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