ip_input.c revision 191259
1/*- 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 30 */ 31 32#include <sys/cdefs.h> 33__FBSDID("$FreeBSD: head/sys/netinet/ip_input.c 191259 2009-04-19 04:44:05Z kmacy $"); 34 35#include "opt_bootp.h" 36#include "opt_ipfw.h" 37#include "opt_ipstealth.h" 38#include "opt_ipsec.h" 39#include "opt_route.h" 40#include "opt_mac.h" 41#include "opt_carp.h" 42 43#include <sys/param.h> 44#include <sys/systm.h> 45#include <sys/callout.h> 46#include <sys/mbuf.h> 47#include <sys/malloc.h> 48#include <sys/domain.h> 49#include <sys/protosw.h> 50#include <sys/socket.h> 51#include <sys/time.h> 52#include <sys/kernel.h> 53#include <sys/lock.h> 54#include <sys/rwlock.h> 55#include <sys/syslog.h> 56#include <sys/sysctl.h> 57#include <sys/vimage.h> 58 59#include <net/pfil.h> 60#include <net/if.h> 61#include <net/if_types.h> 62#include <net/if_var.h> 63#include <net/if_dl.h> 64#include <net/route.h> 65#include <net/netisr.h> 66#include <net/vnet.h> 67#include <net/flowtable.h> 68 69#include <netinet/in.h> 70#include <netinet/in_systm.h> 71#include <netinet/in_var.h> 72#include <netinet/ip.h> 73#include <netinet/in_pcb.h> 74#include <netinet/ip_var.h> 75#include <netinet/ip_icmp.h> 76#include <netinet/ip_options.h> 77#include <machine/in_cksum.h> 78#include <netinet/vinet.h> 79#ifdef DEV_CARP 80#include <netinet/ip_carp.h> 81#endif 82#ifdef IPSEC 83#include <netinet/ip_ipsec.h> 84#endif /* IPSEC */ 85 86#include <sys/socketvar.h> 87 88/* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */ 89#include <netinet/ip_fw.h> 90#include <netinet/ip_dummynet.h> 91 92#include <security/mac/mac_framework.h> 93 94#ifdef CTASSERT 95CTASSERT(sizeof(struct ip) == 20); 96#endif 97 98#ifndef VIMAGE 99#ifndef VIMAGE_GLOBALS 100struct vnet_inet vnet_inet_0; 101#endif 102#endif 103 104#ifdef VIMAGE_GLOBALS 105static int ipsendredirects; 106static int ip_checkinterface; 107static int ip_keepfaith; 108static int ip_sendsourcequench; 109int ip_defttl; 110int ip_do_randomid; 111int ipforwarding; 112struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 113struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 114u_long in_ifaddrhmask; /* mask for hash table */ 115struct ipstat ipstat; 116static int ip_rsvp_on; 117struct socket *ip_rsvpd; 118int rsvp_on; 119static struct ipqhead ipq[IPREASS_NHASH]; 120static int maxnipq; /* Administrative limit on # reass queues. */ 121static int maxfragsperpacket; 122int ipstealth; 123static int nipq; /* Total # of reass queues */ 124#endif 125 126SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_FORWARDING, 127 forwarding, CTLFLAG_RW, ipforwarding, 0, 128 "Enable IP forwarding between interfaces"); 129 130SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_SENDREDIRECTS, 131 redirect, CTLFLAG_RW, ipsendredirects, 0, 132 "Enable sending IP redirects"); 133 134SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_DEFTTL, 135 ttl, CTLFLAG_RW, ip_defttl, 0, "Maximum TTL on IP packets"); 136 137SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_KEEPFAITH, 138 keepfaith, CTLFLAG_RW, ip_keepfaith, 0, 139 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 140 141SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, 142 sendsourcequench, CTLFLAG_RW, ip_sendsourcequench, 0, 143 "Enable the transmission of source quench packets"); 144 145SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, random_id, 146 CTLFLAG_RW, ip_do_randomid, 0, "Assign random ip_id values"); 147 148/* 149 * XXX - Setting ip_checkinterface mostly implements the receive side of 150 * the Strong ES model described in RFC 1122, but since the routing table 151 * and transmit implementation do not implement the Strong ES model, 152 * setting this to 1 results in an odd hybrid. 153 * 154 * XXX - ip_checkinterface currently must be disabled if you use ipnat 155 * to translate the destination address to another local interface. 156 * 157 * XXX - ip_checkinterface must be disabled if you add IP aliases 158 * to the loopback interface instead of the interface where the 159 * packets for those addresses are received. 160 */ 161SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, 162 check_interface, CTLFLAG_RW, ip_checkinterface, 0, 163 "Verify packet arrives on correct interface"); 164 165struct pfil_head inet_pfil_hook; /* Packet filter hooks */ 166 167static struct ifqueue ipintrq; 168static int ipqmaxlen = IFQ_MAXLEN; 169 170extern struct domain inetdomain; 171extern struct protosw inetsw[]; 172u_char ip_protox[IPPROTO_MAX]; 173 174SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 175 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 176SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 177 &ipintrq.ifq_drops, 0, 178 "Number of packets dropped from the IP input queue"); 179 180SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 181 ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 182 183#ifdef VIMAGE_GLOBALS 184static uma_zone_t ipq_zone; 185#endif 186static struct mtx ipqlock; 187 188#define IPQ_LOCK() mtx_lock(&ipqlock) 189#define IPQ_UNLOCK() mtx_unlock(&ipqlock) 190#define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 191#define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 192 193static void maxnipq_update(void); 194static void ipq_zone_change(void *); 195 196SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, fragpackets, 197 CTLFLAG_RD, nipq, 0, 198 "Current number of IPv4 fragment reassembly queue entries"); 199 200SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, maxfragsperpacket, 201 CTLFLAG_RW, maxfragsperpacket, 0, 202 "Maximum number of IPv4 fragments allowed per packet"); 203 204struct callout ipport_tick_callout; 205 206#ifdef IPCTL_DEFMTU 207SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 208 &ip_mtu, 0, "Default MTU"); 209#endif 210 211#ifdef IPSTEALTH 212SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 213 ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding"); 214#endif 215static int ip_output_flowtable_size = 2048; 216TUNABLE_INT("net.inet.ip.output_flowtable_size", &ip_output_flowtable_size); 217SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, output_flowtable_size, 218 CTLFLAG_RDTUN, ip_output_flowtable_size, 2048, 219 "number of entries in the per-cpu output flow caches"); 220 221/* 222 * ipfw_ether and ipfw_bridge hooks. 223 * XXX: Temporary until those are converted to pfil_hooks as well. 224 */ 225ip_fw_chk_t *ip_fw_chk_ptr = NULL; 226ip_dn_io_t *ip_dn_io_ptr = NULL; 227#ifdef VIMAGE_GLOBALS 228int fw_one_pass; 229#endif 230struct flowtable *ip_ft; 231 232static void ip_freef(struct ipqhead *, struct ipq *); 233 234#ifndef VIMAGE_GLOBALS 235static void vnet_inet_register(void); 236 237static const vnet_modinfo_t vnet_inet_modinfo = { 238 .vmi_id = VNET_MOD_INET, 239 .vmi_name = "inet", 240}; 241 242static void vnet_inet_register() 243{ 244 245 vnet_mod_register(&vnet_inet_modinfo); 246} 247 248SYSINIT(inet, SI_SUB_PROTO_BEGIN, SI_ORDER_FIRST, vnet_inet_register, 0); 249#endif 250 251/* 252 * IP initialization: fill in IP protocol switch table. 253 * All protocols not implemented in kernel go to raw IP protocol handler. 254 */ 255void 256ip_init(void) 257{ 258 INIT_VNET_INET(curvnet); 259 struct protosw *pr; 260 int i; 261 262 V_ipsendredirects = 1; /* XXX */ 263 V_ip_checkinterface = 0; 264 V_ip_keepfaith = 0; 265 V_ip_sendsourcequench = 0; 266 V_rsvp_on = 0; 267 V_ip_defttl = IPDEFTTL; 268 V_ip_do_randomid = 0; 269 V_ip_id = time_second & 0xffff; 270 V_ipforwarding = 0; 271 V_ipstealth = 0; 272 V_nipq = 0; /* Total # of reass queues */ 273 274 V_ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */ 275 V_ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */ 276 V_ipport_firstauto = IPPORT_EPHEMERALFIRST; /* 10000 */ 277 V_ipport_lastauto = IPPORT_EPHEMERALLAST; /* 65535 */ 278 V_ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */ 279 V_ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */ 280 V_ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */ 281 V_ipport_reservedlow = 0; 282 V_ipport_randomized = 1; /* user controlled via sysctl */ 283 V_ipport_randomcps = 10; /* user controlled via sysctl */ 284 V_ipport_randomtime = 45; /* user controlled via sysctl */ 285 V_ipport_stoprandom = 0; /* toggled by ipport_tick */ 286 287 V_fw_one_pass = 1; 288 289#ifdef NOTYET 290 /* XXX global static but not instantiated in this file */ 291 V_ipfastforward_active = 0; 292 V_subnetsarelocal = 0; 293 V_sameprefixcarponly = 0; 294#endif 295 296 TAILQ_INIT(&V_in_ifaddrhead); 297 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 298 299 /* Initialize IP reassembly queue. */ 300 for (i = 0; i < IPREASS_NHASH; i++) 301 TAILQ_INIT(&V_ipq[i]); 302 V_maxnipq = nmbclusters / 32; 303 V_maxfragsperpacket = 16; 304 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 305 NULL, UMA_ALIGN_PTR, 0); 306 maxnipq_update(); 307 308 /* Skip initialization of globals for non-default instances. */ 309 if (!IS_DEFAULT_VNET(curvnet)) 310 return; 311 312 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 313 if (pr == NULL) 314 panic("ip_init: PF_INET not found"); 315 316 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 317 for (i = 0; i < IPPROTO_MAX; i++) 318 ip_protox[i] = pr - inetsw; 319 /* 320 * Cycle through IP protocols and put them into the appropriate place 321 * in ip_protox[]. 322 */ 323 for (pr = inetdomain.dom_protosw; 324 pr < inetdomain.dom_protoswNPROTOSW; pr++) 325 if (pr->pr_domain->dom_family == PF_INET && 326 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 327 /* Be careful to only index valid IP protocols. */ 328 if (pr->pr_protocol < IPPROTO_MAX) 329 ip_protox[pr->pr_protocol] = pr - inetsw; 330 } 331 332 /* Initialize packet filter hooks. */ 333 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 334 inet_pfil_hook.ph_af = AF_INET; 335 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 336 printf("%s: WARNING: unable to register pfil hook, " 337 "error %d\n", __func__, i); 338 339 /* Start ipport_tick. */ 340 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE); 341 ipport_tick(NULL); 342 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, 343 SHUTDOWN_PRI_DEFAULT); 344 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, 345 NULL, EVENTHANDLER_PRI_ANY); 346 347 /* Initialize various other remaining things. */ 348 IPQ_LOCK_INIT(); 349 ipintrq.ifq_maxlen = ipqmaxlen; 350 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF); 351 netisr_register(NETISR_IP, ip_input, &ipintrq, 0); 352 353 ip_ft = flowtable_alloc(ip_output_flowtable_size, FL_PCPU); 354} 355 356void 357ip_fini(void *xtp) 358{ 359 360 callout_stop(&ipport_tick_callout); 361} 362 363/* 364 * Ip input routine. Checksum and byte swap header. If fragmented 365 * try to reassemble. Process options. Pass to next level. 366 */ 367void 368ip_input(struct mbuf *m) 369{ 370 INIT_VNET_INET(curvnet); 371 struct ip *ip = NULL; 372 struct in_ifaddr *ia = NULL; 373 struct ifaddr *ifa; 374 int checkif, hlen = 0; 375 u_short sum; 376 int dchg = 0; /* dest changed after fw */ 377 struct in_addr odst; /* original dst address */ 378 379 M_ASSERTPKTHDR(m); 380 381 if (m->m_flags & M_FASTFWD_OURS) { 382 /* 383 * Firewall or NAT changed destination to local. 384 * We expect ip_len and ip_off to be in host byte order. 385 */ 386 m->m_flags &= ~M_FASTFWD_OURS; 387 /* Set up some basics that will be used later. */ 388 ip = mtod(m, struct ip *); 389 hlen = ip->ip_hl << 2; 390 goto ours; 391 } 392 393 IPSTAT_INC(ips_total); 394 395 if (m->m_pkthdr.len < sizeof(struct ip)) 396 goto tooshort; 397 398 if (m->m_len < sizeof (struct ip) && 399 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 400 IPSTAT_INC(ips_toosmall); 401 return; 402 } 403 ip = mtod(m, struct ip *); 404 405 if (ip->ip_v != IPVERSION) { 406 IPSTAT_INC(ips_badvers); 407 goto bad; 408 } 409 410 hlen = ip->ip_hl << 2; 411 if (hlen < sizeof(struct ip)) { /* minimum header length */ 412 IPSTAT_INC(ips_badhlen); 413 goto bad; 414 } 415 if (hlen > m->m_len) { 416 if ((m = m_pullup(m, hlen)) == NULL) { 417 IPSTAT_INC(ips_badhlen); 418 return; 419 } 420 ip = mtod(m, struct ip *); 421 } 422 423 /* 127/8 must not appear on wire - RFC1122 */ 424 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 425 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 426 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 427 IPSTAT_INC(ips_badaddr); 428 goto bad; 429 } 430 } 431 432 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 433 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 434 } else { 435 if (hlen == sizeof(struct ip)) { 436 sum = in_cksum_hdr(ip); 437 } else { 438 sum = in_cksum(m, hlen); 439 } 440 } 441 if (sum) { 442 IPSTAT_INC(ips_badsum); 443 goto bad; 444 } 445 446#ifdef ALTQ 447 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 448 /* packet is dropped by traffic conditioner */ 449 return; 450#endif 451 452 /* 453 * Convert fields to host representation. 454 */ 455 ip->ip_len = ntohs(ip->ip_len); 456 if (ip->ip_len < hlen) { 457 IPSTAT_INC(ips_badlen); 458 goto bad; 459 } 460 ip->ip_off = ntohs(ip->ip_off); 461 462 /* 463 * Check that the amount of data in the buffers 464 * is as at least much as the IP header would have us expect. 465 * Trim mbufs if longer than we expect. 466 * Drop packet if shorter than we expect. 467 */ 468 if (m->m_pkthdr.len < ip->ip_len) { 469tooshort: 470 IPSTAT_INC(ips_tooshort); 471 goto bad; 472 } 473 if (m->m_pkthdr.len > ip->ip_len) { 474 if (m->m_len == m->m_pkthdr.len) { 475 m->m_len = ip->ip_len; 476 m->m_pkthdr.len = ip->ip_len; 477 } else 478 m_adj(m, ip->ip_len - m->m_pkthdr.len); 479 } 480#ifdef IPSEC 481 /* 482 * Bypass packet filtering for packets from a tunnel (gif). 483 */ 484 if (ip_ipsec_filtertunnel(m)) 485 goto passin; 486#endif /* IPSEC */ 487 488 /* 489 * Run through list of hooks for input packets. 490 * 491 * NB: Beware of the destination address changing (e.g. 492 * by NAT rewriting). When this happens, tell 493 * ip_forward to do the right thing. 494 */ 495 496 /* Jump over all PFIL processing if hooks are not active. */ 497 if (!PFIL_HOOKED(&inet_pfil_hook)) 498 goto passin; 499 500 odst = ip->ip_dst; 501 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 502 PFIL_IN, NULL) != 0) 503 return; 504 if (m == NULL) /* consumed by filter */ 505 return; 506 507 ip = mtod(m, struct ip *); 508 dchg = (odst.s_addr != ip->ip_dst.s_addr); 509 510#ifdef IPFIREWALL_FORWARD 511 if (m->m_flags & M_FASTFWD_OURS) { 512 m->m_flags &= ~M_FASTFWD_OURS; 513 goto ours; 514 } 515 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 516 /* 517 * Directly ship on the packet. This allows to forward packets 518 * that were destined for us to some other directly connected 519 * host. 520 */ 521 ip_forward(m, dchg); 522 return; 523 } 524#endif /* IPFIREWALL_FORWARD */ 525 526passin: 527 /* 528 * Process options and, if not destined for us, 529 * ship it on. ip_dooptions returns 1 when an 530 * error was detected (causing an icmp message 531 * to be sent and the original packet to be freed). 532 */ 533 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 534 return; 535 536 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 537 * matter if it is destined to another node, or whether it is 538 * a multicast one, RSVP wants it! and prevents it from being forwarded 539 * anywhere else. Also checks if the rsvp daemon is running before 540 * grabbing the packet. 541 */ 542 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 543 goto ours; 544 545 /* 546 * Check our list of addresses, to see if the packet is for us. 547 * If we don't have any addresses, assume any unicast packet 548 * we receive might be for us (and let the upper layers deal 549 * with it). 550 */ 551 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 552 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 553 goto ours; 554 555 /* 556 * Enable a consistency check between the destination address 557 * and the arrival interface for a unicast packet (the RFC 1122 558 * strong ES model) if IP forwarding is disabled and the packet 559 * is not locally generated and the packet is not subject to 560 * 'ipfw fwd'. 561 * 562 * XXX - Checking also should be disabled if the destination 563 * address is ipnat'ed to a different interface. 564 * 565 * XXX - Checking is incompatible with IP aliases added 566 * to the loopback interface instead of the interface where 567 * the packets are received. 568 * 569 * XXX - This is the case for carp vhost IPs as well so we 570 * insert a workaround. If the packet got here, we already 571 * checked with carp_iamatch() and carp_forus(). 572 */ 573 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 574 m->m_pkthdr.rcvif != NULL && 575 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 576#ifdef DEV_CARP 577 !m->m_pkthdr.rcvif->if_carp && 578#endif 579 (dchg == 0); 580 581 /* 582 * Check for exact addresses in the hash bucket. 583 */ 584 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 585 /* 586 * If the address matches, verify that the packet 587 * arrived via the correct interface if checking is 588 * enabled. 589 */ 590 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 591 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 592 goto ours; 593 } 594 /* 595 * Check for broadcast addresses. 596 * 597 * Only accept broadcast packets that arrive via the matching 598 * interface. Reception of forwarded directed broadcasts would 599 * be handled via ip_forward() and ether_output() with the loopback 600 * into the stack for SIMPLEX interfaces handled by ether_output(). 601 */ 602 if (m->m_pkthdr.rcvif != NULL && 603 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 604 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 605 if (ifa->ifa_addr->sa_family != AF_INET) 606 continue; 607 ia = ifatoia(ifa); 608 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 609 ip->ip_dst.s_addr) 610 goto ours; 611 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) 612 goto ours; 613#ifdef BOOTP_COMPAT 614 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 615 goto ours; 616#endif 617 } 618 } 619 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 620 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 621 IPSTAT_INC(ips_cantforward); 622 m_freem(m); 623 return; 624 } 625 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 626 if (V_ip_mrouter) { 627 /* 628 * If we are acting as a multicast router, all 629 * incoming multicast packets are passed to the 630 * kernel-level multicast forwarding function. 631 * The packet is returned (relatively) intact; if 632 * ip_mforward() returns a non-zero value, the packet 633 * must be discarded, else it may be accepted below. 634 */ 635 if (ip_mforward && 636 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 637 IPSTAT_INC(ips_cantforward); 638 m_freem(m); 639 return; 640 } 641 642 /* 643 * The process-level routing daemon needs to receive 644 * all multicast IGMP packets, whether or not this 645 * host belongs to their destination groups. 646 */ 647 if (ip->ip_p == IPPROTO_IGMP) 648 goto ours; 649 IPSTAT_INC(ips_forward); 650 } 651 /* 652 * Assume the packet is for us, to avoid prematurely taking 653 * a lock on the in_multi hash. Protocols must perform 654 * their own filtering and update statistics accordingly. 655 */ 656 goto ours; 657 } 658 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 659 goto ours; 660 if (ip->ip_dst.s_addr == INADDR_ANY) 661 goto ours; 662 663 /* 664 * FAITH(Firewall Aided Internet Translator) 665 */ 666 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 667 if (V_ip_keepfaith) { 668 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 669 goto ours; 670 } 671 m_freem(m); 672 return; 673 } 674 675 /* 676 * Not for us; forward if possible and desirable. 677 */ 678 if (V_ipforwarding == 0) { 679 IPSTAT_INC(ips_cantforward); 680 m_freem(m); 681 } else { 682#ifdef IPSEC 683 if (ip_ipsec_fwd(m)) 684 goto bad; 685#endif /* IPSEC */ 686 ip_forward(m, dchg); 687 } 688 return; 689 690ours: 691#ifdef IPSTEALTH 692 /* 693 * IPSTEALTH: Process non-routing options only 694 * if the packet is destined for us. 695 */ 696 if (V_ipstealth && hlen > sizeof (struct ip) && 697 ip_dooptions(m, 1)) 698 return; 699#endif /* IPSTEALTH */ 700 701 /* Count the packet in the ip address stats */ 702 if (ia != NULL) { 703 ia->ia_ifa.if_ipackets++; 704 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 705 } 706 707 /* 708 * Attempt reassembly; if it succeeds, proceed. 709 * ip_reass() will return a different mbuf. 710 */ 711 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 712 m = ip_reass(m); 713 if (m == NULL) 714 return; 715 ip = mtod(m, struct ip *); 716 /* Get the header length of the reassembled packet */ 717 hlen = ip->ip_hl << 2; 718 } 719 720 /* 721 * Further protocols expect the packet length to be w/o the 722 * IP header. 723 */ 724 ip->ip_len -= hlen; 725 726#ifdef IPSEC 727 /* 728 * enforce IPsec policy checking if we are seeing last header. 729 * note that we do not visit this with protocols with pcb layer 730 * code - like udp/tcp/raw ip. 731 */ 732 if (ip_ipsec_input(m)) 733 goto bad; 734#endif /* IPSEC */ 735 736 /* 737 * Switch out to protocol's input routine. 738 */ 739 IPSTAT_INC(ips_delivered); 740 741 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 742 return; 743bad: 744 m_freem(m); 745} 746 747/* 748 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 749 * max has slightly different semantics than the sysctl, for historical 750 * reasons. 751 */ 752static void 753maxnipq_update(void) 754{ 755 INIT_VNET_INET(curvnet); 756 757 /* 758 * -1 for unlimited allocation. 759 */ 760 if (V_maxnipq < 0) 761 uma_zone_set_max(V_ipq_zone, 0); 762 /* 763 * Positive number for specific bound. 764 */ 765 if (V_maxnipq > 0) 766 uma_zone_set_max(V_ipq_zone, V_maxnipq); 767 /* 768 * Zero specifies no further fragment queue allocation -- set the 769 * bound very low, but rely on implementation elsewhere to actually 770 * prevent allocation and reclaim current queues. 771 */ 772 if (V_maxnipq == 0) 773 uma_zone_set_max(V_ipq_zone, 1); 774} 775 776static void 777ipq_zone_change(void *tag) 778{ 779 INIT_VNET_INET(curvnet); 780 781 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) { 782 V_maxnipq = nmbclusters / 32; 783 maxnipq_update(); 784 } 785} 786 787static int 788sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 789{ 790 INIT_VNET_INET(curvnet); 791 int error, i; 792 793 i = V_maxnipq; 794 error = sysctl_handle_int(oidp, &i, 0, req); 795 if (error || !req->newptr) 796 return (error); 797 798 /* 799 * XXXRW: Might be a good idea to sanity check the argument and place 800 * an extreme upper bound. 801 */ 802 if (i < -1) 803 return (EINVAL); 804 V_maxnipq = i; 805 maxnipq_update(); 806 return (0); 807} 808 809SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 810 NULL, 0, sysctl_maxnipq, "I", 811 "Maximum number of IPv4 fragment reassembly queue entries"); 812 813/* 814 * Take incoming datagram fragment and try to reassemble it into 815 * whole datagram. If the argument is the first fragment or one 816 * in between the function will return NULL and store the mbuf 817 * in the fragment chain. If the argument is the last fragment 818 * the packet will be reassembled and the pointer to the new 819 * mbuf returned for further processing. Only m_tags attached 820 * to the first packet/fragment are preserved. 821 * The IP header is *NOT* adjusted out of iplen. 822 */ 823struct mbuf * 824ip_reass(struct mbuf *m) 825{ 826 INIT_VNET_INET(curvnet); 827 struct ip *ip; 828 struct mbuf *p, *q, *nq, *t; 829 struct ipq *fp = NULL; 830 struct ipqhead *head; 831 int i, hlen, next; 832 u_int8_t ecn, ecn0; 833 u_short hash; 834 835 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 836 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) { 837 IPSTAT_INC(ips_fragments); 838 IPSTAT_INC(ips_fragdropped); 839 m_freem(m); 840 return (NULL); 841 } 842 843 ip = mtod(m, struct ip *); 844 hlen = ip->ip_hl << 2; 845 846 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 847 head = &V_ipq[hash]; 848 IPQ_LOCK(); 849 850 /* 851 * Look for queue of fragments 852 * of this datagram. 853 */ 854 TAILQ_FOREACH(fp, head, ipq_list) 855 if (ip->ip_id == fp->ipq_id && 856 ip->ip_src.s_addr == fp->ipq_src.s_addr && 857 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 858#ifdef MAC 859 mac_ipq_match(m, fp) && 860#endif 861 ip->ip_p == fp->ipq_p) 862 goto found; 863 864 fp = NULL; 865 866 /* 867 * Attempt to trim the number of allocated fragment queues if it 868 * exceeds the administrative limit. 869 */ 870 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) { 871 /* 872 * drop something from the tail of the current queue 873 * before proceeding further 874 */ 875 struct ipq *q = TAILQ_LAST(head, ipqhead); 876 if (q == NULL) { /* gak */ 877 for (i = 0; i < IPREASS_NHASH; i++) { 878 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead); 879 if (r) { 880 IPSTAT_ADD(ips_fragtimeout, 881 r->ipq_nfrags); 882 ip_freef(&V_ipq[i], r); 883 break; 884 } 885 } 886 } else { 887 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags); 888 ip_freef(head, q); 889 } 890 } 891 892found: 893 /* 894 * Adjust ip_len to not reflect header, 895 * convert offset of this to bytes. 896 */ 897 ip->ip_len -= hlen; 898 if (ip->ip_off & IP_MF) { 899 /* 900 * Make sure that fragments have a data length 901 * that's a non-zero multiple of 8 bytes. 902 */ 903 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 904 IPSTAT_INC(ips_toosmall); /* XXX */ 905 goto dropfrag; 906 } 907 m->m_flags |= M_FRAG; 908 } else 909 m->m_flags &= ~M_FRAG; 910 ip->ip_off <<= 3; 911 912 913 /* 914 * Attempt reassembly; if it succeeds, proceed. 915 * ip_reass() will return a different mbuf. 916 */ 917 IPSTAT_INC(ips_fragments); 918 m->m_pkthdr.header = ip; 919 920 /* Previous ip_reass() started here. */ 921 /* 922 * Presence of header sizes in mbufs 923 * would confuse code below. 924 */ 925 m->m_data += hlen; 926 m->m_len -= hlen; 927 928 /* 929 * If first fragment to arrive, create a reassembly queue. 930 */ 931 if (fp == NULL) { 932 fp = uma_zalloc(V_ipq_zone, M_NOWAIT); 933 if (fp == NULL) 934 goto dropfrag; 935#ifdef MAC 936 if (mac_ipq_init(fp, M_NOWAIT) != 0) { 937 uma_zfree(V_ipq_zone, fp); 938 fp = NULL; 939 goto dropfrag; 940 } 941 mac_ipq_create(m, fp); 942#endif 943 TAILQ_INSERT_HEAD(head, fp, ipq_list); 944 V_nipq++; 945 fp->ipq_nfrags = 1; 946 fp->ipq_ttl = IPFRAGTTL; 947 fp->ipq_p = ip->ip_p; 948 fp->ipq_id = ip->ip_id; 949 fp->ipq_src = ip->ip_src; 950 fp->ipq_dst = ip->ip_dst; 951 fp->ipq_frags = m; 952 m->m_nextpkt = NULL; 953 goto done; 954 } else { 955 fp->ipq_nfrags++; 956#ifdef MAC 957 mac_ipq_update(m, fp); 958#endif 959 } 960 961#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 962 963 /* 964 * Handle ECN by comparing this segment with the first one; 965 * if CE is set, do not lose CE. 966 * drop if CE and not-ECT are mixed for the same packet. 967 */ 968 ecn = ip->ip_tos & IPTOS_ECN_MASK; 969 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 970 if (ecn == IPTOS_ECN_CE) { 971 if (ecn0 == IPTOS_ECN_NOTECT) 972 goto dropfrag; 973 if (ecn0 != IPTOS_ECN_CE) 974 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 975 } 976 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 977 goto dropfrag; 978 979 /* 980 * Find a segment which begins after this one does. 981 */ 982 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 983 if (GETIP(q)->ip_off > ip->ip_off) 984 break; 985 986 /* 987 * If there is a preceding segment, it may provide some of 988 * our data already. If so, drop the data from the incoming 989 * segment. If it provides all of our data, drop us, otherwise 990 * stick new segment in the proper place. 991 * 992 * If some of the data is dropped from the the preceding 993 * segment, then it's checksum is invalidated. 994 */ 995 if (p) { 996 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 997 if (i > 0) { 998 if (i >= ip->ip_len) 999 goto dropfrag; 1000 m_adj(m, i); 1001 m->m_pkthdr.csum_flags = 0; 1002 ip->ip_off += i; 1003 ip->ip_len -= i; 1004 } 1005 m->m_nextpkt = p->m_nextpkt; 1006 p->m_nextpkt = m; 1007 } else { 1008 m->m_nextpkt = fp->ipq_frags; 1009 fp->ipq_frags = m; 1010 } 1011 1012 /* 1013 * While we overlap succeeding segments trim them or, 1014 * if they are completely covered, dequeue them. 1015 */ 1016 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1017 q = nq) { 1018 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1019 if (i < GETIP(q)->ip_len) { 1020 GETIP(q)->ip_len -= i; 1021 GETIP(q)->ip_off += i; 1022 m_adj(q, i); 1023 q->m_pkthdr.csum_flags = 0; 1024 break; 1025 } 1026 nq = q->m_nextpkt; 1027 m->m_nextpkt = nq; 1028 IPSTAT_INC(ips_fragdropped); 1029 fp->ipq_nfrags--; 1030 m_freem(q); 1031 } 1032 1033 /* 1034 * Check for complete reassembly and perform frag per packet 1035 * limiting. 1036 * 1037 * Frag limiting is performed here so that the nth frag has 1038 * a chance to complete the packet before we drop the packet. 1039 * As a result, n+1 frags are actually allowed per packet, but 1040 * only n will ever be stored. (n = maxfragsperpacket.) 1041 * 1042 */ 1043 next = 0; 1044 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1045 if (GETIP(q)->ip_off != next) { 1046 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1047 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1048 ip_freef(head, fp); 1049 } 1050 goto done; 1051 } 1052 next += GETIP(q)->ip_len; 1053 } 1054 /* Make sure the last packet didn't have the IP_MF flag */ 1055 if (p->m_flags & M_FRAG) { 1056 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1057 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1058 ip_freef(head, fp); 1059 } 1060 goto done; 1061 } 1062 1063 /* 1064 * Reassembly is complete. Make sure the packet is a sane size. 1065 */ 1066 q = fp->ipq_frags; 1067 ip = GETIP(q); 1068 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1069 IPSTAT_INC(ips_toolong); 1070 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1071 ip_freef(head, fp); 1072 goto done; 1073 } 1074 1075 /* 1076 * Concatenate fragments. 1077 */ 1078 m = q; 1079 t = m->m_next; 1080 m->m_next = NULL; 1081 m_cat(m, t); 1082 nq = q->m_nextpkt; 1083 q->m_nextpkt = NULL; 1084 for (q = nq; q != NULL; q = nq) { 1085 nq = q->m_nextpkt; 1086 q->m_nextpkt = NULL; 1087 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1088 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1089 m_cat(m, q); 1090 } 1091 /* 1092 * In order to do checksumming faster we do 'end-around carry' here 1093 * (and not in for{} loop), though it implies we are not going to 1094 * reassemble more than 64k fragments. 1095 */ 1096 m->m_pkthdr.csum_data = 1097 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1098#ifdef MAC 1099 mac_ipq_reassemble(fp, m); 1100 mac_ipq_destroy(fp); 1101#endif 1102 1103 /* 1104 * Create header for new ip packet by modifying header of first 1105 * packet; dequeue and discard fragment reassembly header. 1106 * Make header visible. 1107 */ 1108 ip->ip_len = (ip->ip_hl << 2) + next; 1109 ip->ip_src = fp->ipq_src; 1110 ip->ip_dst = fp->ipq_dst; 1111 TAILQ_REMOVE(head, fp, ipq_list); 1112 V_nipq--; 1113 uma_zfree(V_ipq_zone, fp); 1114 m->m_len += (ip->ip_hl << 2); 1115 m->m_data -= (ip->ip_hl << 2); 1116 /* some debugging cruft by sklower, below, will go away soon */ 1117 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1118 m_fixhdr(m); 1119 IPSTAT_INC(ips_reassembled); 1120 IPQ_UNLOCK(); 1121 return (m); 1122 1123dropfrag: 1124 IPSTAT_INC(ips_fragdropped); 1125 if (fp != NULL) 1126 fp->ipq_nfrags--; 1127 m_freem(m); 1128done: 1129 IPQ_UNLOCK(); 1130 return (NULL); 1131 1132#undef GETIP 1133} 1134 1135/* 1136 * Free a fragment reassembly header and all 1137 * associated datagrams. 1138 */ 1139static void 1140ip_freef(struct ipqhead *fhp, struct ipq *fp) 1141{ 1142 INIT_VNET_INET(curvnet); 1143 struct mbuf *q; 1144 1145 IPQ_LOCK_ASSERT(); 1146 1147 while (fp->ipq_frags) { 1148 q = fp->ipq_frags; 1149 fp->ipq_frags = q->m_nextpkt; 1150 m_freem(q); 1151 } 1152 TAILQ_REMOVE(fhp, fp, ipq_list); 1153 uma_zfree(V_ipq_zone, fp); 1154 V_nipq--; 1155} 1156 1157/* 1158 * IP timer processing; 1159 * if a timer expires on a reassembly 1160 * queue, discard it. 1161 */ 1162void 1163ip_slowtimo(void) 1164{ 1165 VNET_ITERATOR_DECL(vnet_iter); 1166 struct ipq *fp; 1167 int i; 1168 1169 IPQ_LOCK(); 1170 VNET_LIST_RLOCK(); 1171 VNET_FOREACH(vnet_iter) { 1172 CURVNET_SET(vnet_iter); 1173 INIT_VNET_INET(vnet_iter); 1174 for (i = 0; i < IPREASS_NHASH; i++) { 1175 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) { 1176 struct ipq *fpp; 1177 1178 fpp = fp; 1179 fp = TAILQ_NEXT(fp, ipq_list); 1180 if(--fpp->ipq_ttl == 0) { 1181 IPSTAT_ADD(ips_fragtimeout, 1182 fpp->ipq_nfrags); 1183 ip_freef(&V_ipq[i], fpp); 1184 } 1185 } 1186 } 1187 /* 1188 * If we are over the maximum number of fragments 1189 * (due to the limit being lowered), drain off 1190 * enough to get down to the new limit. 1191 */ 1192 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) { 1193 for (i = 0; i < IPREASS_NHASH; i++) { 1194 while (V_nipq > V_maxnipq && 1195 !TAILQ_EMPTY(&V_ipq[i])) { 1196 IPSTAT_ADD(ips_fragdropped, 1197 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1198 ip_freef(&V_ipq[i], 1199 TAILQ_FIRST(&V_ipq[i])); 1200 } 1201 } 1202 } 1203 CURVNET_RESTORE(); 1204 } 1205 VNET_LIST_RUNLOCK(); 1206 IPQ_UNLOCK(); 1207} 1208 1209/* 1210 * Drain off all datagram fragments. 1211 */ 1212void 1213ip_drain(void) 1214{ 1215 VNET_ITERATOR_DECL(vnet_iter); 1216 int i; 1217 1218 IPQ_LOCK(); 1219 VNET_LIST_RLOCK(); 1220 VNET_FOREACH(vnet_iter) { 1221 CURVNET_SET(vnet_iter); 1222 INIT_VNET_INET(vnet_iter); 1223 for (i = 0; i < IPREASS_NHASH; i++) { 1224 while(!TAILQ_EMPTY(&V_ipq[i])) { 1225 IPSTAT_ADD(ips_fragdropped, 1226 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1227 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); 1228 } 1229 } 1230 CURVNET_RESTORE(); 1231 } 1232 VNET_LIST_RUNLOCK(); 1233 IPQ_UNLOCK(); 1234 in_rtqdrain(); 1235} 1236 1237/* 1238 * The protocol to be inserted into ip_protox[] must be already registered 1239 * in inetsw[], either statically or through pf_proto_register(). 1240 */ 1241int 1242ipproto_register(u_char ipproto) 1243{ 1244 struct protosw *pr; 1245 1246 /* Sanity checks. */ 1247 if (ipproto == 0) 1248 return (EPROTONOSUPPORT); 1249 1250 /* 1251 * The protocol slot must not be occupied by another protocol 1252 * already. An index pointing to IPPROTO_RAW is unused. 1253 */ 1254 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1255 if (pr == NULL) 1256 return (EPFNOSUPPORT); 1257 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1258 return (EEXIST); 1259 1260 /* Find the protocol position in inetsw[] and set the index. */ 1261 for (pr = inetdomain.dom_protosw; 1262 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1263 if (pr->pr_domain->dom_family == PF_INET && 1264 pr->pr_protocol && pr->pr_protocol == ipproto) { 1265 /* Be careful to only index valid IP protocols. */ 1266 if (pr->pr_protocol < IPPROTO_MAX) { 1267 ip_protox[pr->pr_protocol] = pr - inetsw; 1268 return (0); 1269 } else 1270 return (EINVAL); 1271 } 1272 } 1273 return (EPROTONOSUPPORT); 1274} 1275 1276int 1277ipproto_unregister(u_char ipproto) 1278{ 1279 struct protosw *pr; 1280 1281 /* Sanity checks. */ 1282 if (ipproto == 0) 1283 return (EPROTONOSUPPORT); 1284 1285 /* Check if the protocol was indeed registered. */ 1286 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1287 if (pr == NULL) 1288 return (EPFNOSUPPORT); 1289 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1290 return (ENOENT); 1291 1292 /* Reset the protocol slot to IPPROTO_RAW. */ 1293 ip_protox[ipproto] = pr - inetsw; 1294 return (0); 1295} 1296 1297/* 1298 * Given address of next destination (final or next hop), 1299 * return internet address info of interface to be used to get there. 1300 */ 1301struct in_ifaddr * 1302ip_rtaddr(struct in_addr dst, u_int fibnum) 1303{ 1304 struct route sro; 1305 struct sockaddr_in *sin; 1306 struct in_ifaddr *ifa; 1307 1308 bzero(&sro, sizeof(sro)); 1309 sin = (struct sockaddr_in *)&sro.ro_dst; 1310 sin->sin_family = AF_INET; 1311 sin->sin_len = sizeof(*sin); 1312 sin->sin_addr = dst; 1313 in_rtalloc_ign(&sro, 0, fibnum); 1314 1315 if (sro.ro_rt == NULL) 1316 return (NULL); 1317 1318 ifa = ifatoia(sro.ro_rt->rt_ifa); 1319 RTFREE(sro.ro_rt); 1320 return (ifa); 1321} 1322 1323u_char inetctlerrmap[PRC_NCMDS] = { 1324 0, 0, 0, 0, 1325 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1326 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1327 EMSGSIZE, EHOSTUNREACH, 0, 0, 1328 0, 0, EHOSTUNREACH, 0, 1329 ENOPROTOOPT, ECONNREFUSED 1330}; 1331 1332/* 1333 * Forward a packet. If some error occurs return the sender 1334 * an icmp packet. Note we can't always generate a meaningful 1335 * icmp message because icmp doesn't have a large enough repertoire 1336 * of codes and types. 1337 * 1338 * If not forwarding, just drop the packet. This could be confusing 1339 * if ipforwarding was zero but some routing protocol was advancing 1340 * us as a gateway to somewhere. However, we must let the routing 1341 * protocol deal with that. 1342 * 1343 * The srcrt parameter indicates whether the packet is being forwarded 1344 * via a source route. 1345 */ 1346void 1347ip_forward(struct mbuf *m, int srcrt) 1348{ 1349 INIT_VNET_INET(curvnet); 1350 struct ip *ip = mtod(m, struct ip *); 1351 struct in_ifaddr *ia = NULL; 1352 struct mbuf *mcopy; 1353 struct in_addr dest; 1354 struct route ro; 1355 int error, type = 0, code = 0, mtu = 0; 1356 1357 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1358 IPSTAT_INC(ips_cantforward); 1359 m_freem(m); 1360 return; 1361 } 1362#ifdef IPSTEALTH 1363 if (!V_ipstealth) { 1364#endif 1365 if (ip->ip_ttl <= IPTTLDEC) { 1366 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1367 0, 0); 1368 return; 1369 } 1370#ifdef IPSTEALTH 1371 } 1372#endif 1373 1374 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); 1375 if (!srcrt && ia == NULL) { 1376 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1377 return; 1378 } 1379 1380 /* 1381 * Save the IP header and at most 8 bytes of the payload, 1382 * in case we need to generate an ICMP message to the src. 1383 * 1384 * XXX this can be optimized a lot by saving the data in a local 1385 * buffer on the stack (72 bytes at most), and only allocating the 1386 * mbuf if really necessary. The vast majority of the packets 1387 * are forwarded without having to send an ICMP back (either 1388 * because unnecessary, or because rate limited), so we are 1389 * really we are wasting a lot of work here. 1390 * 1391 * We don't use m_copy() because it might return a reference 1392 * to a shared cluster. Both this function and ip_output() 1393 * assume exclusive access to the IP header in `m', so any 1394 * data in a cluster may change before we reach icmp_error(). 1395 */ 1396 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1397 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1398 /* 1399 * It's probably ok if the pkthdr dup fails (because 1400 * the deep copy of the tag chain failed), but for now 1401 * be conservative and just discard the copy since 1402 * code below may some day want the tags. 1403 */ 1404 m_free(mcopy); 1405 mcopy = NULL; 1406 } 1407 if (mcopy != NULL) { 1408 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy)); 1409 mcopy->m_pkthdr.len = mcopy->m_len; 1410 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1411 } 1412 1413#ifdef IPSTEALTH 1414 if (!V_ipstealth) { 1415#endif 1416 ip->ip_ttl -= IPTTLDEC; 1417#ifdef IPSTEALTH 1418 } 1419#endif 1420 1421 /* 1422 * If forwarding packet using same interface that it came in on, 1423 * perhaps should send a redirect to sender to shortcut a hop. 1424 * Only send redirect if source is sending directly to us, 1425 * and if packet was not source routed (or has any options). 1426 * Also, don't send redirect if forwarding using a default route 1427 * or a route modified by a redirect. 1428 */ 1429 dest.s_addr = 0; 1430 if (!srcrt && V_ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) { 1431 struct sockaddr_in *sin; 1432 struct rtentry *rt; 1433 1434 bzero(&ro, sizeof(ro)); 1435 sin = (struct sockaddr_in *)&ro.ro_dst; 1436 sin->sin_family = AF_INET; 1437 sin->sin_len = sizeof(*sin); 1438 sin->sin_addr = ip->ip_dst; 1439 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 1440 1441 rt = ro.ro_rt; 1442 1443 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1444 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1445#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1446 u_long src = ntohl(ip->ip_src.s_addr); 1447 1448 if (RTA(rt) && 1449 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1450 if (rt->rt_flags & RTF_GATEWAY) 1451 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1452 else 1453 dest.s_addr = ip->ip_dst.s_addr; 1454 /* Router requirements says to only send host redirects */ 1455 type = ICMP_REDIRECT; 1456 code = ICMP_REDIRECT_HOST; 1457 } 1458 } 1459 if (rt) 1460 RTFREE(rt); 1461 } 1462 1463 /* 1464 * Try to cache the route MTU from ip_output so we can consider it for 1465 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. 1466 */ 1467 bzero(&ro, sizeof(ro)); 1468 1469 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1470 1471 if (error == EMSGSIZE && ro.ro_rt) 1472 mtu = ro.ro_rt->rt_rmx.rmx_mtu; 1473 if (ro.ro_rt) 1474 RTFREE(ro.ro_rt); 1475 1476 if (error) 1477 IPSTAT_INC(ips_cantforward); 1478 else { 1479 IPSTAT_INC(ips_forward); 1480 if (type) 1481 IPSTAT_INC(ips_redirectsent); 1482 else { 1483 if (mcopy) 1484 m_freem(mcopy); 1485 return; 1486 } 1487 } 1488 if (mcopy == NULL) 1489 return; 1490 1491 switch (error) { 1492 1493 case 0: /* forwarded, but need redirect */ 1494 /* type, code set above */ 1495 break; 1496 1497 case ENETUNREACH: /* shouldn't happen, checked above */ 1498 case EHOSTUNREACH: 1499 case ENETDOWN: 1500 case EHOSTDOWN: 1501 default: 1502 type = ICMP_UNREACH; 1503 code = ICMP_UNREACH_HOST; 1504 break; 1505 1506 case EMSGSIZE: 1507 type = ICMP_UNREACH; 1508 code = ICMP_UNREACH_NEEDFRAG; 1509 1510#ifdef IPSEC 1511 /* 1512 * If IPsec is configured for this path, 1513 * override any possibly mtu value set by ip_output. 1514 */ 1515 mtu = ip_ipsec_mtu(m, mtu); 1516#endif /* IPSEC */ 1517 /* 1518 * If the MTU was set before make sure we are below the 1519 * interface MTU. 1520 * If the MTU wasn't set before use the interface mtu or 1521 * fall back to the next smaller mtu step compared to the 1522 * current packet size. 1523 */ 1524 if (mtu != 0) { 1525 if (ia != NULL) 1526 mtu = min(mtu, ia->ia_ifp->if_mtu); 1527 } else { 1528 if (ia != NULL) 1529 mtu = ia->ia_ifp->if_mtu; 1530 else 1531 mtu = ip_next_mtu(ip->ip_len, 0); 1532 } 1533 IPSTAT_INC(ips_cantfrag); 1534 break; 1535 1536 case ENOBUFS: 1537 /* 1538 * A router should not generate ICMP_SOURCEQUENCH as 1539 * required in RFC1812 Requirements for IP Version 4 Routers. 1540 * Source quench could be a big problem under DoS attacks, 1541 * or if the underlying interface is rate-limited. 1542 * Those who need source quench packets may re-enable them 1543 * via the net.inet.ip.sendsourcequench sysctl. 1544 */ 1545 if (V_ip_sendsourcequench == 0) { 1546 m_freem(mcopy); 1547 return; 1548 } else { 1549 type = ICMP_SOURCEQUENCH; 1550 code = 0; 1551 } 1552 break; 1553 1554 case EACCES: /* ipfw denied packet */ 1555 m_freem(mcopy); 1556 return; 1557 } 1558 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1559} 1560 1561void 1562ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1563 struct mbuf *m) 1564{ 1565 INIT_VNET_NET(inp->inp_vnet); 1566 1567 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1568 struct bintime bt; 1569 1570 bintime(&bt); 1571 if (inp->inp_socket->so_options & SO_BINTIME) { 1572 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1573 SCM_BINTIME, SOL_SOCKET); 1574 if (*mp) 1575 mp = &(*mp)->m_next; 1576 } 1577 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1578 struct timeval tv; 1579 1580 bintime2timeval(&bt, &tv); 1581 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1582 SCM_TIMESTAMP, SOL_SOCKET); 1583 if (*mp) 1584 mp = &(*mp)->m_next; 1585 } 1586 } 1587 if (inp->inp_flags & INP_RECVDSTADDR) { 1588 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1589 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1590 if (*mp) 1591 mp = &(*mp)->m_next; 1592 } 1593 if (inp->inp_flags & INP_RECVTTL) { 1594 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1595 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1596 if (*mp) 1597 mp = &(*mp)->m_next; 1598 } 1599#ifdef notyet 1600 /* XXX 1601 * Moving these out of udp_input() made them even more broken 1602 * than they already were. 1603 */ 1604 /* options were tossed already */ 1605 if (inp->inp_flags & INP_RECVOPTS) { 1606 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1607 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1608 if (*mp) 1609 mp = &(*mp)->m_next; 1610 } 1611 /* ip_srcroute doesn't do what we want here, need to fix */ 1612 if (inp->inp_flags & INP_RECVRETOPTS) { 1613 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1614 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1615 if (*mp) 1616 mp = &(*mp)->m_next; 1617 } 1618#endif 1619 if (inp->inp_flags & INP_RECVIF) { 1620 struct ifnet *ifp; 1621 struct sdlbuf { 1622 struct sockaddr_dl sdl; 1623 u_char pad[32]; 1624 } sdlbuf; 1625 struct sockaddr_dl *sdp; 1626 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1627 1628 if (((ifp = m->m_pkthdr.rcvif)) 1629 && ( ifp->if_index && (ifp->if_index <= V_if_index))) { 1630 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1631 /* 1632 * Change our mind and don't try copy. 1633 */ 1634 if ((sdp->sdl_family != AF_LINK) 1635 || (sdp->sdl_len > sizeof(sdlbuf))) { 1636 goto makedummy; 1637 } 1638 bcopy(sdp, sdl2, sdp->sdl_len); 1639 } else { 1640makedummy: 1641 sdl2->sdl_len 1642 = offsetof(struct sockaddr_dl, sdl_data[0]); 1643 sdl2->sdl_family = AF_LINK; 1644 sdl2->sdl_index = 0; 1645 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1646 } 1647 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1648 IP_RECVIF, IPPROTO_IP); 1649 if (*mp) 1650 mp = &(*mp)->m_next; 1651 } 1652} 1653 1654/* 1655 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1656 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1657 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1658 * compiled. 1659 */ 1660int 1661ip_rsvp_init(struct socket *so) 1662{ 1663 INIT_VNET_INET(so->so_vnet); 1664 1665 if (so->so_type != SOCK_RAW || 1666 so->so_proto->pr_protocol != IPPROTO_RSVP) 1667 return EOPNOTSUPP; 1668 1669 if (V_ip_rsvpd != NULL) 1670 return EADDRINUSE; 1671 1672 V_ip_rsvpd = so; 1673 /* 1674 * This may seem silly, but we need to be sure we don't over-increment 1675 * the RSVP counter, in case something slips up. 1676 */ 1677 if (!V_ip_rsvp_on) { 1678 V_ip_rsvp_on = 1; 1679 V_rsvp_on++; 1680 } 1681 1682 return 0; 1683} 1684 1685int 1686ip_rsvp_done(void) 1687{ 1688 INIT_VNET_INET(curvnet); 1689 1690 V_ip_rsvpd = NULL; 1691 /* 1692 * This may seem silly, but we need to be sure we don't over-decrement 1693 * the RSVP counter, in case something slips up. 1694 */ 1695 if (V_ip_rsvp_on) { 1696 V_ip_rsvp_on = 0; 1697 V_rsvp_on--; 1698 } 1699 return 0; 1700} 1701 1702void 1703rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1704{ 1705 INIT_VNET_INET(curvnet); 1706 1707 if (rsvp_input_p) { /* call the real one if loaded */ 1708 rsvp_input_p(m, off); 1709 return; 1710 } 1711 1712 /* Can still get packets with rsvp_on = 0 if there is a local member 1713 * of the group to which the RSVP packet is addressed. But in this 1714 * case we want to throw the packet away. 1715 */ 1716 1717 if (!V_rsvp_on) { 1718 m_freem(m); 1719 return; 1720 } 1721 1722 if (V_ip_rsvpd != NULL) { 1723 rip_input(m, off); 1724 return; 1725 } 1726 /* Drop the packet */ 1727 m_freem(m); 1728} 1729