1 /* 2 * Copyright (c) 2003, 2004 Jeffrey M. Hsu. All rights reserved. 3 * Copyright (c) 2003, 2004 The DragonFly Project. All rights reserved. 4 * 5 * This code is derived from software contributed to The DragonFly Project 6 * by Jeffrey M. Hsu. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of The DragonFly Project nor the names of its 17 * contributors may be used to endorse or promote products derived 18 * from this software without specific, prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 25 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING, 26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED 28 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 29 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 30 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 */ 33 34 /* 35 * Copyright (c) 1982, 1986, 1988, 1993 36 * The Regents of the University of California. All rights reserved. 37 * 38 * Redistribution and use in source and binary forms, with or without 39 * modification, are permitted provided that the following conditions 40 * are met: 41 * 1. Redistributions of source code must retain the above copyright 42 * notice, this list of conditions and the following disclaimer. 43 * 2. Redistributions in binary form must reproduce the above copyright 44 * notice, this list of conditions and the following disclaimer in the 45 * documentation and/or other materials provided with the distribution. 46 * 3. All advertising materials mentioning features or use of this software 47 * must display the following acknowledgement: 48 * This product includes software developed by the University of 49 * California, Berkeley and its contributors. 50 * 4. Neither the name of the University nor the names of its contributors 51 * may be used to endorse or promote products derived from this software 52 * without specific prior written permission. 53 * 54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 64 * SUCH DAMAGE. 65 * 66 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 67 * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $ 68 * $DragonFly: src/sys/netinet/ip_input.c,v 1.115 2008/10/28 07:09:26 sephe Exp $ 69 */ 70 71 #define _IP_VHL 72 73 #include "opt_bootp.h" 74 #include "opt_ipfw.h" 75 #include "opt_ipdn.h" 76 #include "opt_ipdivert.h" 77 #include "opt_ipfilter.h" 78 #include "opt_ipstealth.h" 79 #include "opt_ipsec.h" 80 81 #include <sys/param.h> 82 #include <sys/systm.h> 83 #include <sys/mbuf.h> 84 #include <sys/malloc.h> 85 #include <sys/mpipe.h> 86 #include <sys/domain.h> 87 #include <sys/protosw.h> 88 #include <sys/socket.h> 89 #include <sys/time.h> 90 #include <sys/globaldata.h> 91 #include <sys/thread.h> 92 #include <sys/kernel.h> 93 #include <sys/syslog.h> 94 #include <sys/sysctl.h> 95 #include <sys/in_cksum.h> 96 #include <sys/lock.h> 97 98 #include <sys/mplock2.h> 99 100 #include <machine/stdarg.h> 101 102 #include <net/if.h> 103 #include <net/if_types.h> 104 #include <net/if_var.h> 105 #include <net/if_dl.h> 106 #include <net/pfil.h> 107 #include <net/route.h> 108 #include <net/netisr.h> 109 110 #include <netinet/in.h> 111 #include <netinet/in_systm.h> 112 #include <netinet/in_var.h> 113 #include <netinet/ip.h> 114 #include <netinet/in_pcb.h> 115 #include <netinet/ip_var.h> 116 #include <netinet/ip_icmp.h> 117 #include <netinet/ip_divert.h> 118 #include <netinet/ip_flow.h> 119 120 #include <sys/thread2.h> 121 #include <sys/msgport2.h> 122 #include <net/netmsg2.h> 123 124 #include <sys/socketvar.h> 125 126 #include <net/ipfw/ip_fw.h> 127 #include <net/dummynet/ip_dummynet.h> 128 129 #ifdef IPSEC 130 #include <netinet6/ipsec.h> 131 #include <netproto/key/key.h> 132 #endif 133 134 #ifdef FAST_IPSEC 135 #include <netproto/ipsec/ipsec.h> 136 #include <netproto/ipsec/key.h> 137 #endif 138 139 int rsvp_on = 0; 140 static int ip_rsvp_on; 141 struct socket *ip_rsvpd; 142 143 int ipforwarding = 0; 144 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 145 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 146 147 static int ipsendredirects = 1; /* XXX */ 148 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 149 &ipsendredirects, 0, "Enable sending IP redirects"); 150 151 int ip_defttl = IPDEFTTL; 152 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 153 &ip_defttl, 0, "Maximum TTL on IP packets"); 154 155 static int ip_dosourceroute = 0; 156 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 157 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 158 159 static int ip_acceptsourceroute = 0; 160 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 161 CTLFLAG_RW, &ip_acceptsourceroute, 0, 162 "Enable accepting source routed IP packets"); 163 164 static int ip_keepfaith = 0; 165 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 166 &ip_keepfaith, 0, 167 "Enable packet capture for FAITH IPv4->IPv6 translator daemon"); 168 169 static int nipq = 0; /* total # of reass queues */ 170 static int maxnipq; 171 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 172 &maxnipq, 0, 173 "Maximum number of IPv4 fragment reassembly queue entries"); 174 175 static int maxfragsperpacket; 176 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 177 &maxfragsperpacket, 0, 178 "Maximum number of IPv4 fragments allowed per packet"); 179 180 static int ip_sendsourcequench = 0; 181 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 182 &ip_sendsourcequench, 0, 183 "Enable the transmission of source quench packets"); 184 185 int ip_do_randomid = 1; 186 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 187 &ip_do_randomid, 0, 188 "Assign random ip_id values"); 189 /* 190 * XXX - Setting ip_checkinterface mostly implements the receive side of 191 * the Strong ES model described in RFC 1122, but since the routing table 192 * and transmit implementation do not implement the Strong ES model, 193 * setting this to 1 results in an odd hybrid. 194 * 195 * XXX - ip_checkinterface currently must be disabled if you use ipnat 196 * to translate the destination address to another local interface. 197 * 198 * XXX - ip_checkinterface must be disabled if you add IP aliases 199 * to the loopback interface instead of the interface where the 200 * packets for those addresses are received. 201 */ 202 static int ip_checkinterface = 0; 203 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 204 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 205 206 static int ip_dispatch_fast = 0; 207 static int ip_dispatch_slow = 0; 208 static int ip_dispatch_recheck = 0; 209 static int ip_dispatch_software = 0; 210 SYSCTL_INT(_net_inet_ip, OID_AUTO, dispatch_fast_count, CTLFLAG_RW, 211 &ip_dispatch_fast, 0, ""); 212 SYSCTL_INT(_net_inet_ip, OID_AUTO, dispatch_slow_count, CTLFLAG_RW, 213 &ip_dispatch_slow, 0, ""); 214 SYSCTL_INT(_net_inet_ip, OID_AUTO, dispatch_software_count, CTLFLAG_RW, 215 &ip_dispatch_software, 0, ""); 216 SYSCTL_INT(_net_inet_ip, OID_AUTO, dispatch_recheck_count, CTLFLAG_RW, 217 &ip_dispatch_recheck, 0, ""); 218 219 static struct lwkt_token ipq_token = LWKT_TOKEN_MP_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 struct 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 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); 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 ipq[i].next = ipq[i].prev = &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 } 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 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 980 int i, next; 981 u_short sum; 982 983 /* If maxnipq is 0, never accept fragments. */ 984 if (maxnipq == 0) { 985 ipstat.ips_fragments++; 986 ipstat.ips_fragdropped++; 987 m_freem(m); 988 return NULL; 989 } 990 991 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 992 /* 993 * Look for queue of fragments of this datagram. 994 */ 995 lwkt_gettoken(&ipq_token); 996 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) { 997 if (ip->ip_id == fp->ipq_id && 998 ip->ip_src.s_addr == fp->ipq_src.s_addr && 999 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 1000 ip->ip_p == fp->ipq_p) 1001 goto found; 1002 } 1003 1004 fp = NULL; 1005 1006 /* 1007 * Enforce upper bound on number of fragmented packets 1008 * for which we attempt reassembly; 1009 * If maxnipq is -1, accept all fragments without limitation. 1010 */ 1011 if (nipq > maxnipq && maxnipq > 0) { 1012 /* 1013 * drop something from the tail of the current queue 1014 * before proceeding further 1015 */ 1016 if (ipq[sum].prev == &ipq[sum]) { /* gak */ 1017 for (i = 0; i < IPREASS_NHASH; i++) { 1018 if (ipq[i].prev != &ipq[i]) { 1019 ipstat.ips_fragtimeout += 1020 ipq[i].prev->ipq_nfrags; 1021 ip_freef(ipq[i].prev); 1022 break; 1023 } 1024 } 1025 } else { 1026 ipstat.ips_fragtimeout += 1027 ipq[sum].prev->ipq_nfrags; 1028 ip_freef(ipq[sum].prev); 1029 } 1030 } 1031 found: 1032 /* 1033 * Adjust ip_len to not reflect header, 1034 * convert offset of this to bytes. 1035 */ 1036 ip->ip_len -= hlen; 1037 if (ip->ip_off & IP_MF) { 1038 /* 1039 * Make sure that fragments have a data length 1040 * that's a non-zero multiple of 8 bytes. 1041 */ 1042 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 1043 ipstat.ips_toosmall++; /* XXX */ 1044 m_freem(m); 1045 goto done; 1046 } 1047 m->m_flags |= M_FRAG; 1048 } else { 1049 m->m_flags &= ~M_FRAG; 1050 } 1051 ip->ip_off <<= 3; 1052 1053 ipstat.ips_fragments++; 1054 m->m_pkthdr.header = ip; 1055 1056 /* 1057 * If the hardware has not done csum over this fragment 1058 * then csum_data is not valid at all. 1059 */ 1060 if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) 1061 == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) { 1062 m->m_pkthdr.csum_data = 0; 1063 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 1064 } 1065 1066 /* 1067 * Presence of header sizes in mbufs 1068 * would confuse code below. 1069 */ 1070 m->m_data += hlen; 1071 m->m_len -= hlen; 1072 1073 /* 1074 * If first fragment to arrive, create a reassembly queue. 1075 */ 1076 if (fp == NULL) { 1077 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL) 1078 goto dropfrag; 1079 insque(fp, &ipq[sum]); 1080 nipq++; 1081 fp->ipq_nfrags = 1; 1082 fp->ipq_ttl = IPFRAGTTL; 1083 fp->ipq_p = ip->ip_p; 1084 fp->ipq_id = ip->ip_id; 1085 fp->ipq_src = ip->ip_src; 1086 fp->ipq_dst = ip->ip_dst; 1087 fp->ipq_frags = m; 1088 m->m_nextpkt = NULL; 1089 goto inserted; 1090 } else { 1091 fp->ipq_nfrags++; 1092 } 1093 1094 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1095 1096 /* 1097 * Find a segment which begins after this one does. 1098 */ 1099 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1100 if (GETIP(q)->ip_off > ip->ip_off) 1101 break; 1102 } 1103 1104 /* 1105 * If there is a preceding segment, it may provide some of 1106 * our data already. If so, drop the data from the incoming 1107 * segment. If it provides all of our data, drop us, otherwise 1108 * stick new segment in the proper place. 1109 * 1110 * If some of the data is dropped from the the preceding 1111 * segment, then it's checksum is invalidated. 1112 */ 1113 if (p) { 1114 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1115 if (i > 0) { 1116 if (i >= ip->ip_len) 1117 goto dropfrag; 1118 m_adj(m, i); 1119 m->m_pkthdr.csum_flags = 0; 1120 ip->ip_off += i; 1121 ip->ip_len -= i; 1122 } 1123 m->m_nextpkt = p->m_nextpkt; 1124 p->m_nextpkt = m; 1125 } else { 1126 m->m_nextpkt = fp->ipq_frags; 1127 fp->ipq_frags = m; 1128 } 1129 1130 /* 1131 * While we overlap succeeding segments trim them or, 1132 * if they are completely covered, dequeue them. 1133 */ 1134 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1135 q = nq) { 1136 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1137 if (i < GETIP(q)->ip_len) { 1138 GETIP(q)->ip_len -= i; 1139 GETIP(q)->ip_off += i; 1140 m_adj(q, i); 1141 q->m_pkthdr.csum_flags = 0; 1142 break; 1143 } 1144 nq = q->m_nextpkt; 1145 m->m_nextpkt = nq; 1146 ipstat.ips_fragdropped++; 1147 fp->ipq_nfrags--; 1148 q->m_nextpkt = NULL; 1149 m_freem(q); 1150 } 1151 1152 inserted: 1153 /* 1154 * Check for complete reassembly and perform frag per packet 1155 * limiting. 1156 * 1157 * Frag limiting is performed here so that the nth frag has 1158 * a chance to complete the packet before we drop the packet. 1159 * As a result, n+1 frags are actually allowed per packet, but 1160 * only n will ever be stored. (n = maxfragsperpacket.) 1161 * 1162 */ 1163 next = 0; 1164 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1165 if (GETIP(q)->ip_off != next) { 1166 if (fp->ipq_nfrags > maxfragsperpacket) { 1167 ipstat.ips_fragdropped += fp->ipq_nfrags; 1168 ip_freef(fp); 1169 } 1170 goto done; 1171 } 1172 next += GETIP(q)->ip_len; 1173 } 1174 /* Make sure the last packet didn't have the IP_MF flag */ 1175 if (p->m_flags & M_FRAG) { 1176 if (fp->ipq_nfrags > maxfragsperpacket) { 1177 ipstat.ips_fragdropped += fp->ipq_nfrags; 1178 ip_freef(fp); 1179 } 1180 goto done; 1181 } 1182 1183 /* 1184 * Reassembly is complete. Make sure the packet is a sane size. 1185 */ 1186 q = fp->ipq_frags; 1187 ip = GETIP(q); 1188 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 1189 ipstat.ips_toolong++; 1190 ipstat.ips_fragdropped += fp->ipq_nfrags; 1191 ip_freef(fp); 1192 goto done; 1193 } 1194 1195 /* 1196 * Concatenate fragments. 1197 */ 1198 m = q; 1199 n = m->m_next; 1200 m->m_next = NULL; 1201 m_cat(m, n); 1202 nq = q->m_nextpkt; 1203 q->m_nextpkt = NULL; 1204 for (q = nq; q != NULL; q = nq) { 1205 nq = q->m_nextpkt; 1206 q->m_nextpkt = NULL; 1207 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1208 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1209 m_cat(m, q); 1210 } 1211 1212 /* 1213 * Clean up the 1's complement checksum. Carry over 16 bits must 1214 * be added back. This assumes no more then 65535 packet fragments 1215 * were reassembled. A second carry can also occur (but not a third). 1216 */ 1217 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) + 1218 (m->m_pkthdr.csum_data >> 16); 1219 if (m->m_pkthdr.csum_data > 0xFFFF) 1220 m->m_pkthdr.csum_data -= 0xFFFF; 1221 1222 /* 1223 * Create header for new ip packet by 1224 * modifying header of first packet; 1225 * dequeue and discard fragment reassembly header. 1226 * Make header visible. 1227 */ 1228 ip->ip_len = next; 1229 ip->ip_src = fp->ipq_src; 1230 ip->ip_dst = fp->ipq_dst; 1231 remque(fp); 1232 nipq--; 1233 mpipe_free(&ipq_mpipe, fp); 1234 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 1235 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 1236 /* some debugging cruft by sklower, below, will go away soon */ 1237 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1238 int plen = 0; 1239 1240 for (n = m; n; n = n->m_next) 1241 plen += n->m_len; 1242 m->m_pkthdr.len = plen; 1243 } 1244 1245 /* 1246 * Reassembly complete, return the next protocol. 1247 * 1248 * Be sure to clear M_HASH to force the packet 1249 * to be re-characterized. 1250 * 1251 * Clear M_FRAG, we are no longer a fragment. 1252 */ 1253 m->m_flags &= ~(M_HASH | M_FRAG); 1254 1255 ipstat.ips_reassembled++; 1256 lwkt_reltoken(&ipq_token); 1257 return (m); 1258 1259 dropfrag: 1260 ipstat.ips_fragdropped++; 1261 if (fp != NULL) 1262 fp->ipq_nfrags--; 1263 m_freem(m); 1264 done: 1265 lwkt_reltoken(&ipq_token); 1266 return (NULL); 1267 1268 #undef GETIP 1269 } 1270 1271 /* 1272 * Free a fragment reassembly header and all 1273 * associated datagrams. 1274 * 1275 * Called with ipq_token held. 1276 */ 1277 static void 1278 ip_freef(struct ipq *fp) 1279 { 1280 struct mbuf *q; 1281 1282 /* 1283 * Remove first to protect against blocking 1284 */ 1285 remque(fp); 1286 1287 /* 1288 * Clean out at our leisure 1289 */ 1290 while (fp->ipq_frags) { 1291 q = fp->ipq_frags; 1292 fp->ipq_frags = q->m_nextpkt; 1293 q->m_nextpkt = NULL; 1294 m_freem(q); 1295 } 1296 mpipe_free(&ipq_mpipe, fp); 1297 nipq--; 1298 } 1299 1300 /* 1301 * IP timer processing; 1302 * if a timer expires on a reassembly 1303 * queue, discard it. 1304 */ 1305 void 1306 ip_slowtimo(void) 1307 { 1308 struct ipq *fp; 1309 int i; 1310 1311 lwkt_gettoken(&ipq_token); 1312 for (i = 0; i < IPREASS_NHASH; i++) { 1313 fp = ipq[i].next; 1314 if (fp == NULL) 1315 continue; 1316 while (fp != &ipq[i]) { 1317 --fp->ipq_ttl; 1318 fp = fp->next; 1319 if (fp->prev->ipq_ttl == 0) { 1320 ipstat.ips_fragtimeout += fp->prev->ipq_nfrags; 1321 ip_freef(fp->prev); 1322 } 1323 } 1324 } 1325 /* 1326 * If we are over the maximum number of fragments 1327 * (due to the limit being lowered), drain off 1328 * enough to get down to the new limit. 1329 */ 1330 if (maxnipq >= 0 && nipq > maxnipq) { 1331 for (i = 0; i < IPREASS_NHASH; i++) { 1332 while (nipq > maxnipq && 1333 (ipq[i].next != &ipq[i])) { 1334 ipstat.ips_fragdropped += 1335 ipq[i].next->ipq_nfrags; 1336 ip_freef(ipq[i].next); 1337 } 1338 } 1339 } 1340 lwkt_reltoken(&ipq_token); 1341 ipflow_slowtimo(); 1342 } 1343 1344 /* 1345 * Drain off all datagram fragments. 1346 */ 1347 void 1348 ip_drain(void) 1349 { 1350 int i; 1351 1352 lwkt_gettoken(&ipq_token); 1353 for (i = 0; i < IPREASS_NHASH; i++) { 1354 while (ipq[i].next != &ipq[i]) { 1355 ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags; 1356 ip_freef(ipq[i].next); 1357 } 1358 } 1359 lwkt_reltoken(&ipq_token); 1360 in_rtqdrain(); 1361 } 1362 1363 /* 1364 * Do option processing on a datagram, 1365 * possibly discarding it if bad options are encountered, 1366 * or forwarding it if source-routed. 1367 * The pass argument is used when operating in the IPSTEALTH 1368 * mode to tell what options to process: 1369 * [LS]SRR (pass 0) or the others (pass 1). 1370 * The reason for as many as two passes is that when doing IPSTEALTH, 1371 * non-routing options should be processed only if the packet is for us. 1372 * Returns 1 if packet has been forwarded/freed, 1373 * 0 if the packet should be processed further. 1374 */ 1375 static int 1376 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop) 1377 { 1378 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET }; 1379 struct ip *ip = mtod(m, struct ip *); 1380 u_char *cp; 1381 struct in_ifaddr *ia; 1382 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB; 1383 boolean_t forward = FALSE; 1384 struct in_addr *sin, dst; 1385 n_time ntime; 1386 1387 dst = ip->ip_dst; 1388 cp = (u_char *)(ip + 1); 1389 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1390 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1391 opt = cp[IPOPT_OPTVAL]; 1392 if (opt == IPOPT_EOL) 1393 break; 1394 if (opt == IPOPT_NOP) 1395 optlen = 1; 1396 else { 1397 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1398 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1399 goto bad; 1400 } 1401 optlen = cp[IPOPT_OLEN]; 1402 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1403 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1404 goto bad; 1405 } 1406 } 1407 switch (opt) { 1408 1409 default: 1410 break; 1411 1412 /* 1413 * Source routing with record. 1414 * Find interface with current destination address. 1415 * If none on this machine then drop if strictly routed, 1416 * or do nothing if loosely routed. 1417 * Record interface address and bring up next address 1418 * component. If strictly routed make sure next 1419 * address is on directly accessible net. 1420 */ 1421 case IPOPT_LSRR: 1422 case IPOPT_SSRR: 1423 if (ipstealth && pass > 0) 1424 break; 1425 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1426 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1427 goto bad; 1428 } 1429 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1430 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1431 goto bad; 1432 } 1433 ipaddr.sin_addr = ip->ip_dst; 1434 ia = (struct in_ifaddr *) 1435 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1436 if (ia == NULL) { 1437 if (opt == IPOPT_SSRR) { 1438 type = ICMP_UNREACH; 1439 code = ICMP_UNREACH_SRCFAIL; 1440 goto bad; 1441 } 1442 if (!ip_dosourceroute) 1443 goto nosourcerouting; 1444 /* 1445 * Loose routing, and not at next destination 1446 * yet; nothing to do except forward. 1447 */ 1448 break; 1449 } 1450 off--; /* 0 origin */ 1451 if (off > optlen - (int)sizeof(struct in_addr)) { 1452 /* 1453 * End of source route. Should be for us. 1454 */ 1455 if (!ip_acceptsourceroute) 1456 goto nosourcerouting; 1457 save_rte(m, cp, ip->ip_src); 1458 break; 1459 } 1460 if (ipstealth) 1461 goto dropit; 1462 if (!ip_dosourceroute) { 1463 if (ipforwarding) { 1464 char buf[sizeof "aaa.bbb.ccc.ddd"]; 1465 1466 /* 1467 * Acting as a router, so generate ICMP 1468 */ 1469 nosourcerouting: 1470 strcpy(buf, inet_ntoa(ip->ip_dst)); 1471 log(LOG_WARNING, 1472 "attempted source route from %s to %s\n", 1473 inet_ntoa(ip->ip_src), buf); 1474 type = ICMP_UNREACH; 1475 code = ICMP_UNREACH_SRCFAIL; 1476 goto bad; 1477 } else { 1478 /* 1479 * Not acting as a router, 1480 * so silently drop. 1481 */ 1482 dropit: 1483 ipstat.ips_cantforward++; 1484 m_freem(m); 1485 return (1); 1486 } 1487 } 1488 1489 /* 1490 * locate outgoing interface 1491 */ 1492 memcpy(&ipaddr.sin_addr, cp + off, 1493 sizeof ipaddr.sin_addr); 1494 1495 if (opt == IPOPT_SSRR) { 1496 #define INA struct in_ifaddr * 1497 #define SA struct sockaddr * 1498 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) 1499 == NULL) 1500 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1501 } else { 1502 ia = ip_rtaddr(ipaddr.sin_addr, NULL); 1503 } 1504 if (ia == NULL) { 1505 type = ICMP_UNREACH; 1506 code = ICMP_UNREACH_SRCFAIL; 1507 goto bad; 1508 } 1509 ip->ip_dst = ipaddr.sin_addr; 1510 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1511 sizeof(struct in_addr)); 1512 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1513 /* 1514 * Let ip_intr's mcast routing check handle mcast pkts 1515 */ 1516 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1517 break; 1518 1519 case IPOPT_RR: 1520 if (ipstealth && pass == 0) 1521 break; 1522 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1523 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1524 goto bad; 1525 } 1526 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1527 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1528 goto bad; 1529 } 1530 /* 1531 * If no space remains, ignore. 1532 */ 1533 off--; /* 0 origin */ 1534 if (off > optlen - (int)sizeof(struct in_addr)) 1535 break; 1536 memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1537 sizeof ipaddr.sin_addr); 1538 /* 1539 * locate outgoing interface; if we're the destination, 1540 * use the incoming interface (should be same). 1541 */ 1542 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL && 1543 (ia = ip_rtaddr(ipaddr.sin_addr, NULL)) == NULL) { 1544 type = ICMP_UNREACH; 1545 code = ICMP_UNREACH_HOST; 1546 goto bad; 1547 } 1548 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1549 sizeof(struct in_addr)); 1550 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1551 break; 1552 1553 case IPOPT_TS: 1554 if (ipstealth && pass == 0) 1555 break; 1556 code = cp - (u_char *)ip; 1557 if (optlen < 4 || optlen > 40) { 1558 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1559 goto bad; 1560 } 1561 if ((off = cp[IPOPT_OFFSET]) < 5) { 1562 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1563 goto bad; 1564 } 1565 if (off > optlen - (int)sizeof(int32_t)) { 1566 cp[IPOPT_OFFSET + 1] += (1 << 4); 1567 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1568 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1569 goto bad; 1570 } 1571 break; 1572 } 1573 off--; /* 0 origin */ 1574 sin = (struct in_addr *)(cp + off); 1575 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1576 1577 case IPOPT_TS_TSONLY: 1578 break; 1579 1580 case IPOPT_TS_TSANDADDR: 1581 if (off + sizeof(n_time) + 1582 sizeof(struct in_addr) > optlen) { 1583 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1584 goto bad; 1585 } 1586 ipaddr.sin_addr = dst; 1587 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1588 m->m_pkthdr.rcvif); 1589 if (ia == NULL) 1590 continue; 1591 memcpy(sin, &IA_SIN(ia)->sin_addr, 1592 sizeof(struct in_addr)); 1593 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1594 off += sizeof(struct in_addr); 1595 break; 1596 1597 case IPOPT_TS_PRESPEC: 1598 if (off + sizeof(n_time) + 1599 sizeof(struct in_addr) > optlen) { 1600 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1601 goto bad; 1602 } 1603 memcpy(&ipaddr.sin_addr, sin, 1604 sizeof(struct in_addr)); 1605 if (ifa_ifwithaddr((SA)&ipaddr) == NULL) 1606 continue; 1607 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1608 off += sizeof(struct in_addr); 1609 break; 1610 1611 default: 1612 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1613 goto bad; 1614 } 1615 ntime = iptime(); 1616 memcpy(cp + off, &ntime, sizeof(n_time)); 1617 cp[IPOPT_OFFSET] += sizeof(n_time); 1618 } 1619 } 1620 if (forward && ipforwarding) { 1621 ip_forward(m, TRUE, next_hop); 1622 return (1); 1623 } 1624 return (0); 1625 bad: 1626 icmp_error(m, type, code, 0, 0); 1627 ipstat.ips_badoptions++; 1628 return (1); 1629 } 1630 1631 /* 1632 * Given address of next destination (final or next hop), 1633 * return internet address info of interface to be used to get there. 1634 */ 1635 struct in_ifaddr * 1636 ip_rtaddr(struct in_addr dst, struct route *ro0) 1637 { 1638 struct route sro, *ro; 1639 struct sockaddr_in *sin; 1640 struct in_ifaddr *ia; 1641 1642 if (ro0 != NULL) { 1643 ro = ro0; 1644 } else { 1645 bzero(&sro, sizeof(sro)); 1646 ro = &sro; 1647 } 1648 1649 sin = (struct sockaddr_in *)&ro->ro_dst; 1650 1651 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) { 1652 if (ro->ro_rt != NULL) { 1653 RTFREE(ro->ro_rt); 1654 ro->ro_rt = NULL; 1655 } 1656 sin->sin_family = AF_INET; 1657 sin->sin_len = sizeof *sin; 1658 sin->sin_addr = dst; 1659 rtalloc_ign(ro, RTF_PRCLONING); 1660 } 1661 1662 if (ro->ro_rt == NULL) 1663 return (NULL); 1664 1665 ia = ifatoia(ro->ro_rt->rt_ifa); 1666 1667 if (ro == &sro) 1668 RTFREE(ro->ro_rt); 1669 return ia; 1670 } 1671 1672 /* 1673 * Save incoming source route for use in replies, 1674 * to be picked up later by ip_srcroute if the receiver is interested. 1675 */ 1676 static void 1677 save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1678 { 1679 struct m_tag *mtag; 1680 struct ip_srcrt_opt *opt; 1681 unsigned olen; 1682 1683 mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), MB_DONTWAIT); 1684 if (mtag == NULL) 1685 return; 1686 opt = m_tag_data(mtag); 1687 1688 olen = option[IPOPT_OLEN]; 1689 #ifdef DIAGNOSTIC 1690 if (ipprintfs) 1691 kprintf("save_rte: olen %d\n", olen); 1692 #endif 1693 if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) { 1694 m_tag_free(mtag); 1695 return; 1696 } 1697 bcopy(option, opt->ip_srcrt.srcopt, olen); 1698 opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1699 opt->ip_srcrt.dst = dst; 1700 m_tag_prepend(m, mtag); 1701 } 1702 1703 /* 1704 * Retrieve incoming source route for use in replies, 1705 * in the same form used by setsockopt. 1706 * The first hop is placed before the options, will be removed later. 1707 */ 1708 struct mbuf * 1709 ip_srcroute(struct mbuf *m0) 1710 { 1711 struct in_addr *p, *q; 1712 struct mbuf *m; 1713 struct m_tag *mtag; 1714 struct ip_srcrt_opt *opt; 1715 1716 if (m0 == NULL) 1717 return NULL; 1718 1719 mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL); 1720 if (mtag == NULL) 1721 return NULL; 1722 opt = m_tag_data(mtag); 1723 1724 if (opt->ip_nhops == 0) 1725 return (NULL); 1726 m = m_get(MB_DONTWAIT, MT_HEADER); 1727 if (m == NULL) 1728 return (NULL); 1729 1730 #define OPTSIZ (sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt)) 1731 1732 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1733 m->m_len = opt->ip_nhops * sizeof(struct in_addr) + 1734 sizeof(struct in_addr) + OPTSIZ; 1735 #ifdef DIAGNOSTIC 1736 if (ipprintfs) { 1737 kprintf("ip_srcroute: nhops %d mlen %d", 1738 opt->ip_nhops, m->m_len); 1739 } 1740 #endif 1741 1742 /* 1743 * First save first hop for return route 1744 */ 1745 p = &opt->ip_srcrt.route[opt->ip_nhops - 1]; 1746 *(mtod(m, struct in_addr *)) = *p--; 1747 #ifdef DIAGNOSTIC 1748 if (ipprintfs) 1749 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1750 #endif 1751 1752 /* 1753 * Copy option fields and padding (nop) to mbuf. 1754 */ 1755 opt->ip_srcrt.nop = IPOPT_NOP; 1756 opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1757 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop, 1758 OPTSIZ); 1759 q = (struct in_addr *)(mtod(m, caddr_t) + 1760 sizeof(struct in_addr) + OPTSIZ); 1761 #undef OPTSIZ 1762 /* 1763 * Record return path as an IP source route, 1764 * reversing the path (pointers are now aligned). 1765 */ 1766 while (p >= opt->ip_srcrt.route) { 1767 #ifdef DIAGNOSTIC 1768 if (ipprintfs) 1769 kprintf(" %x", ntohl(q->s_addr)); 1770 #endif 1771 *q++ = *p--; 1772 } 1773 /* 1774 * Last hop goes to final destination. 1775 */ 1776 *q = opt->ip_srcrt.dst; 1777 m_tag_delete(m0, mtag); 1778 #ifdef DIAGNOSTIC 1779 if (ipprintfs) 1780 kprintf(" %x\n", ntohl(q->s_addr)); 1781 #endif 1782 return (m); 1783 } 1784 1785 /* 1786 * Strip out IP options. 1787 */ 1788 void 1789 ip_stripoptions(struct mbuf *m) 1790 { 1791 int datalen; 1792 struct ip *ip = mtod(m, struct ip *); 1793 caddr_t opts; 1794 int optlen; 1795 1796 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1797 opts = (caddr_t)(ip + 1); 1798 datalen = m->m_len - (sizeof(struct ip) + optlen); 1799 bcopy(opts + optlen, opts, datalen); 1800 m->m_len -= optlen; 1801 if (m->m_flags & M_PKTHDR) 1802 m->m_pkthdr.len -= optlen; 1803 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1804 } 1805 1806 u_char inetctlerrmap[PRC_NCMDS] = { 1807 0, 0, 0, 0, 1808 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1809 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1810 EMSGSIZE, EHOSTUNREACH, 0, 0, 1811 0, 0, 0, 0, 1812 ENOPROTOOPT, ECONNREFUSED 1813 }; 1814 1815 /* 1816 * Forward a packet. If some error occurs return the sender 1817 * an icmp packet. Note we can't always generate a meaningful 1818 * icmp message because icmp doesn't have a large enough repertoire 1819 * of codes and types. 1820 * 1821 * If not forwarding, just drop the packet. This could be confusing 1822 * if ipforwarding was zero but some routing protocol was advancing 1823 * us as a gateway to somewhere. However, we must let the routing 1824 * protocol deal with that. 1825 * 1826 * The using_srcrt parameter indicates whether the packet is being forwarded 1827 * via a source route. 1828 */ 1829 void 1830 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop) 1831 { 1832 struct ip *ip = mtod(m, struct ip *); 1833 struct rtentry *rt; 1834 struct route fwd_ro; 1835 int error, type = 0, code = 0, destmtu = 0; 1836 struct mbuf *mcopy; 1837 n_long dest; 1838 struct in_addr pkt_dst; 1839 1840 dest = INADDR_ANY; 1841 /* 1842 * Cache the destination address of the packet; this may be 1843 * changed by use of 'ipfw fwd'. 1844 */ 1845 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst; 1846 1847 #ifdef DIAGNOSTIC 1848 if (ipprintfs) 1849 kprintf("forward: src %x dst %x ttl %x\n", 1850 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl); 1851 #endif 1852 1853 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) { 1854 ipstat.ips_cantforward++; 1855 m_freem(m); 1856 return; 1857 } 1858 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) { 1859 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1860 return; 1861 } 1862 1863 bzero(&fwd_ro, sizeof(fwd_ro)); 1864 ip_rtaddr(pkt_dst, &fwd_ro); 1865 if (fwd_ro.ro_rt == NULL) { 1866 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1867 return; 1868 } 1869 rt = fwd_ro.ro_rt; 1870 1871 /* 1872 * Save the IP header and at most 8 bytes of the payload, 1873 * in case we need to generate an ICMP message to the src. 1874 * 1875 * XXX this can be optimized a lot by saving the data in a local 1876 * buffer on the stack (72 bytes at most), and only allocating the 1877 * mbuf if really necessary. The vast majority of the packets 1878 * are forwarded without having to send an ICMP back (either 1879 * because unnecessary, or because rate limited), so we are 1880 * really we are wasting a lot of work here. 1881 * 1882 * We don't use m_copy() because it might return a reference 1883 * to a shared cluster. Both this function and ip_output() 1884 * assume exclusive access to the IP header in `m', so any 1885 * data in a cluster may change before we reach icmp_error(). 1886 */ 1887 MGETHDR(mcopy, MB_DONTWAIT, m->m_type); 1888 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, MB_DONTWAIT)) { 1889 /* 1890 * It's probably ok if the pkthdr dup fails (because 1891 * the deep copy of the tag chain failed), but for now 1892 * be conservative and just discard the copy since 1893 * code below may some day want the tags. 1894 */ 1895 m_free(mcopy); 1896 mcopy = NULL; 1897 } 1898 if (mcopy != NULL) { 1899 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8, 1900 (int)ip->ip_len); 1901 mcopy->m_pkthdr.len = mcopy->m_len; 1902 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1903 } 1904 1905 if (!ipstealth) 1906 ip->ip_ttl -= IPTTLDEC; 1907 1908 /* 1909 * If forwarding packet using same interface that it came in on, 1910 * perhaps should send a redirect to sender to shortcut a hop. 1911 * Only send redirect if source is sending directly to us, 1912 * and if packet was not source routed (or has any options). 1913 * Also, don't send redirect if forwarding using a default route 1914 * or a route modified by a redirect. 1915 */ 1916 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1917 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) && 1918 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY && 1919 ipsendredirects && !using_srcrt && next_hop == NULL) { 1920 u_long src = ntohl(ip->ip_src.s_addr); 1921 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa; 1922 1923 if (rt_ifa != NULL && 1924 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) { 1925 if (rt->rt_flags & RTF_GATEWAY) 1926 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1927 else 1928 dest = pkt_dst.s_addr; 1929 /* 1930 * Router requirements says to only send 1931 * host redirects. 1932 */ 1933 type = ICMP_REDIRECT; 1934 code = ICMP_REDIRECT_HOST; 1935 #ifdef DIAGNOSTIC 1936 if (ipprintfs) 1937 kprintf("redirect (%d) to %x\n", code, dest); 1938 #endif 1939 } 1940 } 1941 1942 error = ip_output(m, NULL, &fwd_ro, IP_FORWARDING, NULL, NULL); 1943 if (error == 0) { 1944 ipstat.ips_forward++; 1945 if (type == 0) { 1946 if (mcopy) { 1947 ipflow_create(&fwd_ro, mcopy); 1948 m_freem(mcopy); 1949 } 1950 goto done; 1951 } else { 1952 ipstat.ips_redirectsent++; 1953 } 1954 } else { 1955 ipstat.ips_cantforward++; 1956 } 1957 1958 if (mcopy == NULL) 1959 goto done; 1960 1961 /* 1962 * Send ICMP message. 1963 */ 1964 1965 switch (error) { 1966 1967 case 0: /* forwarded, but need redirect */ 1968 /* type, code set above */ 1969 break; 1970 1971 case ENETUNREACH: /* shouldn't happen, checked above */ 1972 case EHOSTUNREACH: 1973 case ENETDOWN: 1974 case EHOSTDOWN: 1975 default: 1976 type = ICMP_UNREACH; 1977 code = ICMP_UNREACH_HOST; 1978 break; 1979 1980 case EMSGSIZE: 1981 type = ICMP_UNREACH; 1982 code = ICMP_UNREACH_NEEDFRAG; 1983 #ifdef IPSEC 1984 /* 1985 * If the packet is routed over IPsec tunnel, tell the 1986 * originator the tunnel MTU. 1987 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1988 * XXX quickhack!!! 1989 */ 1990 if (fwd_ro.ro_rt != NULL) { 1991 struct secpolicy *sp = NULL; 1992 int ipsecerror; 1993 int ipsechdr; 1994 struct route *ro; 1995 1996 sp = ipsec4_getpolicybyaddr(mcopy, 1997 IPSEC_DIR_OUTBOUND, 1998 IP_FORWARDING, 1999 &ipsecerror); 2000 2001 if (sp == NULL) 2002 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2003 else { 2004 /* count IPsec header size */ 2005 ipsechdr = ipsec4_hdrsiz(mcopy, 2006 IPSEC_DIR_OUTBOUND, 2007 NULL); 2008 2009 /* 2010 * find the correct route for outer IPv4 2011 * header, compute tunnel MTU. 2012 * 2013 */ 2014 if (sp->req != NULL && sp->req->sav != NULL && 2015 sp->req->sav->sah != NULL) { 2016 ro = &sp->req->sav->sah->sa_route; 2017 if (ro->ro_rt != NULL && 2018 ro->ro_rt->rt_ifp != NULL) { 2019 destmtu = 2020 ro->ro_rt->rt_ifp->if_mtu; 2021 destmtu -= ipsechdr; 2022 } 2023 } 2024 2025 key_freesp(sp); 2026 } 2027 } 2028 #elif FAST_IPSEC 2029 /* 2030 * If the packet is routed over IPsec tunnel, tell the 2031 * originator the tunnel MTU. 2032 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 2033 * XXX quickhack!!! 2034 */ 2035 if (fwd_ro.ro_rt != NULL) { 2036 struct secpolicy *sp = NULL; 2037 int ipsecerror; 2038 int ipsechdr; 2039 struct route *ro; 2040 2041 sp = ipsec_getpolicybyaddr(mcopy, 2042 IPSEC_DIR_OUTBOUND, 2043 IP_FORWARDING, 2044 &ipsecerror); 2045 2046 if (sp == NULL) 2047 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2048 else { 2049 /* count IPsec header size */ 2050 ipsechdr = ipsec4_hdrsiz(mcopy, 2051 IPSEC_DIR_OUTBOUND, 2052 NULL); 2053 2054 /* 2055 * find the correct route for outer IPv4 2056 * header, compute tunnel MTU. 2057 */ 2058 2059 if (sp->req != NULL && 2060 sp->req->sav != NULL && 2061 sp->req->sav->sah != NULL) { 2062 ro = &sp->req->sav->sah->sa_route; 2063 if (ro->ro_rt != NULL && 2064 ro->ro_rt->rt_ifp != NULL) { 2065 destmtu = 2066 ro->ro_rt->rt_ifp->if_mtu; 2067 destmtu -= ipsechdr; 2068 } 2069 } 2070 2071 KEY_FREESP(&sp); 2072 } 2073 } 2074 #else /* !IPSEC && !FAST_IPSEC */ 2075 if (fwd_ro.ro_rt != NULL) 2076 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2077 #endif /*IPSEC*/ 2078 ipstat.ips_cantfrag++; 2079 break; 2080 2081 case ENOBUFS: 2082 /* 2083 * A router should not generate ICMP_SOURCEQUENCH as 2084 * required in RFC1812 Requirements for IP Version 4 Routers. 2085 * Source quench could be a big problem under DoS attacks, 2086 * or if the underlying interface is rate-limited. 2087 * Those who need source quench packets may re-enable them 2088 * via the net.inet.ip.sendsourcequench sysctl. 2089 */ 2090 if (!ip_sendsourcequench) { 2091 m_freem(mcopy); 2092 goto done; 2093 } else { 2094 type = ICMP_SOURCEQUENCH; 2095 code = 0; 2096 } 2097 break; 2098 2099 case EACCES: /* ipfw denied packet */ 2100 m_freem(mcopy); 2101 goto done; 2102 } 2103 icmp_error(mcopy, type, code, dest, destmtu); 2104 done: 2105 if (fwd_ro.ro_rt != NULL) 2106 RTFREE(fwd_ro.ro_rt); 2107 } 2108 2109 void 2110 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2111 struct mbuf *m) 2112 { 2113 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2114 struct timeval tv; 2115 2116 microtime(&tv); 2117 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2118 SCM_TIMESTAMP, SOL_SOCKET); 2119 if (*mp) 2120 mp = &(*mp)->m_next; 2121 } 2122 if (inp->inp_flags & INP_RECVDSTADDR) { 2123 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2124 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2125 if (*mp) 2126 mp = &(*mp)->m_next; 2127 } 2128 if (inp->inp_flags & INP_RECVTTL) { 2129 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 2130 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 2131 if (*mp) 2132 mp = &(*mp)->m_next; 2133 } 2134 #ifdef notyet 2135 /* XXX 2136 * Moving these out of udp_input() made them even more broken 2137 * than they already were. 2138 */ 2139 /* options were tossed already */ 2140 if (inp->inp_flags & INP_RECVOPTS) { 2141 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2142 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2143 if (*mp) 2144 mp = &(*mp)->m_next; 2145 } 2146 /* ip_srcroute doesn't do what we want here, need to fix */ 2147 if (inp->inp_flags & INP_RECVRETOPTS) { 2148 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 2149 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2150 if (*mp) 2151 mp = &(*mp)->m_next; 2152 } 2153 #endif 2154 if (inp->inp_flags & INP_RECVIF) { 2155 struct ifnet *ifp; 2156 struct sdlbuf { 2157 struct sockaddr_dl sdl; 2158 u_char pad[32]; 2159 } sdlbuf; 2160 struct sockaddr_dl *sdp; 2161 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2162 2163 if (((ifp = m->m_pkthdr.rcvif)) && 2164 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) { 2165 sdp = IF_LLSOCKADDR(ifp); 2166 /* 2167 * Change our mind and don't try copy. 2168 */ 2169 if ((sdp->sdl_family != AF_LINK) || 2170 (sdp->sdl_len > sizeof(sdlbuf))) { 2171 goto makedummy; 2172 } 2173 bcopy(sdp, sdl2, sdp->sdl_len); 2174 } else { 2175 makedummy: 2176 sdl2->sdl_len = 2177 offsetof(struct sockaddr_dl, sdl_data[0]); 2178 sdl2->sdl_family = AF_LINK; 2179 sdl2->sdl_index = 0; 2180 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2181 } 2182 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 2183 IP_RECVIF, IPPROTO_IP); 2184 if (*mp) 2185 mp = &(*mp)->m_next; 2186 } 2187 } 2188 2189 /* 2190 * XXX these routines are called from the upper part of the kernel. 2191 * 2192 * They could also be moved to ip_mroute.c, since all the RSVP 2193 * handling is done there already. 2194 */ 2195 int 2196 ip_rsvp_init(struct socket *so) 2197 { 2198 if (so->so_type != SOCK_RAW || 2199 so->so_proto->pr_protocol != IPPROTO_RSVP) 2200 return EOPNOTSUPP; 2201 2202 if (ip_rsvpd != NULL) 2203 return EADDRINUSE; 2204 2205 ip_rsvpd = so; 2206 /* 2207 * This may seem silly, but we need to be sure we don't over-increment 2208 * the RSVP counter, in case something slips up. 2209 */ 2210 if (!ip_rsvp_on) { 2211 ip_rsvp_on = 1; 2212 rsvp_on++; 2213 } 2214 2215 return 0; 2216 } 2217 2218 int 2219 ip_rsvp_done(void) 2220 { 2221 ip_rsvpd = NULL; 2222 /* 2223 * This may seem silly, but we need to be sure we don't over-decrement 2224 * the RSVP counter, in case something slips up. 2225 */ 2226 if (ip_rsvp_on) { 2227 ip_rsvp_on = 0; 2228 rsvp_on--; 2229 } 2230 return 0; 2231 } 2232 2233 int 2234 rsvp_input(struct mbuf **mp, int *offp, int proto) 2235 { 2236 struct mbuf *m = *mp; 2237 int off; 2238 2239 off = *offp; 2240 *mp = NULL; 2241 2242 if (rsvp_input_p) { /* call the real one if loaded */ 2243 *mp = m; 2244 rsvp_input_p(mp, offp, proto); 2245 return(IPPROTO_DONE); 2246 } 2247 2248 /* Can still get packets with rsvp_on = 0 if there is a local member 2249 * of the group to which the RSVP packet is addressed. But in this 2250 * case we want to throw the packet away. 2251 */ 2252 2253 if (!rsvp_on) { 2254 m_freem(m); 2255 return(IPPROTO_DONE); 2256 } 2257 2258 if (ip_rsvpd != NULL) { 2259 *mp = m; 2260 rip_input(mp, offp, proto); 2261 return(IPPROTO_DONE); 2262 } 2263 /* Drop the packet */ 2264 m_freem(m); 2265 return(IPPROTO_DONE); 2266 } 2267