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_carp.h" 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/callout.h> 45 #include <sys/mbuf.h> 46 #include <sys/malloc.h> 47 #include <sys/domain.h> 48 #include <sys/protosw.h> 49 #include <sys/socket.h> 50 #include <sys/time.h> 51 #include <sys/kernel.h> 52 #include <sys/lock.h> 53 #include <sys/rwlock.h> 54 #include <sys/syslog.h> 55 #include <sys/sysctl.h> 56 #include <sys/vimage.h> 57 58 #include <net/pfil.h> 59 #include <net/if.h> 60 #include <net/if_types.h> 61 #include <net/if_var.h> 62 #include <net/if_dl.h> 63 #include <net/route.h> 64 #include <net/netisr.h> 65 #include <net/vnet.h> 66 #include <net/flowtable.h> 67 68 #include <netinet/in.h> 69 #include <netinet/in_systm.h> 70 #include <netinet/in_var.h> 71 #include <netinet/ip.h> 72 #include <netinet/in_pcb.h> 73 #include <netinet/ip_var.h> 74 #include <netinet/ip_icmp.h> 75 #include <netinet/ip_options.h> 76 #include <machine/in_cksum.h> 77 #include <netinet/vinet.h> 78 #ifdef DEV_CARP 79 #include <netinet/ip_carp.h> 80 #endif 81 #ifdef IPSEC 82 #include <netinet/ip_ipsec.h> 83 #endif /* IPSEC */ 84 85 #include <sys/socketvar.h> 86 87 #include <security/mac/mac_framework.h> 88 89 #ifdef CTASSERT 90 CTASSERT(sizeof(struct ip) == 20); 91 #endif 92 93 #ifndef VIMAGE 94 #ifndef VIMAGE_GLOBALS 95 struct vnet_inet vnet_inet_0; 96 #endif 97 #endif 98 99 #ifdef VIMAGE_GLOBALS 100 static int ipsendredirects; 101 static int ip_checkinterface; 102 static int ip_keepfaith; 103 static int ip_sendsourcequench; 104 int ip_defttl; 105 int ip_do_randomid; 106 int ipforwarding; 107 struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 108 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 109 u_long in_ifaddrhmask; /* mask for hash table */ 110 struct ipstat ipstat; 111 static int ip_rsvp_on; 112 struct socket *ip_rsvpd; 113 int rsvp_on; 114 static struct ipqhead ipq[IPREASS_NHASH]; 115 static int maxnipq; /* Administrative limit on # reass queues. */ 116 static int maxfragsperpacket; 117 int ipstealth; 118 static int nipq; /* Total # of reass queues */ 119 #endif 120 121 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_FORWARDING, 122 forwarding, CTLFLAG_RW, ipforwarding, 0, 123 "Enable IP forwarding between interfaces"); 124 125 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_SENDREDIRECTS, 126 redirect, CTLFLAG_RW, ipsendredirects, 0, 127 "Enable sending IP redirects"); 128 129 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_DEFTTL, 130 ttl, CTLFLAG_RW, ip_defttl, 0, "Maximum TTL on IP packets"); 131 132 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_KEEPFAITH, 133 keepfaith, CTLFLAG_RW, ip_keepfaith, 0, 134 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 135 136 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, 137 sendsourcequench, CTLFLAG_RW, ip_sendsourcequench, 0, 138 "Enable the transmission of source quench packets"); 139 140 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, random_id, 141 CTLFLAG_RW, ip_do_randomid, 0, "Assign random ip_id values"); 142 143 /* 144 * XXX - Setting ip_checkinterface mostly implements the receive side of 145 * the Strong ES model described in RFC 1122, but since the routing table 146 * and transmit implementation do not implement the Strong ES model, 147 * setting this to 1 results in an odd hybrid. 148 * 149 * XXX - ip_checkinterface currently must be disabled if you use ipnat 150 * to translate the destination address to another local interface. 151 * 152 * XXX - ip_checkinterface must be disabled if you add IP aliases 153 * to the loopback interface instead of the interface where the 154 * packets for those addresses are received. 155 */ 156 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, 157 check_interface, CTLFLAG_RW, ip_checkinterface, 0, 158 "Verify packet arrives on correct interface"); 159 160 struct pfil_head inet_pfil_hook; /* Packet filter hooks */ 161 162 static struct netisr_handler ip_nh = { 163 .nh_name = "ip", 164 .nh_handler = ip_input, 165 .nh_proto = NETISR_IP, 166 .nh_policy = NETISR_POLICY_FLOW, 167 }; 168 169 extern struct domain inetdomain; 170 extern struct protosw inetsw[]; 171 u_char ip_protox[IPPROTO_MAX]; 172 173 174 SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 175 ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 176 177 #ifdef VIMAGE_GLOBALS 178 static uma_zone_t ipq_zone; 179 #endif 180 static struct mtx ipqlock; 181 182 #define IPQ_LOCK() mtx_lock(&ipqlock) 183 #define IPQ_UNLOCK() mtx_unlock(&ipqlock) 184 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 185 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 186 187 static void maxnipq_update(void); 188 static void ipq_zone_change(void *); 189 190 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, fragpackets, 191 CTLFLAG_RD, nipq, 0, 192 "Current number of IPv4 fragment reassembly queue entries"); 193 194 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, maxfragsperpacket, 195 CTLFLAG_RW, maxfragsperpacket, 0, 196 "Maximum number of IPv4 fragments allowed per packet"); 197 198 struct callout ipport_tick_callout; 199 200 #ifdef IPCTL_DEFMTU 201 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 202 &ip_mtu, 0, "Default MTU"); 203 #endif 204 205 #ifdef IPSTEALTH 206 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 207 ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding"); 208 #endif 209 static int ip_output_flowtable_size = 2048; 210 TUNABLE_INT("net.inet.ip.output_flowtable_size", &ip_output_flowtable_size); 211 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, output_flowtable_size, 212 CTLFLAG_RDTUN, ip_output_flowtable_size, 2048, 213 "number of entries in the per-cpu output flow caches"); 214 215 #ifdef VIMAGE_GLOBALS 216 int fw_one_pass; 217 #endif 218 struct flowtable *ip_ft; 219 220 static void ip_freef(struct ipqhead *, struct ipq *); 221 222 #ifndef VIMAGE_GLOBALS 223 static void vnet_inet_register(void); 224 225 static const vnet_modinfo_t vnet_inet_modinfo = { 226 .vmi_id = VNET_MOD_INET, 227 .vmi_name = "inet", 228 .vmi_size = sizeof(struct vnet_inet) 229 }; 230 231 static void vnet_inet_register() 232 { 233 234 vnet_mod_register(&vnet_inet_modinfo); 235 } 236 237 SYSINIT(inet, SI_SUB_PROTO_BEGIN, SI_ORDER_FIRST, vnet_inet_register, 0); 238 #endif 239 240 static int 241 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 242 { 243 int error, qlimit; 244 245 netisr_getqlimit(&ip_nh, &qlimit); 246 error = sysctl_handle_int(oidp, &qlimit, 0, req); 247 if (error || !req->newptr) 248 return (error); 249 if (qlimit < 1) 250 return (EINVAL); 251 return (netisr_setqlimit(&ip_nh, qlimit)); 252 } 253 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, 254 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I", 255 "Maximum size of the IP input queue"); 256 257 static int 258 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) 259 { 260 u_int64_t qdrops_long; 261 int error, qdrops; 262 263 netisr_getqdrops(&ip_nh, &qdrops_long); 264 qdrops = qdrops_long; 265 error = sysctl_handle_int(oidp, &qdrops, 0, req); 266 if (error || !req->newptr) 267 return (error); 268 if (qdrops != 0) 269 return (EINVAL); 270 netisr_clearqdrops(&ip_nh); 271 return (0); 272 } 273 274 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, 275 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I", 276 "Number of packets dropped from the IP input queue"); 277 278 /* 279 * IP initialization: fill in IP protocol switch table. 280 * All protocols not implemented in kernel go to raw IP protocol handler. 281 */ 282 void 283 ip_init(void) 284 { 285 INIT_VNET_INET(curvnet); 286 struct protosw *pr; 287 int i; 288 289 V_ipsendredirects = 1; /* XXX */ 290 V_ip_checkinterface = 0; 291 V_ip_keepfaith = 0; 292 V_ip_sendsourcequench = 0; 293 V_rsvp_on = 0; 294 V_ip_defttl = IPDEFTTL; 295 V_ip_do_randomid = 0; 296 V_ip_id = time_second & 0xffff; 297 V_ipforwarding = 0; 298 V_ipstealth = 0; 299 V_nipq = 0; /* Total # of reass queues */ 300 301 V_ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */ 302 V_ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */ 303 V_ipport_firstauto = IPPORT_EPHEMERALFIRST; /* 10000 */ 304 V_ipport_lastauto = IPPORT_EPHEMERALLAST; /* 65535 */ 305 V_ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */ 306 V_ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */ 307 V_ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */ 308 V_ipport_reservedlow = 0; 309 V_ipport_randomized = 1; /* user controlled via sysctl */ 310 V_ipport_randomcps = 10; /* user controlled via sysctl */ 311 V_ipport_randomtime = 45; /* user controlled via sysctl */ 312 V_ipport_stoprandom = 0; /* toggled by ipport_tick */ 313 314 V_fw_one_pass = 1; 315 316 #ifdef NOTYET 317 /* XXX global static but not instantiated in this file */ 318 V_ipfastforward_active = 0; 319 V_subnetsarelocal = 0; 320 V_sameprefixcarponly = 0; 321 #endif 322 323 TAILQ_INIT(&V_in_ifaddrhead); 324 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 325 326 /* Initialize IP reassembly queue. */ 327 for (i = 0; i < IPREASS_NHASH; i++) 328 TAILQ_INIT(&V_ipq[i]); 329 V_maxnipq = nmbclusters / 32; 330 V_maxfragsperpacket = 16; 331 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 332 NULL, UMA_ALIGN_PTR, 0); 333 maxnipq_update(); 334 335 /* Skip initialization of globals for non-default instances. */ 336 if (!IS_DEFAULT_VNET(curvnet)) 337 return; 338 339 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 340 if (pr == NULL) 341 panic("ip_init: PF_INET not found"); 342 343 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 344 for (i = 0; i < IPPROTO_MAX; i++) 345 ip_protox[i] = pr - inetsw; 346 /* 347 * Cycle through IP protocols and put them into the appropriate place 348 * in ip_protox[]. 349 */ 350 for (pr = inetdomain.dom_protosw; 351 pr < inetdomain.dom_protoswNPROTOSW; pr++) 352 if (pr->pr_domain->dom_family == PF_INET && 353 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 354 /* Be careful to only index valid IP protocols. */ 355 if (pr->pr_protocol < IPPROTO_MAX) 356 ip_protox[pr->pr_protocol] = pr - inetsw; 357 } 358 359 /* Initialize packet filter hooks. */ 360 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 361 inet_pfil_hook.ph_af = AF_INET; 362 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 363 printf("%s: WARNING: unable to register pfil hook, " 364 "error %d\n", __func__, i); 365 366 /* Start ipport_tick. */ 367 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE); 368 callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL); 369 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, 370 SHUTDOWN_PRI_DEFAULT); 371 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, 372 NULL, EVENTHANDLER_PRI_ANY); 373 374 /* Initialize various other remaining things. */ 375 IPQ_LOCK_INIT(); 376 netisr_register(&ip_nh); 377 ip_ft = flowtable_alloc(ip_output_flowtable_size, FL_PCPU); 378 } 379 380 void 381 ip_fini(void *xtp) 382 { 383 384 callout_stop(&ipport_tick_callout); 385 } 386 387 /* 388 * Ip input routine. Checksum and byte swap header. If fragmented 389 * try to reassemble. Process options. Pass to next level. 390 */ 391 void 392 ip_input(struct mbuf *m) 393 { 394 INIT_VNET_INET(curvnet); 395 struct ip *ip = NULL; 396 struct in_ifaddr *ia = NULL; 397 struct ifaddr *ifa; 398 struct ifnet *ifp; 399 int checkif, hlen = 0; 400 u_short sum; 401 int dchg = 0; /* dest changed after fw */ 402 struct in_addr odst; /* original dst address */ 403 404 M_ASSERTPKTHDR(m); 405 406 if (m->m_flags & M_FASTFWD_OURS) { 407 /* 408 * Firewall or NAT changed destination to local. 409 * We expect ip_len and ip_off to be in host byte order. 410 */ 411 m->m_flags &= ~M_FASTFWD_OURS; 412 /* Set up some basics that will be used later. */ 413 ip = mtod(m, struct ip *); 414 hlen = ip->ip_hl << 2; 415 goto ours; 416 } 417 418 IPSTAT_INC(ips_total); 419 420 if (m->m_pkthdr.len < sizeof(struct ip)) 421 goto tooshort; 422 423 if (m->m_len < sizeof (struct ip) && 424 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 425 IPSTAT_INC(ips_toosmall); 426 return; 427 } 428 ip = mtod(m, struct ip *); 429 430 if (ip->ip_v != IPVERSION) { 431 IPSTAT_INC(ips_badvers); 432 goto bad; 433 } 434 435 hlen = ip->ip_hl << 2; 436 if (hlen < sizeof(struct ip)) { /* minimum header length */ 437 IPSTAT_INC(ips_badhlen); 438 goto bad; 439 } 440 if (hlen > m->m_len) { 441 if ((m = m_pullup(m, hlen)) == NULL) { 442 IPSTAT_INC(ips_badhlen); 443 return; 444 } 445 ip = mtod(m, struct ip *); 446 } 447 448 /* 127/8 must not appear on wire - RFC1122 */ 449 ifp = m->m_pkthdr.rcvif; 450 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 451 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 452 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 453 IPSTAT_INC(ips_badaddr); 454 goto bad; 455 } 456 } 457 458 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 459 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 460 } else { 461 if (hlen == sizeof(struct ip)) { 462 sum = in_cksum_hdr(ip); 463 } else { 464 sum = in_cksum(m, hlen); 465 } 466 } 467 if (sum) { 468 IPSTAT_INC(ips_badsum); 469 goto bad; 470 } 471 472 #ifdef ALTQ 473 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 474 /* packet is dropped by traffic conditioner */ 475 return; 476 #endif 477 478 /* 479 * Convert fields to host representation. 480 */ 481 ip->ip_len = ntohs(ip->ip_len); 482 if (ip->ip_len < hlen) { 483 IPSTAT_INC(ips_badlen); 484 goto bad; 485 } 486 ip->ip_off = ntohs(ip->ip_off); 487 488 /* 489 * Check that the amount of data in the buffers 490 * is as at least much as the IP header would have us expect. 491 * Trim mbufs if longer than we expect. 492 * Drop packet if shorter than we expect. 493 */ 494 if (m->m_pkthdr.len < ip->ip_len) { 495 tooshort: 496 IPSTAT_INC(ips_tooshort); 497 goto bad; 498 } 499 if (m->m_pkthdr.len > ip->ip_len) { 500 if (m->m_len == m->m_pkthdr.len) { 501 m->m_len = ip->ip_len; 502 m->m_pkthdr.len = ip->ip_len; 503 } else 504 m_adj(m, ip->ip_len - m->m_pkthdr.len); 505 } 506 #ifdef IPSEC 507 /* 508 * Bypass packet filtering for packets from a tunnel (gif). 509 */ 510 if (ip_ipsec_filtertunnel(m)) 511 goto passin; 512 #endif /* IPSEC */ 513 514 /* 515 * Run through list of hooks for input packets. 516 * 517 * NB: Beware of the destination address changing (e.g. 518 * by NAT rewriting). When this happens, tell 519 * ip_forward to do the right thing. 520 */ 521 522 /* Jump over all PFIL processing if hooks are not active. */ 523 if (!PFIL_HOOKED(&inet_pfil_hook)) 524 goto passin; 525 526 odst = ip->ip_dst; 527 if (pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0) 528 return; 529 if (m == NULL) /* consumed by filter */ 530 return; 531 532 ip = mtod(m, struct ip *); 533 dchg = (odst.s_addr != ip->ip_dst.s_addr); 534 ifp = m->m_pkthdr.rcvif; 535 536 #ifdef IPFIREWALL_FORWARD 537 if (m->m_flags & M_FASTFWD_OURS) { 538 m->m_flags &= ~M_FASTFWD_OURS; 539 goto ours; 540 } 541 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 542 /* 543 * Directly ship on the packet. This allows to forward packets 544 * that were destined for us to some other directly connected 545 * host. 546 */ 547 ip_forward(m, dchg); 548 return; 549 } 550 #endif /* IPFIREWALL_FORWARD */ 551 552 passin: 553 /* 554 * Process options and, if not destined for us, 555 * ship it on. ip_dooptions returns 1 when an 556 * error was detected (causing an icmp message 557 * to be sent and the original packet to be freed). 558 */ 559 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 560 return; 561 562 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 563 * matter if it is destined to another node, or whether it is 564 * a multicast one, RSVP wants it! and prevents it from being forwarded 565 * anywhere else. Also checks if the rsvp daemon is running before 566 * grabbing the packet. 567 */ 568 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 569 goto ours; 570 571 /* 572 * Check our list of addresses, to see if the packet is for us. 573 * If we don't have any addresses, assume any unicast packet 574 * we receive might be for us (and let the upper layers deal 575 * with it). 576 */ 577 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 578 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 579 goto ours; 580 581 /* 582 * Enable a consistency check between the destination address 583 * and the arrival interface for a unicast packet (the RFC 1122 584 * strong ES model) if IP forwarding is disabled and the packet 585 * is not locally generated and the packet is not subject to 586 * 'ipfw fwd'. 587 * 588 * XXX - Checking also should be disabled if the destination 589 * address is ipnat'ed to a different interface. 590 * 591 * XXX - Checking is incompatible with IP aliases added 592 * to the loopback interface instead of the interface where 593 * the packets are received. 594 * 595 * XXX - This is the case for carp vhost IPs as well so we 596 * insert a workaround. If the packet got here, we already 597 * checked with carp_iamatch() and carp_forus(). 598 */ 599 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 600 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && 601 #ifdef DEV_CARP 602 !ifp->if_carp && 603 #endif 604 (dchg == 0); 605 606 /* 607 * Check for exact addresses in the hash bucket. 608 */ 609 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 610 /* 611 * If the address matches, verify that the packet 612 * arrived via the correct interface if checking is 613 * enabled. 614 */ 615 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 616 (!checkif || ia->ia_ifp == ifp)) 617 goto ours; 618 } 619 /* 620 * Check for broadcast addresses. 621 * 622 * Only accept broadcast packets that arrive via the matching 623 * interface. Reception of forwarded directed broadcasts would 624 * be handled via ip_forward() and ether_output() with the loopback 625 * into the stack for SIMPLEX interfaces handled by ether_output(). 626 */ 627 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 628 IF_ADDR_LOCK(ifp); 629 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 630 if (ifa->ifa_addr->sa_family != AF_INET) 631 continue; 632 ia = ifatoia(ifa); 633 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 634 ip->ip_dst.s_addr) { 635 IF_ADDR_UNLOCK(ifp); 636 goto ours; 637 } 638 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) { 639 IF_ADDR_UNLOCK(ifp); 640 goto ours; 641 } 642 #ifdef BOOTP_COMPAT 643 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { 644 IF_ADDR_UNLOCK(ifp); 645 goto ours; 646 } 647 #endif 648 } 649 IF_ADDR_UNLOCK(ifp); 650 } 651 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 652 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 653 IPSTAT_INC(ips_cantforward); 654 m_freem(m); 655 return; 656 } 657 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 658 if (V_ip_mrouter) { 659 /* 660 * If we are acting as a multicast router, all 661 * incoming multicast packets are passed to the 662 * kernel-level multicast forwarding function. 663 * The packet is returned (relatively) intact; if 664 * ip_mforward() returns a non-zero value, the packet 665 * must be discarded, else it may be accepted below. 666 */ 667 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 668 IPSTAT_INC(ips_cantforward); 669 m_freem(m); 670 return; 671 } 672 673 /* 674 * The process-level routing daemon needs to receive 675 * all multicast IGMP packets, whether or not this 676 * host belongs to their destination groups. 677 */ 678 if (ip->ip_p == IPPROTO_IGMP) 679 goto ours; 680 IPSTAT_INC(ips_forward); 681 } 682 /* 683 * Assume the packet is for us, to avoid prematurely taking 684 * a lock on the in_multi hash. Protocols must perform 685 * their own filtering and update statistics accordingly. 686 */ 687 goto ours; 688 } 689 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 690 goto ours; 691 if (ip->ip_dst.s_addr == INADDR_ANY) 692 goto ours; 693 694 /* 695 * FAITH(Firewall Aided Internet Translator) 696 */ 697 if (ifp && ifp->if_type == IFT_FAITH) { 698 if (V_ip_keepfaith) { 699 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 700 goto ours; 701 } 702 m_freem(m); 703 return; 704 } 705 706 /* 707 * Not for us; forward if possible and desirable. 708 */ 709 if (V_ipforwarding == 0) { 710 IPSTAT_INC(ips_cantforward); 711 m_freem(m); 712 } else { 713 #ifdef IPSEC 714 if (ip_ipsec_fwd(m)) 715 goto bad; 716 #endif /* IPSEC */ 717 ip_forward(m, dchg); 718 } 719 return; 720 721 ours: 722 #ifdef IPSTEALTH 723 /* 724 * IPSTEALTH: Process non-routing options only 725 * if the packet is destined for us. 726 */ 727 if (V_ipstealth && hlen > sizeof (struct ip) && 728 ip_dooptions(m, 1)) 729 return; 730 #endif /* IPSTEALTH */ 731 732 /* Count the packet in the ip address stats */ 733 if (ia != NULL) { 734 ia->ia_ifa.if_ipackets++; 735 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 736 } 737 738 /* 739 * Attempt reassembly; if it succeeds, proceed. 740 * ip_reass() will return a different mbuf. 741 */ 742 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 743 m = ip_reass(m); 744 if (m == NULL) 745 return; 746 ip = mtod(m, struct ip *); 747 /* Get the header length of the reassembled packet */ 748 hlen = ip->ip_hl << 2; 749 } 750 751 /* 752 * Further protocols expect the packet length to be w/o the 753 * IP header. 754 */ 755 ip->ip_len -= hlen; 756 757 #ifdef IPSEC 758 /* 759 * enforce IPsec policy checking if we are seeing last header. 760 * note that we do not visit this with protocols with pcb layer 761 * code - like udp/tcp/raw ip. 762 */ 763 if (ip_ipsec_input(m)) 764 goto bad; 765 #endif /* IPSEC */ 766 767 /* 768 * Switch out to protocol's input routine. 769 */ 770 IPSTAT_INC(ips_delivered); 771 772 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 773 return; 774 bad: 775 m_freem(m); 776 } 777 778 /* 779 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 780 * max has slightly different semantics than the sysctl, for historical 781 * reasons. 782 */ 783 static void 784 maxnipq_update(void) 785 { 786 INIT_VNET_INET(curvnet); 787 788 /* 789 * -1 for unlimited allocation. 790 */ 791 if (V_maxnipq < 0) 792 uma_zone_set_max(V_ipq_zone, 0); 793 /* 794 * Positive number for specific bound. 795 */ 796 if (V_maxnipq > 0) 797 uma_zone_set_max(V_ipq_zone, V_maxnipq); 798 /* 799 * Zero specifies no further fragment queue allocation -- set the 800 * bound very low, but rely on implementation elsewhere to actually 801 * prevent allocation and reclaim current queues. 802 */ 803 if (V_maxnipq == 0) 804 uma_zone_set_max(V_ipq_zone, 1); 805 } 806 807 static void 808 ipq_zone_change(void *tag) 809 { 810 INIT_VNET_INET(curvnet); 811 812 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) { 813 V_maxnipq = nmbclusters / 32; 814 maxnipq_update(); 815 } 816 } 817 818 static int 819 sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 820 { 821 INIT_VNET_INET(curvnet); 822 int error, i; 823 824 i = V_maxnipq; 825 error = sysctl_handle_int(oidp, &i, 0, req); 826 if (error || !req->newptr) 827 return (error); 828 829 /* 830 * XXXRW: Might be a good idea to sanity check the argument and place 831 * an extreme upper bound. 832 */ 833 if (i < -1) 834 return (EINVAL); 835 V_maxnipq = i; 836 maxnipq_update(); 837 return (0); 838 } 839 840 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 841 NULL, 0, sysctl_maxnipq, "I", 842 "Maximum number of IPv4 fragment reassembly queue entries"); 843 844 /* 845 * Take incoming datagram fragment and try to reassemble it into 846 * whole datagram. If the argument is the first fragment or one 847 * in between the function will return NULL and store the mbuf 848 * in the fragment chain. If the argument is the last fragment 849 * the packet will be reassembled and the pointer to the new 850 * mbuf returned for further processing. Only m_tags attached 851 * to the first packet/fragment are preserved. 852 * The IP header is *NOT* adjusted out of iplen. 853 */ 854 struct mbuf * 855 ip_reass(struct mbuf *m) 856 { 857 INIT_VNET_INET(curvnet); 858 struct ip *ip; 859 struct mbuf *p, *q, *nq, *t; 860 struct ipq *fp = NULL; 861 struct ipqhead *head; 862 int i, hlen, next; 863 u_int8_t ecn, ecn0; 864 u_short hash; 865 866 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 867 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) { 868 IPSTAT_INC(ips_fragments); 869 IPSTAT_INC(ips_fragdropped); 870 m_freem(m); 871 return (NULL); 872 } 873 874 ip = mtod(m, struct ip *); 875 hlen = ip->ip_hl << 2; 876 877 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 878 head = &V_ipq[hash]; 879 IPQ_LOCK(); 880 881 /* 882 * Look for queue of fragments 883 * of this datagram. 884 */ 885 TAILQ_FOREACH(fp, head, ipq_list) 886 if (ip->ip_id == fp->ipq_id && 887 ip->ip_src.s_addr == fp->ipq_src.s_addr && 888 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 889 #ifdef MAC 890 mac_ipq_match(m, fp) && 891 #endif 892 ip->ip_p == fp->ipq_p) 893 goto found; 894 895 fp = NULL; 896 897 /* 898 * Attempt to trim the number of allocated fragment queues if it 899 * exceeds the administrative limit. 900 */ 901 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) { 902 /* 903 * drop something from the tail of the current queue 904 * before proceeding further 905 */ 906 struct ipq *q = TAILQ_LAST(head, ipqhead); 907 if (q == NULL) { /* gak */ 908 for (i = 0; i < IPREASS_NHASH; i++) { 909 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead); 910 if (r) { 911 IPSTAT_ADD(ips_fragtimeout, 912 r->ipq_nfrags); 913 ip_freef(&V_ipq[i], r); 914 break; 915 } 916 } 917 } else { 918 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags); 919 ip_freef(head, q); 920 } 921 } 922 923 found: 924 /* 925 * Adjust ip_len to not reflect header, 926 * convert offset of this to bytes. 927 */ 928 ip->ip_len -= hlen; 929 if (ip->ip_off & IP_MF) { 930 /* 931 * Make sure that fragments have a data length 932 * that's a non-zero multiple of 8 bytes. 933 */ 934 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 935 IPSTAT_INC(ips_toosmall); /* XXX */ 936 goto dropfrag; 937 } 938 m->m_flags |= M_FRAG; 939 } else 940 m->m_flags &= ~M_FRAG; 941 ip->ip_off <<= 3; 942 943 944 /* 945 * Attempt reassembly; if it succeeds, proceed. 946 * ip_reass() will return a different mbuf. 947 */ 948 IPSTAT_INC(ips_fragments); 949 m->m_pkthdr.header = ip; 950 951 /* Previous ip_reass() started here. */ 952 /* 953 * Presence of header sizes in mbufs 954 * would confuse code below. 955 */ 956 m->m_data += hlen; 957 m->m_len -= hlen; 958 959 /* 960 * If first fragment to arrive, create a reassembly queue. 961 */ 962 if (fp == NULL) { 963 fp = uma_zalloc(V_ipq_zone, M_NOWAIT); 964 if (fp == NULL) 965 goto dropfrag; 966 #ifdef MAC 967 if (mac_ipq_init(fp, M_NOWAIT) != 0) { 968 uma_zfree(V_ipq_zone, fp); 969 fp = NULL; 970 goto dropfrag; 971 } 972 mac_ipq_create(m, fp); 973 #endif 974 TAILQ_INSERT_HEAD(head, fp, ipq_list); 975 V_nipq++; 976 fp->ipq_nfrags = 1; 977 fp->ipq_ttl = IPFRAGTTL; 978 fp->ipq_p = ip->ip_p; 979 fp->ipq_id = ip->ip_id; 980 fp->ipq_src = ip->ip_src; 981 fp->ipq_dst = ip->ip_dst; 982 fp->ipq_frags = m; 983 m->m_nextpkt = NULL; 984 goto done; 985 } else { 986 fp->ipq_nfrags++; 987 #ifdef MAC 988 mac_ipq_update(m, fp); 989 #endif 990 } 991 992 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 993 994 /* 995 * Handle ECN by comparing this segment with the first one; 996 * if CE is set, do not lose CE. 997 * drop if CE and not-ECT are mixed for the same packet. 998 */ 999 ecn = ip->ip_tos & IPTOS_ECN_MASK; 1000 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 1001 if (ecn == IPTOS_ECN_CE) { 1002 if (ecn0 == IPTOS_ECN_NOTECT) 1003 goto dropfrag; 1004 if (ecn0 != IPTOS_ECN_CE) 1005 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 1006 } 1007 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 1008 goto dropfrag; 1009 1010 /* 1011 * Find a segment which begins after this one does. 1012 */ 1013 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1014 if (GETIP(q)->ip_off > ip->ip_off) 1015 break; 1016 1017 /* 1018 * If there is a preceding segment, it may provide some of 1019 * our data already. If so, drop the data from the incoming 1020 * segment. If it provides all of our data, drop us, otherwise 1021 * stick new segment in the proper place. 1022 * 1023 * If some of the data is dropped from the the preceding 1024 * segment, then it's checksum is invalidated. 1025 */ 1026 if (p) { 1027 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1028 if (i > 0) { 1029 if (i >= ip->ip_len) 1030 goto dropfrag; 1031 m_adj(m, i); 1032 m->m_pkthdr.csum_flags = 0; 1033 ip->ip_off += i; 1034 ip->ip_len -= i; 1035 } 1036 m->m_nextpkt = p->m_nextpkt; 1037 p->m_nextpkt = m; 1038 } else { 1039 m->m_nextpkt = fp->ipq_frags; 1040 fp->ipq_frags = m; 1041 } 1042 1043 /* 1044 * While we overlap succeeding segments trim them or, 1045 * if they are completely covered, dequeue them. 1046 */ 1047 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1048 q = nq) { 1049 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1050 if (i < GETIP(q)->ip_len) { 1051 GETIP(q)->ip_len -= i; 1052 GETIP(q)->ip_off += i; 1053 m_adj(q, i); 1054 q->m_pkthdr.csum_flags = 0; 1055 break; 1056 } 1057 nq = q->m_nextpkt; 1058 m->m_nextpkt = nq; 1059 IPSTAT_INC(ips_fragdropped); 1060 fp->ipq_nfrags--; 1061 m_freem(q); 1062 } 1063 1064 /* 1065 * Check for complete reassembly and perform frag per packet 1066 * limiting. 1067 * 1068 * Frag limiting is performed here so that the nth frag has 1069 * a chance to complete the packet before we drop the packet. 1070 * As a result, n+1 frags are actually allowed per packet, but 1071 * only n will ever be stored. (n = maxfragsperpacket.) 1072 * 1073 */ 1074 next = 0; 1075 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1076 if (GETIP(q)->ip_off != next) { 1077 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1078 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1079 ip_freef(head, fp); 1080 } 1081 goto done; 1082 } 1083 next += GETIP(q)->ip_len; 1084 } 1085 /* Make sure the last packet didn't have the IP_MF flag */ 1086 if (p->m_flags & M_FRAG) { 1087 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1088 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1089 ip_freef(head, fp); 1090 } 1091 goto done; 1092 } 1093 1094 /* 1095 * Reassembly is complete. Make sure the packet is a sane size. 1096 */ 1097 q = fp->ipq_frags; 1098 ip = GETIP(q); 1099 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1100 IPSTAT_INC(ips_toolong); 1101 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1102 ip_freef(head, fp); 1103 goto done; 1104 } 1105 1106 /* 1107 * Concatenate fragments. 1108 */ 1109 m = q; 1110 t = m->m_next; 1111 m->m_next = NULL; 1112 m_cat(m, t); 1113 nq = q->m_nextpkt; 1114 q->m_nextpkt = NULL; 1115 for (q = nq; q != NULL; q = nq) { 1116 nq = q->m_nextpkt; 1117 q->m_nextpkt = NULL; 1118 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1119 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1120 m_cat(m, q); 1121 } 1122 /* 1123 * In order to do checksumming faster we do 'end-around carry' here 1124 * (and not in for{} loop), though it implies we are not going to 1125 * reassemble more than 64k fragments. 1126 */ 1127 m->m_pkthdr.csum_data = 1128 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1129 #ifdef MAC 1130 mac_ipq_reassemble(fp, m); 1131 mac_ipq_destroy(fp); 1132 #endif 1133 1134 /* 1135 * Create header for new ip packet by modifying header of first 1136 * packet; dequeue and discard fragment reassembly header. 1137 * Make header visible. 1138 */ 1139 ip->ip_len = (ip->ip_hl << 2) + next; 1140 ip->ip_src = fp->ipq_src; 1141 ip->ip_dst = fp->ipq_dst; 1142 TAILQ_REMOVE(head, fp, ipq_list); 1143 V_nipq--; 1144 uma_zfree(V_ipq_zone, fp); 1145 m->m_len += (ip->ip_hl << 2); 1146 m->m_data -= (ip->ip_hl << 2); 1147 /* some debugging cruft by sklower, below, will go away soon */ 1148 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1149 m_fixhdr(m); 1150 IPSTAT_INC(ips_reassembled); 1151 IPQ_UNLOCK(); 1152 return (m); 1153 1154 dropfrag: 1155 IPSTAT_INC(ips_fragdropped); 1156 if (fp != NULL) 1157 fp->ipq_nfrags--; 1158 m_freem(m); 1159 done: 1160 IPQ_UNLOCK(); 1161 return (NULL); 1162 1163 #undef GETIP 1164 } 1165 1166 /* 1167 * Free a fragment reassembly header and all 1168 * associated datagrams. 1169 */ 1170 static void 1171 ip_freef(struct ipqhead *fhp, struct ipq *fp) 1172 { 1173 INIT_VNET_INET(curvnet); 1174 struct mbuf *q; 1175 1176 IPQ_LOCK_ASSERT(); 1177 1178 while (fp->ipq_frags) { 1179 q = fp->ipq_frags; 1180 fp->ipq_frags = q->m_nextpkt; 1181 m_freem(q); 1182 } 1183 TAILQ_REMOVE(fhp, fp, ipq_list); 1184 uma_zfree(V_ipq_zone, fp); 1185 V_nipq--; 1186 } 1187 1188 /* 1189 * IP timer processing; 1190 * if a timer expires on a reassembly 1191 * queue, discard it. 1192 */ 1193 void 1194 ip_slowtimo(void) 1195 { 1196 VNET_ITERATOR_DECL(vnet_iter); 1197 struct ipq *fp; 1198 int i; 1199 1200 IPQ_LOCK(); 1201 VNET_LIST_RLOCK(); 1202 VNET_FOREACH(vnet_iter) { 1203 CURVNET_SET(vnet_iter); 1204 INIT_VNET_INET(vnet_iter); 1205 for (i = 0; i < IPREASS_NHASH; i++) { 1206 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) { 1207 struct ipq *fpp; 1208 1209 fpp = fp; 1210 fp = TAILQ_NEXT(fp, ipq_list); 1211 if(--fpp->ipq_ttl == 0) { 1212 IPSTAT_ADD(ips_fragtimeout, 1213 fpp->ipq_nfrags); 1214 ip_freef(&V_ipq[i], fpp); 1215 } 1216 } 1217 } 1218 /* 1219 * If we are over the maximum number of fragments 1220 * (due to the limit being lowered), drain off 1221 * enough to get down to the new limit. 1222 */ 1223 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) { 1224 for (i = 0; i < IPREASS_NHASH; i++) { 1225 while (V_nipq > V_maxnipq && 1226 !TAILQ_EMPTY(&V_ipq[i])) { 1227 IPSTAT_ADD(ips_fragdropped, 1228 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1229 ip_freef(&V_ipq[i], 1230 TAILQ_FIRST(&V_ipq[i])); 1231 } 1232 } 1233 } 1234 CURVNET_RESTORE(); 1235 } 1236 VNET_LIST_RUNLOCK(); 1237 IPQ_UNLOCK(); 1238 } 1239 1240 /* 1241 * Drain off all datagram fragments. 1242 */ 1243 void 1244 ip_drain(void) 1245 { 1246 VNET_ITERATOR_DECL(vnet_iter); 1247 int i; 1248 1249 IPQ_LOCK(); 1250 VNET_LIST_RLOCK(); 1251 VNET_FOREACH(vnet_iter) { 1252 CURVNET_SET(vnet_iter); 1253 INIT_VNET_INET(vnet_iter); 1254 for (i = 0; i < IPREASS_NHASH; i++) { 1255 while(!TAILQ_EMPTY(&V_ipq[i])) { 1256 IPSTAT_ADD(ips_fragdropped, 1257 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1258 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); 1259 } 1260 } 1261 CURVNET_RESTORE(); 1262 } 1263 VNET_LIST_RUNLOCK(); 1264 IPQ_UNLOCK(); 1265 in_rtqdrain(); 1266 } 1267 1268 /* 1269 * The protocol to be inserted into ip_protox[] must be already registered 1270 * in inetsw[], either statically or through pf_proto_register(). 1271 */ 1272 int 1273 ipproto_register(u_char ipproto) 1274 { 1275 struct protosw *pr; 1276 1277 /* Sanity checks. */ 1278 if (ipproto == 0) 1279 return (EPROTONOSUPPORT); 1280 1281 /* 1282 * The protocol slot must not be occupied by another protocol 1283 * already. An index pointing to IPPROTO_RAW is unused. 1284 */ 1285 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1286 if (pr == NULL) 1287 return (EPFNOSUPPORT); 1288 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1289 return (EEXIST); 1290 1291 /* Find the protocol position in inetsw[] and set the index. */ 1292 for (pr = inetdomain.dom_protosw; 1293 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1294 if (pr->pr_domain->dom_family == PF_INET && 1295 pr->pr_protocol && pr->pr_protocol == ipproto) { 1296 /* Be careful to only index valid IP protocols. */ 1297 if (pr->pr_protocol < IPPROTO_MAX) { 1298 ip_protox[pr->pr_protocol] = pr - inetsw; 1299 return (0); 1300 } else 1301 return (EINVAL); 1302 } 1303 } 1304 return (EPROTONOSUPPORT); 1305 } 1306 1307 int 1308 ipproto_unregister(u_char ipproto) 1309 { 1310 struct protosw *pr; 1311 1312 /* Sanity checks. */ 1313 if (ipproto == 0) 1314 return (EPROTONOSUPPORT); 1315 1316 /* Check if the protocol was indeed registered. */ 1317 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1318 if (pr == NULL) 1319 return (EPFNOSUPPORT); 1320 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1321 return (ENOENT); 1322 1323 /* Reset the protocol slot to IPPROTO_RAW. */ 1324 ip_protox[ipproto] = pr - inetsw; 1325 return (0); 1326 } 1327 1328 /* 1329 * Given address of next destination (final or next hop), 1330 * return internet address info of interface to be used to get there. 1331 */ 1332 struct in_ifaddr * 1333 ip_rtaddr(struct in_addr dst, u_int fibnum) 1334 { 1335 struct route sro; 1336 struct sockaddr_in *sin; 1337 struct in_ifaddr *ifa; 1338 1339 bzero(&sro, sizeof(sro)); 1340 sin = (struct sockaddr_in *)&sro.ro_dst; 1341 sin->sin_family = AF_INET; 1342 sin->sin_len = sizeof(*sin); 1343 sin->sin_addr = dst; 1344 in_rtalloc_ign(&sro, 0, fibnum); 1345 1346 if (sro.ro_rt == NULL) 1347 return (NULL); 1348 1349 ifa = ifatoia(sro.ro_rt->rt_ifa); 1350 RTFREE(sro.ro_rt); 1351 return (ifa); 1352 } 1353 1354 u_char inetctlerrmap[PRC_NCMDS] = { 1355 0, 0, 0, 0, 1356 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1357 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1358 EMSGSIZE, EHOSTUNREACH, 0, 0, 1359 0, 0, EHOSTUNREACH, 0, 1360 ENOPROTOOPT, ECONNREFUSED 1361 }; 1362 1363 /* 1364 * Forward a packet. If some error occurs return the sender 1365 * an icmp packet. Note we can't always generate a meaningful 1366 * icmp message because icmp doesn't have a large enough repertoire 1367 * of codes and types. 1368 * 1369 * If not forwarding, just drop the packet. This could be confusing 1370 * if ipforwarding was zero but some routing protocol was advancing 1371 * us as a gateway to somewhere. However, we must let the routing 1372 * protocol deal with that. 1373 * 1374 * The srcrt parameter indicates whether the packet is being forwarded 1375 * via a source route. 1376 */ 1377 void 1378 ip_forward(struct mbuf *m, int srcrt) 1379 { 1380 INIT_VNET_INET(curvnet); 1381 struct ip *ip = mtod(m, struct ip *); 1382 struct in_ifaddr *ia; 1383 struct mbuf *mcopy; 1384 struct in_addr dest; 1385 struct route ro; 1386 int error, type = 0, code = 0, mtu = 0; 1387 1388 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1389 IPSTAT_INC(ips_cantforward); 1390 m_freem(m); 1391 return; 1392 } 1393 #ifdef IPSTEALTH 1394 if (!V_ipstealth) { 1395 #endif 1396 if (ip->ip_ttl <= IPTTLDEC) { 1397 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1398 0, 0); 1399 return; 1400 } 1401 #ifdef IPSTEALTH 1402 } 1403 #endif 1404 1405 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); 1406 #ifndef IPSEC 1407 /* 1408 * 'ia' may be NULL if there is no route for this destination. 1409 * In case of IPsec, Don't discard it just yet, but pass it to 1410 * ip_output in case of outgoing IPsec policy. 1411 */ 1412 if (!srcrt && ia == NULL) { 1413 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1414 return; 1415 } 1416 #endif 1417 1418 /* 1419 * Save the IP header and at most 8 bytes of the payload, 1420 * in case we need to generate an ICMP message to the src. 1421 * 1422 * XXX this can be optimized a lot by saving the data in a local 1423 * buffer on the stack (72 bytes at most), and only allocating the 1424 * mbuf if really necessary. The vast majority of the packets 1425 * are forwarded without having to send an ICMP back (either 1426 * because unnecessary, or because rate limited), so we are 1427 * really we are wasting a lot of work here. 1428 * 1429 * We don't use m_copy() because it might return a reference 1430 * to a shared cluster. Both this function and ip_output() 1431 * assume exclusive access to the IP header in `m', so any 1432 * data in a cluster may change before we reach icmp_error(). 1433 */ 1434 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1435 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1436 /* 1437 * It's probably ok if the pkthdr dup fails (because 1438 * the deep copy of the tag chain failed), but for now 1439 * be conservative and just discard the copy since 1440 * code below may some day want the tags. 1441 */ 1442 m_free(mcopy); 1443 mcopy = NULL; 1444 } 1445 if (mcopy != NULL) { 1446 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy)); 1447 mcopy->m_pkthdr.len = mcopy->m_len; 1448 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1449 } 1450 1451 #ifdef IPSTEALTH 1452 if (!V_ipstealth) { 1453 #endif 1454 ip->ip_ttl -= IPTTLDEC; 1455 #ifdef IPSTEALTH 1456 } 1457 #endif 1458 1459 /* 1460 * If forwarding packet using same interface that it came in on, 1461 * perhaps should send a redirect to sender to shortcut a hop. 1462 * Only send redirect if source is sending directly to us, 1463 * and if packet was not source routed (or has any options). 1464 * Also, don't send redirect if forwarding using a default route 1465 * or a route modified by a redirect. 1466 */ 1467 dest.s_addr = 0; 1468 if (!srcrt && V_ipsendredirects && 1469 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) { 1470 struct sockaddr_in *sin; 1471 struct rtentry *rt; 1472 1473 bzero(&ro, sizeof(ro)); 1474 sin = (struct sockaddr_in *)&ro.ro_dst; 1475 sin->sin_family = AF_INET; 1476 sin->sin_len = sizeof(*sin); 1477 sin->sin_addr = ip->ip_dst; 1478 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 1479 1480 rt = ro.ro_rt; 1481 1482 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1483 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1484 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1485 u_long src = ntohl(ip->ip_src.s_addr); 1486 1487 if (RTA(rt) && 1488 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1489 if (rt->rt_flags & RTF_GATEWAY) 1490 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1491 else 1492 dest.s_addr = ip->ip_dst.s_addr; 1493 /* Router requirements says to only send host redirects */ 1494 type = ICMP_REDIRECT; 1495 code = ICMP_REDIRECT_HOST; 1496 } 1497 } 1498 if (rt) 1499 RTFREE(rt); 1500 } 1501 1502 /* 1503 * Try to cache the route MTU from ip_output so we can consider it for 1504 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. 1505 */ 1506 bzero(&ro, sizeof(ro)); 1507 1508 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1509 1510 if (error == EMSGSIZE && ro.ro_rt) 1511 mtu = ro.ro_rt->rt_rmx.rmx_mtu; 1512 if (ro.ro_rt) 1513 RTFREE(ro.ro_rt); 1514 1515 if (error) 1516 IPSTAT_INC(ips_cantforward); 1517 else { 1518 IPSTAT_INC(ips_forward); 1519 if (type) 1520 IPSTAT_INC(ips_redirectsent); 1521 else { 1522 if (mcopy) 1523 m_freem(mcopy); 1524 return; 1525 } 1526 } 1527 if (mcopy == NULL) 1528 return; 1529 1530 switch (error) { 1531 1532 case 0: /* forwarded, but need redirect */ 1533 /* type, code set above */ 1534 break; 1535 1536 case ENETUNREACH: 1537 case EHOSTUNREACH: 1538 case ENETDOWN: 1539 case EHOSTDOWN: 1540 default: 1541 type = ICMP_UNREACH; 1542 code = ICMP_UNREACH_HOST; 1543 break; 1544 1545 case EMSGSIZE: 1546 type = ICMP_UNREACH; 1547 code = ICMP_UNREACH_NEEDFRAG; 1548 1549 #ifdef IPSEC 1550 /* 1551 * If IPsec is configured for this path, 1552 * override any possibly mtu value set by ip_output. 1553 */ 1554 mtu = ip_ipsec_mtu(m, mtu); 1555 #endif /* IPSEC */ 1556 /* 1557 * If the MTU was set before make sure we are below the 1558 * interface MTU. 1559 * If the MTU wasn't set before use the interface mtu or 1560 * fall back to the next smaller mtu step compared to the 1561 * current packet size. 1562 */ 1563 if (mtu != 0) { 1564 if (ia != NULL) 1565 mtu = min(mtu, ia->ia_ifp->if_mtu); 1566 } else { 1567 if (ia != NULL) 1568 mtu = ia->ia_ifp->if_mtu; 1569 else 1570 mtu = ip_next_mtu(ip->ip_len, 0); 1571 } 1572 IPSTAT_INC(ips_cantfrag); 1573 break; 1574 1575 case ENOBUFS: 1576 /* 1577 * A router should not generate ICMP_SOURCEQUENCH as 1578 * required in RFC1812 Requirements for IP Version 4 Routers. 1579 * Source quench could be a big problem under DoS attacks, 1580 * or if the underlying interface is rate-limited. 1581 * Those who need source quench packets may re-enable them 1582 * via the net.inet.ip.sendsourcequench sysctl. 1583 */ 1584 if (V_ip_sendsourcequench == 0) { 1585 m_freem(mcopy); 1586 return; 1587 } else { 1588 type = ICMP_SOURCEQUENCH; 1589 code = 0; 1590 } 1591 break; 1592 1593 case EACCES: /* ipfw denied packet */ 1594 m_freem(mcopy); 1595 return; 1596 } 1597 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1598 } 1599 1600 void 1601 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1602 struct mbuf *m) 1603 { 1604 INIT_VNET_NET(inp->inp_vnet); 1605 1606 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1607 struct bintime bt; 1608 1609 bintime(&bt); 1610 if (inp->inp_socket->so_options & SO_BINTIME) { 1611 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1612 SCM_BINTIME, SOL_SOCKET); 1613 if (*mp) 1614 mp = &(*mp)->m_next; 1615 } 1616 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1617 struct timeval tv; 1618 1619 bintime2timeval(&bt, &tv); 1620 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1621 SCM_TIMESTAMP, SOL_SOCKET); 1622 if (*mp) 1623 mp = &(*mp)->m_next; 1624 } 1625 } 1626 if (inp->inp_flags & INP_RECVDSTADDR) { 1627 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1628 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1629 if (*mp) 1630 mp = &(*mp)->m_next; 1631 } 1632 if (inp->inp_flags & INP_RECVTTL) { 1633 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1634 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1635 if (*mp) 1636 mp = &(*mp)->m_next; 1637 } 1638 #ifdef notyet 1639 /* XXX 1640 * Moving these out of udp_input() made them even more broken 1641 * than they already were. 1642 */ 1643 /* options were tossed already */ 1644 if (inp->inp_flags & INP_RECVOPTS) { 1645 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1646 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1647 if (*mp) 1648 mp = &(*mp)->m_next; 1649 } 1650 /* ip_srcroute doesn't do what we want here, need to fix */ 1651 if (inp->inp_flags & INP_RECVRETOPTS) { 1652 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1653 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1654 if (*mp) 1655 mp = &(*mp)->m_next; 1656 } 1657 #endif 1658 if (inp->inp_flags & INP_RECVIF) { 1659 struct ifnet *ifp; 1660 struct sdlbuf { 1661 struct sockaddr_dl sdl; 1662 u_char pad[32]; 1663 } sdlbuf; 1664 struct sockaddr_dl *sdp; 1665 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1666 1667 if (((ifp = m->m_pkthdr.rcvif)) 1668 && ( ifp->if_index && (ifp->if_index <= V_if_index))) { 1669 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1670 /* 1671 * Change our mind and don't try copy. 1672 */ 1673 if ((sdp->sdl_family != AF_LINK) 1674 || (sdp->sdl_len > sizeof(sdlbuf))) { 1675 goto makedummy; 1676 } 1677 bcopy(sdp, sdl2, sdp->sdl_len); 1678 } else { 1679 makedummy: 1680 sdl2->sdl_len 1681 = offsetof(struct sockaddr_dl, sdl_data[0]); 1682 sdl2->sdl_family = AF_LINK; 1683 sdl2->sdl_index = 0; 1684 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1685 } 1686 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1687 IP_RECVIF, IPPROTO_IP); 1688 if (*mp) 1689 mp = &(*mp)->m_next; 1690 } 1691 } 1692 1693 /* 1694 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1695 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1696 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1697 * compiled. 1698 */ 1699 int 1700 ip_rsvp_init(struct socket *so) 1701 { 1702 INIT_VNET_INET(so->so_vnet); 1703 1704 if (so->so_type != SOCK_RAW || 1705 so->so_proto->pr_protocol != IPPROTO_RSVP) 1706 return EOPNOTSUPP; 1707 1708 if (V_ip_rsvpd != NULL) 1709 return EADDRINUSE; 1710 1711 V_ip_rsvpd = so; 1712 /* 1713 * This may seem silly, but we need to be sure we don't over-increment 1714 * the RSVP counter, in case something slips up. 1715 */ 1716 if (!V_ip_rsvp_on) { 1717 V_ip_rsvp_on = 1; 1718 V_rsvp_on++; 1719 } 1720 1721 return 0; 1722 } 1723 1724 int 1725 ip_rsvp_done(void) 1726 { 1727 INIT_VNET_INET(curvnet); 1728 1729 V_ip_rsvpd = NULL; 1730 /* 1731 * This may seem silly, but we need to be sure we don't over-decrement 1732 * the RSVP counter, in case something slips up. 1733 */ 1734 if (V_ip_rsvp_on) { 1735 V_ip_rsvp_on = 0; 1736 V_rsvp_on--; 1737 } 1738 return 0; 1739 } 1740 1741 void 1742 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1743 { 1744 INIT_VNET_INET(curvnet); 1745 1746 if (rsvp_input_p) { /* call the real one if loaded */ 1747 rsvp_input_p(m, off); 1748 return; 1749 } 1750 1751 /* Can still get packets with rsvp_on = 0 if there is a local member 1752 * of the group to which the RSVP packet is addressed. But in this 1753 * case we want to throw the packet away. 1754 */ 1755 1756 if (!V_rsvp_on) { 1757 m_freem(m); 1758 return; 1759 } 1760 1761 if (V_ip_rsvpd != NULL) { 1762 rip_input(m, off); 1763 return; 1764 } 1765 /* Drop the packet */ 1766 m_freem(m); 1767 } 1768