ip_reass.c revision 194962
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 194962 2009-06-25 14:44:00Z 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 121struct rwlock in_ifaddr_lock; 122RW_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock"); 123 124SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_FORWARDING, 125 forwarding, CTLFLAG_RW, ipforwarding, 0, 126 "Enable IP forwarding between interfaces"); 127 128SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_SENDREDIRECTS, 129 redirect, CTLFLAG_RW, ipsendredirects, 0, 130 "Enable sending IP redirects"); 131 132SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_DEFTTL, 133 ttl, CTLFLAG_RW, ip_defttl, 0, "Maximum TTL on IP packets"); 134 135SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_KEEPFAITH, 136 keepfaith, CTLFLAG_RW, ip_keepfaith, 0, 137 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 138 139SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, 140 sendsourcequench, CTLFLAG_RW, ip_sendsourcequench, 0, 141 "Enable the transmission of source quench packets"); 142 143SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, random_id, 144 CTLFLAG_RW, ip_do_randomid, 0, "Assign random ip_id values"); 145 146/* 147 * XXX - Setting ip_checkinterface mostly implements the receive side of 148 * the Strong ES model described in RFC 1122, but since the routing table 149 * and transmit implementation do not implement the Strong ES model, 150 * setting this to 1 results in an odd hybrid. 151 * 152 * XXX - ip_checkinterface currently must be disabled if you use ipnat 153 * to translate the destination address to another local interface. 154 * 155 * XXX - ip_checkinterface must be disabled if you add IP aliases 156 * to the loopback interface instead of the interface where the 157 * packets for those addresses are received. 158 */ 159SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, 160 check_interface, CTLFLAG_RW, ip_checkinterface, 0, 161 "Verify packet arrives on correct interface"); 162 163struct pfil_head inet_pfil_hook; /* Packet filter hooks */ 164 165static struct netisr_handler ip_nh = { 166 .nh_name = "ip", 167 .nh_handler = ip_input, 168 .nh_proto = NETISR_IP, 169 .nh_policy = NETISR_POLICY_FLOW, 170}; 171 172extern struct domain inetdomain; 173extern struct protosw inetsw[]; 174u_char ip_protox[IPPROTO_MAX]; 175 176 177SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 178 ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 179 180#ifdef VIMAGE_GLOBALS 181static uma_zone_t ipq_zone; 182#endif 183static struct mtx ipqlock; 184 185#define IPQ_LOCK() mtx_lock(&ipqlock) 186#define IPQ_UNLOCK() mtx_unlock(&ipqlock) 187#define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 188#define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 189 190static void maxnipq_update(void); 191static void ipq_zone_change(void *); 192 193SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, fragpackets, 194 CTLFLAG_RD, nipq, 0, 195 "Current number of IPv4 fragment reassembly queue entries"); 196 197SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, maxfragsperpacket, 198 CTLFLAG_RW, maxfragsperpacket, 0, 199 "Maximum number of IPv4 fragments allowed per packet"); 200 201struct callout ipport_tick_callout; 202 203#ifdef IPCTL_DEFMTU 204SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 205 &ip_mtu, 0, "Default MTU"); 206#endif 207 208#ifdef IPSTEALTH 209SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 210 ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding"); 211#endif 212#ifdef FLOWTABLE 213#ifdef VIMAGE_GLOBALS 214static int ip_output_flowtable_size; 215struct flowtable *ip_ft; 216#endif 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#endif 221 222#ifdef VIMAGE_GLOBALS 223int fw_one_pass; 224#endif 225 226static void ip_freef(struct ipqhead *, struct ipq *); 227 228#ifndef VIMAGE_GLOBALS 229static void vnet_inet_register(void); 230 231static const vnet_modinfo_t vnet_inet_modinfo = { 232 .vmi_id = VNET_MOD_INET, 233 .vmi_name = "inet", 234 .vmi_size = sizeof(struct vnet_inet) 235}; 236 237static void vnet_inet_register() 238{ 239 240 vnet_mod_register(&vnet_inet_modinfo); 241} 242 243SYSINIT(inet, SI_SUB_PROTO_BEGIN, SI_ORDER_FIRST, vnet_inet_register, 0); 244#endif 245 246static int 247sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 248{ 249 int error, qlimit; 250 251 netisr_getqlimit(&ip_nh, &qlimit); 252 error = sysctl_handle_int(oidp, &qlimit, 0, req); 253 if (error || !req->newptr) 254 return (error); 255 if (qlimit < 1) 256 return (EINVAL); 257 return (netisr_setqlimit(&ip_nh, qlimit)); 258} 259SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, 260 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I", 261 "Maximum size of the IP input queue"); 262 263static int 264sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) 265{ 266 u_int64_t qdrops_long; 267 int error, qdrops; 268 269 netisr_getqdrops(&ip_nh, &qdrops_long); 270 qdrops = qdrops_long; 271 error = sysctl_handle_int(oidp, &qdrops, 0, req); 272 if (error || !req->newptr) 273 return (error); 274 if (qdrops != 0) 275 return (EINVAL); 276 netisr_clearqdrops(&ip_nh); 277 return (0); 278} 279 280SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, 281 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I", 282 "Number of packets dropped from the IP input queue"); 283 284/* 285 * IP initialization: fill in IP protocol switch table. 286 * All protocols not implemented in kernel go to raw IP protocol handler. 287 */ 288void 289ip_init(void) 290{ 291 INIT_VNET_INET(curvnet); 292 struct protosw *pr; 293 int i; 294 295 V_ipsendredirects = 1; /* XXX */ 296 V_ip_checkinterface = 0; 297 V_ip_keepfaith = 0; 298 V_ip_sendsourcequench = 0; 299 V_rsvp_on = 0; 300 V_ip_defttl = IPDEFTTL; 301 V_ip_do_randomid = 0; 302 V_ip_id = time_second & 0xffff; 303 V_ipforwarding = 0; 304 V_ipstealth = 0; 305 V_nipq = 0; /* Total # of reass queues */ 306 307 V_ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */ 308 V_ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */ 309 V_ipport_firstauto = IPPORT_EPHEMERALFIRST; /* 10000 */ 310 V_ipport_lastauto = IPPORT_EPHEMERALLAST; /* 65535 */ 311 V_ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */ 312 V_ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */ 313 V_ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */ 314 V_ipport_reservedlow = 0; 315 V_ipport_randomized = 1; /* user controlled via sysctl */ 316 V_ipport_randomcps = 10; /* user controlled via sysctl */ 317 V_ipport_randomtime = 45; /* user controlled via sysctl */ 318 V_ipport_stoprandom = 0; /* toggled by ipport_tick */ 319 320 V_fw_one_pass = 1; 321 322#ifdef NOTYET 323 /* XXX global static but not instantiated in this file */ 324 V_ipfastforward_active = 0; 325 V_subnetsarelocal = 0; 326 V_sameprefixcarponly = 0; 327#endif 328 329 TAILQ_INIT(&V_in_ifaddrhead); 330 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 331 332 /* Initialize IP reassembly queue. */ 333 for (i = 0; i < IPREASS_NHASH; i++) 334 TAILQ_INIT(&V_ipq[i]); 335 V_maxnipq = nmbclusters / 32; 336 V_maxfragsperpacket = 16; 337 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 338 NULL, UMA_ALIGN_PTR, 0); 339 maxnipq_update(); 340 341#ifdef FLOWTABLE 342 V_ip_output_flowtable_size = 2048; 343 TUNABLE_INT_FETCH("net.inet.ip.output_flowtable_size", 344 &V_ip_output_flowtable_size); 345 V_ip_ft = flowtable_alloc(V_ip_output_flowtable_size, FL_PCPU); 346#endif 347 348 /* Skip initialization of globals for non-default instances. */ 349 if (!IS_DEFAULT_VNET(curvnet)) 350 return; 351 352 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 353 if (pr == NULL) 354 panic("ip_init: PF_INET not found"); 355 356 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 357 for (i = 0; i < IPPROTO_MAX; i++) 358 ip_protox[i] = pr - inetsw; 359 /* 360 * Cycle through IP protocols and put them into the appropriate place 361 * in ip_protox[]. 362 */ 363 for (pr = inetdomain.dom_protosw; 364 pr < inetdomain.dom_protoswNPROTOSW; pr++) 365 if (pr->pr_domain->dom_family == PF_INET && 366 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 367 /* Be careful to only index valid IP protocols. */ 368 if (pr->pr_protocol < IPPROTO_MAX) 369 ip_protox[pr->pr_protocol] = pr - inetsw; 370 } 371 372 /* Initialize packet filter hooks. */ 373 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 374 inet_pfil_hook.ph_af = AF_INET; 375 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 376 printf("%s: WARNING: unable to register pfil hook, " 377 "error %d\n", __func__, i); 378 379 /* Start ipport_tick. */ 380 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE); 381 callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL); 382 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, 383 SHUTDOWN_PRI_DEFAULT); 384 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, 385 NULL, EVENTHANDLER_PRI_ANY); 386 387 /* Initialize various other remaining things. */ 388 IPQ_LOCK_INIT(); 389 netisr_register(&ip_nh); 390} 391 392void 393ip_fini(void *xtp) 394{ 395 396 callout_stop(&ipport_tick_callout); 397} 398 399/* 400 * Ip input routine. Checksum and byte swap header. If fragmented 401 * try to reassemble. Process options. Pass to next level. 402 */ 403void 404ip_input(struct mbuf *m) 405{ 406 INIT_VNET_INET(curvnet); 407 struct ip *ip = NULL; 408 struct in_ifaddr *ia = NULL; 409 struct ifaddr *ifa; 410 struct ifnet *ifp; 411 int checkif, hlen = 0; 412 u_short sum; 413 int dchg = 0; /* dest changed after fw */ 414 struct in_addr odst; /* original dst address */ 415 416 M_ASSERTPKTHDR(m); 417 418 if (m->m_flags & M_FASTFWD_OURS) { 419 /* 420 * Firewall or NAT changed destination to local. 421 * We expect ip_len and ip_off to be in host byte order. 422 */ 423 m->m_flags &= ~M_FASTFWD_OURS; 424 /* Set up some basics that will be used later. */ 425 ip = mtod(m, struct ip *); 426 hlen = ip->ip_hl << 2; 427 goto ours; 428 } 429 430 IPSTAT_INC(ips_total); 431 432 if (m->m_pkthdr.len < sizeof(struct ip)) 433 goto tooshort; 434 435 if (m->m_len < sizeof (struct ip) && 436 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 437 IPSTAT_INC(ips_toosmall); 438 return; 439 } 440 ip = mtod(m, struct ip *); 441 442 if (ip->ip_v != IPVERSION) { 443 IPSTAT_INC(ips_badvers); 444 goto bad; 445 } 446 447 hlen = ip->ip_hl << 2; 448 if (hlen < sizeof(struct ip)) { /* minimum header length */ 449 IPSTAT_INC(ips_badhlen); 450 goto bad; 451 } 452 if (hlen > m->m_len) { 453 if ((m = m_pullup(m, hlen)) == NULL) { 454 IPSTAT_INC(ips_badhlen); 455 return; 456 } 457 ip = mtod(m, struct ip *); 458 } 459 460 /* 127/8 must not appear on wire - RFC1122 */ 461 ifp = m->m_pkthdr.rcvif; 462 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 463 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 464 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 465 IPSTAT_INC(ips_badaddr); 466 goto bad; 467 } 468 } 469 470 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 471 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 472 } else { 473 if (hlen == sizeof(struct ip)) { 474 sum = in_cksum_hdr(ip); 475 } else { 476 sum = in_cksum(m, hlen); 477 } 478 } 479 if (sum) { 480 IPSTAT_INC(ips_badsum); 481 goto bad; 482 } 483 484#ifdef ALTQ 485 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 486 /* packet is dropped by traffic conditioner */ 487 return; 488#endif 489 490 /* 491 * Convert fields to host representation. 492 */ 493 ip->ip_len = ntohs(ip->ip_len); 494 if (ip->ip_len < hlen) { 495 IPSTAT_INC(ips_badlen); 496 goto bad; 497 } 498 ip->ip_off = ntohs(ip->ip_off); 499 500 /* 501 * Check that the amount of data in the buffers 502 * is as at least much as the IP header would have us expect. 503 * Trim mbufs if longer than we expect. 504 * Drop packet if shorter than we expect. 505 */ 506 if (m->m_pkthdr.len < ip->ip_len) { 507tooshort: 508 IPSTAT_INC(ips_tooshort); 509 goto bad; 510 } 511 if (m->m_pkthdr.len > ip->ip_len) { 512 if (m->m_len == m->m_pkthdr.len) { 513 m->m_len = ip->ip_len; 514 m->m_pkthdr.len = ip->ip_len; 515 } else 516 m_adj(m, ip->ip_len - m->m_pkthdr.len); 517 } 518#ifdef IPSEC 519 /* 520 * Bypass packet filtering for packets from a tunnel (gif). 521 */ 522 if (ip_ipsec_filtertunnel(m)) 523 goto passin; 524#endif /* IPSEC */ 525 526 /* 527 * Run through list of hooks for input packets. 528 * 529 * NB: Beware of the destination address changing (e.g. 530 * by NAT rewriting). When this happens, tell 531 * ip_forward to do the right thing. 532 */ 533 534 /* Jump over all PFIL processing if hooks are not active. */ 535 if (!PFIL_HOOKED(&inet_pfil_hook)) 536 goto passin; 537 538 odst = ip->ip_dst; 539 if (pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0) 540 return; 541 if (m == NULL) /* consumed by filter */ 542 return; 543 544 ip = mtod(m, struct ip *); 545 dchg = (odst.s_addr != ip->ip_dst.s_addr); 546 ifp = m->m_pkthdr.rcvif; 547 548#ifdef IPFIREWALL_FORWARD 549 if (m->m_flags & M_FASTFWD_OURS) { 550 m->m_flags &= ~M_FASTFWD_OURS; 551 goto ours; 552 } 553 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 554 /* 555 * Directly ship on the packet. This allows to forward packets 556 * that were destined for us to some other directly connected 557 * host. 558 */ 559 ip_forward(m, dchg); 560 return; 561 } 562#endif /* IPFIREWALL_FORWARD */ 563 564passin: 565 /* 566 * Process options and, if not destined for us, 567 * ship it on. ip_dooptions returns 1 when an 568 * error was detected (causing an icmp message 569 * to be sent and the original packet to be freed). 570 */ 571 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 572 return; 573 574 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 575 * matter if it is destined to another node, or whether it is 576 * a multicast one, RSVP wants it! and prevents it from being forwarded 577 * anywhere else. Also checks if the rsvp daemon is running before 578 * grabbing the packet. 579 */ 580 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 581 goto ours; 582 583 /* 584 * Check our list of addresses, to see if the packet is for us. 585 * If we don't have any addresses, assume any unicast packet 586 * we receive might be for us (and let the upper layers deal 587 * with it). 588 */ 589 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 590 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 591 goto ours; 592 593 /* 594 * Enable a consistency check between the destination address 595 * and the arrival interface for a unicast packet (the RFC 1122 596 * strong ES model) if IP forwarding is disabled and the packet 597 * is not locally generated and the packet is not subject to 598 * 'ipfw fwd'. 599 * 600 * XXX - Checking also should be disabled if the destination 601 * address is ipnat'ed to a different interface. 602 * 603 * XXX - Checking is incompatible with IP aliases added 604 * to the loopback interface instead of the interface where 605 * the packets are received. 606 * 607 * XXX - This is the case for carp vhost IPs as well so we 608 * insert a workaround. If the packet got here, we already 609 * checked with carp_iamatch() and carp_forus(). 610 */ 611 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 612 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && 613#ifdef DEV_CARP 614 !ifp->if_carp && 615#endif 616 (dchg == 0); 617 618 /* 619 * Check for exact addresses in the hash bucket. 620 */ 621 /* IN_IFADDR_RLOCK(); */ 622 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 623 /* 624 * If the address matches, verify that the packet 625 * arrived via the correct interface if checking is 626 * enabled. 627 */ 628 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 629 (!checkif || ia->ia_ifp == ifp)) { 630 ifa_ref(&ia->ia_ifa); 631 /* IN_IFADDR_RUNLOCK(); */ 632 goto ours; 633 } 634 } 635 /* IN_IFADDR_RUNLOCK(); */ 636 637 /* 638 * Check for broadcast addresses. 639 * 640 * Only accept broadcast packets that arrive via the matching 641 * interface. Reception of forwarded directed broadcasts would 642 * be handled via ip_forward() and ether_output() with the loopback 643 * into the stack for SIMPLEX interfaces handled by ether_output(). 644 */ 645 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 646 IF_ADDR_LOCK(ifp); 647 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 648 if (ifa->ifa_addr->sa_family != AF_INET) 649 continue; 650 ia = ifatoia(ifa); 651 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 652 ip->ip_dst.s_addr) { 653 ifa_ref(ifa); 654 IF_ADDR_UNLOCK(ifp); 655 goto ours; 656 } 657 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) { 658 ifa_ref(ifa); 659 IF_ADDR_UNLOCK(ifp); 660 goto ours; 661 } 662#ifdef BOOTP_COMPAT 663 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { 664 ifa_ref(ifa); 665 IF_ADDR_UNLOCK(ifp); 666 goto ours; 667 } 668#endif 669 } 670 IF_ADDR_UNLOCK(ifp); 671 ia = NULL; 672 } 673 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 674 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 675 IPSTAT_INC(ips_cantforward); 676 m_freem(m); 677 return; 678 } 679 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 680 if (V_ip_mrouter) { 681 /* 682 * If we are acting as a multicast router, all 683 * incoming multicast packets are passed to the 684 * kernel-level multicast forwarding function. 685 * The packet is returned (relatively) intact; if 686 * ip_mforward() returns a non-zero value, the packet 687 * must be discarded, else it may be accepted below. 688 */ 689 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 690 IPSTAT_INC(ips_cantforward); 691 m_freem(m); 692 return; 693 } 694 695 /* 696 * The process-level routing daemon needs to receive 697 * all multicast IGMP packets, whether or not this 698 * host belongs to their destination groups. 699 */ 700 if (ip->ip_p == IPPROTO_IGMP) 701 goto ours; 702 IPSTAT_INC(ips_forward); 703 } 704 /* 705 * Assume the packet is for us, to avoid prematurely taking 706 * a lock on the in_multi hash. Protocols must perform 707 * their own filtering and update statistics accordingly. 708 */ 709 goto ours; 710 } 711 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 712 goto ours; 713 if (ip->ip_dst.s_addr == INADDR_ANY) 714 goto ours; 715 716 /* 717 * FAITH(Firewall Aided Internet Translator) 718 */ 719 if (ifp && ifp->if_type == IFT_FAITH) { 720 if (V_ip_keepfaith) { 721 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 722 goto ours; 723 } 724 m_freem(m); 725 return; 726 } 727 728 /* 729 * Not for us; forward if possible and desirable. 730 */ 731 if (V_ipforwarding == 0) { 732 IPSTAT_INC(ips_cantforward); 733 m_freem(m); 734 } else { 735#ifdef IPSEC 736 if (ip_ipsec_fwd(m)) 737 goto bad; 738#endif /* IPSEC */ 739 ip_forward(m, dchg); 740 } 741 return; 742 743ours: 744#ifdef IPSTEALTH 745 /* 746 * IPSTEALTH: Process non-routing options only 747 * if the packet is destined for us. 748 */ 749 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) { 750 if (ia != NULL) 751 ifa_free(&ia->ia_ifa); 752 return; 753 } 754#endif /* IPSTEALTH */ 755 756 /* Count the packet in the ip address stats */ 757 if (ia != NULL) { 758 ia->ia_ifa.if_ipackets++; 759 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 760 ifa_free(&ia->ia_ifa); 761 } 762 763 /* 764 * Attempt reassembly; if it succeeds, proceed. 765 * ip_reass() will return a different mbuf. 766 */ 767 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 768 m = ip_reass(m); 769 if (m == NULL) 770 return; 771 ip = mtod(m, struct ip *); 772 /* Get the header length of the reassembled packet */ 773 hlen = ip->ip_hl << 2; 774 } 775 776 /* 777 * Further protocols expect the packet length to be w/o the 778 * IP header. 779 */ 780 ip->ip_len -= hlen; 781 782#ifdef IPSEC 783 /* 784 * enforce IPsec policy checking if we are seeing last header. 785 * note that we do not visit this with protocols with pcb layer 786 * code - like udp/tcp/raw ip. 787 */ 788 if (ip_ipsec_input(m)) 789 goto bad; 790#endif /* IPSEC */ 791 792 /* 793 * Switch out to protocol's input routine. 794 */ 795 IPSTAT_INC(ips_delivered); 796 797 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 798 return; 799bad: 800 m_freem(m); 801} 802 803/* 804 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 805 * max has slightly different semantics than the sysctl, for historical 806 * reasons. 807 */ 808static void 809maxnipq_update(void) 810{ 811 INIT_VNET_INET(curvnet); 812 813 /* 814 * -1 for unlimited allocation. 815 */ 816 if (V_maxnipq < 0) 817 uma_zone_set_max(V_ipq_zone, 0); 818 /* 819 * Positive number for specific bound. 820 */ 821 if (V_maxnipq > 0) 822 uma_zone_set_max(V_ipq_zone, V_maxnipq); 823 /* 824 * Zero specifies no further fragment queue allocation -- set the 825 * bound very low, but rely on implementation elsewhere to actually 826 * prevent allocation and reclaim current queues. 827 */ 828 if (V_maxnipq == 0) 829 uma_zone_set_max(V_ipq_zone, 1); 830} 831 832static void 833ipq_zone_change(void *tag) 834{ 835 INIT_VNET_INET(curvnet); 836 837 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) { 838 V_maxnipq = nmbclusters / 32; 839 maxnipq_update(); 840 } 841} 842 843static int 844sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 845{ 846 INIT_VNET_INET(curvnet); 847 int error, i; 848 849 i = V_maxnipq; 850 error = sysctl_handle_int(oidp, &i, 0, req); 851 if (error || !req->newptr) 852 return (error); 853 854 /* 855 * XXXRW: Might be a good idea to sanity check the argument and place 856 * an extreme upper bound. 857 */ 858 if (i < -1) 859 return (EINVAL); 860 V_maxnipq = i; 861 maxnipq_update(); 862 return (0); 863} 864 865SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 866 NULL, 0, sysctl_maxnipq, "I", 867 "Maximum number of IPv4 fragment reassembly queue entries"); 868 869/* 870 * Take incoming datagram fragment and try to reassemble it into 871 * whole datagram. If the argument is the first fragment or one 872 * in between the function will return NULL and store the mbuf 873 * in the fragment chain. If the argument is the last fragment 874 * the packet will be reassembled and the pointer to the new 875 * mbuf returned for further processing. Only m_tags attached 876 * to the first packet/fragment are preserved. 877 * The IP header is *NOT* adjusted out of iplen. 878 */ 879struct mbuf * 880ip_reass(struct mbuf *m) 881{ 882 INIT_VNET_INET(curvnet); 883 struct ip *ip; 884 struct mbuf *p, *q, *nq, *t; 885 struct ipq *fp = NULL; 886 struct ipqhead *head; 887 int i, hlen, next; 888 u_int8_t ecn, ecn0; 889 u_short hash; 890 891 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 892 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) { 893 IPSTAT_INC(ips_fragments); 894 IPSTAT_INC(ips_fragdropped); 895 m_freem(m); 896 return (NULL); 897 } 898 899 ip = mtod(m, struct ip *); 900 hlen = ip->ip_hl << 2; 901 902 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 903 head = &V_ipq[hash]; 904 IPQ_LOCK(); 905 906 /* 907 * Look for queue of fragments 908 * of this datagram. 909 */ 910 TAILQ_FOREACH(fp, head, ipq_list) 911 if (ip->ip_id == fp->ipq_id && 912 ip->ip_src.s_addr == fp->ipq_src.s_addr && 913 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 914#ifdef MAC 915 mac_ipq_match(m, fp) && 916#endif 917 ip->ip_p == fp->ipq_p) 918 goto found; 919 920 fp = NULL; 921 922 /* 923 * Attempt to trim the number of allocated fragment queues if it 924 * exceeds the administrative limit. 925 */ 926 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) { 927 /* 928 * drop something from the tail of the current queue 929 * before proceeding further 930 */ 931 struct ipq *q = TAILQ_LAST(head, ipqhead); 932 if (q == NULL) { /* gak */ 933 for (i = 0; i < IPREASS_NHASH; i++) { 934 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead); 935 if (r) { 936 IPSTAT_ADD(ips_fragtimeout, 937 r->ipq_nfrags); 938 ip_freef(&V_ipq[i], r); 939 break; 940 } 941 } 942 } else { 943 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags); 944 ip_freef(head, q); 945 } 946 } 947 948found: 949 /* 950 * Adjust ip_len to not reflect header, 951 * convert offset of this to bytes. 952 */ 953 ip->ip_len -= hlen; 954 if (ip->ip_off & IP_MF) { 955 /* 956 * Make sure that fragments have a data length 957 * that's a non-zero multiple of 8 bytes. 958 */ 959 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 960 IPSTAT_INC(ips_toosmall); /* XXX */ 961 goto dropfrag; 962 } 963 m->m_flags |= M_FRAG; 964 } else 965 m->m_flags &= ~M_FRAG; 966 ip->ip_off <<= 3; 967 968 969 /* 970 * Attempt reassembly; if it succeeds, proceed. 971 * ip_reass() will return a different mbuf. 972 */ 973 IPSTAT_INC(ips_fragments); 974 m->m_pkthdr.header = ip; 975 976 /* Previous ip_reass() started here. */ 977 /* 978 * Presence of header sizes in mbufs 979 * would confuse code below. 980 */ 981 m->m_data += hlen; 982 m->m_len -= hlen; 983 984 /* 985 * If first fragment to arrive, create a reassembly queue. 986 */ 987 if (fp == NULL) { 988 fp = uma_zalloc(V_ipq_zone, M_NOWAIT); 989 if (fp == NULL) 990 goto dropfrag; 991#ifdef MAC 992 if (mac_ipq_init(fp, M_NOWAIT) != 0) { 993 uma_zfree(V_ipq_zone, fp); 994 fp = NULL; 995 goto dropfrag; 996 } 997 mac_ipq_create(m, fp); 998#endif 999 TAILQ_INSERT_HEAD(head, fp, ipq_list); 1000 V_nipq++; 1001 fp->ipq_nfrags = 1; 1002 fp->ipq_ttl = IPFRAGTTL; 1003 fp->ipq_p = ip->ip_p; 1004 fp->ipq_id = ip->ip_id; 1005 fp->ipq_src = ip->ip_src; 1006 fp->ipq_dst = ip->ip_dst; 1007 fp->ipq_frags = m; 1008 m->m_nextpkt = NULL; 1009 goto done; 1010 } else { 1011 fp->ipq_nfrags++; 1012#ifdef MAC 1013 mac_ipq_update(m, fp); 1014#endif 1015 } 1016 1017#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1018 1019 /* 1020 * Handle ECN by comparing this segment with the first one; 1021 * if CE is set, do not lose CE. 1022 * drop if CE and not-ECT are mixed for the same packet. 1023 */ 1024 ecn = ip->ip_tos & IPTOS_ECN_MASK; 1025 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 1026 if (ecn == IPTOS_ECN_CE) { 1027 if (ecn0 == IPTOS_ECN_NOTECT) 1028 goto dropfrag; 1029 if (ecn0 != IPTOS_ECN_CE) 1030 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 1031 } 1032 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 1033 goto dropfrag; 1034 1035 /* 1036 * Find a segment which begins after this one does. 1037 */ 1038 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1039 if (GETIP(q)->ip_off > ip->ip_off) 1040 break; 1041 1042 /* 1043 * If there is a preceding segment, it may provide some of 1044 * our data already. If so, drop the data from the incoming 1045 * segment. If it provides all of our data, drop us, otherwise 1046 * stick new segment in the proper place. 1047 * 1048 * If some of the data is dropped from the the preceding 1049 * segment, then it's checksum is invalidated. 1050 */ 1051 if (p) { 1052 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1053 if (i > 0) { 1054 if (i >= ip->ip_len) 1055 goto dropfrag; 1056 m_adj(m, i); 1057 m->m_pkthdr.csum_flags = 0; 1058 ip->ip_off += i; 1059 ip->ip_len -= i; 1060 } 1061 m->m_nextpkt = p->m_nextpkt; 1062 p->m_nextpkt = m; 1063 } else { 1064 m->m_nextpkt = fp->ipq_frags; 1065 fp->ipq_frags = m; 1066 } 1067 1068 /* 1069 * While we overlap succeeding segments trim them or, 1070 * if they are completely covered, dequeue them. 1071 */ 1072 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1073 q = nq) { 1074 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1075 if (i < GETIP(q)->ip_len) { 1076 GETIP(q)->ip_len -= i; 1077 GETIP(q)->ip_off += i; 1078 m_adj(q, i); 1079 q->m_pkthdr.csum_flags = 0; 1080 break; 1081 } 1082 nq = q->m_nextpkt; 1083 m->m_nextpkt = nq; 1084 IPSTAT_INC(ips_fragdropped); 1085 fp->ipq_nfrags--; 1086 m_freem(q); 1087 } 1088 1089 /* 1090 * Check for complete reassembly and perform frag per packet 1091 * limiting. 1092 * 1093 * Frag limiting is performed here so that the nth frag has 1094 * a chance to complete the packet before we drop the packet. 1095 * As a result, n+1 frags are actually allowed per packet, but 1096 * only n will ever be stored. (n = maxfragsperpacket.) 1097 * 1098 */ 1099 next = 0; 1100 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1101 if (GETIP(q)->ip_off != next) { 1102 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1103 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1104 ip_freef(head, fp); 1105 } 1106 goto done; 1107 } 1108 next += GETIP(q)->ip_len; 1109 } 1110 /* Make sure the last packet didn't have the IP_MF flag */ 1111 if (p->m_flags & M_FRAG) { 1112 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1113 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1114 ip_freef(head, fp); 1115 } 1116 goto done; 1117 } 1118 1119 /* 1120 * Reassembly is complete. Make sure the packet is a sane size. 1121 */ 1122 q = fp->ipq_frags; 1123 ip = GETIP(q); 1124 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1125 IPSTAT_INC(ips_toolong); 1126 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1127 ip_freef(head, fp); 1128 goto done; 1129 } 1130 1131 /* 1132 * Concatenate fragments. 1133 */ 1134 m = q; 1135 t = m->m_next; 1136 m->m_next = NULL; 1137 m_cat(m, t); 1138 nq = q->m_nextpkt; 1139 q->m_nextpkt = NULL; 1140 for (q = nq; q != NULL; q = nq) { 1141 nq = q->m_nextpkt; 1142 q->m_nextpkt = NULL; 1143 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1144 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1145 m_cat(m, q); 1146 } 1147 /* 1148 * In order to do checksumming faster we do 'end-around carry' here 1149 * (and not in for{} loop), though it implies we are not going to 1150 * reassemble more than 64k fragments. 1151 */ 1152 m->m_pkthdr.csum_data = 1153 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1154#ifdef MAC 1155 mac_ipq_reassemble(fp, m); 1156 mac_ipq_destroy(fp); 1157#endif 1158 1159 /* 1160 * Create header for new ip packet by modifying header of first 1161 * packet; dequeue and discard fragment reassembly header. 1162 * Make header visible. 1163 */ 1164 ip->ip_len = (ip->ip_hl << 2) + next; 1165 ip->ip_src = fp->ipq_src; 1166 ip->ip_dst = fp->ipq_dst; 1167 TAILQ_REMOVE(head, fp, ipq_list); 1168 V_nipq--; 1169 uma_zfree(V_ipq_zone, fp); 1170 m->m_len += (ip->ip_hl << 2); 1171 m->m_data -= (ip->ip_hl << 2); 1172 /* some debugging cruft by sklower, below, will go away soon */ 1173 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1174 m_fixhdr(m); 1175 IPSTAT_INC(ips_reassembled); 1176 IPQ_UNLOCK(); 1177 return (m); 1178 1179dropfrag: 1180 IPSTAT_INC(ips_fragdropped); 1181 if (fp != NULL) 1182 fp->ipq_nfrags--; 1183 m_freem(m); 1184done: 1185 IPQ_UNLOCK(); 1186 return (NULL); 1187 1188#undef GETIP 1189} 1190 1191/* 1192 * Free a fragment reassembly header and all 1193 * associated datagrams. 1194 */ 1195static void 1196ip_freef(struct ipqhead *fhp, struct ipq *fp) 1197{ 1198 INIT_VNET_INET(curvnet); 1199 struct mbuf *q; 1200 1201 IPQ_LOCK_ASSERT(); 1202 1203 while (fp->ipq_frags) { 1204 q = fp->ipq_frags; 1205 fp->ipq_frags = q->m_nextpkt; 1206 m_freem(q); 1207 } 1208 TAILQ_REMOVE(fhp, fp, ipq_list); 1209 uma_zfree(V_ipq_zone, fp); 1210 V_nipq--; 1211} 1212 1213/* 1214 * IP timer processing; 1215 * if a timer expires on a reassembly 1216 * queue, discard it. 1217 */ 1218void 1219ip_slowtimo(void) 1220{ 1221 VNET_ITERATOR_DECL(vnet_iter); 1222 struct ipq *fp; 1223 int i; 1224 1225 IPQ_LOCK(); 1226 VNET_LIST_RLOCK(); 1227 VNET_FOREACH(vnet_iter) { 1228 CURVNET_SET(vnet_iter); 1229 INIT_VNET_INET(vnet_iter); 1230 for (i = 0; i < IPREASS_NHASH; i++) { 1231 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) { 1232 struct ipq *fpp; 1233 1234 fpp = fp; 1235 fp = TAILQ_NEXT(fp, ipq_list); 1236 if(--fpp->ipq_ttl == 0) { 1237 IPSTAT_ADD(ips_fragtimeout, 1238 fpp->ipq_nfrags); 1239 ip_freef(&V_ipq[i], fpp); 1240 } 1241 } 1242 } 1243 /* 1244 * If we are over the maximum number of fragments 1245 * (due to the limit being lowered), drain off 1246 * enough to get down to the new limit. 1247 */ 1248 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) { 1249 for (i = 0; i < IPREASS_NHASH; i++) { 1250 while (V_nipq > V_maxnipq && 1251 !TAILQ_EMPTY(&V_ipq[i])) { 1252 IPSTAT_ADD(ips_fragdropped, 1253 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1254 ip_freef(&V_ipq[i], 1255 TAILQ_FIRST(&V_ipq[i])); 1256 } 1257 } 1258 } 1259 CURVNET_RESTORE(); 1260 } 1261 VNET_LIST_RUNLOCK(); 1262 IPQ_UNLOCK(); 1263} 1264 1265/* 1266 * Drain off all datagram fragments. 1267 */ 1268void 1269ip_drain(void) 1270{ 1271 VNET_ITERATOR_DECL(vnet_iter); 1272 int i; 1273 1274 IPQ_LOCK(); 1275 VNET_LIST_RLOCK(); 1276 VNET_FOREACH(vnet_iter) { 1277 CURVNET_SET(vnet_iter); 1278 INIT_VNET_INET(vnet_iter); 1279 for (i = 0; i < IPREASS_NHASH; i++) { 1280 while(!TAILQ_EMPTY(&V_ipq[i])) { 1281 IPSTAT_ADD(ips_fragdropped, 1282 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1283 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); 1284 } 1285 } 1286 CURVNET_RESTORE(); 1287 } 1288 VNET_LIST_RUNLOCK(); 1289 IPQ_UNLOCK(); 1290 in_rtqdrain(); 1291} 1292 1293/* 1294 * The protocol to be inserted into ip_protox[] must be already registered 1295 * in inetsw[], either statically or through pf_proto_register(). 1296 */ 1297int 1298ipproto_register(u_char ipproto) 1299{ 1300 struct protosw *pr; 1301 1302 /* Sanity checks. */ 1303 if (ipproto == 0) 1304 return (EPROTONOSUPPORT); 1305 1306 /* 1307 * The protocol slot must not be occupied by another protocol 1308 * already. An index pointing to IPPROTO_RAW is unused. 1309 */ 1310 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1311 if (pr == NULL) 1312 return (EPFNOSUPPORT); 1313 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1314 return (EEXIST); 1315 1316 /* Find the protocol position in inetsw[] and set the index. */ 1317 for (pr = inetdomain.dom_protosw; 1318 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1319 if (pr->pr_domain->dom_family == PF_INET && 1320 pr->pr_protocol && pr->pr_protocol == ipproto) { 1321 /* Be careful to only index valid IP protocols. */ 1322 if (pr->pr_protocol < IPPROTO_MAX) { 1323 ip_protox[pr->pr_protocol] = pr - inetsw; 1324 return (0); 1325 } else 1326 return (EINVAL); 1327 } 1328 } 1329 return (EPROTONOSUPPORT); 1330} 1331 1332int 1333ipproto_unregister(u_char ipproto) 1334{ 1335 struct protosw *pr; 1336 1337 /* Sanity checks. */ 1338 if (ipproto == 0) 1339 return (EPROTONOSUPPORT); 1340 1341 /* Check if the protocol was indeed registered. */ 1342 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1343 if (pr == NULL) 1344 return (EPFNOSUPPORT); 1345 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1346 return (ENOENT); 1347 1348 /* Reset the protocol slot to IPPROTO_RAW. */ 1349 ip_protox[ipproto] = pr - inetsw; 1350 return (0); 1351} 1352 1353/* 1354 * Given address of next destination (final or next hop), return (referenced) 1355 * internet address info of interface to be used to get there. 1356 */ 1357struct in_ifaddr * 1358ip_rtaddr(struct in_addr dst, u_int fibnum) 1359{ 1360 struct route sro; 1361 struct sockaddr_in *sin; 1362 struct in_ifaddr *ia; 1363 1364 bzero(&sro, sizeof(sro)); 1365 sin = (struct sockaddr_in *)&sro.ro_dst; 1366 sin->sin_family = AF_INET; 1367 sin->sin_len = sizeof(*sin); 1368 sin->sin_addr = dst; 1369 in_rtalloc_ign(&sro, 0, fibnum); 1370 1371 if (sro.ro_rt == NULL) 1372 return (NULL); 1373 1374 ia = ifatoia(sro.ro_rt->rt_ifa); 1375 ifa_ref(&ia->ia_ifa); 1376 RTFREE(sro.ro_rt); 1377 return (ia); 1378} 1379 1380u_char inetctlerrmap[PRC_NCMDS] = { 1381 0, 0, 0, 0, 1382 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1383 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1384 EMSGSIZE, EHOSTUNREACH, 0, 0, 1385 0, 0, EHOSTUNREACH, 0, 1386 ENOPROTOOPT, ECONNREFUSED 1387}; 1388 1389/* 1390 * Forward a packet. If some error occurs return the sender 1391 * an icmp packet. Note we can't always generate a meaningful 1392 * icmp message because icmp doesn't have a large enough repertoire 1393 * of codes and types. 1394 * 1395 * If not forwarding, just drop the packet. This could be confusing 1396 * if ipforwarding was zero but some routing protocol was advancing 1397 * us as a gateway to somewhere. However, we must let the routing 1398 * protocol deal with that. 1399 * 1400 * The srcrt parameter indicates whether the packet is being forwarded 1401 * via a source route. 1402 */ 1403void 1404ip_forward(struct mbuf *m, int srcrt) 1405{ 1406 INIT_VNET_INET(curvnet); 1407 struct ip *ip = mtod(m, struct ip *); 1408 struct in_ifaddr *ia; 1409 struct mbuf *mcopy; 1410 struct in_addr dest; 1411 struct route ro; 1412 int error, type = 0, code = 0, mtu = 0; 1413 1414 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1415 IPSTAT_INC(ips_cantforward); 1416 m_freem(m); 1417 return; 1418 } 1419#ifdef IPSTEALTH 1420 if (!V_ipstealth) { 1421#endif 1422 if (ip->ip_ttl <= IPTTLDEC) { 1423 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1424 0, 0); 1425 return; 1426 } 1427#ifdef IPSTEALTH 1428 } 1429#endif 1430 1431 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); 1432#ifndef IPSEC 1433 /* 1434 * 'ia' may be NULL if there is no route for this destination. 1435 * In case of IPsec, Don't discard it just yet, but pass it to 1436 * ip_output in case of outgoing IPsec policy. 1437 */ 1438 if (!srcrt && ia == NULL) { 1439 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1440 return; 1441 } 1442#endif 1443 1444 /* 1445 * Save the IP header and at most 8 bytes of the payload, 1446 * in case we need to generate an ICMP message to the src. 1447 * 1448 * XXX this can be optimized a lot by saving the data in a local 1449 * buffer on the stack (72 bytes at most), and only allocating the 1450 * mbuf if really necessary. The vast majority of the packets 1451 * are forwarded without having to send an ICMP back (either 1452 * because unnecessary, or because rate limited), so we are 1453 * really we are wasting a lot of work here. 1454 * 1455 * We don't use m_copy() because it might return a reference 1456 * to a shared cluster. Both this function and ip_output() 1457 * assume exclusive access to the IP header in `m', so any 1458 * data in a cluster may change before we reach icmp_error(). 1459 */ 1460 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1461 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1462 /* 1463 * It's probably ok if the pkthdr dup fails (because 1464 * the deep copy of the tag chain failed), but for now 1465 * be conservative and just discard the copy since 1466 * code below may some day want the tags. 1467 */ 1468 m_free(mcopy); 1469 mcopy = NULL; 1470 } 1471 if (mcopy != NULL) { 1472 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy)); 1473 mcopy->m_pkthdr.len = mcopy->m_len; 1474 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1475 } 1476 1477#ifdef IPSTEALTH 1478 if (!V_ipstealth) { 1479#endif 1480 ip->ip_ttl -= IPTTLDEC; 1481#ifdef IPSTEALTH 1482 } 1483#endif 1484 1485 /* 1486 * If forwarding packet using same interface that it came in on, 1487 * perhaps should send a redirect to sender to shortcut a hop. 1488 * Only send redirect if source is sending directly to us, 1489 * and if packet was not source routed (or has any options). 1490 * Also, don't send redirect if forwarding using a default route 1491 * or a route modified by a redirect. 1492 */ 1493 dest.s_addr = 0; 1494 if (!srcrt && V_ipsendredirects && 1495 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) { 1496 struct sockaddr_in *sin; 1497 struct rtentry *rt; 1498 1499 bzero(&ro, sizeof(ro)); 1500 sin = (struct sockaddr_in *)&ro.ro_dst; 1501 sin->sin_family = AF_INET; 1502 sin->sin_len = sizeof(*sin); 1503 sin->sin_addr = ip->ip_dst; 1504 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 1505 1506 rt = ro.ro_rt; 1507 1508 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1509 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1510#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1511 u_long src = ntohl(ip->ip_src.s_addr); 1512 1513 if (RTA(rt) && 1514 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1515 if (rt->rt_flags & RTF_GATEWAY) 1516 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1517 else 1518 dest.s_addr = ip->ip_dst.s_addr; 1519 /* Router requirements says to only send host redirects */ 1520 type = ICMP_REDIRECT; 1521 code = ICMP_REDIRECT_HOST; 1522 } 1523 } 1524 if (rt) 1525 RTFREE(rt); 1526 } 1527 1528 /* 1529 * Try to cache the route MTU from ip_output so we can consider it for 1530 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. 1531 */ 1532 bzero(&ro, sizeof(ro)); 1533 1534 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1535 1536 if (error == EMSGSIZE && ro.ro_rt) 1537 mtu = ro.ro_rt->rt_rmx.rmx_mtu; 1538 if (ro.ro_rt) 1539 RTFREE(ro.ro_rt); 1540 1541 if (error) 1542 IPSTAT_INC(ips_cantforward); 1543 else { 1544 IPSTAT_INC(ips_forward); 1545 if (type) 1546 IPSTAT_INC(ips_redirectsent); 1547 else { 1548 if (mcopy) 1549 m_freem(mcopy); 1550 if (ia != NULL) 1551 ifa_free(&ia->ia_ifa); 1552 return; 1553 } 1554 } 1555 if (mcopy == NULL) { 1556 if (ia != NULL) 1557 ifa_free(&ia->ia_ifa); 1558 return; 1559 } 1560 1561 switch (error) { 1562 1563 case 0: /* forwarded, but need redirect */ 1564 /* type, code set above */ 1565 break; 1566 1567 case ENETUNREACH: 1568 case EHOSTUNREACH: 1569 case ENETDOWN: 1570 case EHOSTDOWN: 1571 default: 1572 type = ICMP_UNREACH; 1573 code = ICMP_UNREACH_HOST; 1574 break; 1575 1576 case EMSGSIZE: 1577 type = ICMP_UNREACH; 1578 code = ICMP_UNREACH_NEEDFRAG; 1579 1580#ifdef IPSEC 1581 /* 1582 * If IPsec is configured for this path, 1583 * override any possibly mtu value set by ip_output. 1584 */ 1585 mtu = ip_ipsec_mtu(m, mtu); 1586#endif /* IPSEC */ 1587 /* 1588 * If the MTU was set before make sure we are below the 1589 * interface MTU. 1590 * If the MTU wasn't set before use the interface mtu or 1591 * fall back to the next smaller mtu step compared to the 1592 * current packet size. 1593 */ 1594 if (mtu != 0) { 1595 if (ia != NULL) 1596 mtu = min(mtu, ia->ia_ifp->if_mtu); 1597 } else { 1598 if (ia != NULL) 1599 mtu = ia->ia_ifp->if_mtu; 1600 else 1601 mtu = ip_next_mtu(ip->ip_len, 0); 1602 } 1603 IPSTAT_INC(ips_cantfrag); 1604 break; 1605 1606 case ENOBUFS: 1607 /* 1608 * A router should not generate ICMP_SOURCEQUENCH as 1609 * required in RFC1812 Requirements for IP Version 4 Routers. 1610 * Source quench could be a big problem under DoS attacks, 1611 * or if the underlying interface is rate-limited. 1612 * Those who need source quench packets may re-enable them 1613 * via the net.inet.ip.sendsourcequench sysctl. 1614 */ 1615 if (V_ip_sendsourcequench == 0) { 1616 m_freem(mcopy); 1617 if (ia != NULL) 1618 ifa_free(&ia->ia_ifa); 1619 return; 1620 } else { 1621 type = ICMP_SOURCEQUENCH; 1622 code = 0; 1623 } 1624 break; 1625 1626 case EACCES: /* ipfw denied packet */ 1627 m_freem(mcopy); 1628 if (ia != NULL) 1629 ifa_free(&ia->ia_ifa); 1630 return; 1631 } 1632 if (ia != NULL) 1633 ifa_free(&ia->ia_ifa); 1634 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1635} 1636 1637void 1638ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1639 struct mbuf *m) 1640{ 1641 INIT_VNET_NET(inp->inp_vnet); 1642 1643 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1644 struct bintime bt; 1645 1646 bintime(&bt); 1647 if (inp->inp_socket->so_options & SO_BINTIME) { 1648 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1649 SCM_BINTIME, SOL_SOCKET); 1650 if (*mp) 1651 mp = &(*mp)->m_next; 1652 } 1653 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1654 struct timeval tv; 1655 1656 bintime2timeval(&bt, &tv); 1657 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1658 SCM_TIMESTAMP, SOL_SOCKET); 1659 if (*mp) 1660 mp = &(*mp)->m_next; 1661 } 1662 } 1663 if (inp->inp_flags & INP_RECVDSTADDR) { 1664 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1665 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1666 if (*mp) 1667 mp = &(*mp)->m_next; 1668 } 1669 if (inp->inp_flags & INP_RECVTTL) { 1670 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1671 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1672 if (*mp) 1673 mp = &(*mp)->m_next; 1674 } 1675#ifdef notyet 1676 /* XXX 1677 * Moving these out of udp_input() made them even more broken 1678 * than they already were. 1679 */ 1680 /* options were tossed already */ 1681 if (inp->inp_flags & INP_RECVOPTS) { 1682 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1683 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1684 if (*mp) 1685 mp = &(*mp)->m_next; 1686 } 1687 /* ip_srcroute doesn't do what we want here, need to fix */ 1688 if (inp->inp_flags & INP_RECVRETOPTS) { 1689 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1690 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1691 if (*mp) 1692 mp = &(*mp)->m_next; 1693 } 1694#endif 1695 if (inp->inp_flags & INP_RECVIF) { 1696 struct ifnet *ifp; 1697 struct sdlbuf { 1698 struct sockaddr_dl sdl; 1699 u_char pad[32]; 1700 } sdlbuf; 1701 struct sockaddr_dl *sdp; 1702 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1703 1704 if (((ifp = m->m_pkthdr.rcvif)) 1705 && ( ifp->if_index && (ifp->if_index <= V_if_index))) { 1706 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1707 /* 1708 * Change our mind and don't try copy. 1709 */ 1710 if ((sdp->sdl_family != AF_LINK) 1711 || (sdp->sdl_len > sizeof(sdlbuf))) { 1712 goto makedummy; 1713 } 1714 bcopy(sdp, sdl2, sdp->sdl_len); 1715 } else { 1716makedummy: 1717 sdl2->sdl_len 1718 = offsetof(struct sockaddr_dl, sdl_data[0]); 1719 sdl2->sdl_family = AF_LINK; 1720 sdl2->sdl_index = 0; 1721 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1722 } 1723 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1724 IP_RECVIF, IPPROTO_IP); 1725 if (*mp) 1726 mp = &(*mp)->m_next; 1727 } 1728} 1729 1730/* 1731 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1732 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1733 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1734 * compiled. 1735 */ 1736int 1737ip_rsvp_init(struct socket *so) 1738{ 1739 INIT_VNET_INET(so->so_vnet); 1740 1741 if (so->so_type != SOCK_RAW || 1742 so->so_proto->pr_protocol != IPPROTO_RSVP) 1743 return EOPNOTSUPP; 1744 1745 if (V_ip_rsvpd != NULL) 1746 return EADDRINUSE; 1747 1748 V_ip_rsvpd = so; 1749 /* 1750 * This may seem silly, but we need to be sure we don't over-increment 1751 * the RSVP counter, in case something slips up. 1752 */ 1753 if (!V_ip_rsvp_on) { 1754 V_ip_rsvp_on = 1; 1755 V_rsvp_on++; 1756 } 1757 1758 return 0; 1759} 1760 1761int 1762ip_rsvp_done(void) 1763{ 1764 INIT_VNET_INET(curvnet); 1765 1766 V_ip_rsvpd = NULL; 1767 /* 1768 * This may seem silly, but we need to be sure we don't over-decrement 1769 * the RSVP counter, in case something slips up. 1770 */ 1771 if (V_ip_rsvp_on) { 1772 V_ip_rsvp_on = 0; 1773 V_rsvp_on--; 1774 } 1775 return 0; 1776} 1777 1778void 1779rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1780{ 1781 INIT_VNET_INET(curvnet); 1782 1783 if (rsvp_input_p) { /* call the real one if loaded */ 1784 rsvp_input_p(m, off); 1785 return; 1786 } 1787 1788 /* Can still get packets with rsvp_on = 0 if there is a local member 1789 * of the group to which the RSVP packet is addressed. But in this 1790 * case we want to throw the packet away. 1791 */ 1792 1793 if (!V_rsvp_on) { 1794 m_freem(m); 1795 return; 1796 } 1797 1798 if (V_ip_rsvpd != NULL) { 1799 rip_input(m, off); 1800 return; 1801 } 1802 /* Drop the packet */ 1803 m_freem(m); 1804} 1805