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 */ 69 70 #define _IP_VHL 71 72 #include "opt_bootp.h" 73 #include "opt_ipdn.h" 74 #include "opt_ipdivert.h" 75 #include "opt_ipstealth.h" 76 #include "opt_ipsec.h" 77 #include "opt_rss.h" 78 79 #include <sys/param.h> 80 #include <sys/systm.h> 81 #include <sys/mbuf.h> 82 #include <sys/malloc.h> 83 #include <sys/mpipe.h> 84 #include <sys/domain.h> 85 #include <sys/protosw.h> 86 #include <sys/socket.h> 87 #include <sys/time.h> 88 #include <sys/globaldata.h> 89 #include <sys/thread.h> 90 #include <sys/kernel.h> 91 #include <sys/syslog.h> 92 #include <sys/sysctl.h> 93 #include <sys/in_cksum.h> 94 #include <sys/lock.h> 95 96 #include <sys/mplock2.h> 97 98 #include <machine/stdarg.h> 99 100 #include <net/if.h> 101 #include <net/if_types.h> 102 #include <net/if_var.h> 103 #include <net/if_dl.h> 104 #include <net/pfil.h> 105 #include <net/route.h> 106 #include <net/netisr2.h> 107 108 #include <netinet/in.h> 109 #include <netinet/in_systm.h> 110 #include <netinet/in_var.h> 111 #include <netinet/ip.h> 112 #include <netinet/in_pcb.h> 113 #include <netinet/ip_var.h> 114 #include <netinet/ip_icmp.h> 115 #include <netinet/ip_divert.h> 116 #include <netinet/ip_flow.h> 117 118 #include <sys/thread2.h> 119 #include <sys/msgport2.h> 120 #include <net/netmsg2.h> 121 122 #include <sys/socketvar.h> 123 124 #include <net/ipfw/ip_fw.h> 125 #include <net/dummynet/ip_dummynet.h> 126 127 #ifdef IPSEC 128 #include <netinet6/ipsec.h> 129 #include <netproto/key/key.h> 130 #endif 131 132 #ifdef FAST_IPSEC 133 #include <netproto/ipsec/ipsec.h> 134 #include <netproto/ipsec/key.h> 135 #endif 136 137 int rsvp_on = 0; 138 static int ip_rsvp_on; 139 struct socket *ip_rsvpd; 140 141 int ipforwarding = 0; 142 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 143 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 144 145 static int ipsendredirects = 1; /* XXX */ 146 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 147 &ipsendredirects, 0, "Enable sending IP redirects"); 148 149 int ip_defttl = IPDEFTTL; 150 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 151 &ip_defttl, 0, "Maximum TTL on IP packets"); 152 153 static int ip_dosourceroute = 0; 154 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 155 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 156 157 static int ip_acceptsourceroute = 0; 158 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 159 CTLFLAG_RW, &ip_acceptsourceroute, 0, 160 "Enable accepting source routed IP packets"); 161 162 static int ip_keepfaith = 0; 163 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 164 &ip_keepfaith, 0, 165 "Enable packet capture for FAITH IPv4->IPv6 translator daemon"); 166 167 static int nipq = 0; /* total # of reass queues */ 168 static int maxnipq; 169 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 170 &maxnipq, 0, 171 "Maximum number of IPv4 fragment reassembly queue entries"); 172 173 static int maxfragsperpacket; 174 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 175 &maxfragsperpacket, 0, 176 "Maximum number of IPv4 fragments allowed per packet"); 177 178 static int ip_sendsourcequench = 0; 179 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 180 &ip_sendsourcequench, 0, 181 "Enable the transmission of source quench packets"); 182 183 int ip_do_randomid = 1; 184 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 185 &ip_do_randomid, 0, 186 "Assign random ip_id values"); 187 /* 188 * XXX - Setting ip_checkinterface mostly implements the receive side of 189 * the Strong ES model described in RFC 1122, but since the routing table 190 * and transmit implementation do not implement the Strong ES model, 191 * setting this to 1 results in an odd hybrid. 192 * 193 * XXX - ip_checkinterface currently must be disabled if you use ipnat 194 * to translate the destination address to another local interface. 195 * 196 * XXX - ip_checkinterface must be disabled if you add IP aliases 197 * to the loopback interface instead of the interface where the 198 * packets for those addresses are received. 199 */ 200 static int ip_checkinterface = 0; 201 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 202 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 203 204 static u_long ip_hash_count = 0; 205 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, hash_count, CTLFLAG_RD, 206 &ip_hash_count, 0, "Number of packets hashed by IP"); 207 208 #ifdef RSS_DEBUG 209 static u_long ip_rehash_count = 0; 210 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, rehash_count, CTLFLAG_RD, 211 &ip_rehash_count, 0, "Number of packets rehashed by IP"); 212 213 static u_long ip_dispatch_fast = 0; 214 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_fast_count, CTLFLAG_RD, 215 &ip_dispatch_fast, 0, "Number of packets handled on current CPU"); 216 217 static u_long ip_dispatch_slow = 0; 218 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_slow_count, CTLFLAG_RD, 219 &ip_dispatch_slow, 0, "Number of packets messaged to another CPU"); 220 #endif 221 222 static struct lwkt_token ipq_token = LWKT_TOKEN_INITIALIZER(ipq_token); 223 224 #ifdef DIAGNOSTIC 225 static int ipprintfs = 0; 226 #endif 227 228 extern struct domain inetdomain; 229 extern struct protosw inetsw[]; 230 u_char ip_protox[IPPROTO_MAX]; 231 struct in_ifaddrhead in_ifaddrheads[MAXCPU]; /* first inet address */ 232 struct in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU]; 233 /* inet addr hash table */ 234 u_long in_ifaddrhmask; /* mask for hash table */ 235 236 static struct mbuf *ipforward_mtemp[MAXCPU]; 237 238 struct ip_stats ipstats_percpu[MAXCPU] __cachealign; 239 240 static int 241 sysctl_ipstats(SYSCTL_HANDLER_ARGS) 242 { 243 int cpu, error = 0; 244 245 for (cpu = 0; cpu < ncpus; ++cpu) { 246 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu], 247 sizeof(struct ip_stats)))) 248 break; 249 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu], 250 sizeof(struct ip_stats)))) 251 break; 252 } 253 254 return (error); 255 } 256 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW), 257 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics"); 258 259 /* Packet reassembly stuff */ 260 #define IPREASS_NHASH_LOG2 6 261 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 262 #define IPREASS_HMASK (IPREASS_NHASH - 1) 263 #define IPREASS_HASH(x,y) \ 264 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 265 266 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 267 268 #ifdef IPCTL_DEFMTU 269 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 270 &ip_mtu, 0, "Default MTU"); 271 #endif 272 273 #ifdef IPSTEALTH 274 static int ipstealth = 0; 275 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, ""); 276 #else 277 static const int ipstealth = 0; 278 #endif 279 280 struct mbuf *(*ip_divert_p)(struct mbuf *, int, int); 281 282 struct pfil_head inet_pfil_hook; 283 284 /* 285 * struct ip_srcrt_opt is used to store packet state while it travels 286 * through the stack. 287 * 288 * XXX Note that the code even makes assumptions on the size and 289 * alignment of fields inside struct ip_srcrt so e.g. adding some 290 * fields will break the code. This needs to be fixed. 291 * 292 * We need to save the IP options in case a protocol wants to respond 293 * to an incoming packet over the same route if the packet got here 294 * using IP source routing. This allows connection establishment and 295 * maintenance when the remote end is on a network that is not known 296 * to us. 297 */ 298 struct ip_srcrt { 299 struct in_addr dst; /* final destination */ 300 char nop; /* one NOP to align */ 301 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 302 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 303 }; 304 305 struct ip_srcrt_opt { 306 int ip_nhops; 307 struct ip_srcrt ip_srcrt; 308 }; 309 310 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management"); 311 static struct malloc_pipe ipq_mpipe; 312 313 static void save_rte(struct mbuf *, u_char *, struct in_addr); 314 static int ip_dooptions(struct mbuf *m, int, struct sockaddr_in *); 315 static void ip_freef(struct ipqhead *, struct ipq *); 316 static void ip_input_handler(netmsg_t); 317 318 /* 319 * IP initialization: fill in IP protocol switch table. 320 * All protocols not implemented in kernel go to raw IP protocol handler. 321 */ 322 void 323 ip_init(void) 324 { 325 struct protosw *pr; 326 int i; 327 int cpu; 328 329 /* 330 * Make sure we can handle a reasonable number of fragments but 331 * cap it at 4000 (XXX). 332 */ 333 mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq), 334 IFQ_MAXLEN, 4000, 0, NULL, NULL, NULL); 335 for (i = 0; i < ncpus; ++i) { 336 TAILQ_INIT(&in_ifaddrheads[i]); 337 in_ifaddrhashtbls[i] = 338 hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 339 } 340 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 341 if (pr == NULL) 342 panic("ip_init"); 343 for (i = 0; i < IPPROTO_MAX; i++) 344 ip_protox[i] = pr - inetsw; 345 for (pr = inetdomain.dom_protosw; 346 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 347 if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol) { 348 if (pr->pr_protocol != IPPROTO_RAW) 349 ip_protox[pr->pr_protocol] = pr - inetsw; 350 } 351 } 352 353 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 354 inet_pfil_hook.ph_af = AF_INET; 355 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) { 356 kprintf("%s: WARNING: unable to register pfil hook, " 357 "error %d\n", __func__, i); 358 } 359 360 for (i = 0; i < IPREASS_NHASH; i++) 361 TAILQ_INIT(&ipq[i]); 362 363 maxnipq = nmbclusters / 32; 364 maxfragsperpacket = 16; 365 366 ip_id = time_second & 0xffff; 367 368 for (cpu = 0; cpu < ncpus; ++cpu) { 369 /* 370 * Initialize IP statistics counters for each CPU. 371 */ 372 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats)); 373 374 /* 375 * Preallocate mbuf template for forwarding 376 */ 377 MGETHDR(ipforward_mtemp[cpu], MB_WAIT, MT_DATA); 378 } 379 380 netisr_register(NETISR_IP, ip_input_handler, ip_hashfn_in); 381 netisr_register_hashcheck(NETISR_IP, ip_hashcheck); 382 } 383 384 /* Do transport protocol processing. */ 385 static void 386 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip) 387 { 388 const struct protosw *pr = &inetsw[ip_protox[ip->ip_p]]; 389 390 /* 391 * Switch out to protocol's input routine. 392 */ 393 PR_GET_MPLOCK(pr); 394 pr->pr_input(&m, &hlen, ip->ip_p); 395 PR_REL_MPLOCK(pr); 396 } 397 398 static void 399 transport_processing_handler(netmsg_t msg) 400 { 401 struct netmsg_packet *pmsg = &msg->packet; 402 struct ip *ip; 403 int hlen; 404 405 ip = mtod(pmsg->nm_packet, struct ip *); 406 hlen = pmsg->base.lmsg.u.ms_result; 407 408 transport_processing_oncpu(pmsg->nm_packet, hlen, ip); 409 /* msg was embedded in the mbuf, do not reply! */ 410 } 411 412 static void 413 ip_input_handler(netmsg_t msg) 414 { 415 ip_input(msg->packet.nm_packet); 416 /* msg was embedded in the mbuf, do not reply! */ 417 } 418 419 /* 420 * IP input routine. Checksum and byte swap header. If fragmented 421 * try to reassemble. Process options. Pass to next level. 422 */ 423 void 424 ip_input(struct mbuf *m) 425 { 426 struct ip *ip; 427 struct in_ifaddr *ia = NULL; 428 struct in_ifaddr_container *iac; 429 int hlen, checkif; 430 u_short sum; 431 struct in_addr pkt_dst; 432 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */ 433 struct in_addr odst; /* original dst address(NAT) */ 434 struct m_tag *mtag; 435 struct sockaddr_in *next_hop = NULL; 436 lwkt_port_t port; 437 #ifdef FAST_IPSEC 438 struct tdb_ident *tdbi; 439 struct secpolicy *sp; 440 int error; 441 #endif 442 443 M_ASSERTPKTHDR(m); 444 445 /* 446 * This routine is called from numerous places which may not have 447 * characterized the packet. 448 */ 449 if ((m->m_flags & M_HASH) == 0) { 450 atomic_add_long(&ip_hash_count, 1); 451 ip_hashfn(&m, 0, IP_MPORT_IN); 452 if (m == NULL) 453 return; 454 KKASSERT(m->m_flags & M_HASH); 455 } 456 ip = mtod(m, struct ip *); 457 458 /* 459 * Pull out certain tags 460 */ 461 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 462 /* Next hop */ 463 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 464 KKASSERT(mtag != NULL); 465 next_hop = m_tag_data(mtag); 466 } 467 468 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 469 /* dummynet already filtered us */ 470 ip = mtod(m, struct ip *); 471 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 472 goto iphack; 473 } 474 475 ipstat.ips_total++; 476 477 /* length checks already done in ip_hashfn() */ 478 KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf")); 479 480 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 481 ipstat.ips_badvers++; 482 goto bad; 483 } 484 485 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 486 /* length checks already done in ip_hashfn() */ 487 KASSERT(hlen >= sizeof(struct ip), ("IP header len too small")); 488 KASSERT(m->m_len >= hlen, ("complete IP header not in one mbuf")); 489 490 /* 127/8 must not appear on wire - RFC1122 */ 491 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 492 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 493 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) { 494 ipstat.ips_badaddr++; 495 goto bad; 496 } 497 } 498 499 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 500 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 501 } else { 502 if (hlen == sizeof(struct ip)) 503 sum = in_cksum_hdr(ip); 504 else 505 sum = in_cksum(m, hlen); 506 } 507 if (sum != 0) { 508 ipstat.ips_badsum++; 509 goto bad; 510 } 511 512 #ifdef ALTQ 513 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 514 /* packet is dropped by traffic conditioner */ 515 return; 516 } 517 #endif 518 /* 519 * Convert fields to host representation. 520 */ 521 ip->ip_len = ntohs(ip->ip_len); 522 ip->ip_off = ntohs(ip->ip_off); 523 524 /* length checks already done in ip_hashfn() */ 525 KASSERT(ip->ip_len >= hlen, ("total length less then header length")); 526 KASSERT(m->m_pkthdr.len >= ip->ip_len, ("mbuf too short")); 527 528 /* 529 * Trim mbufs if longer than the IP header would have us expect. 530 */ 531 if (m->m_pkthdr.len > ip->ip_len) { 532 if (m->m_len == m->m_pkthdr.len) { 533 m->m_len = ip->ip_len; 534 m->m_pkthdr.len = ip->ip_len; 535 } else { 536 m_adj(m, ip->ip_len - m->m_pkthdr.len); 537 } 538 } 539 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF) 540 /* 541 * Bypass packet filtering for packets from a tunnel (gif). 542 */ 543 if (ipsec_gethist(m, NULL)) 544 goto pass; 545 #endif 546 547 /* 548 * IpHack's section. 549 * Right now when no processing on packet has done 550 * and it is still fresh out of network we do our black 551 * deals with it. 552 * - Firewall: deny/allow/divert 553 * - Xlate: translate packet's addr/port (NAT). 554 * - Pipe: pass pkt through dummynet. 555 * - Wrap: fake packet's addr/port <unimpl.> 556 * - Encapsulate: put it in another IP and send out. <unimp.> 557 */ 558 559 iphack: 560 /* 561 * If we've been forwarded from the output side, then 562 * skip the firewall a second time 563 */ 564 if (next_hop != NULL) 565 goto ours; 566 567 /* No pfil hooks */ 568 if (!pfil_has_hooks(&inet_pfil_hook)) { 569 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 570 /* 571 * Strip dummynet tags from stranded packets 572 */ 573 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 574 KKASSERT(mtag != NULL); 575 m_tag_delete(m, mtag); 576 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 577 } 578 goto pass; 579 } 580 581 /* 582 * Run through list of hooks for input packets. 583 * 584 * NOTE! If the packet is rewritten pf/ipfw/whoever must 585 * clear M_HASH. 586 */ 587 odst = ip->ip_dst; 588 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN)) 589 return; 590 if (m == NULL) /* consumed by filter */ 591 return; 592 ip = mtod(m, struct ip *); 593 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 594 using_srcrt = (odst.s_addr != ip->ip_dst.s_addr); 595 596 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 597 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 598 KKASSERT(mtag != NULL); 599 next_hop = m_tag_data(mtag); 600 } 601 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 602 ip_dn_queue(m); 603 return; 604 } 605 if (m->m_pkthdr.fw_flags & FW_MBUF_REDISPATCH) { 606 m->m_pkthdr.fw_flags &= ~FW_MBUF_REDISPATCH; 607 } 608 pass: 609 /* 610 * Process options and, if not destined for us, 611 * ship it on. ip_dooptions returns 1 when an 612 * error was detected (causing an icmp message 613 * to be sent and the original packet to be freed). 614 */ 615 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, next_hop)) 616 return; 617 618 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 619 * matter if it is destined to another node, or whether it is 620 * a multicast one, RSVP wants it! and prevents it from being forwarded 621 * anywhere else. Also checks if the rsvp daemon is running before 622 * grabbing the packet. 623 */ 624 if (rsvp_on && ip->ip_p == IPPROTO_RSVP) 625 goto ours; 626 627 /* 628 * Check our list of addresses, to see if the packet is for us. 629 * If we don't have any addresses, assume any unicast packet 630 * we receive might be for us (and let the upper layers deal 631 * with it). 632 */ 633 if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) && 634 !(m->m_flags & (M_MCAST | M_BCAST))) 635 goto ours; 636 637 /* 638 * Cache the destination address of the packet; this may be 639 * changed by use of 'ipfw fwd'. 640 */ 641 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst; 642 643 /* 644 * Enable a consistency check between the destination address 645 * and the arrival interface for a unicast packet (the RFC 1122 646 * strong ES model) if IP forwarding is disabled and the packet 647 * is not locally generated and the packet is not subject to 648 * 'ipfw fwd'. 649 * 650 * XXX - Checking also should be disabled if the destination 651 * address is ipnat'ed to a different interface. 652 * 653 * XXX - Checking is incompatible with IP aliases added 654 * to the loopback interface instead of the interface where 655 * the packets are received. 656 */ 657 checkif = ip_checkinterface && 658 !ipforwarding && 659 m->m_pkthdr.rcvif != NULL && 660 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) && 661 next_hop == NULL; 662 663 /* 664 * Check for exact addresses in the hash bucket. 665 */ 666 LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) { 667 ia = iac->ia; 668 669 /* 670 * If the address matches, verify that the packet 671 * arrived via the correct interface if checking is 672 * enabled. 673 */ 674 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 675 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 676 goto ours; 677 } 678 ia = NULL; 679 680 /* 681 * Check for broadcast addresses. 682 * 683 * Only accept broadcast packets that arrive via the matching 684 * interface. Reception of forwarded directed broadcasts would 685 * be handled via ip_forward() and ether_output() with the loopback 686 * into the stack for SIMPLEX interfaces handled by ether_output(). 687 */ 688 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 689 struct ifaddr_container *ifac; 690 691 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid], 692 ifa_link) { 693 struct ifaddr *ifa = ifac->ifa; 694 695 if (ifa->ifa_addr == NULL) /* shutdown/startup race */ 696 continue; 697 if (ifa->ifa_addr->sa_family != AF_INET) 698 continue; 699 ia = ifatoia(ifa); 700 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 701 pkt_dst.s_addr) 702 goto ours; 703 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 704 goto ours; 705 #ifdef BOOTP_COMPAT 706 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 707 goto ours; 708 #endif 709 } 710 } 711 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 712 struct in_multi *inm; 713 714 /* XXX Multicast is not MPSAFE yet */ 715 get_mplock(); 716 717 if (ip_mrouter != NULL) { 718 /* 719 * If we are acting as a multicast router, all 720 * incoming multicast packets are passed to the 721 * kernel-level multicast forwarding function. 722 * The packet is returned (relatively) intact; if 723 * ip_mforward() returns a non-zero value, the packet 724 * must be discarded, else it may be accepted below. 725 */ 726 if (ip_mforward != NULL && 727 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) { 728 rel_mplock(); 729 ipstat.ips_cantforward++; 730 m_freem(m); 731 return; 732 } 733 734 /* 735 * The process-level routing daemon needs to receive 736 * all multicast IGMP packets, whether or not this 737 * host belongs to their destination groups. 738 */ 739 if (ip->ip_p == IPPROTO_IGMP) { 740 rel_mplock(); 741 goto ours; 742 } 743 ipstat.ips_forward++; 744 } 745 /* 746 * See if we belong to the destination multicast group on the 747 * arrival interface. 748 */ 749 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 750 if (inm == NULL) { 751 rel_mplock(); 752 ipstat.ips_notmember++; 753 m_freem(m); 754 return; 755 } 756 757 rel_mplock(); 758 goto ours; 759 } 760 if (ip->ip_dst.s_addr == INADDR_BROADCAST) 761 goto ours; 762 if (ip->ip_dst.s_addr == INADDR_ANY) 763 goto ours; 764 765 /* 766 * FAITH(Firewall Aided Internet Translator) 767 */ 768 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 769 if (ip_keepfaith) { 770 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 771 goto ours; 772 } 773 m_freem(m); 774 return; 775 } 776 777 /* 778 * Not for us; forward if possible and desirable. 779 */ 780 if (!ipforwarding) { 781 ipstat.ips_cantforward++; 782 m_freem(m); 783 } else { 784 #ifdef IPSEC 785 /* 786 * Enforce inbound IPsec SPD. 787 */ 788 if (ipsec4_in_reject(m, NULL)) { 789 ipsecstat.in_polvio++; 790 goto bad; 791 } 792 #endif 793 #ifdef FAST_IPSEC 794 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 795 crit_enter(); 796 if (mtag != NULL) { 797 tdbi = (struct tdb_ident *)m_tag_data(mtag); 798 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 799 } else { 800 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 801 IP_FORWARDING, &error); 802 } 803 if (sp == NULL) { /* NB: can happen if error */ 804 crit_exit(); 805 /*XXX error stat???*/ 806 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 807 goto bad; 808 } 809 810 /* 811 * Check security policy against packet attributes. 812 */ 813 error = ipsec_in_reject(sp, m); 814 KEY_FREESP(&sp); 815 crit_exit(); 816 if (error) { 817 ipstat.ips_cantforward++; 818 goto bad; 819 } 820 #endif 821 ip_forward(m, using_srcrt, next_hop); 822 } 823 return; 824 825 ours: 826 827 /* 828 * IPSTEALTH: Process non-routing options only 829 * if the packet is destined for us. 830 */ 831 if (ipstealth && 832 hlen > sizeof(struct ip) && 833 ip_dooptions(m, 1, next_hop)) 834 return; 835 836 /* Count the packet in the ip address stats */ 837 if (ia != NULL) { 838 IFA_STAT_INC(&ia->ia_ifa, ipackets, 1); 839 IFA_STAT_INC(&ia->ia_ifa, ibytes, m->m_pkthdr.len); 840 } 841 842 /* 843 * If offset or IP_MF are set, must reassemble. 844 * Otherwise, nothing need be done. 845 * (We could look in the reassembly queue to see 846 * if the packet was previously fragmented, 847 * but it's not worth the time; just let them time out.) 848 */ 849 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 850 /* 851 * Attempt reassembly; if it succeeds, proceed. ip_reass() 852 * will return a different mbuf. 853 * 854 * NOTE: ip_reass() returns m with M_HASH cleared to force 855 * us to recharacterize the packet. 856 */ 857 m = ip_reass(m); 858 if (m == NULL) 859 return; 860 ip = mtod(m, struct ip *); 861 862 /* Get the header length of the reassembled packet */ 863 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 864 } else { 865 ip->ip_len -= hlen; 866 } 867 868 #ifdef IPSEC 869 /* 870 * enforce IPsec policy checking if we are seeing last header. 871 * note that we do not visit this with protocols with pcb layer 872 * code - like udp/tcp/raw ip. 873 */ 874 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) && 875 ipsec4_in_reject(m, NULL)) { 876 ipsecstat.in_polvio++; 877 goto bad; 878 } 879 #endif 880 #if FAST_IPSEC 881 /* 882 * enforce IPsec policy checking if we are seeing last header. 883 * note that we do not visit this with protocols with pcb layer 884 * code - like udp/tcp/raw ip. 885 */ 886 if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) { 887 /* 888 * Check if the packet has already had IPsec processing 889 * done. If so, then just pass it along. This tag gets 890 * set during AH, ESP, etc. input handling, before the 891 * packet is returned to the ip input queue for delivery. 892 */ 893 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 894 crit_enter(); 895 if (mtag != NULL) { 896 tdbi = (struct tdb_ident *)m_tag_data(mtag); 897 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 898 } else { 899 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 900 IP_FORWARDING, &error); 901 } 902 if (sp != NULL) { 903 /* 904 * Check security policy against packet attributes. 905 */ 906 error = ipsec_in_reject(sp, m); 907 KEY_FREESP(&sp); 908 } else { 909 /* XXX error stat??? */ 910 error = EINVAL; 911 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 912 goto bad; 913 } 914 crit_exit(); 915 if (error) 916 goto bad; 917 } 918 #endif /* FAST_IPSEC */ 919 920 /* 921 * We must forward the packet to the correct protocol thread if 922 * we are not already in it. 923 * 924 * NOTE: ip_len is now in host form. ip_len is not adjusted 925 * further for protocol processing, instead we pass hlen 926 * to the protosw and let it deal with it. 927 */ 928 ipstat.ips_delivered++; 929 930 if ((m->m_flags & M_HASH) == 0) { 931 #ifdef RSS_DEBUG 932 atomic_add_long(&ip_rehash_count, 1); 933 #endif 934 ip->ip_len = htons(ip->ip_len + hlen); 935 ip->ip_off = htons(ip->ip_off); 936 937 ip_hashfn(&m, 0, IP_MPORT_IN); 938 if (m == NULL) 939 return; 940 941 ip = mtod(m, struct ip *); 942 ip->ip_len = ntohs(ip->ip_len) - hlen; 943 ip->ip_off = ntohs(ip->ip_off); 944 KKASSERT(m->m_flags & M_HASH); 945 } 946 port = netisr_hashport(m->m_pkthdr.hash); 947 948 if (port != &curthread->td_msgport) { 949 struct netmsg_packet *pmsg; 950 951 #ifdef RSS_DEBUG 952 atomic_add_long(&ip_dispatch_slow, 1); 953 #endif 954 955 pmsg = &m->m_hdr.mh_netmsg; 956 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 957 0, transport_processing_handler); 958 pmsg->nm_packet = m; 959 pmsg->base.lmsg.u.ms_result = hlen; 960 lwkt_sendmsg(port, &pmsg->base.lmsg); 961 } else { 962 #ifdef RSS_DEBUG 963 atomic_add_long(&ip_dispatch_fast, 1); 964 #endif 965 transport_processing_oncpu(m, hlen, ip); 966 } 967 return; 968 969 bad: 970 m_freem(m); 971 } 972 973 /* 974 * Take incoming datagram fragment and try to reassemble it into 975 * whole datagram. If a chain for reassembly of this datagram already 976 * exists, then it is given as fp; otherwise have to make a chain. 977 */ 978 struct mbuf * 979 ip_reass(struct mbuf *m) 980 { 981 struct ip *ip = mtod(m, struct ip *); 982 struct mbuf *p = NULL, *q, *nq; 983 struct mbuf *n; 984 struct ipq *fp = NULL; 985 struct ipqhead *head; 986 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 987 int i, next; 988 u_short sum; 989 990 /* If maxnipq is 0, never accept fragments. */ 991 if (maxnipq == 0) { 992 ipstat.ips_fragments++; 993 ipstat.ips_fragdropped++; 994 m_freem(m); 995 return NULL; 996 } 997 998 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 999 /* 1000 * Look for queue of fragments of this datagram. 1001 */ 1002 lwkt_gettoken(&ipq_token); 1003 head = &ipq[sum]; 1004 TAILQ_FOREACH(fp, head, ipq_list) { 1005 if (ip->ip_id == fp->ipq_id && 1006 ip->ip_src.s_addr == fp->ipq_src.s_addr && 1007 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 1008 ip->ip_p == fp->ipq_p) 1009 goto found; 1010 } 1011 1012 fp = NULL; 1013 1014 /* 1015 * Enforce upper bound on number of fragmented packets 1016 * for which we attempt reassembly; 1017 * If maxnipq is -1, accept all fragments without limitation. 1018 */ 1019 if (nipq > maxnipq && maxnipq > 0) { 1020 /* 1021 * drop something from the tail of the current queue 1022 * before proceeding further 1023 */ 1024 struct ipq *q = TAILQ_LAST(head, ipqhead); 1025 if (q == NULL) { 1026 /* 1027 * The current queue is empty, 1028 * so drop from one of the others. 1029 */ 1030 for (i = 0; i < IPREASS_NHASH; i++) { 1031 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 1032 if (r) { 1033 ipstat.ips_fragtimeout += r->ipq_nfrags; 1034 ip_freef(&ipq[i], r); 1035 break; 1036 } 1037 } 1038 } else { 1039 ipstat.ips_fragtimeout += q->ipq_nfrags; 1040 ip_freef(head, q); 1041 } 1042 } 1043 found: 1044 /* 1045 * Adjust ip_len to not reflect header, 1046 * convert offset of this to bytes. 1047 */ 1048 ip->ip_len -= hlen; 1049 if (ip->ip_off & IP_MF) { 1050 /* 1051 * Make sure that fragments have a data length 1052 * that's a non-zero multiple of 8 bytes. 1053 */ 1054 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 1055 ipstat.ips_toosmall++; /* XXX */ 1056 m_freem(m); 1057 goto done; 1058 } 1059 m->m_flags |= M_FRAG; 1060 } else { 1061 m->m_flags &= ~M_FRAG; 1062 } 1063 ip->ip_off <<= 3; 1064 1065 ipstat.ips_fragments++; 1066 m->m_pkthdr.header = ip; 1067 1068 /* 1069 * If the hardware has not done csum over this fragment 1070 * then csum_data is not valid at all. 1071 */ 1072 if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) 1073 == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) { 1074 m->m_pkthdr.csum_data = 0; 1075 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 1076 } 1077 1078 /* 1079 * Presence of header sizes in mbufs 1080 * would confuse code below. 1081 */ 1082 m->m_data += hlen; 1083 m->m_len -= hlen; 1084 1085 /* 1086 * If first fragment to arrive, create a reassembly queue. 1087 */ 1088 if (fp == NULL) { 1089 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL) 1090 goto dropfrag; 1091 TAILQ_INSERT_HEAD(head, fp, ipq_list); 1092 nipq++; 1093 fp->ipq_nfrags = 1; 1094 fp->ipq_ttl = IPFRAGTTL; 1095 fp->ipq_p = ip->ip_p; 1096 fp->ipq_id = ip->ip_id; 1097 fp->ipq_src = ip->ip_src; 1098 fp->ipq_dst = ip->ip_dst; 1099 fp->ipq_frags = m; 1100 m->m_nextpkt = NULL; 1101 goto inserted; 1102 } else { 1103 fp->ipq_nfrags++; 1104 } 1105 1106 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1107 1108 /* 1109 * Find a segment which begins after this one does. 1110 */ 1111 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1112 if (GETIP(q)->ip_off > ip->ip_off) 1113 break; 1114 } 1115 1116 /* 1117 * If there is a preceding segment, it may provide some of 1118 * our data already. If so, drop the data from the incoming 1119 * segment. If it provides all of our data, drop us, otherwise 1120 * stick new segment in the proper place. 1121 * 1122 * If some of the data is dropped from the the preceding 1123 * segment, then it's checksum is invalidated. 1124 */ 1125 if (p) { 1126 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1127 if (i > 0) { 1128 if (i >= ip->ip_len) 1129 goto dropfrag; 1130 m_adj(m, i); 1131 m->m_pkthdr.csum_flags = 0; 1132 ip->ip_off += i; 1133 ip->ip_len -= i; 1134 } 1135 m->m_nextpkt = p->m_nextpkt; 1136 p->m_nextpkt = m; 1137 } else { 1138 m->m_nextpkt = fp->ipq_frags; 1139 fp->ipq_frags = m; 1140 } 1141 1142 /* 1143 * While we overlap succeeding segments trim them or, 1144 * if they are completely covered, dequeue them. 1145 */ 1146 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1147 q = nq) { 1148 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1149 if (i < GETIP(q)->ip_len) { 1150 GETIP(q)->ip_len -= i; 1151 GETIP(q)->ip_off += i; 1152 m_adj(q, i); 1153 q->m_pkthdr.csum_flags = 0; 1154 break; 1155 } 1156 nq = q->m_nextpkt; 1157 m->m_nextpkt = nq; 1158 ipstat.ips_fragdropped++; 1159 fp->ipq_nfrags--; 1160 q->m_nextpkt = NULL; 1161 m_freem(q); 1162 } 1163 1164 inserted: 1165 /* 1166 * Check for complete reassembly and perform frag per packet 1167 * limiting. 1168 * 1169 * Frag limiting is performed here so that the nth frag has 1170 * a chance to complete the packet before we drop the packet. 1171 * As a result, n+1 frags are actually allowed per packet, but 1172 * only n will ever be stored. (n = maxfragsperpacket.) 1173 * 1174 */ 1175 next = 0; 1176 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1177 if (GETIP(q)->ip_off != next) { 1178 if (fp->ipq_nfrags > maxfragsperpacket) { 1179 ipstat.ips_fragdropped += fp->ipq_nfrags; 1180 ip_freef(head, fp); 1181 } 1182 goto done; 1183 } 1184 next += GETIP(q)->ip_len; 1185 } 1186 /* Make sure the last packet didn't have the IP_MF flag */ 1187 if (p->m_flags & M_FRAG) { 1188 if (fp->ipq_nfrags > maxfragsperpacket) { 1189 ipstat.ips_fragdropped += fp->ipq_nfrags; 1190 ip_freef(head, fp); 1191 } 1192 goto done; 1193 } 1194 1195 /* 1196 * Reassembly is complete. Make sure the packet is a sane size. 1197 */ 1198 q = fp->ipq_frags; 1199 ip = GETIP(q); 1200 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 1201 ipstat.ips_toolong++; 1202 ipstat.ips_fragdropped += fp->ipq_nfrags; 1203 ip_freef(head, fp); 1204 goto done; 1205 } 1206 1207 /* 1208 * Concatenate fragments. 1209 */ 1210 m = q; 1211 n = m->m_next; 1212 m->m_next = NULL; 1213 m_cat(m, n); 1214 nq = q->m_nextpkt; 1215 q->m_nextpkt = NULL; 1216 for (q = nq; q != NULL; q = nq) { 1217 nq = q->m_nextpkt; 1218 q->m_nextpkt = NULL; 1219 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1220 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1221 m_cat(m, q); 1222 } 1223 1224 /* 1225 * Clean up the 1's complement checksum. Carry over 16 bits must 1226 * be added back. This assumes no more then 65535 packet fragments 1227 * were reassembled. A second carry can also occur (but not a third). 1228 */ 1229 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) + 1230 (m->m_pkthdr.csum_data >> 16); 1231 if (m->m_pkthdr.csum_data > 0xFFFF) 1232 m->m_pkthdr.csum_data -= 0xFFFF; 1233 1234 /* 1235 * Create header for new ip packet by 1236 * modifying header of first packet; 1237 * dequeue and discard fragment reassembly header. 1238 * Make header visible. 1239 */ 1240 ip->ip_len = next; 1241 ip->ip_src = fp->ipq_src; 1242 ip->ip_dst = fp->ipq_dst; 1243 TAILQ_REMOVE(head, fp, ipq_list); 1244 nipq--; 1245 mpipe_free(&ipq_mpipe, fp); 1246 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 1247 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 1248 /* some debugging cruft by sklower, below, will go away soon */ 1249 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1250 int plen = 0; 1251 1252 for (n = m; n; n = n->m_next) 1253 plen += n->m_len; 1254 m->m_pkthdr.len = plen; 1255 } 1256 1257 /* 1258 * Reassembly complete, return the next protocol. 1259 * 1260 * Be sure to clear M_HASH to force the packet 1261 * to be re-characterized. 1262 * 1263 * Clear M_FRAG, we are no longer a fragment. 1264 */ 1265 m->m_flags &= ~(M_HASH | M_FRAG); 1266 1267 ipstat.ips_reassembled++; 1268 lwkt_reltoken(&ipq_token); 1269 return (m); 1270 1271 dropfrag: 1272 ipstat.ips_fragdropped++; 1273 if (fp != NULL) 1274 fp->ipq_nfrags--; 1275 m_freem(m); 1276 done: 1277 lwkt_reltoken(&ipq_token); 1278 return (NULL); 1279 1280 #undef GETIP 1281 } 1282 1283 /* 1284 * Free a fragment reassembly header and all 1285 * associated datagrams. 1286 * 1287 * Called with ipq_token held. 1288 */ 1289 static void 1290 ip_freef(struct ipqhead *fhp, struct ipq *fp) 1291 { 1292 struct mbuf *q; 1293 1294 /* 1295 * Remove first to protect against blocking 1296 */ 1297 TAILQ_REMOVE(fhp, fp, ipq_list); 1298 1299 /* 1300 * Clean out at our leisure 1301 */ 1302 while (fp->ipq_frags) { 1303 q = fp->ipq_frags; 1304 fp->ipq_frags = q->m_nextpkt; 1305 q->m_nextpkt = NULL; 1306 m_freem(q); 1307 } 1308 mpipe_free(&ipq_mpipe, fp); 1309 nipq--; 1310 } 1311 1312 /* 1313 * IP timer processing; 1314 * if a timer expires on a reassembly 1315 * queue, discard it. 1316 */ 1317 void 1318 ip_slowtimo(void) 1319 { 1320 struct ipq *fp, *fp_temp; 1321 struct ipqhead *head; 1322 int i; 1323 1324 lwkt_gettoken(&ipq_token); 1325 for (i = 0; i < IPREASS_NHASH; i++) { 1326 head = &ipq[i]; 1327 TAILQ_FOREACH_MUTABLE(fp, head, ipq_list, fp_temp) { 1328 if (--fp->ipq_ttl == 0) { 1329 ipstat.ips_fragtimeout += fp->ipq_nfrags; 1330 ip_freef(head, fp); 1331 } 1332 } 1333 } 1334 /* 1335 * If we are over the maximum number of fragments 1336 * (due to the limit being lowered), drain off 1337 * enough to get down to the new limit. 1338 */ 1339 if (maxnipq >= 0 && nipq > maxnipq) { 1340 for (i = 0; i < IPREASS_NHASH; i++) { 1341 head = &ipq[i]; 1342 while (nipq > maxnipq && !TAILQ_EMPTY(head)) { 1343 ipstat.ips_fragdropped += 1344 TAILQ_FIRST(head)->ipq_nfrags; 1345 ip_freef(head, TAILQ_FIRST(head)); 1346 } 1347 } 1348 } 1349 lwkt_reltoken(&ipq_token); 1350 ipflow_slowtimo(); 1351 } 1352 1353 /* 1354 * Drain off all datagram fragments. 1355 */ 1356 void 1357 ip_drain(void) 1358 { 1359 struct ipqhead *head; 1360 int i; 1361 1362 lwkt_gettoken(&ipq_token); 1363 for (i = 0; i < IPREASS_NHASH; i++) { 1364 head = &ipq[i]; 1365 while (!TAILQ_EMPTY(head)) { 1366 ipstat.ips_fragdropped += TAILQ_FIRST(head)->ipq_nfrags; 1367 ip_freef(head, TAILQ_FIRST(head)); 1368 } 1369 } 1370 lwkt_reltoken(&ipq_token); 1371 in_rtqdrain(); 1372 } 1373 1374 /* 1375 * Do option processing on a datagram, 1376 * possibly discarding it if bad options are encountered, 1377 * or forwarding it if source-routed. 1378 * The pass argument is used when operating in the IPSTEALTH 1379 * mode to tell what options to process: 1380 * [LS]SRR (pass 0) or the others (pass 1). 1381 * The reason for as many as two passes is that when doing IPSTEALTH, 1382 * non-routing options should be processed only if the packet is for us. 1383 * Returns 1 if packet has been forwarded/freed, 1384 * 0 if the packet should be processed further. 1385 */ 1386 static int 1387 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop) 1388 { 1389 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET }; 1390 struct ip *ip = mtod(m, struct ip *); 1391 u_char *cp; 1392 struct in_ifaddr *ia; 1393 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB; 1394 boolean_t forward = FALSE; 1395 struct in_addr *sin, dst; 1396 n_time ntime; 1397 1398 dst = ip->ip_dst; 1399 cp = (u_char *)(ip + 1); 1400 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1401 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1402 opt = cp[IPOPT_OPTVAL]; 1403 if (opt == IPOPT_EOL) 1404 break; 1405 if (opt == IPOPT_NOP) 1406 optlen = 1; 1407 else { 1408 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1409 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1410 goto bad; 1411 } 1412 optlen = cp[IPOPT_OLEN]; 1413 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1414 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1415 goto bad; 1416 } 1417 } 1418 switch (opt) { 1419 1420 default: 1421 break; 1422 1423 /* 1424 * Source routing with record. 1425 * Find interface with current destination address. 1426 * If none on this machine then drop if strictly routed, 1427 * or do nothing if loosely routed. 1428 * Record interface address and bring up next address 1429 * component. If strictly routed make sure next 1430 * address is on directly accessible net. 1431 */ 1432 case IPOPT_LSRR: 1433 case IPOPT_SSRR: 1434 if (ipstealth && pass > 0) 1435 break; 1436 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1437 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1438 goto bad; 1439 } 1440 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1441 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1442 goto bad; 1443 } 1444 ipaddr.sin_addr = ip->ip_dst; 1445 ia = (struct in_ifaddr *) 1446 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1447 if (ia == NULL) { 1448 if (opt == IPOPT_SSRR) { 1449 type = ICMP_UNREACH; 1450 code = ICMP_UNREACH_SRCFAIL; 1451 goto bad; 1452 } 1453 if (!ip_dosourceroute) 1454 goto nosourcerouting; 1455 /* 1456 * Loose routing, and not at next destination 1457 * yet; nothing to do except forward. 1458 */ 1459 break; 1460 } 1461 off--; /* 0 origin */ 1462 if (off > optlen - (int)sizeof(struct in_addr)) { 1463 /* 1464 * End of source route. Should be for us. 1465 */ 1466 if (!ip_acceptsourceroute) 1467 goto nosourcerouting; 1468 save_rte(m, cp, ip->ip_src); 1469 break; 1470 } 1471 if (ipstealth) 1472 goto dropit; 1473 if (!ip_dosourceroute) { 1474 if (ipforwarding) { 1475 char buf[sizeof "aaa.bbb.ccc.ddd"]; 1476 1477 /* 1478 * Acting as a router, so generate ICMP 1479 */ 1480 nosourcerouting: 1481 strcpy(buf, inet_ntoa(ip->ip_dst)); 1482 log(LOG_WARNING, 1483 "attempted source route from %s to %s\n", 1484 inet_ntoa(ip->ip_src), buf); 1485 type = ICMP_UNREACH; 1486 code = ICMP_UNREACH_SRCFAIL; 1487 goto bad; 1488 } else { 1489 /* 1490 * Not acting as a router, 1491 * so silently drop. 1492 */ 1493 dropit: 1494 ipstat.ips_cantforward++; 1495 m_freem(m); 1496 return (1); 1497 } 1498 } 1499 1500 /* 1501 * locate outgoing interface 1502 */ 1503 memcpy(&ipaddr.sin_addr, cp + off, 1504 sizeof ipaddr.sin_addr); 1505 1506 if (opt == IPOPT_SSRR) { 1507 #define INA struct in_ifaddr * 1508 #define SA struct sockaddr * 1509 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) 1510 == NULL) 1511 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1512 } else { 1513 ia = ip_rtaddr(ipaddr.sin_addr, NULL); 1514 } 1515 if (ia == NULL) { 1516 type = ICMP_UNREACH; 1517 code = ICMP_UNREACH_SRCFAIL; 1518 goto bad; 1519 } 1520 ip->ip_dst = ipaddr.sin_addr; 1521 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1522 sizeof(struct in_addr)); 1523 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1524 /* 1525 * Let ip_intr's mcast routing check handle mcast pkts 1526 */ 1527 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1528 break; 1529 1530 case IPOPT_RR: 1531 if (ipstealth && pass == 0) 1532 break; 1533 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1534 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1535 goto bad; 1536 } 1537 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1538 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1539 goto bad; 1540 } 1541 /* 1542 * If no space remains, ignore. 1543 */ 1544 off--; /* 0 origin */ 1545 if (off > optlen - (int)sizeof(struct in_addr)) 1546 break; 1547 memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1548 sizeof ipaddr.sin_addr); 1549 /* 1550 * locate outgoing interface; if we're the destination, 1551 * use the incoming interface (should be same). 1552 */ 1553 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL && 1554 (ia = ip_rtaddr(ipaddr.sin_addr, NULL)) == NULL) { 1555 type = ICMP_UNREACH; 1556 code = ICMP_UNREACH_HOST; 1557 goto bad; 1558 } 1559 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1560 sizeof(struct in_addr)); 1561 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1562 break; 1563 1564 case IPOPT_TS: 1565 if (ipstealth && pass == 0) 1566 break; 1567 code = cp - (u_char *)ip; 1568 if (optlen < 4 || optlen > 40) { 1569 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1570 goto bad; 1571 } 1572 if ((off = cp[IPOPT_OFFSET]) < 5) { 1573 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1574 goto bad; 1575 } 1576 if (off > optlen - (int)sizeof(int32_t)) { 1577 cp[IPOPT_OFFSET + 1] += (1 << 4); 1578 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1579 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1580 goto bad; 1581 } 1582 break; 1583 } 1584 off--; /* 0 origin */ 1585 sin = (struct in_addr *)(cp + off); 1586 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1587 1588 case IPOPT_TS_TSONLY: 1589 break; 1590 1591 case IPOPT_TS_TSANDADDR: 1592 if (off + sizeof(n_time) + 1593 sizeof(struct in_addr) > optlen) { 1594 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1595 goto bad; 1596 } 1597 ipaddr.sin_addr = dst; 1598 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1599 m->m_pkthdr.rcvif); 1600 if (ia == NULL) 1601 continue; 1602 memcpy(sin, &IA_SIN(ia)->sin_addr, 1603 sizeof(struct in_addr)); 1604 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1605 off += sizeof(struct in_addr); 1606 break; 1607 1608 case IPOPT_TS_PRESPEC: 1609 if (off + sizeof(n_time) + 1610 sizeof(struct in_addr) > optlen) { 1611 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1612 goto bad; 1613 } 1614 memcpy(&ipaddr.sin_addr, sin, 1615 sizeof(struct in_addr)); 1616 if (ifa_ifwithaddr((SA)&ipaddr) == NULL) 1617 continue; 1618 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1619 off += sizeof(struct in_addr); 1620 break; 1621 1622 default: 1623 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1624 goto bad; 1625 } 1626 ntime = iptime(); 1627 memcpy(cp + off, &ntime, sizeof(n_time)); 1628 cp[IPOPT_OFFSET] += sizeof(n_time); 1629 } 1630 } 1631 if (forward && ipforwarding) { 1632 ip_forward(m, TRUE, next_hop); 1633 return (1); 1634 } 1635 return (0); 1636 bad: 1637 icmp_error(m, type, code, 0, 0); 1638 ipstat.ips_badoptions++; 1639 return (1); 1640 } 1641 1642 /* 1643 * Given address of next destination (final or next hop), 1644 * return internet address info of interface to be used to get there. 1645 */ 1646 struct in_ifaddr * 1647 ip_rtaddr(struct in_addr dst, struct route *ro0) 1648 { 1649 struct route sro, *ro; 1650 struct sockaddr_in *sin; 1651 struct in_ifaddr *ia; 1652 1653 if (ro0 != NULL) { 1654 ro = ro0; 1655 } else { 1656 bzero(&sro, sizeof(sro)); 1657 ro = &sro; 1658 } 1659 1660 sin = (struct sockaddr_in *)&ro->ro_dst; 1661 1662 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) { 1663 if (ro->ro_rt != NULL) { 1664 RTFREE(ro->ro_rt); 1665 ro->ro_rt = NULL; 1666 } 1667 sin->sin_family = AF_INET; 1668 sin->sin_len = sizeof *sin; 1669 sin->sin_addr = dst; 1670 rtalloc_ign(ro, RTF_PRCLONING); 1671 } 1672 1673 if (ro->ro_rt == NULL) 1674 return (NULL); 1675 1676 ia = ifatoia(ro->ro_rt->rt_ifa); 1677 1678 if (ro == &sro) 1679 RTFREE(ro->ro_rt); 1680 return ia; 1681 } 1682 1683 /* 1684 * Save incoming source route for use in replies, 1685 * to be picked up later by ip_srcroute if the receiver is interested. 1686 */ 1687 static void 1688 save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1689 { 1690 struct m_tag *mtag; 1691 struct ip_srcrt_opt *opt; 1692 unsigned olen; 1693 1694 mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), MB_DONTWAIT); 1695 if (mtag == NULL) 1696 return; 1697 opt = m_tag_data(mtag); 1698 1699 olen = option[IPOPT_OLEN]; 1700 #ifdef DIAGNOSTIC 1701 if (ipprintfs) 1702 kprintf("save_rte: olen %d\n", olen); 1703 #endif 1704 if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) { 1705 m_tag_free(mtag); 1706 return; 1707 } 1708 bcopy(option, opt->ip_srcrt.srcopt, olen); 1709 opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1710 opt->ip_srcrt.dst = dst; 1711 m_tag_prepend(m, mtag); 1712 } 1713 1714 /* 1715 * Retrieve incoming source route for use in replies, 1716 * in the same form used by setsockopt. 1717 * The first hop is placed before the options, will be removed later. 1718 */ 1719 struct mbuf * 1720 ip_srcroute(struct mbuf *m0) 1721 { 1722 struct in_addr *p, *q; 1723 struct mbuf *m; 1724 struct m_tag *mtag; 1725 struct ip_srcrt_opt *opt; 1726 1727 if (m0 == NULL) 1728 return NULL; 1729 1730 mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL); 1731 if (mtag == NULL) 1732 return NULL; 1733 opt = m_tag_data(mtag); 1734 1735 if (opt->ip_nhops == 0) 1736 return (NULL); 1737 m = m_get(MB_DONTWAIT, MT_HEADER); 1738 if (m == NULL) 1739 return (NULL); 1740 1741 #define OPTSIZ (sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt)) 1742 1743 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1744 m->m_len = opt->ip_nhops * sizeof(struct in_addr) + 1745 sizeof(struct in_addr) + OPTSIZ; 1746 #ifdef DIAGNOSTIC 1747 if (ipprintfs) { 1748 kprintf("ip_srcroute: nhops %d mlen %d", 1749 opt->ip_nhops, m->m_len); 1750 } 1751 #endif 1752 1753 /* 1754 * First save first hop for return route 1755 */ 1756 p = &opt->ip_srcrt.route[opt->ip_nhops - 1]; 1757 *(mtod(m, struct in_addr *)) = *p--; 1758 #ifdef DIAGNOSTIC 1759 if (ipprintfs) 1760 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1761 #endif 1762 1763 /* 1764 * Copy option fields and padding (nop) to mbuf. 1765 */ 1766 opt->ip_srcrt.nop = IPOPT_NOP; 1767 opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1768 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop, 1769 OPTSIZ); 1770 q = (struct in_addr *)(mtod(m, caddr_t) + 1771 sizeof(struct in_addr) + OPTSIZ); 1772 #undef OPTSIZ 1773 /* 1774 * Record return path as an IP source route, 1775 * reversing the path (pointers are now aligned). 1776 */ 1777 while (p >= opt->ip_srcrt.route) { 1778 #ifdef DIAGNOSTIC 1779 if (ipprintfs) 1780 kprintf(" %x", ntohl(q->s_addr)); 1781 #endif 1782 *q++ = *p--; 1783 } 1784 /* 1785 * Last hop goes to final destination. 1786 */ 1787 *q = opt->ip_srcrt.dst; 1788 m_tag_delete(m0, mtag); 1789 #ifdef DIAGNOSTIC 1790 if (ipprintfs) 1791 kprintf(" %x\n", ntohl(q->s_addr)); 1792 #endif 1793 return (m); 1794 } 1795 1796 /* 1797 * Strip out IP options. 1798 */ 1799 void 1800 ip_stripoptions(struct mbuf *m) 1801 { 1802 int datalen; 1803 struct ip *ip = mtod(m, struct ip *); 1804 caddr_t opts; 1805 int optlen; 1806 1807 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1808 opts = (caddr_t)(ip + 1); 1809 datalen = m->m_len - (sizeof(struct ip) + optlen); 1810 bcopy(opts + optlen, opts, datalen); 1811 m->m_len -= optlen; 1812 if (m->m_flags & M_PKTHDR) 1813 m->m_pkthdr.len -= optlen; 1814 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1815 } 1816 1817 u_char inetctlerrmap[PRC_NCMDS] = { 1818 0, 0, 0, 0, 1819 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1820 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1821 EMSGSIZE, EHOSTUNREACH, 0, 0, 1822 0, 0, 0, 0, 1823 ENOPROTOOPT, ECONNREFUSED 1824 }; 1825 1826 /* 1827 * Forward a packet. If some error occurs return the sender 1828 * an icmp packet. Note we can't always generate a meaningful 1829 * icmp message because icmp doesn't have a large enough repertoire 1830 * of codes and types. 1831 * 1832 * If not forwarding, just drop the packet. This could be confusing 1833 * if ipforwarding was zero but some routing protocol was advancing 1834 * us as a gateway to somewhere. However, we must let the routing 1835 * protocol deal with that. 1836 * 1837 * The using_srcrt parameter indicates whether the packet is being forwarded 1838 * via a source route. 1839 */ 1840 void 1841 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop) 1842 { 1843 struct ip *ip = mtod(m, struct ip *); 1844 struct rtentry *rt; 1845 struct route fwd_ro; 1846 int error, type = 0, code = 0, destmtu = 0; 1847 struct mbuf *mcopy, *mtemp = NULL; 1848 n_long dest; 1849 struct in_addr pkt_dst; 1850 1851 dest = INADDR_ANY; 1852 /* 1853 * Cache the destination address of the packet; this may be 1854 * changed by use of 'ipfw fwd'. 1855 */ 1856 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst; 1857 1858 #ifdef DIAGNOSTIC 1859 if (ipprintfs) 1860 kprintf("forward: src %x dst %x ttl %x\n", 1861 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl); 1862 #endif 1863 1864 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) { 1865 ipstat.ips_cantforward++; 1866 m_freem(m); 1867 return; 1868 } 1869 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) { 1870 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1871 return; 1872 } 1873 1874 bzero(&fwd_ro, sizeof(fwd_ro)); 1875 ip_rtaddr(pkt_dst, &fwd_ro); 1876 if (fwd_ro.ro_rt == NULL) { 1877 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1878 return; 1879 } 1880 rt = fwd_ro.ro_rt; 1881 1882 if (curthread->td_type == TD_TYPE_NETISR) { 1883 /* 1884 * Save the IP header and at most 8 bytes of the payload, 1885 * in case we need to generate an ICMP message to the src. 1886 */ 1887 mtemp = ipforward_mtemp[mycpuid]; 1888 KASSERT((mtemp->m_flags & M_EXT) == 0 && 1889 mtemp->m_data == mtemp->m_pktdat && 1890 m_tag_first(mtemp) == NULL, 1891 ("ip_forward invalid mtemp1")); 1892 1893 if (!m_dup_pkthdr(mtemp, m, MB_DONTWAIT)) { 1894 /* 1895 * It's probably ok if the pkthdr dup fails (because 1896 * the deep copy of the tag chain failed), but for now 1897 * be conservative and just discard the copy since 1898 * code below may some day want the tags. 1899 */ 1900 mtemp = NULL; 1901 } else { 1902 mtemp->m_type = m->m_type; 1903 mtemp->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8, 1904 (int)ip->ip_len); 1905 mtemp->m_pkthdr.len = mtemp->m_len; 1906 m_copydata(m, 0, mtemp->m_len, mtod(mtemp, caddr_t)); 1907 } 1908 } 1909 1910 if (!ipstealth) 1911 ip->ip_ttl -= IPTTLDEC; 1912 1913 /* 1914 * If forwarding packet using same interface that it came in on, 1915 * perhaps should send a redirect to sender to shortcut a hop. 1916 * Only send redirect if source is sending directly to us, 1917 * and if packet was not source routed (or has any options). 1918 * Also, don't send redirect if forwarding using a default route 1919 * or a route modified by a redirect. 1920 */ 1921 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1922 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) && 1923 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY && 1924 ipsendredirects && !using_srcrt && next_hop == NULL) { 1925 u_long src = ntohl(ip->ip_src.s_addr); 1926 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa; 1927 1928 if (rt_ifa != NULL && 1929 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) { 1930 if (rt->rt_flags & RTF_GATEWAY) 1931 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1932 else 1933 dest = pkt_dst.s_addr; 1934 /* 1935 * Router requirements says to only send 1936 * host redirects. 1937 */ 1938 type = ICMP_REDIRECT; 1939 code = ICMP_REDIRECT_HOST; 1940 #ifdef DIAGNOSTIC 1941 if (ipprintfs) 1942 kprintf("redirect (%d) to %x\n", code, dest); 1943 #endif 1944 } 1945 } 1946 1947 error = ip_output(m, NULL, &fwd_ro, IP_FORWARDING, NULL, NULL); 1948 if (error == 0) { 1949 ipstat.ips_forward++; 1950 if (type == 0) { 1951 if (mtemp) 1952 ipflow_create(&fwd_ro, mtemp); 1953 goto done; 1954 } else { 1955 ipstat.ips_redirectsent++; 1956 } 1957 } else { 1958 ipstat.ips_cantforward++; 1959 } 1960 1961 if (mtemp == NULL) 1962 goto done; 1963 1964 /* 1965 * Errors that do not require generating ICMP message 1966 */ 1967 switch (error) { 1968 case ENOBUFS: 1969 /* 1970 * A router should not generate ICMP_SOURCEQUENCH as 1971 * required in RFC1812 Requirements for IP Version 4 Routers. 1972 * Source quench could be a big problem under DoS attacks, 1973 * or if the underlying interface is rate-limited. 1974 * Those who need source quench packets may re-enable them 1975 * via the net.inet.ip.sendsourcequench sysctl. 1976 */ 1977 if (!ip_sendsourcequench) 1978 goto done; 1979 break; 1980 1981 case EACCES: /* ipfw denied packet */ 1982 goto done; 1983 } 1984 1985 KASSERT((mtemp->m_flags & M_EXT) == 0 && 1986 mtemp->m_data == mtemp->m_pktdat, 1987 ("ip_forward invalid mtemp2")); 1988 mcopy = m_copym(mtemp, 0, mtemp->m_len, MB_DONTWAIT); 1989 if (mcopy == NULL) 1990 goto done; 1991 1992 /* 1993 * Send ICMP message. 1994 */ 1995 switch (error) { 1996 case 0: /* forwarded, but need redirect */ 1997 /* type, code set above */ 1998 break; 1999 2000 case ENETUNREACH: /* shouldn't happen, checked above */ 2001 case EHOSTUNREACH: 2002 case ENETDOWN: 2003 case EHOSTDOWN: 2004 default: 2005 type = ICMP_UNREACH; 2006 code = ICMP_UNREACH_HOST; 2007 break; 2008 2009 case EMSGSIZE: 2010 type = ICMP_UNREACH; 2011 code = ICMP_UNREACH_NEEDFRAG; 2012 #ifdef IPSEC 2013 /* 2014 * If the packet is routed over IPsec tunnel, tell the 2015 * originator the tunnel MTU. 2016 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 2017 * XXX quickhack!!! 2018 */ 2019 if (fwd_ro.ro_rt != NULL) { 2020 struct secpolicy *sp = NULL; 2021 int ipsecerror; 2022 int ipsechdr; 2023 struct route *ro; 2024 2025 sp = ipsec4_getpolicybyaddr(mcopy, 2026 IPSEC_DIR_OUTBOUND, 2027 IP_FORWARDING, 2028 &ipsecerror); 2029 2030 if (sp == NULL) 2031 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2032 else { 2033 /* count IPsec header size */ 2034 ipsechdr = ipsec4_hdrsiz(mcopy, 2035 IPSEC_DIR_OUTBOUND, 2036 NULL); 2037 2038 /* 2039 * find the correct route for outer IPv4 2040 * header, compute tunnel MTU. 2041 * 2042 */ 2043 if (sp->req != NULL && sp->req->sav != NULL && 2044 sp->req->sav->sah != NULL) { 2045 ro = &sp->req->sav->sah->sa_route; 2046 if (ro->ro_rt != NULL && 2047 ro->ro_rt->rt_ifp != NULL) { 2048 destmtu = 2049 ro->ro_rt->rt_ifp->if_mtu; 2050 destmtu -= ipsechdr; 2051 } 2052 } 2053 2054 key_freesp(sp); 2055 } 2056 } 2057 #elif FAST_IPSEC 2058 /* 2059 * If the packet is routed over IPsec tunnel, tell the 2060 * originator the tunnel MTU. 2061 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 2062 * XXX quickhack!!! 2063 */ 2064 if (fwd_ro.ro_rt != NULL) { 2065 struct secpolicy *sp = NULL; 2066 int ipsecerror; 2067 int ipsechdr; 2068 struct route *ro; 2069 2070 sp = ipsec_getpolicybyaddr(mcopy, 2071 IPSEC_DIR_OUTBOUND, 2072 IP_FORWARDING, 2073 &ipsecerror); 2074 2075 if (sp == NULL) 2076 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2077 else { 2078 /* count IPsec header size */ 2079 ipsechdr = ipsec4_hdrsiz(mcopy, 2080 IPSEC_DIR_OUTBOUND, 2081 NULL); 2082 2083 /* 2084 * find the correct route for outer IPv4 2085 * header, compute tunnel MTU. 2086 */ 2087 2088 if (sp->req != NULL && 2089 sp->req->sav != NULL && 2090 sp->req->sav->sah != NULL) { 2091 ro = &sp->req->sav->sah->sa_route; 2092 if (ro->ro_rt != NULL && 2093 ro->ro_rt->rt_ifp != NULL) { 2094 destmtu = 2095 ro->ro_rt->rt_ifp->if_mtu; 2096 destmtu -= ipsechdr; 2097 } 2098 } 2099 2100 KEY_FREESP(&sp); 2101 } 2102 } 2103 #else /* !IPSEC && !FAST_IPSEC */ 2104 if (fwd_ro.ro_rt != NULL) 2105 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2106 #endif /*IPSEC*/ 2107 ipstat.ips_cantfrag++; 2108 break; 2109 2110 case ENOBUFS: 2111 type = ICMP_SOURCEQUENCH; 2112 code = 0; 2113 break; 2114 2115 case EACCES: /* ipfw denied packet */ 2116 panic("ip_forward EACCES should not reach"); 2117 } 2118 icmp_error(mcopy, type, code, dest, destmtu); 2119 done: 2120 if (mtemp != NULL) 2121 m_tag_delete_chain(mtemp); 2122 if (fwd_ro.ro_rt != NULL) 2123 RTFREE(fwd_ro.ro_rt); 2124 } 2125 2126 void 2127 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2128 struct mbuf *m) 2129 { 2130 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2131 struct timeval tv; 2132 2133 microtime(&tv); 2134 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2135 SCM_TIMESTAMP, SOL_SOCKET); 2136 if (*mp) 2137 mp = &(*mp)->m_next; 2138 } 2139 if (inp->inp_flags & INP_RECVDSTADDR) { 2140 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2141 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2142 if (*mp) 2143 mp = &(*mp)->m_next; 2144 } 2145 if (inp->inp_flags & INP_RECVTTL) { 2146 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 2147 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 2148 if (*mp) 2149 mp = &(*mp)->m_next; 2150 } 2151 #ifdef notyet 2152 /* XXX 2153 * Moving these out of udp_input() made them even more broken 2154 * than they already were. 2155 */ 2156 /* options were tossed already */ 2157 if (inp->inp_flags & INP_RECVOPTS) { 2158 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2159 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2160 if (*mp) 2161 mp = &(*mp)->m_next; 2162 } 2163 /* ip_srcroute doesn't do what we want here, need to fix */ 2164 if (inp->inp_flags & INP_RECVRETOPTS) { 2165 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 2166 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2167 if (*mp) 2168 mp = &(*mp)->m_next; 2169 } 2170 #endif 2171 if (inp->inp_flags & INP_RECVIF) { 2172 struct ifnet *ifp; 2173 struct sdlbuf { 2174 struct sockaddr_dl sdl; 2175 u_char pad[32]; 2176 } sdlbuf; 2177 struct sockaddr_dl *sdp; 2178 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2179 2180 if (((ifp = m->m_pkthdr.rcvif)) && 2181 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) { 2182 sdp = IF_LLSOCKADDR(ifp); 2183 /* 2184 * Change our mind and don't try copy. 2185 */ 2186 if ((sdp->sdl_family != AF_LINK) || 2187 (sdp->sdl_len > sizeof(sdlbuf))) { 2188 goto makedummy; 2189 } 2190 bcopy(sdp, sdl2, sdp->sdl_len); 2191 } else { 2192 makedummy: 2193 sdl2->sdl_len = 2194 offsetof(struct sockaddr_dl, sdl_data[0]); 2195 sdl2->sdl_family = AF_LINK; 2196 sdl2->sdl_index = 0; 2197 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2198 } 2199 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 2200 IP_RECVIF, IPPROTO_IP); 2201 if (*mp) 2202 mp = &(*mp)->m_next; 2203 } 2204 } 2205 2206 /* 2207 * XXX these routines are called from the upper part of the kernel. 2208 * 2209 * They could also be moved to ip_mroute.c, since all the RSVP 2210 * handling is done there already. 2211 */ 2212 int 2213 ip_rsvp_init(struct socket *so) 2214 { 2215 if (so->so_type != SOCK_RAW || 2216 so->so_proto->pr_protocol != IPPROTO_RSVP) 2217 return EOPNOTSUPP; 2218 2219 if (ip_rsvpd != NULL) 2220 return EADDRINUSE; 2221 2222 ip_rsvpd = so; 2223 /* 2224 * This may seem silly, but we need to be sure we don't over-increment 2225 * the RSVP counter, in case something slips up. 2226 */ 2227 if (!ip_rsvp_on) { 2228 ip_rsvp_on = 1; 2229 rsvp_on++; 2230 } 2231 2232 return 0; 2233 } 2234 2235 int 2236 ip_rsvp_done(void) 2237 { 2238 ip_rsvpd = NULL; 2239 /* 2240 * This may seem silly, but we need to be sure we don't over-decrement 2241 * the RSVP counter, in case something slips up. 2242 */ 2243 if (ip_rsvp_on) { 2244 ip_rsvp_on = 0; 2245 rsvp_on--; 2246 } 2247 return 0; 2248 } 2249 2250 int 2251 rsvp_input(struct mbuf **mp, int *offp, int proto) 2252 { 2253 struct mbuf *m = *mp; 2254 2255 *mp = NULL; 2256 2257 if (rsvp_input_p) { /* call the real one if loaded */ 2258 *mp = m; 2259 rsvp_input_p(mp, offp, proto); 2260 return(IPPROTO_DONE); 2261 } 2262 2263 /* Can still get packets with rsvp_on = 0 if there is a local member 2264 * of the group to which the RSVP packet is addressed. But in this 2265 * case we want to throw the packet away. 2266 */ 2267 2268 if (!rsvp_on) { 2269 m_freem(m); 2270 return(IPPROTO_DONE); 2271 } 2272 2273 if (ip_rsvpd != NULL) { 2274 *mp = m; 2275 rip_input(mp, offp, proto); 2276 return(IPPROTO_DONE); 2277 } 2278 /* Drop the packet */ 2279 m_freem(m); 2280 return(IPPROTO_DONE); 2281 } 2282