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