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("$FreeBSD$"); 34 35 #include "opt_bootp.h" 36 #include "opt_ipfw.h" 37 #include "opt_ipstealth.h" 38 #include "opt_ipsec.h" 39 #include "opt_route.h" 40 #include "opt_mac.h" 41 #include "opt_carp.h" 42 43 #include <sys/param.h> 44 #include <sys/systm.h> 45 #include <sys/callout.h> 46 #include <sys/mbuf.h> 47 #include <sys/malloc.h> 48 #include <sys/domain.h> 49 #include <sys/protosw.h> 50 #include <sys/socket.h> 51 #include <sys/time.h> 52 #include <sys/kernel.h> 53 #include <sys/lock.h> 54 #include <sys/rwlock.h> 55 #include <sys/syslog.h> 56 #include <sys/sysctl.h> 57 #include <sys/vimage.h> 58 59 #include <net/pfil.h> 60 #include <net/if.h> 61 #include <net/if_types.h> 62 #include <net/if_var.h> 63 #include <net/if_dl.h> 64 #include <net/route.h> 65 #include <net/netisr.h> 66 #include <net/vnet.h> 67 #include <net/flowtable.h> 68 69 #include <netinet/in.h> 70 #include <netinet/in_systm.h> 71 #include <netinet/in_var.h> 72 #include <netinet/ip.h> 73 #include <netinet/in_pcb.h> 74 #include <netinet/ip_var.h> 75 #include <netinet/ip_icmp.h> 76 #include <netinet/ip_options.h> 77 #include <machine/in_cksum.h> 78 #include <netinet/vinet.h> 79 #ifdef DEV_CARP 80 #include <netinet/ip_carp.h> 81 #endif 82 #ifdef IPSEC 83 #include <netinet/ip_ipsec.h> 84 #endif /* IPSEC */ 85 86 #include <sys/socketvar.h> 87 88 /* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */ 89 #include <netinet/ip_fw.h> 90 #include <netinet/ip_dummynet.h> 91 92 #include <security/mac/mac_framework.h> 93 94 #ifdef CTASSERT 95 CTASSERT(sizeof(struct ip) == 20); 96 #endif 97 98 #ifndef VIMAGE 99 #ifndef VIMAGE_GLOBALS 100 struct vnet_inet vnet_inet_0; 101 #endif 102 #endif 103 104 #ifdef VIMAGE_GLOBALS 105 static int ipsendredirects; 106 static int ip_checkinterface; 107 static int ip_keepfaith; 108 static int ip_sendsourcequench; 109 int ip_defttl; 110 int ip_do_randomid; 111 int ipforwarding; 112 struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 113 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 114 u_long in_ifaddrhmask; /* mask for hash table */ 115 struct ipstat ipstat; 116 static int ip_rsvp_on; 117 struct socket *ip_rsvpd; 118 int rsvp_on; 119 static struct ipqhead ipq[IPREASS_NHASH]; 120 static int maxnipq; /* Administrative limit on # reass queues. */ 121 static int maxfragsperpacket; 122 int ipstealth; 123 static int nipq; /* Total # of reass queues */ 124 #endif 125 126 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_FORWARDING, 127 forwarding, CTLFLAG_RW, ipforwarding, 0, 128 "Enable IP forwarding between interfaces"); 129 130 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_SENDREDIRECTS, 131 redirect, CTLFLAG_RW, ipsendredirects, 0, 132 "Enable sending IP redirects"); 133 134 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_DEFTTL, 135 ttl, CTLFLAG_RW, ip_defttl, 0, "Maximum TTL on IP packets"); 136 137 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_KEEPFAITH, 138 keepfaith, CTLFLAG_RW, ip_keepfaith, 0, 139 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 140 141 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, 142 sendsourcequench, CTLFLAG_RW, ip_sendsourcequench, 0, 143 "Enable the transmission of source quench packets"); 144 145 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, random_id, 146 CTLFLAG_RW, ip_do_randomid, 0, "Assign random ip_id values"); 147 148 /* 149 * XXX - Setting ip_checkinterface mostly implements the receive side of 150 * the Strong ES model described in RFC 1122, but since the routing table 151 * and transmit implementation do not implement the Strong ES model, 152 * setting this to 1 results in an odd hybrid. 153 * 154 * XXX - ip_checkinterface currently must be disabled if you use ipnat 155 * to translate the destination address to another local interface. 156 * 157 * XXX - ip_checkinterface must be disabled if you add IP aliases 158 * to the loopback interface instead of the interface where the 159 * packets for those addresses are received. 160 */ 161 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, 162 check_interface, CTLFLAG_RW, ip_checkinterface, 0, 163 "Verify packet arrives on correct interface"); 164 165 struct pfil_head inet_pfil_hook; /* Packet filter hooks */ 166 167 static struct ifqueue ipintrq; 168 static int ipqmaxlen = IFQ_MAXLEN; 169 170 extern struct domain inetdomain; 171 extern struct protosw inetsw[]; 172 u_char ip_protox[IPPROTO_MAX]; 173 174 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 175 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 176 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 177 &ipintrq.ifq_drops, 0, 178 "Number of packets dropped from the IP input queue"); 179 180 SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 181 ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 182 183 #ifdef VIMAGE_GLOBALS 184 static uma_zone_t ipq_zone; 185 #endif 186 static struct mtx ipqlock; 187 188 #define IPQ_LOCK() mtx_lock(&ipqlock) 189 #define IPQ_UNLOCK() mtx_unlock(&ipqlock) 190 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 191 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 192 193 static void maxnipq_update(void); 194 static void ipq_zone_change(void *); 195 196 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, fragpackets, 197 CTLFLAG_RD, nipq, 0, 198 "Current number of IPv4 fragment reassembly queue entries"); 199 200 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, maxfragsperpacket, 201 CTLFLAG_RW, maxfragsperpacket, 0, 202 "Maximum number of IPv4 fragments allowed per packet"); 203 204 struct callout ipport_tick_callout; 205 206 #ifdef IPCTL_DEFMTU 207 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 208 &ip_mtu, 0, "Default MTU"); 209 #endif 210 211 #ifdef IPSTEALTH 212 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 213 ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding"); 214 #endif 215 static int ip_output_flowtable_size = 2048; 216 TUNABLE_INT("net.inet.ip.output_flowtable_size", &ip_output_flowtable_size); 217 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, output_flowtable_size, 218 CTLFLAG_RDTUN, ip_output_flowtable_size, 2048, 219 "number of entries in the per-cpu output flow caches"); 220 221 /* 222 * ipfw_ether and ipfw_bridge hooks. 223 * XXX: Temporary until those are converted to pfil_hooks as well. 224 */ 225 ip_fw_chk_t *ip_fw_chk_ptr = NULL; 226 ip_dn_io_t *ip_dn_io_ptr = NULL; 227 #ifdef VIMAGE_GLOBALS 228 int fw_one_pass; 229 #endif 230 struct flowtable *ip_ft; 231 232 static void ip_freef(struct ipqhead *, struct ipq *); 233 234 #ifndef VIMAGE_GLOBALS 235 static void vnet_inet_register(void); 236 237 static const vnet_modinfo_t vnet_inet_modinfo = { 238 .vmi_id = VNET_MOD_INET, 239 .vmi_name = "inet", 240 .vmi_size = sizeof(struct vnet_inet) 241 }; 242 243 static void vnet_inet_register() 244 { 245 246 vnet_mod_register(&vnet_inet_modinfo); 247 } 248 249 SYSINIT(inet, SI_SUB_PROTO_BEGIN, SI_ORDER_FIRST, vnet_inet_register, 0); 250 #endif 251 252 /* 253 * IP initialization: fill in IP protocol switch table. 254 * All protocols not implemented in kernel go to raw IP protocol handler. 255 */ 256 void 257 ip_init(void) 258 { 259 INIT_VNET_INET(curvnet); 260 struct protosw *pr; 261 int i; 262 263 V_ipsendredirects = 1; /* XXX */ 264 V_ip_checkinterface = 0; 265 V_ip_keepfaith = 0; 266 V_ip_sendsourcequench = 0; 267 V_rsvp_on = 0; 268 V_ip_defttl = IPDEFTTL; 269 V_ip_do_randomid = 0; 270 V_ip_id = time_second & 0xffff; 271 V_ipforwarding = 0; 272 V_ipstealth = 0; 273 V_nipq = 0; /* Total # of reass queues */ 274 275 V_ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */ 276 V_ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */ 277 V_ipport_firstauto = IPPORT_EPHEMERALFIRST; /* 10000 */ 278 V_ipport_lastauto = IPPORT_EPHEMERALLAST; /* 65535 */ 279 V_ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */ 280 V_ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */ 281 V_ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */ 282 V_ipport_reservedlow = 0; 283 V_ipport_randomized = 1; /* user controlled via sysctl */ 284 V_ipport_randomcps = 10; /* user controlled via sysctl */ 285 V_ipport_randomtime = 45; /* user controlled via sysctl */ 286 V_ipport_stoprandom = 0; /* toggled by ipport_tick */ 287 288 V_fw_one_pass = 1; 289 290 #ifdef NOTYET 291 /* XXX global static but not instantiated in this file */ 292 V_ipfastforward_active = 0; 293 V_subnetsarelocal = 0; 294 V_sameprefixcarponly = 0; 295 #endif 296 297 TAILQ_INIT(&V_in_ifaddrhead); 298 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 299 300 /* Initialize IP reassembly queue. */ 301 for (i = 0; i < IPREASS_NHASH; i++) 302 TAILQ_INIT(&V_ipq[i]); 303 V_maxnipq = nmbclusters / 32; 304 V_maxfragsperpacket = 16; 305 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 306 NULL, UMA_ALIGN_PTR, 0); 307 maxnipq_update(); 308 309 /* Skip initialization of globals for non-default instances. */ 310 if (!IS_DEFAULT_VNET(curvnet)) 311 return; 312 313 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 314 if (pr == NULL) 315 panic("ip_init: PF_INET not found"); 316 317 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 318 for (i = 0; i < IPPROTO_MAX; i++) 319 ip_protox[i] = pr - inetsw; 320 /* 321 * Cycle through IP protocols and put them into the appropriate place 322 * in ip_protox[]. 323 */ 324 for (pr = inetdomain.dom_protosw; 325 pr < inetdomain.dom_protoswNPROTOSW; pr++) 326 if (pr->pr_domain->dom_family == PF_INET && 327 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 328 /* Be careful to only index valid IP protocols. */ 329 if (pr->pr_protocol < IPPROTO_MAX) 330 ip_protox[pr->pr_protocol] = pr - inetsw; 331 } 332 333 /* Initialize packet filter hooks. */ 334 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 335 inet_pfil_hook.ph_af = AF_INET; 336 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 337 printf("%s: WARNING: unable to register pfil hook, " 338 "error %d\n", __func__, i); 339 340 /* Start ipport_tick. */ 341 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE); 342 callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL); 343 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, 344 SHUTDOWN_PRI_DEFAULT); 345 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, 346 NULL, EVENTHANDLER_PRI_ANY); 347 348 /* Initialize various other remaining things. */ 349 IPQ_LOCK_INIT(); 350 ipintrq.ifq_maxlen = ipqmaxlen; 351 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF); 352 netisr_register(NETISR_IP, ip_input, &ipintrq, 0); 353 354 ip_ft = flowtable_alloc(ip_output_flowtable_size, FL_PCPU); 355 } 356 357 void 358 ip_fini(void *xtp) 359 { 360 361 callout_stop(&ipport_tick_callout); 362 } 363 364 /* 365 * Ip input routine. Checksum and byte swap header. If fragmented 366 * try to reassemble. Process options. Pass to next level. 367 */ 368 void 369 ip_input(struct mbuf *m) 370 { 371 INIT_VNET_INET(curvnet); 372 struct ip *ip = NULL; 373 struct in_ifaddr *ia = NULL; 374 struct ifaddr *ifa; 375 struct ifnet *ifp; 376 int checkif, hlen = 0; 377 u_short sum; 378 int dchg = 0; /* dest changed after fw */ 379 struct in_addr odst; /* original dst address */ 380 381 M_ASSERTPKTHDR(m); 382 383 if (m->m_flags & M_FASTFWD_OURS) { 384 /* 385 * Firewall or NAT changed destination to local. 386 * We expect ip_len and ip_off to be in host byte order. 387 */ 388 m->m_flags &= ~M_FASTFWD_OURS; 389 /* Set up some basics that will be used later. */ 390 ip = mtod(m, struct ip *); 391 hlen = ip->ip_hl << 2; 392 goto ours; 393 } 394 395 IPSTAT_INC(ips_total); 396 397 if (m->m_pkthdr.len < sizeof(struct ip)) 398 goto tooshort; 399 400 if (m->m_len < sizeof (struct ip) && 401 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 402 IPSTAT_INC(ips_toosmall); 403 return; 404 } 405 ip = mtod(m, struct ip *); 406 407 if (ip->ip_v != IPVERSION) { 408 IPSTAT_INC(ips_badvers); 409 goto bad; 410 } 411 412 hlen = ip->ip_hl << 2; 413 if (hlen < sizeof(struct ip)) { /* minimum header length */ 414 IPSTAT_INC(ips_badhlen); 415 goto bad; 416 } 417 if (hlen > m->m_len) { 418 if ((m = m_pullup(m, hlen)) == NULL) { 419 IPSTAT_INC(ips_badhlen); 420 return; 421 } 422 ip = mtod(m, struct ip *); 423 } 424 425 /* 127/8 must not appear on wire - RFC1122 */ 426 ifp = m->m_pkthdr.rcvif; 427 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 428 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 429 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 430 IPSTAT_INC(ips_badaddr); 431 goto bad; 432 } 433 } 434 435 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 436 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 437 } else { 438 if (hlen == sizeof(struct ip)) { 439 sum = in_cksum_hdr(ip); 440 } else { 441 sum = in_cksum(m, hlen); 442 } 443 } 444 if (sum) { 445 IPSTAT_INC(ips_badsum); 446 goto bad; 447 } 448 449 #ifdef ALTQ 450 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 451 /* packet is dropped by traffic conditioner */ 452 return; 453 #endif 454 455 /* 456 * Convert fields to host representation. 457 */ 458 ip->ip_len = ntohs(ip->ip_len); 459 if (ip->ip_len < hlen) { 460 IPSTAT_INC(ips_badlen); 461 goto bad; 462 } 463 ip->ip_off = ntohs(ip->ip_off); 464 465 /* 466 * Check that the amount of data in the buffers 467 * is as at least much as the IP header would have us expect. 468 * Trim mbufs if longer than we expect. 469 * Drop packet if shorter than we expect. 470 */ 471 if (m->m_pkthdr.len < ip->ip_len) { 472 tooshort: 473 IPSTAT_INC(ips_tooshort); 474 goto bad; 475 } 476 if (m->m_pkthdr.len > ip->ip_len) { 477 if (m->m_len == m->m_pkthdr.len) { 478 m->m_len = ip->ip_len; 479 m->m_pkthdr.len = ip->ip_len; 480 } else 481 m_adj(m, ip->ip_len - m->m_pkthdr.len); 482 } 483 #ifdef IPSEC 484 /* 485 * Bypass packet filtering for packets from a tunnel (gif). 486 */ 487 if (ip_ipsec_filtertunnel(m)) 488 goto passin; 489 #endif /* IPSEC */ 490 491 /* 492 * Run through list of hooks for input packets. 493 * 494 * NB: Beware of the destination address changing (e.g. 495 * by NAT rewriting). When this happens, tell 496 * ip_forward to do the right thing. 497 */ 498 499 /* Jump over all PFIL processing if hooks are not active. */ 500 if (!PFIL_HOOKED(&inet_pfil_hook)) 501 goto passin; 502 503 odst = ip->ip_dst; 504 if (pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0) 505 return; 506 if (m == NULL) /* consumed by filter */ 507 return; 508 509 ip = mtod(m, struct ip *); 510 dchg = (odst.s_addr != ip->ip_dst.s_addr); 511 ifp = m->m_pkthdr.rcvif; 512 513 #ifdef IPFIREWALL_FORWARD 514 if (m->m_flags & M_FASTFWD_OURS) { 515 m->m_flags &= ~M_FASTFWD_OURS; 516 goto ours; 517 } 518 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 519 /* 520 * Directly ship on the packet. This allows to forward packets 521 * that were destined for us to some other directly connected 522 * host. 523 */ 524 ip_forward(m, dchg); 525 return; 526 } 527 #endif /* IPFIREWALL_FORWARD */ 528 529 passin: 530 /* 531 * Process options and, if not destined for us, 532 * ship it on. ip_dooptions returns 1 when an 533 * error was detected (causing an icmp message 534 * to be sent and the original packet to be freed). 535 */ 536 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 537 return; 538 539 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 540 * matter if it is destined to another node, or whether it is 541 * a multicast one, RSVP wants it! and prevents it from being forwarded 542 * anywhere else. Also checks if the rsvp daemon is running before 543 * grabbing the packet. 544 */ 545 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 546 goto ours; 547 548 /* 549 * Check our list of addresses, to see if the packet is for us. 550 * If we don't have any addresses, assume any unicast packet 551 * we receive might be for us (and let the upper layers deal 552 * with it). 553 */ 554 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 555 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 556 goto ours; 557 558 /* 559 * Enable a consistency check between the destination address 560 * and the arrival interface for a unicast packet (the RFC 1122 561 * strong ES model) if IP forwarding is disabled and the packet 562 * is not locally generated and the packet is not subject to 563 * 'ipfw fwd'. 564 * 565 * XXX - Checking also should be disabled if the destination 566 * address is ipnat'ed to a different interface. 567 * 568 * XXX - Checking is incompatible with IP aliases added 569 * to the loopback interface instead of the interface where 570 * the packets are received. 571 * 572 * XXX - This is the case for carp vhost IPs as well so we 573 * insert a workaround. If the packet got here, we already 574 * checked with carp_iamatch() and carp_forus(). 575 */ 576 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 577 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && 578 #ifdef DEV_CARP 579 !ifp->if_carp && 580 #endif 581 (dchg == 0); 582 583 /* 584 * Check for exact addresses in the hash bucket. 585 */ 586 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 587 /* 588 * If the address matches, verify that the packet 589 * arrived via the correct interface if checking is 590 * enabled. 591 */ 592 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 593 (!checkif || ia->ia_ifp == ifp)) 594 goto ours; 595 } 596 /* 597 * Check for broadcast addresses. 598 * 599 * Only accept broadcast packets that arrive via the matching 600 * interface. Reception of forwarded directed broadcasts would 601 * be handled via ip_forward() and ether_output() with the loopback 602 * into the stack for SIMPLEX interfaces handled by ether_output(). 603 */ 604 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 605 IF_ADDR_LOCK(ifp); 606 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 607 if (ifa->ifa_addr->sa_family != AF_INET) 608 continue; 609 ia = ifatoia(ifa); 610 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 611 ip->ip_dst.s_addr) { 612 IF_ADDR_UNLOCK(ifp); 613 goto ours; 614 } 615 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) { 616 IF_ADDR_UNLOCK(ifp); 617 goto ours; 618 } 619 #ifdef BOOTP_COMPAT 620 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { 621 IF_ADDR_UNLOCK(ifp); 622 goto ours; 623 } 624 #endif 625 } 626 IF_ADDR_UNLOCK(ifp); 627 } 628 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 629 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 630 IPSTAT_INC(ips_cantforward); 631 m_freem(m); 632 return; 633 } 634 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 635 if (V_ip_mrouter) { 636 /* 637 * If we are acting as a multicast router, all 638 * incoming multicast packets are passed to the 639 * kernel-level multicast forwarding function. 640 * The packet is returned (relatively) intact; if 641 * ip_mforward() returns a non-zero value, the packet 642 * must be discarded, else it may be accepted below. 643 */ 644 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 645 IPSTAT_INC(ips_cantforward); 646 m_freem(m); 647 return; 648 } 649 650 /* 651 * The process-level routing daemon needs to receive 652 * all multicast IGMP packets, whether or not this 653 * host belongs to their destination groups. 654 */ 655 if (ip->ip_p == IPPROTO_IGMP) 656 goto ours; 657 IPSTAT_INC(ips_forward); 658 } 659 /* 660 * Assume the packet is for us, to avoid prematurely taking 661 * a lock on the in_multi hash. Protocols must perform 662 * their own filtering and update statistics accordingly. 663 */ 664 goto ours; 665 } 666 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 667 goto ours; 668 if (ip->ip_dst.s_addr == INADDR_ANY) 669 goto ours; 670 671 /* 672 * FAITH(Firewall Aided Internet Translator) 673 */ 674 if (ifp && ifp->if_type == IFT_FAITH) { 675 if (V_ip_keepfaith) { 676 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 677 goto ours; 678 } 679 m_freem(m); 680 return; 681 } 682 683 /* 684 * Not for us; forward if possible and desirable. 685 */ 686 if (V_ipforwarding == 0) { 687 IPSTAT_INC(ips_cantforward); 688 m_freem(m); 689 } else { 690 #ifdef IPSEC 691 if (ip_ipsec_fwd(m)) 692 goto bad; 693 #endif /* IPSEC */ 694 ip_forward(m, dchg); 695 } 696 return; 697 698 ours: 699 #ifdef IPSTEALTH 700 /* 701 * IPSTEALTH: Process non-routing options only 702 * if the packet is destined for us. 703 */ 704 if (V_ipstealth && hlen > sizeof (struct ip) && 705 ip_dooptions(m, 1)) 706 return; 707 #endif /* IPSTEALTH */ 708 709 /* Count the packet in the ip address stats */ 710 if (ia != NULL) { 711 ia->ia_ifa.if_ipackets++; 712 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 713 } 714 715 /* 716 * Attempt reassembly; if it succeeds, proceed. 717 * ip_reass() will return a different mbuf. 718 */ 719 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 720 m = ip_reass(m); 721 if (m == NULL) 722 return; 723 ip = mtod(m, struct ip *); 724 /* Get the header length of the reassembled packet */ 725 hlen = ip->ip_hl << 2; 726 } 727 728 /* 729 * Further protocols expect the packet length to be w/o the 730 * IP header. 731 */ 732 ip->ip_len -= hlen; 733 734 #ifdef IPSEC 735 /* 736 * enforce IPsec policy checking if we are seeing last header. 737 * note that we do not visit this with protocols with pcb layer 738 * code - like udp/tcp/raw ip. 739 */ 740 if (ip_ipsec_input(m)) 741 goto bad; 742 #endif /* IPSEC */ 743 744 /* 745 * Switch out to protocol's input routine. 746 */ 747 IPSTAT_INC(ips_delivered); 748 749 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 750 return; 751 bad: 752 m_freem(m); 753 } 754 755 /* 756 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 757 * max has slightly different semantics than the sysctl, for historical 758 * reasons. 759 */ 760 static void 761 maxnipq_update(void) 762 { 763 INIT_VNET_INET(curvnet); 764 765 /* 766 * -1 for unlimited allocation. 767 */ 768 if (V_maxnipq < 0) 769 uma_zone_set_max(V_ipq_zone, 0); 770 /* 771 * Positive number for specific bound. 772 */ 773 if (V_maxnipq > 0) 774 uma_zone_set_max(V_ipq_zone, V_maxnipq); 775 /* 776 * Zero specifies no further fragment queue allocation -- set the 777 * bound very low, but rely on implementation elsewhere to actually 778 * prevent allocation and reclaim current queues. 779 */ 780 if (V_maxnipq == 0) 781 uma_zone_set_max(V_ipq_zone, 1); 782 } 783 784 static void 785 ipq_zone_change(void *tag) 786 { 787 INIT_VNET_INET(curvnet); 788 789 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) { 790 V_maxnipq = nmbclusters / 32; 791 maxnipq_update(); 792 } 793 } 794 795 static int 796 sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 797 { 798 INIT_VNET_INET(curvnet); 799 int error, i; 800 801 i = V_maxnipq; 802 error = sysctl_handle_int(oidp, &i, 0, req); 803 if (error || !req->newptr) 804 return (error); 805 806 /* 807 * XXXRW: Might be a good idea to sanity check the argument and place 808 * an extreme upper bound. 809 */ 810 if (i < -1) 811 return (EINVAL); 812 V_maxnipq = i; 813 maxnipq_update(); 814 return (0); 815 } 816 817 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 818 NULL, 0, sysctl_maxnipq, "I", 819 "Maximum number of IPv4 fragment reassembly queue entries"); 820 821 /* 822 * Take incoming datagram fragment and try to reassemble it into 823 * whole datagram. If the argument is the first fragment or one 824 * in between the function will return NULL and store the mbuf 825 * in the fragment chain. If the argument is the last fragment 826 * the packet will be reassembled and the pointer to the new 827 * mbuf returned for further processing. Only m_tags attached 828 * to the first packet/fragment are preserved. 829 * The IP header is *NOT* adjusted out of iplen. 830 */ 831 struct mbuf * 832 ip_reass(struct mbuf *m) 833 { 834 INIT_VNET_INET(curvnet); 835 struct ip *ip; 836 struct mbuf *p, *q, *nq, *t; 837 struct ipq *fp = NULL; 838 struct ipqhead *head; 839 int i, hlen, next; 840 u_int8_t ecn, ecn0; 841 u_short hash; 842 843 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 844 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) { 845 IPSTAT_INC(ips_fragments); 846 IPSTAT_INC(ips_fragdropped); 847 m_freem(m); 848 return (NULL); 849 } 850 851 ip = mtod(m, struct ip *); 852 hlen = ip->ip_hl << 2; 853 854 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 855 head = &V_ipq[hash]; 856 IPQ_LOCK(); 857 858 /* 859 * Look for queue of fragments 860 * of this datagram. 861 */ 862 TAILQ_FOREACH(fp, head, ipq_list) 863 if (ip->ip_id == fp->ipq_id && 864 ip->ip_src.s_addr == fp->ipq_src.s_addr && 865 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 866 #ifdef MAC 867 mac_ipq_match(m, fp) && 868 #endif 869 ip->ip_p == fp->ipq_p) 870 goto found; 871 872 fp = NULL; 873 874 /* 875 * Attempt to trim the number of allocated fragment queues if it 876 * exceeds the administrative limit. 877 */ 878 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) { 879 /* 880 * drop something from the tail of the current queue 881 * before proceeding further 882 */ 883 struct ipq *q = TAILQ_LAST(head, ipqhead); 884 if (q == NULL) { /* gak */ 885 for (i = 0; i < IPREASS_NHASH; i++) { 886 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead); 887 if (r) { 888 IPSTAT_ADD(ips_fragtimeout, 889 r->ipq_nfrags); 890 ip_freef(&V_ipq[i], r); 891 break; 892 } 893 } 894 } else { 895 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags); 896 ip_freef(head, q); 897 } 898 } 899 900 found: 901 /* 902 * Adjust ip_len to not reflect header, 903 * convert offset of this to bytes. 904 */ 905 ip->ip_len -= hlen; 906 if (ip->ip_off & IP_MF) { 907 /* 908 * Make sure that fragments have a data length 909 * that's a non-zero multiple of 8 bytes. 910 */ 911 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 912 IPSTAT_INC(ips_toosmall); /* XXX */ 913 goto dropfrag; 914 } 915 m->m_flags |= M_FRAG; 916 } else 917 m->m_flags &= ~M_FRAG; 918 ip->ip_off <<= 3; 919 920 921 /* 922 * Attempt reassembly; if it succeeds, proceed. 923 * ip_reass() will return a different mbuf. 924 */ 925 IPSTAT_INC(ips_fragments); 926 m->m_pkthdr.header = ip; 927 928 /* Previous ip_reass() started here. */ 929 /* 930 * Presence of header sizes in mbufs 931 * would confuse code below. 932 */ 933 m->m_data += hlen; 934 m->m_len -= hlen; 935 936 /* 937 * If first fragment to arrive, create a reassembly queue. 938 */ 939 if (fp == NULL) { 940 fp = uma_zalloc(V_ipq_zone, M_NOWAIT); 941 if (fp == NULL) 942 goto dropfrag; 943 #ifdef MAC 944 if (mac_ipq_init(fp, M_NOWAIT) != 0) { 945 uma_zfree(V_ipq_zone, fp); 946 fp = NULL; 947 goto dropfrag; 948 } 949 mac_ipq_create(m, fp); 950 #endif 951 TAILQ_INSERT_HEAD(head, fp, ipq_list); 952 V_nipq++; 953 fp->ipq_nfrags = 1; 954 fp->ipq_ttl = IPFRAGTTL; 955 fp->ipq_p = ip->ip_p; 956 fp->ipq_id = ip->ip_id; 957 fp->ipq_src = ip->ip_src; 958 fp->ipq_dst = ip->ip_dst; 959 fp->ipq_frags = m; 960 m->m_nextpkt = NULL; 961 goto done; 962 } else { 963 fp->ipq_nfrags++; 964 #ifdef MAC 965 mac_ipq_update(m, fp); 966 #endif 967 } 968 969 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 970 971 /* 972 * Handle ECN by comparing this segment with the first one; 973 * if CE is set, do not lose CE. 974 * drop if CE and not-ECT are mixed for the same packet. 975 */ 976 ecn = ip->ip_tos & IPTOS_ECN_MASK; 977 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 978 if (ecn == IPTOS_ECN_CE) { 979 if (ecn0 == IPTOS_ECN_NOTECT) 980 goto dropfrag; 981 if (ecn0 != IPTOS_ECN_CE) 982 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 983 } 984 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 985 goto dropfrag; 986 987 /* 988 * Find a segment which begins after this one does. 989 */ 990 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 991 if (GETIP(q)->ip_off > ip->ip_off) 992 break; 993 994 /* 995 * If there is a preceding segment, it may provide some of 996 * our data already. If so, drop the data from the incoming 997 * segment. If it provides all of our data, drop us, otherwise 998 * stick new segment in the proper place. 999 * 1000 * If some of the data is dropped from the the preceding 1001 * segment, then it's checksum is invalidated. 1002 */ 1003 if (p) { 1004 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1005 if (i > 0) { 1006 if (i >= ip->ip_len) 1007 goto dropfrag; 1008 m_adj(m, i); 1009 m->m_pkthdr.csum_flags = 0; 1010 ip->ip_off += i; 1011 ip->ip_len -= i; 1012 } 1013 m->m_nextpkt = p->m_nextpkt; 1014 p->m_nextpkt = m; 1015 } else { 1016 m->m_nextpkt = fp->ipq_frags; 1017 fp->ipq_frags = m; 1018 } 1019 1020 /* 1021 * While we overlap succeeding segments trim them or, 1022 * if they are completely covered, dequeue them. 1023 */ 1024 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1025 q = nq) { 1026 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1027 if (i < GETIP(q)->ip_len) { 1028 GETIP(q)->ip_len -= i; 1029 GETIP(q)->ip_off += i; 1030 m_adj(q, i); 1031 q->m_pkthdr.csum_flags = 0; 1032 break; 1033 } 1034 nq = q->m_nextpkt; 1035 m->m_nextpkt = nq; 1036 IPSTAT_INC(ips_fragdropped); 1037 fp->ipq_nfrags--; 1038 m_freem(q); 1039 } 1040 1041 /* 1042 * Check for complete reassembly and perform frag per packet 1043 * limiting. 1044 * 1045 * Frag limiting is performed here so that the nth frag has 1046 * a chance to complete the packet before we drop the packet. 1047 * As a result, n+1 frags are actually allowed per packet, but 1048 * only n will ever be stored. (n = maxfragsperpacket.) 1049 * 1050 */ 1051 next = 0; 1052 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1053 if (GETIP(q)->ip_off != next) { 1054 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1055 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1056 ip_freef(head, fp); 1057 } 1058 goto done; 1059 } 1060 next += GETIP(q)->ip_len; 1061 } 1062 /* Make sure the last packet didn't have the IP_MF flag */ 1063 if (p->m_flags & M_FRAG) { 1064 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1065 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1066 ip_freef(head, fp); 1067 } 1068 goto done; 1069 } 1070 1071 /* 1072 * Reassembly is complete. Make sure the packet is a sane size. 1073 */ 1074 q = fp->ipq_frags; 1075 ip = GETIP(q); 1076 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1077 IPSTAT_INC(ips_toolong); 1078 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1079 ip_freef(head, fp); 1080 goto done; 1081 } 1082 1083 /* 1084 * Concatenate fragments. 1085 */ 1086 m = q; 1087 t = m->m_next; 1088 m->m_next = NULL; 1089 m_cat(m, t); 1090 nq = q->m_nextpkt; 1091 q->m_nextpkt = NULL; 1092 for (q = nq; q != NULL; q = nq) { 1093 nq = q->m_nextpkt; 1094 q->m_nextpkt = NULL; 1095 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1096 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1097 m_cat(m, q); 1098 } 1099 /* 1100 * In order to do checksumming faster we do 'end-around carry' here 1101 * (and not in for{} loop), though it implies we are not going to 1102 * reassemble more than 64k fragments. 1103 */ 1104 m->m_pkthdr.csum_data = 1105 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1106 #ifdef MAC 1107 mac_ipq_reassemble(fp, m); 1108 mac_ipq_destroy(fp); 1109 #endif 1110 1111 /* 1112 * Create header for new ip packet by modifying header of first 1113 * packet; dequeue and discard fragment reassembly header. 1114 * Make header visible. 1115 */ 1116 ip->ip_len = (ip->ip_hl << 2) + next; 1117 ip->ip_src = fp->ipq_src; 1118 ip->ip_dst = fp->ipq_dst; 1119 TAILQ_REMOVE(head, fp, ipq_list); 1120 V_nipq--; 1121 uma_zfree(V_ipq_zone, fp); 1122 m->m_len += (ip->ip_hl << 2); 1123 m->m_data -= (ip->ip_hl << 2); 1124 /* some debugging cruft by sklower, below, will go away soon */ 1125 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1126 m_fixhdr(m); 1127 IPSTAT_INC(ips_reassembled); 1128 IPQ_UNLOCK(); 1129 return (m); 1130 1131 dropfrag: 1132 IPSTAT_INC(ips_fragdropped); 1133 if (fp != NULL) 1134 fp->ipq_nfrags--; 1135 m_freem(m); 1136 done: 1137 IPQ_UNLOCK(); 1138 return (NULL); 1139 1140 #undef GETIP 1141 } 1142 1143 /* 1144 * Free a fragment reassembly header and all 1145 * associated datagrams. 1146 */ 1147 static void 1148 ip_freef(struct ipqhead *fhp, struct ipq *fp) 1149 { 1150 INIT_VNET_INET(curvnet); 1151 struct mbuf *q; 1152 1153 IPQ_LOCK_ASSERT(); 1154 1155 while (fp->ipq_frags) { 1156 q = fp->ipq_frags; 1157 fp->ipq_frags = q->m_nextpkt; 1158 m_freem(q); 1159 } 1160 TAILQ_REMOVE(fhp, fp, ipq_list); 1161 uma_zfree(V_ipq_zone, fp); 1162 V_nipq--; 1163 } 1164 1165 /* 1166 * IP timer processing; 1167 * if a timer expires on a reassembly 1168 * queue, discard it. 1169 */ 1170 void 1171 ip_slowtimo(void) 1172 { 1173 VNET_ITERATOR_DECL(vnet_iter); 1174 struct ipq *fp; 1175 int i; 1176 1177 IPQ_LOCK(); 1178 VNET_LIST_RLOCK(); 1179 VNET_FOREACH(vnet_iter) { 1180 CURVNET_SET(vnet_iter); 1181 INIT_VNET_INET(vnet_iter); 1182 for (i = 0; i < IPREASS_NHASH; i++) { 1183 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) { 1184 struct ipq *fpp; 1185 1186 fpp = fp; 1187 fp = TAILQ_NEXT(fp, ipq_list); 1188 if(--fpp->ipq_ttl == 0) { 1189 IPSTAT_ADD(ips_fragtimeout, 1190 fpp->ipq_nfrags); 1191 ip_freef(&V_ipq[i], fpp); 1192 } 1193 } 1194 } 1195 /* 1196 * If we are over the maximum number of fragments 1197 * (due to the limit being lowered), drain off 1198 * enough to get down to the new limit. 1199 */ 1200 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) { 1201 for (i = 0; i < IPREASS_NHASH; i++) { 1202 while (V_nipq > V_maxnipq && 1203 !TAILQ_EMPTY(&V_ipq[i])) { 1204 IPSTAT_ADD(ips_fragdropped, 1205 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1206 ip_freef(&V_ipq[i], 1207 TAILQ_FIRST(&V_ipq[i])); 1208 } 1209 } 1210 } 1211 CURVNET_RESTORE(); 1212 } 1213 VNET_LIST_RUNLOCK(); 1214 IPQ_UNLOCK(); 1215 } 1216 1217 /* 1218 * Drain off all datagram fragments. 1219 */ 1220 void 1221 ip_drain(void) 1222 { 1223 VNET_ITERATOR_DECL(vnet_iter); 1224 int i; 1225 1226 IPQ_LOCK(); 1227 VNET_LIST_RLOCK(); 1228 VNET_FOREACH(vnet_iter) { 1229 CURVNET_SET(vnet_iter); 1230 INIT_VNET_INET(vnet_iter); 1231 for (i = 0; i < IPREASS_NHASH; i++) { 1232 while(!TAILQ_EMPTY(&V_ipq[i])) { 1233 IPSTAT_ADD(ips_fragdropped, 1234 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1235 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); 1236 } 1237 } 1238 CURVNET_RESTORE(); 1239 } 1240 VNET_LIST_RUNLOCK(); 1241 IPQ_UNLOCK(); 1242 in_rtqdrain(); 1243 } 1244 1245 /* 1246 * The protocol to be inserted into ip_protox[] must be already registered 1247 * in inetsw[], either statically or through pf_proto_register(). 1248 */ 1249 int 1250 ipproto_register(u_char ipproto) 1251 { 1252 struct protosw *pr; 1253 1254 /* Sanity checks. */ 1255 if (ipproto == 0) 1256 return (EPROTONOSUPPORT); 1257 1258 /* 1259 * The protocol slot must not be occupied by another protocol 1260 * already. An index pointing to IPPROTO_RAW is unused. 1261 */ 1262 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1263 if (pr == NULL) 1264 return (EPFNOSUPPORT); 1265 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1266 return (EEXIST); 1267 1268 /* Find the protocol position in inetsw[] and set the index. */ 1269 for (pr = inetdomain.dom_protosw; 1270 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1271 if (pr->pr_domain->dom_family == PF_INET && 1272 pr->pr_protocol && pr->pr_protocol == ipproto) { 1273 /* Be careful to only index valid IP protocols. */ 1274 if (pr->pr_protocol < IPPROTO_MAX) { 1275 ip_protox[pr->pr_protocol] = pr - inetsw; 1276 return (0); 1277 } else 1278 return (EINVAL); 1279 } 1280 } 1281 return (EPROTONOSUPPORT); 1282 } 1283 1284 int 1285 ipproto_unregister(u_char ipproto) 1286 { 1287 struct protosw *pr; 1288 1289 /* Sanity checks. */ 1290 if (ipproto == 0) 1291 return (EPROTONOSUPPORT); 1292 1293 /* Check if the protocol was indeed registered. */ 1294 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1295 if (pr == NULL) 1296 return (EPFNOSUPPORT); 1297 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1298 return (ENOENT); 1299 1300 /* Reset the protocol slot to IPPROTO_RAW. */ 1301 ip_protox[ipproto] = pr - inetsw; 1302 return (0); 1303 } 1304 1305 /* 1306 * Given address of next destination (final or next hop), 1307 * return internet address info of interface to be used to get there. 1308 */ 1309 struct in_ifaddr * 1310 ip_rtaddr(struct in_addr dst, u_int fibnum) 1311 { 1312 struct route sro; 1313 struct sockaddr_in *sin; 1314 struct in_ifaddr *ifa; 1315 1316 bzero(&sro, sizeof(sro)); 1317 sin = (struct sockaddr_in *)&sro.ro_dst; 1318 sin->sin_family = AF_INET; 1319 sin->sin_len = sizeof(*sin); 1320 sin->sin_addr = dst; 1321 in_rtalloc_ign(&sro, 0, fibnum); 1322 1323 if (sro.ro_rt == NULL) 1324 return (NULL); 1325 1326 ifa = ifatoia(sro.ro_rt->rt_ifa); 1327 RTFREE(sro.ro_rt); 1328 return (ifa); 1329 } 1330 1331 u_char inetctlerrmap[PRC_NCMDS] = { 1332 0, 0, 0, 0, 1333 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1334 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1335 EMSGSIZE, EHOSTUNREACH, 0, 0, 1336 0, 0, EHOSTUNREACH, 0, 1337 ENOPROTOOPT, ECONNREFUSED 1338 }; 1339 1340 /* 1341 * Forward a packet. If some error occurs return the sender 1342 * an icmp packet. Note we can't always generate a meaningful 1343 * icmp message because icmp doesn't have a large enough repertoire 1344 * of codes and types. 1345 * 1346 * If not forwarding, just drop the packet. This could be confusing 1347 * if ipforwarding was zero but some routing protocol was advancing 1348 * us as a gateway to somewhere. However, we must let the routing 1349 * protocol deal with that. 1350 * 1351 * The srcrt parameter indicates whether the packet is being forwarded 1352 * via a source route. 1353 */ 1354 void 1355 ip_forward(struct mbuf *m, int srcrt) 1356 { 1357 INIT_VNET_INET(curvnet); 1358 struct ip *ip = mtod(m, struct ip *); 1359 struct in_ifaddr *ia = NULL; 1360 struct mbuf *mcopy; 1361 struct in_addr dest; 1362 struct route ro; 1363 int error, type = 0, code = 0, mtu = 0; 1364 1365 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1366 IPSTAT_INC(ips_cantforward); 1367 m_freem(m); 1368 return; 1369 } 1370 #ifdef IPSTEALTH 1371 if (!V_ipstealth) { 1372 #endif 1373 if (ip->ip_ttl <= IPTTLDEC) { 1374 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1375 0, 0); 1376 return; 1377 } 1378 #ifdef IPSTEALTH 1379 } 1380 #endif 1381 1382 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); 1383 if (!srcrt && ia == NULL) { 1384 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1385 return; 1386 } 1387 1388 /* 1389 * Save the IP header and at most 8 bytes of the payload, 1390 * in case we need to generate an ICMP message to the src. 1391 * 1392 * XXX this can be optimized a lot by saving the data in a local 1393 * buffer on the stack (72 bytes at most), and only allocating the 1394 * mbuf if really necessary. The vast majority of the packets 1395 * are forwarded without having to send an ICMP back (either 1396 * because unnecessary, or because rate limited), so we are 1397 * really we are wasting a lot of work here. 1398 * 1399 * We don't use m_copy() because it might return a reference 1400 * to a shared cluster. Both this function and ip_output() 1401 * assume exclusive access to the IP header in `m', so any 1402 * data in a cluster may change before we reach icmp_error(). 1403 */ 1404 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1405 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1406 /* 1407 * It's probably ok if the pkthdr dup fails (because 1408 * the deep copy of the tag chain failed), but for now 1409 * be conservative and just discard the copy since 1410 * code below may some day want the tags. 1411 */ 1412 m_free(mcopy); 1413 mcopy = NULL; 1414 } 1415 if (mcopy != NULL) { 1416 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy)); 1417 mcopy->m_pkthdr.len = mcopy->m_len; 1418 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1419 } 1420 1421 #ifdef IPSTEALTH 1422 if (!V_ipstealth) { 1423 #endif 1424 ip->ip_ttl -= IPTTLDEC; 1425 #ifdef IPSTEALTH 1426 } 1427 #endif 1428 1429 /* 1430 * If forwarding packet using same interface that it came in on, 1431 * perhaps should send a redirect to sender to shortcut a hop. 1432 * Only send redirect if source is sending directly to us, 1433 * and if packet was not source routed (or has any options). 1434 * Also, don't send redirect if forwarding using a default route 1435 * or a route modified by a redirect. 1436 */ 1437 dest.s_addr = 0; 1438 if (!srcrt && V_ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) { 1439 struct sockaddr_in *sin; 1440 struct rtentry *rt; 1441 1442 bzero(&ro, sizeof(ro)); 1443 sin = (struct sockaddr_in *)&ro.ro_dst; 1444 sin->sin_family = AF_INET; 1445 sin->sin_len = sizeof(*sin); 1446 sin->sin_addr = ip->ip_dst; 1447 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 1448 1449 rt = ro.ro_rt; 1450 1451 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1452 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1453 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1454 u_long src = ntohl(ip->ip_src.s_addr); 1455 1456 if (RTA(rt) && 1457 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1458 if (rt->rt_flags & RTF_GATEWAY) 1459 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1460 else 1461 dest.s_addr = ip->ip_dst.s_addr; 1462 /* Router requirements says to only send host redirects */ 1463 type = ICMP_REDIRECT; 1464 code = ICMP_REDIRECT_HOST; 1465 } 1466 } 1467 if (rt) 1468 RTFREE(rt); 1469 } 1470 1471 /* 1472 * Try to cache the route MTU from ip_output so we can consider it for 1473 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. 1474 */ 1475 bzero(&ro, sizeof(ro)); 1476 1477 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1478 1479 if (error == EMSGSIZE && ro.ro_rt) 1480 mtu = ro.ro_rt->rt_rmx.rmx_mtu; 1481 if (ro.ro_rt) 1482 RTFREE(ro.ro_rt); 1483 1484 if (error) 1485 IPSTAT_INC(ips_cantforward); 1486 else { 1487 IPSTAT_INC(ips_forward); 1488 if (type) 1489 IPSTAT_INC(ips_redirectsent); 1490 else { 1491 if (mcopy) 1492 m_freem(mcopy); 1493 return; 1494 } 1495 } 1496 if (mcopy == NULL) 1497 return; 1498 1499 switch (error) { 1500 1501 case 0: /* forwarded, but need redirect */ 1502 /* type, code set above */ 1503 break; 1504 1505 case ENETUNREACH: /* shouldn't happen, checked above */ 1506 case EHOSTUNREACH: 1507 case ENETDOWN: 1508 case EHOSTDOWN: 1509 default: 1510 type = ICMP_UNREACH; 1511 code = ICMP_UNREACH_HOST; 1512 break; 1513 1514 case EMSGSIZE: 1515 type = ICMP_UNREACH; 1516 code = ICMP_UNREACH_NEEDFRAG; 1517 1518 #ifdef IPSEC 1519 /* 1520 * If IPsec is configured for this path, 1521 * override any possibly mtu value set by ip_output. 1522 */ 1523 mtu = ip_ipsec_mtu(m, mtu); 1524 #endif /* IPSEC */ 1525 /* 1526 * If the MTU was set before make sure we are below the 1527 * interface MTU. 1528 * If the MTU wasn't set before use the interface mtu or 1529 * fall back to the next smaller mtu step compared to the 1530 * current packet size. 1531 */ 1532 if (mtu != 0) { 1533 if (ia != NULL) 1534 mtu = min(mtu, ia->ia_ifp->if_mtu); 1535 } else { 1536 if (ia != NULL) 1537 mtu = ia->ia_ifp->if_mtu; 1538 else 1539 mtu = ip_next_mtu(ip->ip_len, 0); 1540 } 1541 IPSTAT_INC(ips_cantfrag); 1542 break; 1543 1544 case ENOBUFS: 1545 /* 1546 * A router should not generate ICMP_SOURCEQUENCH as 1547 * required in RFC1812 Requirements for IP Version 4 Routers. 1548 * Source quench could be a big problem under DoS attacks, 1549 * or if the underlying interface is rate-limited. 1550 * Those who need source quench packets may re-enable them 1551 * via the net.inet.ip.sendsourcequench sysctl. 1552 */ 1553 if (V_ip_sendsourcequench == 0) { 1554 m_freem(mcopy); 1555 return; 1556 } else { 1557 type = ICMP_SOURCEQUENCH; 1558 code = 0; 1559 } 1560 break; 1561 1562 case EACCES: /* ipfw denied packet */ 1563 m_freem(mcopy); 1564 return; 1565 } 1566 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1567 } 1568 1569 void 1570 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1571 struct mbuf *m) 1572 { 1573 INIT_VNET_NET(inp->inp_vnet); 1574 1575 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1576 struct bintime bt; 1577 1578 bintime(&bt); 1579 if (inp->inp_socket->so_options & SO_BINTIME) { 1580 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1581 SCM_BINTIME, SOL_SOCKET); 1582 if (*mp) 1583 mp = &(*mp)->m_next; 1584 } 1585 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1586 struct timeval tv; 1587 1588 bintime2timeval(&bt, &tv); 1589 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1590 SCM_TIMESTAMP, SOL_SOCKET); 1591 if (*mp) 1592 mp = &(*mp)->m_next; 1593 } 1594 } 1595 if (inp->inp_flags & INP_RECVDSTADDR) { 1596 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1597 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1598 if (*mp) 1599 mp = &(*mp)->m_next; 1600 } 1601 if (inp->inp_flags & INP_RECVTTL) { 1602 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1603 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1604 if (*mp) 1605 mp = &(*mp)->m_next; 1606 } 1607 #ifdef notyet 1608 /* XXX 1609 * Moving these out of udp_input() made them even more broken 1610 * than they already were. 1611 */ 1612 /* options were tossed already */ 1613 if (inp->inp_flags & INP_RECVOPTS) { 1614 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1615 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1616 if (*mp) 1617 mp = &(*mp)->m_next; 1618 } 1619 /* ip_srcroute doesn't do what we want here, need to fix */ 1620 if (inp->inp_flags & INP_RECVRETOPTS) { 1621 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1622 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1623 if (*mp) 1624 mp = &(*mp)->m_next; 1625 } 1626 #endif 1627 if (inp->inp_flags & INP_RECVIF) { 1628 struct ifnet *ifp; 1629 struct sdlbuf { 1630 struct sockaddr_dl sdl; 1631 u_char pad[32]; 1632 } sdlbuf; 1633 struct sockaddr_dl *sdp; 1634 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1635 1636 if (((ifp = m->m_pkthdr.rcvif)) 1637 && ( ifp->if_index && (ifp->if_index <= V_if_index))) { 1638 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1639 /* 1640 * Change our mind and don't try copy. 1641 */ 1642 if ((sdp->sdl_family != AF_LINK) 1643 || (sdp->sdl_len > sizeof(sdlbuf))) { 1644 goto makedummy; 1645 } 1646 bcopy(sdp, sdl2, sdp->sdl_len); 1647 } else { 1648 makedummy: 1649 sdl2->sdl_len 1650 = offsetof(struct sockaddr_dl, sdl_data[0]); 1651 sdl2->sdl_family = AF_LINK; 1652 sdl2->sdl_index = 0; 1653 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1654 } 1655 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1656 IP_RECVIF, IPPROTO_IP); 1657 if (*mp) 1658 mp = &(*mp)->m_next; 1659 } 1660 } 1661 1662 /* 1663 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1664 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1665 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1666 * compiled. 1667 */ 1668 int 1669 ip_rsvp_init(struct socket *so) 1670 { 1671 INIT_VNET_INET(so->so_vnet); 1672 1673 if (so->so_type != SOCK_RAW || 1674 so->so_proto->pr_protocol != IPPROTO_RSVP) 1675 return EOPNOTSUPP; 1676 1677 if (V_ip_rsvpd != NULL) 1678 return EADDRINUSE; 1679 1680 V_ip_rsvpd = so; 1681 /* 1682 * This may seem silly, but we need to be sure we don't over-increment 1683 * the RSVP counter, in case something slips up. 1684 */ 1685 if (!V_ip_rsvp_on) { 1686 V_ip_rsvp_on = 1; 1687 V_rsvp_on++; 1688 } 1689 1690 return 0; 1691 } 1692 1693 int 1694 ip_rsvp_done(void) 1695 { 1696 INIT_VNET_INET(curvnet); 1697 1698 V_ip_rsvpd = NULL; 1699 /* 1700 * This may seem silly, but we need to be sure we don't over-decrement 1701 * the RSVP counter, in case something slips up. 1702 */ 1703 if (V_ip_rsvp_on) { 1704 V_ip_rsvp_on = 0; 1705 V_rsvp_on--; 1706 } 1707 return 0; 1708 } 1709 1710 void 1711 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1712 { 1713 INIT_VNET_INET(curvnet); 1714 1715 if (rsvp_input_p) { /* call the real one if loaded */ 1716 rsvp_input_p(m, off); 1717 return; 1718 } 1719 1720 /* Can still get packets with rsvp_on = 0 if there is a local member 1721 * of the group to which the RSVP packet is addressed. But in this 1722 * case we want to throw the packet away. 1723 */ 1724 1725 if (!V_rsvp_on) { 1726 m_freem(m); 1727 return; 1728 } 1729 1730 if (V_ip_rsvpd != NULL) { 1731 rip_input(m, off); 1732 return; 1733 } 1734 /* Drop the packet */ 1735 m_freem(m); 1736 } 1737