ip_input.c revision 195699
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 195699 2009-07-14 22:48:30Z 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_fw.h> 75#include <netinet/ip_icmp.h> 76#include <netinet/ip_options.h> 77#include <machine/in_cksum.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 93static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */ 94static VNET_DEFINE(int, ip_checkinterface); 95static VNET_DEFINE(int, ip_keepfaith); 96static VNET_DEFINE(int, ip_sendsourcequench); 97 98#define V_ipsendredirects VNET_GET(ipsendredirects) 99#define V_ip_checkinterface VNET_GET(ip_checkinterface) 100#define V_ip_keepfaith VNET_GET(ip_keepfaith) 101#define V_ip_sendsourcequench VNET_GET(ip_sendsourcequench) 102 103VNET_DEFINE(int, ip_defttl) = IPDEFTTL; 104VNET_DEFINE(int, ip_do_randomid); 105VNET_DEFINE(int, ipforwarding); 106 107VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */ 108VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */ 109VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */ 110VNET_DEFINE(struct ipstat, ipstat); 111 112static VNET_DEFINE(int, ip_rsvp_on); 113VNET_DEFINE(struct socket *, ip_rsvpd); 114VNET_DEFINE(int, rsvp_on); 115 116#define V_ip_rsvp_on VNET_GET(ip_rsvp_on) 117 118static VNET_DEFINE(TAILQ_HEAD(ipqhead, ipq), ipq[IPREASS_NHASH]); 119static VNET_DEFINE(int, maxnipq); /* Administrative limit on # reass queues. */ 120static VNET_DEFINE(int, maxfragsperpacket); 121static VNET_DEFINE(int, nipq); /* Total # of reass queues */ 122 123#define V_ipq VNET_GET(ipq) 124#define V_maxnipq VNET_GET(maxnipq) 125#define V_maxfragsperpacket VNET_GET(maxfragsperpacket) 126#define V_nipq VNET_GET(nipq) 127 128VNET_DEFINE(int, ipstealth); 129 130struct rwlock in_ifaddr_lock; 131RW_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock"); 132 133SYSCTL_VNET_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 134 &VNET_NAME(ipforwarding), 0, 135 "Enable IP forwarding between interfaces"); 136 137SYSCTL_VNET_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 138 &VNET_NAME(ipsendredirects), 0, 139 "Enable sending IP redirects"); 140 141SYSCTL_VNET_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 142 &VNET_NAME(ip_defttl), 0, 143 "Maximum TTL on IP packets"); 144 145SYSCTL_VNET_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 146 &VNET_NAME(ip_keepfaith), 0, 147 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 148 149SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 150 &VNET_NAME(ip_sendsourcequench), 0, 151 "Enable the transmission of source quench packets"); 152 153SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 154 &VNET_NAME(ip_do_randomid), 0, 155 "Assign random ip_id values"); 156 157/* 158 * XXX - Setting ip_checkinterface mostly implements the receive side of 159 * the Strong ES model described in RFC 1122, but since the routing table 160 * and transmit implementation do not implement the Strong ES model, 161 * setting this to 1 results in an odd hybrid. 162 * 163 * XXX - ip_checkinterface currently must be disabled if you use ipnat 164 * to translate the destination address to another local interface. 165 * 166 * XXX - ip_checkinterface must be disabled if you add IP aliases 167 * to the loopback interface instead of the interface where the 168 * packets for those addresses are received. 169 */ 170SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 171 &VNET_NAME(ip_checkinterface), 0, 172 "Verify packet arrives on correct interface"); 173 174struct pfil_head inet_pfil_hook; /* Packet filter hooks */ 175 176static struct netisr_handler ip_nh = { 177 .nh_name = "ip", 178 .nh_handler = ip_input, 179 .nh_proto = NETISR_IP, 180 .nh_policy = NETISR_POLICY_FLOW, 181}; 182 183extern struct domain inetdomain; 184extern struct protosw inetsw[]; 185u_char ip_protox[IPPROTO_MAX]; 186 187SYSCTL_VNET_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 188 &VNET_NAME(ipstat), ipstat, 189 "IP statistics (struct ipstat, netinet/ip_var.h)"); 190 191static VNET_DEFINE(uma_zone_t, ipq_zone); 192#define V_ipq_zone VNET_GET(ipq_zone) 193 194static struct mtx ipqlock; 195 196#define IPQ_LOCK() mtx_lock(&ipqlock) 197#define IPQ_UNLOCK() mtx_unlock(&ipqlock) 198#define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 199#define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 200 201static void maxnipq_update(void); 202static void ipq_zone_change(void *); 203 204SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, 205 &VNET_NAME(nipq), 0, 206 "Current number of IPv4 fragment reassembly queue entries"); 207 208SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 209 &VNET_NAME(maxfragsperpacket), 0, 210 "Maximum number of IPv4 fragments allowed per packet"); 211 212struct callout ipport_tick_callout; 213 214#ifdef IPCTL_DEFMTU 215SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 216 &ip_mtu, 0, "Default MTU"); 217#endif 218 219#ifdef IPSTEALTH 220SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 221 &VNET_NAME(ipstealth), 0, 222 "IP stealth mode, no TTL decrementation on forwarding"); 223#endif 224 225#ifdef FLOWTABLE 226static VNET_DEFINE(int, ip_output_flowtable_size) = 2048; 227VNET_DEFINE(struct flowtable *, ip_ft); 228#define V_ip_output_flowtable_size VNET_GET(ip_output_flowtable_size) 229 230SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, output_flowtable_size, CTLFLAG_RDTUN, 231 &VNET_NAME(ip_output_flowtable_size), 2048, 232 "number of entries in the per-cpu output flow caches"); 233#endif 234 235VNET_DEFINE(int, fw_one_pass) = 1; 236 237static void ip_freef(struct ipqhead *, struct ipq *); 238 239#ifdef VIMAGE 240/* XXX only has to stay for .vmi_dependson elsewhere. */ 241static void vnet_inet_register(void); 242 243static const vnet_modinfo_t vnet_inet_modinfo = { 244 .vmi_id = VNET_MOD_INET, 245 .vmi_name = "inet", 246}; 247 248static void vnet_inet_register() 249{ 250 251 vnet_mod_register(&vnet_inet_modinfo); 252} 253 254SYSINIT(inet, SI_SUB_PROTO_BEGIN, SI_ORDER_FIRST, vnet_inet_register, 0); 255#endif 256 257static int 258sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 259{ 260 int error, qlimit; 261 262 netisr_getqlimit(&ip_nh, &qlimit); 263 error = sysctl_handle_int(oidp, &qlimit, 0, req); 264 if (error || !req->newptr) 265 return (error); 266 if (qlimit < 1) 267 return (EINVAL); 268 return (netisr_setqlimit(&ip_nh, qlimit)); 269} 270SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, 271 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I", 272 "Maximum size of the IP input queue"); 273 274static int 275sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) 276{ 277 u_int64_t qdrops_long; 278 int error, qdrops; 279 280 netisr_getqdrops(&ip_nh, &qdrops_long); 281 qdrops = qdrops_long; 282 error = sysctl_handle_int(oidp, &qdrops, 0, req); 283 if (error || !req->newptr) 284 return (error); 285 if (qdrops != 0) 286 return (EINVAL); 287 netisr_clearqdrops(&ip_nh); 288 return (0); 289} 290 291SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, 292 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I", 293 "Number of packets dropped from the IP input queue"); 294 295/* 296 * IP initialization: fill in IP protocol switch table. 297 * All protocols not implemented in kernel go to raw IP protocol handler. 298 */ 299void 300ip_init(void) 301{ 302 struct protosw *pr; 303 int i; 304 305 V_ip_id = time_second & 0xffff; 306 307 TAILQ_INIT(&V_in_ifaddrhead); 308 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 309 310 /* Initialize IP reassembly queue. */ 311 for (i = 0; i < IPREASS_NHASH; i++) 312 TAILQ_INIT(&V_ipq[i]); 313 V_maxnipq = nmbclusters / 32; 314 V_maxfragsperpacket = 16; 315 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 316 NULL, UMA_ALIGN_PTR, 0); 317 maxnipq_update(); 318 319#ifdef FLOWTABLE 320 TUNABLE_INT_FETCH("net.inet.ip.output_flowtable_size", 321 &V_ip_output_flowtable_size); 322 V_ip_ft = flowtable_alloc(V_ip_output_flowtable_size, FL_PCPU); 323#endif 324 325 /* Skip initialization of globals for non-default instances. */ 326 if (!IS_DEFAULT_VNET(curvnet)) 327 return; 328 329 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 330 if (pr == NULL) 331 panic("ip_init: PF_INET not found"); 332 333 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 334 for (i = 0; i < IPPROTO_MAX; i++) 335 ip_protox[i] = pr - inetsw; 336 /* 337 * Cycle through IP protocols and put them into the appropriate place 338 * in ip_protox[]. 339 */ 340 for (pr = inetdomain.dom_protosw; 341 pr < inetdomain.dom_protoswNPROTOSW; pr++) 342 if (pr->pr_domain->dom_family == PF_INET && 343 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 344 /* Be careful to only index valid IP protocols. */ 345 if (pr->pr_protocol < IPPROTO_MAX) 346 ip_protox[pr->pr_protocol] = pr - inetsw; 347 } 348 349 /* Initialize packet filter hooks. */ 350 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 351 inet_pfil_hook.ph_af = AF_INET; 352 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 353 printf("%s: WARNING: unable to register pfil hook, " 354 "error %d\n", __func__, i); 355 356 /* Start ipport_tick. */ 357 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE); 358 callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL); 359 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, 360 SHUTDOWN_PRI_DEFAULT); 361 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, 362 NULL, EVENTHANDLER_PRI_ANY); 363 364 /* Initialize various other remaining things. */ 365 IPQ_LOCK_INIT(); 366 netisr_register(&ip_nh); 367} 368 369void 370ip_fini(void *xtp) 371{ 372 373 callout_stop(&ipport_tick_callout); 374} 375 376/* 377 * Ip input routine. Checksum and byte swap header. If fragmented 378 * try to reassemble. Process options. Pass to next level. 379 */ 380void 381ip_input(struct mbuf *m) 382{ 383 struct ip *ip = NULL; 384 struct in_ifaddr *ia = NULL; 385 struct ifaddr *ifa; 386 struct ifnet *ifp; 387 int checkif, hlen = 0; 388 u_short sum; 389 int dchg = 0; /* dest changed after fw */ 390 struct in_addr odst; /* original dst address */ 391 392 M_ASSERTPKTHDR(m); 393 394 if (m->m_flags & M_FASTFWD_OURS) { 395 /* 396 * Firewall or NAT changed destination to local. 397 * We expect ip_len and ip_off to be in host byte order. 398 */ 399 m->m_flags &= ~M_FASTFWD_OURS; 400 /* Set up some basics that will be used later. */ 401 ip = mtod(m, struct ip *); 402 hlen = ip->ip_hl << 2; 403 goto ours; 404 } 405 406 IPSTAT_INC(ips_total); 407 408 if (m->m_pkthdr.len < sizeof(struct ip)) 409 goto tooshort; 410 411 if (m->m_len < sizeof (struct ip) && 412 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 413 IPSTAT_INC(ips_toosmall); 414 return; 415 } 416 ip = mtod(m, struct ip *); 417 418 if (ip->ip_v != IPVERSION) { 419 IPSTAT_INC(ips_badvers); 420 goto bad; 421 } 422 423 hlen = ip->ip_hl << 2; 424 if (hlen < sizeof(struct ip)) { /* minimum header length */ 425 IPSTAT_INC(ips_badhlen); 426 goto bad; 427 } 428 if (hlen > m->m_len) { 429 if ((m = m_pullup(m, hlen)) == NULL) { 430 IPSTAT_INC(ips_badhlen); 431 return; 432 } 433 ip = mtod(m, struct ip *); 434 } 435 436 /* 127/8 must not appear on wire - RFC1122 */ 437 ifp = m->m_pkthdr.rcvif; 438 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 439 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 440 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 441 IPSTAT_INC(ips_badaddr); 442 goto bad; 443 } 444 } 445 446 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 447 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 448 } else { 449 if (hlen == sizeof(struct ip)) { 450 sum = in_cksum_hdr(ip); 451 } else { 452 sum = in_cksum(m, hlen); 453 } 454 } 455 if (sum) { 456 IPSTAT_INC(ips_badsum); 457 goto bad; 458 } 459 460#ifdef ALTQ 461 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 462 /* packet is dropped by traffic conditioner */ 463 return; 464#endif 465 466 /* 467 * Convert fields to host representation. 468 */ 469 ip->ip_len = ntohs(ip->ip_len); 470 if (ip->ip_len < hlen) { 471 IPSTAT_INC(ips_badlen); 472 goto bad; 473 } 474 ip->ip_off = ntohs(ip->ip_off); 475 476 /* 477 * Check that the amount of data in the buffers 478 * is as at least much as the IP header would have us expect. 479 * Trim mbufs if longer than we expect. 480 * Drop packet if shorter than we expect. 481 */ 482 if (m->m_pkthdr.len < ip->ip_len) { 483tooshort: 484 IPSTAT_INC(ips_tooshort); 485 goto bad; 486 } 487 if (m->m_pkthdr.len > ip->ip_len) { 488 if (m->m_len == m->m_pkthdr.len) { 489 m->m_len = ip->ip_len; 490 m->m_pkthdr.len = ip->ip_len; 491 } else 492 m_adj(m, ip->ip_len - m->m_pkthdr.len); 493 } 494#ifdef IPSEC 495 /* 496 * Bypass packet filtering for packets from a tunnel (gif). 497 */ 498 if (ip_ipsec_filtertunnel(m)) 499 goto passin; 500#endif /* IPSEC */ 501 502 /* 503 * Run through list of hooks for input packets. 504 * 505 * NB: Beware of the destination address changing (e.g. 506 * by NAT rewriting). When this happens, tell 507 * ip_forward to do the right thing. 508 */ 509 510 /* Jump over all PFIL processing if hooks are not active. */ 511 if (!PFIL_HOOKED(&inet_pfil_hook)) 512 goto passin; 513 514 odst = ip->ip_dst; 515 if (pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0) 516 return; 517 if (m == NULL) /* consumed by filter */ 518 return; 519 520 ip = mtod(m, struct ip *); 521 dchg = (odst.s_addr != ip->ip_dst.s_addr); 522 ifp = m->m_pkthdr.rcvif; 523 524#ifdef IPFIREWALL_FORWARD 525 if (m->m_flags & M_FASTFWD_OURS) { 526 m->m_flags &= ~M_FASTFWD_OURS; 527 goto ours; 528 } 529 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 530 /* 531 * Directly ship on the packet. This allows to forward packets 532 * that were destined for us to some other directly connected 533 * host. 534 */ 535 ip_forward(m, dchg); 536 return; 537 } 538#endif /* IPFIREWALL_FORWARD */ 539 540passin: 541 /* 542 * Process options and, if not destined for us, 543 * ship it on. ip_dooptions returns 1 when an 544 * error was detected (causing an icmp message 545 * to be sent and the original packet to be freed). 546 */ 547 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 548 return; 549 550 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 551 * matter if it is destined to another node, or whether it is 552 * a multicast one, RSVP wants it! and prevents it from being forwarded 553 * anywhere else. Also checks if the rsvp daemon is running before 554 * grabbing the packet. 555 */ 556 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 557 goto ours; 558 559 /* 560 * Check our list of addresses, to see if the packet is for us. 561 * If we don't have any addresses, assume any unicast packet 562 * we receive might be for us (and let the upper layers deal 563 * with it). 564 */ 565 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 566 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 567 goto ours; 568 569 /* 570 * Enable a consistency check between the destination address 571 * and the arrival interface for a unicast packet (the RFC 1122 572 * strong ES model) if IP forwarding is disabled and the packet 573 * is not locally generated and the packet is not subject to 574 * 'ipfw fwd'. 575 * 576 * XXX - Checking also should be disabled if the destination 577 * address is ipnat'ed to a different interface. 578 * 579 * XXX - Checking is incompatible with IP aliases added 580 * to the loopback interface instead of the interface where 581 * the packets are received. 582 * 583 * XXX - This is the case for carp vhost IPs as well so we 584 * insert a workaround. If the packet got here, we already 585 * checked with carp_iamatch() and carp_forus(). 586 */ 587 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 588 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && 589#ifdef DEV_CARP 590 !ifp->if_carp && 591#endif 592 (dchg == 0); 593 594 /* 595 * Check for exact addresses in the hash bucket. 596 */ 597 /* IN_IFADDR_RLOCK(); */ 598 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 599 /* 600 * If the address matches, verify that the packet 601 * arrived via the correct interface if checking is 602 * enabled. 603 */ 604 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 605 (!checkif || ia->ia_ifp == ifp)) { 606 ifa_ref(&ia->ia_ifa); 607 /* IN_IFADDR_RUNLOCK(); */ 608 goto ours; 609 } 610 } 611 /* IN_IFADDR_RUNLOCK(); */ 612 613 /* 614 * Check for broadcast addresses. 615 * 616 * Only accept broadcast packets that arrive via the matching 617 * interface. Reception of forwarded directed broadcasts would 618 * be handled via ip_forward() and ether_output() with the loopback 619 * into the stack for SIMPLEX interfaces handled by ether_output(). 620 */ 621 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 622 IF_ADDR_LOCK(ifp); 623 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 624 if (ifa->ifa_addr->sa_family != AF_INET) 625 continue; 626 ia = ifatoia(ifa); 627 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 628 ip->ip_dst.s_addr) { 629 ifa_ref(ifa); 630 IF_ADDR_UNLOCK(ifp); 631 goto ours; 632 } 633 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) { 634 ifa_ref(ifa); 635 IF_ADDR_UNLOCK(ifp); 636 goto ours; 637 } 638#ifdef BOOTP_COMPAT 639 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { 640 ifa_ref(ifa); 641 IF_ADDR_UNLOCK(ifp); 642 goto ours; 643 } 644#endif 645 } 646 IF_ADDR_UNLOCK(ifp); 647 ia = NULL; 648 } 649 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 650 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 651 IPSTAT_INC(ips_cantforward); 652 m_freem(m); 653 return; 654 } 655 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 656 if (V_ip_mrouter) { 657 /* 658 * If we are acting as a multicast router, all 659 * incoming multicast packets are passed to the 660 * kernel-level multicast forwarding function. 661 * The packet is returned (relatively) intact; if 662 * ip_mforward() returns a non-zero value, the packet 663 * must be discarded, else it may be accepted below. 664 */ 665 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 666 IPSTAT_INC(ips_cantforward); 667 m_freem(m); 668 return; 669 } 670 671 /* 672 * The process-level routing daemon needs to receive 673 * all multicast IGMP packets, whether or not this 674 * host belongs to their destination groups. 675 */ 676 if (ip->ip_p == IPPROTO_IGMP) 677 goto ours; 678 IPSTAT_INC(ips_forward); 679 } 680 /* 681 * Assume the packet is for us, to avoid prematurely taking 682 * a lock on the in_multi hash. Protocols must perform 683 * their own filtering and update statistics accordingly. 684 */ 685 goto ours; 686 } 687 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 688 goto ours; 689 if (ip->ip_dst.s_addr == INADDR_ANY) 690 goto ours; 691 692 /* 693 * FAITH(Firewall Aided Internet Translator) 694 */ 695 if (ifp && ifp->if_type == IFT_FAITH) { 696 if (V_ip_keepfaith) { 697 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 698 goto ours; 699 } 700 m_freem(m); 701 return; 702 } 703 704 /* 705 * Not for us; forward if possible and desirable. 706 */ 707 if (V_ipforwarding == 0) { 708 IPSTAT_INC(ips_cantforward); 709 m_freem(m); 710 } else { 711#ifdef IPSEC 712 if (ip_ipsec_fwd(m)) 713 goto bad; 714#endif /* IPSEC */ 715 ip_forward(m, dchg); 716 } 717 return; 718 719ours: 720#ifdef IPSTEALTH 721 /* 722 * IPSTEALTH: Process non-routing options only 723 * if the packet is destined for us. 724 */ 725 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) { 726 if (ia != NULL) 727 ifa_free(&ia->ia_ifa); 728 return; 729 } 730#endif /* IPSTEALTH */ 731 732 /* Count the packet in the ip address stats */ 733 if (ia != NULL) { 734 ia->ia_ifa.if_ipackets++; 735 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 736 ifa_free(&ia->ia_ifa); 737 } 738 739 /* 740 * Attempt reassembly; if it succeeds, proceed. 741 * ip_reass() will return a different mbuf. 742 */ 743 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 744 m = ip_reass(m); 745 if (m == NULL) 746 return; 747 ip = mtod(m, struct ip *); 748 /* Get the header length of the reassembled packet */ 749 hlen = ip->ip_hl << 2; 750 } 751 752 /* 753 * Further protocols expect the packet length to be w/o the 754 * IP header. 755 */ 756 ip->ip_len -= hlen; 757 758#ifdef IPSEC 759 /* 760 * enforce IPsec policy checking if we are seeing last header. 761 * note that we do not visit this with protocols with pcb layer 762 * code - like udp/tcp/raw ip. 763 */ 764 if (ip_ipsec_input(m)) 765 goto bad; 766#endif /* IPSEC */ 767 768 /* 769 * Switch out to protocol's input routine. 770 */ 771 IPSTAT_INC(ips_delivered); 772 773 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 774 return; 775bad: 776 m_freem(m); 777} 778 779/* 780 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 781 * max has slightly different semantics than the sysctl, for historical 782 * reasons. 783 */ 784static void 785maxnipq_update(void) 786{ 787 788 /* 789 * -1 for unlimited allocation. 790 */ 791 if (V_maxnipq < 0) 792 uma_zone_set_max(V_ipq_zone, 0); 793 /* 794 * Positive number for specific bound. 795 */ 796 if (V_maxnipq > 0) 797 uma_zone_set_max(V_ipq_zone, V_maxnipq); 798 /* 799 * Zero specifies no further fragment queue allocation -- set the 800 * bound very low, but rely on implementation elsewhere to actually 801 * prevent allocation and reclaim current queues. 802 */ 803 if (V_maxnipq == 0) 804 uma_zone_set_max(V_ipq_zone, 1); 805} 806 807static void 808ipq_zone_change(void *tag) 809{ 810 811 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) { 812 V_maxnipq = nmbclusters / 32; 813 maxnipq_update(); 814 } 815} 816 817static int 818sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 819{ 820 int error, i; 821 822 i = V_maxnipq; 823 error = sysctl_handle_int(oidp, &i, 0, req); 824 if (error || !req->newptr) 825 return (error); 826 827 /* 828 * XXXRW: Might be a good idea to sanity check the argument and place 829 * an extreme upper bound. 830 */ 831 if (i < -1) 832 return (EINVAL); 833 V_maxnipq = i; 834 maxnipq_update(); 835 return (0); 836} 837 838SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 839 NULL, 0, sysctl_maxnipq, "I", 840 "Maximum number of IPv4 fragment reassembly queue entries"); 841 842/* 843 * Take incoming datagram fragment and try to reassemble it into 844 * whole datagram. If the argument is the first fragment or one 845 * in between the function will return NULL and store the mbuf 846 * in the fragment chain. If the argument is the last fragment 847 * the packet will be reassembled and the pointer to the new 848 * mbuf returned for further processing. Only m_tags attached 849 * to the first packet/fragment are preserved. 850 * The IP header is *NOT* adjusted out of iplen. 851 */ 852struct mbuf * 853ip_reass(struct mbuf *m) 854{ 855 struct ip *ip; 856 struct mbuf *p, *q, *nq, *t; 857 struct ipq *fp = NULL; 858 struct ipqhead *head; 859 int i, hlen, next; 860 u_int8_t ecn, ecn0; 861 u_short hash; 862 863 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 864 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) { 865 IPSTAT_INC(ips_fragments); 866 IPSTAT_INC(ips_fragdropped); 867 m_freem(m); 868 return (NULL); 869 } 870 871 ip = mtod(m, struct ip *); 872 hlen = ip->ip_hl << 2; 873 874 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 875 head = &V_ipq[hash]; 876 IPQ_LOCK(); 877 878 /* 879 * Look for queue of fragments 880 * of this datagram. 881 */ 882 TAILQ_FOREACH(fp, head, ipq_list) 883 if (ip->ip_id == fp->ipq_id && 884 ip->ip_src.s_addr == fp->ipq_src.s_addr && 885 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 886#ifdef MAC 887 mac_ipq_match(m, fp) && 888#endif 889 ip->ip_p == fp->ipq_p) 890 goto found; 891 892 fp = NULL; 893 894 /* 895 * Attempt to trim the number of allocated fragment queues if it 896 * exceeds the administrative limit. 897 */ 898 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) { 899 /* 900 * drop something from the tail of the current queue 901 * before proceeding further 902 */ 903 struct ipq *q = TAILQ_LAST(head, ipqhead); 904 if (q == NULL) { /* gak */ 905 for (i = 0; i < IPREASS_NHASH; i++) { 906 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead); 907 if (r) { 908 IPSTAT_ADD(ips_fragtimeout, 909 r->ipq_nfrags); 910 ip_freef(&V_ipq[i], r); 911 break; 912 } 913 } 914 } else { 915 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags); 916 ip_freef(head, q); 917 } 918 } 919 920found: 921 /* 922 * Adjust ip_len to not reflect header, 923 * convert offset of this to bytes. 924 */ 925 ip->ip_len -= hlen; 926 if (ip->ip_off & IP_MF) { 927 /* 928 * Make sure that fragments have a data length 929 * that's a non-zero multiple of 8 bytes. 930 */ 931 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 932 IPSTAT_INC(ips_toosmall); /* XXX */ 933 goto dropfrag; 934 } 935 m->m_flags |= M_FRAG; 936 } else 937 m->m_flags &= ~M_FRAG; 938 ip->ip_off <<= 3; 939 940 941 /* 942 * Attempt reassembly; if it succeeds, proceed. 943 * ip_reass() will return a different mbuf. 944 */ 945 IPSTAT_INC(ips_fragments); 946 m->m_pkthdr.header = ip; 947 948 /* Previous ip_reass() started here. */ 949 /* 950 * Presence of header sizes in mbufs 951 * would confuse code below. 952 */ 953 m->m_data += hlen; 954 m->m_len -= hlen; 955 956 /* 957 * If first fragment to arrive, create a reassembly queue. 958 */ 959 if (fp == NULL) { 960 fp = uma_zalloc(V_ipq_zone, M_NOWAIT); 961 if (fp == NULL) 962 goto dropfrag; 963#ifdef MAC 964 if (mac_ipq_init(fp, M_NOWAIT) != 0) { 965 uma_zfree(V_ipq_zone, fp); 966 fp = NULL; 967 goto dropfrag; 968 } 969 mac_ipq_create(m, fp); 970#endif 971 TAILQ_INSERT_HEAD(head, fp, ipq_list); 972 V_nipq++; 973 fp->ipq_nfrags = 1; 974 fp->ipq_ttl = IPFRAGTTL; 975 fp->ipq_p = ip->ip_p; 976 fp->ipq_id = ip->ip_id; 977 fp->ipq_src = ip->ip_src; 978 fp->ipq_dst = ip->ip_dst; 979 fp->ipq_frags = m; 980 m->m_nextpkt = NULL; 981 goto done; 982 } else { 983 fp->ipq_nfrags++; 984#ifdef MAC 985 mac_ipq_update(m, fp); 986#endif 987 } 988 989#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 990 991 /* 992 * Handle ECN by comparing this segment with the first one; 993 * if CE is set, do not lose CE. 994 * drop if CE and not-ECT are mixed for the same packet. 995 */ 996 ecn = ip->ip_tos & IPTOS_ECN_MASK; 997 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 998 if (ecn == IPTOS_ECN_CE) { 999 if (ecn0 == IPTOS_ECN_NOTECT) 1000 goto dropfrag; 1001 if (ecn0 != IPTOS_ECN_CE) 1002 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 1003 } 1004 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 1005 goto dropfrag; 1006 1007 /* 1008 * Find a segment which begins after this one does. 1009 */ 1010 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1011 if (GETIP(q)->ip_off > ip->ip_off) 1012 break; 1013 1014 /* 1015 * If there is a preceding segment, it may provide some of 1016 * our data already. If so, drop the data from the incoming 1017 * segment. If it provides all of our data, drop us, otherwise 1018 * stick new segment in the proper place. 1019 * 1020 * If some of the data is dropped from the the preceding 1021 * segment, then it's checksum is invalidated. 1022 */ 1023 if (p) { 1024 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1025 if (i > 0) { 1026 if (i >= ip->ip_len) 1027 goto dropfrag; 1028 m_adj(m, i); 1029 m->m_pkthdr.csum_flags = 0; 1030 ip->ip_off += i; 1031 ip->ip_len -= i; 1032 } 1033 m->m_nextpkt = p->m_nextpkt; 1034 p->m_nextpkt = m; 1035 } else { 1036 m->m_nextpkt = fp->ipq_frags; 1037 fp->ipq_frags = m; 1038 } 1039 1040 /* 1041 * While we overlap succeeding segments trim them or, 1042 * if they are completely covered, dequeue them. 1043 */ 1044 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1045 q = nq) { 1046 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1047 if (i < GETIP(q)->ip_len) { 1048 GETIP(q)->ip_len -= i; 1049 GETIP(q)->ip_off += i; 1050 m_adj(q, i); 1051 q->m_pkthdr.csum_flags = 0; 1052 break; 1053 } 1054 nq = q->m_nextpkt; 1055 m->m_nextpkt = nq; 1056 IPSTAT_INC(ips_fragdropped); 1057 fp->ipq_nfrags--; 1058 m_freem(q); 1059 } 1060 1061 /* 1062 * Check for complete reassembly and perform frag per packet 1063 * limiting. 1064 * 1065 * Frag limiting is performed here so that the nth frag has 1066 * a chance to complete the packet before we drop the packet. 1067 * As a result, n+1 frags are actually allowed per packet, but 1068 * only n will ever be stored. (n = maxfragsperpacket.) 1069 * 1070 */ 1071 next = 0; 1072 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1073 if (GETIP(q)->ip_off != next) { 1074 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1075 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1076 ip_freef(head, fp); 1077 } 1078 goto done; 1079 } 1080 next += GETIP(q)->ip_len; 1081 } 1082 /* Make sure the last packet didn't have the IP_MF flag */ 1083 if (p->m_flags & M_FRAG) { 1084 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1085 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1086 ip_freef(head, fp); 1087 } 1088 goto done; 1089 } 1090 1091 /* 1092 * Reassembly is complete. Make sure the packet is a sane size. 1093 */ 1094 q = fp->ipq_frags; 1095 ip = GETIP(q); 1096 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1097 IPSTAT_INC(ips_toolong); 1098 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1099 ip_freef(head, fp); 1100 goto done; 1101 } 1102 1103 /* 1104 * Concatenate fragments. 1105 */ 1106 m = q; 1107 t = m->m_next; 1108 m->m_next = NULL; 1109 m_cat(m, t); 1110 nq = q->m_nextpkt; 1111 q->m_nextpkt = NULL; 1112 for (q = nq; q != NULL; q = nq) { 1113 nq = q->m_nextpkt; 1114 q->m_nextpkt = NULL; 1115 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1116 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1117 m_cat(m, q); 1118 } 1119 /* 1120 * In order to do checksumming faster we do 'end-around carry' here 1121 * (and not in for{} loop), though it implies we are not going to 1122 * reassemble more than 64k fragments. 1123 */ 1124 m->m_pkthdr.csum_data = 1125 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1126#ifdef MAC 1127 mac_ipq_reassemble(fp, m); 1128 mac_ipq_destroy(fp); 1129#endif 1130 1131 /* 1132 * Create header for new ip packet by modifying header of first 1133 * packet; dequeue and discard fragment reassembly header. 1134 * Make header visible. 1135 */ 1136 ip->ip_len = (ip->ip_hl << 2) + next; 1137 ip->ip_src = fp->ipq_src; 1138 ip->ip_dst = fp->ipq_dst; 1139 TAILQ_REMOVE(head, fp, ipq_list); 1140 V_nipq--; 1141 uma_zfree(V_ipq_zone, fp); 1142 m->m_len += (ip->ip_hl << 2); 1143 m->m_data -= (ip->ip_hl << 2); 1144 /* some debugging cruft by sklower, below, will go away soon */ 1145 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1146 m_fixhdr(m); 1147 IPSTAT_INC(ips_reassembled); 1148 IPQ_UNLOCK(); 1149 return (m); 1150 1151dropfrag: 1152 IPSTAT_INC(ips_fragdropped); 1153 if (fp != NULL) 1154 fp->ipq_nfrags--; 1155 m_freem(m); 1156done: 1157 IPQ_UNLOCK(); 1158 return (NULL); 1159 1160#undef GETIP 1161} 1162 1163/* 1164 * Free a fragment reassembly header and all 1165 * associated datagrams. 1166 */ 1167static void 1168ip_freef(struct ipqhead *fhp, struct ipq *fp) 1169{ 1170 struct mbuf *q; 1171 1172 IPQ_LOCK_ASSERT(); 1173 1174 while (fp->ipq_frags) { 1175 q = fp->ipq_frags; 1176 fp->ipq_frags = q->m_nextpkt; 1177 m_freem(q); 1178 } 1179 TAILQ_REMOVE(fhp, fp, ipq_list); 1180 uma_zfree(V_ipq_zone, fp); 1181 V_nipq--; 1182} 1183 1184/* 1185 * IP timer processing; 1186 * if a timer expires on a reassembly 1187 * queue, discard it. 1188 */ 1189void 1190ip_slowtimo(void) 1191{ 1192 VNET_ITERATOR_DECL(vnet_iter); 1193 struct ipq *fp; 1194 int i; 1195 1196 IPQ_LOCK(); 1197 VNET_LIST_RLOCK(); 1198 VNET_FOREACH(vnet_iter) { 1199 CURVNET_SET(vnet_iter); 1200 for (i = 0; i < IPREASS_NHASH; i++) { 1201 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) { 1202 struct ipq *fpp; 1203 1204 fpp = fp; 1205 fp = TAILQ_NEXT(fp, ipq_list); 1206 if(--fpp->ipq_ttl == 0) { 1207 IPSTAT_ADD(ips_fragtimeout, 1208 fpp->ipq_nfrags); 1209 ip_freef(&V_ipq[i], fpp); 1210 } 1211 } 1212 } 1213 /* 1214 * If we are over the maximum number of fragments 1215 * (due to the limit being lowered), drain off 1216 * enough to get down to the new limit. 1217 */ 1218 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) { 1219 for (i = 0; i < IPREASS_NHASH; i++) { 1220 while (V_nipq > V_maxnipq && 1221 !TAILQ_EMPTY(&V_ipq[i])) { 1222 IPSTAT_ADD(ips_fragdropped, 1223 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1224 ip_freef(&V_ipq[i], 1225 TAILQ_FIRST(&V_ipq[i])); 1226 } 1227 } 1228 } 1229 CURVNET_RESTORE(); 1230 } 1231 VNET_LIST_RUNLOCK(); 1232 IPQ_UNLOCK(); 1233} 1234 1235/* 1236 * Drain off all datagram fragments. 1237 */ 1238void 1239ip_drain(void) 1240{ 1241 VNET_ITERATOR_DECL(vnet_iter); 1242 int i; 1243 1244 IPQ_LOCK(); 1245 VNET_LIST_RLOCK(); 1246 VNET_FOREACH(vnet_iter) { 1247 CURVNET_SET(vnet_iter); 1248 for (i = 0; i < IPREASS_NHASH; i++) { 1249 while(!TAILQ_EMPTY(&V_ipq[i])) { 1250 IPSTAT_ADD(ips_fragdropped, 1251 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1252 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); 1253 } 1254 } 1255 CURVNET_RESTORE(); 1256 } 1257 VNET_LIST_RUNLOCK(); 1258 IPQ_UNLOCK(); 1259 in_rtqdrain(); 1260} 1261 1262/* 1263 * The protocol to be inserted into ip_protox[] must be already registered 1264 * in inetsw[], either statically or through pf_proto_register(). 1265 */ 1266int 1267ipproto_register(u_char ipproto) 1268{ 1269 struct protosw *pr; 1270 1271 /* Sanity checks. */ 1272 if (ipproto == 0) 1273 return (EPROTONOSUPPORT); 1274 1275 /* 1276 * The protocol slot must not be occupied by another protocol 1277 * already. An index pointing to IPPROTO_RAW is unused. 1278 */ 1279 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1280 if (pr == NULL) 1281 return (EPFNOSUPPORT); 1282 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1283 return (EEXIST); 1284 1285 /* Find the protocol position in inetsw[] and set the index. */ 1286 for (pr = inetdomain.dom_protosw; 1287 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1288 if (pr->pr_domain->dom_family == PF_INET && 1289 pr->pr_protocol && pr->pr_protocol == ipproto) { 1290 /* Be careful to only index valid IP protocols. */ 1291 if (pr->pr_protocol < IPPROTO_MAX) { 1292 ip_protox[pr->pr_protocol] = pr - inetsw; 1293 return (0); 1294 } else 1295 return (EINVAL); 1296 } 1297 } 1298 return (EPROTONOSUPPORT); 1299} 1300 1301int 1302ipproto_unregister(u_char ipproto) 1303{ 1304 struct protosw *pr; 1305 1306 /* Sanity checks. */ 1307 if (ipproto == 0) 1308 return (EPROTONOSUPPORT); 1309 1310 /* Check if the protocol was indeed registered. */ 1311 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1312 if (pr == NULL) 1313 return (EPFNOSUPPORT); 1314 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1315 return (ENOENT); 1316 1317 /* Reset the protocol slot to IPPROTO_RAW. */ 1318 ip_protox[ipproto] = pr - inetsw; 1319 return (0); 1320} 1321 1322/* 1323 * Given address of next destination (final or next hop), return (referenced) 1324 * internet address info of interface to be used to get there. 1325 */ 1326struct in_ifaddr * 1327ip_rtaddr(struct in_addr dst, u_int fibnum) 1328{ 1329 struct route sro; 1330 struct sockaddr_in *sin; 1331 struct in_ifaddr *ia; 1332 1333 bzero(&sro, sizeof(sro)); 1334 sin = (struct sockaddr_in *)&sro.ro_dst; 1335 sin->sin_family = AF_INET; 1336 sin->sin_len = sizeof(*sin); 1337 sin->sin_addr = dst; 1338 in_rtalloc_ign(&sro, 0, fibnum); 1339 1340 if (sro.ro_rt == NULL) 1341 return (NULL); 1342 1343 ia = ifatoia(sro.ro_rt->rt_ifa); 1344 ifa_ref(&ia->ia_ifa); 1345 RTFREE(sro.ro_rt); 1346 return (ia); 1347} 1348 1349u_char inetctlerrmap[PRC_NCMDS] = { 1350 0, 0, 0, 0, 1351 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1352 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1353 EMSGSIZE, EHOSTUNREACH, 0, 0, 1354 0, 0, EHOSTUNREACH, 0, 1355 ENOPROTOOPT, ECONNREFUSED 1356}; 1357 1358/* 1359 * Forward a packet. If some error occurs return the sender 1360 * an icmp packet. Note we can't always generate a meaningful 1361 * icmp message because icmp doesn't have a large enough repertoire 1362 * of codes and types. 1363 * 1364 * If not forwarding, just drop the packet. This could be confusing 1365 * if ipforwarding was zero but some routing protocol was advancing 1366 * us as a gateway to somewhere. However, we must let the routing 1367 * protocol deal with that. 1368 * 1369 * The srcrt parameter indicates whether the packet is being forwarded 1370 * via a source route. 1371 */ 1372void 1373ip_forward(struct mbuf *m, int srcrt) 1374{ 1375 struct ip *ip = mtod(m, struct ip *); 1376 struct in_ifaddr *ia; 1377 struct mbuf *mcopy; 1378 struct in_addr dest; 1379 struct route ro; 1380 int error, type = 0, code = 0, mtu = 0; 1381 1382 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1383 IPSTAT_INC(ips_cantforward); 1384 m_freem(m); 1385 return; 1386 } 1387#ifdef IPSTEALTH 1388 if (!V_ipstealth) { 1389#endif 1390 if (ip->ip_ttl <= IPTTLDEC) { 1391 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1392 0, 0); 1393 return; 1394 } 1395#ifdef IPSTEALTH 1396 } 1397#endif 1398 1399 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); 1400#ifndef IPSEC 1401 /* 1402 * 'ia' may be NULL if there is no route for this destination. 1403 * In case of IPsec, Don't discard it just yet, but pass it to 1404 * ip_output in case of outgoing IPsec policy. 1405 */ 1406 if (!srcrt && ia == NULL) { 1407 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1408 return; 1409 } 1410#endif 1411 1412 /* 1413 * Save the IP header and at most 8 bytes of the payload, 1414 * in case we need to generate an ICMP message to the src. 1415 * 1416 * XXX this can be optimized a lot by saving the data in a local 1417 * buffer on the stack (72 bytes at most), and only allocating the 1418 * mbuf if really necessary. The vast majority of the packets 1419 * are forwarded without having to send an ICMP back (either 1420 * because unnecessary, or because rate limited), so we are 1421 * really we are wasting a lot of work here. 1422 * 1423 * We don't use m_copy() because it might return a reference 1424 * to a shared cluster. Both this function and ip_output() 1425 * assume exclusive access to the IP header in `m', so any 1426 * data in a cluster may change before we reach icmp_error(). 1427 */ 1428 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1429 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1430 /* 1431 * It's probably ok if the pkthdr dup fails (because 1432 * the deep copy of the tag chain failed), but for now 1433 * be conservative and just discard the copy since 1434 * code below may some day want the tags. 1435 */ 1436 m_free(mcopy); 1437 mcopy = NULL; 1438 } 1439 if (mcopy != NULL) { 1440 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy)); 1441 mcopy->m_pkthdr.len = mcopy->m_len; 1442 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1443 } 1444 1445#ifdef IPSTEALTH 1446 if (!V_ipstealth) { 1447#endif 1448 ip->ip_ttl -= IPTTLDEC; 1449#ifdef IPSTEALTH 1450 } 1451#endif 1452 1453 /* 1454 * If forwarding packet using same interface that it came in on, 1455 * perhaps should send a redirect to sender to shortcut a hop. 1456 * Only send redirect if source is sending directly to us, 1457 * and if packet was not source routed (or has any options). 1458 * Also, don't send redirect if forwarding using a default route 1459 * or a route modified by a redirect. 1460 */ 1461 dest.s_addr = 0; 1462 if (!srcrt && V_ipsendredirects && 1463 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) { 1464 struct sockaddr_in *sin; 1465 struct rtentry *rt; 1466 1467 bzero(&ro, sizeof(ro)); 1468 sin = (struct sockaddr_in *)&ro.ro_dst; 1469 sin->sin_family = AF_INET; 1470 sin->sin_len = sizeof(*sin); 1471 sin->sin_addr = ip->ip_dst; 1472 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 1473 1474 rt = ro.ro_rt; 1475 1476 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1477 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1478#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1479 u_long src = ntohl(ip->ip_src.s_addr); 1480 1481 if (RTA(rt) && 1482 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1483 if (rt->rt_flags & RTF_GATEWAY) 1484 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1485 else 1486 dest.s_addr = ip->ip_dst.s_addr; 1487 /* Router requirements says to only send host redirects */ 1488 type = ICMP_REDIRECT; 1489 code = ICMP_REDIRECT_HOST; 1490 } 1491 } 1492 if (rt) 1493 RTFREE(rt); 1494 } 1495 1496 /* 1497 * Try to cache the route MTU from ip_output so we can consider it for 1498 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. 1499 */ 1500 bzero(&ro, sizeof(ro)); 1501 1502 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1503 1504 if (error == EMSGSIZE && ro.ro_rt) 1505 mtu = ro.ro_rt->rt_rmx.rmx_mtu; 1506 if (ro.ro_rt) 1507 RTFREE(ro.ro_rt); 1508 1509 if (error) 1510 IPSTAT_INC(ips_cantforward); 1511 else { 1512 IPSTAT_INC(ips_forward); 1513 if (type) 1514 IPSTAT_INC(ips_redirectsent); 1515 else { 1516 if (mcopy) 1517 m_freem(mcopy); 1518 if (ia != NULL) 1519 ifa_free(&ia->ia_ifa); 1520 return; 1521 } 1522 } 1523 if (mcopy == NULL) { 1524 if (ia != NULL) 1525 ifa_free(&ia->ia_ifa); 1526 return; 1527 } 1528 1529 switch (error) { 1530 1531 case 0: /* forwarded, but need redirect */ 1532 /* type, code set above */ 1533 break; 1534 1535 case ENETUNREACH: 1536 case EHOSTUNREACH: 1537 case ENETDOWN: 1538 case EHOSTDOWN: 1539 default: 1540 type = ICMP_UNREACH; 1541 code = ICMP_UNREACH_HOST; 1542 break; 1543 1544 case EMSGSIZE: 1545 type = ICMP_UNREACH; 1546 code = ICMP_UNREACH_NEEDFRAG; 1547 1548#ifdef IPSEC 1549 /* 1550 * If IPsec is configured for this path, 1551 * override any possibly mtu value set by ip_output. 1552 */ 1553 mtu = ip_ipsec_mtu(m, mtu); 1554#endif /* IPSEC */ 1555 /* 1556 * If the MTU was set before make sure we are below the 1557 * interface MTU. 1558 * If the MTU wasn't set before use the interface mtu or 1559 * fall back to the next smaller mtu step compared to the 1560 * current packet size. 1561 */ 1562 if (mtu != 0) { 1563 if (ia != NULL) 1564 mtu = min(mtu, ia->ia_ifp->if_mtu); 1565 } else { 1566 if (ia != NULL) 1567 mtu = ia->ia_ifp->if_mtu; 1568 else 1569 mtu = ip_next_mtu(ip->ip_len, 0); 1570 } 1571 IPSTAT_INC(ips_cantfrag); 1572 break; 1573 1574 case ENOBUFS: 1575 /* 1576 * A router should not generate ICMP_SOURCEQUENCH as 1577 * required in RFC1812 Requirements for IP Version 4 Routers. 1578 * Source quench could be a big problem under DoS attacks, 1579 * or if the underlying interface is rate-limited. 1580 * Those who need source quench packets may re-enable them 1581 * via the net.inet.ip.sendsourcequench sysctl. 1582 */ 1583 if (V_ip_sendsourcequench == 0) { 1584 m_freem(mcopy); 1585 if (ia != NULL) 1586 ifa_free(&ia->ia_ifa); 1587 return; 1588 } else { 1589 type = ICMP_SOURCEQUENCH; 1590 code = 0; 1591 } 1592 break; 1593 1594 case EACCES: /* ipfw denied packet */ 1595 m_freem(mcopy); 1596 if (ia != NULL) 1597 ifa_free(&ia->ia_ifa); 1598 return; 1599 } 1600 if (ia != NULL) 1601 ifa_free(&ia->ia_ifa); 1602 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1603} 1604 1605void 1606ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1607 struct mbuf *m) 1608{ 1609 1610 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1611 struct bintime bt; 1612 1613 bintime(&bt); 1614 if (inp->inp_socket->so_options & SO_BINTIME) { 1615 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1616 SCM_BINTIME, SOL_SOCKET); 1617 if (*mp) 1618 mp = &(*mp)->m_next; 1619 } 1620 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1621 struct timeval tv; 1622 1623 bintime2timeval(&bt, &tv); 1624 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1625 SCM_TIMESTAMP, SOL_SOCKET); 1626 if (*mp) 1627 mp = &(*mp)->m_next; 1628 } 1629 } 1630 if (inp->inp_flags & INP_RECVDSTADDR) { 1631 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1632 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1633 if (*mp) 1634 mp = &(*mp)->m_next; 1635 } 1636 if (inp->inp_flags & INP_RECVTTL) { 1637 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1638 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1639 if (*mp) 1640 mp = &(*mp)->m_next; 1641 } 1642#ifdef notyet 1643 /* XXX 1644 * Moving these out of udp_input() made them even more broken 1645 * than they already were. 1646 */ 1647 /* options were tossed already */ 1648 if (inp->inp_flags & INP_RECVOPTS) { 1649 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1650 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1651 if (*mp) 1652 mp = &(*mp)->m_next; 1653 } 1654 /* ip_srcroute doesn't do what we want here, need to fix */ 1655 if (inp->inp_flags & INP_RECVRETOPTS) { 1656 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1657 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1658 if (*mp) 1659 mp = &(*mp)->m_next; 1660 } 1661#endif 1662 if (inp->inp_flags & INP_RECVIF) { 1663 struct ifnet *ifp; 1664 struct sdlbuf { 1665 struct sockaddr_dl sdl; 1666 u_char pad[32]; 1667 } sdlbuf; 1668 struct sockaddr_dl *sdp; 1669 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1670 1671 if (((ifp = m->m_pkthdr.rcvif)) 1672 && ( ifp->if_index && (ifp->if_index <= V_if_index))) { 1673 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1674 /* 1675 * Change our mind and don't try copy. 1676 */ 1677 if ((sdp->sdl_family != AF_LINK) 1678 || (sdp->sdl_len > sizeof(sdlbuf))) { 1679 goto makedummy; 1680 } 1681 bcopy(sdp, sdl2, sdp->sdl_len); 1682 } else { 1683makedummy: 1684 sdl2->sdl_len 1685 = offsetof(struct sockaddr_dl, sdl_data[0]); 1686 sdl2->sdl_family = AF_LINK; 1687 sdl2->sdl_index = 0; 1688 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1689 } 1690 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1691 IP_RECVIF, IPPROTO_IP); 1692 if (*mp) 1693 mp = &(*mp)->m_next; 1694 } 1695} 1696 1697/* 1698 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1699 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1700 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1701 * compiled. 1702 */ 1703int 1704ip_rsvp_init(struct socket *so) 1705{ 1706 1707 if (so->so_type != SOCK_RAW || 1708 so->so_proto->pr_protocol != IPPROTO_RSVP) 1709 return EOPNOTSUPP; 1710 1711 if (V_ip_rsvpd != NULL) 1712 return EADDRINUSE; 1713 1714 V_ip_rsvpd = so; 1715 /* 1716 * This may seem silly, but we need to be sure we don't over-increment 1717 * the RSVP counter, in case something slips up. 1718 */ 1719 if (!V_ip_rsvp_on) { 1720 V_ip_rsvp_on = 1; 1721 V_rsvp_on++; 1722 } 1723 1724 return 0; 1725} 1726 1727int 1728ip_rsvp_done(void) 1729{ 1730 1731 V_ip_rsvpd = NULL; 1732 /* 1733 * This may seem silly, but we need to be sure we don't over-decrement 1734 * the RSVP counter, in case something slips up. 1735 */ 1736 if (V_ip_rsvp_on) { 1737 V_ip_rsvp_on = 0; 1738 V_rsvp_on--; 1739 } 1740 return 0; 1741} 1742 1743void 1744rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1745{ 1746 1747 if (rsvp_input_p) { /* call the real one if loaded */ 1748 rsvp_input_p(m, off); 1749 return; 1750 } 1751 1752 /* Can still get packets with rsvp_on = 0 if there is a local member 1753 * of the group to which the RSVP packet is addressed. But in this 1754 * case we want to throw the packet away. 1755 */ 1756 1757 if (!V_rsvp_on) { 1758 m_freem(m); 1759 return; 1760 } 1761 1762 if (V_ip_rsvpd != NULL) { 1763 rip_input(m, off); 1764 return; 1765 } 1766 /* Drop the packet */ 1767 m_freem(m); 1768} 1769