } 817 #endif 818 ip_forward(m, using_srcrt, next_hop); 819 } 820 return; 821 822 ours: 823 824 /* 825 * IPSTEALTH: Process non-routing options only 826 * if the packet is destined for us. 827 */ 828 if (ipstealth && 829 hlen > sizeof(struct ip) && 830 ip_dooptions(m, 1, next_hop)) 831 return; 832 833 /* Count the packet in the ip address stats */ 834 if (ia != NULL) { 835 IFA_STAT_INC(&ia->ia_ifa, ipackets, 1); 836 IFA_STAT_INC(&ia->ia_ifa, ibytes, m->m_pkthdr.len); 837 } 838 839 /* 840 * If offset or IP_MF are set, must reassemble. 841 * Otherwise, nothing need be done. 842 * (We could look in the reassembly queue to see 843 * if the packet was previously fragmented, 844 * but it's not worth the time; just let them time out.) 845 */ 846 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 847 /* 848 * Attempt reassembly; if it succeeds, proceed. ip_reass() 849 * will return a different mbuf. 850 * 851 * NOTE: ip_reass() returns m with M_HASH cleared to force 852 * us to recharacterize the packet. 853 */ 854 m = ip_reass(m); 855 if (m == NULL) 856 return; 857 ip = mtod(m, struct ip *); 858 859 /* Get the header length of the reassembled packet */ 860 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 861 } else { 862 ip->ip_len -= hlen; 863 } 864 865 #ifdef IPSEC 866 /* 867 * enforce IPsec policy checking if we are seeing last header. 868 * note that we do not visit this with protocols with pcb layer 869 * code - like udp/tcp/raw ip. 870 */ 871 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) && 872 ipsec4_in_reject(m, NULL)) { 873 ipsecstat.in_polvio++; 874 goto bad; 875 } 876 #endif 877 #if FAST_IPSEC 878 /* 879 * enforce IPsec policy checking if we are seeing last header. 880 * note that we do not visit this with protocols with pcb layer 881 * code - like udp/tcp/raw ip. 882 */ 883 if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) { 884 /* 885 * Check if the packet has already had IPsec processing 886 * done. If so, then just pass it along. This tag gets 887 * set during AH, ESP, etc. input handling, before the 888 * packet is returned to the ip input queue for delivery. 889 */ 890 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 891 crit_enter(); 892 if (mtag != NULL) { 893 tdbi = (struct tdb_ident *)m_tag_data(mtag); 894 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 895 } else { 896 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 897 IP_FORWARDING, &error); 898 } 899 if (sp != NULL) { 900 /* 901 * Check security policy against packet attributes. 902 */ 903 error = ipsec_in_reject(sp, m); 904 KEY_FREESP(&sp); 905 } else { 906 /* XXX error stat??? */ 907 error = EINVAL; 908 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 909 goto bad; 910 } 911 crit_exit(); 912 if (error) 913 goto bad; 914 } 915 #endif /* FAST_IPSEC */ 916 917 /* 918 * We must forward the packet to the correct protocol thread if 919 * we are not already in it. 920 * 921 * NOTE: ip_len is now in host form. ip_len is not adjusted 922 * further for protocol processing, instead we pass hlen 923 * to the protosw and let it deal with it. 924 */ 925 ipstat.ips_delivered++; 926 927 if ((m->m_flags & M_HASH) == 0) { 928 #ifdef RSS_DEBUG 929 atomic_add_long(&ip_rehash_count, 1); 930 #endif 931 ip->ip_len = htons(ip->ip_len + hlen); 932 ip->ip_off = htons(ip->ip_off); 933 934 ip_hashfn(&m, 0, IP_MPORT_IN); 935 if (m == NULL) 936 return; 937 938 ip = mtod(m, struct ip *); 939 ip->ip_len = ntohs(ip->ip_len) - hlen; 940 ip->ip_off = ntohs(ip->ip_off); 941 KKASSERT(m->m_flags & M_HASH); 942 } 943 port = netisr_hashport(m->m_pkthdr.hash); 944 945 if (port != &curthread->td_msgport) { 946 struct netmsg_packet *pmsg; 947 948 #ifdef RSS_DEBUG 949 atomic_add_long(&ip_dispatch_slow, 1); 950 #endif 951 952 pmsg = &m->m_hdr.mh_netmsg; 953 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 954 0, transport_processing_handler); 955 pmsg->nm_packet = m; 956 pmsg->base.lmsg.u.ms_result = hlen; 957 lwkt_sendmsg(port, &pmsg->base.lmsg); 958 } else { 959 #ifdef RSS_DEBUG 960 atomic_add_long(&ip_dispatch_fast, 1); 961 #endif 962 transport_processing_oncpu(m, hlen, ip); 963 } 964 return; 965 966 bad: 967 m_freem(m); 968 } 969 970 /* 971 * Take incoming datagram fragment and try to reassemble it into 972 * whole datagram. If a chain for reassembly of this datagram already 973 * exists, then it is given as fp; otherwise have to make a chain. 974 */ 975 struct mbuf * 976 ip_reass(struct mbuf *m) 977 { 978 struct ip *ip = mtod(m, struct ip *); 979 struct mbuf *p = NULL, *q, *nq; 980 struct mbuf *n; 981 struct ipq *fp = NULL; 982 struct ipqhead *head; 983 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 984 int i, next; 985 u_short sum; 986 987 /* If maxnipq is 0, never accept fragments. */ 988 if (maxnipq == 0) { 989 ipstat.ips_fragments++; 990 ipstat.ips_fragdropped++; 991 m_freem(m); 992 return NULL; 993 } 994 995 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 996 /* 997 * Look for queue of fragments of this datagram. 998 */ 999 lwkt_gettoken(&ipq_token); 1000 head = &ipq[sum]; 1001 TAILQ_FOREACH(fp, head, ipq_list) { 1002 if (ip->ip_id == fp->ipq_id && 1003 ip->ip_src.s_addr == fp->ipq_src.s_addr && 1004 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 1005 ip->ip_p == fp->ipq_p) 1006 goto found; 1007 } 1008 1009 fp = NULL; 1010 1011 /* 1012 * Enforce upper bound on number of fragmented packets 1013 * for which we attempt reassembly; 1014 * If maxnipq is -1, accept all fragments without limitation. 1015 */ 1016 if (nipq > maxnipq && maxnipq > 0) { 1017 /* 1018 * drop something from the tail of the current queue 1019 * before proceeding further 1020 */ 1021 struct ipq *q = TAILQ_LAST(head, ipqhead); 1022 if (q == NULL) { 1023 /* 1024 * The current queue is empty, 1025 * so drop from one of the others. 1026 */ 1027 for (i = 0; i < IPREASS_NHASH; i++) { 1028 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 1029 if (r) { 1030 ipstat.ips_fragtimeout += r->ipq_nfrags; 1031 ip_freef(&ipq[i], r); 1032 break; 1033 } 1034 } 1035 } else { 1036 ipstat.ips_fragtimeout += q->ipq_nfrags; 1037 ip_freef(head, q); 1038 } 1039 } 1040 found: 1041 /* 1042 * Adjust ip_len to not reflect header, 1043 * convert offset of this to bytes. 1044 */ 1045 ip->ip_len -= hlen; 1046 if (ip->ip_off & IP_MF) { 1047 /* 1048 * Make sure that fragments have a data length 1049 * that's a non-zero multiple of 8 bytes. 1050 */ 1051 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 1052 ipstat.ips_toosmall++; /* XXX */ 1053 m_freem(m); 1054 goto done; 1055 } 1056 m->m_flags |= M_FRAG; 1057 } else { 1058 m->m_flags &= ~M_FRAG; 1059 } 1060 ip->ip_off <<= 3; 1061 1062 ipstat.ips_fragments++; 1063 m->m_pkthdr.header = ip; 1064 1065 /* 1066 * If the hardware has not done csum over this fragment 1067 * then csum_data is not valid at all. 1068 */ 1069 if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) 1070 == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) { 1071 m->m_pkthdr.csum_data = 0; 1072 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 1073 } 1074 1075 /* 1076 * Presence of header sizes in mbufs 1077 * would confuse code below. 1078 */ 1079 m->m_data += hlen; 1080 m->m_len -= hlen; 1081 1082 /* 1083 * If first fragment to arrive, create a reassembly queue. 1084 */ 1085 if (fp == NULL) { 1086 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL) 1087 goto dropfrag; 1088 TAILQ_INSERT_HEAD(head, fp, ipq_list); 1089 nipq++; 1090 fp->ipq_nfrags = 1; 1091 fp->ipq_ttl = IPFRAGTTL; 1092 fp->ipq_p = ip->ip_p; 1093 fp->ipq_id = ip->ip_id; 1094 fp->ipq_src = ip->ip_src; 1095 fp->ipq_dst = ip->ip_dst; 1096 fp->ipq_frags = m; 1097 m->m_nextpkt = NULL; 1098 goto inserted; 1099 } else { 1100 fp->ipq_nfrags++; 1101 } 1102 1103 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1104 1105 /* 1106 * Find a segment which begins after this one does. 1107 */ 1108 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1109 if (GETIP(q)->ip_off > ip->ip_off) 1110 break; 1111 } 1112 1113 /* 1114 * If there is a preceding segment, it may provide some of 1115 * our data already. If so, drop the data from the incoming 1116 * segment. If it provides all of our data, drop us, otherwise 1117 * stick new segment in the proper place. 1118 * 1119 * If some of the data is dropped from the the preceding 1120 * segment, then it's checksum is invalidated. 1121 */ 1122 if (p) { 1123 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1124 if (i > 0) { 1125 if (i >= ip->ip_len) 1126 goto dropfrag; 1127 m_adj(m, i); 1128 m->m_pkthdr.csum_flags = 0; 1129 ip->ip_off += i; 1130 ip->ip_len -= i; 1131 } 1132 m->m_nextpkt = p->m_nextpkt; 1133 p->m_nextpkt = m; 1134 } else { 1135 m->m_nextpkt = fp->ipq_frags; 1136 fp->ipq_frags = m; 1137 } 1138 1139 /* 1140 * While we overlap succeeding segments trim them or, 1141 * if they are completely covered, dequeue them. 1142 */ 1143 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1144 q = nq) { 1145 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1146 if (i < GETIP(q)->ip_len) { 1147 GETIP(q)->ip_len -= i; 1148 GETIP(q)->ip_off += i; 1149 m_adj(q, i); 1150 q->m_pkthdr.csum_flags = 0; 1151 break; 1152 } 1153 nq = q->m_nextpkt; 1154 m->m_nextpkt = nq; 1155 ipstat.ips_fragdropped++; 1156 fp->ipq_nfrags--; 1157 q->m_nextpkt = NULL; 1158 m_freem(q); 1159 } 1160 1161 inserted: 1162 /* 1163 * Check for complete reassembly and perform frag per packet 1164 * limiting. 1165 * 1166 * Frag limiting is performed here so that the nth frag has 1167 * a chance to complete the packet before we drop the packet. 1168 * As a result, n+1 frags are actually allowed per packet, but 1169 * only n will ever be stored. (n = maxfragsperpacket.) 1170 * 1171 */ 1172 next = 0; 1173 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1174 if (GETIP(q)->ip_off != next) { 1175 if (fp->ipq_nfrags > maxfragsperpacket) { 1176 ipstat.ips_fragdropped += fp->ipq_nfrags; 1177 ip_freef(head, fp); 1178 } 1179 goto done; 1180 } 1181 next += GETIP(q)->ip_len; 1182 } 1183 /* Make sure the last packet didn't have the IP_MF flag */ 1184 if (p->m_flags & M_FRAG) { 1185 if (fp->ipq_nfrags > maxfragsperpacket) { 1186 ipstat.ips_fragdropped += fp->ipq_nfrags; 1187 ip_freef(head, fp); 1188 } 1189 goto done; 1190 } 1191 1192 /* 1193 * Reassembly is complete. Make sure the packet is a sane size. 1194 */ 1195 q = fp->ipq_frags; 1196 ip = GETIP(q); 1197 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 1198 ipstat.ips_toolong++; 1199 ipstat.ips_fragdropped += fp->ipq_nfrags; 1200 ip_freef(head, fp); 1201 goto done; 1202 } 1203 1204 /* 1205 * Concatenate fragments. 1206 */ 1207 m = q; 1208 n = m->m_next; 1209 m->m_next = NULL; 1210 m_cat(m, n); 1211 nq = q->m_nextpkt; 1212 q->m_nextpkt = NULL; 1213 for (q = nq; q != NULL; q = nq) { 1214 nq = q->m_nextpkt; 1215 q->m_nextpkt = NULL; 1216 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1217 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1218 m_cat(m, q); 1219 } 1220 1221 /* 1222 * Clean up the 1's complement checksum. Carry over 16 bits must 1223 * be added back. This assumes no more then 65535 packet fragments 1224 * were reassembled. A second carry can also occur (but not a third). 1225 */ 1226 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) + 1227 (m->m_pkthdr.csum_data >> 16); 1228 if (m->m_pkthdr.csum_data > 0xFFFF) 1229 m->m_pkthdr.csum_data -= 0xFFFF; 1230 1231 /* 1232 * Create header for new ip packet by 1233 * modifying header of first packet; 1234 * dequeue and discard fragment reassembly header. 1235 * Make header visible. 1236 */ 1237 ip->ip_len = next; 1238 ip->ip_src = fp->ipq_src; 1239 ip->ip_dst = fp->ipq_dst; 1240 TAILQ_REMOVE(head, fp, ipq_list); 1241 nipq--; 1242 mpipe_free(&ipq_mpipe, fp); 1243 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 1244 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 1245 /* some debugging cruft by sklower, below, will go away soon */ 1246 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1247 int plen = 0; 1248 1249 for (n = m; n; n = n->m_next) 1250 plen += n->m_len; 1251 m->m_pkthdr.len = plen; 1252 } 1253 1254 /* 1255 * Reassembly complete, return the next protocol. 1256 * 1257 * Be sure to clear M_HASH to force the packet 1258 * to be re-characterized. 1259 * 1260 * Clear M_FRAG, we are no longer a fragment. 1261 */ 1262 m->m_flags &= ~(M_HASH | M_FRAG); 1263 1264 ipstat.ips_reassembled++; 1265 lwkt_reltoken(&ipq_token); 1266 return (m); 1267 1268 dropfrag: 1269 ipstat.ips_fragdropped++; 1270 if (fp != NULL) 1271 fp->ipq_nfrags--; 1272 m_freem(m); 1273 done: 1274 lwkt_reltoken(&ipq_token); 1275 return (NULL); 1276 1277 #undef GETIP 1278 } 1279 1280 /* 1281 * Free a fragment reassembly header and all 1282 * associated datagrams. 1283 * 1284 * Called with ipq_token held. 1285 */ 1286 static void 1287 ip_freef(struct ipqhead *fhp, struct ipq *fp) 1288 { 1289 struct mbuf *q; 1290 1291 /* 1292 * Remove first to protect against blocking 1293 */ 1294 TAILQ_REMOVE(fhp, fp, ipq_list); 1295 1296 /* 1297 * Clean out at our leisure 1298 */ 1299 while (fp->ipq_frags) { 1300 q = fp->ipq_frags; 1301 fp->ipq_frags = q->m_nextpkt; 1302 q->m_nextpkt = NULL; 1303 m_freem(q); 1304 } 1305 mpipe_free(&ipq_mpipe, fp); 1306 nipq--; 1307 } 1308 1309 /* 1310 * IP timer processing; 1311 * if a timer expires on a reassembly 1312 * queue, discard it. 1313 */ 1314 void 1315 ip_slowtimo(void) 1316 { 1317 struct ipq *fp, *fp_temp; 1318 struct ipqhead *head; 1319 int i; 1320 1321 lwkt_gettoken(&ipq_token); 1322 for (i = 0; i < IPREASS_NHASH; i++) { 1323 head = &ipq[i]; 1324 TAILQ_FOREACH_MUTABLE(fp, head, ipq_list, fp_temp) { 1325 if (--fp->ipq_ttl == 0) { 1326 ipstat.ips_fragtimeout += fp->ipq_nfrags; 1327 ip_freef(head, fp); 1328 } 1329 } 1330 } 1331 /* 1332 * If we are over the maximum number of fragments 1333 * (due to the limit being lowered), drain off 1334 * enough to get down to the new limit. 1335 */ 1336 if (maxnipq >= 0 && nipq > maxnipq) { 1337 for (i = 0; i < IPREASS_NHASH; i++) { 1338 head = &ipq[i]; 1339 while (nipq > maxnipq && !TAILQ_EMPTY(head)) { 1340 ipstat.ips_fragdropped += 1341 TAILQ_FIRST(head)->ipq_nfrags; 1342 ip_freef(head, TAILQ_FIRST(head)); 1343 } 1344 } 1345 } 1346 lwkt_reltoken(&ipq_token); 1347 ipflow_slowtimo(); 1348 } 1349 1350 /* 1351 * Drain off all datagram fragments. 1352 */ 1353 void 1354 ip_drain(void) 1355 { 1356 struct ipqhead *head; 1357 int i; 1358 1359 lwkt_gettoken(&ipq_token); 1360 for (i = 0; i < IPREASS_NHASH; i++) { 1361 head = &ipq[i]; 1362 while (!TAILQ_EMPTY(head)) { 1363 ipstat.ips_fragdropped += TAILQ_FIRST(head)->ipq_nfrags; 1364 ip_freef(head, TAILQ_FIRST(head)); 1365 } 1366 } 1367 lwkt_reltoken(&ipq_token); 1368 in_rtqdrain(); 1369 } 1370 1371 /* 1372 * Do option processing on a datagram, 1373 * possibly discarding it if bad options are encountered, 1374 * or forwarding it if source-routed. 1375 * The pass argument is used when operating in the IPSTEALTH 1376 * mode to tell what options to process: 1377 * [LS]SRR (pass 0) or the others (pass 1). 1378 * The reason for as many as two passes is that when doing IPSTEALTH, 1379 * non-routing options should be processed only if the packet is for us. 1380 * Returns 1 if packet has been forwarded/freed, 1381 * 0 if the packet should be processed further. 1382 */ 1383 static int 1384 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop) 1385 { 1386 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET }; 1387 struct ip *ip = mtod(m, struct ip *); 1388 u_char *cp; 1389 struct in_ifaddr *ia; 1390 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB; 1391 boolean_t forward = FALSE; 1392 struct in_addr *sin, dst; 1393 n_time ntime; 1394 1395 dst = ip->ip_dst; 1396 cp = (u_char *)(ip + 1); 1397 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1398 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1399 opt = cp[IPOPT_OPTVAL]; 1400 if (opt == IPOPT_EOL) 1401 break; 1402 if (opt == IPOPT_NOP) 1403 optlen = 1; 1404 else { 1405 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1406 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1407 goto bad; 1408 } 1409 optlen = cp[IPOPT_OLEN]; 1410 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1411 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1412 goto bad; 1413 } 1414 } 1415 switch (opt) { 1416 1417 default: 1418 break; 1419 1420 /* 1421 * Source routing with record. 1422 * Find interface with current destination address. 1423 * If none on this machine then drop if strictly routed, 1424 * or do nothing if loosely routed. 1425 * Record interface address and bring up next address 1426 * component. If strictly routed make sure next 1427 * address is on directly accessible net. 1428 */ 1429 case IPOPT_LSRR: 1430 case IPOPT_SSRR: 1431 if (ipstealth && pass > 0) 1432 break; 1433 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1434 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1435 goto bad; 1436 } 1437 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1438 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1439 goto bad; 1440 } 1441 ipaddr.sin_addr = ip->ip_dst; 1442 ia = (struct in_ifaddr *) 1443 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1444 if (ia == NULL) { 1445 if (opt == IPOPT_SSRR) { 1446 type = ICMP_UNREACH; 1447 code = ICMP_UNREACH_SRCFAIL; 1448 goto bad; 1449 } 1450 if (!ip_dosourceroute) 1451 goto nosourcerouting; 1452 /* 1453 * Loose routing, and not at next destination 1454 * yet; nothing to do except forward. 1455 */ 1456 break; 1457 } 1458 off--; /* 0 origin */ 1459 if (off > optlen - (int)sizeof(struct in_addr)) { 1460 /* 1461 * End of source route. Should be for us. 1462 */ 1463 if (!ip_acceptsourceroute) 1464 goto nosourcerouting; 1465 save_rte(m, cp, ip->ip_src); 1466 break; 1467 } 1468 if (ipstealth) 1469 goto dropit; 1470 if (!ip_dosourceroute) { 1471 if (ipforwarding) { 1472 char buf[sizeof "aaa.bbb.ccc.ddd"]; 1473 1474 /* 1475 * Acting as a router, so generate ICMP 1476 */ 1477 nosourcerouting: 1478 strcpy(buf, inet_ntoa(ip->ip_dst)); 1479 log(LOG_WARNING, 1480 "attempted source route from %s to %s\n", 1481 inet_ntoa(ip->ip_src), buf); 1482 type = ICMP_UNREACH; 1483 code = ICMP_UNREACH_SRCFAIL; 1484 goto bad; 1485 } else { 1486 /* 1487 * Not acting as a router, 1488 * so silently drop. 1489 */ 1490 dropit: 1491 ipstat.ips_cantforward++; 1492 m_freem(m); 1493 return (1); 1494 } 1495 } 1496 1497 /* 1498 * locate outgoing interface 1499 */ 1500 memcpy(&ipaddr.sin_addr, cp + off, 1501 sizeof ipaddr.sin_addr); 1502 1503 if (opt == IPOPT_SSRR) { 1504 #define INA struct in_ifaddr * 1505 #define SA struct sockaddr * 1506 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) 1507 == NULL) 1508 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1509 } else { 1510 ia = ip_rtaddr(ipaddr.sin_addr, NULL); 1511 } 1512 if (ia == NULL) { 1513 type = ICMP_UNREACH; 1514 code = ICMP_UNREACH_SRCFAIL; 1515 goto bad; 1516 } 1517 ip->ip_dst = ipaddr.sin_addr; 1518 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1519 sizeof(struct in_addr)); 1520 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1521 /* 1522 * Let ip_intr's mcast routing check handle mcast pkts 1523 */ 1524 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1525 break; 1526 1527 case IPOPT_RR: 1528 if (ipstealth && pass == 0) 1529 break; 1530 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1531 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1532 goto bad; 1533 } 1534 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1535 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1536 goto bad; 1537 } 1538 /* 1539 * If no space remains, ignore. 1540 */ 1541 off--; /* 0 origin */ 1542 if (off > optlen - (int)sizeof(struct in_addr)) 1543 break; 1544 memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1545 sizeof ipaddr.sin_addr); 1546 /* 1547 * locate outgoing interface; if we're the destination, 1548 * use the incoming interface (should be same). 1549 */ 1550 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL && 1551 (ia = ip_rtaddr(ipaddr.sin_addr, NULL)) == NULL) { 1552 type = ICMP_UNREACH; 1553 code = ICMP_UNREACH_HOST; 1554 goto bad; 1555 } 1556 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1557 sizeof(struct in_addr)); 1558 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1559 break; 1560 1561 case IPOPT_TS: 1562 if (ipstealth && pass == 0) 1563 break; 1564 code = cp - (u_char *)ip; 1565 if (optlen < 4 || optlen > 40) { 1566 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1567 goto bad; 1568 } 1569 if ((off = cp[IPOPT_OFFSET]) < 5) { 1570 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1571 goto bad; 1572 } 1573 if (off > optlen - (int)sizeof(int32_t)) { 1574 cp[IPOPT_OFFSET + 1] += (1 << 4); 1575 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1576 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1577 goto bad; 1578 } 1579 break; 1580 } 1581 off--; /* 0 origin */ 1582 sin = (struct in_addr *)(cp + off); 1583 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1584 1585 case IPOPT_TS_TSONLY: 1586 break; 1587 1588 case IPOPT_TS_TSANDADDR: 1589 if (off + sizeof(n_time) + 1590 sizeof(struct in_addr) > optlen) { 1591 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1592 goto bad; 1593 } 1594 ipaddr.sin_addr = dst; 1595 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1596 m->m_pkthdr.rcvif); 1597 if (ia == NULL) 1598 continue; 1599 memcpy(sin, &IA_SIN(ia)->sin_addr, 1600 sizeof(struct in_addr)); 1601 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1602 off += sizeof(struct in_addr); 1603 break; 1604 1605 case IPOPT_TS_PRESPEC: 1606 if (off + sizeof(n_time) + 1607 sizeof(struct in_addr) > optlen) { 1608 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1609 goto bad; 1610 } 1611 memcpy(&ipaddr.sin_addr, sin, 1612 sizeof(struct in_addr)); 1613 if (ifa_ifwithaddr((SA)&ipaddr) == NULL) 1614 continue; 1615 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1616 off += sizeof(struct in_addr); 1617 break; 1618 1619 default: 1620 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1621 goto bad; 1622 } 1623 ntime = iptime(); 1624 memcpy(cp + off, &ntime, sizeof(n_time)); 1625 cp[IPOPT_OFFSET] += sizeof(n_time); 1626 } 1627 } 1628 if (forward && ipforwarding) { 1629 ip_forward(m, TRUE, next_hop); 1630 return (1); 1631 } 1632 return (0); 1633 bad: 1634 icmp_error(m, type, code, 0, 0); 1635 ipstat.ips_badoptions++; 1636 return (1); 1637 } 1638 1639 /* 1640 * Given address of next destination (final or next hop), 1641 * return internet address info of interface to be used to get there. 1642 */ 1643 struct in_ifaddr * 1644 ip_rtaddr(struct in_addr dst, struct route *ro0) 1645 { 1646 struct route sro, *ro; 1647 struct sockaddr_in *sin; 1648 struct in_ifaddr *ia; 1649 1650 if (ro0 != NULL) { 1651 ro = ro0; 1652 } else { 1653 bzero(&sro, sizeof(sro)); 1654 ro = &sro; 1655 } 1656 1657 sin = (struct sockaddr_in *)&ro->ro_dst; 1658 1659 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) { 1660 if (ro->ro_rt != NULL) { 1661 RTFREE(ro->ro_rt); 1662 ro->ro_rt = NULL; 1663 } 1664 sin->sin_family = AF_INET; 1665 sin->sin_len = sizeof *sin; 1666 sin->sin_addr = dst; 1667 rtalloc_ign(ro, RTF_PRCLONING); 1668 } 1669 1670 if (ro->ro_rt == NULL) 1671 return (NULL); 1672 1673 ia = ifatoia(ro->ro_rt->rt_ifa); 1674 1675 if (ro == &sro) 1676 RTFREE(ro->ro_rt); 1677 return ia; 1678 } 1679 1680 /* 1681 * Save incoming source route for use in replies, 1682 * to be picked up later by ip_srcroute if the receiver is interested. 1683 */ 1684 static void 1685 save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1686 { 1687 struct m_tag *mtag; 1688 struct ip_srcrt_opt *opt; 1689 unsigned olen; 1690 1691 mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), MB_DONTWAIT); 1692 if (mtag == NULL) 1693 return; 1694 opt = m_tag_data(mtag); 1695 1696 olen = option[IPOPT_OLEN]; 1697 #ifdef DIAGNOSTIC 1698 if (ipprintfs) 1699 kprintf("save_rte: olen %d\n", olen); 1700 #endif 1701 if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) { 1702 m_tag_free(mtag); 1703 return; 1704 } 1705 bcopy(option, opt->ip_srcrt.srcopt, olen); 1706 opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1707 opt->ip_srcrt.dst = dst; 1708 m_tag_prepend(m, mtag); 1709 } 1710 1711 /* 1712 * Retrieve incoming source route for use in replies, 1713 * in the same form used by setsockopt. 1714 * The first hop is placed before the options, will be removed later. 1715 */ 1716 struct mbuf * 1717 ip_srcroute(struct mbuf *m0) 1718 { 1719 struct in_addr *p, *q; 1720 struct mbuf *m; 1721 struct m_tag *mtag; 1722 struct ip_srcrt_opt *opt; 1723 1724 if (m0 == NULL) 1725 return NULL; 1726 1727 mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL); 1728 if (mtag == NULL) 1729 return NULL; 1730 opt = m_tag_data(mtag); 1731 1732 if (opt->ip_nhops == 0) 1733 return (NULL); 1734 m = m_get(MB_DONTWAIT, MT_HEADER); 1735 if (m == NULL) 1736 return (NULL); 1737 1738 #define OPTSIZ (sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt)) 1739 1740 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1741 m->m_len = opt->ip_nhops * sizeof(struct in_addr) + 1742 sizeof(struct in_addr) + OPTSIZ; 1743 #ifdef DIAGNOSTIC 1744 if (ipprintfs) { 1745 kprintf("ip_srcroute: nhops %d mlen %d", 1746 opt->ip_nhops, m->m_len); 1747 } 1748 #endif 1749 1750 /* 1751 * First save first hop for return route 1752 */ 1753 p = &opt->ip_srcrt.route[opt->ip_nhops - 1]; 1754 *(mtod(m, struct in_addr *)) = *p--; 1755 #ifdef DIAGNOSTIC 1756 if (ipprintfs) 1757 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1758 #endif 1759 1760 /* 1761 * Copy option fields and padding (nop) to mbuf. 1762 */ 1763 opt->ip_srcrt.nop = IPOPT_NOP; 1764 opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1765 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop, 1766 OPTSIZ); 1767 q = (struct in_addr *)(mtod(m, caddr_t) + 1768 sizeof(struct in_addr) + OPTSIZ); 1769 #undef OPTSIZ 1770 /* 1771 * Record return path as an IP source route, 1772 * reversing the path (pointers are now aligned). 1773 */ 1774 while (p >= opt->ip_srcrt.route) { 1775 #ifdef DIAGNOSTIC 1776 if (ipprintfs) 1777 kprintf(" %x", ntohl(q->s_addr)); 1778 #endif 1779 *q++ = *p--; 1780 } 1781 /* 1782 * Last hop goes to final destination. 1783 */ 1784 *q = opt->ip_srcrt.dst; 1785 m_tag_delete(m0, mtag); 1786 #ifdef DIAGNOSTIC 1787 if (ipprintfs) 1788 kprintf(" %x\n", ntohl(q->s_addr)); 1789 #endif 1790 return (m); 1791 } 1792 1793 /* 1794 * Strip out IP options. 1795 */ 1796 void 1797 ip_stripoptions(struct mbuf *m) 1798 { 1799 int datalen; 1800 struct ip *ip = mtod(m, struct ip *); 1801 caddr_t opts; 1802 int optlen; 1803 1804 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1805 opts = (caddr_t)(ip + 1); 1806 datalen = m->m_len - (sizeof(struct ip) + optlen); 1807 bcopy(opts + optlen, opts, datalen); 1808 m->m_len -= optlen; 1809 if (m->m_flags & M_PKTHDR) 1810 m->m_pkthdr.len -= optlen; 1811 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1812 } 1813 1814 u_char inetctlerrmap[PRC_NCMDS] = { 1815 0, 0, 0, 0, 1816 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1817 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1818 EMSGSIZE, EHOSTUNREACH, 0, 0, 1819 0, 0, 0, 0, 1820 ENOPROTOOPT, ECONNREFUSED 1821 }; 1822 1823 /* 1824 * Forward a packet. If some error occurs return the sender 1825 * an icmp packet. Note we can't always generate a meaningful 1826 * icmp message because icmp doesn't have a large enough repertoire 1827 * of codes and types. 1828 * 1829 * If not forwarding, just drop the packet. This could be confusing 1830 * if ipforwarding was zero but some routing protocol was advancing 1831 * us as a gateway to somewhere. However, we must let the routing 1832 * protocol deal with that. 1833 * 1834 * The using_srcrt parameter indicates whether the packet is being forwarded 1835 * via a source route. 1836 */ 1837 void 1838 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop) 1839 { 1840 struct ip *ip = mtod(m, struct ip *); 1841 struct rtentry *rt; 1842 struct route fwd_ro; 1843 int error, type = 0, code = 0, destmtu = 0; 1844 struct mbuf *mcopy, *mtemp = NULL; 1845 n_long dest; 1846 struct in_addr pkt_dst; 1847 1848 dest = INADDR_ANY; 1849 /* 1850 * Cache the destination address of the packet; this may be 1851 * changed by use of 'ipfw fwd'. 1852 */ 1853 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst; 1854 1855 #ifdef DIAGNOSTIC 1856 if (ipprintfs) 1857 kprintf("forward: src %x dst %x ttl %x\n", 1858 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl); 1859 #endif 1860 1861 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) { 1862 ipstat.ips_cantforward++; 1863 m_freem(m); 1864 return; 1865 } 1866 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) { 1867 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1868 return; 1869 } 1870 1871 bzero(&fwd_ro, sizeof(fwd_ro)); 1872 ip_rtaddr(pkt_dst, &fwd_ro); 1873 if (fwd_ro.ro_rt == NULL) { 1874 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1875 return; 1876 } 1877 rt = fwd_ro.ro_rt; 1878 1879 if (curthread->td_type == TD_TYPE_NETISR) { 1880 /* 1881 * Save the IP header and at most 8 bytes of the payload, 1882 * in case we need to generate an ICMP message to the src. 1883 */ 1884 mtemp = ipforward_mtemp[mycpuid]; 1885 KASSERT((mtemp->m_flags & M_EXT) == 0 && 1886 mtemp->m_data == mtemp->m_pktdat && 1887 m_tag_first(mtemp) == NULL, 1888 ("ip_forward invalid mtemp1")); 1889 1890 if (!m_dup_pkthdr(mtemp, m, MB_DONTWAIT)) { 1891 /* 1892 * It's probably ok if the pkthdr dup fails (because 1893 * the deep copy of the tag chain failed), but for now 1894 * be conservative and just discard the copy since 1895 * code below may some day want the tags. 1896 */ 1897 mtemp = NULL; 1898 } else { 1899 mtemp->m_type = m->m_type; 1900 mtemp->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8, 1901 (int)ip->ip_len); 1902 mtemp->m_pkthdr.len = mtemp->m_len; 1903 m_copydata(m, 0, mtemp->m_len, mtod(mtemp, caddr_t)); 1904 } 1905 } 1906 1907 if (!ipstealth) 1908 ip->ip_ttl -= IPTTLDEC; 1909 1910 /* 1911 * If forwarding packet using same interface that it came in on, 1912 * perhaps should send a redirect to sender to shortcut a hop. 1913 * Only send redirect if source is sending directly to us, 1914 * and if packet was not source routed (or has any options). 1915 * Also, don't send redirect if forwarding using a default route 1916 * or a route modified by a redirect. 1917 */ 1918 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1919 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) && 1920 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY && 1921 ipsendredirects && !using_srcrt && next_hop == NULL) { 1922 u_long src = ntohl(ip->ip_src.s_addr); 1923 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa; 1924 1925 if (rt_ifa != NULL && 1926 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) { 1927 if (rt->rt_flags & RTF_GATEWAY) 1928 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1929 else 1930 dest = pkt_dst.s_addr; 1931 /* 1932 * Router requirements says to only send 1933 * host redirects. 1934 */ 1935 type = ICMP_REDIRECT; 1936 code = ICMP_REDIRECT_HOST; 1937 #ifdef DIAGNOSTIC 1938 if (ipprintfs) 1939 kprintf("redirect (%d) to %x\n", code, dest); 1940 #endif 1941 } 1942 } 1943 1944 error = ip_output(m, NULL, &fwd_ro, IP_FORWARDING, NULL, NULL); 1945 if (error == 0) { 1946 ipstat.ips_forward++; 1947 if (type == 0) { 1948 if (mtemp) 1949 ipflow_create(&fwd_ro, mtemp); 1950 goto done; 1951 } else { 1952 ipstat.ips_redirectsent++; 1953 } 1954 } else { 1955 ipstat.ips_cantforward++; 1956 } 1957 1958 if (mtemp == NULL) 1959 goto done; 1960 1961 /* 1962 * Errors that do not require generating ICMP message 1963 */ 1964 switch (error) { 1965 case ENOBUFS: 1966 /* 1967 * A router should not generate ICMP_SOURCEQUENCH as 1968 * required in RFC1812 Requirements for IP Version 4 Routers. 1969 * Source quench could be a big problem under DoS attacks, 1970 * or if the underlying interface is rate-limited. 1971 * Those who need source quench packets may re-enable them 1972 * via the net.inet.ip.sendsourcequench sysctl. 1973 */ 1974 if (!ip_sendsourcequench) 1975 goto done; 1976 break; 1977 1978 case EACCES: /* ipfw denied packet */ 1979 goto done; 1980 } 1981 1982 KASSERT((mtemp->m_flags & M_EXT) == 0 && 1983 mtemp->m_data == mtemp->m_pktdat, 1984 ("ip_forward invalid mtemp2")); 1985 mcopy = m_copym(mtemp, 0, mtemp->m_len, MB_DONTWAIT); 1986 if (mcopy == NULL) 1987 goto done; 1988 1989 /* 1990 * Send ICMP message. 1991 */ 1992 switch (error) { 1993 case 0: /* forwarded, but need redirect */ 1994 /* type, code set above */ 1995 break; 1996 1997 case ENETUNREACH: /* shouldn't happen, checked above */ 1998 case EHOSTUNREACH: 1999 case ENETDOWN: 2000 case EHOSTDOWN: 2001 default: 2002 type = ICMP_UNREACH; 2003 code = ICMP_UNREACH_HOST; 2004 break; 2005 2006 case EMSGSIZE: 2007 type = ICMP_UNREACH; 2008 code = ICMP_UNREACH_NEEDFRAG; 2009 #ifdef IPSEC 2010 /* 2011 * If the packet is routed over IPsec tunnel, tell the 2012 * originator the tunnel MTU. 2013 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 2014 * XXX quickhack!!! 2015 */ 2016 if (fwd_ro.ro_rt != NULL) { 2017 struct secpolicy *sp = NULL; 2018 int ipsecerror; 2019 int ipsechdr; 2020 struct route *ro; 2021 2022 sp = ipsec4_getpolicybyaddr(mcopy, 2023 IPSEC_DIR_OUTBOUND, 2024 IP_FORWARDING, 2025 &ipsecerror); 2026 2027 if (sp == NULL) 2028 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2029 else { 2030 /* count IPsec header size */ 2031 ipsechdr = ipsec4_hdrsiz(mcopy, 2032 IPSEC_DIR_OUTBOUND, 2033 NULL); 2034 2035 /* 2036 * find the correct route for outer IPv4 2037 * header, compute tunnel MTU. 2038 * 2039 */ 2040 if (sp->req != NULL && sp->req->sav != NULL && 2041 sp->req->sav->sah != NULL) { 2042 ro = &sp->req->sav->sah->sa_route; 2043 if (ro->ro_rt != NULL && 2044 ro->ro_rt->rt_ifp != NULL) { 2045 destmtu = 2046 ro->ro_rt->rt_ifp->if_mtu; 2047 destmtu -= ipsechdr; 2048 } 2049 } 2050 2051 key_freesp(sp); 2052 } 2053 } 2054 #elif FAST_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 (fwd_ro.ro_rt != NULL) { 2062 struct secpolicy *sp = NULL; 2063 int ipsecerror; 2064 int ipsechdr; 2065 struct route *ro; 2066 2067 sp = ipsec_getpolicybyaddr(mcopy, 2068 IPSEC_DIR_OUTBOUND, 2069 IP_FORWARDING, 2070 &ipsecerror); 2071 2072 if (sp == NULL) 2073 destmtu = fwd_ro.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 && 2086 sp->req->sav != NULL && 2087 sp->req->sav->sah != NULL) { 2088 ro = &sp->req->sav->sah->sa_route; 2089 if (ro->ro_rt != NULL && 2090 ro->ro_rt->rt_ifp != NULL) { 2091 destmtu = 2092 ro->ro_rt->rt_ifp->if_mtu; 2093 destmtu -= ipsechdr; 2094 } 2095 } 2096 2097 KEY_FREESP(&sp); 2098 } 2099 } 2100 #else /* !IPSEC && !FAST_IPSEC */ 2101 if (fwd_ro.ro_rt != NULL) 2102 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2103 #endif /*IPSEC*/ 2104 ipstat.ips_cantfrag++; 2105 break; 2106 2107 case ENOBUFS: 2108 type = ICMP_SOURCEQUENCH; 2109 code = 0; 2110 break; 2111 2112 case EACCES: /* ipfw denied packet */ 2113 panic("ip_forward EACCES should not reach"); 2114 } 2115 icmp_error(mcopy, type, code, dest, destmtu); 2116 done: 2117 if (mtemp != NULL) 2118 m_tag_delete_chain(mtemp); 2119 if (fwd_ro.ro_rt != NULL) 2120 RTFREE(fwd_ro.ro_rt); 2121 } 2122 2123 void 2124 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2125 struct mbuf *m) 2126 { 2127 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2128 struct timeval tv; 2129 2130 microtime(&tv); 2131 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2132 SCM_TIMESTAMP, SOL_SOCKET); 2133 if (*mp) 2134 mp = &(*mp)->m_next; 2135 } 2136 if (inp->inp_flags & INP_RECVDSTADDR) { 2137 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2138 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2139 if (*mp) 2140 mp = &(*mp)->m_next; 2141 } 2142 if (inp->inp_flags & INP_RECVTTL) { 2143 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 2144 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 2145 if (*mp) 2146 mp = &(*mp)->m_next; 2147 } 2148 #ifdef notyet 2149 /* XXX 2150 * Moving these out of udp_input() made them even more broken 2151 * than they already were. 2152 */ 2153 /* options were tossed already */ 2154 if (inp->inp_flags & INP_RECVOPTS) { 2155 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2156 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2157 if (*mp) 2158 mp = &(*mp)->m_next; 2159 } 2160 /* ip_srcroute doesn't do what we want here, need to fix */ 2161 if (inp->inp_flags & INP_RECVRETOPTS) { 2162 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 2163 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2164 if (*mp) 2165 mp = &(*mp)->m_next; 2166 } 2167 #endif 2168 if (inp->inp_flags & INP_RECVIF) { 2169 struct ifnet *ifp; 2170 struct sdlbuf { 2171 struct sockaddr_dl sdl; 2172 u_char pad[32]; 2173 } sdlbuf; 2174 struct sockaddr_dl *sdp; 2175 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2176 2177 if (((ifp = m->m_pkthdr.rcvif)) && 2178 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) { 2179 sdp = IF_LLSOCKADDR(ifp); 2180 /* 2181 * Change our mind and don't try copy. 2182 */ 2183 if ((sdp->sdl_family != AF_LINK) || 2184 (sdp->sdl_len > sizeof(sdlbuf))) { 2185 goto makedummy; 2186 } 2187 bcopy(sdp, sdl2, sdp->sdl_len); 2188 } else { 2189 makedummy: 2190 sdl2->sdl_len = 2191 offsetof(struct sockaddr_dl, sdl_data[0]); 2192 sdl2->sdl_family = AF_LINK; 2193 sdl2->sdl_index = 0; 2194 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2195 } 2196 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 2197 IP_RECVIF, IPPROTO_IP); 2198 if (*mp) 2199 mp = &(*mp)->m_next; 2200 } 2201 } 2202 2203 /* 2204 * XXX these routines are called from the upper part of the kernel. 2205 * 2206 * They could also be moved to ip_mroute.c, since all the RSVP 2207 * handling is done there already. 2208 */ 2209 int 2210 ip_rsvp_init(struct socket *so) 2211 { 2212 if (so->so_type != SOCK_RAW || 2213 so->so_proto->pr_protocol != IPPROTO_RSVP) 2214 return EOPNOTSUPP; 2215 2216 if (ip_rsvpd != NULL) 2217 return EADDRINUSE; 2218 2219 ip_rsvpd = so; 2220 /* 2221 * This may seem silly, but we need to be sure we don't over-increment 2222 * the RSVP counter, in case something slips up. 2223 */ 2224 if (!ip_rsvp_on) { 2225 ip_rsvp_on = 1; 2226 rsvp_on++; 2227 } 2228 2229 return 0; 2230 } 2231 2232 int 2233 ip_rsvp_done(void) 2234 { 2235 ip_rsvpd = NULL; 2236 /* 2237 * This may seem silly, but we need to be sure we don't over-decrement 2238 * the RSVP counter, in case something slips up. 2239 */ 2240 if (ip_rsvp_on) { 2241 ip_rsvp_on = 0; 2242 rsvp_on--; 2243 } 2244 return 0; 2245 } 2246 2247 int 2248 rsvp_input(struct mbuf **mp, int *offp, int proto) 2249 { 2250 struct mbuf *m = *mp; 2251 2252 *mp = NULL; 2253 2254 if (rsvp_input_p) { /* call the real one if loaded */ 2255 *mp = m; 2256 rsvp_input_p(mp, offp, proto); 2257 return(IPPROTO_DONE); 2258 } 2259 2260 /* Can still get packets with rsvp_on = 0 if there is a local member 2261 * of the group to which the RSVP packet is addressed. But in this 2262 * case we want to throw the packet away. 2263 */ 2264 2265 if (!rsvp_on) { 2266 m_freem(m); 2267 return(IPPROTO_DONE); 2268 } 2269 2270 if (ip_rsvpd != NULL) { 2271 *mp = m; 2272 rip_input(mp, offp, proto); 2273 return(IPPROTO_DONE); 2274 } 2275 /* Drop the packet */ 2276 m_freem(m); 2277 return(IPPROTO_DONE); 2278 } 2279