1 /*- 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include "opt_bootp.h" 36 #include "opt_ipfw.h" 37 #include "opt_ipstealth.h" 38 #include "opt_ipsec.h" 39 #include "opt_route.h" 40 #include "opt_mac.h" 41 #include "opt_carp.h" 42 43 #include <sys/param.h> 44 #include <sys/systm.h> 45 #include <sys/callout.h> 46 #include <sys/mbuf.h> 47 #include <sys/malloc.h> 48 #include <sys/domain.h> 49 #include <sys/protosw.h> 50 #include <sys/socket.h> 51 #include <sys/time.h> 52 #include <sys/kernel.h> 53 #include <sys/lock.h> 54 #include <sys/rwlock.h> 55 #include <sys/syslog.h> 56 #include <sys/sysctl.h> 57 #include <sys/vimage.h> 58 59 #include <net/pfil.h> 60 #include <net/if.h> 61 #include <net/if_types.h> 62 #include <net/if_var.h> 63 #include <net/if_dl.h> 64 #include <net/route.h> 65 #include <net/netisr.h> 66 #include <net/vnet.h> 67 #include <net/flowtable.h> 68 69 #include <netinet/in.h> 70 #include <netinet/in_systm.h> 71 #include <netinet/in_var.h> 72 #include <netinet/ip.h> 73 #include <netinet/in_pcb.h> 74 #include <netinet/ip_var.h> 75 #include <netinet/ip_icmp.h> 76 #include <netinet/ip_options.h> 77 #include <machine/in_cksum.h> 78 #include <netinet/vinet.h> 79 #ifdef DEV_CARP 80 #include <netinet/ip_carp.h> 81 #endif 82 #ifdef IPSEC 83 #include <netinet/ip_ipsec.h> 84 #endif /* IPSEC */ 85 86 #include <sys/socketvar.h> 87 88 /* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */ 89 #include <netinet/ip_fw.h> 90 #include <netinet/ip_dummynet.h> 91 92 #include <security/mac/mac_framework.h> 93 94 #ifdef CTASSERT 95 CTASSERT(sizeof(struct ip) == 20); 96 #endif 97 98 #ifndef VIMAGE 99 #ifndef VIMAGE_GLOBALS 100 struct vnet_inet vnet_inet_0; 101 #endif 102 #endif 103 104 #ifdef VIMAGE_GLOBALS 105 static int ipsendredirects; 106 static int ip_checkinterface; 107 static int ip_keepfaith; 108 static int ip_sendsourcequench; 109 int ip_defttl; 110 int ip_do_randomid; 111 int ipforwarding; 112 struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 113 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 114 u_long in_ifaddrhmask; /* mask for hash table */ 115 struct ipstat ipstat; 116 static int ip_rsvp_on; 117 struct socket *ip_rsvpd; 118 int rsvp_on; 119 static struct ipqhead ipq[IPREASS_NHASH]; 120 static int maxnipq; /* Administrative limit on # reass queues. */ 121 static int maxfragsperpacket; 122 int ipstealth; 123 static int nipq; /* Total # of reass queues */ 124 #endif 125 126 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_FORWARDING, 127 forwarding, CTLFLAG_RW, ipforwarding, 0, 128 "Enable IP forwarding between interfaces"); 129 130 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_SENDREDIRECTS, 131 redirect, CTLFLAG_RW, ipsendredirects, 0, 132 "Enable sending IP redirects"); 133 134 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_DEFTTL, 135 ttl, CTLFLAG_RW, ip_defttl, 0, "Maximum TTL on IP packets"); 136 137 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_KEEPFAITH, 138 keepfaith, CTLFLAG_RW, ip_keepfaith, 0, 139 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 140 141 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, 142 sendsourcequench, CTLFLAG_RW, ip_sendsourcequench, 0, 143 "Enable the transmission of source quench packets"); 144 145 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, random_id, 146 CTLFLAG_RW, ip_do_randomid, 0, "Assign random ip_id values"); 147 148 /* 149 * XXX - Setting ip_checkinterface mostly implements the receive side of 150 * the Strong ES model described in RFC 1122, but since the routing table 151 * and transmit implementation do not implement the Strong ES model, 152 * setting this to 1 results in an odd hybrid. 153 * 154 * XXX - ip_checkinterface currently must be disabled if you use ipnat 155 * to translate the destination address to another local interface. 156 * 157 * XXX - ip_checkinterface must be disabled if you add IP aliases 158 * to the loopback interface instead of the interface where the 159 * packets for those addresses are received. 160 */ 161 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, 162 check_interface, CTLFLAG_RW, ip_checkinterface, 0, 163 "Verify packet arrives on correct interface"); 164 165 struct pfil_head inet_pfil_hook; /* Packet filter hooks */ 166 167 static struct netisr_handler ip_nh = { 168 .nh_name = "ip", 169 .nh_handler = ip_input, 170 .nh_proto = NETISR_IP, 171 .nh_policy = NETISR_POLICY_FLOW, 172 }; 173 174 extern struct domain inetdomain; 175 extern struct protosw inetsw[]; 176 u_char ip_protox[IPPROTO_MAX]; 177 178 179 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 static int ip_output_flowtable_size = 2048; 215 TUNABLE_INT("net.inet.ip.output_flowtable_size", &ip_output_flowtable_size); 216 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, output_flowtable_size, 217 CTLFLAG_RDTUN, ip_output_flowtable_size, 2048, 218 "number of entries in the per-cpu output flow caches"); 219 220 /* 221 * ipfw_ether and ipfw_bridge hooks. 222 * XXX: Temporary until those are converted to pfil_hooks as well. 223 */ 224 ip_fw_chk_t *ip_fw_chk_ptr = NULL; 225 ip_dn_io_t *ip_dn_io_ptr = NULL; 226 #ifdef VIMAGE_GLOBALS 227 int fw_one_pass; 228 #endif 229 struct flowtable *ip_ft; 230 231 static void ip_freef(struct ipqhead *, struct ipq *); 232 233 #ifndef VIMAGE_GLOBALS 234 static void vnet_inet_register(void); 235 236 static const vnet_modinfo_t vnet_inet_modinfo = { 237 .vmi_id = VNET_MOD_INET, 238 .vmi_name = "inet", 239 .vmi_size = sizeof(struct vnet_inet) 240 }; 241 242 static void vnet_inet_register() 243 { 244 245 vnet_mod_register(&vnet_inet_modinfo); 246 } 247 248 SYSINIT(inet, SI_SUB_PROTO_BEGIN, SI_ORDER_FIRST, vnet_inet_register, 0); 249 #endif 250 251 static int 252 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 253 { 254 int error, qlimit; 255 256 netisr_getqlimit(&ip_nh, &qlimit); 257 error = sysctl_handle_int(oidp, &qlimit, 0, req); 258 if (error || !req->newptr) 259 return (error); 260 if (qlimit < 1) 261 return (EINVAL); 262 return (netisr_setqlimit(&ip_nh, qlimit)); 263 } 264 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, 265 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I", 266 "Maximum size of the IP input queue"); 267 268 static int 269 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) 270 { 271 u_int64_t qdrops_long; 272 int error, qdrops; 273 274 netisr_getqdrops(&ip_nh, &qdrops_long); 275 qdrops = qdrops_long; 276 error = sysctl_handle_int(oidp, &qdrops, 0, req); 277 if (error || !req->newptr) 278 return (error); 279 if (qdrops != 0) 280 return (EINVAL); 281 netisr_clearqdrops(&ip_nh); 282 return (0); 283 } 284 285 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, 286 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I", 287 "Number of packets dropped from the IP input queue"); 288 289 /* 290 * IP initialization: fill in IP protocol switch table. 291 * All protocols not implemented in kernel go to raw IP protocol handler. 292 */ 293 void 294 ip_init(void) 295 { 296 INIT_VNET_INET(curvnet); 297 struct protosw *pr; 298 int i; 299 300 V_ipsendredirects = 1; /* XXX */ 301 V_ip_checkinterface = 0; 302 V_ip_keepfaith = 0; 303 V_ip_sendsourcequench = 0; 304 V_rsvp_on = 0; 305 V_ip_defttl = IPDEFTTL; 306 V_ip_do_randomid = 0; 307 V_ip_id = time_second & 0xffff; 308 V_ipforwarding = 0; 309 V_ipstealth = 0; 310 V_nipq = 0; /* Total # of reass queues */ 311 312 V_ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */ 313 V_ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */ 314 V_ipport_firstauto = IPPORT_EPHEMERALFIRST; /* 10000 */ 315 V_ipport_lastauto = IPPORT_EPHEMERALLAST; /* 65535 */ 316 V_ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */ 317 V_ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */ 318 V_ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */ 319 V_ipport_reservedlow = 0; 320 V_ipport_randomized = 1; /* user controlled via sysctl */ 321 V_ipport_randomcps = 10; /* user controlled via sysctl */ 322 V_ipport_randomtime = 45; /* user controlled via sysctl */ 323 V_ipport_stoprandom = 0; /* toggled by ipport_tick */ 324 325 V_fw_one_pass = 1; 326 327 #ifdef NOTYET 328 /* XXX global static but not instantiated in this file */ 329 V_ipfastforward_active = 0; 330 V_subnetsarelocal = 0; 331 V_sameprefixcarponly = 0; 332 #endif 333 334 TAILQ_INIT(&V_in_ifaddrhead); 335 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 336 337 /* Initialize IP reassembly queue. */ 338 for (i = 0; i < IPREASS_NHASH; i++) 339 TAILQ_INIT(&V_ipq[i]); 340 V_maxnipq = nmbclusters / 32; 341 V_maxfragsperpacket = 16; 342 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 343 NULL, UMA_ALIGN_PTR, 0); 344 maxnipq_update(); 345 346 /* Skip initialization of globals for non-default instances. */ 347 if (!IS_DEFAULT_VNET(curvnet)) 348 return; 349 350 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 351 if (pr == NULL) 352 panic("ip_init: PF_INET not found"); 353 354 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 355 for (i = 0; i < IPPROTO_MAX; i++) 356 ip_protox[i] = pr - inetsw; 357 /* 358 * Cycle through IP protocols and put them into the appropriate place 359 * in ip_protox[]. 360 */ 361 for (pr = inetdomain.dom_protosw; 362 pr < inetdomain.dom_protoswNPROTOSW; pr++) 363 if (pr->pr_domain->dom_family == PF_INET && 364 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 365 /* Be careful to only index valid IP protocols. */ 366 if (pr->pr_protocol < IPPROTO_MAX) 367 ip_protox[pr->pr_protocol] = pr - inetsw; 368 } 369 370 /* Initialize packet filter hooks. */ 371 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 372 inet_pfil_hook.ph_af = AF_INET; 373 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 374 printf("%s: WARNING: unable to register pfil hook, " 375 "error %d\n", __func__, i); 376 377 /* Start ipport_tick. */ 378 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE); 379 callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL); 380 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, 381 SHUTDOWN_PRI_DEFAULT); 382 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, 383 NULL, EVENTHANDLER_PRI_ANY); 384 385 /* Initialize various other remaining things. */ 386 IPQ_LOCK_INIT(); 387 netisr_register(&ip_nh); 388 ip_ft = flowtable_alloc(ip_output_flowtable_size, FL_PCPU); 389 } 390 391 void 392 ip_fini(void *xtp) 393 { 394 395 callout_stop(&ipport_tick_callout); 396 } 397 398 /* 399 * Ip input routine. Checksum and byte swap header. If fragmented 400 * try to reassemble. Process options. Pass to next level. 401 */ 402 void 403 ip_input(struct mbuf *m) 404 { 405 INIT_VNET_INET(curvnet); 406 struct ip *ip = NULL; 407 struct in_ifaddr *ia = NULL; 408 struct ifaddr *ifa; 409 struct ifnet *ifp; 410 int checkif, hlen = 0; 411 u_short sum; 412 int dchg = 0; /* dest changed after fw */ 413 struct in_addr odst; /* original dst address */ 414 415 M_ASSERTPKTHDR(m); 416 417 if (m->m_flags & M_FASTFWD_OURS) { 418 /* 419 * Firewall or NAT changed destination to local. 420 * We expect ip_len and ip_off to be in host byte order. 421 */ 422 m->m_flags &= ~M_FASTFWD_OURS; 423 /* Set up some basics that will be used later. */ 424 ip = mtod(m, struct ip *); 425 hlen = ip->ip_hl << 2; 426 goto ours; 427 } 428 429 IPSTAT_INC(ips_total); 430 431 if (m->m_pkthdr.len < sizeof(struct ip)) 432 goto tooshort; 433 434 if (m->m_len < sizeof (struct ip) && 435 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 436 IPSTAT_INC(ips_toosmall); 437 return; 438 } 439 ip = mtod(m, struct ip *); 440 441 if (ip->ip_v != IPVERSION) { 442 IPSTAT_INC(ips_badvers); 443 goto bad; 444 } 445 446 hlen = ip->ip_hl << 2; 447 if (hlen < sizeof(struct ip)) { /* minimum header length */ 448 IPSTAT_INC(ips_badhlen); 449 goto bad; 450 } 451 if (hlen > m->m_len) { 452 if ((m = m_pullup(m, hlen)) == NULL) { 453 IPSTAT_INC(ips_badhlen); 454 return; 455 } 456 ip = mtod(m, struct ip *); 457 } 458 459 /* 127/8 must not appear on wire - RFC1122 */ 460 ifp = m->m_pkthdr.rcvif; 461 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 462 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 463 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 464 IPSTAT_INC(ips_badaddr); 465 goto bad; 466 } 467 } 468 469 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 470 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 471 } else { 472 if (hlen == sizeof(struct ip)) { 473 sum = in_cksum_hdr(ip); 474 } else { 475 sum = in_cksum(m, hlen); 476 } 477 } 478 if (sum) { 479 IPSTAT_INC(ips_badsum); 480 goto bad; 481 } 482 483 #ifdef ALTQ 484 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 485 /* packet is dropped by traffic conditioner */ 486 return; 487 #endif 488 489 /* 490 * Convert fields to host representation. 491 */ 492 ip->ip_len = ntohs(ip->ip_len); 493 if (ip->ip_len < hlen) { 494 IPSTAT_INC(ips_badlen); 495 goto bad; 496 } 497 ip->ip_off = ntohs(ip->ip_off); 498 499 /* 500 * Check that the amount of data in the buffers 501 * is as at least much as the IP header would have us expect. 502 * Trim mbufs if longer than we expect. 503 * Drop packet if shorter than we expect. 504 */ 505 if (m->m_pkthdr.len < ip->ip_len) { 506 tooshort: 507 IPSTAT_INC(ips_tooshort); 508 goto bad; 509 } 510 if (m->m_pkthdr.len > ip->ip_len) { 511 if (m->m_len == m->m_pkthdr.len) { 512 m->m_len = ip->ip_len; 513 m->m_pkthdr.len = ip->ip_len; 514 } else 515 m_adj(m, ip->ip_len - m->m_pkthdr.len); 516 } 517 #ifdef IPSEC 518 /* 519 * Bypass packet filtering for packets from a tunnel (gif). 520 */ 521 if (ip_ipsec_filtertunnel(m)) 522 goto passin; 523 #endif /* IPSEC */ 524 525 /* 526 * Run through list of hooks for input packets. 527 * 528 * NB: Beware of the destination address changing (e.g. 529 * by NAT rewriting). When this happens, tell 530 * ip_forward to do the right thing. 531 */ 532 533 /* Jump over all PFIL processing if hooks are not active. */ 534 if (!PFIL_HOOKED(&inet_pfil_hook)) 535 goto passin; 536 537 odst = ip->ip_dst; 538 if (pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0) 539 return; 540 if (m == NULL) /* consumed by filter */ 541 return; 542 543 ip = mtod(m, struct ip *); 544 dchg = (odst.s_addr != ip->ip_dst.s_addr); 545 ifp = m->m_pkthdr.rcvif; 546 547 #ifdef IPFIREWALL_FORWARD 548 if (m->m_flags & M_FASTFWD_OURS) { 549 m->m_flags &= ~M_FASTFWD_OURS; 550 goto ours; 551 } 552 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 553 /* 554 * Directly ship on the packet. This allows to forward packets 555 * that were destined for us to some other directly connected 556 * host. 557 */ 558 ip_forward(m, dchg); 559 return; 560 } 561 #endif /* IPFIREWALL_FORWARD */ 562 563 passin: 564 /* 565 * Process options and, if not destined for us, 566 * ship it on. ip_dooptions returns 1 when an 567 * error was detected (causing an icmp message 568 * to be sent and the original packet to be freed). 569 */ 570 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 571 return; 572 573 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 574 * matter if it is destined to another node, or whether it is 575 * a multicast one, RSVP wants it! and prevents it from being forwarded 576 * anywhere else. Also checks if the rsvp daemon is running before 577 * grabbing the packet. 578 */ 579 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 580 goto ours; 581 582 /* 583 * Check our list of addresses, to see if the packet is for us. 584 * If we don't have any addresses, assume any unicast packet 585 * we receive might be for us (and let the upper layers deal 586 * with it). 587 */ 588 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 589 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 590 goto ours; 591 592 /* 593 * Enable a consistency check between the destination address 594 * and the arrival interface for a unicast packet (the RFC 1122 595 * strong ES model) if IP forwarding is disabled and the packet 596 * is not locally generated and the packet is not subject to 597 * 'ipfw fwd'. 598 * 599 * XXX - Checking also should be disabled if the destination 600 * address is ipnat'ed to a different interface. 601 * 602 * XXX - Checking is incompatible with IP aliases added 603 * to the loopback interface instead of the interface where 604 * the packets are received. 605 * 606 * XXX - This is the case for carp vhost IPs as well so we 607 * insert a workaround. If the packet got here, we already 608 * checked with carp_iamatch() and carp_forus(). 609 */ 610 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 611 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && 612 #ifdef DEV_CARP 613 !ifp->if_carp && 614 #endif 615 (dchg == 0); 616 617 /* 618 * Check for exact addresses in the hash bucket. 619 */ 620 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 621 /* 622 * If the address matches, verify that the packet 623 * arrived via the correct interface if checking is 624 * enabled. 625 */ 626 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 627 (!checkif || ia->ia_ifp == ifp)) 628 goto ours; 629 } 630 /* 631 * Check for broadcast addresses. 632 * 633 * Only accept broadcast packets that arrive via the matching 634 * interface. Reception of forwarded directed broadcasts would 635 * be handled via ip_forward() and ether_output() with the loopback 636 * into the stack for SIMPLEX interfaces handled by ether_output(). 637 */ 638 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 639 IF_ADDR_LOCK(ifp); 640 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 641 if (ifa->ifa_addr->sa_family != AF_INET) 642 continue; 643 ia = ifatoia(ifa); 644 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 645 ip->ip_dst.s_addr) { 646 IF_ADDR_UNLOCK(ifp); 647 goto ours; 648 } 649 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) { 650 IF_ADDR_UNLOCK(ifp); 651 goto ours; 652 } 653 #ifdef BOOTP_COMPAT 654 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { 655 IF_ADDR_UNLOCK(ifp); 656 goto ours; 657 } 658 #endif 659 } 660 IF_ADDR_UNLOCK(ifp); 661 } 662 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 663 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 664 IPSTAT_INC(ips_cantforward); 665 m_freem(m); 666 return; 667 } 668 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 669 if (V_ip_mrouter) { 670 /* 671 * If we are acting as a multicast router, all 672 * incoming multicast packets are passed to the 673 * kernel-level multicast forwarding function. 674 * The packet is returned (relatively) intact; if 675 * ip_mforward() returns a non-zero value, the packet 676 * must be discarded, else it may be accepted below. 677 */ 678 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 679 IPSTAT_INC(ips_cantforward); 680 m_freem(m); 681 return; 682 } 683 684 /* 685 * The process-level routing daemon needs to receive 686 * all multicast IGMP packets, whether or not this 687 * host belongs to their destination groups. 688 */ 689 if (ip->ip_p == IPPROTO_IGMP) 690 goto ours; 691 IPSTAT_INC(ips_forward); 692 } 693 /* 694 * Assume the packet is for us, to avoid prematurely taking 695 * a lock on the in_multi hash. Protocols must perform 696 * their own filtering and update statistics accordingly. 697 */ 698 goto ours; 699 } 700 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 701 goto ours; 702 if (ip->ip_dst.s_addr == INADDR_ANY) 703 goto ours; 704 705 /* 706 * FAITH(Firewall Aided Internet Translator) 707 */ 708 if (ifp && ifp->if_type == IFT_FAITH) { 709 if (V_ip_keepfaith) { 710 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 711 goto ours; 712 } 713 m_freem(m); 714 return; 715 } 716 717 /* 718 * Not for us; forward if possible and desirable. 719 */ 720 if (V_ipforwarding == 0) { 721 IPSTAT_INC(ips_cantforward); 722 m_freem(m); 723 } else { 724 #ifdef IPSEC 725 if (ip_ipsec_fwd(m)) 726 goto bad; 727 #endif /* IPSEC */ 728 ip_forward(m, dchg); 729 } 730 return; 731 732 ours: 733 #ifdef IPSTEALTH 734 /* 735 * IPSTEALTH: Process non-routing options only 736 * if the packet is destined for us. 737 */ 738 if (V_ipstealth && hlen > sizeof (struct ip) && 739 ip_dooptions(m, 1)) 740 return; 741 #endif /* IPSTEALTH */ 742 743 /* Count the packet in the ip address stats */ 744 if (ia != NULL) { 745 ia->ia_ifa.if_ipackets++; 746 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 747 } 748 749 /* 750 * Attempt reassembly; if it succeeds, proceed. 751 * ip_reass() will return a different mbuf. 752 */ 753 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 754 m = ip_reass(m); 755 if (m == NULL) 756 return; 757 ip = mtod(m, struct ip *); 758 /* Get the header length of the reassembled packet */ 759 hlen = ip->ip_hl << 2; 760 } 761 762 /* 763 * Further protocols expect the packet length to be w/o the 764 * IP header. 765 */ 766 ip->ip_len -= hlen; 767 768 #ifdef IPSEC 769 /* 770 * enforce IPsec policy checking if we are seeing last header. 771 * note that we do not visit this with protocols with pcb layer 772 * code - like udp/tcp/raw ip. 773 */ 774 if (ip_ipsec_input(m)) 775 goto bad; 776 #endif /* IPSEC */ 777 778 /* 779 * Switch out to protocol's input routine. 780 */ 781 IPSTAT_INC(ips_delivered); 782 783 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 784 return; 785 bad: 786 m_freem(m); 787 } 788 789 /* 790 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 791 * max has slightly different semantics than the sysctl, for historical 792 * reasons. 793 */ 794 static void 795 maxnipq_update(void) 796 { 797 INIT_VNET_INET(curvnet); 798 799 /* 800 * -1 for unlimited allocation. 801 */ 802 if (V_maxnipq < 0) 803 uma_zone_set_max(V_ipq_zone, 0); 804 /* 805 * Positive number for specific bound. 806 */ 807 if (V_maxnipq > 0) 808 uma_zone_set_max(V_ipq_zone, V_maxnipq); 809 /* 810 * Zero specifies no further fragment queue allocation -- set the 811 * bound very low, but rely on implementation elsewhere to actually 812 * prevent allocation and reclaim current queues. 813 */ 814 if (V_maxnipq == 0) 815 uma_zone_set_max(V_ipq_zone, 1); 816 } 817 818 static void 819 ipq_zone_change(void *tag) 820 { 821 INIT_VNET_INET(curvnet); 822 823 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) { 824 V_maxnipq = nmbclusters / 32; 825 maxnipq_update(); 826 } 827 } 828 829 static int 830 sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 831 { 832 INIT_VNET_INET(curvnet); 833 int error, i; 834 835 i = V_maxnipq; 836 error = sysctl_handle_int(oidp, &i, 0, req); 837 if (error || !req->newptr) 838 return (error); 839 840 /* 841 * XXXRW: Might be a good idea to sanity check the argument and place 842 * an extreme upper bound. 843 */ 844 if (i < -1) 845 return (EINVAL); 846 V_maxnipq = i; 847 maxnipq_update(); 848 return (0); 849 } 850 851 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 852 NULL, 0, sysctl_maxnipq, "I", 853 "Maximum number of IPv4 fragment reassembly queue entries"); 854 855 /* 856 * Take incoming datagram fragment and try to reassemble it into 857 * whole datagram. If the argument is the first fragment or one 858 * in between the function will return NULL and store the mbuf 859 * in the fragment chain. If the argument is the last fragment 860 * the packet will be reassembled and the pointer to the new 861 * mbuf returned for further processing. Only m_tags attached 862 * to the first packet/fragment are preserved. 863 * The IP header is *NOT* adjusted out of iplen. 864 */ 865 struct mbuf * 866 ip_reass(struct mbuf *m) 867 { 868 INIT_VNET_INET(curvnet); 869 struct ip *ip; 870 struct mbuf *p, *q, *nq, *t; 871 struct ipq *fp = NULL; 872 struct ipqhead *head; 873 int i, hlen, next; 874 u_int8_t ecn, ecn0; 875 u_short hash; 876 877 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 878 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) { 879 IPSTAT_INC(ips_fragments); 880 IPSTAT_INC(ips_fragdropped); 881 m_freem(m); 882 return (NULL); 883 } 884 885 ip = mtod(m, struct ip *); 886 hlen = ip->ip_hl << 2; 887 888 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 889 head = &V_ipq[hash]; 890 IPQ_LOCK(); 891 892 /* 893 * Look for queue of fragments 894 * of this datagram. 895 */ 896 TAILQ_FOREACH(fp, head, ipq_list) 897 if (ip->ip_id == fp->ipq_id && 898 ip->ip_src.s_addr == fp->ipq_src.s_addr && 899 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 900 #ifdef MAC 901 mac_ipq_match(m, fp) && 902 #endif 903 ip->ip_p == fp->ipq_p) 904 goto found; 905 906 fp = NULL; 907 908 /* 909 * Attempt to trim the number of allocated fragment queues if it 910 * exceeds the administrative limit. 911 */ 912 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) { 913 /* 914 * drop something from the tail of the current queue 915 * before proceeding further 916 */ 917 struct ipq *q = TAILQ_LAST(head, ipqhead); 918 if (q == NULL) { /* gak */ 919 for (i = 0; i < IPREASS_NHASH; i++) { 920 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead); 921 if (r) { 922 IPSTAT_ADD(ips_fragtimeout, 923 r->ipq_nfrags); 924 ip_freef(&V_ipq[i], r); 925 break; 926 } 927 } 928 } else { 929 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags); 930 ip_freef(head, q); 931 } 932 } 933 934 found: 935 /* 936 * Adjust ip_len to not reflect header, 937 * convert offset of this to bytes. 938 */ 939 ip->ip_len -= hlen; 940 if (ip->ip_off & IP_MF) { 941 /* 942 * Make sure that fragments have a data length 943 * that's a non-zero multiple of 8 bytes. 944 */ 945 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 946 IPSTAT_INC(ips_toosmall); /* XXX */ 947 goto dropfrag; 948 } 949 m->m_flags |= M_FRAG; 950 } else 951 m->m_flags &= ~M_FRAG; 952 ip->ip_off <<= 3; 953 954 955 /* 956 * Attempt reassembly; if it succeeds, proceed. 957 * ip_reass() will return a different mbuf. 958 */ 959 IPSTAT_INC(ips_fragments); 960 m->m_pkthdr.header = ip; 961 962 /* Previous ip_reass() started here. */ 963 /* 964 * Presence of header sizes in mbufs 965 * would confuse code below. 966 */ 967 m->m_data += hlen; 968 m->m_len -= hlen; 969 970 /* 971 * If first fragment to arrive, create a reassembly queue. 972 */ 973 if (fp == NULL) { 974 fp = uma_zalloc(V_ipq_zone, M_NOWAIT); 975 if (fp == NULL) 976 goto dropfrag; 977 #ifdef MAC 978 if (mac_ipq_init(fp, M_NOWAIT) != 0) { 979 uma_zfree(V_ipq_zone, fp); 980 fp = NULL; 981 goto dropfrag; 982 } 983 mac_ipq_create(m, fp); 984 #endif 985 TAILQ_INSERT_HEAD(head, fp, ipq_list); 986 V_nipq++; 987 fp->ipq_nfrags = 1; 988 fp->ipq_ttl = IPFRAGTTL; 989 fp->ipq_p = ip->ip_p; 990 fp->ipq_id = ip->ip_id; 991 fp->ipq_src = ip->ip_src; 992 fp->ipq_dst = ip->ip_dst; 993 fp->ipq_frags = m; 994 m->m_nextpkt = NULL; 995 goto done; 996 } else { 997 fp->ipq_nfrags++; 998 #ifdef MAC 999 mac_ipq_update(m, fp); 1000 #endif 1001 } 1002 1003 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1004 1005 /* 1006 * Handle ECN by comparing this segment with the first one; 1007 * if CE is set, do not lose CE. 1008 * drop if CE and not-ECT are mixed for the same packet. 1009 */ 1010 ecn = ip->ip_tos & IPTOS_ECN_MASK; 1011 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 1012 if (ecn == IPTOS_ECN_CE) { 1013 if (ecn0 == IPTOS_ECN_NOTECT) 1014 goto dropfrag; 1015 if (ecn0 != IPTOS_ECN_CE) 1016 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 1017 } 1018 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 1019 goto dropfrag; 1020 1021 /* 1022 * Find a segment which begins after this one does. 1023 */ 1024 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1025 if (GETIP(q)->ip_off > ip->ip_off) 1026 break; 1027 1028 /* 1029 * If there is a preceding segment, it may provide some of 1030 * our data already. If so, drop the data from the incoming 1031 * segment. If it provides all of our data, drop us, otherwise 1032 * stick new segment in the proper place. 1033 * 1034 * If some of the data is dropped from the the preceding 1035 * segment, then it's checksum is invalidated. 1036 */ 1037 if (p) { 1038 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1039 if (i > 0) { 1040 if (i >= ip->ip_len) 1041 goto dropfrag; 1042 m_adj(m, i); 1043 m->m_pkthdr.csum_flags = 0; 1044 ip->ip_off += i; 1045 ip->ip_len -= i; 1046 } 1047 m->m_nextpkt = p->m_nextpkt; 1048 p->m_nextpkt = m; 1049 } else { 1050 m->m_nextpkt = fp->ipq_frags; 1051 fp->ipq_frags = m; 1052 } 1053 1054 /* 1055 * While we overlap succeeding segments trim them or, 1056 * if they are completely covered, dequeue them. 1057 */ 1058 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1059 q = nq) { 1060 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1061 if (i < GETIP(q)->ip_len) { 1062 GETIP(q)->ip_len -= i; 1063 GETIP(q)->ip_off += i; 1064 m_adj(q, i); 1065 q->m_pkthdr.csum_flags = 0; 1066 break; 1067 } 1068 nq = q->m_nextpkt; 1069 m->m_nextpkt = nq; 1070 IPSTAT_INC(ips_fragdropped); 1071 fp->ipq_nfrags--; 1072 m_freem(q); 1073 } 1074 1075 /* 1076 * Check for complete reassembly and perform frag per packet 1077 * limiting. 1078 * 1079 * Frag limiting is performed here so that the nth frag has 1080 * a chance to complete the packet before we drop the packet. 1081 * As a result, n+1 frags are actually allowed per packet, but 1082 * only n will ever be stored. (n = maxfragsperpacket.) 1083 * 1084 */ 1085 next = 0; 1086 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1087 if (GETIP(q)->ip_off != next) { 1088 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1089 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1090 ip_freef(head, fp); 1091 } 1092 goto done; 1093 } 1094 next += GETIP(q)->ip_len; 1095 } 1096 /* Make sure the last packet didn't have the IP_MF flag */ 1097 if (p->m_flags & M_FRAG) { 1098 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1099 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1100 ip_freef(head, fp); 1101 } 1102 goto done; 1103 } 1104 1105 /* 1106 * Reassembly is complete. Make sure the packet is a sane size. 1107 */ 1108 q = fp->ipq_frags; 1109 ip = GETIP(q); 1110 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1111 IPSTAT_INC(ips_toolong); 1112 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1113 ip_freef(head, fp); 1114 goto done; 1115 } 1116 1117 /* 1118 * Concatenate fragments. 1119 */ 1120 m = q; 1121 t = m->m_next; 1122 m->m_next = NULL; 1123 m_cat(m, t); 1124 nq = q->m_nextpkt; 1125 q->m_nextpkt = NULL; 1126 for (q = nq; q != NULL; q = nq) { 1127 nq = q->m_nextpkt; 1128 q->m_nextpkt = NULL; 1129 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1130 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1131 m_cat(m, q); 1132 } 1133 /* 1134 * In order to do checksumming faster we do 'end-around carry' here 1135 * (and not in for{} loop), though it implies we are not going to 1136 * reassemble more than 64k fragments. 1137 */ 1138 m->m_pkthdr.csum_data = 1139 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1140 #ifdef MAC 1141 mac_ipq_reassemble(fp, m); 1142 mac_ipq_destroy(fp); 1143 #endif 1144 1145 /* 1146 * Create header for new ip packet by modifying header of first 1147 * packet; dequeue and discard fragment reassembly header. 1148 * Make header visible. 1149 */ 1150 ip->ip_len = (ip->ip_hl << 2) + next; 1151 ip->ip_src = fp->ipq_src; 1152 ip->ip_dst = fp->ipq_dst; 1153 TAILQ_REMOVE(head, fp, ipq_list); 1154 V_nipq--; 1155 uma_zfree(V_ipq_zone, fp); 1156 m->m_len += (ip->ip_hl << 2); 1157 m->m_data -= (ip->ip_hl << 2); 1158 /* some debugging cruft by sklower, below, will go away soon */ 1159 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1160 m_fixhdr(m); 1161 IPSTAT_INC(ips_reassembled); 1162 IPQ_UNLOCK(); 1163 return (m); 1164 1165 dropfrag: 1166 IPSTAT_INC(ips_fragdropped); 1167 if (fp != NULL) 1168 fp->ipq_nfrags--; 1169 m_freem(m); 1170 done: 1171 IPQ_UNLOCK(); 1172 return (NULL); 1173 1174 #undef GETIP 1175 } 1176 1177 /* 1178 * Free a fragment reassembly header and all 1179 * associated datagrams. 1180 */ 1181 static void 1182 ip_freef(struct ipqhead *fhp, struct ipq *fp) 1183 { 1184 INIT_VNET_INET(curvnet); 1185 struct mbuf *q; 1186 1187 IPQ_LOCK_ASSERT(); 1188 1189 while (fp->ipq_frags) { 1190 q = fp->ipq_frags; 1191 fp->ipq_frags = q->m_nextpkt; 1192 m_freem(q); 1193 } 1194 TAILQ_REMOVE(fhp, fp, ipq_list); 1195 uma_zfree(V_ipq_zone, fp); 1196 V_nipq--; 1197 } 1198 1199 /* 1200 * IP timer processing; 1201 * if a timer expires on a reassembly 1202 * queue, discard it. 1203 */ 1204 void 1205 ip_slowtimo(void) 1206 { 1207 VNET_ITERATOR_DECL(vnet_iter); 1208 struct ipq *fp; 1209 int i; 1210 1211 IPQ_LOCK(); 1212 VNET_LIST_RLOCK(); 1213 VNET_FOREACH(vnet_iter) { 1214 CURVNET_SET(vnet_iter); 1215 INIT_VNET_INET(vnet_iter); 1216 for (i = 0; i < IPREASS_NHASH; i++) { 1217 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) { 1218 struct ipq *fpp; 1219 1220 fpp = fp; 1221 fp = TAILQ_NEXT(fp, ipq_list); 1222 if(--fpp->ipq_ttl == 0) { 1223 IPSTAT_ADD(ips_fragtimeout, 1224 fpp->ipq_nfrags); 1225 ip_freef(&V_ipq[i], fpp); 1226 } 1227 } 1228 } 1229 /* 1230 * If we are over the maximum number of fragments 1231 * (due to the limit being lowered), drain off 1232 * enough to get down to the new limit. 1233 */ 1234 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) { 1235 for (i = 0; i < IPREASS_NHASH; i++) { 1236 while (V_nipq > V_maxnipq && 1237 !TAILQ_EMPTY(&V_ipq[i])) { 1238 IPSTAT_ADD(ips_fragdropped, 1239 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1240 ip_freef(&V_ipq[i], 1241 TAILQ_FIRST(&V_ipq[i])); 1242 } 1243 } 1244 } 1245 CURVNET_RESTORE(); 1246 } 1247 VNET_LIST_RUNLOCK(); 1248 IPQ_UNLOCK(); 1249 } 1250 1251 /* 1252 * Drain off all datagram fragments. 1253 */ 1254 void 1255 ip_drain(void) 1256 { 1257 VNET_ITERATOR_DECL(vnet_iter); 1258 int i; 1259 1260 IPQ_LOCK(); 1261 VNET_LIST_RLOCK(); 1262 VNET_FOREACH(vnet_iter) { 1263 CURVNET_SET(vnet_iter); 1264 INIT_VNET_INET(vnet_iter); 1265 for (i = 0; i < IPREASS_NHASH; i++) { 1266 while(!TAILQ_EMPTY(&V_ipq[i])) { 1267 IPSTAT_ADD(ips_fragdropped, 1268 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1269 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); 1270 } 1271 } 1272 CURVNET_RESTORE(); 1273 } 1274 VNET_LIST_RUNLOCK(); 1275 IPQ_UNLOCK(); 1276 in_rtqdrain(); 1277 } 1278 1279 /* 1280 * The protocol to be inserted into ip_protox[] must be already registered 1281 * in inetsw[], either statically or through pf_proto_register(). 1282 */ 1283 int 1284 ipproto_register(u_char ipproto) 1285 { 1286 struct protosw *pr; 1287 1288 /* Sanity checks. */ 1289 if (ipproto == 0) 1290 return (EPROTONOSUPPORT); 1291 1292 /* 1293 * The protocol slot must not be occupied by another protocol 1294 * already. An index pointing to IPPROTO_RAW is unused. 1295 */ 1296 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1297 if (pr == NULL) 1298 return (EPFNOSUPPORT); 1299 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1300 return (EEXIST); 1301 1302 /* Find the protocol position in inetsw[] and set the index. */ 1303 for (pr = inetdomain.dom_protosw; 1304 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1305 if (pr->pr_domain->dom_family == PF_INET && 1306 pr->pr_protocol && pr->pr_protocol == ipproto) { 1307 /* Be careful to only index valid IP protocols. */ 1308 if (pr->pr_protocol < IPPROTO_MAX) { 1309 ip_protox[pr->pr_protocol] = pr - inetsw; 1310 return (0); 1311 } else 1312 return (EINVAL); 1313 } 1314 } 1315 return (EPROTONOSUPPORT); 1316 } 1317 1318 int 1319 ipproto_unregister(u_char ipproto) 1320 { 1321 struct protosw *pr; 1322 1323 /* Sanity checks. */ 1324 if (ipproto == 0) 1325 return (EPROTONOSUPPORT); 1326 1327 /* Check if the protocol was indeed registered. */ 1328 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1329 if (pr == NULL) 1330 return (EPFNOSUPPORT); 1331 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1332 return (ENOENT); 1333 1334 /* Reset the protocol slot to IPPROTO_RAW. */ 1335 ip_protox[ipproto] = pr - inetsw; 1336 return (0); 1337 } 1338 1339 /* 1340 * Given address of next destination (final or next hop), 1341 * return internet address info of interface to be used to get there. 1342 */ 1343 struct in_ifaddr * 1344 ip_rtaddr(struct in_addr dst, u_int fibnum) 1345 { 1346 struct route sro; 1347 struct sockaddr_in *sin; 1348 struct in_ifaddr *ifa; 1349 1350 bzero(&sro, sizeof(sro)); 1351 sin = (struct sockaddr_in *)&sro.ro_dst; 1352 sin->sin_family = AF_INET; 1353 sin->sin_len = sizeof(*sin); 1354 sin->sin_addr = dst; 1355 in_rtalloc_ign(&sro, 0, fibnum); 1356 1357 if (sro.ro_rt == NULL) 1358 return (NULL); 1359 1360 ifa = ifatoia(sro.ro_rt->rt_ifa); 1361 RTFREE(sro.ro_rt); 1362 return (ifa); 1363 } 1364 1365 u_char inetctlerrmap[PRC_NCMDS] = { 1366 0, 0, 0, 0, 1367 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1368 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1369 EMSGSIZE, EHOSTUNREACH, 0, 0, 1370 0, 0, EHOSTUNREACH, 0, 1371 ENOPROTOOPT, ECONNREFUSED 1372 }; 1373 1374 /* 1375 * Forward a packet. If some error occurs return the sender 1376 * an icmp packet. Note we can't always generate a meaningful 1377 * icmp message because icmp doesn't have a large enough repertoire 1378 * of codes and types. 1379 * 1380 * If not forwarding, just drop the packet. This could be confusing 1381 * if ipforwarding was zero but some routing protocol was advancing 1382 * us as a gateway to somewhere. However, we must let the routing 1383 * protocol deal with that. 1384 * 1385 * The srcrt parameter indicates whether the packet is being forwarded 1386 * via a source route. 1387 */ 1388 void 1389 ip_forward(struct mbuf *m, int srcrt) 1390 { 1391 INIT_VNET_INET(curvnet); 1392 struct ip *ip = mtod(m, struct ip *); 1393 struct in_ifaddr *ia; 1394 struct mbuf *mcopy; 1395 struct in_addr dest; 1396 struct route ro; 1397 int error, type = 0, code = 0, mtu = 0; 1398 1399 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1400 IPSTAT_INC(ips_cantforward); 1401 m_freem(m); 1402 return; 1403 } 1404 #ifdef IPSTEALTH 1405 if (!V_ipstealth) { 1406 #endif 1407 if (ip->ip_ttl <= IPTTLDEC) { 1408 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1409 0, 0); 1410 return; 1411 } 1412 #ifdef IPSTEALTH 1413 } 1414 #endif 1415 1416 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); 1417 #ifndef IPSEC 1418 /* 1419 * 'ia' may be NULL if there is no route for this destination. 1420 * In case of IPsec, Don't discard it just yet, but pass it to 1421 * ip_output in case of outgoing IPsec policy. 1422 */ 1423 if (!srcrt && ia == NULL) { 1424 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1425 return; 1426 } 1427 #endif 1428 1429 /* 1430 * Save the IP header and at most 8 bytes of the payload, 1431 * in case we need to generate an ICMP message to the src. 1432 * 1433 * XXX this can be optimized a lot by saving the data in a local 1434 * buffer on the stack (72 bytes at most), and only allocating the 1435 * mbuf if really necessary. The vast majority of the packets 1436 * are forwarded without having to send an ICMP back (either 1437 * because unnecessary, or because rate limited), so we are 1438 * really we are wasting a lot of work here. 1439 * 1440 * We don't use m_copy() because it might return a reference 1441 * to a shared cluster. Both this function and ip_output() 1442 * assume exclusive access to the IP header in `m', so any 1443 * data in a cluster may change before we reach icmp_error(). 1444 */ 1445 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1446 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1447 /* 1448 * It's probably ok if the pkthdr dup fails (because 1449 * the deep copy of the tag chain failed), but for now 1450 * be conservative and just discard the copy since 1451 * code below may some day want the tags. 1452 */ 1453 m_free(mcopy); 1454 mcopy = NULL; 1455 } 1456 if (mcopy != NULL) { 1457 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy)); 1458 mcopy->m_pkthdr.len = mcopy->m_len; 1459 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1460 } 1461 1462 #ifdef IPSTEALTH 1463 if (!V_ipstealth) { 1464 #endif 1465 ip->ip_ttl -= IPTTLDEC; 1466 #ifdef IPSTEALTH 1467 } 1468 #endif 1469 1470 /* 1471 * If forwarding packet using same interface that it came in on, 1472 * perhaps should send a redirect to sender to shortcut a hop. 1473 * Only send redirect if source is sending directly to us, 1474 * and if packet was not source routed (or has any options). 1475 * Also, don't send redirect if forwarding using a default route 1476 * or a route modified by a redirect. 1477 */ 1478 dest.s_addr = 0; 1479 if (!srcrt && V_ipsendredirects && 1480 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) { 1481 struct sockaddr_in *sin; 1482 struct rtentry *rt; 1483 1484 bzero(&ro, sizeof(ro)); 1485 sin = (struct sockaddr_in *)&ro.ro_dst; 1486 sin->sin_family = AF_INET; 1487 sin->sin_len = sizeof(*sin); 1488 sin->sin_addr = ip->ip_dst; 1489 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 1490 1491 rt = ro.ro_rt; 1492 1493 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1494 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1495 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1496 u_long src = ntohl(ip->ip_src.s_addr); 1497 1498 if (RTA(rt) && 1499 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1500 if (rt->rt_flags & RTF_GATEWAY) 1501 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1502 else 1503 dest.s_addr = ip->ip_dst.s_addr; 1504 /* Router requirements says to only send host redirects */ 1505 type = ICMP_REDIRECT; 1506 code = ICMP_REDIRECT_HOST; 1507 } 1508 } 1509 if (rt) 1510 RTFREE(rt); 1511 } 1512 1513 /* 1514 * Try to cache the route MTU from ip_output so we can consider it for 1515 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. 1516 */ 1517 bzero(&ro, sizeof(ro)); 1518 1519 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1520 1521 if (error == EMSGSIZE && ro.ro_rt) 1522 mtu = ro.ro_rt->rt_rmx.rmx_mtu; 1523 if (ro.ro_rt) 1524 RTFREE(ro.ro_rt); 1525 1526 if (error) 1527 IPSTAT_INC(ips_cantforward); 1528 else { 1529 IPSTAT_INC(ips_forward); 1530 if (type) 1531 IPSTAT_INC(ips_redirectsent); 1532 else { 1533 if (mcopy) 1534 m_freem(mcopy); 1535 return; 1536 } 1537 } 1538 if (mcopy == NULL) 1539 return; 1540 1541 switch (error) { 1542 1543 case 0: /* forwarded, but need redirect */ 1544 /* type, code set above */ 1545 break; 1546 1547 case ENETUNREACH: 1548 case EHOSTUNREACH: 1549 case ENETDOWN: 1550 case EHOSTDOWN: 1551 default: 1552 type = ICMP_UNREACH; 1553 code = ICMP_UNREACH_HOST; 1554 break; 1555 1556 case EMSGSIZE: 1557 type = ICMP_UNREACH; 1558 code = ICMP_UNREACH_NEEDFRAG; 1559 1560 #ifdef IPSEC 1561 /* 1562 * If IPsec is configured for this path, 1563 * override any possibly mtu value set by ip_output. 1564 */ 1565 mtu = ip_ipsec_mtu(m, mtu); 1566 #endif /* IPSEC */ 1567 /* 1568 * If the MTU was set before make sure we are below the 1569 * interface MTU. 1570 * If the MTU wasn't set before use the interface mtu or 1571 * fall back to the next smaller mtu step compared to the 1572 * current packet size. 1573 */ 1574 if (mtu != 0) { 1575 if (ia != NULL) 1576 mtu = min(mtu, ia->ia_ifp->if_mtu); 1577 } else { 1578 if (ia != NULL) 1579 mtu = ia->ia_ifp->if_mtu; 1580 else 1581 mtu = ip_next_mtu(ip->ip_len, 0); 1582 } 1583 IPSTAT_INC(ips_cantfrag); 1584 break; 1585 1586 case ENOBUFS: 1587 /* 1588 * A router should not generate ICMP_SOURCEQUENCH as 1589 * required in RFC1812 Requirements for IP Version 4 Routers. 1590 * Source quench could be a big problem under DoS attacks, 1591 * or if the underlying interface is rate-limited. 1592 * Those who need source quench packets may re-enable them 1593 * via the net.inet.ip.sendsourcequench sysctl. 1594 */ 1595 if (V_ip_sendsourcequench == 0) { 1596 m_freem(mcopy); 1597 return; 1598 } else { 1599 type = ICMP_SOURCEQUENCH; 1600 code = 0; 1601 } 1602 break; 1603 1604 case EACCES: /* ipfw denied packet */ 1605 m_freem(mcopy); 1606 return; 1607 } 1608 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1609 } 1610 1611 void 1612 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1613 struct mbuf *m) 1614 { 1615 INIT_VNET_NET(inp->inp_vnet); 1616 1617 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1618 struct bintime bt; 1619 1620 bintime(&bt); 1621 if (inp->inp_socket->so_options & SO_BINTIME) { 1622 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1623 SCM_BINTIME, SOL_SOCKET); 1624 if (*mp) 1625 mp = &(*mp)->m_next; 1626 } 1627 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1628 struct timeval tv; 1629 1630 bintime2timeval(&bt, &tv); 1631 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1632 SCM_TIMESTAMP, SOL_SOCKET); 1633 if (*mp) 1634 mp = &(*mp)->m_next; 1635 } 1636 } 1637 if (inp->inp_flags & INP_RECVDSTADDR) { 1638 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1639 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1640 if (*mp) 1641 mp = &(*mp)->m_next; 1642 } 1643 if (inp->inp_flags & INP_RECVTTL) { 1644 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1645 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1646 if (*mp) 1647 mp = &(*mp)->m_next; 1648 } 1649 #ifdef notyet 1650 /* XXX 1651 * Moving these out of udp_input() made them even more broken 1652 * than they already were. 1653 */ 1654 /* options were tossed already */ 1655 if (inp->inp_flags & INP_RECVOPTS) { 1656 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1657 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1658 if (*mp) 1659 mp = &(*mp)->m_next; 1660 } 1661 /* ip_srcroute doesn't do what we want here, need to fix */ 1662 if (inp->inp_flags & INP_RECVRETOPTS) { 1663 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1664 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1665 if (*mp) 1666 mp = &(*mp)->m_next; 1667 } 1668 #endif 1669 if (inp->inp_flags & INP_RECVIF) { 1670 struct ifnet *ifp; 1671 struct sdlbuf { 1672 struct sockaddr_dl sdl; 1673 u_char pad[32]; 1674 } sdlbuf; 1675 struct sockaddr_dl *sdp; 1676 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1677 1678 if (((ifp = m->m_pkthdr.rcvif)) 1679 && ( ifp->if_index && (ifp->if_index <= V_if_index))) { 1680 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1681 /* 1682 * Change our mind and don't try copy. 1683 */ 1684 if ((sdp->sdl_family != AF_LINK) 1685 || (sdp->sdl_len > sizeof(sdlbuf))) { 1686 goto makedummy; 1687 } 1688 bcopy(sdp, sdl2, sdp->sdl_len); 1689 } else { 1690 makedummy: 1691 sdl2->sdl_len 1692 = offsetof(struct sockaddr_dl, sdl_data[0]); 1693 sdl2->sdl_family = AF_LINK; 1694 sdl2->sdl_index = 0; 1695 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1696 } 1697 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1698 IP_RECVIF, IPPROTO_IP); 1699 if (*mp) 1700 mp = &(*mp)->m_next; 1701 } 1702 } 1703 1704 /* 1705 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1706 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1707 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1708 * compiled. 1709 */ 1710 int 1711 ip_rsvp_init(struct socket *so) 1712 { 1713 INIT_VNET_INET(so->so_vnet); 1714 1715 if (so->so_type != SOCK_RAW || 1716 so->so_proto->pr_protocol != IPPROTO_RSVP) 1717 return EOPNOTSUPP; 1718 1719 if (V_ip_rsvpd != NULL) 1720 return EADDRINUSE; 1721 1722 V_ip_rsvpd = so; 1723 /* 1724 * This may seem silly, but we need to be sure we don't over-increment 1725 * the RSVP counter, in case something slips up. 1726 */ 1727 if (!V_ip_rsvp_on) { 1728 V_ip_rsvp_on = 1; 1729 V_rsvp_on++; 1730 } 1731 1732 return 0; 1733 } 1734 1735 int 1736 ip_rsvp_done(void) 1737 { 1738 INIT_VNET_INET(curvnet); 1739 1740 V_ip_rsvpd = NULL; 1741 /* 1742 * This may seem silly, but we need to be sure we don't over-decrement 1743 * the RSVP counter, in case something slips up. 1744 */ 1745 if (V_ip_rsvp_on) { 1746 V_ip_rsvp_on = 0; 1747 V_rsvp_on--; 1748 } 1749 return 0; 1750 } 1751 1752 void 1753 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1754 { 1755 INIT_VNET_INET(curvnet); 1756 1757 if (rsvp_input_p) { /* call the real one if loaded */ 1758 rsvp_input_p(m, off); 1759 return; 1760 } 1761 1762 /* Can still get packets with rsvp_on = 0 if there is a local member 1763 * of the group to which the RSVP packet is addressed. But in this 1764 * case we want to throw the packet away. 1765 */ 1766 1767 if (!V_rsvp_on) { 1768 m_freem(m); 1769 return; 1770 } 1771 1772 if (V_ip_rsvpd != NULL) { 1773 rip_input(m, off); 1774 return; 1775 } 1776 /* Drop the packet */ 1777 m_freem(m); 1778 } 1779