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