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. Neither the name of the University nor the names of its contributors 47 * may be used to endorse or promote products derived from this software 48 * without specific prior written permission. 49 * 50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 60 * SUCH DAMAGE. 61 * 62 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 63 * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $ 64 */ 65 66 #define _IP_VHL 67 68 #include "opt_bootp.h" 69 #include "opt_ipdn.h" 70 #include "opt_ipdivert.h" 71 #include "opt_ipstealth.h" 72 #include "opt_ipsec.h" 73 #include "opt_rss.h" 74 75 #include <sys/param.h> 76 #include <sys/systm.h> 77 #include <sys/mbuf.h> 78 #include <sys/malloc.h> 79 #include <sys/mpipe.h> 80 #include <sys/domain.h> 81 #include <sys/protosw.h> 82 #include <sys/socket.h> 83 #include <sys/time.h> 84 #include <sys/globaldata.h> 85 #include <sys/thread.h> 86 #include <sys/kernel.h> 87 #include <sys/syslog.h> 88 #include <sys/sysctl.h> 89 #include <sys/in_cksum.h> 90 #include <sys/lock.h> 91 92 #include <sys/mplock2.h> 93 94 #include <machine/stdarg.h> 95 96 #include <net/if.h> 97 #include <net/if_types.h> 98 #include <net/if_var.h> 99 #include <net/if_dl.h> 100 #include <net/pfil.h> 101 #include <net/route.h> 102 #include <net/netisr2.h> 103 104 #include <netinet/in.h> 105 #include <netinet/in_systm.h> 106 #include <netinet/in_var.h> 107 #include <netinet/ip.h> 108 #include <netinet/in_pcb.h> 109 #include <netinet/ip_var.h> 110 #include <netinet/ip_icmp.h> 111 #include <netinet/ip_divert.h> 112 #include <netinet/ip_flow.h> 113 114 #include <sys/thread2.h> 115 #include <sys/msgport2.h> 116 #include <net/netmsg2.h> 117 118 #include <sys/socketvar.h> 119 120 #include <net/ipfw/ip_fw.h> 121 #include <net/dummynet/ip_dummynet.h> 122 123 #ifdef IPSEC 124 #include <netinet6/ipsec.h> 125 #include <netproto/key/key.h> 126 #endif 127 128 #ifdef FAST_IPSEC 129 #include <netproto/ipsec/ipsec.h> 130 #include <netproto/ipsec/key.h> 131 #endif 132 133 int rsvp_on = 0; 134 static int ip_rsvp_on; 135 struct socket *ip_rsvpd; 136 137 int ipforwarding = 0; 138 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 139 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 140 141 static int ipsendredirects = 1; /* XXX */ 142 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 143 &ipsendredirects, 0, "Enable sending IP redirects"); 144 145 int ip_defttl = IPDEFTTL; 146 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 147 &ip_defttl, 0, "Maximum TTL on IP packets"); 148 149 static int ip_dosourceroute = 0; 150 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 151 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 152 153 static int ip_acceptsourceroute = 0; 154 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 155 CTLFLAG_RW, &ip_acceptsourceroute, 0, 156 "Enable accepting source routed IP packets"); 157 158 static int ip_keepfaith = 0; 159 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 160 &ip_keepfaith, 0, 161 "Enable packet capture for FAITH IPv4->IPv6 translator daemon"); 162 163 static int maxnipq; 164 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 165 &maxnipq, 0, 166 "Maximum number of IPv4 fragment reassembly queue entries"); 167 168 static int maxfragsperpacket; 169 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 170 &maxfragsperpacket, 0, 171 "Maximum number of IPv4 fragments allowed per packet"); 172 173 static int ip_sendsourcequench = 0; 174 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 175 &ip_sendsourcequench, 0, 176 "Enable the transmission of source quench packets"); 177 178 int ip_do_randomid = 1; 179 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 180 &ip_do_randomid, 0, 181 "Assign random ip_id values"); 182 /* 183 * XXX - Setting ip_checkinterface mostly implements the receive side of 184 * the Strong ES model described in RFC 1122, but since the routing table 185 * and transmit implementation do not implement the Strong ES model, 186 * setting this to 1 results in an odd hybrid. 187 * 188 * XXX - ip_checkinterface currently must be disabled if you use ipnat 189 * to translate the destination address to another local interface. 190 * 191 * XXX - ip_checkinterface must be disabled if you add IP aliases 192 * to the loopback interface instead of the interface where the 193 * packets for those addresses are received. 194 */ 195 static int ip_checkinterface = 0; 196 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 197 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 198 199 static u_long ip_hash_count = 0; 200 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, hash_count, CTLFLAG_RD, 201 &ip_hash_count, 0, "Number of packets hashed by IP"); 202 203 #ifdef RSS_DEBUG 204 static u_long ip_rehash_count = 0; 205 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, rehash_count, CTLFLAG_RD, 206 &ip_rehash_count, 0, "Number of packets rehashed by IP"); 207 208 static u_long ip_dispatch_fast = 0; 209 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_fast_count, CTLFLAG_RD, 210 &ip_dispatch_fast, 0, "Number of packets handled on current CPU"); 211 212 static u_long ip_dispatch_slow = 0; 213 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_slow_count, CTLFLAG_RD, 214 &ip_dispatch_slow, 0, "Number of packets messaged to another CPU"); 215 #endif 216 217 #ifdef DIAGNOSTIC 218 static int ipprintfs = 0; 219 #endif 220 221 extern struct domain inetdomain; 222 extern struct protosw inetsw[]; 223 u_char ip_protox[IPPROTO_MAX]; 224 struct in_ifaddrhead in_ifaddrheads[MAXCPU]; /* first inet address */ 225 struct in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU]; 226 /* inet addr hash table */ 227 u_long in_ifaddrhmask; /* mask for hash table */ 228 229 static struct mbuf *ipforward_mtemp[MAXCPU]; 230 231 struct ip_stats ipstats_percpu[MAXCPU] __cachealign; 232 233 static int 234 sysctl_ipstats(SYSCTL_HANDLER_ARGS) 235 { 236 int cpu, error = 0; 237 238 for (cpu = 0; cpu < ncpus; ++cpu) { 239 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu], 240 sizeof(struct ip_stats)))) 241 break; 242 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu], 243 sizeof(struct ip_stats)))) 244 break; 245 } 246 247 return (error); 248 } 249 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW), 250 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics"); 251 252 /* Packet reassembly stuff */ 253 #define IPREASS_NHASH_LOG2 6 254 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 255 #define IPREASS_HMASK (IPREASS_NHASH - 1) 256 #define IPREASS_HASH(x,y) \ 257 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 258 259 TAILQ_HEAD(ipqhead, ipq); 260 struct ipfrag_queue { 261 int nipq; 262 int timeo_inprog; 263 struct netmsg_base timeo_netmsg; 264 struct netmsg_base drain_netmsg; 265 struct ipqhead ipq[IPREASS_NHASH]; 266 } __cachealign; 267 268 static struct ipfrag_queue ipfrag_queue_pcpu[MAXCPU]; 269 270 #ifdef IPCTL_DEFMTU 271 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 272 &ip_mtu, 0, "Default MTU"); 273 #endif 274 275 #ifdef IPSTEALTH 276 static int ipstealth = 0; 277 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, ""); 278 #else 279 static const int ipstealth = 0; 280 #endif 281 282 struct mbuf *(*ip_divert_p)(struct mbuf *, int, int); 283 284 struct pfil_head inet_pfil_hook; 285 286 /* 287 * struct ip_srcrt_opt is used to store packet state while it travels 288 * through the stack. 289 * 290 * XXX Note that the code even makes assumptions on the size and 291 * alignment of fields inside struct ip_srcrt so e.g. adding some 292 * fields will break the code. This needs to be fixed. 293 * 294 * We need to save the IP options in case a protocol wants to respond 295 * to an incoming packet over the same route if the packet got here 296 * using IP source routing. This allows connection establishment and 297 * maintenance when the remote end is on a network that is not known 298 * to us. 299 */ 300 struct ip_srcrt { 301 struct in_addr dst; /* final destination */ 302 char nop; /* one NOP to align */ 303 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 304 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 305 }; 306 307 struct ip_srcrt_opt { 308 int ip_nhops; 309 struct ip_srcrt ip_srcrt; 310 }; 311 312 #define IPFRAG_MPIPE_MAX 4096 313 #define MAXIPFRAG_MIN ((IPFRAG_MPIPE_MAX * 2) / 256) 314 315 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management"); 316 static struct malloc_pipe ipq_mpipe; 317 318 static void save_rte(struct mbuf *, u_char *, struct in_addr); 319 static int ip_dooptions(struct mbuf *m, int, struct sockaddr_in *); 320 static void ip_freef(struct ipfrag_queue *, struct ipqhead *, 321 struct ipq *); 322 static void ip_input_handler(netmsg_t); 323 324 static void ipfrag_timeo_dispatch(netmsg_t); 325 static void ipfrag_drain_dispatch(netmsg_t); 326 327 /* 328 * IP initialization: fill in IP protocol switch table. 329 * All protocols not implemented in kernel go to raw IP protocol handler. 330 */ 331 void 332 ip_init(void) 333 { 334 struct protosw *pr; 335 int cpu, i; 336 337 /* 338 * Make sure we can handle a reasonable number of fragments but 339 * cap it at IPFRAG_MPIPE_MAX. 340 */ 341 mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq), 342 IFQ_MAXLEN, IPFRAG_MPIPE_MAX, 0, NULL, NULL, NULL); 343 for (cpu = 0; cpu < ncpus; ++cpu) { 344 TAILQ_INIT(&in_ifaddrheads[cpu]); 345 in_ifaddrhashtbls[cpu] = 346 hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 347 } 348 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 349 if (pr == NULL) 350 panic("ip_init"); 351 for (i = 0; i < IPPROTO_MAX; i++) 352 ip_protox[i] = pr - inetsw; 353 for (pr = inetdomain.dom_protosw; 354 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 355 if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol) { 356 if (pr->pr_protocol != IPPROTO_RAW) 357 ip_protox[pr->pr_protocol] = pr - inetsw; 358 } 359 } 360 361 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 362 inet_pfil_hook.ph_af = AF_INET; 363 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) { 364 kprintf("%s: WARNING: unable to register pfil hook, " 365 "error %d\n", __func__, i); 366 } 367 368 maxnipq = (nmbclusters / 32) / ncpus2; 369 if (maxnipq < MAXIPFRAG_MIN) 370 maxnipq = MAXIPFRAG_MIN; 371 maxfragsperpacket = 16; 372 373 ip_id = time_second & 0xffff; /* time_second survives reboots */ 374 375 for (cpu = 0; cpu < ncpus; ++cpu) { 376 /* 377 * Initialize IP statistics counters for each CPU. 378 */ 379 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats)); 380 381 /* 382 * Preallocate mbuf template for forwarding 383 */ 384 MGETHDR(ipforward_mtemp[cpu], M_WAITOK, MT_DATA); 385 386 /* 387 * Initialize per-cpu ip fragments queues 388 */ 389 for (i = 0; i < IPREASS_NHASH; i++) { 390 struct ipfrag_queue *fragq = &ipfrag_queue_pcpu[cpu]; 391 392 TAILQ_INIT(&fragq->ipq[i]); 393 netmsg_init(&fragq->timeo_netmsg, NULL, 394 &netisr_adone_rport, MSGF_PRIORITY, 395 ipfrag_timeo_dispatch); 396 netmsg_init(&fragq->drain_netmsg, NULL, 397 &netisr_adone_rport, MSGF_PRIORITY, 398 ipfrag_drain_dispatch); 399 } 400 } 401 402 netisr_register(NETISR_IP, ip_input_handler, ip_hashfn); 403 netisr_register_hashcheck(NETISR_IP, ip_hashcheck); 404 } 405 406 /* Do transport protocol processing. */ 407 static void 408 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip) 409 { 410 const struct protosw *pr = &inetsw[ip_protox[ip->ip_p]]; 411 412 /* 413 * Switch out to protocol's input routine. 414 */ 415 PR_GET_MPLOCK(pr); 416 pr->pr_input(&m, &hlen, ip->ip_p); 417 PR_REL_MPLOCK(pr); 418 } 419 420 static void 421 transport_processing_handler(netmsg_t msg) 422 { 423 struct netmsg_packet *pmsg = &msg->packet; 424 struct ip *ip; 425 int hlen; 426 427 ip = mtod(pmsg->nm_packet, struct ip *); 428 hlen = pmsg->base.lmsg.u.ms_result; 429 430 transport_processing_oncpu(pmsg->nm_packet, hlen, ip); 431 /* msg was embedded in the mbuf, do not reply! */ 432 } 433 434 static void 435 ip_input_handler(netmsg_t msg) 436 { 437 ip_input(msg->packet.nm_packet); 438 /* msg was embedded in the mbuf, do not reply! */ 439 } 440 441 /* 442 * IP input routine. Checksum and byte swap header. If fragmented 443 * try to reassemble. Process options. Pass to next level. 444 */ 445 void 446 ip_input(struct mbuf *m) 447 { 448 struct ip *ip; 449 struct in_ifaddr *ia = NULL; 450 struct in_ifaddr_container *iac; 451 int hlen, checkif; 452 u_short sum; 453 struct in_addr pkt_dst; 454 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */ 455 struct in_addr odst; /* original dst address(NAT) */ 456 struct m_tag *mtag; 457 struct sockaddr_in *next_hop = NULL; 458 lwkt_port_t port; 459 #ifdef FAST_IPSEC 460 struct tdb_ident *tdbi; 461 struct secpolicy *sp; 462 int error; 463 #endif 464 465 M_ASSERTPKTHDR(m); 466 467 /* 468 * This routine is called from numerous places which may not have 469 * characterized the packet. 470 */ 471 ip = mtod(m, struct ip *); 472 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || 473 ntohs(ip->ip_off) & (IP_MF | IP_OFFMASK)) { 474 /* 475 * Force hash recalculation for fragments and multicast 476 * packets; hardware may not do it correctly. 477 * XXX add flag to indicate the hash is from hardware 478 */ 479 m->m_flags &= ~M_HASH; 480 } 481 if ((m->m_flags & M_HASH) == 0) { 482 ip_hashfn(&m, 0); 483 if (m == NULL) 484 return; 485 KKASSERT(m->m_flags & M_HASH); 486 487 if (&curthread->td_msgport != 488 netisr_hashport(m->m_pkthdr.hash)) { 489 netisr_queue(NETISR_IP, m); 490 /* Requeued to other netisr msgport; done */ 491 return; 492 } 493 494 /* mbuf could have been changed */ 495 ip = mtod(m, struct ip *); 496 } 497 498 /* 499 * Pull out certain tags 500 */ 501 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 502 /* Next hop */ 503 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 504 KKASSERT(mtag != NULL); 505 next_hop = m_tag_data(mtag); 506 } 507 508 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 509 /* dummynet already filtered us */ 510 ip = mtod(m, struct ip *); 511 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 512 goto iphack; 513 } 514 515 ipstat.ips_total++; 516 517 /* length checks already done in ip_hashfn() */ 518 KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf")); 519 520 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 521 ipstat.ips_badvers++; 522 goto bad; 523 } 524 525 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 526 /* length checks already done in ip_hashfn() */ 527 KASSERT(hlen >= sizeof(struct ip), ("IP header len too small")); 528 KASSERT(m->m_len >= hlen, ("complete IP header not in one mbuf")); 529 530 /* 127/8 must not appear on wire - RFC1122 */ 531 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 532 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 533 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) { 534 ipstat.ips_badaddr++; 535 goto bad; 536 } 537 } 538 539 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 540 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 541 } else { 542 if (hlen == sizeof(struct ip)) 543 sum = in_cksum_hdr(ip); 544 else 545 sum = in_cksum(m, hlen); 546 } 547 if (sum != 0) { 548 ipstat.ips_badsum++; 549 goto bad; 550 } 551 552 #ifdef ALTQ 553 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 554 /* packet is dropped by traffic conditioner */ 555 return; 556 } 557 #endif 558 /* 559 * Convert fields to host representation. 560 */ 561 ip->ip_len = ntohs(ip->ip_len); 562 ip->ip_off = ntohs(ip->ip_off); 563 564 /* length checks already done in ip_hashfn() */ 565 KASSERT(ip->ip_len >= hlen, ("total length less then header length")); 566 KASSERT(m->m_pkthdr.len >= ip->ip_len, ("mbuf too short")); 567 568 /* 569 * Trim mbufs if longer than the IP header would have us expect. 570 */ 571 if (m->m_pkthdr.len > ip->ip_len) { 572 if (m->m_len == m->m_pkthdr.len) { 573 m->m_len = ip->ip_len; 574 m->m_pkthdr.len = ip->ip_len; 575 } else { 576 m_adj(m, ip->ip_len - m->m_pkthdr.len); 577 } 578 } 579 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF) 580 /* 581 * Bypass packet filtering for packets from a tunnel (gif). 582 */ 583 if (ipsec_gethist(m, NULL)) 584 goto pass; 585 #endif 586 587 /* 588 * IpHack's section. 589 * Right now when no processing on packet has done 590 * and it is still fresh out of network we do our black 591 * deals with it. 592 * - Firewall: deny/allow/divert 593 * - Xlate: translate packet's addr/port (NAT). 594 * - Pipe: pass pkt through dummynet. 595 * - Wrap: fake packet's addr/port <unimpl.> 596 * - Encapsulate: put it in another IP and send out. <unimp.> 597 */ 598 599 iphack: 600 /* 601 * If we've been forwarded from the output side, then 602 * skip the firewall a second time 603 */ 604 if (next_hop != NULL) 605 goto ours; 606 607 /* No pfil hooks */ 608 if (!pfil_has_hooks(&inet_pfil_hook)) { 609 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 610 /* 611 * Strip dummynet tags from stranded packets 612 */ 613 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 614 KKASSERT(mtag != NULL); 615 m_tag_delete(m, mtag); 616 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 617 } 618 goto pass; 619 } 620 621 /* 622 * Run through list of hooks for input packets. 623 * 624 * NOTE! If the packet is rewritten pf/ipfw/whoever must 625 * clear M_HASH. 626 */ 627 odst = ip->ip_dst; 628 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN)) 629 return; 630 if (m == NULL) /* consumed by filter */ 631 return; 632 ip = mtod(m, struct ip *); 633 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 634 using_srcrt = (odst.s_addr != ip->ip_dst.s_addr); 635 636 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 637 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 638 KKASSERT(mtag != NULL); 639 next_hop = m_tag_data(mtag); 640 } 641 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 642 ip_dn_queue(m); 643 return; 644 } 645 if (m->m_pkthdr.fw_flags & FW_MBUF_REDISPATCH) { 646 m->m_pkthdr.fw_flags &= ~FW_MBUF_REDISPATCH; 647 } 648 pass: 649 /* 650 * Process options and, if not destined for us, 651 * ship it on. ip_dooptions returns 1 when an 652 * error was detected (causing an icmp message 653 * to be sent and the original packet to be freed). 654 */ 655 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, next_hop)) 656 return; 657 658 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 659 * matter if it is destined to another node, or whether it is 660 * a multicast one, RSVP wants it! and prevents it from being forwarded 661 * anywhere else. Also checks if the rsvp daemon is running before 662 * grabbing the packet. 663 */ 664 if (rsvp_on && ip->ip_p == IPPROTO_RSVP) 665 goto ours; 666 667 /* 668 * Check our list of addresses, to see if the packet is for us. 669 * If we don't have any addresses, assume any unicast packet 670 * we receive might be for us (and let the upper layers deal 671 * with it). 672 */ 673 if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) && 674 !(m->m_flags & (M_MCAST | M_BCAST))) 675 goto ours; 676 677 /* 678 * Cache the destination address of the packet; this may be 679 * changed by use of 'ipfw fwd'. 680 */ 681 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst; 682 683 /* 684 * Enable a consistency check between the destination address 685 * and the arrival interface for a unicast packet (the RFC 1122 686 * strong ES model) if IP forwarding is disabled and the packet 687 * is not locally generated and the packet is not subject to 688 * 'ipfw fwd'. 689 * 690 * XXX - Checking also should be disabled if the destination 691 * address is ipnat'ed to a different interface. 692 * 693 * XXX - Checking is incompatible with IP aliases added 694 * to the loopback interface instead of the interface where 695 * the packets are received. 696 */ 697 checkif = ip_checkinterface && 698 !ipforwarding && 699 m->m_pkthdr.rcvif != NULL && 700 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) && 701 next_hop == NULL; 702 703 /* 704 * Check for exact addresses in the hash bucket. 705 */ 706 LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) { 707 ia = iac->ia; 708 709 /* 710 * If the address matches, verify that the packet 711 * arrived via the correct interface if checking is 712 * enabled. 713 */ 714 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 715 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 716 goto ours; 717 } 718 ia = NULL; 719 720 /* 721 * Check for broadcast addresses. 722 * 723 * Only accept broadcast packets that arrive via the matching 724 * interface. Reception of forwarded directed broadcasts would 725 * be handled via ip_forward() and ether_output() with the loopback 726 * into the stack for SIMPLEX interfaces handled by ether_output(). 727 */ 728 if (m->m_pkthdr.rcvif != NULL && 729 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 730 struct ifaddr_container *ifac; 731 732 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid], 733 ifa_link) { 734 struct ifaddr *ifa = ifac->ifa; 735 736 if (ifa->ifa_addr == NULL) /* shutdown/startup race */ 737 continue; 738 if (ifa->ifa_addr->sa_family != AF_INET) 739 continue; 740 ia = ifatoia(ifa); 741 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 742 pkt_dst.s_addr) 743 goto ours; 744 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 745 goto ours; 746 #ifdef BOOTP_COMPAT 747 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 748 goto ours; 749 #endif 750 } 751 } 752 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 753 struct in_multi *inm; 754 755 if (ip_mrouter != NULL) { 756 /* XXX Multicast routing is not MPSAFE yet */ 757 get_mplock(); 758 759 /* 760 * If we are acting as a multicast router, all 761 * incoming multicast packets are passed to the 762 * kernel-level multicast forwarding function. 763 * The packet is returned (relatively) intact; if 764 * ip_mforward() returns a non-zero value, the packet 765 * must be discarded, else it may be accepted below. 766 */ 767 if (ip_mforward != NULL && 768 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) { 769 rel_mplock(); 770 ipstat.ips_cantforward++; 771 m_freem(m); 772 return; 773 } 774 775 rel_mplock(); 776 777 /* 778 * The process-level routing daemon needs to receive 779 * all multicast IGMP packets, whether or not this 780 * host belongs to their destination groups. 781 */ 782 if (ip->ip_p == IPPROTO_IGMP) 783 goto ours; 784 ipstat.ips_forward++; 785 } 786 /* 787 * See if we belong to the destination multicast group on the 788 * arrival interface. 789 */ 790 inm = IN_LOOKUP_MULTI(&ip->ip_dst, m->m_pkthdr.rcvif); 791 if (inm == NULL) { 792 ipstat.ips_notmember++; 793 m_freem(m); 794 return; 795 } 796 goto ours; 797 } 798 if (ip->ip_dst.s_addr == INADDR_BROADCAST) 799 goto ours; 800 if (ip->ip_dst.s_addr == INADDR_ANY) 801 goto ours; 802 803 /* 804 * FAITH(Firewall Aided Internet Translator) 805 */ 806 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 807 if (ip_keepfaith) { 808 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 809 goto ours; 810 } 811 m_freem(m); 812 return; 813 } 814 815 /* 816 * Not for us; forward if possible and desirable. 817 */ 818 if (!ipforwarding) { 819 ipstat.ips_cantforward++; 820 m_freem(m); 821 } else { 822 #ifdef IPSEC 823 /* 824 * Enforce inbound IPsec SPD. 825 */ 826 if (ipsec4_in_reject(m, NULL)) { 827 ipsecstat.in_polvio++; 828 goto bad; 829 } 830 #endif 831 #ifdef FAST_IPSEC 832 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 833 crit_enter(); 834 if (mtag != NULL) { 835 tdbi = (struct tdb_ident *)m_tag_data(mtag); 836 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 837 } else { 838 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 839 IP_FORWARDING, &error); 840 } 841 if (sp == NULL) { /* NB: can happen if error */ 842 crit_exit(); 843 /*XXX error stat???*/ 844 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 845 goto bad; 846 } 847 848 /* 849 * Check security policy against packet attributes. 850 */ 851 error = ipsec_in_reject(sp, m); 852 KEY_FREESP(&sp); 853 crit_exit(); 854 if (error) { 855 ipstat.ips_cantforward++; 856 goto bad; 857 } 858 #endif 859 ip_forward(m, using_srcrt, next_hop); 860 } 861 return; 862 863 ours: 864 865 /* 866 * IPSTEALTH: Process non-routing options only 867 * if the packet is destined for us. 868 */ 869 if (ipstealth && 870 hlen > sizeof(struct ip) && 871 ip_dooptions(m, 1, next_hop)) 872 return; 873 874 /* Count the packet in the ip address stats */ 875 if (ia != NULL) { 876 IFA_STAT_INC(&ia->ia_ifa, ipackets, 1); 877 IFA_STAT_INC(&ia->ia_ifa, ibytes, m->m_pkthdr.len); 878 } 879 880 /* 881 * If offset or IP_MF are set, must reassemble. 882 * Otherwise, nothing need be done. 883 * (We could look in the reassembly queue to see 884 * if the packet was previously fragmented, 885 * but it's not worth the time; just let them time out.) 886 */ 887 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 888 /* 889 * Attempt reassembly; if it succeeds, proceed. ip_reass() 890 * will return a different mbuf. 891 * 892 * NOTE: ip_reass() returns m with M_HASH cleared to force 893 * us to recharacterize the packet. 894 */ 895 m = ip_reass(m); 896 if (m == NULL) 897 return; 898 ip = mtod(m, struct ip *); 899 900 /* Get the header length of the reassembled packet */ 901 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 902 } else { 903 ip->ip_len -= hlen; 904 } 905 906 #ifdef IPSEC 907 /* 908 * enforce IPsec policy checking if we are seeing last header. 909 * note that we do not visit this with protocols with pcb layer 910 * code - like udp/tcp/raw ip. 911 */ 912 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) && 913 ipsec4_in_reject(m, NULL)) { 914 ipsecstat.in_polvio++; 915 goto bad; 916 } 917 #endif 918 #ifdef FAST_IPSEC 919 /* 920 * enforce IPsec policy checking if we are seeing last header. 921 * note that we do not visit this with protocols with pcb layer 922 * code - like udp/tcp/raw ip. 923 */ 924 if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) { 925 /* 926 * Check if the packet has already had IPsec processing 927 * done. If so, then just pass it along. This tag gets 928 * set during AH, ESP, etc. input handling, before the 929 * packet is returned to the ip input queue for delivery. 930 */ 931 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 932 crit_enter(); 933 if (mtag != NULL) { 934 tdbi = (struct tdb_ident *)m_tag_data(mtag); 935 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 936 } else { 937 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 938 IP_FORWARDING, &error); 939 } 940 if (sp != NULL) { 941 /* 942 * Check security policy against packet attributes. 943 */ 944 error = ipsec_in_reject(sp, m); 945 KEY_FREESP(&sp); 946 } else { 947 /* XXX error stat??? */ 948 error = EINVAL; 949 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 950 crit_exit(); 951 goto bad; 952 } 953 crit_exit(); 954 if (error) 955 goto bad; 956 } 957 #endif /* FAST_IPSEC */ 958 959 /* 960 * We must forward the packet to the correct protocol thread if 961 * we are not already in it. 962 * 963 * NOTE: ip_len is now in host form. ip_len is not adjusted 964 * further for protocol processing, instead we pass hlen 965 * to the protosw and let it deal with it. 966 */ 967 ipstat.ips_delivered++; 968 969 if ((m->m_flags & M_HASH) == 0) { 970 #ifdef RSS_DEBUG 971 atomic_add_long(&ip_rehash_count, 1); 972 #endif 973 ip->ip_len = htons(ip->ip_len + hlen); 974 ip->ip_off = htons(ip->ip_off); 975 976 ip_hashfn(&m, 0); 977 if (m == NULL) 978 return; 979 980 ip = mtod(m, struct ip *); 981 ip->ip_len = ntohs(ip->ip_len) - hlen; 982 ip->ip_off = ntohs(ip->ip_off); 983 KKASSERT(m->m_flags & M_HASH); 984 } 985 port = netisr_hashport(m->m_pkthdr.hash); 986 987 if (port != &curthread->td_msgport) { 988 struct netmsg_packet *pmsg; 989 990 #ifdef RSS_DEBUG 991 atomic_add_long(&ip_dispatch_slow, 1); 992 #endif 993 994 pmsg = &m->m_hdr.mh_netmsg; 995 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 996 0, transport_processing_handler); 997 pmsg->nm_packet = m; 998 pmsg->base.lmsg.u.ms_result = hlen; 999 lwkt_sendmsg(port, &pmsg->base.lmsg); 1000 } else { 1001 #ifdef RSS_DEBUG 1002 atomic_add_long(&ip_dispatch_fast, 1); 1003 #endif 1004 transport_processing_oncpu(m, hlen, ip); 1005 } 1006 return; 1007 1008 bad: 1009 m_freem(m); 1010 } 1011 1012 /* 1013 * Take incoming datagram fragment and try to reassemble it into 1014 * whole datagram. If a chain for reassembly of this datagram already 1015 * exists, then it is given as fp; otherwise have to make a chain. 1016 */ 1017 struct mbuf * 1018 ip_reass(struct mbuf *m) 1019 { 1020 struct ipfrag_queue *fragq = &ipfrag_queue_pcpu[mycpuid]; 1021 struct ip *ip = mtod(m, struct ip *); 1022 struct mbuf *p = NULL, *q, *nq; 1023 struct mbuf *n; 1024 struct ipq *fp = NULL; 1025 struct ipqhead *head; 1026 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 1027 int i, next; 1028 u_short sum; 1029 1030 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 1031 if (maxnipq == 0 || maxfragsperpacket == 0) { 1032 ipstat.ips_fragments++; 1033 ipstat.ips_fragdropped++; 1034 m_freem(m); 1035 return NULL; 1036 } 1037 1038 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 1039 /* 1040 * Look for queue of fragments of this datagram. 1041 */ 1042 head = &fragq->ipq[sum]; 1043 TAILQ_FOREACH(fp, head, ipq_list) { 1044 if (ip->ip_id == fp->ipq_id && 1045 ip->ip_src.s_addr == fp->ipq_src.s_addr && 1046 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 1047 ip->ip_p == fp->ipq_p) 1048 goto found; 1049 } 1050 1051 fp = NULL; 1052 1053 /* 1054 * Enforce upper bound on number of fragmented packets 1055 * for which we attempt reassembly; 1056 * If maxnipq is -1, accept all fragments without limitation. 1057 */ 1058 if (fragq->nipq > maxnipq && maxnipq > 0) { 1059 /* 1060 * drop something from the tail of the current queue 1061 * before proceeding further 1062 */ 1063 struct ipq *q = TAILQ_LAST(head, ipqhead); 1064 if (q == NULL) { 1065 /* 1066 * The current queue is empty, 1067 * so drop from one of the others. 1068 */ 1069 for (i = 0; i < IPREASS_NHASH; i++) { 1070 struct ipq *r = TAILQ_LAST(&fragq->ipq[i], 1071 ipqhead); 1072 if (r) { 1073 ipstat.ips_fragtimeout += r->ipq_nfrags; 1074 ip_freef(fragq, &fragq->ipq[i], r); 1075 break; 1076 } 1077 } 1078 } else { 1079 ipstat.ips_fragtimeout += q->ipq_nfrags; 1080 ip_freef(fragq, head, q); 1081 } 1082 } 1083 found: 1084 /* 1085 * Adjust ip_len to not reflect header, 1086 * convert offset of this to bytes. 1087 */ 1088 ip->ip_len -= hlen; 1089 if (ip->ip_off & IP_MF) { 1090 /* 1091 * Make sure that fragments have a data length 1092 * that's a non-zero multiple of 8 bytes. 1093 */ 1094 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 1095 ipstat.ips_toosmall++; /* XXX */ 1096 m_freem(m); 1097 goto done; 1098 } 1099 m->m_flags |= M_FRAG; 1100 } else { 1101 m->m_flags &= ~M_FRAG; 1102 } 1103 ip->ip_off <<= 3; 1104 1105 ipstat.ips_fragments++; 1106 m->m_pkthdr.header = ip; 1107 1108 /* 1109 * If the hardware has not done csum over this fragment 1110 * then csum_data is not valid at all. 1111 */ 1112 if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) 1113 == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) { 1114 m->m_pkthdr.csum_data = 0; 1115 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 1116 } 1117 1118 /* 1119 * Presence of header sizes in mbufs 1120 * would confuse code below. 1121 */ 1122 m->m_data += hlen; 1123 m->m_len -= hlen; 1124 1125 /* 1126 * If first fragment to arrive, create a reassembly queue. 1127 */ 1128 if (fp == NULL) { 1129 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL) 1130 goto dropfrag; 1131 TAILQ_INSERT_HEAD(head, fp, ipq_list); 1132 fragq->nipq++; 1133 fp->ipq_nfrags = 1; 1134 fp->ipq_ttl = IPFRAGTTL; 1135 fp->ipq_p = ip->ip_p; 1136 fp->ipq_id = ip->ip_id; 1137 fp->ipq_src = ip->ip_src; 1138 fp->ipq_dst = ip->ip_dst; 1139 fp->ipq_frags = m; 1140 m->m_nextpkt = NULL; 1141 goto inserted; 1142 } 1143 fp->ipq_nfrags++; 1144 1145 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1146 1147 /* 1148 * Find a segment which begins after this one does. 1149 */ 1150 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1151 if (GETIP(q)->ip_off > ip->ip_off) 1152 break; 1153 } 1154 1155 /* 1156 * If there is a preceding segment, it may provide some of 1157 * our data already. If so, drop the data from the incoming 1158 * segment. If it provides all of our data, drop us, otherwise 1159 * stick new segment in the proper place. 1160 * 1161 * If some of the data is dropped from the the preceding 1162 * segment, then it's checksum is invalidated. 1163 */ 1164 if (p) { 1165 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1166 if (i > 0) { 1167 if (i >= ip->ip_len) 1168 goto dropfrag; 1169 m_adj(m, i); 1170 m->m_pkthdr.csum_flags = 0; 1171 ip->ip_off += i; 1172 ip->ip_len -= i; 1173 } 1174 m->m_nextpkt = p->m_nextpkt; 1175 p->m_nextpkt = m; 1176 } else { 1177 m->m_nextpkt = fp->ipq_frags; 1178 fp->ipq_frags = m; 1179 } 1180 1181 /* 1182 * While we overlap succeeding segments trim them or, 1183 * if they are completely covered, dequeue them. 1184 */ 1185 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1186 q = nq) { 1187 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1188 if (i < GETIP(q)->ip_len) { 1189 GETIP(q)->ip_len -= i; 1190 GETIP(q)->ip_off += i; 1191 m_adj(q, i); 1192 q->m_pkthdr.csum_flags = 0; 1193 break; 1194 } 1195 nq = q->m_nextpkt; 1196 m->m_nextpkt = nq; 1197 ipstat.ips_fragdropped++; 1198 fp->ipq_nfrags--; 1199 q->m_nextpkt = NULL; 1200 m_freem(q); 1201 } 1202 1203 inserted: 1204 /* 1205 * Check for complete reassembly and perform frag per packet 1206 * limiting. 1207 * 1208 * Frag limiting is performed here so that the nth frag has 1209 * a chance to complete the packet before we drop the packet. 1210 * As a result, n+1 frags are actually allowed per packet, but 1211 * only n will ever be stored. (n = maxfragsperpacket.) 1212 * 1213 */ 1214 next = 0; 1215 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1216 if (GETIP(q)->ip_off != next) { 1217 if (fp->ipq_nfrags > maxfragsperpacket) { 1218 ipstat.ips_fragdropped += fp->ipq_nfrags; 1219 ip_freef(fragq, head, fp); 1220 } 1221 goto done; 1222 } 1223 next += GETIP(q)->ip_len; 1224 } 1225 /* Make sure the last packet didn't have the IP_MF flag */ 1226 if (p->m_flags & M_FRAG) { 1227 if (fp->ipq_nfrags > maxfragsperpacket) { 1228 ipstat.ips_fragdropped += fp->ipq_nfrags; 1229 ip_freef(fragq, head, fp); 1230 } 1231 goto done; 1232 } 1233 1234 /* 1235 * Reassembly is complete. Make sure the packet is a sane size. 1236 */ 1237 q = fp->ipq_frags; 1238 ip = GETIP(q); 1239 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 1240 ipstat.ips_toolong++; 1241 ipstat.ips_fragdropped += fp->ipq_nfrags; 1242 ip_freef(fragq, head, fp); 1243 goto done; 1244 } 1245 1246 /* 1247 * Concatenate fragments. 1248 */ 1249 m = q; 1250 n = m->m_next; 1251 m->m_next = NULL; 1252 m_cat(m, n); 1253 nq = q->m_nextpkt; 1254 q->m_nextpkt = NULL; 1255 for (q = nq; q != NULL; q = nq) { 1256 nq = q->m_nextpkt; 1257 q->m_nextpkt = NULL; 1258 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1259 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1260 m_cat(m, q); 1261 } 1262 1263 /* 1264 * Clean up the 1's complement checksum. Carry over 16 bits must 1265 * be added back. This assumes no more then 65535 packet fragments 1266 * were reassembled. A second carry can also occur (but not a third). 1267 */ 1268 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) + 1269 (m->m_pkthdr.csum_data >> 16); 1270 if (m->m_pkthdr.csum_data > 0xFFFF) 1271 m->m_pkthdr.csum_data -= 0xFFFF; 1272 1273 /* 1274 * Create header for new ip packet by 1275 * modifying header of first packet; 1276 * dequeue and discard fragment reassembly header. 1277 * Make header visible. 1278 */ 1279 ip->ip_len = next; 1280 ip->ip_src = fp->ipq_src; 1281 ip->ip_dst = fp->ipq_dst; 1282 TAILQ_REMOVE(head, fp, ipq_list); 1283 fragq->nipq--; 1284 mpipe_free(&ipq_mpipe, fp); 1285 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 1286 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 1287 /* some debugging cruft by sklower, below, will go away soon */ 1288 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1289 int plen = 0; 1290 1291 for (n = m; n; n = n->m_next) 1292 plen += n->m_len; 1293 m->m_pkthdr.len = plen; 1294 } 1295 1296 /* 1297 * Reassembly complete, return the next protocol. 1298 * 1299 * Be sure to clear M_HASH to force the packet 1300 * to be re-characterized. 1301 * 1302 * Clear M_FRAG, we are no longer a fragment. 1303 */ 1304 m->m_flags &= ~(M_HASH | M_FRAG); 1305 1306 ipstat.ips_reassembled++; 1307 return (m); 1308 1309 dropfrag: 1310 ipstat.ips_fragdropped++; 1311 if (fp != NULL) 1312 fp->ipq_nfrags--; 1313 m_freem(m); 1314 done: 1315 return (NULL); 1316 1317 #undef GETIP 1318 } 1319 1320 /* 1321 * Free a fragment reassembly header and all 1322 * associated datagrams. 1323 */ 1324 static void 1325 ip_freef(struct ipfrag_queue *fragq, struct ipqhead *fhp, struct ipq *fp) 1326 { 1327 struct mbuf *q; 1328 1329 /* 1330 * Remove first to protect against blocking 1331 */ 1332 TAILQ_REMOVE(fhp, fp, ipq_list); 1333 1334 /* 1335 * Clean out at our leisure 1336 */ 1337 while (fp->ipq_frags) { 1338 q = fp->ipq_frags; 1339 fp->ipq_frags = q->m_nextpkt; 1340 q->m_nextpkt = NULL; 1341 m_freem(q); 1342 } 1343 mpipe_free(&ipq_mpipe, fp); 1344 fragq->nipq--; 1345 } 1346 1347 /* 1348 * If a timer expires on a reassembly queue, discard it. 1349 */ 1350 static void 1351 ipfrag_timeo_dispatch(netmsg_t nmsg) 1352 { 1353 struct ipfrag_queue *fragq = &ipfrag_queue_pcpu[mycpuid]; 1354 struct ipq *fp, *fp_temp; 1355 struct ipqhead *head; 1356 int i; 1357 1358 crit_enter(); 1359 lwkt_replymsg(&nmsg->lmsg, 0); /* reply ASAP */ 1360 crit_exit(); 1361 1362 for (i = 0; i < IPREASS_NHASH; i++) { 1363 head = &fragq->ipq[i]; 1364 TAILQ_FOREACH_MUTABLE(fp, head, ipq_list, fp_temp) { 1365 if (--fp->ipq_ttl == 0) { 1366 ipstat.ips_fragtimeout += fp->ipq_nfrags; 1367 ip_freef(fragq, head, fp); 1368 } 1369 } 1370 } 1371 /* 1372 * If we are over the maximum number of fragments 1373 * (due to the limit being lowered), drain off 1374 * enough to get down to the new limit. 1375 */ 1376 if (maxnipq >= 0 && fragq->nipq > maxnipq) { 1377 for (i = 0; i < IPREASS_NHASH; i++) { 1378 head = &fragq->ipq[i]; 1379 while (fragq->nipq > maxnipq && !TAILQ_EMPTY(head)) { 1380 ipstat.ips_fragdropped += 1381 TAILQ_FIRST(head)->ipq_nfrags; 1382 ip_freef(fragq, head, TAILQ_FIRST(head)); 1383 } 1384 } 1385 } 1386 } 1387 1388 static void 1389 ipfrag_timeo_ipi(void *arg __unused) 1390 { 1391 int cpu = mycpuid; 1392 struct lwkt_msg *msg = &ipfrag_queue_pcpu[cpu].timeo_netmsg.lmsg; 1393 1394 ipfrag_queue_pcpu[cpu].timeo_inprog = 0; 1395 crit_enter(); 1396 if (msg->ms_flags & MSGF_DONE) 1397 lwkt_sendmsg_oncpu(netisr_cpuport(cpu), msg); 1398 crit_exit(); 1399 } 1400 1401 static void 1402 ipfrag_slowtimo(void) 1403 { 1404 cpumask_t mask; 1405 int i; 1406 1407 CPUMASK_ASSZERO(mask); 1408 for (i = 0; i < ncpus; ++i) { 1409 if (ipfrag_queue_pcpu[i].nipq && 1410 ipfrag_queue_pcpu[i].timeo_inprog == 0) { 1411 ipfrag_queue_pcpu[i].timeo_inprog = 1; 1412 CPUMASK_ORBIT(mask, i); 1413 } 1414 } 1415 CPUMASK_ANDMASK(mask, smp_active_mask); 1416 if (CPUMASK_TESTNZERO(mask)) 1417 lwkt_send_ipiq_mask(mask, ipfrag_timeo_ipi, NULL); 1418 } 1419 1420 /* 1421 * IP timer processing 1422 */ 1423 void 1424 ip_slowtimo(void) 1425 { 1426 ipfrag_slowtimo(); 1427 ipflow_slowtimo(); 1428 } 1429 1430 /* 1431 * Drain off all datagram fragments. 1432 */ 1433 static void 1434 ipfrag_drain_dispatch(netmsg_t nmsg) 1435 { 1436 struct ipfrag_queue *fragq = &ipfrag_queue_pcpu[mycpuid]; 1437 struct ipqhead *head; 1438 int i; 1439 1440 crit_enter(); 1441 lwkt_replymsg(&nmsg->lmsg, 0); /* reply ASAP */ 1442 crit_exit(); 1443 1444 for (i = 0; i < IPREASS_NHASH; i++) { 1445 head = &fragq->ipq[i]; 1446 while (!TAILQ_EMPTY(head)) { 1447 ipstat.ips_fragdropped += TAILQ_FIRST(head)->ipq_nfrags; 1448 ip_freef(fragq, head, TAILQ_FIRST(head)); 1449 } 1450 } 1451 } 1452 1453 static void 1454 ipfrag_drain_ipi(void *arg __unused) 1455 { 1456 int cpu = mycpuid; 1457 struct lwkt_msg *msg = &ipfrag_queue_pcpu[cpu].drain_netmsg.lmsg; 1458 1459 crit_enter(); 1460 if (msg->ms_flags & MSGF_DONE) 1461 lwkt_sendmsg_oncpu(netisr_cpuport(cpu), msg); 1462 crit_exit(); 1463 } 1464 1465 static void 1466 ipfrag_drain(void) 1467 { 1468 cpumask_t mask; 1469 1470 CPUMASK_ASSBMASK(mask, ncpus); 1471 CPUMASK_ANDMASK(mask, smp_active_mask); 1472 if (CPUMASK_TESTNZERO(mask)) 1473 lwkt_send_ipiq_mask(mask, ipfrag_drain_ipi, NULL); 1474 } 1475 1476 void 1477 ip_drain(void) 1478 { 1479 ipfrag_drain(); 1480 in_rtqdrain(); 1481 } 1482 1483 /* 1484 * Do option processing on a datagram, 1485 * possibly discarding it if bad options are encountered, 1486 * or forwarding it if source-routed. 1487 * The pass argument is used when operating in the IPSTEALTH 1488 * mode to tell what options to process: 1489 * [LS]SRR (pass 0) or the others (pass 1). 1490 * The reason for as many as two passes is that when doing IPSTEALTH, 1491 * non-routing options should be processed only if the packet is for us. 1492 * Returns 1 if packet has been forwarded/freed, 1493 * 0 if the packet should be processed further. 1494 */ 1495 static int 1496 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop) 1497 { 1498 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET }; 1499 struct ip *ip = mtod(m, struct ip *); 1500 u_char *cp; 1501 struct in_ifaddr *ia; 1502 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB; 1503 boolean_t forward = FALSE; 1504 struct in_addr *sin, dst; 1505 n_time ntime; 1506 1507 dst = ip->ip_dst; 1508 cp = (u_char *)(ip + 1); 1509 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1510 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1511 opt = cp[IPOPT_OPTVAL]; 1512 if (opt == IPOPT_EOL) 1513 break; 1514 if (opt == IPOPT_NOP) 1515 optlen = 1; 1516 else { 1517 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1518 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1519 goto bad; 1520 } 1521 optlen = cp[IPOPT_OLEN]; 1522 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1523 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1524 goto bad; 1525 } 1526 } 1527 switch (opt) { 1528 1529 default: 1530 break; 1531 1532 /* 1533 * Source routing with record. 1534 * Find interface with current destination address. 1535 * If none on this machine then drop if strictly routed, 1536 * or do nothing if loosely routed. 1537 * Record interface address and bring up next address 1538 * component. If strictly routed make sure next 1539 * address is on directly accessible net. 1540 */ 1541 case IPOPT_LSRR: 1542 case IPOPT_SSRR: 1543 if (ipstealth && pass > 0) 1544 break; 1545 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1546 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1547 goto bad; 1548 } 1549 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1550 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1551 goto bad; 1552 } 1553 ipaddr.sin_addr = ip->ip_dst; 1554 ia = (struct in_ifaddr *) 1555 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1556 if (ia == NULL) { 1557 if (opt == IPOPT_SSRR) { 1558 type = ICMP_UNREACH; 1559 code = ICMP_UNREACH_SRCFAIL; 1560 goto bad; 1561 } 1562 if (!ip_dosourceroute) 1563 goto nosourcerouting; 1564 /* 1565 * Loose routing, and not at next destination 1566 * yet; nothing to do except forward. 1567 */ 1568 break; 1569 } 1570 off--; /* 0 origin */ 1571 if (off > optlen - (int)sizeof(struct in_addr)) { 1572 /* 1573 * End of source route. Should be for us. 1574 */ 1575 if (!ip_acceptsourceroute) 1576 goto nosourcerouting; 1577 save_rte(m, cp, ip->ip_src); 1578 break; 1579 } 1580 if (ipstealth) 1581 goto dropit; 1582 if (!ip_dosourceroute) { 1583 if (ipforwarding) { 1584 char buf[sizeof "aaa.bbb.ccc.ddd"]; 1585 1586 /* 1587 * Acting as a router, so generate ICMP 1588 */ 1589 nosourcerouting: 1590 strcpy(buf, inet_ntoa(ip->ip_dst)); 1591 log(LOG_WARNING, 1592 "attempted source route from %s to %s\n", 1593 inet_ntoa(ip->ip_src), buf); 1594 type = ICMP_UNREACH; 1595 code = ICMP_UNREACH_SRCFAIL; 1596 goto bad; 1597 } else { 1598 /* 1599 * Not acting as a router, 1600 * so silently drop. 1601 */ 1602 dropit: 1603 ipstat.ips_cantforward++; 1604 m_freem(m); 1605 return (1); 1606 } 1607 } 1608 1609 /* 1610 * locate outgoing interface 1611 */ 1612 memcpy(&ipaddr.sin_addr, cp + off, 1613 sizeof ipaddr.sin_addr); 1614 1615 if (opt == IPOPT_SSRR) { 1616 #define INA struct in_ifaddr * 1617 #define SA struct sockaddr * 1618 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) 1619 == NULL) 1620 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1621 } else { 1622 ia = ip_rtaddr(ipaddr.sin_addr, NULL); 1623 } 1624 if (ia == NULL) { 1625 type = ICMP_UNREACH; 1626 code = ICMP_UNREACH_SRCFAIL; 1627 goto bad; 1628 } 1629 ip->ip_dst = ipaddr.sin_addr; 1630 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1631 sizeof(struct in_addr)); 1632 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1633 /* 1634 * Let ip_intr's mcast routing check handle mcast pkts 1635 */ 1636 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1637 break; 1638 1639 case IPOPT_RR: 1640 if (ipstealth && pass == 0) 1641 break; 1642 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1643 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1644 goto bad; 1645 } 1646 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1647 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1648 goto bad; 1649 } 1650 /* 1651 * If no space remains, ignore. 1652 */ 1653 off--; /* 0 origin */ 1654 if (off > optlen - (int)sizeof(struct in_addr)) 1655 break; 1656 memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1657 sizeof ipaddr.sin_addr); 1658 /* 1659 * locate outgoing interface; if we're the destination, 1660 * use the incoming interface (should be same). 1661 */ 1662 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL && 1663 (ia = ip_rtaddr(ipaddr.sin_addr, NULL)) == NULL) { 1664 type = ICMP_UNREACH; 1665 code = ICMP_UNREACH_HOST; 1666 goto bad; 1667 } 1668 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1669 sizeof(struct in_addr)); 1670 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1671 break; 1672 1673 case IPOPT_TS: 1674 if (ipstealth && pass == 0) 1675 break; 1676 code = cp - (u_char *)ip; 1677 if (optlen < 4 || optlen > 40) { 1678 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1679 goto bad; 1680 } 1681 if ((off = cp[IPOPT_OFFSET]) < 5) { 1682 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1683 goto bad; 1684 } 1685 if (off > optlen - (int)sizeof(int32_t)) { 1686 cp[IPOPT_OFFSET + 1] += (1 << 4); 1687 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1688 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1689 goto bad; 1690 } 1691 break; 1692 } 1693 off--; /* 0 origin */ 1694 sin = (struct in_addr *)(cp + off); 1695 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1696 1697 case IPOPT_TS_TSONLY: 1698 break; 1699 1700 case IPOPT_TS_TSANDADDR: 1701 if (off + sizeof(n_time) + 1702 sizeof(struct in_addr) > optlen) { 1703 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1704 goto bad; 1705 } 1706 ipaddr.sin_addr = dst; 1707 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1708 m->m_pkthdr.rcvif); 1709 if (ia == NULL) 1710 continue; 1711 memcpy(sin, &IA_SIN(ia)->sin_addr, 1712 sizeof(struct in_addr)); 1713 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1714 off += sizeof(struct in_addr); 1715 break; 1716 1717 case IPOPT_TS_PRESPEC: 1718 if (off + sizeof(n_time) + 1719 sizeof(struct in_addr) > optlen) { 1720 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1721 goto bad; 1722 } 1723 memcpy(&ipaddr.sin_addr, sin, 1724 sizeof(struct in_addr)); 1725 if (ifa_ifwithaddr((SA)&ipaddr) == NULL) 1726 continue; 1727 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1728 off += sizeof(struct in_addr); 1729 break; 1730 1731 default: 1732 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1733 goto bad; 1734 } 1735 ntime = iptime(); 1736 memcpy(cp + off, &ntime, sizeof(n_time)); 1737 cp[IPOPT_OFFSET] += sizeof(n_time); 1738 } 1739 } 1740 if (forward && ipforwarding) { 1741 ip_forward(m, TRUE, next_hop); 1742 return (1); 1743 } 1744 return (0); 1745 bad: 1746 icmp_error(m, type, code, 0, 0); 1747 ipstat.ips_badoptions++; 1748 return (1); 1749 } 1750 1751 /* 1752 * Given address of next destination (final or next hop), 1753 * return internet address info of interface to be used to get there. 1754 */ 1755 struct in_ifaddr * 1756 ip_rtaddr(struct in_addr dst, struct route *ro0) 1757 { 1758 struct route sro, *ro; 1759 struct sockaddr_in *sin; 1760 struct in_ifaddr *ia; 1761 1762 if (ro0 != NULL) { 1763 ro = ro0; 1764 } else { 1765 bzero(&sro, sizeof(sro)); 1766 ro = &sro; 1767 } 1768 1769 sin = (struct sockaddr_in *)&ro->ro_dst; 1770 1771 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) { 1772 if (ro->ro_rt != NULL) { 1773 RTFREE(ro->ro_rt); 1774 ro->ro_rt = NULL; 1775 } 1776 sin->sin_family = AF_INET; 1777 sin->sin_len = sizeof *sin; 1778 sin->sin_addr = dst; 1779 rtalloc_ign(ro, RTF_PRCLONING); 1780 } 1781 1782 if (ro->ro_rt == NULL) 1783 return (NULL); 1784 1785 ia = ifatoia(ro->ro_rt->rt_ifa); 1786 1787 if (ro == &sro) 1788 RTFREE(ro->ro_rt); 1789 return ia; 1790 } 1791 1792 /* 1793 * Save incoming source route for use in replies, 1794 * to be picked up later by ip_srcroute if the receiver is interested. 1795 */ 1796 static void 1797 save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1798 { 1799 struct m_tag *mtag; 1800 struct ip_srcrt_opt *opt; 1801 unsigned olen; 1802 1803 mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), M_NOWAIT); 1804 if (mtag == NULL) 1805 return; 1806 opt = m_tag_data(mtag); 1807 1808 olen = option[IPOPT_OLEN]; 1809 #ifdef DIAGNOSTIC 1810 if (ipprintfs) 1811 kprintf("save_rte: olen %d\n", olen); 1812 #endif 1813 if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) { 1814 m_tag_free(mtag); 1815 return; 1816 } 1817 bcopy(option, opt->ip_srcrt.srcopt, olen); 1818 opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1819 opt->ip_srcrt.dst = dst; 1820 m_tag_prepend(m, mtag); 1821 } 1822 1823 /* 1824 * Retrieve incoming source route for use in replies, 1825 * in the same form used by setsockopt. 1826 * The first hop is placed before the options, will be removed later. 1827 */ 1828 struct mbuf * 1829 ip_srcroute(struct mbuf *m0) 1830 { 1831 struct in_addr *p, *q; 1832 struct mbuf *m; 1833 struct m_tag *mtag; 1834 struct ip_srcrt_opt *opt; 1835 1836 if (m0 == NULL) 1837 return NULL; 1838 1839 mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL); 1840 if (mtag == NULL) 1841 return NULL; 1842 opt = m_tag_data(mtag); 1843 1844 if (opt->ip_nhops == 0) 1845 return (NULL); 1846 m = m_get(M_NOWAIT, MT_HEADER); 1847 if (m == NULL) 1848 return (NULL); 1849 1850 #define OPTSIZ (sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt)) 1851 1852 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1853 m->m_len = opt->ip_nhops * sizeof(struct in_addr) + 1854 sizeof(struct in_addr) + OPTSIZ; 1855 #ifdef DIAGNOSTIC 1856 if (ipprintfs) { 1857 kprintf("ip_srcroute: nhops %d mlen %d", 1858 opt->ip_nhops, m->m_len); 1859 } 1860 #endif 1861 1862 /* 1863 * First save first hop for return route 1864 */ 1865 p = &opt->ip_srcrt.route[opt->ip_nhops - 1]; 1866 *(mtod(m, struct in_addr *)) = *p--; 1867 #ifdef DIAGNOSTIC 1868 if (ipprintfs) 1869 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1870 #endif 1871 1872 /* 1873 * Copy option fields and padding (nop) to mbuf. 1874 */ 1875 opt->ip_srcrt.nop = IPOPT_NOP; 1876 opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1877 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop, 1878 OPTSIZ); 1879 q = (struct in_addr *)(mtod(m, caddr_t) + 1880 sizeof(struct in_addr) + OPTSIZ); 1881 #undef OPTSIZ 1882 /* 1883 * Record return path as an IP source route, 1884 * reversing the path (pointers are now aligned). 1885 */ 1886 while (p >= opt->ip_srcrt.route) { 1887 #ifdef DIAGNOSTIC 1888 if (ipprintfs) 1889 kprintf(" %x", ntohl(q->s_addr)); 1890 #endif 1891 *q++ = *p--; 1892 } 1893 /* 1894 * Last hop goes to final destination. 1895 */ 1896 *q = opt->ip_srcrt.dst; 1897 m_tag_delete(m0, mtag); 1898 #ifdef DIAGNOSTIC 1899 if (ipprintfs) 1900 kprintf(" %x\n", ntohl(q->s_addr)); 1901 #endif 1902 return (m); 1903 } 1904 1905 /* 1906 * Strip out IP options. 1907 */ 1908 void 1909 ip_stripoptions(struct mbuf *m) 1910 { 1911 int datalen; 1912 struct ip *ip = mtod(m, struct ip *); 1913 caddr_t opts; 1914 int optlen; 1915 1916 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1917 opts = (caddr_t)(ip + 1); 1918 datalen = m->m_len - (sizeof(struct ip) + optlen); 1919 bcopy(opts + optlen, opts, datalen); 1920 m->m_len -= optlen; 1921 if (m->m_flags & M_PKTHDR) 1922 m->m_pkthdr.len -= optlen; 1923 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1924 } 1925 1926 u_char inetctlerrmap[PRC_NCMDS] = { 1927 0, 0, 0, 0, 1928 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1929 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1930 EMSGSIZE, EHOSTUNREACH, 0, 0, 1931 0, 0, 0, 0, 1932 ENOPROTOOPT, ECONNREFUSED 1933 }; 1934 1935 /* 1936 * Forward a packet. If some error occurs return the sender 1937 * an icmp packet. Note we can't always generate a meaningful 1938 * icmp message because icmp doesn't have a large enough repertoire 1939 * of codes and types. 1940 * 1941 * If not forwarding, just drop the packet. This could be confusing 1942 * if ipforwarding was zero but some routing protocol was advancing 1943 * us as a gateway to somewhere. However, we must let the routing 1944 * protocol deal with that. 1945 * 1946 * The using_srcrt parameter indicates whether the packet is being forwarded 1947 * via a source route. 1948 */ 1949 void 1950 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop) 1951 { 1952 struct ip *ip = mtod(m, struct ip *); 1953 struct rtentry *rt; 1954 struct route fwd_ro; 1955 int error, type = 0, code = 0, destmtu = 0; 1956 struct mbuf *mcopy, *mtemp = NULL; 1957 n_long dest; 1958 struct in_addr pkt_dst; 1959 1960 dest = INADDR_ANY; 1961 /* 1962 * Cache the destination address of the packet; this may be 1963 * changed by use of 'ipfw fwd'. 1964 */ 1965 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst; 1966 1967 #ifdef DIAGNOSTIC 1968 if (ipprintfs) 1969 kprintf("forward: src %x dst %x ttl %x\n", 1970 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl); 1971 #endif 1972 1973 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) { 1974 ipstat.ips_cantforward++; 1975 m_freem(m); 1976 return; 1977 } 1978 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) { 1979 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1980 return; 1981 } 1982 1983 bzero(&fwd_ro, sizeof(fwd_ro)); 1984 ip_rtaddr(pkt_dst, &fwd_ro); 1985 if (fwd_ro.ro_rt == NULL) { 1986 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1987 return; 1988 } 1989 rt = fwd_ro.ro_rt; 1990 1991 if (curthread->td_type == TD_TYPE_NETISR) { 1992 /* 1993 * Save the IP header and at most 8 bytes of the payload, 1994 * in case we need to generate an ICMP message to the src. 1995 */ 1996 mtemp = ipforward_mtemp[mycpuid]; 1997 KASSERT((mtemp->m_flags & M_EXT) == 0 && 1998 mtemp->m_data == mtemp->m_pktdat && 1999 m_tag_first(mtemp) == NULL, 2000 ("ip_forward invalid mtemp1")); 2001 2002 if (!m_dup_pkthdr(mtemp, m, M_NOWAIT)) { 2003 /* 2004 * It's probably ok if the pkthdr dup fails (because 2005 * the deep copy of the tag chain failed), but for now 2006 * be conservative and just discard the copy since 2007 * code below may some day want the tags. 2008 */ 2009 mtemp = NULL; 2010 } else { 2011 mtemp->m_type = m->m_type; 2012 mtemp->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8, 2013 (int)ip->ip_len); 2014 mtemp->m_pkthdr.len = mtemp->m_len; 2015 m_copydata(m, 0, mtemp->m_len, mtod(mtemp, caddr_t)); 2016 } 2017 } 2018 2019 if (!ipstealth) 2020 ip->ip_ttl -= IPTTLDEC; 2021 2022 /* 2023 * If forwarding packet using same interface that it came in on, 2024 * perhaps should send a redirect to sender to shortcut a hop. 2025 * Only send redirect if source is sending directly to us, 2026 * and if packet was not source routed (or has any options). 2027 * Also, don't send redirect if forwarding using a default route 2028 * or a route modified by a redirect. 2029 */ 2030 if (rt->rt_ifp == m->m_pkthdr.rcvif && 2031 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) && 2032 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY && 2033 ipsendredirects && !using_srcrt && next_hop == NULL) { 2034 u_long src = ntohl(ip->ip_src.s_addr); 2035 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa; 2036 2037 if (rt_ifa != NULL && 2038 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) { 2039 if (rt->rt_flags & RTF_GATEWAY) 2040 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 2041 else 2042 dest = pkt_dst.s_addr; 2043 /* 2044 * Router requirements says to only send 2045 * host redirects. 2046 */ 2047 type = ICMP_REDIRECT; 2048 code = ICMP_REDIRECT_HOST; 2049 #ifdef DIAGNOSTIC 2050 if (ipprintfs) 2051 kprintf("redirect (%d) to %x\n", code, dest); 2052 #endif 2053 } 2054 } 2055 2056 error = ip_output(m, NULL, &fwd_ro, IP_FORWARDING, NULL, NULL); 2057 if (error == 0) { 2058 ipstat.ips_forward++; 2059 if (type == 0) { 2060 if (mtemp) 2061 ipflow_create(&fwd_ro, mtemp); 2062 goto done; 2063 } 2064 ipstat.ips_redirectsent++; 2065 } else { 2066 ipstat.ips_cantforward++; 2067 } 2068 2069 if (mtemp == NULL) 2070 goto done; 2071 2072 /* 2073 * Errors that do not require generating ICMP message 2074 */ 2075 switch (error) { 2076 case ENOBUFS: 2077 /* 2078 * A router should not generate ICMP_SOURCEQUENCH as 2079 * required in RFC1812 Requirements for IP Version 4 Routers. 2080 * Source quench could be a big problem under DoS attacks, 2081 * or if the underlying interface is rate-limited. 2082 * Those who need source quench packets may re-enable them 2083 * via the net.inet.ip.sendsourcequench sysctl. 2084 */ 2085 if (!ip_sendsourcequench) 2086 goto done; 2087 break; 2088 2089 case EACCES: /* ipfw denied packet */ 2090 goto done; 2091 } 2092 2093 KASSERT((mtemp->m_flags & M_EXT) == 0 && 2094 mtemp->m_data == mtemp->m_pktdat, 2095 ("ip_forward invalid mtemp2")); 2096 mcopy = m_copym(mtemp, 0, mtemp->m_len, M_NOWAIT); 2097 if (mcopy == NULL) 2098 goto done; 2099 2100 /* 2101 * Send ICMP message. 2102 */ 2103 switch (error) { 2104 case 0: /* forwarded, but need redirect */ 2105 /* type, code set above */ 2106 break; 2107 2108 case ENETUNREACH: /* shouldn't happen, checked above */ 2109 case EHOSTUNREACH: 2110 case ENETDOWN: 2111 case EHOSTDOWN: 2112 default: 2113 type = ICMP_UNREACH; 2114 code = ICMP_UNREACH_HOST; 2115 break; 2116 2117 case EMSGSIZE: 2118 type = ICMP_UNREACH; 2119 code = ICMP_UNREACH_NEEDFRAG; 2120 #ifdef IPSEC 2121 /* 2122 * If the packet is routed over IPsec tunnel, tell the 2123 * originator the tunnel MTU. 2124 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 2125 * XXX quickhack!!! 2126 */ 2127 if (fwd_ro.ro_rt != NULL) { 2128 struct secpolicy *sp = NULL; 2129 int ipsecerror; 2130 int ipsechdr; 2131 struct route *ro; 2132 2133 sp = ipsec4_getpolicybyaddr(mcopy, 2134 IPSEC_DIR_OUTBOUND, 2135 IP_FORWARDING, 2136 &ipsecerror); 2137 2138 if (sp == NULL) 2139 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2140 else { 2141 /* count IPsec header size */ 2142 ipsechdr = ipsec4_hdrsiz(mcopy, 2143 IPSEC_DIR_OUTBOUND, 2144 NULL); 2145 2146 /* 2147 * find the correct route for outer IPv4 2148 * header, compute tunnel MTU. 2149 * 2150 */ 2151 if (sp->req != NULL && sp->req->sav != NULL && 2152 sp->req->sav->sah != NULL) { 2153 ro = &sp->req->sav->sah->sa_route; 2154 if (ro->ro_rt != NULL && 2155 ro->ro_rt->rt_ifp != NULL) { 2156 destmtu = 2157 ro->ro_rt->rt_ifp->if_mtu; 2158 destmtu -= ipsechdr; 2159 } 2160 } 2161 2162 key_freesp(sp); 2163 } 2164 } 2165 #elif defined(FAST_IPSEC) 2166 /* 2167 * If the packet is routed over IPsec tunnel, tell the 2168 * originator the tunnel MTU. 2169 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 2170 * XXX quickhack!!! 2171 */ 2172 if (fwd_ro.ro_rt != NULL) { 2173 struct secpolicy *sp = NULL; 2174 int ipsecerror; 2175 int ipsechdr; 2176 struct route *ro; 2177 2178 sp = ipsec_getpolicybyaddr(mcopy, 2179 IPSEC_DIR_OUTBOUND, 2180 IP_FORWARDING, 2181 &ipsecerror); 2182 2183 if (sp == NULL) 2184 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2185 else { 2186 /* count IPsec header size */ 2187 ipsechdr = ipsec4_hdrsiz(mcopy, 2188 IPSEC_DIR_OUTBOUND, 2189 NULL); 2190 2191 /* 2192 * find the correct route for outer IPv4 2193 * header, compute tunnel MTU. 2194 */ 2195 2196 if (sp->req != NULL && 2197 sp->req->sav != NULL && 2198 sp->req->sav->sah != NULL) { 2199 ro = &sp->req->sav->sah->sa_route; 2200 if (ro->ro_rt != NULL && 2201 ro->ro_rt->rt_ifp != NULL) { 2202 destmtu = 2203 ro->ro_rt->rt_ifp->if_mtu; 2204 destmtu -= ipsechdr; 2205 } 2206 } 2207 2208 KEY_FREESP(&sp); 2209 } 2210 } 2211 #else /* !IPSEC && !FAST_IPSEC */ 2212 if (fwd_ro.ro_rt != NULL) 2213 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2214 #endif /*IPSEC*/ 2215 ipstat.ips_cantfrag++; 2216 break; 2217 2218 case ENOBUFS: 2219 type = ICMP_SOURCEQUENCH; 2220 code = 0; 2221 break; 2222 2223 case EACCES: /* ipfw denied packet */ 2224 panic("ip_forward EACCES should not reach"); 2225 } 2226 icmp_error(mcopy, type, code, dest, destmtu); 2227 done: 2228 if (mtemp != NULL) 2229 m_tag_delete_chain(mtemp); 2230 if (fwd_ro.ro_rt != NULL) 2231 RTFREE(fwd_ro.ro_rt); 2232 } 2233 2234 void 2235 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2236 struct mbuf *m) 2237 { 2238 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2239 struct timeval tv; 2240 2241 microtime(&tv); 2242 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2243 SCM_TIMESTAMP, SOL_SOCKET); 2244 if (*mp) 2245 mp = &(*mp)->m_next; 2246 } 2247 if (inp->inp_flags & INP_RECVDSTADDR) { 2248 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2249 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2250 if (*mp) 2251 mp = &(*mp)->m_next; 2252 } 2253 if (inp->inp_flags & INP_RECVTTL) { 2254 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 2255 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 2256 if (*mp) 2257 mp = &(*mp)->m_next; 2258 } 2259 #ifdef notyet 2260 /* XXX 2261 * Moving these out of udp_input() made them even more broken 2262 * than they already were. 2263 */ 2264 /* options were tossed already */ 2265 if (inp->inp_flags & INP_RECVOPTS) { 2266 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2267 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2268 if (*mp) 2269 mp = &(*mp)->m_next; 2270 } 2271 /* ip_srcroute doesn't do what we want here, need to fix */ 2272 if (inp->inp_flags & INP_RECVRETOPTS) { 2273 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 2274 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2275 if (*mp) 2276 mp = &(*mp)->m_next; 2277 } 2278 #endif 2279 if (inp->inp_flags & INP_RECVIF) { 2280 struct ifnet *ifp; 2281 struct sdlbuf { 2282 struct sockaddr_dl sdl; 2283 u_char pad[32]; 2284 } sdlbuf; 2285 struct sockaddr_dl *sdp; 2286 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2287 2288 if (((ifp = m->m_pkthdr.rcvif)) && 2289 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) { 2290 sdp = IF_LLSOCKADDR(ifp); 2291 /* 2292 * Change our mind and don't try copy. 2293 */ 2294 if ((sdp->sdl_family != AF_LINK) || 2295 (sdp->sdl_len > sizeof(sdlbuf))) { 2296 goto makedummy; 2297 } 2298 bcopy(sdp, sdl2, sdp->sdl_len); 2299 } else { 2300 makedummy: 2301 sdl2->sdl_len = 2302 offsetof(struct sockaddr_dl, sdl_data[0]); 2303 sdl2->sdl_family = AF_LINK; 2304 sdl2->sdl_index = 0; 2305 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2306 } 2307 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 2308 IP_RECVIF, IPPROTO_IP); 2309 if (*mp) 2310 mp = &(*mp)->m_next; 2311 } 2312 } 2313 2314 /* 2315 * XXX these routines are called from the upper part of the kernel. 2316 * 2317 * They could also be moved to ip_mroute.c, since all the RSVP 2318 * handling is done there already. 2319 */ 2320 int 2321 ip_rsvp_init(struct socket *so) 2322 { 2323 if (so->so_type != SOCK_RAW || 2324 so->so_proto->pr_protocol != IPPROTO_RSVP) 2325 return EOPNOTSUPP; 2326 2327 if (ip_rsvpd != NULL) 2328 return EADDRINUSE; 2329 2330 ip_rsvpd = so; 2331 /* 2332 * This may seem silly, but we need to be sure we don't over-increment 2333 * the RSVP counter, in case something slips up. 2334 */ 2335 if (!ip_rsvp_on) { 2336 ip_rsvp_on = 1; 2337 rsvp_on++; 2338 } 2339 2340 return 0; 2341 } 2342 2343 int 2344 ip_rsvp_done(void) 2345 { 2346 ip_rsvpd = NULL; 2347 /* 2348 * This may seem silly, but we need to be sure we don't over-decrement 2349 * the RSVP counter, in case something slips up. 2350 */ 2351 if (ip_rsvp_on) { 2352 ip_rsvp_on = 0; 2353 rsvp_on--; 2354 } 2355 return 0; 2356 } 2357 2358 int 2359 rsvp_input(struct mbuf **mp, int *offp, int proto) 2360 { 2361 struct mbuf *m = *mp; 2362 2363 *mp = NULL; 2364 2365 if (rsvp_input_p) { /* call the real one if loaded */ 2366 *mp = m; 2367 rsvp_input_p(mp, offp, proto); 2368 return(IPPROTO_DONE); 2369 } 2370 2371 /* Can still get packets with rsvp_on = 0 if there is a local member 2372 * of the group to which the RSVP packet is addressed. But in this 2373 * case we want to throw the packet away. 2374 */ 2375 2376 if (!rsvp_on) { 2377 m_freem(m); 2378 return(IPPROTO_DONE); 2379 } 2380 2381 if (ip_rsvpd != NULL) { 2382 *mp = m; 2383 rip_input(mp, offp, proto); 2384 return(IPPROTO_DONE); 2385 } 2386 /* Drop the packet */ 2387 m_freem(m); 2388 return(IPPROTO_DONE); 2389 } 2390