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