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 <machine/stdarg.h> 99 100 #include <net/if.h> 101 #include <net/if_types.h> 102 #include <net/if_var.h> 103 #include <net/if_dl.h> 104 #include <net/pfil.h> 105 #include <net/route.h> 106 #include <net/netisr.h> 107 108 #include <netinet/in.h> 109 #include <netinet/in_systm.h> 110 #include <netinet/in_var.h> 111 #include <netinet/ip.h> 112 #include <netinet/in_pcb.h> 113 #include <netinet/ip_var.h> 114 #include <netinet/ip_icmp.h> 115 #include <netinet/ip_divert.h> 116 #include <netinet/ip_flow.h> 117 118 #include <sys/thread2.h> 119 #include <sys/msgport2.h> 120 #include <net/netmsg2.h> 121 122 #include <sys/socketvar.h> 123 124 #include <net/ipfw/ip_fw.h> 125 #include <net/dummynet/ip_dummynet.h> 126 127 #ifdef IPSEC 128 #include <netinet6/ipsec.h> 129 #include <netproto/key/key.h> 130 #endif 131 132 #ifdef FAST_IPSEC 133 #include <netproto/ipsec/ipsec.h> 134 #include <netproto/ipsec/key.h> 135 #endif 136 137 int rsvp_on = 0; 138 static int ip_rsvp_on; 139 struct socket *ip_rsvpd; 140 141 int ip_mpsafe = 1; 142 TUNABLE_INT("net.inet.ip.mpsafe", &ip_mpsafe); 143 144 int ipforwarding = 0; 145 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 146 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 147 148 static int ipsendredirects = 1; /* XXX */ 149 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 150 &ipsendredirects, 0, "Enable sending IP redirects"); 151 152 int ip_defttl = IPDEFTTL; 153 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 154 &ip_defttl, 0, "Maximum TTL on IP packets"); 155 156 static int ip_dosourceroute = 0; 157 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 158 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 159 160 static int ip_acceptsourceroute = 0; 161 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 162 CTLFLAG_RW, &ip_acceptsourceroute, 0, 163 "Enable accepting source routed IP packets"); 164 165 static int ip_keepfaith = 0; 166 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 167 &ip_keepfaith, 0, 168 "Enable packet capture for FAITH IPv4->IPv6 translator daemon"); 169 170 static int nipq = 0; /* total # of reass queues */ 171 static int maxnipq; 172 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 173 &maxnipq, 0, 174 "Maximum number of IPv4 fragment reassembly queue entries"); 175 176 static int maxfragsperpacket; 177 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 178 &maxfragsperpacket, 0, 179 "Maximum number of IPv4 fragments allowed per packet"); 180 181 static int ip_sendsourcequench = 0; 182 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 183 &ip_sendsourcequench, 0, 184 "Enable the transmission of source quench packets"); 185 186 int ip_do_randomid = 1; 187 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 188 &ip_do_randomid, 0, 189 "Assign random ip_id values"); 190 /* 191 * XXX - Setting ip_checkinterface mostly implements the receive side of 192 * the Strong ES model described in RFC 1122, but since the routing table 193 * and transmit implementation do not implement the Strong ES model, 194 * setting this to 1 results in an odd hybrid. 195 * 196 * XXX - ip_checkinterface currently must be disabled if you use ipnat 197 * to translate the destination address to another local interface. 198 * 199 * XXX - ip_checkinterface must be disabled if you add IP aliases 200 * to the loopback interface instead of the interface where the 201 * packets for those addresses are received. 202 */ 203 static int ip_checkinterface = 0; 204 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 205 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 206 207 #ifdef DIAGNOSTIC 208 static int ipprintfs = 0; 209 #endif 210 211 extern int udp_mpsafe_proto; 212 extern int tcp_mpsafe_proto; 213 214 extern struct domain inetdomain; 215 extern struct protosw inetsw[]; 216 u_char ip_protox[IPPROTO_MAX]; 217 struct in_ifaddrhead in_ifaddrheads[MAXCPU]; /* first inet address */ 218 struct in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU]; 219 /* inet addr hash table */ 220 u_long in_ifaddrhmask; /* mask for hash table */ 221 222 struct ip_stats ipstats_percpu[MAXCPU]; 223 #ifdef SMP 224 static int 225 sysctl_ipstats(SYSCTL_HANDLER_ARGS) 226 { 227 int cpu, error = 0; 228 229 for (cpu = 0; cpu < ncpus; ++cpu) { 230 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu], 231 sizeof(struct ip_stats)))) 232 break; 233 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu], 234 sizeof(struct ip_stats)))) 235 break; 236 } 237 238 return (error); 239 } 240 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW), 241 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics"); 242 #else 243 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 244 &ipstat, ip_stats, "IP statistics"); 245 #endif 246 247 /* Packet reassembly stuff */ 248 #define IPREASS_NHASH_LOG2 6 249 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 250 #define IPREASS_HMASK (IPREASS_NHASH - 1) 251 #define IPREASS_HASH(x,y) \ 252 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 253 254 static struct ipq ipq[IPREASS_NHASH]; 255 256 #ifdef IPCTL_DEFMTU 257 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 258 &ip_mtu, 0, "Default MTU"); 259 #endif 260 261 #ifdef IPSTEALTH 262 static int ipstealth = 0; 263 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, ""); 264 #else 265 static const int ipstealth = 0; 266 #endif 267 268 struct mbuf *(*ip_divert_p)(struct mbuf *, int, int); 269 270 struct pfil_head inet_pfil_hook; 271 272 /* 273 * struct ip_srcrt_opt is used to store packet state while it travels 274 * through the stack. 275 * 276 * XXX Note that the code even makes assumptions on the size and 277 * alignment of fields inside struct ip_srcrt so e.g. adding some 278 * fields will break the code. This needs to be fixed. 279 * 280 * We need to save the IP options in case a protocol wants to respond 281 * to an incoming packet over the same route if the packet got here 282 * using IP source routing. This allows connection establishment and 283 * maintenance when the remote end is on a network that is not known 284 * to us. 285 */ 286 struct ip_srcrt { 287 struct in_addr dst; /* final destination */ 288 char nop; /* one NOP to align */ 289 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 290 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 291 }; 292 293 struct ip_srcrt_opt { 294 int ip_nhops; 295 struct ip_srcrt ip_srcrt; 296 }; 297 298 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management"); 299 static struct malloc_pipe ipq_mpipe; 300 301 static void save_rte(struct mbuf *, u_char *, struct in_addr); 302 static int ip_dooptions(struct mbuf *m, int, struct sockaddr_in *); 303 static void ip_freef(struct ipq *); 304 static void ip_input_handler(struct netmsg *); 305 306 /* 307 * IP initialization: fill in IP protocol switch table. 308 * All protocols not implemented in kernel go to raw IP protocol handler. 309 */ 310 void 311 ip_init(void) 312 { 313 struct protosw *pr; 314 uint32_t flags; 315 int i; 316 #ifdef SMP 317 int cpu; 318 #endif 319 320 /* 321 * Make sure we can handle a reasonable number of fragments but 322 * cap it at 4000 (XXX). 323 */ 324 mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq), 325 IFQ_MAXLEN, 4000, 0, NULL); 326 for (i = 0; i < ncpus; ++i) { 327 TAILQ_INIT(&in_ifaddrheads[i]); 328 in_ifaddrhashtbls[i] = 329 hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 330 } 331 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 332 if (pr == NULL) 333 panic("ip_init"); 334 for (i = 0; i < IPPROTO_MAX; i++) 335 ip_protox[i] = pr - inetsw; 336 for (pr = inetdomain.dom_protosw; 337 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 338 if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol) { 339 if (pr->pr_protocol != IPPROTO_RAW) 340 ip_protox[pr->pr_protocol] = pr - inetsw; 341 342 /* XXX */ 343 switch (pr->pr_protocol) { 344 case IPPROTO_TCP: 345 if (tcp_mpsafe_proto) 346 pr->pr_flags |= PR_MPSAFE; 347 break; 348 349 case IPPROTO_UDP: 350 if (udp_mpsafe_proto) 351 pr->pr_flags |= PR_MPSAFE; 352 break; 353 } 354 } 355 } 356 357 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 358 inet_pfil_hook.ph_af = AF_INET; 359 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) { 360 kprintf("%s: WARNING: unable to register pfil hook, " 361 "error %d\n", __func__, i); 362 } 363 364 for (i = 0; i < IPREASS_NHASH; i++) 365 ipq[i].next = ipq[i].prev = &ipq[i]; 366 367 maxnipq = nmbclusters / 32; 368 maxfragsperpacket = 16; 369 370 ip_id = time_second & 0xffff; 371 372 /* 373 * Initialize IP statistics counters for each CPU. 374 * 375 */ 376 #ifdef SMP 377 for (cpu = 0; cpu < ncpus; ++cpu) { 378 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats)); 379 } 380 #else 381 bzero(&ipstat, sizeof(struct ip_stats)); 382 #endif 383 384 #if defined(IPSEC) || defined(FAST_IPSEC) 385 /* XXX IPSEC is not MPSAFE yet */ 386 flags = NETISR_FLAG_NOTMPSAFE; 387 #else 388 if (ip_mpsafe) { 389 kprintf("ip: MPSAFE\n"); 390 flags = NETISR_FLAG_MPSAFE; 391 } else { 392 flags = NETISR_FLAG_NOTMPSAFE; 393 } 394 #endif 395 netisr_register(NETISR_IP, ip_mport_in, ip_mport_pktinfo, 396 ip_input_handler, flags); 397 } 398 399 /* Do transport protocol processing. */ 400 static void 401 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip) 402 { 403 const struct protosw *pr = &inetsw[ip_protox[ip->ip_p]]; 404 405 /* 406 * Switch out to protocol's input routine. 407 */ 408 PR_GET_MPLOCK(pr); 409 pr->pr_input(m, hlen, ip->ip_p); 410 PR_REL_MPLOCK(pr); 411 } 412 413 static void 414 transport_processing_handler(netmsg_t netmsg) 415 { 416 struct netmsg_packet *pmsg = (struct netmsg_packet *)netmsg; 417 struct ip *ip; 418 int hlen; 419 420 ip = mtod(pmsg->nm_packet, struct ip *); 421 hlen = pmsg->nm_netmsg.nm_lmsg.u.ms_result; 422 423 transport_processing_oncpu(pmsg->nm_packet, hlen, ip); 424 /* netmsg was embedded in the mbuf, do not reply! */ 425 } 426 427 static void 428 ip_input_handler(struct netmsg *msg0) 429 { 430 struct mbuf *m = ((struct netmsg_packet *)msg0)->nm_packet; 431 432 ip_input(m); 433 /* msg0 was embedded in the mbuf, do not reply! */ 434 } 435 436 /* 437 * IP input routine. Checksum and byte swap header. If fragmented 438 * try to reassemble. Process options. Pass to next level. 439 */ 440 void 441 ip_input(struct mbuf *m) 442 { 443 struct ip *ip; 444 struct in_ifaddr *ia = NULL; 445 struct in_ifaddr_container *iac; 446 int hlen, checkif; 447 u_short sum; 448 struct in_addr pkt_dst; 449 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */ 450 boolean_t needredispatch = FALSE; 451 struct in_addr odst; /* original dst address(NAT) */ 452 struct m_tag *mtag; 453 struct sockaddr_in *next_hop = NULL; 454 #ifdef FAST_IPSEC 455 struct tdb_ident *tdbi; 456 struct secpolicy *sp; 457 int error; 458 #endif 459 460 M_ASSERTPKTHDR(m); 461 462 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 463 /* Next hop */ 464 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 465 KKASSERT(mtag != NULL); 466 next_hop = m_tag_data(mtag); 467 } 468 469 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 470 /* dummynet already filtered us */ 471 ip = mtod(m, struct ip *); 472 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 473 goto iphack; 474 } 475 476 ipstat.ips_total++; 477 478 /* length checks already done in ip_mport() */ 479 KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf")); 480 ip = mtod(m, struct ip *); 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_mport() */ 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_mport() */ 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 * NB: Beware of the destination address changing (e.g. 587 * by NAT rewriting). When this happens, tell 588 * ip_forward to do the right thing. 589 */ 590 odst = ip->ip_dst; 591 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN)) 592 return; 593 if (m == NULL) /* consumed by filter */ 594 return; 595 ip = mtod(m, struct ip *); 596 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 597 using_srcrt = (odst.s_addr != ip->ip_dst.s_addr); 598 599 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 600 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 601 KKASSERT(mtag != NULL); 602 next_hop = m_tag_data(mtag); 603 } 604 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 605 ip_dn_queue(m); 606 return; 607 } 608 if (m->m_pkthdr.fw_flags & FW_MBUF_REDISPATCH) { 609 needredispatch = TRUE; 610 m->m_pkthdr.fw_flags &= ~FW_MBUF_REDISPATCH; 611 } 612 pass: 613 /* 614 * Process options and, if not destined for us, 615 * ship it on. ip_dooptions returns 1 when an 616 * error was detected (causing an icmp message 617 * to be sent and the original packet to be freed). 618 */ 619 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, next_hop)) 620 return; 621 622 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 623 * matter if it is destined to another node, or whether it is 624 * a multicast one, RSVP wants it! and prevents it from being forwarded 625 * anywhere else. Also checks if the rsvp daemon is running before 626 * grabbing the packet. 627 */ 628 if (rsvp_on && ip->ip_p == IPPROTO_RSVP) 629 goto ours; 630 631 /* 632 * Check our list of addresses, to see if the packet is for us. 633 * If we don't have any addresses, assume any unicast packet 634 * we receive might be for us (and let the upper layers deal 635 * with it). 636 */ 637 if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) && 638 !(m->m_flags & (M_MCAST | M_BCAST))) 639 goto ours; 640 641 /* 642 * Cache the destination address of the packet; this may be 643 * changed by use of 'ipfw fwd'. 644 */ 645 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst; 646 647 /* 648 * Enable a consistency check between the destination address 649 * and the arrival interface for a unicast packet (the RFC 1122 650 * strong ES model) if IP forwarding is disabled and the packet 651 * is not locally generated and the packet is not subject to 652 * 'ipfw fwd'. 653 * 654 * XXX - Checking also should be disabled if the destination 655 * address is ipnat'ed to a different interface. 656 * 657 * XXX - Checking is incompatible with IP aliases added 658 * to the loopback interface instead of the interface where 659 * the packets are received. 660 */ 661 checkif = ip_checkinterface && 662 !ipforwarding && 663 m->m_pkthdr.rcvif != NULL && 664 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) && 665 next_hop == NULL; 666 667 /* 668 * Check for exact addresses in the hash bucket. 669 */ 670 LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) { 671 ia = iac->ia; 672 673 /* 674 * If the address matches, verify that the packet 675 * arrived via the correct interface if checking is 676 * enabled. 677 */ 678 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 679 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 680 goto ours; 681 } 682 ia = NULL; 683 684 /* 685 * Check for broadcast addresses. 686 * 687 * Only accept broadcast packets that arrive via the matching 688 * interface. Reception of forwarded directed broadcasts would 689 * be handled via ip_forward() and ether_output() with the loopback 690 * into the stack for SIMPLEX interfaces handled by ether_output(). 691 */ 692 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 693 struct ifaddr_container *ifac; 694 695 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid], 696 ifa_link) { 697 struct ifaddr *ifa = ifac->ifa; 698 699 if (ifa->ifa_addr == NULL) /* shutdown/startup race */ 700 continue; 701 if (ifa->ifa_addr->sa_family != AF_INET) 702 continue; 703 ia = ifatoia(ifa); 704 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 705 pkt_dst.s_addr) 706 goto ours; 707 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 708 goto ours; 709 #ifdef BOOTP_COMPAT 710 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 711 goto ours; 712 #endif 713 } 714 } 715 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 716 struct in_multi *inm; 717 718 /* XXX Multicast is not MPSAFE yet */ 719 get_mplock(); 720 721 if (ip_mrouter != NULL) { 722 /* 723 * If we are acting as a multicast router, all 724 * incoming multicast packets are passed to the 725 * kernel-level multicast forwarding function. 726 * The packet is returned (relatively) intact; if 727 * ip_mforward() returns a non-zero value, the packet 728 * must be discarded, else it may be accepted below. 729 */ 730 if (ip_mforward != NULL && 731 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) { 732 rel_mplock(); 733 ipstat.ips_cantforward++; 734 m_freem(m); 735 return; 736 } 737 738 /* 739 * The process-level routing daemon needs to receive 740 * all multicast IGMP packets, whether or not this 741 * host belongs to their destination groups. 742 */ 743 if (ip->ip_p == IPPROTO_IGMP) { 744 rel_mplock(); 745 goto ours; 746 } 747 ipstat.ips_forward++; 748 } 749 /* 750 * See if we belong to the destination multicast group on the 751 * arrival interface. 752 */ 753 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 754 if (inm == NULL) { 755 rel_mplock(); 756 ipstat.ips_notmember++; 757 m_freem(m); 758 return; 759 } 760 761 rel_mplock(); 762 goto ours; 763 } 764 if (ip->ip_dst.s_addr == INADDR_BROADCAST) 765 goto ours; 766 if (ip->ip_dst.s_addr == INADDR_ANY) 767 goto ours; 768 769 /* 770 * FAITH(Firewall Aided Internet Translator) 771 */ 772 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 773 if (ip_keepfaith) { 774 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 775 goto ours; 776 } 777 m_freem(m); 778 return; 779 } 780 781 /* 782 * Not for us; forward if possible and desirable. 783 */ 784 if (!ipforwarding) { 785 ipstat.ips_cantforward++; 786 m_freem(m); 787 } else { 788 #ifdef IPSEC 789 /* 790 * Enforce inbound IPsec SPD. 791 */ 792 if (ipsec4_in_reject(m, NULL)) { 793 ipsecstat.in_polvio++; 794 goto bad; 795 } 796 #endif 797 #ifdef FAST_IPSEC 798 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 799 crit_enter(); 800 if (mtag != NULL) { 801 tdbi = (struct tdb_ident *)m_tag_data(mtag); 802 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 803 } else { 804 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 805 IP_FORWARDING, &error); 806 } 807 if (sp == NULL) { /* NB: can happen if error */ 808 crit_exit(); 809 /*XXX error stat???*/ 810 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 811 goto bad; 812 } 813 814 /* 815 * Check security policy against packet attributes. 816 */ 817 error = ipsec_in_reject(sp, m); 818 KEY_FREESP(&sp); 819 crit_exit(); 820 if (error) { 821 ipstat.ips_cantforward++; 822 goto bad; 823 } 824 #endif 825 ip_forward(m, using_srcrt, next_hop); 826 } 827 return; 828 829 ours: 830 831 /* 832 * IPSTEALTH: Process non-routing options only 833 * if the packet is destined for us. 834 */ 835 if (ipstealth && 836 hlen > sizeof(struct ip) && 837 ip_dooptions(m, 1, next_hop)) 838 return; 839 840 /* Count the packet in the ip address stats */ 841 if (ia != NULL) { 842 ia->ia_ifa.if_ipackets++; 843 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 844 } 845 846 /* 847 * If offset or IP_MF are set, must reassemble. 848 * Otherwise, nothing need be done. 849 * (We could look in the reassembly queue to see 850 * if the packet was previously fragmented, 851 * but it's not worth the time; just let them time out.) 852 */ 853 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 854 /* 855 * Attempt reassembly; if it succeeds, proceed. 856 * ip_reass() will return a different mbuf. 857 */ 858 m = ip_reass(m); 859 if (m == NULL) 860 return; 861 ip = mtod(m, struct ip *); 862 863 /* Get the header length of the reassembled packet */ 864 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 865 866 needredispatch = TRUE; 867 } else { 868 ip->ip_len -= hlen; 869 } 870 871 #ifdef IPSEC 872 /* 873 * enforce IPsec policy checking if we are seeing last header. 874 * note that we do not visit this with protocols with pcb layer 875 * code - like udp/tcp/raw ip. 876 */ 877 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) && 878 ipsec4_in_reject(m, NULL)) { 879 ipsecstat.in_polvio++; 880 goto bad; 881 } 882 #endif 883 #if FAST_IPSEC 884 /* 885 * enforce IPsec policy checking if we are seeing last header. 886 * note that we do not visit this with protocols with pcb layer 887 * code - like udp/tcp/raw ip. 888 */ 889 if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) { 890 /* 891 * Check if the packet has already had IPsec processing 892 * done. If so, then just pass it along. This tag gets 893 * set during AH, ESP, etc. input handling, before the 894 * packet is returned to the ip input queue for delivery. 895 */ 896 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 897 crit_enter(); 898 if (mtag != NULL) { 899 tdbi = (struct tdb_ident *)m_tag_data(mtag); 900 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 901 } else { 902 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 903 IP_FORWARDING, &error); 904 } 905 if (sp != NULL) { 906 /* 907 * Check security policy against packet attributes. 908 */ 909 error = ipsec_in_reject(sp, m); 910 KEY_FREESP(&sp); 911 } else { 912 /* XXX error stat??? */ 913 error = EINVAL; 914 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 915 goto bad; 916 } 917 crit_exit(); 918 if (error) 919 goto bad; 920 } 921 #endif /* FAST_IPSEC */ 922 923 /* 924 * NOTE: ip_len in host form and adjusted down by hlen for 925 * protocol processing. 926 */ 927 ipstat.ips_delivered++; 928 if (needredispatch) { 929 struct netmsg_packet *pmsg; 930 lwkt_port_t port; 931 932 ip->ip_off = htons(ip->ip_off); 933 ip->ip_len = htons(ip->ip_len + hlen); 934 port = ip_mport_in(&m); 935 if (port == NULL) 936 return; 937 938 pmsg = &m->m_hdr.mh_netmsg; 939 netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport, MSGF_MPSAFE, 940 transport_processing_handler); 941 pmsg->nm_packet = m; 942 pmsg->nm_netmsg.nm_lmsg.u.ms_result = hlen; 943 944 ip = mtod(m, struct ip *); 945 ip->ip_len = ntohs(ip->ip_len) - hlen; 946 ip->ip_off = ntohs(ip->ip_off); 947 lwkt_sendmsg(port, &pmsg->nm_netmsg.nm_lmsg); 948 } else { 949 transport_processing_oncpu(m, hlen, ip); 950 } 951 return; 952 953 bad: 954 m_freem(m); 955 } 956 957 /* 958 * Take incoming datagram fragment and try to reassemble it into 959 * whole datagram. If a chain for reassembly of this datagram already 960 * exists, then it is given as fp; otherwise have to make a chain. 961 */ 962 struct mbuf * 963 ip_reass(struct mbuf *m) 964 { 965 struct ip *ip = mtod(m, struct ip *); 966 struct mbuf *p = NULL, *q, *nq; 967 struct mbuf *n; 968 struct ipq *fp = NULL; 969 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 970 int i, next; 971 u_short sum; 972 973 /* If maxnipq is 0, never accept fragments. */ 974 if (maxnipq == 0) { 975 ipstat.ips_fragments++; 976 ipstat.ips_fragdropped++; 977 m_freem(m); 978 return NULL; 979 } 980 981 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 982 /* 983 * Look for queue of fragments of this datagram. 984 */ 985 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) 986 if (ip->ip_id == fp->ipq_id && 987 ip->ip_src.s_addr == fp->ipq_src.s_addr && 988 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 989 ip->ip_p == fp->ipq_p) 990 goto found; 991 992 fp = NULL; 993 994 /* 995 * Enforce upper bound on number of fragmented packets 996 * for which we attempt reassembly; 997 * If maxnipq is -1, accept all fragments without limitation. 998 */ 999 if (nipq > maxnipq && maxnipq > 0) { 1000 /* 1001 * drop something from the tail of the current queue 1002 * before proceeding further 1003 */ 1004 if (ipq[sum].prev == &ipq[sum]) { /* gak */ 1005 for (i = 0; i < IPREASS_NHASH; i++) { 1006 if (ipq[i].prev != &ipq[i]) { 1007 ipstat.ips_fragtimeout += 1008 ipq[i].prev->ipq_nfrags; 1009 ip_freef(ipq[i].prev); 1010 break; 1011 } 1012 } 1013 } else { 1014 ipstat.ips_fragtimeout += 1015 ipq[sum].prev->ipq_nfrags; 1016 ip_freef(ipq[sum].prev); 1017 } 1018 } 1019 found: 1020 /* 1021 * Adjust ip_len to not reflect header, 1022 * convert offset of this to bytes. 1023 */ 1024 ip->ip_len -= hlen; 1025 if (ip->ip_off & IP_MF) { 1026 /* 1027 * Make sure that fragments have a data length 1028 * that's a non-zero multiple of 8 bytes. 1029 */ 1030 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 1031 ipstat.ips_toosmall++; /* XXX */ 1032 m_freem(m); 1033 return NULL; 1034 } 1035 m->m_flags |= M_FRAG; 1036 } else 1037 m->m_flags &= ~M_FRAG; 1038 ip->ip_off <<= 3; 1039 1040 ipstat.ips_fragments++; 1041 m->m_pkthdr.header = ip; 1042 1043 /* 1044 * If the hardware has not done csum over this fragment 1045 * then csum_data is not valid at all. 1046 */ 1047 if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) 1048 == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) { 1049 m->m_pkthdr.csum_data = 0; 1050 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 1051 } 1052 1053 /* 1054 * Presence of header sizes in mbufs 1055 * would confuse code below. 1056 */ 1057 m->m_data += hlen; 1058 m->m_len -= hlen; 1059 1060 /* 1061 * If first fragment to arrive, create a reassembly queue. 1062 */ 1063 if (fp == NULL) { 1064 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL) 1065 goto dropfrag; 1066 insque(fp, &ipq[sum]); 1067 nipq++; 1068 fp->ipq_nfrags = 1; 1069 fp->ipq_ttl = IPFRAGTTL; 1070 fp->ipq_p = ip->ip_p; 1071 fp->ipq_id = ip->ip_id; 1072 fp->ipq_src = ip->ip_src; 1073 fp->ipq_dst = ip->ip_dst; 1074 fp->ipq_frags = m; 1075 m->m_nextpkt = NULL; 1076 goto inserted; 1077 } else { 1078 fp->ipq_nfrags++; 1079 } 1080 1081 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1082 1083 /* 1084 * Find a segment which begins after this one does. 1085 */ 1086 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1087 if (GETIP(q)->ip_off > ip->ip_off) 1088 break; 1089 1090 /* 1091 * If there is a preceding segment, it may provide some of 1092 * our data already. If so, drop the data from the incoming 1093 * segment. If it provides all of our data, drop us, otherwise 1094 * stick new segment in the proper place. 1095 * 1096 * If some of the data is dropped from the the preceding 1097 * segment, then it's checksum is invalidated. 1098 */ 1099 if (p) { 1100 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1101 if (i > 0) { 1102 if (i >= ip->ip_len) 1103 goto dropfrag; 1104 m_adj(m, i); 1105 m->m_pkthdr.csum_flags = 0; 1106 ip->ip_off += i; 1107 ip->ip_len -= i; 1108 } 1109 m->m_nextpkt = p->m_nextpkt; 1110 p->m_nextpkt = m; 1111 } else { 1112 m->m_nextpkt = fp->ipq_frags; 1113 fp->ipq_frags = m; 1114 } 1115 1116 /* 1117 * While we overlap succeeding segments trim them or, 1118 * if they are completely covered, dequeue them. 1119 */ 1120 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1121 q = nq) { 1122 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1123 if (i < GETIP(q)->ip_len) { 1124 GETIP(q)->ip_len -= i; 1125 GETIP(q)->ip_off += i; 1126 m_adj(q, i); 1127 q->m_pkthdr.csum_flags = 0; 1128 break; 1129 } 1130 nq = q->m_nextpkt; 1131 m->m_nextpkt = nq; 1132 ipstat.ips_fragdropped++; 1133 fp->ipq_nfrags--; 1134 q->m_nextpkt = NULL; 1135 m_freem(q); 1136 } 1137 1138 inserted: 1139 /* 1140 * Check for complete reassembly and perform frag per packet 1141 * limiting. 1142 * 1143 * Frag limiting is performed here so that the nth frag has 1144 * a chance to complete the packet before we drop the packet. 1145 * As a result, n+1 frags are actually allowed per packet, but 1146 * only n will ever be stored. (n = maxfragsperpacket.) 1147 * 1148 */ 1149 next = 0; 1150 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1151 if (GETIP(q)->ip_off != next) { 1152 if (fp->ipq_nfrags > maxfragsperpacket) { 1153 ipstat.ips_fragdropped += fp->ipq_nfrags; 1154 ip_freef(fp); 1155 } 1156 return (NULL); 1157 } 1158 next += GETIP(q)->ip_len; 1159 } 1160 /* Make sure the last packet didn't have the IP_MF flag */ 1161 if (p->m_flags & M_FRAG) { 1162 if (fp->ipq_nfrags > maxfragsperpacket) { 1163 ipstat.ips_fragdropped += fp->ipq_nfrags; 1164 ip_freef(fp); 1165 } 1166 return (NULL); 1167 } 1168 1169 /* 1170 * Reassembly is complete. Make sure the packet is a sane size. 1171 */ 1172 q = fp->ipq_frags; 1173 ip = GETIP(q); 1174 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 1175 ipstat.ips_toolong++; 1176 ipstat.ips_fragdropped += fp->ipq_nfrags; 1177 ip_freef(fp); 1178 return (NULL); 1179 } 1180 1181 /* 1182 * Concatenate fragments. 1183 */ 1184 m = q; 1185 n = m->m_next; 1186 m->m_next = NULL; 1187 m_cat(m, n); 1188 nq = q->m_nextpkt; 1189 q->m_nextpkt = NULL; 1190 for (q = nq; q != NULL; q = nq) { 1191 nq = q->m_nextpkt; 1192 q->m_nextpkt = NULL; 1193 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1194 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1195 m_cat(m, q); 1196 } 1197 1198 /* 1199 * Clean up the 1's complement checksum. Carry over 16 bits must 1200 * be added back. This assumes no more then 65535 packet fragments 1201 * were reassembled. A second carry can also occur (but not a third). 1202 */ 1203 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) + 1204 (m->m_pkthdr.csum_data >> 16); 1205 if (m->m_pkthdr.csum_data > 0xFFFF) 1206 m->m_pkthdr.csum_data -= 0xFFFF; 1207 1208 /* 1209 * Create header for new ip packet by 1210 * modifying header of first packet; 1211 * dequeue and discard fragment reassembly header. 1212 * Make header visible. 1213 */ 1214 ip->ip_len = next; 1215 ip->ip_src = fp->ipq_src; 1216 ip->ip_dst = fp->ipq_dst; 1217 remque(fp); 1218 nipq--; 1219 mpipe_free(&ipq_mpipe, fp); 1220 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 1221 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 1222 /* some debugging cruft by sklower, below, will go away soon */ 1223 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1224 int plen = 0; 1225 1226 for (n = m; n; n = n->m_next) 1227 plen += n->m_len; 1228 m->m_pkthdr.len = plen; 1229 } 1230 1231 ipstat.ips_reassembled++; 1232 return (m); 1233 1234 dropfrag: 1235 ipstat.ips_fragdropped++; 1236 if (fp != NULL) 1237 fp->ipq_nfrags--; 1238 m_freem(m); 1239 return (NULL); 1240 1241 #undef GETIP 1242 } 1243 1244 /* 1245 * Free a fragment reassembly header and all 1246 * associated datagrams. 1247 */ 1248 static void 1249 ip_freef(struct ipq *fp) 1250 { 1251 struct mbuf *q; 1252 1253 while (fp->ipq_frags) { 1254 q = fp->ipq_frags; 1255 fp->ipq_frags = q->m_nextpkt; 1256 q->m_nextpkt = NULL; 1257 m_freem(q); 1258 } 1259 remque(fp); 1260 mpipe_free(&ipq_mpipe, fp); 1261 nipq--; 1262 } 1263 1264 /* 1265 * IP timer processing; 1266 * if a timer expires on a reassembly 1267 * queue, discard it. 1268 */ 1269 void 1270 ip_slowtimo(void) 1271 { 1272 struct ipq *fp; 1273 int i; 1274 1275 crit_enter(); 1276 for (i = 0; i < IPREASS_NHASH; i++) { 1277 fp = ipq[i].next; 1278 if (fp == NULL) 1279 continue; 1280 while (fp != &ipq[i]) { 1281 --fp->ipq_ttl; 1282 fp = fp->next; 1283 if (fp->prev->ipq_ttl == 0) { 1284 ipstat.ips_fragtimeout += fp->prev->ipq_nfrags; 1285 ip_freef(fp->prev); 1286 } 1287 } 1288 } 1289 /* 1290 * If we are over the maximum number of fragments 1291 * (due to the limit being lowered), drain off 1292 * enough to get down to the new limit. 1293 */ 1294 if (maxnipq >= 0 && nipq > maxnipq) { 1295 for (i = 0; i < IPREASS_NHASH; i++) { 1296 while (nipq > maxnipq && 1297 (ipq[i].next != &ipq[i])) { 1298 ipstat.ips_fragdropped += 1299 ipq[i].next->ipq_nfrags; 1300 ip_freef(ipq[i].next); 1301 } 1302 } 1303 } 1304 ipflow_slowtimo(); 1305 crit_exit(); 1306 } 1307 1308 /* 1309 * Drain off all datagram fragments. 1310 */ 1311 void 1312 ip_drain(void) 1313 { 1314 int i; 1315 1316 for (i = 0; i < IPREASS_NHASH; i++) { 1317 while (ipq[i].next != &ipq[i]) { 1318 ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags; 1319 ip_freef(ipq[i].next); 1320 } 1321 } 1322 in_rtqdrain(); 1323 } 1324 1325 /* 1326 * Do option processing on a datagram, 1327 * possibly discarding it if bad options are encountered, 1328 * or forwarding it if source-routed. 1329 * The pass argument is used when operating in the IPSTEALTH 1330 * mode to tell what options to process: 1331 * [LS]SRR (pass 0) or the others (pass 1). 1332 * The reason for as many as two passes is that when doing IPSTEALTH, 1333 * non-routing options should be processed only if the packet is for us. 1334 * Returns 1 if packet has been forwarded/freed, 1335 * 0 if the packet should be processed further. 1336 */ 1337 static int 1338 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop) 1339 { 1340 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET }; 1341 struct ip *ip = mtod(m, struct ip *); 1342 u_char *cp; 1343 struct in_ifaddr *ia; 1344 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB; 1345 boolean_t forward = FALSE; 1346 struct in_addr *sin, dst; 1347 n_time ntime; 1348 1349 dst = ip->ip_dst; 1350 cp = (u_char *)(ip + 1); 1351 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1352 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1353 opt = cp[IPOPT_OPTVAL]; 1354 if (opt == IPOPT_EOL) 1355 break; 1356 if (opt == IPOPT_NOP) 1357 optlen = 1; 1358 else { 1359 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1360 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1361 goto bad; 1362 } 1363 optlen = cp[IPOPT_OLEN]; 1364 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1365 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1366 goto bad; 1367 } 1368 } 1369 switch (opt) { 1370 1371 default: 1372 break; 1373 1374 /* 1375 * Source routing with record. 1376 * Find interface with current destination address. 1377 * If none on this machine then drop if strictly routed, 1378 * or do nothing if loosely routed. 1379 * Record interface address and bring up next address 1380 * component. If strictly routed make sure next 1381 * address is on directly accessible net. 1382 */ 1383 case IPOPT_LSRR: 1384 case IPOPT_SSRR: 1385 if (ipstealth && pass > 0) 1386 break; 1387 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1388 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1389 goto bad; 1390 } 1391 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1392 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1393 goto bad; 1394 } 1395 ipaddr.sin_addr = ip->ip_dst; 1396 ia = (struct in_ifaddr *) 1397 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1398 if (ia == NULL) { 1399 if (opt == IPOPT_SSRR) { 1400 type = ICMP_UNREACH; 1401 code = ICMP_UNREACH_SRCFAIL; 1402 goto bad; 1403 } 1404 if (!ip_dosourceroute) 1405 goto nosourcerouting; 1406 /* 1407 * Loose routing, and not at next destination 1408 * yet; nothing to do except forward. 1409 */ 1410 break; 1411 } 1412 off--; /* 0 origin */ 1413 if (off > optlen - (int)sizeof(struct in_addr)) { 1414 /* 1415 * End of source route. Should be for us. 1416 */ 1417 if (!ip_acceptsourceroute) 1418 goto nosourcerouting; 1419 save_rte(m, cp, ip->ip_src); 1420 break; 1421 } 1422 if (ipstealth) 1423 goto dropit; 1424 if (!ip_dosourceroute) { 1425 if (ipforwarding) { 1426 char buf[sizeof "aaa.bbb.ccc.ddd"]; 1427 1428 /* 1429 * Acting as a router, so generate ICMP 1430 */ 1431 nosourcerouting: 1432 strcpy(buf, inet_ntoa(ip->ip_dst)); 1433 log(LOG_WARNING, 1434 "attempted source route from %s to %s\n", 1435 inet_ntoa(ip->ip_src), buf); 1436 type = ICMP_UNREACH; 1437 code = ICMP_UNREACH_SRCFAIL; 1438 goto bad; 1439 } else { 1440 /* 1441 * Not acting as a router, 1442 * so silently drop. 1443 */ 1444 dropit: 1445 ipstat.ips_cantforward++; 1446 m_freem(m); 1447 return (1); 1448 } 1449 } 1450 1451 /* 1452 * locate outgoing interface 1453 */ 1454 memcpy(&ipaddr.sin_addr, cp + off, 1455 sizeof ipaddr.sin_addr); 1456 1457 if (opt == IPOPT_SSRR) { 1458 #define INA struct in_ifaddr * 1459 #define SA struct sockaddr * 1460 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) 1461 == NULL) 1462 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1463 } else { 1464 ia = ip_rtaddr(ipaddr.sin_addr, NULL); 1465 } 1466 if (ia == NULL) { 1467 type = ICMP_UNREACH; 1468 code = ICMP_UNREACH_SRCFAIL; 1469 goto bad; 1470 } 1471 ip->ip_dst = ipaddr.sin_addr; 1472 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1473 sizeof(struct in_addr)); 1474 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1475 /* 1476 * Let ip_intr's mcast routing check handle mcast pkts 1477 */ 1478 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1479 break; 1480 1481 case IPOPT_RR: 1482 if (ipstealth && pass == 0) 1483 break; 1484 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1485 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1486 goto bad; 1487 } 1488 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1489 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1490 goto bad; 1491 } 1492 /* 1493 * If no space remains, ignore. 1494 */ 1495 off--; /* 0 origin */ 1496 if (off > optlen - (int)sizeof(struct in_addr)) 1497 break; 1498 memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1499 sizeof ipaddr.sin_addr); 1500 /* 1501 * locate outgoing interface; if we're the destination, 1502 * use the incoming interface (should be same). 1503 */ 1504 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL && 1505 (ia = ip_rtaddr(ipaddr.sin_addr, NULL)) == NULL) { 1506 type = ICMP_UNREACH; 1507 code = ICMP_UNREACH_HOST; 1508 goto bad; 1509 } 1510 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1511 sizeof(struct in_addr)); 1512 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1513 break; 1514 1515 case IPOPT_TS: 1516 if (ipstealth && pass == 0) 1517 break; 1518 code = cp - (u_char *)ip; 1519 if (optlen < 4 || optlen > 40) { 1520 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1521 goto bad; 1522 } 1523 if ((off = cp[IPOPT_OFFSET]) < 5) { 1524 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1525 goto bad; 1526 } 1527 if (off > optlen - (int)sizeof(int32_t)) { 1528 cp[IPOPT_OFFSET + 1] += (1 << 4); 1529 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1530 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1531 goto bad; 1532 } 1533 break; 1534 } 1535 off--; /* 0 origin */ 1536 sin = (struct in_addr *)(cp + off); 1537 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1538 1539 case IPOPT_TS_TSONLY: 1540 break; 1541 1542 case IPOPT_TS_TSANDADDR: 1543 if (off + sizeof(n_time) + 1544 sizeof(struct in_addr) > optlen) { 1545 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1546 goto bad; 1547 } 1548 ipaddr.sin_addr = dst; 1549 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1550 m->m_pkthdr.rcvif); 1551 if (ia == NULL) 1552 continue; 1553 memcpy(sin, &IA_SIN(ia)->sin_addr, 1554 sizeof(struct in_addr)); 1555 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1556 off += sizeof(struct in_addr); 1557 break; 1558 1559 case IPOPT_TS_PRESPEC: 1560 if (off + sizeof(n_time) + 1561 sizeof(struct in_addr) > optlen) { 1562 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1563 goto bad; 1564 } 1565 memcpy(&ipaddr.sin_addr, sin, 1566 sizeof(struct in_addr)); 1567 if (ifa_ifwithaddr((SA)&ipaddr) == NULL) 1568 continue; 1569 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1570 off += sizeof(struct in_addr); 1571 break; 1572 1573 default: 1574 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1575 goto bad; 1576 } 1577 ntime = iptime(); 1578 memcpy(cp + off, &ntime, sizeof(n_time)); 1579 cp[IPOPT_OFFSET] += sizeof(n_time); 1580 } 1581 } 1582 if (forward && ipforwarding) { 1583 ip_forward(m, TRUE, next_hop); 1584 return (1); 1585 } 1586 return (0); 1587 bad: 1588 icmp_error(m, type, code, 0, 0); 1589 ipstat.ips_badoptions++; 1590 return (1); 1591 } 1592 1593 /* 1594 * Given address of next destination (final or next hop), 1595 * return internet address info of interface to be used to get there. 1596 */ 1597 struct in_ifaddr * 1598 ip_rtaddr(struct in_addr dst, struct route *ro0) 1599 { 1600 struct route sro, *ro; 1601 struct sockaddr_in *sin; 1602 struct in_ifaddr *ia; 1603 1604 if (ro0 != NULL) { 1605 ro = ro0; 1606 } else { 1607 bzero(&sro, sizeof(sro)); 1608 ro = &sro; 1609 } 1610 1611 sin = (struct sockaddr_in *)&ro->ro_dst; 1612 1613 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) { 1614 if (ro->ro_rt != NULL) { 1615 RTFREE(ro->ro_rt); 1616 ro->ro_rt = NULL; 1617 } 1618 sin->sin_family = AF_INET; 1619 sin->sin_len = sizeof *sin; 1620 sin->sin_addr = dst; 1621 rtalloc_ign(ro, RTF_PRCLONING); 1622 } 1623 1624 if (ro->ro_rt == NULL) 1625 return (NULL); 1626 1627 ia = ifatoia(ro->ro_rt->rt_ifa); 1628 1629 if (ro == &sro) 1630 RTFREE(ro->ro_rt); 1631 return ia; 1632 } 1633 1634 /* 1635 * Save incoming source route for use in replies, 1636 * to be picked up later by ip_srcroute if the receiver is interested. 1637 */ 1638 static void 1639 save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1640 { 1641 struct m_tag *mtag; 1642 struct ip_srcrt_opt *opt; 1643 unsigned olen; 1644 1645 mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), MB_DONTWAIT); 1646 if (mtag == NULL) 1647 return; 1648 opt = m_tag_data(mtag); 1649 1650 olen = option[IPOPT_OLEN]; 1651 #ifdef DIAGNOSTIC 1652 if (ipprintfs) 1653 kprintf("save_rte: olen %d\n", olen); 1654 #endif 1655 if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) { 1656 m_tag_free(mtag); 1657 return; 1658 } 1659 bcopy(option, opt->ip_srcrt.srcopt, olen); 1660 opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1661 opt->ip_srcrt.dst = dst; 1662 m_tag_prepend(m, mtag); 1663 } 1664 1665 /* 1666 * Retrieve incoming source route for use in replies, 1667 * in the same form used by setsockopt. 1668 * The first hop is placed before the options, will be removed later. 1669 */ 1670 struct mbuf * 1671 ip_srcroute(struct mbuf *m0) 1672 { 1673 struct in_addr *p, *q; 1674 struct mbuf *m; 1675 struct m_tag *mtag; 1676 struct ip_srcrt_opt *opt; 1677 1678 if (m0 == NULL) 1679 return NULL; 1680 1681 mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL); 1682 if (mtag == NULL) 1683 return NULL; 1684 opt = m_tag_data(mtag); 1685 1686 if (opt->ip_nhops == 0) 1687 return (NULL); 1688 m = m_get(MB_DONTWAIT, MT_HEADER); 1689 if (m == NULL) 1690 return (NULL); 1691 1692 #define OPTSIZ (sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt)) 1693 1694 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1695 m->m_len = opt->ip_nhops * sizeof(struct in_addr) + 1696 sizeof(struct in_addr) + OPTSIZ; 1697 #ifdef DIAGNOSTIC 1698 if (ipprintfs) { 1699 kprintf("ip_srcroute: nhops %d mlen %d", 1700 opt->ip_nhops, m->m_len); 1701 } 1702 #endif 1703 1704 /* 1705 * First save first hop for return route 1706 */ 1707 p = &opt->ip_srcrt.route[opt->ip_nhops - 1]; 1708 *(mtod(m, struct in_addr *)) = *p--; 1709 #ifdef DIAGNOSTIC 1710 if (ipprintfs) 1711 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1712 #endif 1713 1714 /* 1715 * Copy option fields and padding (nop) to mbuf. 1716 */ 1717 opt->ip_srcrt.nop = IPOPT_NOP; 1718 opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1719 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop, 1720 OPTSIZ); 1721 q = (struct in_addr *)(mtod(m, caddr_t) + 1722 sizeof(struct in_addr) + OPTSIZ); 1723 #undef OPTSIZ 1724 /* 1725 * Record return path as an IP source route, 1726 * reversing the path (pointers are now aligned). 1727 */ 1728 while (p >= opt->ip_srcrt.route) { 1729 #ifdef DIAGNOSTIC 1730 if (ipprintfs) 1731 kprintf(" %x", ntohl(q->s_addr)); 1732 #endif 1733 *q++ = *p--; 1734 } 1735 /* 1736 * Last hop goes to final destination. 1737 */ 1738 *q = opt->ip_srcrt.dst; 1739 m_tag_delete(m0, mtag); 1740 #ifdef DIAGNOSTIC 1741 if (ipprintfs) 1742 kprintf(" %x\n", ntohl(q->s_addr)); 1743 #endif 1744 return (m); 1745 } 1746 1747 /* 1748 * Strip out IP options. 1749 */ 1750 void 1751 ip_stripoptions(struct mbuf *m) 1752 { 1753 int datalen; 1754 struct ip *ip = mtod(m, struct ip *); 1755 caddr_t opts; 1756 int optlen; 1757 1758 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1759 opts = (caddr_t)(ip + 1); 1760 datalen = m->m_len - (sizeof(struct ip) + optlen); 1761 bcopy(opts + optlen, opts, datalen); 1762 m->m_len -= optlen; 1763 if (m->m_flags & M_PKTHDR) 1764 m->m_pkthdr.len -= optlen; 1765 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1766 } 1767 1768 u_char inetctlerrmap[PRC_NCMDS] = { 1769 0, 0, 0, 0, 1770 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1771 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1772 EMSGSIZE, EHOSTUNREACH, 0, 0, 1773 0, 0, 0, 0, 1774 ENOPROTOOPT, ECONNREFUSED 1775 }; 1776 1777 /* 1778 * Forward a packet. If some error occurs return the sender 1779 * an icmp packet. Note we can't always generate a meaningful 1780 * icmp message because icmp doesn't have a large enough repertoire 1781 * of codes and types. 1782 * 1783 * If not forwarding, just drop the packet. This could be confusing 1784 * if ipforwarding was zero but some routing protocol was advancing 1785 * us as a gateway to somewhere. However, we must let the routing 1786 * protocol deal with that. 1787 * 1788 * The using_srcrt parameter indicates whether the packet is being forwarded 1789 * via a source route. 1790 */ 1791 void 1792 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop) 1793 { 1794 struct ip *ip = mtod(m, struct ip *); 1795 struct rtentry *rt; 1796 struct route fwd_ro; 1797 int error, type = 0, code = 0, destmtu = 0; 1798 struct mbuf *mcopy; 1799 n_long dest; 1800 struct in_addr pkt_dst; 1801 1802 dest = INADDR_ANY; 1803 /* 1804 * Cache the destination address of the packet; this may be 1805 * changed by use of 'ipfw fwd'. 1806 */ 1807 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst; 1808 1809 #ifdef DIAGNOSTIC 1810 if (ipprintfs) 1811 kprintf("forward: src %x dst %x ttl %x\n", 1812 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl); 1813 #endif 1814 1815 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) { 1816 ipstat.ips_cantforward++; 1817 m_freem(m); 1818 return; 1819 } 1820 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) { 1821 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1822 return; 1823 } 1824 1825 bzero(&fwd_ro, sizeof(fwd_ro)); 1826 ip_rtaddr(pkt_dst, &fwd_ro); 1827 if (fwd_ro.ro_rt == NULL) { 1828 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1829 return; 1830 } 1831 rt = fwd_ro.ro_rt; 1832 1833 /* 1834 * Save the IP header and at most 8 bytes of the payload, 1835 * in case we need to generate an ICMP message to the src. 1836 * 1837 * XXX this can be optimized a lot by saving the data in a local 1838 * buffer on the stack (72 bytes at most), and only allocating the 1839 * mbuf if really necessary. The vast majority of the packets 1840 * are forwarded without having to send an ICMP back (either 1841 * because unnecessary, or because rate limited), so we are 1842 * really we are wasting a lot of work here. 1843 * 1844 * We don't use m_copy() because it might return a reference 1845 * to a shared cluster. Both this function and ip_output() 1846 * assume exclusive access to the IP header in `m', so any 1847 * data in a cluster may change before we reach icmp_error(). 1848 */ 1849 MGETHDR(mcopy, MB_DONTWAIT, m->m_type); 1850 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, MB_DONTWAIT)) { 1851 /* 1852 * It's probably ok if the pkthdr dup fails (because 1853 * the deep copy of the tag chain failed), but for now 1854 * be conservative and just discard the copy since 1855 * code below may some day want the tags. 1856 */ 1857 m_free(mcopy); 1858 mcopy = NULL; 1859 } 1860 if (mcopy != NULL) { 1861 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8, 1862 (int)ip->ip_len); 1863 mcopy->m_pkthdr.len = mcopy->m_len; 1864 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1865 } 1866 1867 if (!ipstealth) 1868 ip->ip_ttl -= IPTTLDEC; 1869 1870 /* 1871 * If forwarding packet using same interface that it came in on, 1872 * perhaps should send a redirect to sender to shortcut a hop. 1873 * Only send redirect if source is sending directly to us, 1874 * and if packet was not source routed (or has any options). 1875 * Also, don't send redirect if forwarding using a default route 1876 * or a route modified by a redirect. 1877 */ 1878 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1879 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) && 1880 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY && 1881 ipsendredirects && !using_srcrt && next_hop == NULL) { 1882 u_long src = ntohl(ip->ip_src.s_addr); 1883 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa; 1884 1885 if (rt_ifa != NULL && 1886 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) { 1887 if (rt->rt_flags & RTF_GATEWAY) 1888 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1889 else 1890 dest = pkt_dst.s_addr; 1891 /* 1892 * Router requirements says to only send 1893 * host redirects. 1894 */ 1895 type = ICMP_REDIRECT; 1896 code = ICMP_REDIRECT_HOST; 1897 #ifdef DIAGNOSTIC 1898 if (ipprintfs) 1899 kprintf("redirect (%d) to %x\n", code, dest); 1900 #endif 1901 } 1902 } 1903 1904 error = ip_output(m, NULL, &fwd_ro, IP_FORWARDING, NULL, NULL); 1905 if (error == 0) { 1906 ipstat.ips_forward++; 1907 if (type == 0) { 1908 if (mcopy) { 1909 ipflow_create(&fwd_ro, mcopy); 1910 m_freem(mcopy); 1911 } 1912 goto done; 1913 } else { 1914 ipstat.ips_redirectsent++; 1915 } 1916 } else { 1917 ipstat.ips_cantforward++; 1918 } 1919 1920 if (mcopy == NULL) 1921 goto done; 1922 1923 /* 1924 * Send ICMP message. 1925 */ 1926 1927 switch (error) { 1928 1929 case 0: /* forwarded, but need redirect */ 1930 /* type, code set above */ 1931 break; 1932 1933 case ENETUNREACH: /* shouldn't happen, checked above */ 1934 case EHOSTUNREACH: 1935 case ENETDOWN: 1936 case EHOSTDOWN: 1937 default: 1938 type = ICMP_UNREACH; 1939 code = ICMP_UNREACH_HOST; 1940 break; 1941 1942 case EMSGSIZE: 1943 type = ICMP_UNREACH; 1944 code = ICMP_UNREACH_NEEDFRAG; 1945 #ifdef IPSEC 1946 /* 1947 * If the packet is routed over IPsec tunnel, tell the 1948 * originator the tunnel MTU. 1949 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1950 * XXX quickhack!!! 1951 */ 1952 if (fwd_ro.ro_rt != NULL) { 1953 struct secpolicy *sp = NULL; 1954 int ipsecerror; 1955 int ipsechdr; 1956 struct route *ro; 1957 1958 sp = ipsec4_getpolicybyaddr(mcopy, 1959 IPSEC_DIR_OUTBOUND, 1960 IP_FORWARDING, 1961 &ipsecerror); 1962 1963 if (sp == NULL) 1964 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 1965 else { 1966 /* count IPsec header size */ 1967 ipsechdr = ipsec4_hdrsiz(mcopy, 1968 IPSEC_DIR_OUTBOUND, 1969 NULL); 1970 1971 /* 1972 * find the correct route for outer IPv4 1973 * header, compute tunnel MTU. 1974 * 1975 */ 1976 if (sp->req != NULL && sp->req->sav != NULL && 1977 sp->req->sav->sah != NULL) { 1978 ro = &sp->req->sav->sah->sa_route; 1979 if (ro->ro_rt != NULL && 1980 ro->ro_rt->rt_ifp != NULL) { 1981 destmtu = 1982 ro->ro_rt->rt_ifp->if_mtu; 1983 destmtu -= ipsechdr; 1984 } 1985 } 1986 1987 key_freesp(sp); 1988 } 1989 } 1990 #elif FAST_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 = ipsec_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 && 2022 sp->req->sav != NULL && 2023 sp->req->sav->sah != NULL) { 2024 ro = &sp->req->sav->sah->sa_route; 2025 if (ro->ro_rt != NULL && 2026 ro->ro_rt->rt_ifp != NULL) { 2027 destmtu = 2028 ro->ro_rt->rt_ifp->if_mtu; 2029 destmtu -= ipsechdr; 2030 } 2031 } 2032 2033 KEY_FREESP(&sp); 2034 } 2035 } 2036 #else /* !IPSEC && !FAST_IPSEC */ 2037 if (fwd_ro.ro_rt != NULL) 2038 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2039 #endif /*IPSEC*/ 2040 ipstat.ips_cantfrag++; 2041 break; 2042 2043 case ENOBUFS: 2044 /* 2045 * A router should not generate ICMP_SOURCEQUENCH as 2046 * required in RFC1812 Requirements for IP Version 4 Routers. 2047 * Source quench could be a big problem under DoS attacks, 2048 * or if the underlying interface is rate-limited. 2049 * Those who need source quench packets may re-enable them 2050 * via the net.inet.ip.sendsourcequench sysctl. 2051 */ 2052 if (!ip_sendsourcequench) { 2053 m_freem(mcopy); 2054 goto done; 2055 } else { 2056 type = ICMP_SOURCEQUENCH; 2057 code = 0; 2058 } 2059 break; 2060 2061 case EACCES: /* ipfw denied packet */ 2062 m_freem(mcopy); 2063 goto done; 2064 } 2065 icmp_error(mcopy, type, code, dest, destmtu); 2066 done: 2067 if (fwd_ro.ro_rt != NULL) 2068 RTFREE(fwd_ro.ro_rt); 2069 } 2070 2071 void 2072 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2073 struct mbuf *m) 2074 { 2075 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2076 struct timeval tv; 2077 2078 microtime(&tv); 2079 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2080 SCM_TIMESTAMP, SOL_SOCKET); 2081 if (*mp) 2082 mp = &(*mp)->m_next; 2083 } 2084 if (inp->inp_flags & INP_RECVDSTADDR) { 2085 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2086 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2087 if (*mp) 2088 mp = &(*mp)->m_next; 2089 } 2090 if (inp->inp_flags & INP_RECVTTL) { 2091 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 2092 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 2093 if (*mp) 2094 mp = &(*mp)->m_next; 2095 } 2096 #ifdef notyet 2097 /* XXX 2098 * Moving these out of udp_input() made them even more broken 2099 * than they already were. 2100 */ 2101 /* options were tossed already */ 2102 if (inp->inp_flags & INP_RECVOPTS) { 2103 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2104 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2105 if (*mp) 2106 mp = &(*mp)->m_next; 2107 } 2108 /* ip_srcroute doesn't do what we want here, need to fix */ 2109 if (inp->inp_flags & INP_RECVRETOPTS) { 2110 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 2111 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2112 if (*mp) 2113 mp = &(*mp)->m_next; 2114 } 2115 #endif 2116 if (inp->inp_flags & INP_RECVIF) { 2117 struct ifnet *ifp; 2118 struct sdlbuf { 2119 struct sockaddr_dl sdl; 2120 u_char pad[32]; 2121 } sdlbuf; 2122 struct sockaddr_dl *sdp; 2123 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2124 2125 if (((ifp = m->m_pkthdr.rcvif)) && 2126 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) { 2127 sdp = IF_LLSOCKADDR(ifp); 2128 /* 2129 * Change our mind and don't try copy. 2130 */ 2131 if ((sdp->sdl_family != AF_LINK) || 2132 (sdp->sdl_len > sizeof(sdlbuf))) { 2133 goto makedummy; 2134 } 2135 bcopy(sdp, sdl2, sdp->sdl_len); 2136 } else { 2137 makedummy: 2138 sdl2->sdl_len = 2139 offsetof(struct sockaddr_dl, sdl_data[0]); 2140 sdl2->sdl_family = AF_LINK; 2141 sdl2->sdl_index = 0; 2142 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2143 } 2144 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 2145 IP_RECVIF, IPPROTO_IP); 2146 if (*mp) 2147 mp = &(*mp)->m_next; 2148 } 2149 } 2150 2151 /* 2152 * XXX these routines are called from the upper part of the kernel. 2153 * 2154 * They could also be moved to ip_mroute.c, since all the RSVP 2155 * handling is done there already. 2156 */ 2157 int 2158 ip_rsvp_init(struct socket *so) 2159 { 2160 if (so->so_type != SOCK_RAW || 2161 so->so_proto->pr_protocol != IPPROTO_RSVP) 2162 return EOPNOTSUPP; 2163 2164 if (ip_rsvpd != NULL) 2165 return EADDRINUSE; 2166 2167 ip_rsvpd = so; 2168 /* 2169 * This may seem silly, but we need to be sure we don't over-increment 2170 * the RSVP counter, in case something slips up. 2171 */ 2172 if (!ip_rsvp_on) { 2173 ip_rsvp_on = 1; 2174 rsvp_on++; 2175 } 2176 2177 return 0; 2178 } 2179 2180 int 2181 ip_rsvp_done(void) 2182 { 2183 ip_rsvpd = NULL; 2184 /* 2185 * This may seem silly, but we need to be sure we don't over-decrement 2186 * the RSVP counter, in case something slips up. 2187 */ 2188 if (ip_rsvp_on) { 2189 ip_rsvp_on = 0; 2190 rsvp_on--; 2191 } 2192 return 0; 2193 } 2194 2195 void 2196 rsvp_input(struct mbuf *m, ...) /* XXX must fixup manually */ 2197 { 2198 int off, proto; 2199 __va_list ap; 2200 2201 __va_start(ap, m); 2202 off = __va_arg(ap, int); 2203 proto = __va_arg(ap, int); 2204 __va_end(ap); 2205 2206 if (rsvp_input_p) { /* call the real one if loaded */ 2207 rsvp_input_p(m, off, proto); 2208 return; 2209 } 2210 2211 /* Can still get packets with rsvp_on = 0 if there is a local member 2212 * of the group to which the RSVP packet is addressed. But in this 2213 * case we want to throw the packet away. 2214 */ 2215 2216 if (!rsvp_on) { 2217 m_freem(m); 2218 return; 2219 } 2220 2221 if (ip_rsvpd != NULL) { 2222 rip_input(m, off, proto); 2223 return; 2224 } 2225 /* Drop the packet */ 2226 m_freem(m); 2227 } 2228