ip_input.c revision 189106
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 189106 2009-02-27 14:12:05Z bz $"); 34 35#include "opt_bootp.h" 36#include "opt_ipfw.h" 37#include "opt_ipstealth.h" 38#include "opt_ipsec.h" 39#include "opt_route.h" 40#include "opt_mac.h" 41#include "opt_carp.h" 42 43#include <sys/param.h> 44#include <sys/systm.h> 45#include <sys/callout.h> 46#include <sys/mbuf.h> 47#include <sys/malloc.h> 48#include <sys/domain.h> 49#include <sys/protosw.h> 50#include <sys/socket.h> 51#include <sys/time.h> 52#include <sys/kernel.h> 53#include <sys/lock.h> 54#include <sys/rwlock.h> 55#include <sys/syslog.h> 56#include <sys/sysctl.h> 57#include <sys/vimage.h> 58 59#include <net/pfil.h> 60#include <net/if.h> 61#include <net/if_types.h> 62#include <net/if_var.h> 63#include <net/if_dl.h> 64#include <net/route.h> 65#include <net/netisr.h> 66#include <net/vnet.h> 67 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/* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */ 88#include <netinet/ip_fw.h> 89#include <netinet/ip_dummynet.h> 90 91#include <security/mac/mac_framework.h> 92 93#ifdef CTASSERT 94CTASSERT(sizeof(struct ip) == 20); 95#endif 96 97#ifndef VIMAGE 98#ifndef VIMAGE_GLOBALS 99struct vnet_inet vnet_inet_0; 100#endif 101#endif 102 103#ifdef VIMAGE_GLOBALS 104static int ipsendredirects; 105static int ip_checkinterface; 106static int ip_keepfaith; 107static int ip_sendsourcequench; 108int ip_defttl; 109int ip_do_randomid; 110int ipforwarding; 111struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 112struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 113u_long in_ifaddrhmask; /* mask for hash table */ 114struct ipstat ipstat; 115static int ip_rsvp_on; 116struct socket *ip_rsvpd; 117int rsvp_on; 118static struct ipqhead ipq[IPREASS_NHASH]; 119static int maxnipq; /* Administrative limit on # reass queues. */ 120static int maxfragsperpacket; 121int ipstealth; 122static int nipq; /* Total # of reass queues */ 123#endif 124 125SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_FORWARDING, 126 forwarding, CTLFLAG_RW, ipforwarding, 0, 127 "Enable IP forwarding between interfaces"); 128 129SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_SENDREDIRECTS, 130 redirect, CTLFLAG_RW, ipsendredirects, 0, 131 "Enable sending IP redirects"); 132 133SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_DEFTTL, 134 ttl, CTLFLAG_RW, ip_defttl, 0, "Maximum TTL on IP packets"); 135 136SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_KEEPFAITH, 137 keepfaith, CTLFLAG_RW, ip_keepfaith, 0, 138 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 139 140SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, 141 sendsourcequench, CTLFLAG_RW, ip_sendsourcequench, 0, 142 "Enable the transmission of source quench packets"); 143 144SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, random_id, 145 CTLFLAG_RW, ip_do_randomid, 0, "Assign random ip_id values"); 146 147/* 148 * XXX - Setting ip_checkinterface mostly implements the receive side of 149 * the Strong ES model described in RFC 1122, but since the routing table 150 * and transmit implementation do not implement the Strong ES model, 151 * setting this to 1 results in an odd hybrid. 152 * 153 * XXX - ip_checkinterface currently must be disabled if you use ipnat 154 * to translate the destination address to another local interface. 155 * 156 * XXX - ip_checkinterface must be disabled if you add IP aliases 157 * to the loopback interface instead of the interface where the 158 * packets for those addresses are received. 159 */ 160SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, 161 check_interface, CTLFLAG_RW, ip_checkinterface, 0, 162 "Verify packet arrives on correct interface"); 163 164struct pfil_head inet_pfil_hook; /* Packet filter hooks */ 165 166static struct ifqueue ipintrq; 167static int ipqmaxlen = IFQ_MAXLEN; 168 169extern struct domain inetdomain; 170extern struct protosw inetsw[]; 171u_char ip_protox[IPPROTO_MAX]; 172 173SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 174 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 175SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 176 &ipintrq.ifq_drops, 0, 177 "Number of packets dropped from the IP input queue"); 178 179SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 180 ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 181 182#ifdef VIMAGE_GLOBALS 183static uma_zone_t ipq_zone; 184#endif 185static struct mtx ipqlock; 186 187#define IPQ_LOCK() mtx_lock(&ipqlock) 188#define IPQ_UNLOCK() mtx_unlock(&ipqlock) 189#define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 190#define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 191 192static void maxnipq_update(void); 193static void ipq_zone_change(void *); 194 195SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, fragpackets, 196 CTLFLAG_RD, nipq, 0, 197 "Current number of IPv4 fragment reassembly queue entries"); 198 199SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, maxfragsperpacket, 200 CTLFLAG_RW, maxfragsperpacket, 0, 201 "Maximum number of IPv4 fragments allowed per packet"); 202 203struct callout ipport_tick_callout; 204 205#ifdef IPCTL_DEFMTU 206SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 207 &ip_mtu, 0, "Default MTU"); 208#endif 209 210#ifdef IPSTEALTH 211SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 212 ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding"); 213#endif 214 215/* 216 * ipfw_ether and ipfw_bridge hooks. 217 * XXX: Temporary until those are converted to pfil_hooks as well. 218 */ 219ip_fw_chk_t *ip_fw_chk_ptr = NULL; 220ip_dn_io_t *ip_dn_io_ptr = NULL; 221#ifdef VIMAGE_GLOBALS 222int fw_one_pass; 223#endif 224 225static void ip_freef(struct ipqhead *, struct ipq *); 226 227/* 228 * IP initialization: fill in IP protocol switch table. 229 * All protocols not implemented in kernel go to raw IP protocol handler. 230 */ 231void 232ip_init(void) 233{ 234 INIT_VNET_INET(curvnet); 235 struct protosw *pr; 236 int i; 237 238 V_ipsendredirects = 1; /* XXX */ 239 V_ip_checkinterface = 0; 240 V_ip_keepfaith = 0; 241 V_ip_sendsourcequench = 0; 242 V_rsvp_on = 0; 243 V_ip_defttl = IPDEFTTL; 244 V_ip_do_randomid = 0; 245 V_ipforwarding = 0; 246 V_ipstealth = 0; 247 V_nipq = 0; /* Total # of reass queues */ 248 249 V_ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */ 250 V_ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */ 251 V_ipport_firstauto = IPPORT_EPHEMERALFIRST; /* 10000 */ 252 V_ipport_lastauto = IPPORT_EPHEMERALLAST; /* 65535 */ 253 V_ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */ 254 V_ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */ 255 V_ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */ 256 V_ipport_reservedlow = 0; 257 V_ipport_randomized = 1; /* user controlled via sysctl */ 258 V_ipport_randomcps = 10; /* user controlled via sysctl */ 259 V_ipport_randomtime = 45; /* user controlled via sysctl */ 260 V_ipport_stoprandom = 0; /* toggled by ipport_tick */ 261 262 V_fw_one_pass = 1; 263 264#ifdef NOTYET 265 /* XXX global static but not instantiated in this file */ 266 V_ipfastforward_active = 0; 267 V_subnetsarelocal = 0; 268 V_sameprefixcarponly = 0; 269#endif 270 271 TAILQ_INIT(&V_in_ifaddrhead); 272 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 273 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 274 if (pr == NULL) 275 panic("ip_init: PF_INET not found"); 276 277 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 278 for (i = 0; i < IPPROTO_MAX; i++) 279 ip_protox[i] = pr - inetsw; 280 /* 281 * Cycle through IP protocols and put them into the appropriate place 282 * in ip_protox[]. 283 */ 284 for (pr = inetdomain.dom_protosw; 285 pr < inetdomain.dom_protoswNPROTOSW; pr++) 286 if (pr->pr_domain->dom_family == PF_INET && 287 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 288 /* Be careful to only index valid IP protocols. */ 289 if (pr->pr_protocol < IPPROTO_MAX) 290 ip_protox[pr->pr_protocol] = pr - inetsw; 291 } 292 293 /* Initialize packet filter hooks. */ 294 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 295 inet_pfil_hook.ph_af = AF_INET; 296 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 297 printf("%s: WARNING: unable to register pfil hook, " 298 "error %d\n", __func__, i); 299 300 /* Initialize IP reassembly queue. */ 301 IPQ_LOCK_INIT(); 302 for (i = 0; i < IPREASS_NHASH; i++) 303 TAILQ_INIT(&V_ipq[i]); 304 V_maxnipq = nmbclusters / 32; 305 V_maxfragsperpacket = 16; 306 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 307 NULL, UMA_ALIGN_PTR, 0); 308 maxnipq_update(); 309 310 /* Start ipport_tick. */ 311 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE); 312 ipport_tick(NULL); 313 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, 314 SHUTDOWN_PRI_DEFAULT); 315 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, 316 NULL, EVENTHANDLER_PRI_ANY); 317 318 /* Initialize various other remaining things. */ 319 V_ip_id = time_second & 0xffff; 320 ipintrq.ifq_maxlen = ipqmaxlen; 321 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF); 322 netisr_register(NETISR_IP, ip_input, &ipintrq, 0); 323} 324 325void 326ip_fini(void *xtp) 327{ 328 329 callout_stop(&ipport_tick_callout); 330} 331 332/* 333 * Ip input routine. Checksum and byte swap header. If fragmented 334 * try to reassemble. Process options. Pass to next level. 335 */ 336void 337ip_input(struct mbuf *m) 338{ 339 INIT_VNET_INET(curvnet); 340 struct ip *ip = NULL; 341 struct in_ifaddr *ia = NULL; 342 struct ifaddr *ifa; 343 int checkif, hlen = 0; 344 u_short sum; 345 int dchg = 0; /* dest changed after fw */ 346 struct in_addr odst; /* original dst address */ 347 348 M_ASSERTPKTHDR(m); 349 350 if (m->m_flags & M_FASTFWD_OURS) { 351 /* 352 * Firewall or NAT changed destination to local. 353 * We expect ip_len and ip_off to be in host byte order. 354 */ 355 m->m_flags &= ~M_FASTFWD_OURS; 356 /* Set up some basics that will be used later. */ 357 ip = mtod(m, struct ip *); 358 hlen = ip->ip_hl << 2; 359 goto ours; 360 } 361 362 V_ipstat.ips_total++; 363 364 if (m->m_pkthdr.len < sizeof(struct ip)) 365 goto tooshort; 366 367 if (m->m_len < sizeof (struct ip) && 368 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 369 V_ipstat.ips_toosmall++; 370 return; 371 } 372 ip = mtod(m, struct ip *); 373 374 if (ip->ip_v != IPVERSION) { 375 V_ipstat.ips_badvers++; 376 goto bad; 377 } 378 379 hlen = ip->ip_hl << 2; 380 if (hlen < sizeof(struct ip)) { /* minimum header length */ 381 V_ipstat.ips_badhlen++; 382 goto bad; 383 } 384 if (hlen > m->m_len) { 385 if ((m = m_pullup(m, hlen)) == NULL) { 386 V_ipstat.ips_badhlen++; 387 return; 388 } 389 ip = mtod(m, struct ip *); 390 } 391 392 /* 127/8 must not appear on wire - RFC1122 */ 393 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 394 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 395 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 396 V_ipstat.ips_badaddr++; 397 goto bad; 398 } 399 } 400 401 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 402 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 403 } else { 404 if (hlen == sizeof(struct ip)) { 405 sum = in_cksum_hdr(ip); 406 } else { 407 sum = in_cksum(m, hlen); 408 } 409 } 410 if (sum) { 411 V_ipstat.ips_badsum++; 412 goto bad; 413 } 414 415#ifdef ALTQ 416 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 417 /* packet is dropped by traffic conditioner */ 418 return; 419#endif 420 421 /* 422 * Convert fields to host representation. 423 */ 424 ip->ip_len = ntohs(ip->ip_len); 425 if (ip->ip_len < hlen) { 426 V_ipstat.ips_badlen++; 427 goto bad; 428 } 429 ip->ip_off = ntohs(ip->ip_off); 430 431 /* 432 * Check that the amount of data in the buffers 433 * is as at least much as the IP header would have us expect. 434 * Trim mbufs if longer than we expect. 435 * Drop packet if shorter than we expect. 436 */ 437 if (m->m_pkthdr.len < ip->ip_len) { 438tooshort: 439 V_ipstat.ips_tooshort++; 440 goto bad; 441 } 442 if (m->m_pkthdr.len > ip->ip_len) { 443 if (m->m_len == m->m_pkthdr.len) { 444 m->m_len = ip->ip_len; 445 m->m_pkthdr.len = ip->ip_len; 446 } else 447 m_adj(m, ip->ip_len - m->m_pkthdr.len); 448 } 449#ifdef IPSEC 450 /* 451 * Bypass packet filtering for packets from a tunnel (gif). 452 */ 453 if (ip_ipsec_filtertunnel(m)) 454 goto passin; 455#endif /* IPSEC */ 456 457 /* 458 * Run through list of hooks for input packets. 459 * 460 * NB: Beware of the destination address changing (e.g. 461 * by NAT rewriting). When this happens, tell 462 * ip_forward to do the right thing. 463 */ 464 465 /* Jump over all PFIL processing if hooks are not active. */ 466 if (!PFIL_HOOKED(&inet_pfil_hook)) 467 goto passin; 468 469 odst = ip->ip_dst; 470 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 471 PFIL_IN, NULL) != 0) 472 return; 473 if (m == NULL) /* consumed by filter */ 474 return; 475 476 ip = mtod(m, struct ip *); 477 dchg = (odst.s_addr != ip->ip_dst.s_addr); 478 479#ifdef IPFIREWALL_FORWARD 480 if (m->m_flags & M_FASTFWD_OURS) { 481 m->m_flags &= ~M_FASTFWD_OURS; 482 goto ours; 483 } 484 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 485 /* 486 * Directly ship on the packet. This allows to forward packets 487 * that were destined for us to some other directly connected 488 * host. 489 */ 490 ip_forward(m, dchg); 491 return; 492 } 493#endif /* IPFIREWALL_FORWARD */ 494 495passin: 496 /* 497 * Process options and, if not destined for us, 498 * ship it on. ip_dooptions returns 1 when an 499 * error was detected (causing an icmp message 500 * to be sent and the original packet to be freed). 501 */ 502 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 503 return; 504 505 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 506 * matter if it is destined to another node, or whether it is 507 * a multicast one, RSVP wants it! and prevents it from being forwarded 508 * anywhere else. Also checks if the rsvp daemon is running before 509 * grabbing the packet. 510 */ 511 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 512 goto ours; 513 514 /* 515 * Check our list of addresses, to see if the packet is for us. 516 * If we don't have any addresses, assume any unicast packet 517 * we receive might be for us (and let the upper layers deal 518 * with it). 519 */ 520 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 521 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 522 goto ours; 523 524 /* 525 * Enable a consistency check between the destination address 526 * and the arrival interface for a unicast packet (the RFC 1122 527 * strong ES model) if IP forwarding is disabled and the packet 528 * is not locally generated and the packet is not subject to 529 * 'ipfw fwd'. 530 * 531 * XXX - Checking also should be disabled if the destination 532 * address is ipnat'ed to a different interface. 533 * 534 * XXX - Checking is incompatible with IP aliases added 535 * to the loopback interface instead of the interface where 536 * the packets are received. 537 * 538 * XXX - This is the case for carp vhost IPs as well so we 539 * insert a workaround. If the packet got here, we already 540 * checked with carp_iamatch() and carp_forus(). 541 */ 542 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 543 m->m_pkthdr.rcvif != NULL && 544 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 545#ifdef DEV_CARP 546 !m->m_pkthdr.rcvif->if_carp && 547#endif 548 (dchg == 0); 549 550 /* 551 * Check for exact addresses in the hash bucket. 552 */ 553 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 554 /* 555 * If the address matches, verify that the packet 556 * arrived via the correct interface if checking is 557 * enabled. 558 */ 559 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 560 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 561 goto ours; 562 } 563 /* 564 * Check for broadcast addresses. 565 * 566 * Only accept broadcast packets that arrive via the matching 567 * interface. Reception of forwarded directed broadcasts would 568 * be handled via ip_forward() and ether_output() with the loopback 569 * into the stack for SIMPLEX interfaces handled by ether_output(). 570 */ 571 if (m->m_pkthdr.rcvif != NULL && 572 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 573 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 574 if (ifa->ifa_addr->sa_family != AF_INET) 575 continue; 576 ia = ifatoia(ifa); 577 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 578 ip->ip_dst.s_addr) 579 goto ours; 580 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) 581 goto ours; 582#ifdef BOOTP_COMPAT 583 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 584 goto ours; 585#endif 586 } 587 } 588 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 589 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 590 V_ipstat.ips_cantforward++; 591 m_freem(m); 592 return; 593 } 594 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 595 struct in_multi *inm; 596 if (V_ip_mrouter) { 597 /* 598 * If we are acting as a multicast router, all 599 * incoming multicast packets are passed to the 600 * kernel-level multicast forwarding function. 601 * The packet is returned (relatively) intact; if 602 * ip_mforward() returns a non-zero value, the packet 603 * must be discarded, else it may be accepted below. 604 */ 605 if (ip_mforward && 606 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 607 V_ipstat.ips_cantforward++; 608 m_freem(m); 609 return; 610 } 611 612 /* 613 * The process-level routing daemon needs to receive 614 * all multicast IGMP packets, whether or not this 615 * host belongs to their destination groups. 616 */ 617 if (ip->ip_p == IPPROTO_IGMP) 618 goto ours; 619 V_ipstat.ips_forward++; 620 } 621 /* 622 * See if we belong to the destination multicast group on the 623 * arrival interface. 624 */ 625 IN_MULTI_LOCK(); 626 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 627 IN_MULTI_UNLOCK(); 628 if (inm == NULL) { 629 V_ipstat.ips_notmember++; 630 m_freem(m); 631 return; 632 } 633 goto ours; 634 } 635 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 636 goto ours; 637 if (ip->ip_dst.s_addr == INADDR_ANY) 638 goto ours; 639 640 /* 641 * FAITH(Firewall Aided Internet Translator) 642 */ 643 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 644 if (V_ip_keepfaith) { 645 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 646 goto ours; 647 } 648 m_freem(m); 649 return; 650 } 651 652 /* 653 * Not for us; forward if possible and desirable. 654 */ 655 if (V_ipforwarding == 0) { 656 V_ipstat.ips_cantforward++; 657 m_freem(m); 658 } else { 659#ifdef IPSEC 660 if (ip_ipsec_fwd(m)) 661 goto bad; 662#endif /* IPSEC */ 663 ip_forward(m, dchg); 664 } 665 return; 666 667ours: 668#ifdef IPSTEALTH 669 /* 670 * IPSTEALTH: Process non-routing options only 671 * if the packet is destined for us. 672 */ 673 if (V_ipstealth && hlen > sizeof (struct ip) && 674 ip_dooptions(m, 1)) 675 return; 676#endif /* IPSTEALTH */ 677 678 /* Count the packet in the ip address stats */ 679 if (ia != NULL) { 680 ia->ia_ifa.if_ipackets++; 681 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 682 } 683 684 /* 685 * Attempt reassembly; if it succeeds, proceed. 686 * ip_reass() will return a different mbuf. 687 */ 688 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 689 m = ip_reass(m); 690 if (m == NULL) 691 return; 692 ip = mtod(m, struct ip *); 693 /* Get the header length of the reassembled packet */ 694 hlen = ip->ip_hl << 2; 695 } 696 697 /* 698 * Further protocols expect the packet length to be w/o the 699 * IP header. 700 */ 701 ip->ip_len -= hlen; 702 703#ifdef IPSEC 704 /* 705 * enforce IPsec policy checking if we are seeing last header. 706 * note that we do not visit this with protocols with pcb layer 707 * code - like udp/tcp/raw ip. 708 */ 709 if (ip_ipsec_input(m)) 710 goto bad; 711#endif /* IPSEC */ 712 713 /* 714 * Switch out to protocol's input routine. 715 */ 716 V_ipstat.ips_delivered++; 717 718 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 719 return; 720bad: 721 m_freem(m); 722} 723 724/* 725 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 726 * max has slightly different semantics than the sysctl, for historical 727 * reasons. 728 */ 729static void 730maxnipq_update(void) 731{ 732 INIT_VNET_INET(curvnet); 733 734 /* 735 * -1 for unlimited allocation. 736 */ 737 if (V_maxnipq < 0) 738 uma_zone_set_max(V_ipq_zone, 0); 739 /* 740 * Positive number for specific bound. 741 */ 742 if (V_maxnipq > 0) 743 uma_zone_set_max(V_ipq_zone, V_maxnipq); 744 /* 745 * Zero specifies no further fragment queue allocation -- set the 746 * bound very low, but rely on implementation elsewhere to actually 747 * prevent allocation and reclaim current queues. 748 */ 749 if (V_maxnipq == 0) 750 uma_zone_set_max(V_ipq_zone, 1); 751} 752 753static void 754ipq_zone_change(void *tag) 755{ 756 INIT_VNET_INET(curvnet); 757 758 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) { 759 V_maxnipq = nmbclusters / 32; 760 maxnipq_update(); 761 } 762} 763 764static int 765sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 766{ 767 INIT_VNET_INET(curvnet); 768 int error, i; 769 770 i = V_maxnipq; 771 error = sysctl_handle_int(oidp, &i, 0, req); 772 if (error || !req->newptr) 773 return (error); 774 775 /* 776 * XXXRW: Might be a good idea to sanity check the argument and place 777 * an extreme upper bound. 778 */ 779 if (i < -1) 780 return (EINVAL); 781 V_maxnipq = i; 782 maxnipq_update(); 783 return (0); 784} 785 786SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 787 NULL, 0, sysctl_maxnipq, "I", 788 "Maximum number of IPv4 fragment reassembly queue entries"); 789 790/* 791 * Take incoming datagram fragment and try to reassemble it into 792 * whole datagram. If the argument is the first fragment or one 793 * in between the function will return NULL and store the mbuf 794 * in the fragment chain. If the argument is the last fragment 795 * the packet will be reassembled and the pointer to the new 796 * mbuf returned for further processing. Only m_tags attached 797 * to the first packet/fragment are preserved. 798 * The IP header is *NOT* adjusted out of iplen. 799 */ 800struct mbuf * 801ip_reass(struct mbuf *m) 802{ 803 INIT_VNET_INET(curvnet); 804 struct ip *ip; 805 struct mbuf *p, *q, *nq, *t; 806 struct ipq *fp = NULL; 807 struct ipqhead *head; 808 int i, hlen, next; 809 u_int8_t ecn, ecn0; 810 u_short hash; 811 812 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 813 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) { 814 V_ipstat.ips_fragments++; 815 V_ipstat.ips_fragdropped++; 816 m_freem(m); 817 return (NULL); 818 } 819 820 ip = mtod(m, struct ip *); 821 hlen = ip->ip_hl << 2; 822 823 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 824 head = &V_ipq[hash]; 825 IPQ_LOCK(); 826 827 /* 828 * Look for queue of fragments 829 * of this datagram. 830 */ 831 TAILQ_FOREACH(fp, head, ipq_list) 832 if (ip->ip_id == fp->ipq_id && 833 ip->ip_src.s_addr == fp->ipq_src.s_addr && 834 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 835#ifdef MAC 836 mac_ipq_match(m, fp) && 837#endif 838 ip->ip_p == fp->ipq_p) 839 goto found; 840 841 fp = NULL; 842 843 /* 844 * Attempt to trim the number of allocated fragment queues if it 845 * exceeds the administrative limit. 846 */ 847 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) { 848 /* 849 * drop something from the tail of the current queue 850 * before proceeding further 851 */ 852 struct ipq *q = TAILQ_LAST(head, ipqhead); 853 if (q == NULL) { /* gak */ 854 for (i = 0; i < IPREASS_NHASH; i++) { 855 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead); 856 if (r) { 857 V_ipstat.ips_fragtimeout += 858 r->ipq_nfrags; 859 ip_freef(&V_ipq[i], r); 860 break; 861 } 862 } 863 } else { 864 V_ipstat.ips_fragtimeout += q->ipq_nfrags; 865 ip_freef(head, q); 866 } 867 } 868 869found: 870 /* 871 * Adjust ip_len to not reflect header, 872 * convert offset of this to bytes. 873 */ 874 ip->ip_len -= hlen; 875 if (ip->ip_off & IP_MF) { 876 /* 877 * Make sure that fragments have a data length 878 * that's a non-zero multiple of 8 bytes. 879 */ 880 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 881 V_ipstat.ips_toosmall++; /* XXX */ 882 goto dropfrag; 883 } 884 m->m_flags |= M_FRAG; 885 } else 886 m->m_flags &= ~M_FRAG; 887 ip->ip_off <<= 3; 888 889 890 /* 891 * Attempt reassembly; if it succeeds, proceed. 892 * ip_reass() will return a different mbuf. 893 */ 894 V_ipstat.ips_fragments++; 895 m->m_pkthdr.header = ip; 896 897 /* Previous ip_reass() started here. */ 898 /* 899 * Presence of header sizes in mbufs 900 * would confuse code below. 901 */ 902 m->m_data += hlen; 903 m->m_len -= hlen; 904 905 /* 906 * If first fragment to arrive, create a reassembly queue. 907 */ 908 if (fp == NULL) { 909 fp = uma_zalloc(V_ipq_zone, M_NOWAIT); 910 if (fp == NULL) 911 goto dropfrag; 912#ifdef MAC 913 if (mac_ipq_init(fp, M_NOWAIT) != 0) { 914 uma_zfree(V_ipq_zone, fp); 915 fp = NULL; 916 goto dropfrag; 917 } 918 mac_ipq_create(m, fp); 919#endif 920 TAILQ_INSERT_HEAD(head, fp, ipq_list); 921 V_nipq++; 922 fp->ipq_nfrags = 1; 923 fp->ipq_ttl = IPFRAGTTL; 924 fp->ipq_p = ip->ip_p; 925 fp->ipq_id = ip->ip_id; 926 fp->ipq_src = ip->ip_src; 927 fp->ipq_dst = ip->ip_dst; 928 fp->ipq_frags = m; 929 m->m_nextpkt = NULL; 930 goto done; 931 } else { 932 fp->ipq_nfrags++; 933#ifdef MAC 934 mac_ipq_update(m, fp); 935#endif 936 } 937 938#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 939 940 /* 941 * Handle ECN by comparing this segment with the first one; 942 * if CE is set, do not lose CE. 943 * drop if CE and not-ECT are mixed for the same packet. 944 */ 945 ecn = ip->ip_tos & IPTOS_ECN_MASK; 946 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 947 if (ecn == IPTOS_ECN_CE) { 948 if (ecn0 == IPTOS_ECN_NOTECT) 949 goto dropfrag; 950 if (ecn0 != IPTOS_ECN_CE) 951 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 952 } 953 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 954 goto dropfrag; 955 956 /* 957 * Find a segment which begins after this one does. 958 */ 959 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 960 if (GETIP(q)->ip_off > ip->ip_off) 961 break; 962 963 /* 964 * If there is a preceding segment, it may provide some of 965 * our data already. If so, drop the data from the incoming 966 * segment. If it provides all of our data, drop us, otherwise 967 * stick new segment in the proper place. 968 * 969 * If some of the data is dropped from the the preceding 970 * segment, then it's checksum is invalidated. 971 */ 972 if (p) { 973 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 974 if (i > 0) { 975 if (i >= ip->ip_len) 976 goto dropfrag; 977 m_adj(m, i); 978 m->m_pkthdr.csum_flags = 0; 979 ip->ip_off += i; 980 ip->ip_len -= i; 981 } 982 m->m_nextpkt = p->m_nextpkt; 983 p->m_nextpkt = m; 984 } else { 985 m->m_nextpkt = fp->ipq_frags; 986 fp->ipq_frags = m; 987 } 988 989 /* 990 * While we overlap succeeding segments trim them or, 991 * if they are completely covered, dequeue them. 992 */ 993 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 994 q = nq) { 995 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 996 if (i < GETIP(q)->ip_len) { 997 GETIP(q)->ip_len -= i; 998 GETIP(q)->ip_off += i; 999 m_adj(q, i); 1000 q->m_pkthdr.csum_flags = 0; 1001 break; 1002 } 1003 nq = q->m_nextpkt; 1004 m->m_nextpkt = nq; 1005 V_ipstat.ips_fragdropped++; 1006 fp->ipq_nfrags--; 1007 m_freem(q); 1008 } 1009 1010 /* 1011 * Check for complete reassembly and perform frag per packet 1012 * limiting. 1013 * 1014 * Frag limiting is performed here so that the nth frag has 1015 * a chance to complete the packet before we drop the packet. 1016 * As a result, n+1 frags are actually allowed per packet, but 1017 * only n will ever be stored. (n = maxfragsperpacket.) 1018 * 1019 */ 1020 next = 0; 1021 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1022 if (GETIP(q)->ip_off != next) { 1023 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1024 V_ipstat.ips_fragdropped += fp->ipq_nfrags; 1025 ip_freef(head, fp); 1026 } 1027 goto done; 1028 } 1029 next += GETIP(q)->ip_len; 1030 } 1031 /* Make sure the last packet didn't have the IP_MF flag */ 1032 if (p->m_flags & M_FRAG) { 1033 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1034 V_ipstat.ips_fragdropped += fp->ipq_nfrags; 1035 ip_freef(head, fp); 1036 } 1037 goto done; 1038 } 1039 1040 /* 1041 * Reassembly is complete. Make sure the packet is a sane size. 1042 */ 1043 q = fp->ipq_frags; 1044 ip = GETIP(q); 1045 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1046 V_ipstat.ips_toolong++; 1047 V_ipstat.ips_fragdropped += fp->ipq_nfrags; 1048 ip_freef(head, fp); 1049 goto done; 1050 } 1051 1052 /* 1053 * Concatenate fragments. 1054 */ 1055 m = q; 1056 t = m->m_next; 1057 m->m_next = NULL; 1058 m_cat(m, t); 1059 nq = q->m_nextpkt; 1060 q->m_nextpkt = NULL; 1061 for (q = nq; q != NULL; q = nq) { 1062 nq = q->m_nextpkt; 1063 q->m_nextpkt = NULL; 1064 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1065 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1066 m_cat(m, q); 1067 } 1068 /* 1069 * In order to do checksumming faster we do 'end-around carry' here 1070 * (and not in for{} loop), though it implies we are not going to 1071 * reassemble more than 64k fragments. 1072 */ 1073 m->m_pkthdr.csum_data = 1074 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1075#ifdef MAC 1076 mac_ipq_reassemble(fp, m); 1077 mac_ipq_destroy(fp); 1078#endif 1079 1080 /* 1081 * Create header for new ip packet by modifying header of first 1082 * packet; dequeue and discard fragment reassembly header. 1083 * Make header visible. 1084 */ 1085 ip->ip_len = (ip->ip_hl << 2) + next; 1086 ip->ip_src = fp->ipq_src; 1087 ip->ip_dst = fp->ipq_dst; 1088 TAILQ_REMOVE(head, fp, ipq_list); 1089 V_nipq--; 1090 uma_zfree(V_ipq_zone, fp); 1091 m->m_len += (ip->ip_hl << 2); 1092 m->m_data -= (ip->ip_hl << 2); 1093 /* some debugging cruft by sklower, below, will go away soon */ 1094 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1095 m_fixhdr(m); 1096 V_ipstat.ips_reassembled++; 1097 IPQ_UNLOCK(); 1098 return (m); 1099 1100dropfrag: 1101 V_ipstat.ips_fragdropped++; 1102 if (fp != NULL) 1103 fp->ipq_nfrags--; 1104 m_freem(m); 1105done: 1106 IPQ_UNLOCK(); 1107 return (NULL); 1108 1109#undef GETIP 1110} 1111 1112/* 1113 * Free a fragment reassembly header and all 1114 * associated datagrams. 1115 */ 1116static void 1117ip_freef(struct ipqhead *fhp, struct ipq *fp) 1118{ 1119 INIT_VNET_INET(curvnet); 1120 struct mbuf *q; 1121 1122 IPQ_LOCK_ASSERT(); 1123 1124 while (fp->ipq_frags) { 1125 q = fp->ipq_frags; 1126 fp->ipq_frags = q->m_nextpkt; 1127 m_freem(q); 1128 } 1129 TAILQ_REMOVE(fhp, fp, ipq_list); 1130 uma_zfree(V_ipq_zone, fp); 1131 V_nipq--; 1132} 1133 1134/* 1135 * IP timer processing; 1136 * if a timer expires on a reassembly 1137 * queue, discard it. 1138 */ 1139void 1140ip_slowtimo(void) 1141{ 1142 VNET_ITERATOR_DECL(vnet_iter); 1143 struct ipq *fp; 1144 int i; 1145 1146 IPQ_LOCK(); 1147 VNET_LIST_RLOCK(); 1148 VNET_FOREACH(vnet_iter) { 1149 CURVNET_SET(vnet_iter); 1150 INIT_VNET_INET(vnet_iter); 1151 for (i = 0; i < IPREASS_NHASH; i++) { 1152 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) { 1153 struct ipq *fpp; 1154 1155 fpp = fp; 1156 fp = TAILQ_NEXT(fp, ipq_list); 1157 if(--fpp->ipq_ttl == 0) { 1158 V_ipstat.ips_fragtimeout += 1159 fpp->ipq_nfrags; 1160 ip_freef(&V_ipq[i], fpp); 1161 } 1162 } 1163 } 1164 /* 1165 * If we are over the maximum number of fragments 1166 * (due to the limit being lowered), drain off 1167 * enough to get down to the new limit. 1168 */ 1169 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) { 1170 for (i = 0; i < IPREASS_NHASH; i++) { 1171 while (V_nipq > V_maxnipq && 1172 !TAILQ_EMPTY(&V_ipq[i])) { 1173 V_ipstat.ips_fragdropped += 1174 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags; 1175 ip_freef(&V_ipq[i], 1176 TAILQ_FIRST(&V_ipq[i])); 1177 } 1178 } 1179 } 1180 CURVNET_RESTORE(); 1181 } 1182 VNET_LIST_RUNLOCK(); 1183 IPQ_UNLOCK(); 1184} 1185 1186/* 1187 * Drain off all datagram fragments. 1188 */ 1189void 1190ip_drain(void) 1191{ 1192 VNET_ITERATOR_DECL(vnet_iter); 1193 int i; 1194 1195 IPQ_LOCK(); 1196 VNET_LIST_RLOCK(); 1197 VNET_FOREACH(vnet_iter) { 1198 CURVNET_SET(vnet_iter); 1199 INIT_VNET_INET(vnet_iter); 1200 for (i = 0; i < IPREASS_NHASH; i++) { 1201 while(!TAILQ_EMPTY(&V_ipq[i])) { 1202 V_ipstat.ips_fragdropped += 1203 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags; 1204 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); 1205 } 1206 } 1207 CURVNET_RESTORE(); 1208 } 1209 VNET_LIST_RUNLOCK(); 1210 IPQ_UNLOCK(); 1211 in_rtqdrain(); 1212} 1213 1214/* 1215 * The protocol to be inserted into ip_protox[] must be already registered 1216 * in inetsw[], either statically or through pf_proto_register(). 1217 */ 1218int 1219ipproto_register(u_char ipproto) 1220{ 1221 struct protosw *pr; 1222 1223 /* Sanity checks. */ 1224 if (ipproto == 0) 1225 return (EPROTONOSUPPORT); 1226 1227 /* 1228 * The protocol slot must not be occupied by another protocol 1229 * already. An index pointing to IPPROTO_RAW is unused. 1230 */ 1231 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1232 if (pr == NULL) 1233 return (EPFNOSUPPORT); 1234 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1235 return (EEXIST); 1236 1237 /* Find the protocol position in inetsw[] and set the index. */ 1238 for (pr = inetdomain.dom_protosw; 1239 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1240 if (pr->pr_domain->dom_family == PF_INET && 1241 pr->pr_protocol && pr->pr_protocol == ipproto) { 1242 /* Be careful to only index valid IP protocols. */ 1243 if (pr->pr_protocol < IPPROTO_MAX) { 1244 ip_protox[pr->pr_protocol] = pr - inetsw; 1245 return (0); 1246 } else 1247 return (EINVAL); 1248 } 1249 } 1250 return (EPROTONOSUPPORT); 1251} 1252 1253int 1254ipproto_unregister(u_char ipproto) 1255{ 1256 struct protosw *pr; 1257 1258 /* Sanity checks. */ 1259 if (ipproto == 0) 1260 return (EPROTONOSUPPORT); 1261 1262 /* Check if the protocol was indeed registered. */ 1263 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1264 if (pr == NULL) 1265 return (EPFNOSUPPORT); 1266 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1267 return (ENOENT); 1268 1269 /* Reset the protocol slot to IPPROTO_RAW. */ 1270 ip_protox[ipproto] = pr - inetsw; 1271 return (0); 1272} 1273 1274/* 1275 * Given address of next destination (final or next hop), 1276 * return internet address info of interface to be used to get there. 1277 */ 1278struct in_ifaddr * 1279ip_rtaddr(struct in_addr dst, u_int fibnum) 1280{ 1281 struct route sro; 1282 struct sockaddr_in *sin; 1283 struct in_ifaddr *ifa; 1284 1285 bzero(&sro, sizeof(sro)); 1286 sin = (struct sockaddr_in *)&sro.ro_dst; 1287 sin->sin_family = AF_INET; 1288 sin->sin_len = sizeof(*sin); 1289 sin->sin_addr = dst; 1290 in_rtalloc_ign(&sro, 0, fibnum); 1291 1292 if (sro.ro_rt == NULL) 1293 return (NULL); 1294 1295 ifa = ifatoia(sro.ro_rt->rt_ifa); 1296 RTFREE(sro.ro_rt); 1297 return (ifa); 1298} 1299 1300u_char inetctlerrmap[PRC_NCMDS] = { 1301 0, 0, 0, 0, 1302 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1303 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1304 EMSGSIZE, EHOSTUNREACH, 0, 0, 1305 0, 0, EHOSTUNREACH, 0, 1306 ENOPROTOOPT, ECONNREFUSED 1307}; 1308 1309/* 1310 * Forward a packet. If some error occurs return the sender 1311 * an icmp packet. Note we can't always generate a meaningful 1312 * icmp message because icmp doesn't have a large enough repertoire 1313 * of codes and types. 1314 * 1315 * If not forwarding, just drop the packet. This could be confusing 1316 * if ipforwarding was zero but some routing protocol was advancing 1317 * us as a gateway to somewhere. However, we must let the routing 1318 * protocol deal with that. 1319 * 1320 * The srcrt parameter indicates whether the packet is being forwarded 1321 * via a source route. 1322 */ 1323void 1324ip_forward(struct mbuf *m, int srcrt) 1325{ 1326 INIT_VNET_INET(curvnet); 1327 struct ip *ip = mtod(m, struct ip *); 1328 struct in_ifaddr *ia = NULL; 1329 struct mbuf *mcopy; 1330 struct in_addr dest; 1331 struct route ro; 1332 int error, type = 0, code = 0, mtu = 0; 1333 1334 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1335 V_ipstat.ips_cantforward++; 1336 m_freem(m); 1337 return; 1338 } 1339#ifdef IPSTEALTH 1340 if (!V_ipstealth) { 1341#endif 1342 if (ip->ip_ttl <= IPTTLDEC) { 1343 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1344 0, 0); 1345 return; 1346 } 1347#ifdef IPSTEALTH 1348 } 1349#endif 1350 1351 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); 1352 if (!srcrt && ia == NULL) { 1353 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1354 return; 1355 } 1356 1357 /* 1358 * Save the IP header and at most 8 bytes of the payload, 1359 * in case we need to generate an ICMP message to the src. 1360 * 1361 * XXX this can be optimized a lot by saving the data in a local 1362 * buffer on the stack (72 bytes at most), and only allocating the 1363 * mbuf if really necessary. The vast majority of the packets 1364 * are forwarded without having to send an ICMP back (either 1365 * because unnecessary, or because rate limited), so we are 1366 * really we are wasting a lot of work here. 1367 * 1368 * We don't use m_copy() because it might return a reference 1369 * to a shared cluster. Both this function and ip_output() 1370 * assume exclusive access to the IP header in `m', so any 1371 * data in a cluster may change before we reach icmp_error(). 1372 */ 1373 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1374 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1375 /* 1376 * It's probably ok if the pkthdr dup fails (because 1377 * the deep copy of the tag chain failed), but for now 1378 * be conservative and just discard the copy since 1379 * code below may some day want the tags. 1380 */ 1381 m_free(mcopy); 1382 mcopy = NULL; 1383 } 1384 if (mcopy != NULL) { 1385 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy)); 1386 mcopy->m_pkthdr.len = mcopy->m_len; 1387 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1388 } 1389 1390#ifdef IPSTEALTH 1391 if (!V_ipstealth) { 1392#endif 1393 ip->ip_ttl -= IPTTLDEC; 1394#ifdef IPSTEALTH 1395 } 1396#endif 1397 1398 /* 1399 * If forwarding packet using same interface that it came in on, 1400 * perhaps should send a redirect to sender to shortcut a hop. 1401 * Only send redirect if source is sending directly to us, 1402 * and if packet was not source routed (or has any options). 1403 * Also, don't send redirect if forwarding using a default route 1404 * or a route modified by a redirect. 1405 */ 1406 dest.s_addr = 0; 1407 if (!srcrt && V_ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) { 1408 struct sockaddr_in *sin; 1409 struct rtentry *rt; 1410 1411 bzero(&ro, sizeof(ro)); 1412 sin = (struct sockaddr_in *)&ro.ro_dst; 1413 sin->sin_family = AF_INET; 1414 sin->sin_len = sizeof(*sin); 1415 sin->sin_addr = ip->ip_dst; 1416 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 1417 1418 rt = ro.ro_rt; 1419 1420 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1421 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1422#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1423 u_long src = ntohl(ip->ip_src.s_addr); 1424 1425 if (RTA(rt) && 1426 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1427 if (rt->rt_flags & RTF_GATEWAY) 1428 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1429 else 1430 dest.s_addr = ip->ip_dst.s_addr; 1431 /* Router requirements says to only send host redirects */ 1432 type = ICMP_REDIRECT; 1433 code = ICMP_REDIRECT_HOST; 1434 } 1435 } 1436 if (rt) 1437 RTFREE(rt); 1438 } 1439 1440 /* 1441 * Try to cache the route MTU from ip_output so we can consider it for 1442 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. 1443 */ 1444 bzero(&ro, sizeof(ro)); 1445 1446 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1447 1448 if (error == EMSGSIZE && ro.ro_rt) 1449 mtu = ro.ro_rt->rt_rmx.rmx_mtu; 1450 if (ro.ro_rt) 1451 RTFREE(ro.ro_rt); 1452 1453 if (error) 1454 V_ipstat.ips_cantforward++; 1455 else { 1456 V_ipstat.ips_forward++; 1457 if (type) 1458 V_ipstat.ips_redirectsent++; 1459 else { 1460 if (mcopy) 1461 m_freem(mcopy); 1462 return; 1463 } 1464 } 1465 if (mcopy == NULL) 1466 return; 1467 1468 switch (error) { 1469 1470 case 0: /* forwarded, but need redirect */ 1471 /* type, code set above */ 1472 break; 1473 1474 case ENETUNREACH: /* shouldn't happen, checked above */ 1475 case EHOSTUNREACH: 1476 case ENETDOWN: 1477 case EHOSTDOWN: 1478 default: 1479 type = ICMP_UNREACH; 1480 code = ICMP_UNREACH_HOST; 1481 break; 1482 1483 case EMSGSIZE: 1484 type = ICMP_UNREACH; 1485 code = ICMP_UNREACH_NEEDFRAG; 1486 1487#ifdef IPSEC 1488 /* 1489 * If IPsec is configured for this path, 1490 * override any possibly mtu value set by ip_output. 1491 */ 1492 mtu = ip_ipsec_mtu(m, mtu); 1493#endif /* IPSEC */ 1494 /* 1495 * If the MTU was set before make sure we are below the 1496 * interface MTU. 1497 * If the MTU wasn't set before use the interface mtu or 1498 * fall back to the next smaller mtu step compared to the 1499 * current packet size. 1500 */ 1501 if (mtu != 0) { 1502 if (ia != NULL) 1503 mtu = min(mtu, ia->ia_ifp->if_mtu); 1504 } else { 1505 if (ia != NULL) 1506 mtu = ia->ia_ifp->if_mtu; 1507 else 1508 mtu = ip_next_mtu(ip->ip_len, 0); 1509 } 1510 V_ipstat.ips_cantfrag++; 1511 break; 1512 1513 case ENOBUFS: 1514 /* 1515 * A router should not generate ICMP_SOURCEQUENCH as 1516 * required in RFC1812 Requirements for IP Version 4 Routers. 1517 * Source quench could be a big problem under DoS attacks, 1518 * or if the underlying interface is rate-limited. 1519 * Those who need source quench packets may re-enable them 1520 * via the net.inet.ip.sendsourcequench sysctl. 1521 */ 1522 if (V_ip_sendsourcequench == 0) { 1523 m_freem(mcopy); 1524 return; 1525 } else { 1526 type = ICMP_SOURCEQUENCH; 1527 code = 0; 1528 } 1529 break; 1530 1531 case EACCES: /* ipfw denied packet */ 1532 m_freem(mcopy); 1533 return; 1534 } 1535 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1536} 1537 1538void 1539ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1540 struct mbuf *m) 1541{ 1542 INIT_VNET_NET(inp->inp_vnet); 1543 1544 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1545 struct bintime bt; 1546 1547 bintime(&bt); 1548 if (inp->inp_socket->so_options & SO_BINTIME) { 1549 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1550 SCM_BINTIME, SOL_SOCKET); 1551 if (*mp) 1552 mp = &(*mp)->m_next; 1553 } 1554 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1555 struct timeval tv; 1556 1557 bintime2timeval(&bt, &tv); 1558 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1559 SCM_TIMESTAMP, SOL_SOCKET); 1560 if (*mp) 1561 mp = &(*mp)->m_next; 1562 } 1563 } 1564 if (inp->inp_flags & INP_RECVDSTADDR) { 1565 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1566 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1567 if (*mp) 1568 mp = &(*mp)->m_next; 1569 } 1570 if (inp->inp_flags & INP_RECVTTL) { 1571 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1572 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1573 if (*mp) 1574 mp = &(*mp)->m_next; 1575 } 1576#ifdef notyet 1577 /* XXX 1578 * Moving these out of udp_input() made them even more broken 1579 * than they already were. 1580 */ 1581 /* options were tossed already */ 1582 if (inp->inp_flags & INP_RECVOPTS) { 1583 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1584 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1585 if (*mp) 1586 mp = &(*mp)->m_next; 1587 } 1588 /* ip_srcroute doesn't do what we want here, need to fix */ 1589 if (inp->inp_flags & INP_RECVRETOPTS) { 1590 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1591 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1592 if (*mp) 1593 mp = &(*mp)->m_next; 1594 } 1595#endif 1596 if (inp->inp_flags & INP_RECVIF) { 1597 struct ifnet *ifp; 1598 struct sdlbuf { 1599 struct sockaddr_dl sdl; 1600 u_char pad[32]; 1601 } sdlbuf; 1602 struct sockaddr_dl *sdp; 1603 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1604 1605 if (((ifp = m->m_pkthdr.rcvif)) 1606 && ( ifp->if_index && (ifp->if_index <= V_if_index))) { 1607 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1608 /* 1609 * Change our mind and don't try copy. 1610 */ 1611 if ((sdp->sdl_family != AF_LINK) 1612 || (sdp->sdl_len > sizeof(sdlbuf))) { 1613 goto makedummy; 1614 } 1615 bcopy(sdp, sdl2, sdp->sdl_len); 1616 } else { 1617makedummy: 1618 sdl2->sdl_len 1619 = offsetof(struct sockaddr_dl, sdl_data[0]); 1620 sdl2->sdl_family = AF_LINK; 1621 sdl2->sdl_index = 0; 1622 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1623 } 1624 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1625 IP_RECVIF, IPPROTO_IP); 1626 if (*mp) 1627 mp = &(*mp)->m_next; 1628 } 1629} 1630 1631/* 1632 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1633 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1634 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1635 * compiled. 1636 */ 1637int 1638ip_rsvp_init(struct socket *so) 1639{ 1640 INIT_VNET_INET(so->so_vnet); 1641 1642 if (so->so_type != SOCK_RAW || 1643 so->so_proto->pr_protocol != IPPROTO_RSVP) 1644 return EOPNOTSUPP; 1645 1646 if (V_ip_rsvpd != NULL) 1647 return EADDRINUSE; 1648 1649 V_ip_rsvpd = so; 1650 /* 1651 * This may seem silly, but we need to be sure we don't over-increment 1652 * the RSVP counter, in case something slips up. 1653 */ 1654 if (!V_ip_rsvp_on) { 1655 V_ip_rsvp_on = 1; 1656 V_rsvp_on++; 1657 } 1658 1659 return 0; 1660} 1661 1662int 1663ip_rsvp_done(void) 1664{ 1665 INIT_VNET_INET(curvnet); 1666 1667 V_ip_rsvpd = NULL; 1668 /* 1669 * This may seem silly, but we need to be sure we don't over-decrement 1670 * the RSVP counter, in case something slips up. 1671 */ 1672 if (V_ip_rsvp_on) { 1673 V_ip_rsvp_on = 0; 1674 V_rsvp_on--; 1675 } 1676 return 0; 1677} 1678 1679void 1680rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1681{ 1682 INIT_VNET_INET(curvnet); 1683 1684 if (rsvp_input_p) { /* call the real one if loaded */ 1685 rsvp_input_p(m, off); 1686 return; 1687 } 1688 1689 /* Can still get packets with rsvp_on = 0 if there is a local member 1690 * of the group to which the RSVP packet is addressed. But in this 1691 * case we want to throw the packet away. 1692 */ 1693 1694 if (!V_rsvp_on) { 1695 m_freem(m); 1696 return; 1697 } 1698 1699 if (V_ip_rsvpd != NULL) { 1700 rip_input(m, off); 1701 return; 1702 } 1703 /* Drop the packet */ 1704 m_freem(m); 1705} 1706