ip_input.c revision 194835
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 194835 2009-06-24 14:29:40Z rwatson $"); 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_carp.h" 41 42#include <sys/param.h> 43#include <sys/systm.h> 44#include <sys/callout.h> 45#include <sys/mbuf.h> 46#include <sys/malloc.h> 47#include <sys/domain.h> 48#include <sys/protosw.h> 49#include <sys/socket.h> 50#include <sys/time.h> 51#include <sys/kernel.h> 52#include <sys/lock.h> 53#include <sys/rwlock.h> 54#include <sys/syslog.h> 55#include <sys/sysctl.h> 56#include <sys/vimage.h> 57 58#include <net/pfil.h> 59#include <net/if.h> 60#include <net/if_types.h> 61#include <net/if_var.h> 62#include <net/if_dl.h> 63#include <net/route.h> 64#include <net/netisr.h> 65#include <net/vnet.h> 66#include <net/flowtable.h> 67 68#include <netinet/in.h> 69#include <netinet/in_systm.h> 70#include <netinet/in_var.h> 71#include <netinet/ip.h> 72#include <netinet/in_pcb.h> 73#include <netinet/ip_var.h> 74#include <netinet/ip_icmp.h> 75#include <netinet/ip_options.h> 76#include <machine/in_cksum.h> 77#include <netinet/vinet.h> 78#ifdef DEV_CARP 79#include <netinet/ip_carp.h> 80#endif 81#ifdef IPSEC 82#include <netinet/ip_ipsec.h> 83#endif /* IPSEC */ 84 85#include <sys/socketvar.h> 86 87#include <security/mac/mac_framework.h> 88 89#ifdef CTASSERT 90CTASSERT(sizeof(struct ip) == 20); 91#endif 92 93#ifndef VIMAGE 94#ifndef VIMAGE_GLOBALS 95struct vnet_inet vnet_inet_0; 96#endif 97#endif 98 99#ifdef VIMAGE_GLOBALS 100static int ipsendredirects; 101static int ip_checkinterface; 102static int ip_keepfaith; 103static int ip_sendsourcequench; 104int ip_defttl; 105int ip_do_randomid; 106int ipforwarding; 107struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 108struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 109u_long in_ifaddrhmask; /* mask for hash table */ 110struct ipstat ipstat; 111static int ip_rsvp_on; 112struct socket *ip_rsvpd; 113int rsvp_on; 114static struct ipqhead ipq[IPREASS_NHASH]; 115static int maxnipq; /* Administrative limit on # reass queues. */ 116static int maxfragsperpacket; 117int ipstealth; 118static int nipq; /* Total # of reass queues */ 119#endif 120 121SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_FORWARDING, 122 forwarding, CTLFLAG_RW, ipforwarding, 0, 123 "Enable IP forwarding between interfaces"); 124 125SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_SENDREDIRECTS, 126 redirect, CTLFLAG_RW, ipsendredirects, 0, 127 "Enable sending IP redirects"); 128 129SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_DEFTTL, 130 ttl, CTLFLAG_RW, ip_defttl, 0, "Maximum TTL on IP packets"); 131 132SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_KEEPFAITH, 133 keepfaith, CTLFLAG_RW, ip_keepfaith, 0, 134 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 135 136SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, 137 sendsourcequench, CTLFLAG_RW, ip_sendsourcequench, 0, 138 "Enable the transmission of source quench packets"); 139 140SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, random_id, 141 CTLFLAG_RW, ip_do_randomid, 0, "Assign random ip_id values"); 142 143/* 144 * XXX - Setting ip_checkinterface mostly implements the receive side of 145 * the Strong ES model described in RFC 1122, but since the routing table 146 * and transmit implementation do not implement the Strong ES model, 147 * setting this to 1 results in an odd hybrid. 148 * 149 * XXX - ip_checkinterface currently must be disabled if you use ipnat 150 * to translate the destination address to another local interface. 151 * 152 * XXX - ip_checkinterface must be disabled if you add IP aliases 153 * to the loopback interface instead of the interface where the 154 * packets for those addresses are received. 155 */ 156SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, 157 check_interface, CTLFLAG_RW, ip_checkinterface, 0, 158 "Verify packet arrives on correct interface"); 159 160struct pfil_head inet_pfil_hook; /* Packet filter hooks */ 161 162static struct netisr_handler ip_nh = { 163 .nh_name = "ip", 164 .nh_handler = ip_input, 165 .nh_proto = NETISR_IP, 166 .nh_policy = NETISR_POLICY_FLOW, 167}; 168 169extern struct domain inetdomain; 170extern struct protosw inetsw[]; 171u_char ip_protox[IPPROTO_MAX]; 172 173 174SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 175 ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 176 177#ifdef VIMAGE_GLOBALS 178static uma_zone_t ipq_zone; 179#endif 180static struct mtx ipqlock; 181 182#define IPQ_LOCK() mtx_lock(&ipqlock) 183#define IPQ_UNLOCK() mtx_unlock(&ipqlock) 184#define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 185#define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 186 187static void maxnipq_update(void); 188static void ipq_zone_change(void *); 189 190SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, fragpackets, 191 CTLFLAG_RD, nipq, 0, 192 "Current number of IPv4 fragment reassembly queue entries"); 193 194SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, maxfragsperpacket, 195 CTLFLAG_RW, maxfragsperpacket, 0, 196 "Maximum number of IPv4 fragments allowed per packet"); 197 198struct callout ipport_tick_callout; 199 200#ifdef IPCTL_DEFMTU 201SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 202 &ip_mtu, 0, "Default MTU"); 203#endif 204 205#ifdef IPSTEALTH 206SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 207 ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding"); 208#endif 209#ifdef FLOWTABLE 210#ifdef VIMAGE_GLOBALS 211static int ip_output_flowtable_size; 212struct flowtable *ip_ft; 213#endif 214SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, output_flowtable_size, 215 CTLFLAG_RDTUN, ip_output_flowtable_size, 2048, 216 "number of entries in the per-cpu output flow caches"); 217#endif 218 219#ifdef VIMAGE_GLOBALS 220int fw_one_pass; 221#endif 222 223static void ip_freef(struct ipqhead *, struct ipq *); 224 225#ifndef VIMAGE_GLOBALS 226static void vnet_inet_register(void); 227 228static const vnet_modinfo_t vnet_inet_modinfo = { 229 .vmi_id = VNET_MOD_INET, 230 .vmi_name = "inet", 231 .vmi_size = sizeof(struct vnet_inet) 232}; 233 234static void vnet_inet_register() 235{ 236 237 vnet_mod_register(&vnet_inet_modinfo); 238} 239 240SYSINIT(inet, SI_SUB_PROTO_BEGIN, SI_ORDER_FIRST, vnet_inet_register, 0); 241#endif 242 243static int 244sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 245{ 246 int error, qlimit; 247 248 netisr_getqlimit(&ip_nh, &qlimit); 249 error = sysctl_handle_int(oidp, &qlimit, 0, req); 250 if (error || !req->newptr) 251 return (error); 252 if (qlimit < 1) 253 return (EINVAL); 254 return (netisr_setqlimit(&ip_nh, qlimit)); 255} 256SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, 257 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I", 258 "Maximum size of the IP input queue"); 259 260static int 261sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) 262{ 263 u_int64_t qdrops_long; 264 int error, qdrops; 265 266 netisr_getqdrops(&ip_nh, &qdrops_long); 267 qdrops = qdrops_long; 268 error = sysctl_handle_int(oidp, &qdrops, 0, req); 269 if (error || !req->newptr) 270 return (error); 271 if (qdrops != 0) 272 return (EINVAL); 273 netisr_clearqdrops(&ip_nh); 274 return (0); 275} 276 277SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, 278 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I", 279 "Number of packets dropped from the IP input queue"); 280 281/* 282 * IP initialization: fill in IP protocol switch table. 283 * All protocols not implemented in kernel go to raw IP protocol handler. 284 */ 285void 286ip_init(void) 287{ 288 INIT_VNET_INET(curvnet); 289 struct protosw *pr; 290 int i; 291 292 V_ipsendredirects = 1; /* XXX */ 293 V_ip_checkinterface = 0; 294 V_ip_keepfaith = 0; 295 V_ip_sendsourcequench = 0; 296 V_rsvp_on = 0; 297 V_ip_defttl = IPDEFTTL; 298 V_ip_do_randomid = 0; 299 V_ip_id = time_second & 0xffff; 300 V_ipforwarding = 0; 301 V_ipstealth = 0; 302 V_nipq = 0; /* Total # of reass queues */ 303 304 V_ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */ 305 V_ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */ 306 V_ipport_firstauto = IPPORT_EPHEMERALFIRST; /* 10000 */ 307 V_ipport_lastauto = IPPORT_EPHEMERALLAST; /* 65535 */ 308 V_ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */ 309 V_ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */ 310 V_ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */ 311 V_ipport_reservedlow = 0; 312 V_ipport_randomized = 1; /* user controlled via sysctl */ 313 V_ipport_randomcps = 10; /* user controlled via sysctl */ 314 V_ipport_randomtime = 45; /* user controlled via sysctl */ 315 V_ipport_stoprandom = 0; /* toggled by ipport_tick */ 316 317 V_fw_one_pass = 1; 318 319#ifdef NOTYET 320 /* XXX global static but not instantiated in this file */ 321 V_ipfastforward_active = 0; 322 V_subnetsarelocal = 0; 323 V_sameprefixcarponly = 0; 324#endif 325 326 TAILQ_INIT(&V_in_ifaddrhead); 327 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 328 329 /* Initialize IP reassembly queue. */ 330 for (i = 0; i < IPREASS_NHASH; i++) 331 TAILQ_INIT(&V_ipq[i]); 332 V_maxnipq = nmbclusters / 32; 333 V_maxfragsperpacket = 16; 334 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 335 NULL, UMA_ALIGN_PTR, 0); 336 maxnipq_update(); 337 338#ifdef FLOWTABLE 339 V_ip_output_flowtable_size = 2048; 340 TUNABLE_INT_FETCH("net.inet.ip.output_flowtable_size", 341 &V_ip_output_flowtable_size); 342 V_ip_ft = flowtable_alloc(V_ip_output_flowtable_size, FL_PCPU); 343#endif 344 345 /* Skip initialization of globals for non-default instances. */ 346 if (!IS_DEFAULT_VNET(curvnet)) 347 return; 348 349 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 350 if (pr == NULL) 351 panic("ip_init: PF_INET not found"); 352 353 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 354 for (i = 0; i < IPPROTO_MAX; i++) 355 ip_protox[i] = pr - inetsw; 356 /* 357 * Cycle through IP protocols and put them into the appropriate place 358 * in ip_protox[]. 359 */ 360 for (pr = inetdomain.dom_protosw; 361 pr < inetdomain.dom_protoswNPROTOSW; pr++) 362 if (pr->pr_domain->dom_family == PF_INET && 363 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 364 /* Be careful to only index valid IP protocols. */ 365 if (pr->pr_protocol < IPPROTO_MAX) 366 ip_protox[pr->pr_protocol] = pr - inetsw; 367 } 368 369 /* Initialize packet filter hooks. */ 370 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 371 inet_pfil_hook.ph_af = AF_INET; 372 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 373 printf("%s: WARNING: unable to register pfil hook, " 374 "error %d\n", __func__, i); 375 376 /* Start ipport_tick. */ 377 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE); 378 callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL); 379 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, 380 SHUTDOWN_PRI_DEFAULT); 381 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, 382 NULL, EVENTHANDLER_PRI_ANY); 383 384 /* Initialize various other remaining things. */ 385 IPQ_LOCK_INIT(); 386 netisr_register(&ip_nh); 387} 388 389void 390ip_fini(void *xtp) 391{ 392 393 callout_stop(&ipport_tick_callout); 394} 395 396/* 397 * Ip input routine. Checksum and byte swap header. If fragmented 398 * try to reassemble. Process options. Pass to next level. 399 */ 400void 401ip_input(struct mbuf *m) 402{ 403 INIT_VNET_INET(curvnet); 404 struct ip *ip = NULL; 405 struct in_ifaddr *ia = NULL; 406 struct ifaddr *ifa; 407 struct ifnet *ifp; 408 int checkif, hlen = 0; 409 u_short sum; 410 int dchg = 0; /* dest changed after fw */ 411 struct in_addr odst; /* original dst address */ 412 413 M_ASSERTPKTHDR(m); 414 415 if (m->m_flags & M_FASTFWD_OURS) { 416 /* 417 * Firewall or NAT changed destination to local. 418 * We expect ip_len and ip_off to be in host byte order. 419 */ 420 m->m_flags &= ~M_FASTFWD_OURS; 421 /* Set up some basics that will be used later. */ 422 ip = mtod(m, struct ip *); 423 hlen = ip->ip_hl << 2; 424 goto ours; 425 } 426 427 IPSTAT_INC(ips_total); 428 429 if (m->m_pkthdr.len < sizeof(struct ip)) 430 goto tooshort; 431 432 if (m->m_len < sizeof (struct ip) && 433 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 434 IPSTAT_INC(ips_toosmall); 435 return; 436 } 437 ip = mtod(m, struct ip *); 438 439 if (ip->ip_v != IPVERSION) { 440 IPSTAT_INC(ips_badvers); 441 goto bad; 442 } 443 444 hlen = ip->ip_hl << 2; 445 if (hlen < sizeof(struct ip)) { /* minimum header length */ 446 IPSTAT_INC(ips_badhlen); 447 goto bad; 448 } 449 if (hlen > m->m_len) { 450 if ((m = m_pullup(m, hlen)) == NULL) { 451 IPSTAT_INC(ips_badhlen); 452 return; 453 } 454 ip = mtod(m, struct ip *); 455 } 456 457 /* 127/8 must not appear on wire - RFC1122 */ 458 ifp = m->m_pkthdr.rcvif; 459 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 460 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 461 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 462 IPSTAT_INC(ips_badaddr); 463 goto bad; 464 } 465 } 466 467 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 468 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 469 } else { 470 if (hlen == sizeof(struct ip)) { 471 sum = in_cksum_hdr(ip); 472 } else { 473 sum = in_cksum(m, hlen); 474 } 475 } 476 if (sum) { 477 IPSTAT_INC(ips_badsum); 478 goto bad; 479 } 480 481#ifdef ALTQ 482 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 483 /* packet is dropped by traffic conditioner */ 484 return; 485#endif 486 487 /* 488 * Convert fields to host representation. 489 */ 490 ip->ip_len = ntohs(ip->ip_len); 491 if (ip->ip_len < hlen) { 492 IPSTAT_INC(ips_badlen); 493 goto bad; 494 } 495 ip->ip_off = ntohs(ip->ip_off); 496 497 /* 498 * Check that the amount of data in the buffers 499 * is as at least much as the IP header would have us expect. 500 * Trim mbufs if longer than we expect. 501 * Drop packet if shorter than we expect. 502 */ 503 if (m->m_pkthdr.len < ip->ip_len) { 504tooshort: 505 IPSTAT_INC(ips_tooshort); 506 goto bad; 507 } 508 if (m->m_pkthdr.len > ip->ip_len) { 509 if (m->m_len == m->m_pkthdr.len) { 510 m->m_len = ip->ip_len; 511 m->m_pkthdr.len = ip->ip_len; 512 } else 513 m_adj(m, ip->ip_len - m->m_pkthdr.len); 514 } 515#ifdef IPSEC 516 /* 517 * Bypass packet filtering for packets from a tunnel (gif). 518 */ 519 if (ip_ipsec_filtertunnel(m)) 520 goto passin; 521#endif /* IPSEC */ 522 523 /* 524 * Run through list of hooks for input packets. 525 * 526 * NB: Beware of the destination address changing (e.g. 527 * by NAT rewriting). When this happens, tell 528 * ip_forward to do the right thing. 529 */ 530 531 /* Jump over all PFIL processing if hooks are not active. */ 532 if (!PFIL_HOOKED(&inet_pfil_hook)) 533 goto passin; 534 535 odst = ip->ip_dst; 536 if (pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0) 537 return; 538 if (m == NULL) /* consumed by filter */ 539 return; 540 541 ip = mtod(m, struct ip *); 542 dchg = (odst.s_addr != ip->ip_dst.s_addr); 543 ifp = m->m_pkthdr.rcvif; 544 545#ifdef IPFIREWALL_FORWARD 546 if (m->m_flags & M_FASTFWD_OURS) { 547 m->m_flags &= ~M_FASTFWD_OURS; 548 goto ours; 549 } 550 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 551 /* 552 * Directly ship on the packet. This allows to forward packets 553 * that were destined for us to some other directly connected 554 * host. 555 */ 556 ip_forward(m, dchg); 557 return; 558 } 559#endif /* IPFIREWALL_FORWARD */ 560 561passin: 562 /* 563 * Process options and, if not destined for us, 564 * ship it on. ip_dooptions returns 1 when an 565 * error was detected (causing an icmp message 566 * to be sent and the original packet to be freed). 567 */ 568 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 569 return; 570 571 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 572 * matter if it is destined to another node, or whether it is 573 * a multicast one, RSVP wants it! and prevents it from being forwarded 574 * anywhere else. Also checks if the rsvp daemon is running before 575 * grabbing the packet. 576 */ 577 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 578 goto ours; 579 580 /* 581 * Check our list of addresses, to see if the packet is for us. 582 * If we don't have any addresses, assume any unicast packet 583 * we receive might be for us (and let the upper layers deal 584 * with it). 585 */ 586 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 587 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 588 goto ours; 589 590 /* 591 * Enable a consistency check between the destination address 592 * and the arrival interface for a unicast packet (the RFC 1122 593 * strong ES model) if IP forwarding is disabled and the packet 594 * is not locally generated and the packet is not subject to 595 * 'ipfw fwd'. 596 * 597 * XXX - Checking also should be disabled if the destination 598 * address is ipnat'ed to a different interface. 599 * 600 * XXX - Checking is incompatible with IP aliases added 601 * to the loopback interface instead of the interface where 602 * the packets are received. 603 * 604 * XXX - This is the case for carp vhost IPs as well so we 605 * insert a workaround. If the packet got here, we already 606 * checked with carp_iamatch() and carp_forus(). 607 */ 608 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 609 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && 610#ifdef DEV_CARP 611 !ifp->if_carp && 612#endif 613 (dchg == 0); 614 615 /* 616 * Check for exact addresses in the hash bucket. 617 */ 618 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 619 /* 620 * If the address matches, verify that the packet 621 * arrived via the correct interface if checking is 622 * enabled. 623 */ 624 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 625 (!checkif || ia->ia_ifp == ifp)) { 626 ifa_ref(&ia->ia_ifa); 627 goto ours; 628 } 629 } 630 /* 631 * Check for broadcast addresses. 632 * 633 * Only accept broadcast packets that arrive via the matching 634 * interface. Reception of forwarded directed broadcasts would 635 * be handled via ip_forward() and ether_output() with the loopback 636 * into the stack for SIMPLEX interfaces handled by ether_output(). 637 */ 638 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 639 IF_ADDR_LOCK(ifp); 640 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 641 if (ifa->ifa_addr->sa_family != AF_INET) 642 continue; 643 ia = ifatoia(ifa); 644 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 645 ip->ip_dst.s_addr) { 646 ifa_ref(ifa); 647 IF_ADDR_UNLOCK(ifp); 648 goto ours; 649 } 650 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) { 651 ifa_ref(ifa); 652 IF_ADDR_UNLOCK(ifp); 653 goto ours; 654 } 655#ifdef BOOTP_COMPAT 656 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { 657 ifa_ref(ifa); 658 IF_ADDR_UNLOCK(ifp); 659 goto ours; 660 } 661#endif 662 } 663 IF_ADDR_UNLOCK(ifp); 664 ia = NULL; 665 } 666 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 667 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 668 IPSTAT_INC(ips_cantforward); 669 m_freem(m); 670 return; 671 } 672 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 673 if (V_ip_mrouter) { 674 /* 675 * If we are acting as a multicast router, all 676 * incoming multicast packets are passed to the 677 * kernel-level multicast forwarding function. 678 * The packet is returned (relatively) intact; if 679 * ip_mforward() returns a non-zero value, the packet 680 * must be discarded, else it may be accepted below. 681 */ 682 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 683 IPSTAT_INC(ips_cantforward); 684 m_freem(m); 685 return; 686 } 687 688 /* 689 * The process-level routing daemon needs to receive 690 * all multicast IGMP packets, whether or not this 691 * host belongs to their destination groups. 692 */ 693 if (ip->ip_p == IPPROTO_IGMP) 694 goto ours; 695 IPSTAT_INC(ips_forward); 696 } 697 /* 698 * Assume the packet is for us, to avoid prematurely taking 699 * a lock on the in_multi hash. Protocols must perform 700 * their own filtering and update statistics accordingly. 701 */ 702 goto ours; 703 } 704 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 705 goto ours; 706 if (ip->ip_dst.s_addr == INADDR_ANY) 707 goto ours; 708 709 /* 710 * FAITH(Firewall Aided Internet Translator) 711 */ 712 if (ifp && ifp->if_type == IFT_FAITH) { 713 if (V_ip_keepfaith) { 714 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 715 goto ours; 716 } 717 m_freem(m); 718 return; 719 } 720 721 /* 722 * Not for us; forward if possible and desirable. 723 */ 724 if (V_ipforwarding == 0) { 725 IPSTAT_INC(ips_cantforward); 726 m_freem(m); 727 } else { 728#ifdef IPSEC 729 if (ip_ipsec_fwd(m)) 730 goto bad; 731#endif /* IPSEC */ 732 ip_forward(m, dchg); 733 } 734 return; 735 736ours: 737#ifdef IPSTEALTH 738 /* 739 * IPSTEALTH: Process non-routing options only 740 * if the packet is destined for us. 741 */ 742 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) { 743 if (ia != NULL) 744 ifa_free(&ia->ia_ifa); 745 return; 746 } 747#endif /* IPSTEALTH */ 748 749 /* Count the packet in the ip address stats */ 750 if (ia != NULL) { 751 ia->ia_ifa.if_ipackets++; 752 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 753 ifa_free(&ia->ia_ifa); 754 } 755 756 /* 757 * Attempt reassembly; if it succeeds, proceed. 758 * ip_reass() will return a different mbuf. 759 */ 760 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 761 m = ip_reass(m); 762 if (m == NULL) 763 return; 764 ip = mtod(m, struct ip *); 765 /* Get the header length of the reassembled packet */ 766 hlen = ip->ip_hl << 2; 767 } 768 769 /* 770 * Further protocols expect the packet length to be w/o the 771 * IP header. 772 */ 773 ip->ip_len -= hlen; 774 775#ifdef IPSEC 776 /* 777 * enforce IPsec policy checking if we are seeing last header. 778 * note that we do not visit this with protocols with pcb layer 779 * code - like udp/tcp/raw ip. 780 */ 781 if (ip_ipsec_input(m)) 782 goto bad; 783#endif /* IPSEC */ 784 785 /* 786 * Switch out to protocol's input routine. 787 */ 788 IPSTAT_INC(ips_delivered); 789 790 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 791 return; 792bad: 793 m_freem(m); 794} 795 796/* 797 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 798 * max has slightly different semantics than the sysctl, for historical 799 * reasons. 800 */ 801static void 802maxnipq_update(void) 803{ 804 INIT_VNET_INET(curvnet); 805 806 /* 807 * -1 for unlimited allocation. 808 */ 809 if (V_maxnipq < 0) 810 uma_zone_set_max(V_ipq_zone, 0); 811 /* 812 * Positive number for specific bound. 813 */ 814 if (V_maxnipq > 0) 815 uma_zone_set_max(V_ipq_zone, V_maxnipq); 816 /* 817 * Zero specifies no further fragment queue allocation -- set the 818 * bound very low, but rely on implementation elsewhere to actually 819 * prevent allocation and reclaim current queues. 820 */ 821 if (V_maxnipq == 0) 822 uma_zone_set_max(V_ipq_zone, 1); 823} 824 825static void 826ipq_zone_change(void *tag) 827{ 828 INIT_VNET_INET(curvnet); 829 830 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) { 831 V_maxnipq = nmbclusters / 32; 832 maxnipq_update(); 833 } 834} 835 836static int 837sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 838{ 839 INIT_VNET_INET(curvnet); 840 int error, i; 841 842 i = V_maxnipq; 843 error = sysctl_handle_int(oidp, &i, 0, req); 844 if (error || !req->newptr) 845 return (error); 846 847 /* 848 * XXXRW: Might be a good idea to sanity check the argument and place 849 * an extreme upper bound. 850 */ 851 if (i < -1) 852 return (EINVAL); 853 V_maxnipq = i; 854 maxnipq_update(); 855 return (0); 856} 857 858SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 859 NULL, 0, sysctl_maxnipq, "I", 860 "Maximum number of IPv4 fragment reassembly queue entries"); 861 862/* 863 * Take incoming datagram fragment and try to reassemble it into 864 * whole datagram. If the argument is the first fragment or one 865 * in between the function will return NULL and store the mbuf 866 * in the fragment chain. If the argument is the last fragment 867 * the packet will be reassembled and the pointer to the new 868 * mbuf returned for further processing. Only m_tags attached 869 * to the first packet/fragment are preserved. 870 * The IP header is *NOT* adjusted out of iplen. 871 */ 872struct mbuf * 873ip_reass(struct mbuf *m) 874{ 875 INIT_VNET_INET(curvnet); 876 struct ip *ip; 877 struct mbuf *p, *q, *nq, *t; 878 struct ipq *fp = NULL; 879 struct ipqhead *head; 880 int i, hlen, next; 881 u_int8_t ecn, ecn0; 882 u_short hash; 883 884 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 885 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) { 886 IPSTAT_INC(ips_fragments); 887 IPSTAT_INC(ips_fragdropped); 888 m_freem(m); 889 return (NULL); 890 } 891 892 ip = mtod(m, struct ip *); 893 hlen = ip->ip_hl << 2; 894 895 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 896 head = &V_ipq[hash]; 897 IPQ_LOCK(); 898 899 /* 900 * Look for queue of fragments 901 * of this datagram. 902 */ 903 TAILQ_FOREACH(fp, head, ipq_list) 904 if (ip->ip_id == fp->ipq_id && 905 ip->ip_src.s_addr == fp->ipq_src.s_addr && 906 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 907#ifdef MAC 908 mac_ipq_match(m, fp) && 909#endif 910 ip->ip_p == fp->ipq_p) 911 goto found; 912 913 fp = NULL; 914 915 /* 916 * Attempt to trim the number of allocated fragment queues if it 917 * exceeds the administrative limit. 918 */ 919 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) { 920 /* 921 * drop something from the tail of the current queue 922 * before proceeding further 923 */ 924 struct ipq *q = TAILQ_LAST(head, ipqhead); 925 if (q == NULL) { /* gak */ 926 for (i = 0; i < IPREASS_NHASH; i++) { 927 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead); 928 if (r) { 929 IPSTAT_ADD(ips_fragtimeout, 930 r->ipq_nfrags); 931 ip_freef(&V_ipq[i], r); 932 break; 933 } 934 } 935 } else { 936 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags); 937 ip_freef(head, q); 938 } 939 } 940 941found: 942 /* 943 * Adjust ip_len to not reflect header, 944 * convert offset of this to bytes. 945 */ 946 ip->ip_len -= hlen; 947 if (ip->ip_off & IP_MF) { 948 /* 949 * Make sure that fragments have a data length 950 * that's a non-zero multiple of 8 bytes. 951 */ 952 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 953 IPSTAT_INC(ips_toosmall); /* XXX */ 954 goto dropfrag; 955 } 956 m->m_flags |= M_FRAG; 957 } else 958 m->m_flags &= ~M_FRAG; 959 ip->ip_off <<= 3; 960 961 962 /* 963 * Attempt reassembly; if it succeeds, proceed. 964 * ip_reass() will return a different mbuf. 965 */ 966 IPSTAT_INC(ips_fragments); 967 m->m_pkthdr.header = ip; 968 969 /* Previous ip_reass() started here. */ 970 /* 971 * Presence of header sizes in mbufs 972 * would confuse code below. 973 */ 974 m->m_data += hlen; 975 m->m_len -= hlen; 976 977 /* 978 * If first fragment to arrive, create a reassembly queue. 979 */ 980 if (fp == NULL) { 981 fp = uma_zalloc(V_ipq_zone, M_NOWAIT); 982 if (fp == NULL) 983 goto dropfrag; 984#ifdef MAC 985 if (mac_ipq_init(fp, M_NOWAIT) != 0) { 986 uma_zfree(V_ipq_zone, fp); 987 fp = NULL; 988 goto dropfrag; 989 } 990 mac_ipq_create(m, fp); 991#endif 992 TAILQ_INSERT_HEAD(head, fp, ipq_list); 993 V_nipq++; 994 fp->ipq_nfrags = 1; 995 fp->ipq_ttl = IPFRAGTTL; 996 fp->ipq_p = ip->ip_p; 997 fp->ipq_id = ip->ip_id; 998 fp->ipq_src = ip->ip_src; 999 fp->ipq_dst = ip->ip_dst; 1000 fp->ipq_frags = m; 1001 m->m_nextpkt = NULL; 1002 goto done; 1003 } else { 1004 fp->ipq_nfrags++; 1005#ifdef MAC 1006 mac_ipq_update(m, fp); 1007#endif 1008 } 1009 1010#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1011 1012 /* 1013 * Handle ECN by comparing this segment with the first one; 1014 * if CE is set, do not lose CE. 1015 * drop if CE and not-ECT are mixed for the same packet. 1016 */ 1017 ecn = ip->ip_tos & IPTOS_ECN_MASK; 1018 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 1019 if (ecn == IPTOS_ECN_CE) { 1020 if (ecn0 == IPTOS_ECN_NOTECT) 1021 goto dropfrag; 1022 if (ecn0 != IPTOS_ECN_CE) 1023 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 1024 } 1025 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 1026 goto dropfrag; 1027 1028 /* 1029 * Find a segment which begins after this one does. 1030 */ 1031 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1032 if (GETIP(q)->ip_off > ip->ip_off) 1033 break; 1034 1035 /* 1036 * If there is a preceding segment, it may provide some of 1037 * our data already. If so, drop the data from the incoming 1038 * segment. If it provides all of our data, drop us, otherwise 1039 * stick new segment in the proper place. 1040 * 1041 * If some of the data is dropped from the the preceding 1042 * segment, then it's checksum is invalidated. 1043 */ 1044 if (p) { 1045 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1046 if (i > 0) { 1047 if (i >= ip->ip_len) 1048 goto dropfrag; 1049 m_adj(m, i); 1050 m->m_pkthdr.csum_flags = 0; 1051 ip->ip_off += i; 1052 ip->ip_len -= i; 1053 } 1054 m->m_nextpkt = p->m_nextpkt; 1055 p->m_nextpkt = m; 1056 } else { 1057 m->m_nextpkt = fp->ipq_frags; 1058 fp->ipq_frags = m; 1059 } 1060 1061 /* 1062 * While we overlap succeeding segments trim them or, 1063 * if they are completely covered, dequeue them. 1064 */ 1065 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1066 q = nq) { 1067 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1068 if (i < GETIP(q)->ip_len) { 1069 GETIP(q)->ip_len -= i; 1070 GETIP(q)->ip_off += i; 1071 m_adj(q, i); 1072 q->m_pkthdr.csum_flags = 0; 1073 break; 1074 } 1075 nq = q->m_nextpkt; 1076 m->m_nextpkt = nq; 1077 IPSTAT_INC(ips_fragdropped); 1078 fp->ipq_nfrags--; 1079 m_freem(q); 1080 } 1081 1082 /* 1083 * Check for complete reassembly and perform frag per packet 1084 * limiting. 1085 * 1086 * Frag limiting is performed here so that the nth frag has 1087 * a chance to complete the packet before we drop the packet. 1088 * As a result, n+1 frags are actually allowed per packet, but 1089 * only n will ever be stored. (n = maxfragsperpacket.) 1090 * 1091 */ 1092 next = 0; 1093 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1094 if (GETIP(q)->ip_off != next) { 1095 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1096 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1097 ip_freef(head, fp); 1098 } 1099 goto done; 1100 } 1101 next += GETIP(q)->ip_len; 1102 } 1103 /* Make sure the last packet didn't have the IP_MF flag */ 1104 if (p->m_flags & M_FRAG) { 1105 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1106 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1107 ip_freef(head, fp); 1108 } 1109 goto done; 1110 } 1111 1112 /* 1113 * Reassembly is complete. Make sure the packet is a sane size. 1114 */ 1115 q = fp->ipq_frags; 1116 ip = GETIP(q); 1117 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1118 IPSTAT_INC(ips_toolong); 1119 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1120 ip_freef(head, fp); 1121 goto done; 1122 } 1123 1124 /* 1125 * Concatenate fragments. 1126 */ 1127 m = q; 1128 t = m->m_next; 1129 m->m_next = NULL; 1130 m_cat(m, t); 1131 nq = q->m_nextpkt; 1132 q->m_nextpkt = NULL; 1133 for (q = nq; q != NULL; q = nq) { 1134 nq = q->m_nextpkt; 1135 q->m_nextpkt = NULL; 1136 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1137 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1138 m_cat(m, q); 1139 } 1140 /* 1141 * In order to do checksumming faster we do 'end-around carry' here 1142 * (and not in for{} loop), though it implies we are not going to 1143 * reassemble more than 64k fragments. 1144 */ 1145 m->m_pkthdr.csum_data = 1146 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1147#ifdef MAC 1148 mac_ipq_reassemble(fp, m); 1149 mac_ipq_destroy(fp); 1150#endif 1151 1152 /* 1153 * Create header for new ip packet by modifying header of first 1154 * packet; dequeue and discard fragment reassembly header. 1155 * Make header visible. 1156 */ 1157 ip->ip_len = (ip->ip_hl << 2) + next; 1158 ip->ip_src = fp->ipq_src; 1159 ip->ip_dst = fp->ipq_dst; 1160 TAILQ_REMOVE(head, fp, ipq_list); 1161 V_nipq--; 1162 uma_zfree(V_ipq_zone, fp); 1163 m->m_len += (ip->ip_hl << 2); 1164 m->m_data -= (ip->ip_hl << 2); 1165 /* some debugging cruft by sklower, below, will go away soon */ 1166 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1167 m_fixhdr(m); 1168 IPSTAT_INC(ips_reassembled); 1169 IPQ_UNLOCK(); 1170 return (m); 1171 1172dropfrag: 1173 IPSTAT_INC(ips_fragdropped); 1174 if (fp != NULL) 1175 fp->ipq_nfrags--; 1176 m_freem(m); 1177done: 1178 IPQ_UNLOCK(); 1179 return (NULL); 1180 1181#undef GETIP 1182} 1183 1184/* 1185 * Free a fragment reassembly header and all 1186 * associated datagrams. 1187 */ 1188static void 1189ip_freef(struct ipqhead *fhp, struct ipq *fp) 1190{ 1191 INIT_VNET_INET(curvnet); 1192 struct mbuf *q; 1193 1194 IPQ_LOCK_ASSERT(); 1195 1196 while (fp->ipq_frags) { 1197 q = fp->ipq_frags; 1198 fp->ipq_frags = q->m_nextpkt; 1199 m_freem(q); 1200 } 1201 TAILQ_REMOVE(fhp, fp, ipq_list); 1202 uma_zfree(V_ipq_zone, fp); 1203 V_nipq--; 1204} 1205 1206/* 1207 * IP timer processing; 1208 * if a timer expires on a reassembly 1209 * queue, discard it. 1210 */ 1211void 1212ip_slowtimo(void) 1213{ 1214 VNET_ITERATOR_DECL(vnet_iter); 1215 struct ipq *fp; 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 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) { 1225 struct ipq *fpp; 1226 1227 fpp = fp; 1228 fp = TAILQ_NEXT(fp, ipq_list); 1229 if(--fpp->ipq_ttl == 0) { 1230 IPSTAT_ADD(ips_fragtimeout, 1231 fpp->ipq_nfrags); 1232 ip_freef(&V_ipq[i], fpp); 1233 } 1234 } 1235 } 1236 /* 1237 * If we are over the maximum number of fragments 1238 * (due to the limit being lowered), drain off 1239 * enough to get down to the new limit. 1240 */ 1241 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) { 1242 for (i = 0; i < IPREASS_NHASH; i++) { 1243 while (V_nipq > V_maxnipq && 1244 !TAILQ_EMPTY(&V_ipq[i])) { 1245 IPSTAT_ADD(ips_fragdropped, 1246 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1247 ip_freef(&V_ipq[i], 1248 TAILQ_FIRST(&V_ipq[i])); 1249 } 1250 } 1251 } 1252 CURVNET_RESTORE(); 1253 } 1254 VNET_LIST_RUNLOCK(); 1255 IPQ_UNLOCK(); 1256} 1257 1258/* 1259 * Drain off all datagram fragments. 1260 */ 1261void 1262ip_drain(void) 1263{ 1264 VNET_ITERATOR_DECL(vnet_iter); 1265 int i; 1266 1267 IPQ_LOCK(); 1268 VNET_LIST_RLOCK(); 1269 VNET_FOREACH(vnet_iter) { 1270 CURVNET_SET(vnet_iter); 1271 INIT_VNET_INET(vnet_iter); 1272 for (i = 0; i < IPREASS_NHASH; i++) { 1273 while(!TAILQ_EMPTY(&V_ipq[i])) { 1274 IPSTAT_ADD(ips_fragdropped, 1275 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1276 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); 1277 } 1278 } 1279 CURVNET_RESTORE(); 1280 } 1281 VNET_LIST_RUNLOCK(); 1282 IPQ_UNLOCK(); 1283 in_rtqdrain(); 1284} 1285 1286/* 1287 * The protocol to be inserted into ip_protox[] must be already registered 1288 * in inetsw[], either statically or through pf_proto_register(). 1289 */ 1290int 1291ipproto_register(u_char ipproto) 1292{ 1293 struct protosw *pr; 1294 1295 /* Sanity checks. */ 1296 if (ipproto == 0) 1297 return (EPROTONOSUPPORT); 1298 1299 /* 1300 * The protocol slot must not be occupied by another protocol 1301 * already. An index pointing to IPPROTO_RAW is unused. 1302 */ 1303 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1304 if (pr == NULL) 1305 return (EPFNOSUPPORT); 1306 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1307 return (EEXIST); 1308 1309 /* Find the protocol position in inetsw[] and set the index. */ 1310 for (pr = inetdomain.dom_protosw; 1311 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1312 if (pr->pr_domain->dom_family == PF_INET && 1313 pr->pr_protocol && pr->pr_protocol == ipproto) { 1314 /* Be careful to only index valid IP protocols. */ 1315 if (pr->pr_protocol < IPPROTO_MAX) { 1316 ip_protox[pr->pr_protocol] = pr - inetsw; 1317 return (0); 1318 } else 1319 return (EINVAL); 1320 } 1321 } 1322 return (EPROTONOSUPPORT); 1323} 1324 1325int 1326ipproto_unregister(u_char ipproto) 1327{ 1328 struct protosw *pr; 1329 1330 /* Sanity checks. */ 1331 if (ipproto == 0) 1332 return (EPROTONOSUPPORT); 1333 1334 /* Check if the protocol was indeed registered. */ 1335 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1336 if (pr == NULL) 1337 return (EPFNOSUPPORT); 1338 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1339 return (ENOENT); 1340 1341 /* Reset the protocol slot to IPPROTO_RAW. */ 1342 ip_protox[ipproto] = pr - inetsw; 1343 return (0); 1344} 1345 1346/* 1347 * Given address of next destination (final or next hop), return (referenced) 1348 * internet address info of interface to be used to get there. 1349 */ 1350struct in_ifaddr * 1351ip_rtaddr(struct in_addr dst, u_int fibnum) 1352{ 1353 struct route sro; 1354 struct sockaddr_in *sin; 1355 struct in_ifaddr *ia; 1356 1357 bzero(&sro, sizeof(sro)); 1358 sin = (struct sockaddr_in *)&sro.ro_dst; 1359 sin->sin_family = AF_INET; 1360 sin->sin_len = sizeof(*sin); 1361 sin->sin_addr = dst; 1362 in_rtalloc_ign(&sro, 0, fibnum); 1363 1364 if (sro.ro_rt == NULL) 1365 return (NULL); 1366 1367 ia = ifatoia(sro.ro_rt->rt_ifa); 1368 ifa_ref(&ia->ia_ifa); 1369 RTFREE(sro.ro_rt); 1370 return (ia); 1371} 1372 1373u_char inetctlerrmap[PRC_NCMDS] = { 1374 0, 0, 0, 0, 1375 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1376 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1377 EMSGSIZE, EHOSTUNREACH, 0, 0, 1378 0, 0, EHOSTUNREACH, 0, 1379 ENOPROTOOPT, ECONNREFUSED 1380}; 1381 1382/* 1383 * Forward a packet. If some error occurs return the sender 1384 * an icmp packet. Note we can't always generate a meaningful 1385 * icmp message because icmp doesn't have a large enough repertoire 1386 * of codes and types. 1387 * 1388 * If not forwarding, just drop the packet. This could be confusing 1389 * if ipforwarding was zero but some routing protocol was advancing 1390 * us as a gateway to somewhere. However, we must let the routing 1391 * protocol deal with that. 1392 * 1393 * The srcrt parameter indicates whether the packet is being forwarded 1394 * via a source route. 1395 */ 1396void 1397ip_forward(struct mbuf *m, int srcrt) 1398{ 1399 INIT_VNET_INET(curvnet); 1400 struct ip *ip = mtod(m, struct ip *); 1401 struct in_ifaddr *ia; 1402 struct mbuf *mcopy; 1403 struct in_addr dest; 1404 struct route ro; 1405 int error, type = 0, code = 0, mtu = 0; 1406 1407 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1408 IPSTAT_INC(ips_cantforward); 1409 m_freem(m); 1410 return; 1411 } 1412#ifdef IPSTEALTH 1413 if (!V_ipstealth) { 1414#endif 1415 if (ip->ip_ttl <= IPTTLDEC) { 1416 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1417 0, 0); 1418 return; 1419 } 1420#ifdef IPSTEALTH 1421 } 1422#endif 1423 1424 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); 1425#ifndef IPSEC 1426 /* 1427 * 'ia' may be NULL if there is no route for this destination. 1428 * In case of IPsec, Don't discard it just yet, but pass it to 1429 * ip_output in case of outgoing IPsec policy. 1430 */ 1431 if (!srcrt && ia == NULL) { 1432 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1433 return; 1434 } 1435#endif 1436 1437 /* 1438 * Save the IP header and at most 8 bytes of the payload, 1439 * in case we need to generate an ICMP message to the src. 1440 * 1441 * XXX this can be optimized a lot by saving the data in a local 1442 * buffer on the stack (72 bytes at most), and only allocating the 1443 * mbuf if really necessary. The vast majority of the packets 1444 * are forwarded without having to send an ICMP back (either 1445 * because unnecessary, or because rate limited), so we are 1446 * really we are wasting a lot of work here. 1447 * 1448 * We don't use m_copy() because it might return a reference 1449 * to a shared cluster. Both this function and ip_output() 1450 * assume exclusive access to the IP header in `m', so any 1451 * data in a cluster may change before we reach icmp_error(). 1452 */ 1453 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1454 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1455 /* 1456 * It's probably ok if the pkthdr dup fails (because 1457 * the deep copy of the tag chain failed), but for now 1458 * be conservative and just discard the copy since 1459 * code below may some day want the tags. 1460 */ 1461 m_free(mcopy); 1462 mcopy = NULL; 1463 } 1464 if (mcopy != NULL) { 1465 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy)); 1466 mcopy->m_pkthdr.len = mcopy->m_len; 1467 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1468 } 1469 1470#ifdef IPSTEALTH 1471 if (!V_ipstealth) { 1472#endif 1473 ip->ip_ttl -= IPTTLDEC; 1474#ifdef IPSTEALTH 1475 } 1476#endif 1477 1478 /* 1479 * If forwarding packet using same interface that it came in on, 1480 * perhaps should send a redirect to sender to shortcut a hop. 1481 * Only send redirect if source is sending directly to us, 1482 * and if packet was not source routed (or has any options). 1483 * Also, don't send redirect if forwarding using a default route 1484 * or a route modified by a redirect. 1485 */ 1486 dest.s_addr = 0; 1487 if (!srcrt && V_ipsendredirects && 1488 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) { 1489 struct sockaddr_in *sin; 1490 struct rtentry *rt; 1491 1492 bzero(&ro, sizeof(ro)); 1493 sin = (struct sockaddr_in *)&ro.ro_dst; 1494 sin->sin_family = AF_INET; 1495 sin->sin_len = sizeof(*sin); 1496 sin->sin_addr = ip->ip_dst; 1497 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 1498 1499 rt = ro.ro_rt; 1500 1501 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1502 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1503#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1504 u_long src = ntohl(ip->ip_src.s_addr); 1505 1506 if (RTA(rt) && 1507 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1508 if (rt->rt_flags & RTF_GATEWAY) 1509 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1510 else 1511 dest.s_addr = ip->ip_dst.s_addr; 1512 /* Router requirements says to only send host redirects */ 1513 type = ICMP_REDIRECT; 1514 code = ICMP_REDIRECT_HOST; 1515 } 1516 } 1517 if (rt) 1518 RTFREE(rt); 1519 } 1520 1521 /* 1522 * Try to cache the route MTU from ip_output so we can consider it for 1523 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. 1524 */ 1525 bzero(&ro, sizeof(ro)); 1526 1527 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1528 1529 if (error == EMSGSIZE && ro.ro_rt) 1530 mtu = ro.ro_rt->rt_rmx.rmx_mtu; 1531 if (ro.ro_rt) 1532 RTFREE(ro.ro_rt); 1533 1534 if (error) 1535 IPSTAT_INC(ips_cantforward); 1536 else { 1537 IPSTAT_INC(ips_forward); 1538 if (type) 1539 IPSTAT_INC(ips_redirectsent); 1540 else { 1541 if (mcopy) 1542 m_freem(mcopy); 1543 if (ia != NULL) 1544 ifa_free(&ia->ia_ifa); 1545 return; 1546 } 1547 } 1548 if (mcopy == NULL) { 1549 if (ia != NULL) 1550 ifa_free(&ia->ia_ifa); 1551 return; 1552 } 1553 1554 switch (error) { 1555 1556 case 0: /* forwarded, but need redirect */ 1557 /* type, code set above */ 1558 break; 1559 1560 case ENETUNREACH: 1561 case EHOSTUNREACH: 1562 case ENETDOWN: 1563 case EHOSTDOWN: 1564 default: 1565 type = ICMP_UNREACH; 1566 code = ICMP_UNREACH_HOST; 1567 break; 1568 1569 case EMSGSIZE: 1570 type = ICMP_UNREACH; 1571 code = ICMP_UNREACH_NEEDFRAG; 1572 1573#ifdef IPSEC 1574 /* 1575 * If IPsec is configured for this path, 1576 * override any possibly mtu value set by ip_output. 1577 */ 1578 mtu = ip_ipsec_mtu(m, mtu); 1579#endif /* IPSEC */ 1580 /* 1581 * If the MTU was set before make sure we are below the 1582 * interface MTU. 1583 * If the MTU wasn't set before use the interface mtu or 1584 * fall back to the next smaller mtu step compared to the 1585 * current packet size. 1586 */ 1587 if (mtu != 0) { 1588 if (ia != NULL) 1589 mtu = min(mtu, ia->ia_ifp->if_mtu); 1590 } else { 1591 if (ia != NULL) 1592 mtu = ia->ia_ifp->if_mtu; 1593 else 1594 mtu = ip_next_mtu(ip->ip_len, 0); 1595 } 1596 IPSTAT_INC(ips_cantfrag); 1597 break; 1598 1599 case ENOBUFS: 1600 /* 1601 * A router should not generate ICMP_SOURCEQUENCH as 1602 * required in RFC1812 Requirements for IP Version 4 Routers. 1603 * Source quench could be a big problem under DoS attacks, 1604 * or if the underlying interface is rate-limited. 1605 * Those who need source quench packets may re-enable them 1606 * via the net.inet.ip.sendsourcequench sysctl. 1607 */ 1608 if (V_ip_sendsourcequench == 0) { 1609 m_freem(mcopy); 1610 if (ia != NULL) 1611 ifa_free(&ia->ia_ifa); 1612 return; 1613 } else { 1614 type = ICMP_SOURCEQUENCH; 1615 code = 0; 1616 } 1617 break; 1618 1619 case EACCES: /* ipfw denied packet */ 1620 m_freem(mcopy); 1621 if (ia != NULL) 1622 ifa_free(&ia->ia_ifa); 1623 return; 1624 } 1625 if (ia != NULL) 1626 ifa_free(&ia->ia_ifa); 1627 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1628} 1629 1630void 1631ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1632 struct mbuf *m) 1633{ 1634 INIT_VNET_NET(inp->inp_vnet); 1635 1636 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1637 struct bintime bt; 1638 1639 bintime(&bt); 1640 if (inp->inp_socket->so_options & SO_BINTIME) { 1641 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1642 SCM_BINTIME, SOL_SOCKET); 1643 if (*mp) 1644 mp = &(*mp)->m_next; 1645 } 1646 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1647 struct timeval tv; 1648 1649 bintime2timeval(&bt, &tv); 1650 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1651 SCM_TIMESTAMP, SOL_SOCKET); 1652 if (*mp) 1653 mp = &(*mp)->m_next; 1654 } 1655 } 1656 if (inp->inp_flags & INP_RECVDSTADDR) { 1657 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1658 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1659 if (*mp) 1660 mp = &(*mp)->m_next; 1661 } 1662 if (inp->inp_flags & INP_RECVTTL) { 1663 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1664 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1665 if (*mp) 1666 mp = &(*mp)->m_next; 1667 } 1668#ifdef notyet 1669 /* XXX 1670 * Moving these out of udp_input() made them even more broken 1671 * than they already were. 1672 */ 1673 /* options were tossed already */ 1674 if (inp->inp_flags & INP_RECVOPTS) { 1675 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1676 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1677 if (*mp) 1678 mp = &(*mp)->m_next; 1679 } 1680 /* ip_srcroute doesn't do what we want here, need to fix */ 1681 if (inp->inp_flags & INP_RECVRETOPTS) { 1682 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1683 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1684 if (*mp) 1685 mp = &(*mp)->m_next; 1686 } 1687#endif 1688 if (inp->inp_flags & INP_RECVIF) { 1689 struct ifnet *ifp; 1690 struct sdlbuf { 1691 struct sockaddr_dl sdl; 1692 u_char pad[32]; 1693 } sdlbuf; 1694 struct sockaddr_dl *sdp; 1695 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1696 1697 if (((ifp = m->m_pkthdr.rcvif)) 1698 && ( ifp->if_index && (ifp->if_index <= V_if_index))) { 1699 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1700 /* 1701 * Change our mind and don't try copy. 1702 */ 1703 if ((sdp->sdl_family != AF_LINK) 1704 || (sdp->sdl_len > sizeof(sdlbuf))) { 1705 goto makedummy; 1706 } 1707 bcopy(sdp, sdl2, sdp->sdl_len); 1708 } else { 1709makedummy: 1710 sdl2->sdl_len 1711 = offsetof(struct sockaddr_dl, sdl_data[0]); 1712 sdl2->sdl_family = AF_LINK; 1713 sdl2->sdl_index = 0; 1714 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1715 } 1716 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1717 IP_RECVIF, IPPROTO_IP); 1718 if (*mp) 1719 mp = &(*mp)->m_next; 1720 } 1721} 1722 1723/* 1724 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1725 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1726 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1727 * compiled. 1728 */ 1729int 1730ip_rsvp_init(struct socket *so) 1731{ 1732 INIT_VNET_INET(so->so_vnet); 1733 1734 if (so->so_type != SOCK_RAW || 1735 so->so_proto->pr_protocol != IPPROTO_RSVP) 1736 return EOPNOTSUPP; 1737 1738 if (V_ip_rsvpd != NULL) 1739 return EADDRINUSE; 1740 1741 V_ip_rsvpd = so; 1742 /* 1743 * This may seem silly, but we need to be sure we don't over-increment 1744 * the RSVP counter, in case something slips up. 1745 */ 1746 if (!V_ip_rsvp_on) { 1747 V_ip_rsvp_on = 1; 1748 V_rsvp_on++; 1749 } 1750 1751 return 0; 1752} 1753 1754int 1755ip_rsvp_done(void) 1756{ 1757 INIT_VNET_INET(curvnet); 1758 1759 V_ip_rsvpd = NULL; 1760 /* 1761 * This may seem silly, but we need to be sure we don't over-decrement 1762 * the RSVP counter, in case something slips up. 1763 */ 1764 if (V_ip_rsvp_on) { 1765 V_ip_rsvp_on = 0; 1766 V_rsvp_on--; 1767 } 1768 return 0; 1769} 1770 1771void 1772rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1773{ 1774 INIT_VNET_INET(curvnet); 1775 1776 if (rsvp_input_p) { /* call the real one if loaded */ 1777 rsvp_input_p(m, off); 1778 return; 1779 } 1780 1781 /* Can still get packets with rsvp_on = 0 if there is a local member 1782 * of the group to which the RSVP packet is addressed. But in this 1783 * case we want to throw the packet away. 1784 */ 1785 1786 if (!V_rsvp_on) { 1787 m_freem(m); 1788 return; 1789 } 1790 1791 if (V_ip_rsvpd != NULL) { 1792 rip_input(m, off); 1793 return; 1794 } 1795 /* Drop the packet */ 1796 m_freem(m); 1797} 1798