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 78 #include <sys/param.h> 79 #include <sys/systm.h> 80 #include <sys/mbuf.h> 81 #include <sys/malloc.h> 82 #include <sys/mpipe.h> 83 #include <sys/domain.h> 84 #include <sys/protosw.h> 85 #include <sys/socket.h> 86 #include <sys/time.h> 87 #include <sys/globaldata.h> 88 #include <sys/thread.h> 89 #include <sys/kernel.h> 90 #include <sys/syslog.h> 91 #include <sys/sysctl.h> 92 #include <sys/in_cksum.h> 93 #include <sys/lock.h> 94 95 #include <sys/mplock2.h> 96 97 #include <machine/stdarg.h> 98 99 #include <net/if.h> 100 #include <net/if_types.h> 101 #include <net/if_var.h> 102 #include <net/if_dl.h> 103 #include <net/pfil.h> 104 #include <net/route.h> 105 #include <net/netisr.h> 106 107 #include <netinet/in.h> 108 #include <netinet/in_systm.h> 109 #include <netinet/in_var.h> 110 #include <netinet/ip.h> 111 #include <netinet/in_pcb.h> 112 #include <netinet/ip_var.h> 113 #include <netinet/ip_icmp.h> 114 #include <netinet/ip_divert.h> 115 #include <netinet/ip_flow.h> 116 117 #include <sys/thread2.h> 118 #include <sys/msgport2.h> 119 #include <net/netmsg2.h> 120 121 #include <sys/socketvar.h> 122 123 #include <net/ipfw/ip_fw.h> 124 #include <net/dummynet/ip_dummynet.h> 125 126 #ifdef IPSEC 127 #include <netinet6/ipsec.h> 128 #include <netproto/key/key.h> 129 #endif 130 131 #ifdef FAST_IPSEC 132 #include <netproto/ipsec/ipsec.h> 133 #include <netproto/ipsec/key.h> 134 #endif 135 136 int rsvp_on = 0; 137 static int ip_rsvp_on; 138 struct socket *ip_rsvpd; 139 140 int ipforwarding = 0; 141 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 142 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 143 144 static int ipsendredirects = 1; /* XXX */ 145 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 146 &ipsendredirects, 0, "Enable sending IP redirects"); 147 148 int ip_defttl = IPDEFTTL; 149 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 150 &ip_defttl, 0, "Maximum TTL on IP packets"); 151 152 static int ip_dosourceroute = 0; 153 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 154 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 155 156 static int ip_acceptsourceroute = 0; 157 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 158 CTLFLAG_RW, &ip_acceptsourceroute, 0, 159 "Enable accepting source routed IP packets"); 160 161 static int ip_keepfaith = 0; 162 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 163 &ip_keepfaith, 0, 164 "Enable packet capture for FAITH IPv4->IPv6 translator daemon"); 165 166 static int nipq = 0; /* total # of reass queues */ 167 static int maxnipq; 168 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 169 &maxnipq, 0, 170 "Maximum number of IPv4 fragment reassembly queue entries"); 171 172 static int maxfragsperpacket; 173 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 174 &maxfragsperpacket, 0, 175 "Maximum number of IPv4 fragments allowed per packet"); 176 177 static int ip_sendsourcequench = 0; 178 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 179 &ip_sendsourcequench, 0, 180 "Enable the transmission of source quench packets"); 181 182 int ip_do_randomid = 1; 183 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 184 &ip_do_randomid, 0, 185 "Assign random ip_id values"); 186 /* 187 * XXX - Setting ip_checkinterface mostly implements the receive side of 188 * the Strong ES model described in RFC 1122, but since the routing table 189 * and transmit implementation do not implement the Strong ES model, 190 * setting this to 1 results in an odd hybrid. 191 * 192 * XXX - ip_checkinterface currently must be disabled if you use ipnat 193 * to translate the destination address to another local interface. 194 * 195 * XXX - ip_checkinterface must be disabled if you add IP aliases 196 * to the loopback interface instead of the interface where the 197 * packets for those addresses are received. 198 */ 199 static int ip_checkinterface = 0; 200 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 201 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 202 203 static int ip_dispatch_fast = 0; 204 static int ip_dispatch_slow = 0; 205 static int ip_dispatch_recheck = 0; 206 static int ip_dispatch_software = 0; 207 SYSCTL_INT(_net_inet_ip, OID_AUTO, dispatch_fast_count, CTLFLAG_RD, 208 &ip_dispatch_fast, 0, 209 "Number of IP dispatches handled on current CPU"); 210 SYSCTL_INT(_net_inet_ip, OID_AUTO, dispatch_slow_count, CTLFLAG_RD, 211 &ip_dispatch_slow, 0, 212 "Number of IP dispatches messaged to another CPU"); 213 SYSCTL_INT(_net_inet_ip, OID_AUTO, dispatch_software_count, CTLFLAG_RD, 214 &ip_dispatch_software, 0, ""); 215 SYSCTL_INT(_net_inet_ip, OID_AUTO, dispatch_recheck_count, CTLFLAG_RD, 216 &ip_dispatch_recheck, 0, ""); 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 struct ip_stats ipstats_percpu[MAXCPU]; 233 #ifdef SMP 234 static int 235 sysctl_ipstats(SYSCTL_HANDLER_ARGS) 236 { 237 int cpu, error = 0; 238 239 for (cpu = 0; cpu < ncpus; ++cpu) { 240 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu], 241 sizeof(struct ip_stats)))) 242 break; 243 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu], 244 sizeof(struct ip_stats)))) 245 break; 246 } 247 248 return (error); 249 } 250 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW), 251 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics"); 252 #else 253 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 254 &ipstat, ip_stats, "IP statistics"); 255 #endif 256 257 /* Packet reassembly stuff */ 258 #define IPREASS_NHASH_LOG2 6 259 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 260 #define IPREASS_HMASK (IPREASS_NHASH - 1) 261 #define IPREASS_HASH(x,y) \ 262 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 263 264 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 265 266 #ifdef IPCTL_DEFMTU 267 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 268 &ip_mtu, 0, "Default MTU"); 269 #endif 270 271 #ifdef IPSTEALTH 272 static int ipstealth = 0; 273 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, ""); 274 #else 275 static const int ipstealth = 0; 276 #endif 277 278 struct mbuf *(*ip_divert_p)(struct mbuf *, int, int); 279 280 struct pfil_head inet_pfil_hook; 281 282 /* 283 * struct ip_srcrt_opt is used to store packet state while it travels 284 * through the stack. 285 * 286 * XXX Note that the code even makes assumptions on the size and 287 * alignment of fields inside struct ip_srcrt so e.g. adding some 288 * fields will break the code. This needs to be fixed. 289 * 290 * We need to save the IP options in case a protocol wants to respond 291 * to an incoming packet over the same route if the packet got here 292 * using IP source routing. This allows connection establishment and 293 * maintenance when the remote end is on a network that is not known 294 * to us. 295 */ 296 struct ip_srcrt { 297 struct in_addr dst; /* final destination */ 298 char nop; /* one NOP to align */ 299 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 300 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 301 }; 302 303 struct ip_srcrt_opt { 304 int ip_nhops; 305 struct ip_srcrt ip_srcrt; 306 }; 307 308 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management"); 309 static struct malloc_pipe ipq_mpipe; 310 311 static void save_rte(struct mbuf *, u_char *, struct in_addr); 312 static int ip_dooptions(struct mbuf *m, int, struct sockaddr_in *); 313 static void ip_freef(struct ipqhead *, struct ipq *); 314 static void ip_input_handler(netmsg_t); 315 316 /* 317 * IP initialization: fill in IP protocol switch table. 318 * All protocols not implemented in kernel go to raw IP protocol handler. 319 */ 320 void 321 ip_init(void) 322 { 323 struct protosw *pr; 324 int i; 325 #ifdef SMP 326 int cpu; 327 #endif 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 /* 369 * Initialize IP statistics counters for each CPU. 370 * 371 */ 372 #ifdef SMP 373 for (cpu = 0; cpu < ncpus; ++cpu) { 374 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats)); 375 } 376 #else 377 bzero(&ipstat, sizeof(struct ip_stats)); 378 #endif 379 380 netisr_register(NETISR_IP, ip_input_handler, ip_cpufn_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 ++ip_dispatch_software; 451 ip_cpufn(&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_cpufn() */ 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_cpufn() */ 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_cpufn() */ 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 ia->ia_ifa.if_ipackets++; 839 ia->ia_ifa.if_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 ++ip_dispatch_recheck; 932 ip->ip_len = htons(ip->ip_len + hlen); 933 ip->ip_off = htons(ip->ip_off); 934 935 ip_cpufn(&m, 0, IP_MPORT_IN); 936 if (m == NULL) 937 return; 938 939 ip = mtod(m, struct ip *); 940 ip->ip_len = ntohs(ip->ip_len) - hlen; 941 ip->ip_off = ntohs(ip->ip_off); 942 KKASSERT(m->m_flags & M_HASH); 943 } 944 port = cpu_portfn(m->m_pkthdr.hash); 945 946 if (port != &curthread->td_msgport) { 947 struct netmsg_packet *pmsg; 948 949 ++ip_dispatch_slow; 950 951 pmsg = &m->m_hdr.mh_netmsg; 952 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 953 0, transport_processing_handler); 954 pmsg->nm_packet = m; 955 pmsg->base.lmsg.u.ms_result = hlen; 956 lwkt_sendmsg(port, &pmsg->base.lmsg); 957 } else { 958 ++ip_dispatch_fast; 959 transport_processing_oncpu(m, hlen, ip); 960 } 961 return; 962 963 bad: 964 m_freem(m); 965 } 966 967 /* 968 * Take incoming datagram fragment and try to reassemble it into 969 * whole datagram. If a chain for reassembly of this datagram already 970 * exists, then it is given as fp; otherwise have to make a chain. 971 */ 972 struct mbuf * 973 ip_reass(struct mbuf *m) 974 { 975 struct ip *ip = mtod(m, struct ip *); 976 struct mbuf *p = NULL, *q, *nq; 977 struct mbuf *n; 978 struct ipq *fp = NULL; 979 struct ipqhead *head; 980 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 981 int i, next; 982 u_short sum; 983 984 /* If maxnipq is 0, never accept fragments. */ 985 if (maxnipq == 0) { 986 ipstat.ips_fragments++; 987 ipstat.ips_fragdropped++; 988 m_freem(m); 989 return NULL; 990 } 991 992 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 993 /* 994 * Look for queue of fragments of this datagram. 995 */ 996 lwkt_gettoken(&ipq_token); 997 head = &ipq[sum]; 998 TAILQ_FOREACH(fp, head, ipq_list) { 999 if (ip->ip_id == fp->ipq_id && 1000 ip->ip_src.s_addr == fp->ipq_src.s_addr && 1001 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 1002 ip->ip_p == fp->ipq_p) 1003 goto found; 1004 } 1005 1006 fp = NULL; 1007 1008 /* 1009 * Enforce upper bound on number of fragmented packets 1010 * for which we attempt reassembly; 1011 * If maxnipq is -1, accept all fragments without limitation. 1012 */ 1013 if (nipq > maxnipq && maxnipq > 0) { 1014 /* 1015 * drop something from the tail of the current queue 1016 * before proceeding further 1017 */ 1018 struct ipq *q = TAILQ_LAST(head, ipqhead); 1019 if (q == NULL) { 1020 /* 1021 * The current queue is empty, 1022 * so drop from one of the others. 1023 */ 1024 for (i = 0; i < IPREASS_NHASH; i++) { 1025 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 1026 if (r) { 1027 ipstat.ips_fragtimeout += r->ipq_nfrags; 1028 ip_freef(&ipq[i], r); 1029 break; 1030 } 1031 } 1032 } else { 1033 ipstat.ips_fragtimeout += q->ipq_nfrags; 1034 ip_freef(head, q); 1035 } 1036 } 1037 found: 1038 /* 1039 * Adjust ip_len to not reflect header, 1040 * convert offset of this to bytes. 1041 */ 1042 ip->ip_len -= hlen; 1043 if (ip->ip_off & IP_MF) { 1044 /* 1045 * Make sure that fragments have a data length 1046 * that's a non-zero multiple of 8 bytes. 1047 */ 1048 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 1049 ipstat.ips_toosmall++; /* XXX */ 1050 m_freem(m); 1051 goto done; 1052 } 1053 m->m_flags |= M_FRAG; 1054 } else { 1055 m->m_flags &= ~M_FRAG; 1056 } 1057 ip->ip_off <<= 3; 1058 1059 ipstat.ips_fragments++; 1060 m->m_pkthdr.header = ip; 1061 1062 /* 1063 * If the hardware has not done csum over this fragment 1064 * then csum_data is not valid at all. 1065 */ 1066 if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) 1067 == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) { 1068 m->m_pkthdr.csum_data = 0; 1069 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 1070 } 1071 1072 /* 1073 * Presence of header sizes in mbufs 1074 * would confuse code below. 1075 */ 1076 m->m_data += hlen; 1077 m->m_len -= hlen; 1078 1079 /* 1080 * If first fragment to arrive, create a reassembly queue. 1081 */ 1082 if (fp == NULL) { 1083 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL) 1084 goto dropfrag; 1085 TAILQ_INSERT_HEAD(head, fp, ipq_list); 1086 nipq++; 1087 fp->ipq_nfrags = 1; 1088 fp->ipq_ttl = IPFRAGTTL; 1089 fp->ipq_p = ip->ip_p; 1090 fp->ipq_id = ip->ip_id; 1091 fp->ipq_src = ip->ip_src; 1092 fp->ipq_dst = ip->ip_dst; 1093 fp->ipq_frags = m; 1094 m->m_nextpkt = NULL; 1095 goto inserted; 1096 } else { 1097 fp->ipq_nfrags++; 1098 } 1099 1100 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1101 1102 /* 1103 * Find a segment which begins after this one does. 1104 */ 1105 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1106 if (GETIP(q)->ip_off > ip->ip_off) 1107 break; 1108 } 1109 1110 /* 1111 * If there is a preceding segment, it may provide some of 1112 * our data already. If so, drop the data from the incoming 1113 * segment. If it provides all of our data, drop us, otherwise 1114 * stick new segment in the proper place. 1115 * 1116 * If some of the data is dropped from the the preceding 1117 * segment, then it's checksum is invalidated. 1118 */ 1119 if (p) { 1120 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1121 if (i > 0) { 1122 if (i >= ip->ip_len) 1123 goto dropfrag; 1124 m_adj(m, i); 1125 m->m_pkthdr.csum_flags = 0; 1126 ip->ip_off += i; 1127 ip->ip_len -= i; 1128 } 1129 m->m_nextpkt = p->m_nextpkt; 1130 p->m_nextpkt = m; 1131 } else { 1132 m->m_nextpkt = fp->ipq_frags; 1133 fp->ipq_frags = m; 1134 } 1135 1136 /* 1137 * While we overlap succeeding segments trim them or, 1138 * if they are completely covered, dequeue them. 1139 */ 1140 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1141 q = nq) { 1142 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1143 if (i < GETIP(q)->ip_len) { 1144 GETIP(q)->ip_len -= i; 1145 GETIP(q)->ip_off += i; 1146 m_adj(q, i); 1147 q->m_pkthdr.csum_flags = 0; 1148 break; 1149 } 1150 nq = q->m_nextpkt; 1151 m->m_nextpkt = nq; 1152 ipstat.ips_fragdropped++; 1153 fp->ipq_nfrags--; 1154 q->m_nextpkt = NULL; 1155 m_freem(q); 1156 } 1157 1158 inserted: 1159 /* 1160 * Check for complete reassembly and perform frag per packet 1161 * limiting. 1162 * 1163 * Frag limiting is performed here so that the nth frag has 1164 * a chance to complete the packet before we drop the packet. 1165 * As a result, n+1 frags are actually allowed per packet, but 1166 * only n will ever be stored. (n = maxfragsperpacket.) 1167 * 1168 */ 1169 next = 0; 1170 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1171 if (GETIP(q)->ip_off != next) { 1172 if (fp->ipq_nfrags > maxfragsperpacket) { 1173 ipstat.ips_fragdropped += fp->ipq_nfrags; 1174 ip_freef(head, fp); 1175 } 1176 goto done; 1177 } 1178 next += GETIP(q)->ip_len; 1179 } 1180 /* Make sure the last packet didn't have the IP_MF flag */ 1181 if (p->m_flags & M_FRAG) { 1182 if (fp->ipq_nfrags > maxfragsperpacket) { 1183 ipstat.ips_fragdropped += fp->ipq_nfrags; 1184 ip_freef(head, fp); 1185 } 1186 goto done; 1187 } 1188 1189 /* 1190 * Reassembly is complete. Make sure the packet is a sane size. 1191 */ 1192 q = fp->ipq_frags; 1193 ip = GETIP(q); 1194 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 1195 ipstat.ips_toolong++; 1196 ipstat.ips_fragdropped += fp->ipq_nfrags; 1197 ip_freef(head, fp); 1198 goto done; 1199 } 1200 1201 /* 1202 * Concatenate fragments. 1203 */ 1204 m = q; 1205 n = m->m_next; 1206 m->m_next = NULL; 1207 m_cat(m, n); 1208 nq = q->m_nextpkt; 1209 q->m_nextpkt = NULL; 1210 for (q = nq; q != NULL; q = nq) { 1211 nq = q->m_nextpkt; 1212 q->m_nextpkt = NULL; 1213 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1214 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1215 m_cat(m, q); 1216 } 1217 1218 /* 1219 * Clean up the 1's complement checksum. Carry over 16 bits must 1220 * be added back. This assumes no more then 65535 packet fragments 1221 * were reassembled. A second carry can also occur (but not a third). 1222 */ 1223 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) + 1224 (m->m_pkthdr.csum_data >> 16); 1225 if (m->m_pkthdr.csum_data > 0xFFFF) 1226 m->m_pkthdr.csum_data -= 0xFFFF; 1227 1228 /* 1229 * Create header for new ip packet by 1230 * modifying header of first packet; 1231 * dequeue and discard fragment reassembly header. 1232 * Make header visible. 1233 */ 1234 ip->ip_len = next; 1235 ip->ip_src = fp->ipq_src; 1236 ip->ip_dst = fp->ipq_dst; 1237 TAILQ_REMOVE(head, fp, ipq_list); 1238 nipq--; 1239 mpipe_free(&ipq_mpipe, fp); 1240 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 1241 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 1242 /* some debugging cruft by sklower, below, will go away soon */ 1243 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1244 int plen = 0; 1245 1246 for (n = m; n; n = n->m_next) 1247 plen += n->m_len; 1248 m->m_pkthdr.len = plen; 1249 } 1250 1251 /* 1252 * Reassembly complete, return the next protocol. 1253 * 1254 * Be sure to clear M_HASH to force the packet 1255 * to be re-characterized. 1256 * 1257 * Clear M_FRAG, we are no longer a fragment. 1258 */ 1259 m->m_flags &= ~(M_HASH | M_FRAG); 1260 1261 ipstat.ips_reassembled++; 1262 lwkt_reltoken(&ipq_token); 1263 return (m); 1264 1265 dropfrag: 1266 ipstat.ips_fragdropped++; 1267 if (fp != NULL) 1268 fp->ipq_nfrags--; 1269 m_freem(m); 1270 done: 1271 lwkt_reltoken(&ipq_token); 1272 return (NULL); 1273 1274 #undef GETIP 1275 } 1276 1277 /* 1278 * Free a fragment reassembly header and all 1279 * associated datagrams. 1280 * 1281 * Called with ipq_token held. 1282 */ 1283 static void 1284 ip_freef(struct ipqhead *fhp, struct ipq *fp) 1285 { 1286 struct mbuf *q; 1287 1288 /* 1289 * Remove first to protect against blocking 1290 */ 1291 TAILQ_REMOVE(fhp, fp, ipq_list); 1292 1293 /* 1294 * Clean out at our leisure 1295 */ 1296 while (fp->ipq_frags) { 1297 q = fp->ipq_frags; 1298 fp->ipq_frags = q->m_nextpkt; 1299 q->m_nextpkt = NULL; 1300 m_freem(q); 1301 } 1302 mpipe_free(&ipq_mpipe, fp); 1303 nipq--; 1304 } 1305 1306 /* 1307 * IP timer processing; 1308 * if a timer expires on a reassembly 1309 * queue, discard it. 1310 */ 1311 void 1312 ip_slowtimo(void) 1313 { 1314 struct ipq *fp, *fp_temp; 1315 struct ipqhead *head; 1316 int i; 1317 1318 lwkt_gettoken(&ipq_token); 1319 for (i = 0; i < IPREASS_NHASH; i++) { 1320 head = &ipq[i]; 1321 TAILQ_FOREACH_MUTABLE(fp, head, ipq_list, fp_temp) { 1322 if (--fp->ipq_ttl == 0) { 1323 ipstat.ips_fragtimeout += fp->ipq_nfrags; 1324 ip_freef(head, fp); 1325 } 1326 } 1327 } 1328 /* 1329 * If we are over the maximum number of fragments 1330 * (due to the limit being lowered), drain off 1331 * enough to get down to the new limit. 1332 */ 1333 if (maxnipq >= 0 && nipq > maxnipq) { 1334 for (i = 0; i < IPREASS_NHASH; i++) { 1335 head = &ipq[i]; 1336 while (nipq > maxnipq && !TAILQ_EMPTY(head)) { 1337 ipstat.ips_fragdropped += 1338 TAILQ_FIRST(head)->ipq_nfrags; 1339 ip_freef(head, TAILQ_FIRST(head)); 1340 } 1341 } 1342 } 1343 lwkt_reltoken(&ipq_token); 1344 ipflow_slowtimo(); 1345 } 1346 1347 /* 1348 * Drain off all datagram fragments. 1349 */ 1350 void 1351 ip_drain(void) 1352 { 1353 struct ipqhead *head; 1354 int i; 1355 1356 lwkt_gettoken(&ipq_token); 1357 for (i = 0; i < IPREASS_NHASH; i++) { 1358 head = &ipq[i]; 1359 while (!TAILQ_EMPTY(head)) { 1360 ipstat.ips_fragdropped += TAILQ_FIRST(head)->ipq_nfrags; 1361 ip_freef(head, TAILQ_FIRST(head)); 1362 } 1363 } 1364 lwkt_reltoken(&ipq_token); 1365 in_rtqdrain(); 1366 } 1367 1368 /* 1369 * Do option processing on a datagram, 1370 * possibly discarding it if bad options are encountered, 1371 * or forwarding it if source-routed. 1372 * The pass argument is used when operating in the IPSTEALTH 1373 * mode to tell what options to process: 1374 * [LS]SRR (pass 0) or the others (pass 1). 1375 * The reason for as many as two passes is that when doing IPSTEALTH, 1376 * non-routing options should be processed only if the packet is for us. 1377 * Returns 1 if packet has been forwarded/freed, 1378 * 0 if the packet should be processed further. 1379 */ 1380 static int 1381 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop) 1382 { 1383 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET }; 1384 struct ip *ip = mtod(m, struct ip *); 1385 u_char *cp; 1386 struct in_ifaddr *ia; 1387 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB; 1388 boolean_t forward = FALSE; 1389 struct in_addr *sin, dst; 1390 n_time ntime; 1391 1392 dst = ip->ip_dst; 1393 cp = (u_char *)(ip + 1); 1394 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1395 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1396 opt = cp[IPOPT_OPTVAL]; 1397 if (opt == IPOPT_EOL) 1398 break; 1399 if (opt == IPOPT_NOP) 1400 optlen = 1; 1401 else { 1402 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1403 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1404 goto bad; 1405 } 1406 optlen = cp[IPOPT_OLEN]; 1407 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1408 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1409 goto bad; 1410 } 1411 } 1412 switch (opt) { 1413 1414 default: 1415 break; 1416 1417 /* 1418 * Source routing with record. 1419 * Find interface with current destination address. 1420 * If none on this machine then drop if strictly routed, 1421 * or do nothing if loosely routed. 1422 * Record interface address and bring up next address 1423 * component. If strictly routed make sure next 1424 * address is on directly accessible net. 1425 */ 1426 case IPOPT_LSRR: 1427 case IPOPT_SSRR: 1428 if (ipstealth && pass > 0) 1429 break; 1430 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1431 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1432 goto bad; 1433 } 1434 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1435 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1436 goto bad; 1437 } 1438 ipaddr.sin_addr = ip->ip_dst; 1439 ia = (struct in_ifaddr *) 1440 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1441 if (ia == NULL) { 1442 if (opt == IPOPT_SSRR) { 1443 type = ICMP_UNREACH; 1444 code = ICMP_UNREACH_SRCFAIL; 1445 goto bad; 1446 } 1447 if (!ip_dosourceroute) 1448 goto nosourcerouting; 1449 /* 1450 * Loose routing, and not at next destination 1451 * yet; nothing to do except forward. 1452 */ 1453 break; 1454 } 1455 off--; /* 0 origin */ 1456 if (off > optlen - (int)sizeof(struct in_addr)) { 1457 /* 1458 * End of source route. Should be for us. 1459 */ 1460 if (!ip_acceptsourceroute) 1461 goto nosourcerouting; 1462 save_rte(m, cp, ip->ip_src); 1463 break; 1464 } 1465 if (ipstealth) 1466 goto dropit; 1467 if (!ip_dosourceroute) { 1468 if (ipforwarding) { 1469 char buf[sizeof "aaa.bbb.ccc.ddd"]; 1470 1471 /* 1472 * Acting as a router, so generate ICMP 1473 */ 1474 nosourcerouting: 1475 strcpy(buf, inet_ntoa(ip->ip_dst)); 1476 log(LOG_WARNING, 1477 "attempted source route from %s to %s\n", 1478 inet_ntoa(ip->ip_src), buf); 1479 type = ICMP_UNREACH; 1480 code = ICMP_UNREACH_SRCFAIL; 1481 goto bad; 1482 } else { 1483 /* 1484 * Not acting as a router, 1485 * so silently drop. 1486 */ 1487 dropit: 1488 ipstat.ips_cantforward++; 1489 m_freem(m); 1490 return (1); 1491 } 1492 } 1493 1494 /* 1495 * locate outgoing interface 1496 */ 1497 memcpy(&ipaddr.sin_addr, cp + off, 1498 sizeof ipaddr.sin_addr); 1499 1500 if (opt == IPOPT_SSRR) { 1501 #define INA struct in_ifaddr * 1502 #define SA struct sockaddr * 1503 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) 1504 == NULL) 1505 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1506 } else { 1507 ia = ip_rtaddr(ipaddr.sin_addr, NULL); 1508 } 1509 if (ia == NULL) { 1510 type = ICMP_UNREACH; 1511 code = ICMP_UNREACH_SRCFAIL; 1512 goto bad; 1513 } 1514 ip->ip_dst = ipaddr.sin_addr; 1515 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1516 sizeof(struct in_addr)); 1517 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1518 /* 1519 * Let ip_intr's mcast routing check handle mcast pkts 1520 */ 1521 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1522 break; 1523 1524 case IPOPT_RR: 1525 if (ipstealth && pass == 0) 1526 break; 1527 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1528 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1529 goto bad; 1530 } 1531 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1532 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1533 goto bad; 1534 } 1535 /* 1536 * If no space remains, ignore. 1537 */ 1538 off--; /* 0 origin */ 1539 if (off > optlen - (int)sizeof(struct in_addr)) 1540 break; 1541 memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1542 sizeof ipaddr.sin_addr); 1543 /* 1544 * locate outgoing interface; if we're the destination, 1545 * use the incoming interface (should be same). 1546 */ 1547 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL && 1548 (ia = ip_rtaddr(ipaddr.sin_addr, NULL)) == NULL) { 1549 type = ICMP_UNREACH; 1550 code = ICMP_UNREACH_HOST; 1551 goto bad; 1552 } 1553 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1554 sizeof(struct in_addr)); 1555 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1556 break; 1557 1558 case IPOPT_TS: 1559 if (ipstealth && pass == 0) 1560 break; 1561 code = cp - (u_char *)ip; 1562 if (optlen < 4 || optlen > 40) { 1563 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1564 goto bad; 1565 } 1566 if ((off = cp[IPOPT_OFFSET]) < 5) { 1567 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1568 goto bad; 1569 } 1570 if (off > optlen - (int)sizeof(int32_t)) { 1571 cp[IPOPT_OFFSET + 1] += (1 << 4); 1572 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1573 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1574 goto bad; 1575 } 1576 break; 1577 } 1578 off--; /* 0 origin */ 1579 sin = (struct in_addr *)(cp + off); 1580 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1581 1582 case IPOPT_TS_TSONLY: 1583 break; 1584 1585 case IPOPT_TS_TSANDADDR: 1586 if (off + sizeof(n_time) + 1587 sizeof(struct in_addr) > optlen) { 1588 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1589 goto bad; 1590 } 1591 ipaddr.sin_addr = dst; 1592 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1593 m->m_pkthdr.rcvif); 1594 if (ia == NULL) 1595 continue; 1596 memcpy(sin, &IA_SIN(ia)->sin_addr, 1597 sizeof(struct in_addr)); 1598 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1599 off += sizeof(struct in_addr); 1600 break; 1601 1602 case IPOPT_TS_PRESPEC: 1603 if (off + sizeof(n_time) + 1604 sizeof(struct in_addr) > optlen) { 1605 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1606 goto bad; 1607 } 1608 memcpy(&ipaddr.sin_addr, sin, 1609 sizeof(struct in_addr)); 1610 if (ifa_ifwithaddr((SA)&ipaddr) == NULL) 1611 continue; 1612 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1613 off += sizeof(struct in_addr); 1614 break; 1615 1616 default: 1617 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1618 goto bad; 1619 } 1620 ntime = iptime(); 1621 memcpy(cp + off, &ntime, sizeof(n_time)); 1622 cp[IPOPT_OFFSET] += sizeof(n_time); 1623 } 1624 } 1625 if (forward && ipforwarding) { 1626 ip_forward(m, TRUE, next_hop); 1627 return (1); 1628 } 1629 return (0); 1630 bad: 1631 icmp_error(m, type, code, 0, 0); 1632 ipstat.ips_badoptions++; 1633 return (1); 1634 } 1635 1636 /* 1637 * Given address of next destination (final or next hop), 1638 * return internet address info of interface to be used to get there. 1639 */ 1640 struct in_ifaddr * 1641 ip_rtaddr(struct in_addr dst, struct route *ro0) 1642 { 1643 struct route sro, *ro; 1644 struct sockaddr_in *sin; 1645 struct in_ifaddr *ia; 1646 1647 if (ro0 != NULL) { 1648 ro = ro0; 1649 } else { 1650 bzero(&sro, sizeof(sro)); 1651 ro = &sro; 1652 } 1653 1654 sin = (struct sockaddr_in *)&ro->ro_dst; 1655 1656 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) { 1657 if (ro->ro_rt != NULL) { 1658 RTFREE(ro->ro_rt); 1659 ro->ro_rt = NULL; 1660 } 1661 sin->sin_family = AF_INET; 1662 sin->sin_len = sizeof *sin; 1663 sin->sin_addr = dst; 1664 rtalloc_ign(ro, RTF_PRCLONING); 1665 } 1666 1667 if (ro->ro_rt == NULL) 1668 return (NULL); 1669 1670 ia = ifatoia(ro->ro_rt->rt_ifa); 1671 1672 if (ro == &sro) 1673 RTFREE(ro->ro_rt); 1674 return ia; 1675 } 1676 1677 /* 1678 * Save incoming source route for use in replies, 1679 * to be picked up later by ip_srcroute if the receiver is interested. 1680 */ 1681 static void 1682 save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1683 { 1684 struct m_tag *mtag; 1685 struct ip_srcrt_opt *opt; 1686 unsigned olen; 1687 1688 mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), MB_DONTWAIT); 1689 if (mtag == NULL) 1690 return; 1691 opt = m_tag_data(mtag); 1692 1693 olen = option[IPOPT_OLEN]; 1694 #ifdef DIAGNOSTIC 1695 if (ipprintfs) 1696 kprintf("save_rte: olen %d\n", olen); 1697 #endif 1698 if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) { 1699 m_tag_free(mtag); 1700 return; 1701 } 1702 bcopy(option, opt->ip_srcrt.srcopt, olen); 1703 opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1704 opt->ip_srcrt.dst = dst; 1705 m_tag_prepend(m, mtag); 1706 } 1707 1708 /* 1709 * Retrieve incoming source route for use in replies, 1710 * in the same form used by setsockopt. 1711 * The first hop is placed before the options, will be removed later. 1712 */ 1713 struct mbuf * 1714 ip_srcroute(struct mbuf *m0) 1715 { 1716 struct in_addr *p, *q; 1717 struct mbuf *m; 1718 struct m_tag *mtag; 1719 struct ip_srcrt_opt *opt; 1720 1721 if (m0 == NULL) 1722 return NULL; 1723 1724 mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL); 1725 if (mtag == NULL) 1726 return NULL; 1727 opt = m_tag_data(mtag); 1728 1729 if (opt->ip_nhops == 0) 1730 return (NULL); 1731 m = m_get(MB_DONTWAIT, MT_HEADER); 1732 if (m == NULL) 1733 return (NULL); 1734 1735 #define OPTSIZ (sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt)) 1736 1737 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1738 m->m_len = opt->ip_nhops * sizeof(struct in_addr) + 1739 sizeof(struct in_addr) + OPTSIZ; 1740 #ifdef DIAGNOSTIC 1741 if (ipprintfs) { 1742 kprintf("ip_srcroute: nhops %d mlen %d", 1743 opt->ip_nhops, m->m_len); 1744 } 1745 #endif 1746 1747 /* 1748 * First save first hop for return route 1749 */ 1750 p = &opt->ip_srcrt.route[opt->ip_nhops - 1]; 1751 *(mtod(m, struct in_addr *)) = *p--; 1752 #ifdef DIAGNOSTIC 1753 if (ipprintfs) 1754 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1755 #endif 1756 1757 /* 1758 * Copy option fields and padding (nop) to mbuf. 1759 */ 1760 opt->ip_srcrt.nop = IPOPT_NOP; 1761 opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1762 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop, 1763 OPTSIZ); 1764 q = (struct in_addr *)(mtod(m, caddr_t) + 1765 sizeof(struct in_addr) + OPTSIZ); 1766 #undef OPTSIZ 1767 /* 1768 * Record return path as an IP source route, 1769 * reversing the path (pointers are now aligned). 1770 */ 1771 while (p >= opt->ip_srcrt.route) { 1772 #ifdef DIAGNOSTIC 1773 if (ipprintfs) 1774 kprintf(" %x", ntohl(q->s_addr)); 1775 #endif 1776 *q++ = *p--; 1777 } 1778 /* 1779 * Last hop goes to final destination. 1780 */ 1781 *q = opt->ip_srcrt.dst; 1782 m_tag_delete(m0, mtag); 1783 #ifdef DIAGNOSTIC 1784 if (ipprintfs) 1785 kprintf(" %x\n", ntohl(q->s_addr)); 1786 #endif 1787 return (m); 1788 } 1789 1790 /* 1791 * Strip out IP options. 1792 */ 1793 void 1794 ip_stripoptions(struct mbuf *m) 1795 { 1796 int datalen; 1797 struct ip *ip = mtod(m, struct ip *); 1798 caddr_t opts; 1799 int optlen; 1800 1801 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1802 opts = (caddr_t)(ip + 1); 1803 datalen = m->m_len - (sizeof(struct ip) + optlen); 1804 bcopy(opts + optlen, opts, datalen); 1805 m->m_len -= optlen; 1806 if (m->m_flags & M_PKTHDR) 1807 m->m_pkthdr.len -= optlen; 1808 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1809 } 1810 1811 u_char inetctlerrmap[PRC_NCMDS] = { 1812 0, 0, 0, 0, 1813 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1814 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1815 EMSGSIZE, EHOSTUNREACH, 0, 0, 1816 0, 0, 0, 0, 1817 ENOPROTOOPT, ECONNREFUSED 1818 }; 1819 1820 /* 1821 * Forward a packet. If some error occurs return the sender 1822 * an icmp packet. Note we can't always generate a meaningful 1823 * icmp message because icmp doesn't have a large enough repertoire 1824 * of codes and types. 1825 * 1826 * If not forwarding, just drop the packet. This could be confusing 1827 * if ipforwarding was zero but some routing protocol was advancing 1828 * us as a gateway to somewhere. However, we must let the routing 1829 * protocol deal with that. 1830 * 1831 * The using_srcrt parameter indicates whether the packet is being forwarded 1832 * via a source route. 1833 */ 1834 void 1835 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop) 1836 { 1837 struct ip *ip = mtod(m, struct ip *); 1838 struct rtentry *rt; 1839 struct route fwd_ro; 1840 int error, type = 0, code = 0, destmtu = 0; 1841 struct mbuf *mcopy; 1842 n_long dest; 1843 struct in_addr pkt_dst; 1844 1845 dest = INADDR_ANY; 1846 /* 1847 * Cache the destination address of the packet; this may be 1848 * changed by use of 'ipfw fwd'. 1849 */ 1850 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst; 1851 1852 #ifdef DIAGNOSTIC 1853 if (ipprintfs) 1854 kprintf("forward: src %x dst %x ttl %x\n", 1855 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl); 1856 #endif 1857 1858 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) { 1859 ipstat.ips_cantforward++; 1860 m_freem(m); 1861 return; 1862 } 1863 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) { 1864 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1865 return; 1866 } 1867 1868 bzero(&fwd_ro, sizeof(fwd_ro)); 1869 ip_rtaddr(pkt_dst, &fwd_ro); 1870 if (fwd_ro.ro_rt == NULL) { 1871 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1872 return; 1873 } 1874 rt = fwd_ro.ro_rt; 1875 1876 /* 1877 * Save the IP header and at most 8 bytes of the payload, 1878 * in case we need to generate an ICMP message to the src. 1879 * 1880 * XXX this can be optimized a lot by saving the data in a local 1881 * buffer on the stack (72 bytes at most), and only allocating the 1882 * mbuf if really necessary. The vast majority of the packets 1883 * are forwarded without having to send an ICMP back (either 1884 * because unnecessary, or because rate limited), so we are 1885 * really we are wasting a lot of work here. 1886 * 1887 * We don't use m_copy() because it might return a reference 1888 * to a shared cluster. Both this function and ip_output() 1889 * assume exclusive access to the IP header in `m', so any 1890 * data in a cluster may change before we reach icmp_error(). 1891 */ 1892 MGETHDR(mcopy, MB_DONTWAIT, m->m_type); 1893 if (mcopy != NULL && !m_dup_pkthdr(mcopy, 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 m_free(mcopy); 1901 mcopy = NULL; 1902 } 1903 if (mcopy != NULL) { 1904 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8, 1905 (int)ip->ip_len); 1906 mcopy->m_pkthdr.len = mcopy->m_len; 1907 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 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 (mcopy) { 1952 ipflow_create(&fwd_ro, mcopy); 1953 m_freem(mcopy); 1954 } 1955 goto done; 1956 } else { 1957 ipstat.ips_redirectsent++; 1958 } 1959 } else { 1960 ipstat.ips_cantforward++; 1961 } 1962 1963 if (mcopy == NULL) 1964 goto done; 1965 1966 /* 1967 * Send ICMP message. 1968 */ 1969 1970 switch (error) { 1971 1972 case 0: /* forwarded, but need redirect */ 1973 /* type, code set above */ 1974 break; 1975 1976 case ENETUNREACH: /* shouldn't happen, checked above */ 1977 case EHOSTUNREACH: 1978 case ENETDOWN: 1979 case EHOSTDOWN: 1980 default: 1981 type = ICMP_UNREACH; 1982 code = ICMP_UNREACH_HOST; 1983 break; 1984 1985 case EMSGSIZE: 1986 type = ICMP_UNREACH; 1987 code = ICMP_UNREACH_NEEDFRAG; 1988 #ifdef IPSEC 1989 /* 1990 * If the packet is routed over IPsec tunnel, tell the 1991 * originator the tunnel MTU. 1992 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1993 * XXX quickhack!!! 1994 */ 1995 if (fwd_ro.ro_rt != NULL) { 1996 struct secpolicy *sp = NULL; 1997 int ipsecerror; 1998 int ipsechdr; 1999 struct route *ro; 2000 2001 sp = ipsec4_getpolicybyaddr(mcopy, 2002 IPSEC_DIR_OUTBOUND, 2003 IP_FORWARDING, 2004 &ipsecerror); 2005 2006 if (sp == NULL) 2007 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2008 else { 2009 /* count IPsec header size */ 2010 ipsechdr = ipsec4_hdrsiz(mcopy, 2011 IPSEC_DIR_OUTBOUND, 2012 NULL); 2013 2014 /* 2015 * find the correct route for outer IPv4 2016 * header, compute tunnel MTU. 2017 * 2018 */ 2019 if (sp->req != NULL && sp->req->sav != NULL && 2020 sp->req->sav->sah != NULL) { 2021 ro = &sp->req->sav->sah->sa_route; 2022 if (ro->ro_rt != NULL && 2023 ro->ro_rt->rt_ifp != NULL) { 2024 destmtu = 2025 ro->ro_rt->rt_ifp->if_mtu; 2026 destmtu -= ipsechdr; 2027 } 2028 } 2029 2030 key_freesp(sp); 2031 } 2032 } 2033 #elif FAST_IPSEC 2034 /* 2035 * If the packet is routed over IPsec tunnel, tell the 2036 * originator the tunnel MTU. 2037 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 2038 * XXX quickhack!!! 2039 */ 2040 if (fwd_ro.ro_rt != NULL) { 2041 struct secpolicy *sp = NULL; 2042 int ipsecerror; 2043 int ipsechdr; 2044 struct route *ro; 2045 2046 sp = ipsec_getpolicybyaddr(mcopy, 2047 IPSEC_DIR_OUTBOUND, 2048 IP_FORWARDING, 2049 &ipsecerror); 2050 2051 if (sp == NULL) 2052 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2053 else { 2054 /* count IPsec header size */ 2055 ipsechdr = ipsec4_hdrsiz(mcopy, 2056 IPSEC_DIR_OUTBOUND, 2057 NULL); 2058 2059 /* 2060 * find the correct route for outer IPv4 2061 * header, compute tunnel MTU. 2062 */ 2063 2064 if (sp->req != NULL && 2065 sp->req->sav != NULL && 2066 sp->req->sav->sah != NULL) { 2067 ro = &sp->req->sav->sah->sa_route; 2068 if (ro->ro_rt != NULL && 2069 ro->ro_rt->rt_ifp != NULL) { 2070 destmtu = 2071 ro->ro_rt->rt_ifp->if_mtu; 2072 destmtu -= ipsechdr; 2073 } 2074 } 2075 2076 KEY_FREESP(&sp); 2077 } 2078 } 2079 #else /* !IPSEC && !FAST_IPSEC */ 2080 if (fwd_ro.ro_rt != NULL) 2081 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2082 #endif /*IPSEC*/ 2083 ipstat.ips_cantfrag++; 2084 break; 2085 2086 case ENOBUFS: 2087 /* 2088 * A router should not generate ICMP_SOURCEQUENCH as 2089 * required in RFC1812 Requirements for IP Version 4 Routers. 2090 * Source quench could be a big problem under DoS attacks, 2091 * or if the underlying interface is rate-limited. 2092 * Those who need source quench packets may re-enable them 2093 * via the net.inet.ip.sendsourcequench sysctl. 2094 */ 2095 if (!ip_sendsourcequench) { 2096 m_freem(mcopy); 2097 goto done; 2098 } else { 2099 type = ICMP_SOURCEQUENCH; 2100 code = 0; 2101 } 2102 break; 2103 2104 case EACCES: /* ipfw denied packet */ 2105 m_freem(mcopy); 2106 goto done; 2107 } 2108 icmp_error(mcopy, type, code, dest, destmtu); 2109 done: 2110 if (fwd_ro.ro_rt != NULL) 2111 RTFREE(fwd_ro.ro_rt); 2112 } 2113 2114 void 2115 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2116 struct mbuf *m) 2117 { 2118 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2119 struct timeval tv; 2120 2121 microtime(&tv); 2122 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2123 SCM_TIMESTAMP, SOL_SOCKET); 2124 if (*mp) 2125 mp = &(*mp)->m_next; 2126 } 2127 if (inp->inp_flags & INP_RECVDSTADDR) { 2128 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2129 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2130 if (*mp) 2131 mp = &(*mp)->m_next; 2132 } 2133 if (inp->inp_flags & INP_RECVTTL) { 2134 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 2135 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 2136 if (*mp) 2137 mp = &(*mp)->m_next; 2138 } 2139 #ifdef notyet 2140 /* XXX 2141 * Moving these out of udp_input() made them even more broken 2142 * than they already were. 2143 */ 2144 /* options were tossed already */ 2145 if (inp->inp_flags & INP_RECVOPTS) { 2146 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2147 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2148 if (*mp) 2149 mp = &(*mp)->m_next; 2150 } 2151 /* ip_srcroute doesn't do what we want here, need to fix */ 2152 if (inp->inp_flags & INP_RECVRETOPTS) { 2153 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 2154 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2155 if (*mp) 2156 mp = &(*mp)->m_next; 2157 } 2158 #endif 2159 if (inp->inp_flags & INP_RECVIF) { 2160 struct ifnet *ifp; 2161 struct sdlbuf { 2162 struct sockaddr_dl sdl; 2163 u_char pad[32]; 2164 } sdlbuf; 2165 struct sockaddr_dl *sdp; 2166 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2167 2168 if (((ifp = m->m_pkthdr.rcvif)) && 2169 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) { 2170 sdp = IF_LLSOCKADDR(ifp); 2171 /* 2172 * Change our mind and don't try copy. 2173 */ 2174 if ((sdp->sdl_family != AF_LINK) || 2175 (sdp->sdl_len > sizeof(sdlbuf))) { 2176 goto makedummy; 2177 } 2178 bcopy(sdp, sdl2, sdp->sdl_len); 2179 } else { 2180 makedummy: 2181 sdl2->sdl_len = 2182 offsetof(struct sockaddr_dl, sdl_data[0]); 2183 sdl2->sdl_family = AF_LINK; 2184 sdl2->sdl_index = 0; 2185 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2186 } 2187 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 2188 IP_RECVIF, IPPROTO_IP); 2189 if (*mp) 2190 mp = &(*mp)->m_next; 2191 } 2192 } 2193 2194 /* 2195 * XXX these routines are called from the upper part of the kernel. 2196 * 2197 * They could also be moved to ip_mroute.c, since all the RSVP 2198 * handling is done there already. 2199 */ 2200 int 2201 ip_rsvp_init(struct socket *so) 2202 { 2203 if (so->so_type != SOCK_RAW || 2204 so->so_proto->pr_protocol != IPPROTO_RSVP) 2205 return EOPNOTSUPP; 2206 2207 if (ip_rsvpd != NULL) 2208 return EADDRINUSE; 2209 2210 ip_rsvpd = so; 2211 /* 2212 * This may seem silly, but we need to be sure we don't over-increment 2213 * the RSVP counter, in case something slips up. 2214 */ 2215 if (!ip_rsvp_on) { 2216 ip_rsvp_on = 1; 2217 rsvp_on++; 2218 } 2219 2220 return 0; 2221 } 2222 2223 int 2224 ip_rsvp_done(void) 2225 { 2226 ip_rsvpd = NULL; 2227 /* 2228 * This may seem silly, but we need to be sure we don't over-decrement 2229 * the RSVP counter, in case something slips up. 2230 */ 2231 if (ip_rsvp_on) { 2232 ip_rsvp_on = 0; 2233 rsvp_on--; 2234 } 2235 return 0; 2236 } 2237 2238 int 2239 rsvp_input(struct mbuf **mp, int *offp, int proto) 2240 { 2241 struct mbuf *m = *mp; 2242 int off; 2243 2244 off = *offp; 2245 *mp = NULL; 2246 2247 if (rsvp_input_p) { /* call the real one if loaded */ 2248 *mp = m; 2249 rsvp_input_p(mp, offp, proto); 2250 return(IPPROTO_DONE); 2251 } 2252 2253 /* Can still get packets with rsvp_on = 0 if there is a local member 2254 * of the group to which the RSVP packet is addressed. But in this 2255 * case we want to throw the packet away. 2256 */ 2257 2258 if (!rsvp_on) { 2259 m_freem(m); 2260 return(IPPROTO_DONE); 2261 } 2262 2263 if (ip_rsvpd != NULL) { 2264 *mp = m; 2265 rip_input(mp, offp, proto); 2266 return(IPPROTO_DONE); 2267 } 2268 /* Drop the packet */ 2269 m_freem(m); 2270 return(IPPROTO_DONE); 2271 } 2272