ip_input.c revision 242079
1/*- 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 30 */ 31 32#include <sys/cdefs.h> 33__FBSDID("$FreeBSD: head/sys/netinet/ip_input.c 242079 2012-10-25 09:39:14Z ae $"); 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 41#include <sys/param.h> 42#include <sys/systm.h> 43#include <sys/mbuf.h> 44#include <sys/malloc.h> 45#include <sys/domain.h> 46#include <sys/protosw.h> 47#include <sys/socket.h> 48#include <sys/time.h> 49#include <sys/kernel.h> 50#include <sys/lock.h> 51#include <sys/rwlock.h> 52#include <sys/syslog.h> 53#include <sys/sysctl.h> 54 55#include <net/pfil.h> 56#include <net/if.h> 57#include <net/if_types.h> 58#include <net/if_var.h> 59#include <net/if_dl.h> 60#include <net/route.h> 61#include <net/netisr.h> 62#include <net/vnet.h> 63#include <net/flowtable.h> 64 65#include <netinet/in.h> 66#include <netinet/in_systm.h> 67#include <netinet/in_var.h> 68#include <netinet/ip.h> 69#include <netinet/in_pcb.h> 70#include <netinet/ip_var.h> 71#include <netinet/ip_fw.h> 72#include <netinet/ip_icmp.h> 73#include <netinet/ip_options.h> 74#include <machine/in_cksum.h> 75#include <netinet/ip_carp.h> 76#ifdef IPSEC 77#include <netinet/ip_ipsec.h> 78#endif /* IPSEC */ 79 80#include <sys/socketvar.h> 81 82#include <security/mac/mac_framework.h> 83 84#ifdef CTASSERT 85CTASSERT(sizeof(struct ip) == 20); 86#endif 87 88struct rwlock in_ifaddr_lock; 89RW_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock"); 90 91VNET_DEFINE(int, rsvp_on); 92 93VNET_DEFINE(int, ipforwarding); 94SYSCTL_VNET_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 95 &VNET_NAME(ipforwarding), 0, 96 "Enable IP forwarding between interfaces"); 97 98static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */ 99#define V_ipsendredirects VNET(ipsendredirects) 100SYSCTL_VNET_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 101 &VNET_NAME(ipsendredirects), 0, 102 "Enable sending IP redirects"); 103 104static VNET_DEFINE(int, ip_keepfaith); 105#define V_ip_keepfaith VNET(ip_keepfaith) 106SYSCTL_VNET_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 107 &VNET_NAME(ip_keepfaith), 0, 108 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 109 110static VNET_DEFINE(int, ip_sendsourcequench); 111#define V_ip_sendsourcequench VNET(ip_sendsourcequench) 112SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 113 &VNET_NAME(ip_sendsourcequench), 0, 114 "Enable the transmission of source quench packets"); 115 116VNET_DEFINE(int, ip_do_randomid); 117SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 118 &VNET_NAME(ip_do_randomid), 0, 119 "Assign random ip_id values"); 120 121/* 122 * XXX - Setting ip_checkinterface mostly implements the receive side of 123 * the Strong ES model described in RFC 1122, but since the routing table 124 * and transmit implementation do not implement the Strong ES model, 125 * setting this to 1 results in an odd hybrid. 126 * 127 * XXX - ip_checkinterface currently must be disabled if you use ipnat 128 * to translate the destination address to another local interface. 129 * 130 * XXX - ip_checkinterface must be disabled if you add IP aliases 131 * to the loopback interface instead of the interface where the 132 * packets for those addresses are received. 133 */ 134static VNET_DEFINE(int, ip_checkinterface); 135#define V_ip_checkinterface VNET(ip_checkinterface) 136SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 137 &VNET_NAME(ip_checkinterface), 0, 138 "Verify packet arrives on correct interface"); 139 140VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */ 141 142static struct netisr_handler ip_nh = { 143 .nh_name = "ip", 144 .nh_handler = ip_input, 145 .nh_proto = NETISR_IP, 146 .nh_policy = NETISR_POLICY_FLOW, 147}; 148 149extern struct domain inetdomain; 150extern struct protosw inetsw[]; 151u_char ip_protox[IPPROTO_MAX]; 152VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */ 153VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */ 154VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */ 155 156VNET_DEFINE(struct ipstat, ipstat); 157SYSCTL_VNET_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 158 &VNET_NAME(ipstat), ipstat, 159 "IP statistics (struct ipstat, netinet/ip_var.h)"); 160 161static VNET_DEFINE(uma_zone_t, ipq_zone); 162static VNET_DEFINE(TAILQ_HEAD(ipqhead, ipq), ipq[IPREASS_NHASH]); 163static struct mtx ipqlock; 164 165#define V_ipq_zone VNET(ipq_zone) 166#define V_ipq VNET(ipq) 167 168#define IPQ_LOCK() mtx_lock(&ipqlock) 169#define IPQ_UNLOCK() mtx_unlock(&ipqlock) 170#define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 171#define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 172 173static void maxnipq_update(void); 174static void ipq_zone_change(void *); 175static void ip_drain_locked(void); 176 177static VNET_DEFINE(int, maxnipq); /* Administrative limit on # reass queues. */ 178static VNET_DEFINE(int, nipq); /* Total # of reass queues */ 179#define V_maxnipq VNET(maxnipq) 180#define V_nipq VNET(nipq) 181SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, 182 &VNET_NAME(nipq), 0, 183 "Current number of IPv4 fragment reassembly queue entries"); 184 185static VNET_DEFINE(int, maxfragsperpacket); 186#define V_maxfragsperpacket VNET(maxfragsperpacket) 187SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 188 &VNET_NAME(maxfragsperpacket), 0, 189 "Maximum number of IPv4 fragments allowed per packet"); 190 191#ifdef IPCTL_DEFMTU 192SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 193 &ip_mtu, 0, "Default MTU"); 194#endif 195 196#ifdef IPSTEALTH 197VNET_DEFINE(int, ipstealth); 198SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 199 &VNET_NAME(ipstealth), 0, 200 "IP stealth mode, no TTL decrementation on forwarding"); 201#endif 202 203#ifdef FLOWTABLE 204static VNET_DEFINE(int, ip_output_flowtable_size) = 2048; 205VNET_DEFINE(struct flowtable *, ip_ft); 206#define V_ip_output_flowtable_size VNET(ip_output_flowtable_size) 207 208SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, output_flowtable_size, CTLFLAG_RDTUN, 209 &VNET_NAME(ip_output_flowtable_size), 2048, 210 "number of entries in the per-cpu output flow caches"); 211#endif 212 213static void ip_freef(struct ipqhead *, struct ipq *); 214 215/* 216 * Kernel module interface for updating ipstat. The argument is an index 217 * into ipstat treated as an array of u_long. While this encodes the general 218 * layout of ipstat into the caller, it doesn't encode its location, so that 219 * future changes to add, for example, per-CPU stats support won't cause 220 * binary compatibility problems for kernel modules. 221 */ 222void 223kmod_ipstat_inc(int statnum) 224{ 225 226 (*((u_long *)&V_ipstat + statnum))++; 227} 228 229void 230kmod_ipstat_dec(int statnum) 231{ 232 233 (*((u_long *)&V_ipstat + statnum))--; 234} 235 236static int 237sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 238{ 239 int error, qlimit; 240 241 netisr_getqlimit(&ip_nh, &qlimit); 242 error = sysctl_handle_int(oidp, &qlimit, 0, req); 243 if (error || !req->newptr) 244 return (error); 245 if (qlimit < 1) 246 return (EINVAL); 247 return (netisr_setqlimit(&ip_nh, qlimit)); 248} 249SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, 250 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I", 251 "Maximum size of the IP input queue"); 252 253static int 254sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) 255{ 256 u_int64_t qdrops_long; 257 int error, qdrops; 258 259 netisr_getqdrops(&ip_nh, &qdrops_long); 260 qdrops = qdrops_long; 261 error = sysctl_handle_int(oidp, &qdrops, 0, req); 262 if (error || !req->newptr) 263 return (error); 264 if (qdrops != 0) 265 return (EINVAL); 266 netisr_clearqdrops(&ip_nh); 267 return (0); 268} 269 270SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, 271 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I", 272 "Number of packets dropped from the IP input queue"); 273 274/* 275 * IP initialization: fill in IP protocol switch table. 276 * All protocols not implemented in kernel go to raw IP protocol handler. 277 */ 278void 279ip_init(void) 280{ 281 struct protosw *pr; 282 int i; 283 284 V_ip_id = time_second & 0xffff; 285 286 TAILQ_INIT(&V_in_ifaddrhead); 287 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 288 289 /* Initialize IP reassembly queue. */ 290 for (i = 0; i < IPREASS_NHASH; i++) 291 TAILQ_INIT(&V_ipq[i]); 292 V_maxnipq = nmbclusters / 32; 293 V_maxfragsperpacket = 16; 294 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 295 NULL, UMA_ALIGN_PTR, 0); 296 maxnipq_update(); 297 298 /* Initialize packet filter hooks. */ 299 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF; 300 V_inet_pfil_hook.ph_af = AF_INET; 301 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0) 302 printf("%s: WARNING: unable to register pfil hook, " 303 "error %d\n", __func__, i); 304 305#ifdef FLOWTABLE 306 if (TUNABLE_INT_FETCH("net.inet.ip.output_flowtable_size", 307 &V_ip_output_flowtable_size)) { 308 if (V_ip_output_flowtable_size < 256) 309 V_ip_output_flowtable_size = 256; 310 if (!powerof2(V_ip_output_flowtable_size)) { 311 printf("flowtable must be power of 2 size\n"); 312 V_ip_output_flowtable_size = 2048; 313 } 314 } else { 315 /* 316 * round up to the next power of 2 317 */ 318 V_ip_output_flowtable_size = 1 << fls((1024 + maxusers * 64)-1); 319 } 320 V_ip_ft = flowtable_alloc("ipv4", V_ip_output_flowtable_size, FL_PCPU); 321#endif 322 323 /* Skip initialization of globals for non-default instances. */ 324 if (!IS_DEFAULT_VNET(curvnet)) 325 return; 326 327 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 328 if (pr == NULL) 329 panic("ip_init: PF_INET not found"); 330 331 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 332 for (i = 0; i < IPPROTO_MAX; i++) 333 ip_protox[i] = pr - inetsw; 334 /* 335 * Cycle through IP protocols and put them into the appropriate place 336 * in ip_protox[]. 337 */ 338 for (pr = inetdomain.dom_protosw; 339 pr < inetdomain.dom_protoswNPROTOSW; pr++) 340 if (pr->pr_domain->dom_family == PF_INET && 341 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 342 /* Be careful to only index valid IP protocols. */ 343 if (pr->pr_protocol < IPPROTO_MAX) 344 ip_protox[pr->pr_protocol] = pr - inetsw; 345 } 346 347 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, 348 NULL, EVENTHANDLER_PRI_ANY); 349 350 /* Initialize various other remaining things. */ 351 IPQ_LOCK_INIT(); 352 netisr_register(&ip_nh); 353} 354 355#ifdef VIMAGE 356void 357ip_destroy(void) 358{ 359 360 /* Cleanup in_ifaddr hash table; should be empty. */ 361 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask); 362 363 IPQ_LOCK(); 364 ip_drain_locked(); 365 IPQ_UNLOCK(); 366 367 uma_zdestroy(V_ipq_zone); 368} 369#endif 370 371/* 372 * Ip input routine. Checksum and byte swap header. If fragmented 373 * try to reassemble. Process options. Pass to next level. 374 */ 375void 376ip_input(struct mbuf *m) 377{ 378 struct ip *ip = NULL; 379 struct in_ifaddr *ia = NULL; 380 struct ifaddr *ifa; 381 struct ifnet *ifp; 382 int checkif, hlen = 0; 383 uint16_t sum, ip_len; 384 int dchg = 0; /* dest changed after fw */ 385 struct in_addr odst; /* original dst address */ 386 387 M_ASSERTPKTHDR(m); 388 389 if (m->m_flags & M_FASTFWD_OURS) { 390 m->m_flags &= ~M_FASTFWD_OURS; 391 /* Set up some basics that will be used later. */ 392 ip = mtod(m, struct ip *); 393 hlen = ip->ip_hl << 2; 394 ip_len = ntohs(ip->ip_len); 395 goto ours; 396 } 397 398 IPSTAT_INC(ips_total); 399 400 if (m->m_pkthdr.len < sizeof(struct ip)) 401 goto tooshort; 402 403 if (m->m_len < sizeof (struct ip) && 404 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 405 IPSTAT_INC(ips_toosmall); 406 return; 407 } 408 ip = mtod(m, struct ip *); 409 410 if (ip->ip_v != IPVERSION) { 411 IPSTAT_INC(ips_badvers); 412 goto bad; 413 } 414 415 hlen = ip->ip_hl << 2; 416 if (hlen < sizeof(struct ip)) { /* minimum header length */ 417 IPSTAT_INC(ips_badhlen); 418 goto bad; 419 } 420 if (hlen > m->m_len) { 421 if ((m = m_pullup(m, hlen)) == NULL) { 422 IPSTAT_INC(ips_badhlen); 423 return; 424 } 425 ip = mtod(m, struct ip *); 426 } 427 428 /* 127/8 must not appear on wire - RFC1122 */ 429 ifp = m->m_pkthdr.rcvif; 430 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 431 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 432 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 433 IPSTAT_INC(ips_badaddr); 434 goto bad; 435 } 436 } 437 438 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 439 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 440 } else { 441 if (hlen == sizeof(struct ip)) { 442 sum = in_cksum_hdr(ip); 443 } else { 444 sum = in_cksum(m, hlen); 445 } 446 } 447 if (sum) { 448 IPSTAT_INC(ips_badsum); 449 goto bad; 450 } 451 452#ifdef ALTQ 453 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 454 /* packet is dropped by traffic conditioner */ 455 return; 456#endif 457 458 ip_len = ntohs(ip->ip_len); 459 if (ip_len < hlen) { 460 IPSTAT_INC(ips_badlen); 461 goto bad; 462 } 463 464 /* 465 * Check that the amount of data in the buffers 466 * is as at least much as the IP header would have us expect. 467 * Trim mbufs if longer than we expect. 468 * Drop packet if shorter than we expect. 469 */ 470 if (m->m_pkthdr.len < ip_len) { 471tooshort: 472 IPSTAT_INC(ips_tooshort); 473 goto bad; 474 } 475 if (m->m_pkthdr.len > ip_len) { 476 if (m->m_len == m->m_pkthdr.len) { 477 m->m_len = ip_len; 478 m->m_pkthdr.len = ip_len; 479 } else 480 m_adj(m, ip_len - m->m_pkthdr.len); 481 } 482#ifdef IPSEC 483 /* 484 * Bypass packet filtering for packets previously handled by IPsec. 485 */ 486 if (ip_ipsec_filtertunnel(m)) 487 goto passin; 488#endif /* IPSEC */ 489 490 /* 491 * Run through list of hooks for input packets. 492 * 493 * NB: Beware of the destination address changing (e.g. 494 * by NAT rewriting). When this happens, tell 495 * ip_forward to do the right thing. 496 */ 497 498 /* Jump over all PFIL processing if hooks are not active. */ 499 if (!PFIL_HOOKED(&V_inet_pfil_hook)) 500 goto passin; 501 502 odst = ip->ip_dst; 503 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0) 504 return; 505 if (m == NULL) /* consumed by filter */ 506 return; 507 508 ip = mtod(m, struct ip *); 509 dchg = (odst.s_addr != ip->ip_dst.s_addr); 510 ifp = m->m_pkthdr.rcvif; 511 512 if (V_pfilforward == 0) 513 goto passin; 514 515 if (m->m_flags & M_FASTFWD_OURS) { 516 m->m_flags &= ~M_FASTFWD_OURS; 517 goto ours; 518 } 519 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 520 /* 521 * Directly ship the packet on. This allows forwarding 522 * packets originally destined to us to some other directly 523 * connected host. 524 */ 525 ip_forward(m, dchg); 526 return; 527 } 528 529passin: 530 531 /* 532 * Process options and, if not destined for us, 533 * ship it on. ip_dooptions returns 1 when an 534 * error was detected (causing an icmp message 535 * to be sent and the original packet to be freed). 536 */ 537 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 538 return; 539 540 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 541 * matter if it is destined to another node, or whether it is 542 * a multicast one, RSVP wants it! and prevents it from being forwarded 543 * anywhere else. Also checks if the rsvp daemon is running before 544 * grabbing the packet. 545 */ 546 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 547 goto ours; 548 549 /* 550 * Check our list of addresses, to see if the packet is for us. 551 * If we don't have any addresses, assume any unicast packet 552 * we receive might be for us (and let the upper layers deal 553 * with it). 554 */ 555 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 556 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 557 goto ours; 558 559 /* 560 * Enable a consistency check between the destination address 561 * and the arrival interface for a unicast packet (the RFC 1122 562 * strong ES model) if IP forwarding is disabled and the packet 563 * is not locally generated and the packet is not subject to 564 * 'ipfw fwd'. 565 * 566 * XXX - Checking also should be disabled if the destination 567 * address is ipnat'ed to a different interface. 568 * 569 * XXX - Checking is incompatible with IP aliases added 570 * to the loopback interface instead of the interface where 571 * the packets are received. 572 * 573 * XXX - This is the case for carp vhost IPs as well so we 574 * insert a workaround. If the packet got here, we already 575 * checked with carp_iamatch() and carp_forus(). 576 */ 577 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 578 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && 579 ifp->if_carp == NULL && (dchg == 0); 580 581 /* 582 * Check for exact addresses in the hash bucket. 583 */ 584 /* IN_IFADDR_RLOCK(); */ 585 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 586 /* 587 * If the address matches, verify that the packet 588 * arrived via the correct interface if checking is 589 * enabled. 590 */ 591 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 592 (!checkif || ia->ia_ifp == ifp)) { 593 ifa_ref(&ia->ia_ifa); 594 /* IN_IFADDR_RUNLOCK(); */ 595 goto ours; 596 } 597 } 598 /* IN_IFADDR_RUNLOCK(); */ 599 600 /* 601 * Check for broadcast addresses. 602 * 603 * Only accept broadcast packets that arrive via the matching 604 * interface. Reception of forwarded directed broadcasts would 605 * be handled via ip_forward() and ether_output() with the loopback 606 * into the stack for SIMPLEX interfaces handled by ether_output(). 607 */ 608 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 609 IF_ADDR_RLOCK(ifp); 610 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 611 if (ifa->ifa_addr->sa_family != AF_INET) 612 continue; 613 ia = ifatoia(ifa); 614 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 615 ip->ip_dst.s_addr) { 616 ifa_ref(ifa); 617 IF_ADDR_RUNLOCK(ifp); 618 goto ours; 619 } 620#ifdef BOOTP_COMPAT 621 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { 622 ifa_ref(ifa); 623 IF_ADDR_RUNLOCK(ifp); 624 goto ours; 625 } 626#endif 627 } 628 IF_ADDR_RUNLOCK(ifp); 629 ia = NULL; 630 } 631 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 632 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 633 IPSTAT_INC(ips_cantforward); 634 m_freem(m); 635 return; 636 } 637 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 638 if (V_ip_mrouter) { 639 /* 640 * If we are acting as a multicast router, all 641 * incoming multicast packets are passed to the 642 * kernel-level multicast forwarding function. 643 * The packet is returned (relatively) intact; if 644 * ip_mforward() returns a non-zero value, the packet 645 * must be discarded, else it may be accepted below. 646 */ 647 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 648 IPSTAT_INC(ips_cantforward); 649 m_freem(m); 650 return; 651 } 652 653 /* 654 * The process-level routing daemon needs to receive 655 * all multicast IGMP packets, whether or not this 656 * host belongs to their destination groups. 657 */ 658 if (ip->ip_p == IPPROTO_IGMP) 659 goto ours; 660 IPSTAT_INC(ips_forward); 661 } 662 /* 663 * Assume the packet is for us, to avoid prematurely taking 664 * a lock on the in_multi hash. Protocols must perform 665 * their own filtering and update statistics accordingly. 666 */ 667 goto ours; 668 } 669 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 670 goto ours; 671 if (ip->ip_dst.s_addr == INADDR_ANY) 672 goto ours; 673 674 /* 675 * FAITH(Firewall Aided Internet Translator) 676 */ 677 if (ifp && ifp->if_type == IFT_FAITH) { 678 if (V_ip_keepfaith) { 679 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 680 goto ours; 681 } 682 m_freem(m); 683 return; 684 } 685 686 /* 687 * Not for us; forward if possible and desirable. 688 */ 689 if (V_ipforwarding == 0) { 690 IPSTAT_INC(ips_cantforward); 691 m_freem(m); 692 } else { 693#ifdef IPSEC 694 if (ip_ipsec_fwd(m)) 695 goto bad; 696#endif /* IPSEC */ 697 ip_forward(m, dchg); 698 } 699 return; 700 701ours: 702#ifdef IPSTEALTH 703 /* 704 * IPSTEALTH: Process non-routing options only 705 * if the packet is destined for us. 706 */ 707 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) { 708 if (ia != NULL) 709 ifa_free(&ia->ia_ifa); 710 return; 711 } 712#endif /* IPSTEALTH */ 713 714 /* Count the packet in the ip address stats */ 715 if (ia != NULL) { 716 ia->ia_ifa.if_ipackets++; 717 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 718 ifa_free(&ia->ia_ifa); 719 } 720 721 /* 722 * Attempt reassembly; if it succeeds, proceed. 723 * ip_reass() will return a different mbuf. 724 */ 725 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) { 726 m = ip_reass(m); 727 if (m == NULL) 728 return; 729 ip = mtod(m, struct ip *); 730 /* Get the header length of the reassembled packet */ 731 hlen = ip->ip_hl << 2; 732 } 733 734#ifdef IPSEC 735 /* 736 * enforce IPsec policy checking if we are seeing last header. 737 * note that we do not visit this with protocols with pcb layer 738 * code - like udp/tcp/raw ip. 739 */ 740 if (ip_ipsec_input(m)) 741 goto bad; 742#endif /* IPSEC */ 743 744 /* 745 * Switch out to protocol's input routine. 746 */ 747 IPSTAT_INC(ips_delivered); 748 749 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 750 return; 751bad: 752 m_freem(m); 753} 754 755/* 756 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 757 * max has slightly different semantics than the sysctl, for historical 758 * reasons. 759 */ 760static void 761maxnipq_update(void) 762{ 763 764 /* 765 * -1 for unlimited allocation. 766 */ 767 if (V_maxnipq < 0) 768 uma_zone_set_max(V_ipq_zone, 0); 769 /* 770 * Positive number for specific bound. 771 */ 772 if (V_maxnipq > 0) 773 uma_zone_set_max(V_ipq_zone, V_maxnipq); 774 /* 775 * Zero specifies no further fragment queue allocation -- set the 776 * bound very low, but rely on implementation elsewhere to actually 777 * prevent allocation and reclaim current queues. 778 */ 779 if (V_maxnipq == 0) 780 uma_zone_set_max(V_ipq_zone, 1); 781} 782 783static void 784ipq_zone_change(void *tag) 785{ 786 787 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) { 788 V_maxnipq = nmbclusters / 32; 789 maxnipq_update(); 790 } 791} 792 793static int 794sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 795{ 796 int error, i; 797 798 i = V_maxnipq; 799 error = sysctl_handle_int(oidp, &i, 0, req); 800 if (error || !req->newptr) 801 return (error); 802 803 /* 804 * XXXRW: Might be a good idea to sanity check the argument and place 805 * an extreme upper bound. 806 */ 807 if (i < -1) 808 return (EINVAL); 809 V_maxnipq = i; 810 maxnipq_update(); 811 return (0); 812} 813 814SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 815 NULL, 0, sysctl_maxnipq, "I", 816 "Maximum number of IPv4 fragment reassembly queue entries"); 817 818/* 819 * Take incoming datagram fragment and try to reassemble it into 820 * whole datagram. If the argument is the first fragment or one 821 * in between the function will return NULL and store the mbuf 822 * in the fragment chain. If the argument is the last fragment 823 * the packet will be reassembled and the pointer to the new 824 * mbuf returned for further processing. Only m_tags attached 825 * to the first packet/fragment are preserved. 826 * The IP header is *NOT* adjusted out of iplen. 827 */ 828struct mbuf * 829ip_reass(struct mbuf *m) 830{ 831 struct ip *ip; 832 struct mbuf *p, *q, *nq, *t; 833 struct ipq *fp = NULL; 834 struct ipqhead *head; 835 int i, hlen, next; 836 u_int8_t ecn, ecn0; 837 u_short hash; 838 839 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 840 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) { 841 IPSTAT_INC(ips_fragments); 842 IPSTAT_INC(ips_fragdropped); 843 m_freem(m); 844 return (NULL); 845 } 846 847 ip = mtod(m, struct ip *); 848 hlen = ip->ip_hl << 2; 849 850 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 851 head = &V_ipq[hash]; 852 IPQ_LOCK(); 853 854 /* 855 * Look for queue of fragments 856 * of this datagram. 857 */ 858 TAILQ_FOREACH(fp, head, ipq_list) 859 if (ip->ip_id == fp->ipq_id && 860 ip->ip_src.s_addr == fp->ipq_src.s_addr && 861 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 862#ifdef MAC 863 mac_ipq_match(m, fp) && 864#endif 865 ip->ip_p == fp->ipq_p) 866 goto found; 867 868 fp = NULL; 869 870 /* 871 * Attempt to trim the number of allocated fragment queues if it 872 * exceeds the administrative limit. 873 */ 874 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) { 875 /* 876 * drop something from the tail of the current queue 877 * before proceeding further 878 */ 879 struct ipq *q = TAILQ_LAST(head, ipqhead); 880 if (q == NULL) { /* gak */ 881 for (i = 0; i < IPREASS_NHASH; i++) { 882 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead); 883 if (r) { 884 IPSTAT_ADD(ips_fragtimeout, 885 r->ipq_nfrags); 886 ip_freef(&V_ipq[i], r); 887 break; 888 } 889 } 890 } else { 891 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags); 892 ip_freef(head, q); 893 } 894 } 895 896found: 897 /* 898 * Adjust ip_len to not reflect header, 899 * convert offset of this to bytes. 900 */ 901 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 902 if (ip->ip_off & htons(IP_MF)) { 903 /* 904 * Make sure that fragments have a data length 905 * that's a non-zero multiple of 8 bytes. 906 */ 907 if (ip->ip_len == htons(0) || (ntohs(ip->ip_len) & 0x7) != 0) { 908 IPSTAT_INC(ips_toosmall); /* XXX */ 909 goto dropfrag; 910 } 911 m->m_flags |= M_FRAG; 912 } else 913 m->m_flags &= ~M_FRAG; 914 ip->ip_off = htons(ntohs(ip->ip_off) << 3); 915 916 /* 917 * Attempt reassembly; if it succeeds, proceed. 918 * ip_reass() will return a different mbuf. 919 */ 920 IPSTAT_INC(ips_fragments); 921 m->m_pkthdr.header = ip; 922 923 /* Previous ip_reass() started here. */ 924 /* 925 * Presence of header sizes in mbufs 926 * would confuse code below. 927 */ 928 m->m_data += hlen; 929 m->m_len -= hlen; 930 931 /* 932 * If first fragment to arrive, create a reassembly queue. 933 */ 934 if (fp == NULL) { 935 fp = uma_zalloc(V_ipq_zone, M_NOWAIT); 936 if (fp == NULL) 937 goto dropfrag; 938#ifdef MAC 939 if (mac_ipq_init(fp, M_NOWAIT) != 0) { 940 uma_zfree(V_ipq_zone, fp); 941 fp = NULL; 942 goto dropfrag; 943 } 944 mac_ipq_create(m, fp); 945#endif 946 TAILQ_INSERT_HEAD(head, fp, ipq_list); 947 V_nipq++; 948 fp->ipq_nfrags = 1; 949 fp->ipq_ttl = IPFRAGTTL; 950 fp->ipq_p = ip->ip_p; 951 fp->ipq_id = ip->ip_id; 952 fp->ipq_src = ip->ip_src; 953 fp->ipq_dst = ip->ip_dst; 954 fp->ipq_frags = m; 955 m->m_nextpkt = NULL; 956 goto done; 957 } else { 958 fp->ipq_nfrags++; 959#ifdef MAC 960 mac_ipq_update(m, fp); 961#endif 962 } 963 964#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 965 966 /* 967 * Handle ECN by comparing this segment with the first one; 968 * if CE is set, do not lose CE. 969 * drop if CE and not-ECT are mixed for the same packet. 970 */ 971 ecn = ip->ip_tos & IPTOS_ECN_MASK; 972 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 973 if (ecn == IPTOS_ECN_CE) { 974 if (ecn0 == IPTOS_ECN_NOTECT) 975 goto dropfrag; 976 if (ecn0 != IPTOS_ECN_CE) 977 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 978 } 979 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 980 goto dropfrag; 981 982 /* 983 * Find a segment which begins after this one does. 984 */ 985 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 986 if (ntohs(GETIP(q)->ip_off) > ntohs(ip->ip_off)) 987 break; 988 989 /* 990 * If there is a preceding segment, it may provide some of 991 * our data already. If so, drop the data from the incoming 992 * segment. If it provides all of our data, drop us, otherwise 993 * stick new segment in the proper place. 994 * 995 * If some of the data is dropped from the preceding 996 * segment, then it's checksum is invalidated. 997 */ 998 if (p) { 999 i = ntohs(GETIP(p)->ip_off) + ntohs(GETIP(p)->ip_len) - 1000 ntohs(ip->ip_off); 1001 if (i > 0) { 1002 if (i >= ntohs(ip->ip_len)) 1003 goto dropfrag; 1004 m_adj(m, i); 1005 m->m_pkthdr.csum_flags = 0; 1006 ip->ip_off = htons(ntohs(ip->ip_off) + i); 1007 ip->ip_len = htons(ntohs(ip->ip_len) - i); 1008 } 1009 m->m_nextpkt = p->m_nextpkt; 1010 p->m_nextpkt = m; 1011 } else { 1012 m->m_nextpkt = fp->ipq_frags; 1013 fp->ipq_frags = m; 1014 } 1015 1016 /* 1017 * While we overlap succeeding segments trim them or, 1018 * if they are completely covered, dequeue them. 1019 */ 1020 for (; q != NULL && ntohs(ip->ip_off) + ntohs(ip->ip_len) > 1021 ntohs(GETIP(q)->ip_off); q = nq) { 1022 i = (ntohs(ip->ip_off) + ntohs(ip->ip_len)) - 1023 ntohs(GETIP(q)->ip_off); 1024 if (i < ntohs(GETIP(q)->ip_len)) { 1025 GETIP(q)->ip_len = htons(ntohs(GETIP(q)->ip_len) - i); 1026 GETIP(q)->ip_off = htons(ntohs(GETIP(q)->ip_off) + i); 1027 m_adj(q, i); 1028 q->m_pkthdr.csum_flags = 0; 1029 break; 1030 } 1031 nq = q->m_nextpkt; 1032 m->m_nextpkt = nq; 1033 IPSTAT_INC(ips_fragdropped); 1034 fp->ipq_nfrags--; 1035 m_freem(q); 1036 } 1037 1038 /* 1039 * Check for complete reassembly and perform frag per packet 1040 * limiting. 1041 * 1042 * Frag limiting is performed here so that the nth frag has 1043 * a chance to complete the packet before we drop the packet. 1044 * As a result, n+1 frags are actually allowed per packet, but 1045 * only n will ever be stored. (n = maxfragsperpacket.) 1046 * 1047 */ 1048 next = 0; 1049 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1050 if (ntohs(GETIP(q)->ip_off) != next) { 1051 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1052 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1053 ip_freef(head, fp); 1054 } 1055 goto done; 1056 } 1057 next += ntohs(GETIP(q)->ip_len); 1058 } 1059 /* Make sure the last packet didn't have the IP_MF flag */ 1060 if (p->m_flags & M_FRAG) { 1061 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1062 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1063 ip_freef(head, fp); 1064 } 1065 goto done; 1066 } 1067 1068 /* 1069 * Reassembly is complete. Make sure the packet is a sane size. 1070 */ 1071 q = fp->ipq_frags; 1072 ip = GETIP(q); 1073 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1074 IPSTAT_INC(ips_toolong); 1075 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1076 ip_freef(head, fp); 1077 goto done; 1078 } 1079 1080 /* 1081 * Concatenate fragments. 1082 */ 1083 m = q; 1084 t = m->m_next; 1085 m->m_next = NULL; 1086 m_cat(m, t); 1087 nq = q->m_nextpkt; 1088 q->m_nextpkt = NULL; 1089 for (q = nq; q != NULL; q = nq) { 1090 nq = q->m_nextpkt; 1091 q->m_nextpkt = NULL; 1092 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1093 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1094 m_cat(m, q); 1095 } 1096 /* 1097 * In order to do checksumming faster we do 'end-around carry' here 1098 * (and not in for{} loop), though it implies we are not going to 1099 * reassemble more than 64k fragments. 1100 */ 1101 m->m_pkthdr.csum_data = 1102 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1103#ifdef MAC 1104 mac_ipq_reassemble(fp, m); 1105 mac_ipq_destroy(fp); 1106#endif 1107 1108 /* 1109 * Create header for new ip packet by modifying header of first 1110 * packet; dequeue and discard fragment reassembly header. 1111 * Make header visible. 1112 */ 1113 ip->ip_len = htons((ip->ip_hl << 2) + next); 1114 ip->ip_src = fp->ipq_src; 1115 ip->ip_dst = fp->ipq_dst; 1116 TAILQ_REMOVE(head, fp, ipq_list); 1117 V_nipq--; 1118 uma_zfree(V_ipq_zone, fp); 1119 m->m_len += (ip->ip_hl << 2); 1120 m->m_data -= (ip->ip_hl << 2); 1121 /* some debugging cruft by sklower, below, will go away soon */ 1122 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1123 m_fixhdr(m); 1124 IPSTAT_INC(ips_reassembled); 1125 IPQ_UNLOCK(); 1126 return (m); 1127 1128dropfrag: 1129 IPSTAT_INC(ips_fragdropped); 1130 if (fp != NULL) 1131 fp->ipq_nfrags--; 1132 m_freem(m); 1133done: 1134 IPQ_UNLOCK(); 1135 return (NULL); 1136 1137#undef GETIP 1138} 1139 1140/* 1141 * Free a fragment reassembly header and all 1142 * associated datagrams. 1143 */ 1144static void 1145ip_freef(struct ipqhead *fhp, struct ipq *fp) 1146{ 1147 struct mbuf *q; 1148 1149 IPQ_LOCK_ASSERT(); 1150 1151 while (fp->ipq_frags) { 1152 q = fp->ipq_frags; 1153 fp->ipq_frags = q->m_nextpkt; 1154 m_freem(q); 1155 } 1156 TAILQ_REMOVE(fhp, fp, ipq_list); 1157 uma_zfree(V_ipq_zone, fp); 1158 V_nipq--; 1159} 1160 1161/* 1162 * IP timer processing; 1163 * if a timer expires on a reassembly 1164 * queue, discard it. 1165 */ 1166void 1167ip_slowtimo(void) 1168{ 1169 VNET_ITERATOR_DECL(vnet_iter); 1170 struct ipq *fp; 1171 int i; 1172 1173 VNET_LIST_RLOCK_NOSLEEP(); 1174 IPQ_LOCK(); 1175 VNET_FOREACH(vnet_iter) { 1176 CURVNET_SET(vnet_iter); 1177 for (i = 0; i < IPREASS_NHASH; i++) { 1178 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) { 1179 struct ipq *fpp; 1180 1181 fpp = fp; 1182 fp = TAILQ_NEXT(fp, ipq_list); 1183 if(--fpp->ipq_ttl == 0) { 1184 IPSTAT_ADD(ips_fragtimeout, 1185 fpp->ipq_nfrags); 1186 ip_freef(&V_ipq[i], fpp); 1187 } 1188 } 1189 } 1190 /* 1191 * If we are over the maximum number of fragments 1192 * (due to the limit being lowered), drain off 1193 * enough to get down to the new limit. 1194 */ 1195 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) { 1196 for (i = 0; i < IPREASS_NHASH; i++) { 1197 while (V_nipq > V_maxnipq && 1198 !TAILQ_EMPTY(&V_ipq[i])) { 1199 IPSTAT_ADD(ips_fragdropped, 1200 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1201 ip_freef(&V_ipq[i], 1202 TAILQ_FIRST(&V_ipq[i])); 1203 } 1204 } 1205 } 1206 CURVNET_RESTORE(); 1207 } 1208 IPQ_UNLOCK(); 1209 VNET_LIST_RUNLOCK_NOSLEEP(); 1210} 1211 1212/* 1213 * Drain off all datagram fragments. 1214 */ 1215static void 1216ip_drain_locked(void) 1217{ 1218 int i; 1219 1220 IPQ_LOCK_ASSERT(); 1221 1222 for (i = 0; i < IPREASS_NHASH; i++) { 1223 while(!TAILQ_EMPTY(&V_ipq[i])) { 1224 IPSTAT_ADD(ips_fragdropped, 1225 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1226 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); 1227 } 1228 } 1229} 1230 1231void 1232ip_drain(void) 1233{ 1234 VNET_ITERATOR_DECL(vnet_iter); 1235 1236 VNET_LIST_RLOCK_NOSLEEP(); 1237 IPQ_LOCK(); 1238 VNET_FOREACH(vnet_iter) { 1239 CURVNET_SET(vnet_iter); 1240 ip_drain_locked(); 1241 CURVNET_RESTORE(); 1242 } 1243 IPQ_UNLOCK(); 1244 VNET_LIST_RUNLOCK_NOSLEEP(); 1245 in_rtqdrain(); 1246} 1247 1248/* 1249 * The protocol to be inserted into ip_protox[] must be already registered 1250 * in inetsw[], either statically or through pf_proto_register(). 1251 */ 1252int 1253ipproto_register(short ipproto) 1254{ 1255 struct protosw *pr; 1256 1257 /* Sanity checks. */ 1258 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 1259 return (EPROTONOSUPPORT); 1260 1261 /* 1262 * The protocol slot must not be occupied by another protocol 1263 * already. An index pointing to IPPROTO_RAW is unused. 1264 */ 1265 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1266 if (pr == NULL) 1267 return (EPFNOSUPPORT); 1268 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1269 return (EEXIST); 1270 1271 /* Find the protocol position in inetsw[] and set the index. */ 1272 for (pr = inetdomain.dom_protosw; 1273 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1274 if (pr->pr_domain->dom_family == PF_INET && 1275 pr->pr_protocol && pr->pr_protocol == ipproto) { 1276 ip_protox[pr->pr_protocol] = pr - inetsw; 1277 return (0); 1278 } 1279 } 1280 return (EPROTONOSUPPORT); 1281} 1282 1283int 1284ipproto_unregister(short ipproto) 1285{ 1286 struct protosw *pr; 1287 1288 /* Sanity checks. */ 1289 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 1290 return (EPROTONOSUPPORT); 1291 1292 /* Check if the protocol was indeed registered. */ 1293 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1294 if (pr == NULL) 1295 return (EPFNOSUPPORT); 1296 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1297 return (ENOENT); 1298 1299 /* Reset the protocol slot to IPPROTO_RAW. */ 1300 ip_protox[ipproto] = pr - inetsw; 1301 return (0); 1302} 1303 1304/* 1305 * Given address of next destination (final or next hop), return (referenced) 1306 * internet address info of interface to be used to get there. 1307 */ 1308struct in_ifaddr * 1309ip_rtaddr(struct in_addr dst, u_int fibnum) 1310{ 1311 struct route sro; 1312 struct sockaddr_in *sin; 1313 struct in_ifaddr *ia; 1314 1315 bzero(&sro, sizeof(sro)); 1316 sin = (struct sockaddr_in *)&sro.ro_dst; 1317 sin->sin_family = AF_INET; 1318 sin->sin_len = sizeof(*sin); 1319 sin->sin_addr = dst; 1320 in_rtalloc_ign(&sro, 0, fibnum); 1321 1322 if (sro.ro_rt == NULL) 1323 return (NULL); 1324 1325 ia = ifatoia(sro.ro_rt->rt_ifa); 1326 ifa_ref(&ia->ia_ifa); 1327 RTFREE(sro.ro_rt); 1328 return (ia); 1329} 1330 1331u_char inetctlerrmap[PRC_NCMDS] = { 1332 0, 0, 0, 0, 1333 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1334 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1335 EMSGSIZE, EHOSTUNREACH, 0, 0, 1336 0, 0, EHOSTUNREACH, 0, 1337 ENOPROTOOPT, ECONNREFUSED 1338}; 1339 1340/* 1341 * Forward a packet. If some error occurs return the sender 1342 * an icmp packet. Note we can't always generate a meaningful 1343 * icmp message because icmp doesn't have a large enough repertoire 1344 * of codes and types. 1345 * 1346 * If not forwarding, just drop the packet. This could be confusing 1347 * if ipforwarding was zero but some routing protocol was advancing 1348 * us as a gateway to somewhere. However, we must let the routing 1349 * protocol deal with that. 1350 * 1351 * The srcrt parameter indicates whether the packet is being forwarded 1352 * via a source route. 1353 */ 1354void 1355ip_forward(struct mbuf *m, int srcrt) 1356{ 1357 struct ip *ip = mtod(m, struct ip *); 1358 struct in_ifaddr *ia; 1359 struct mbuf *mcopy; 1360 struct in_addr dest; 1361 struct route ro; 1362 int error, type = 0, code = 0, mtu = 0; 1363 1364 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1365 IPSTAT_INC(ips_cantforward); 1366 m_freem(m); 1367 return; 1368 } 1369#ifdef IPSTEALTH 1370 if (!V_ipstealth) { 1371#endif 1372 if (ip->ip_ttl <= IPTTLDEC) { 1373 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1374 0, 0); 1375 return; 1376 } 1377#ifdef IPSTEALTH 1378 } 1379#endif 1380 1381 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); 1382#ifndef IPSEC 1383 /* 1384 * 'ia' may be NULL if there is no route for this destination. 1385 * In case of IPsec, Don't discard it just yet, but pass it to 1386 * ip_output in case of outgoing IPsec policy. 1387 */ 1388 if (!srcrt && ia == NULL) { 1389 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1390 return; 1391 } 1392#endif 1393 1394 /* 1395 * Save the IP header and at most 8 bytes of the payload, 1396 * in case we need to generate an ICMP message to the src. 1397 * 1398 * XXX this can be optimized a lot by saving the data in a local 1399 * buffer on the stack (72 bytes at most), and only allocating the 1400 * mbuf if really necessary. The vast majority of the packets 1401 * are forwarded without having to send an ICMP back (either 1402 * because unnecessary, or because rate limited), so we are 1403 * really we are wasting a lot of work here. 1404 * 1405 * We don't use m_copy() because it might return a reference 1406 * to a shared cluster. Both this function and ip_output() 1407 * assume exclusive access to the IP header in `m', so any 1408 * data in a cluster may change before we reach icmp_error(). 1409 */ 1410 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1411 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1412 /* 1413 * It's probably ok if the pkthdr dup fails (because 1414 * the deep copy of the tag chain failed), but for now 1415 * be conservative and just discard the copy since 1416 * code below may some day want the tags. 1417 */ 1418 m_free(mcopy); 1419 mcopy = NULL; 1420 } 1421 if (mcopy != NULL) { 1422 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy)); 1423 mcopy->m_pkthdr.len = mcopy->m_len; 1424 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1425 } 1426 1427#ifdef IPSTEALTH 1428 if (!V_ipstealth) { 1429#endif 1430 ip->ip_ttl -= IPTTLDEC; 1431#ifdef IPSTEALTH 1432 } 1433#endif 1434 1435 /* 1436 * If forwarding packet using same interface that it came in on, 1437 * perhaps should send a redirect to sender to shortcut a hop. 1438 * Only send redirect if source is sending directly to us, 1439 * and if packet was not source routed (or has any options). 1440 * Also, don't send redirect if forwarding using a default route 1441 * or a route modified by a redirect. 1442 */ 1443 dest.s_addr = 0; 1444 if (!srcrt && V_ipsendredirects && 1445 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) { 1446 struct sockaddr_in *sin; 1447 struct rtentry *rt; 1448 1449 bzero(&ro, sizeof(ro)); 1450 sin = (struct sockaddr_in *)&ro.ro_dst; 1451 sin->sin_family = AF_INET; 1452 sin->sin_len = sizeof(*sin); 1453 sin->sin_addr = ip->ip_dst; 1454 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 1455 1456 rt = ro.ro_rt; 1457 1458 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1459 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1460#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1461 u_long src = ntohl(ip->ip_src.s_addr); 1462 1463 if (RTA(rt) && 1464 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1465 if (rt->rt_flags & RTF_GATEWAY) 1466 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1467 else 1468 dest.s_addr = ip->ip_dst.s_addr; 1469 /* Router requirements says to only send host redirects */ 1470 type = ICMP_REDIRECT; 1471 code = ICMP_REDIRECT_HOST; 1472 } 1473 } 1474 if (rt) 1475 RTFREE(rt); 1476 } 1477 1478 /* 1479 * Try to cache the route MTU from ip_output so we can consider it for 1480 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. 1481 */ 1482 bzero(&ro, sizeof(ro)); 1483 1484 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1485 1486 if (error == EMSGSIZE && ro.ro_rt) 1487 mtu = ro.ro_rt->rt_rmx.rmx_mtu; 1488 RO_RTFREE(&ro); 1489 1490 if (error) 1491 IPSTAT_INC(ips_cantforward); 1492 else { 1493 IPSTAT_INC(ips_forward); 1494 if (type) 1495 IPSTAT_INC(ips_redirectsent); 1496 else { 1497 if (mcopy) 1498 m_freem(mcopy); 1499 if (ia != NULL) 1500 ifa_free(&ia->ia_ifa); 1501 return; 1502 } 1503 } 1504 if (mcopy == NULL) { 1505 if (ia != NULL) 1506 ifa_free(&ia->ia_ifa); 1507 return; 1508 } 1509 1510 switch (error) { 1511 1512 case 0: /* forwarded, but need redirect */ 1513 /* type, code set above */ 1514 break; 1515 1516 case ENETUNREACH: 1517 case EHOSTUNREACH: 1518 case ENETDOWN: 1519 case EHOSTDOWN: 1520 default: 1521 type = ICMP_UNREACH; 1522 code = ICMP_UNREACH_HOST; 1523 break; 1524 1525 case EMSGSIZE: 1526 type = ICMP_UNREACH; 1527 code = ICMP_UNREACH_NEEDFRAG; 1528 1529#ifdef IPSEC 1530 /* 1531 * If IPsec is configured for this path, 1532 * override any possibly mtu value set by ip_output. 1533 */ 1534 mtu = ip_ipsec_mtu(mcopy, mtu); 1535#endif /* IPSEC */ 1536 /* 1537 * If the MTU was set before make sure we are below the 1538 * interface MTU. 1539 * If the MTU wasn't set before use the interface mtu or 1540 * fall back to the next smaller mtu step compared to the 1541 * current packet size. 1542 */ 1543 if (mtu != 0) { 1544 if (ia != NULL) 1545 mtu = min(mtu, ia->ia_ifp->if_mtu); 1546 } else { 1547 if (ia != NULL) 1548 mtu = ia->ia_ifp->if_mtu; 1549 else 1550 mtu = ip_next_mtu(ntohs(ip->ip_len), 0); 1551 } 1552 IPSTAT_INC(ips_cantfrag); 1553 break; 1554 1555 case ENOBUFS: 1556 /* 1557 * A router should not generate ICMP_SOURCEQUENCH as 1558 * required in RFC1812 Requirements for IP Version 4 Routers. 1559 * Source quench could be a big problem under DoS attacks, 1560 * or if the underlying interface is rate-limited. 1561 * Those who need source quench packets may re-enable them 1562 * via the net.inet.ip.sendsourcequench sysctl. 1563 */ 1564 if (V_ip_sendsourcequench == 0) { 1565 m_freem(mcopy); 1566 if (ia != NULL) 1567 ifa_free(&ia->ia_ifa); 1568 return; 1569 } else { 1570 type = ICMP_SOURCEQUENCH; 1571 code = 0; 1572 } 1573 break; 1574 1575 case EACCES: /* ipfw denied packet */ 1576 m_freem(mcopy); 1577 if (ia != NULL) 1578 ifa_free(&ia->ia_ifa); 1579 return; 1580 } 1581 if (ia != NULL) 1582 ifa_free(&ia->ia_ifa); 1583 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1584} 1585 1586void 1587ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1588 struct mbuf *m) 1589{ 1590 1591 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1592 struct bintime bt; 1593 1594 bintime(&bt); 1595 if (inp->inp_socket->so_options & SO_BINTIME) { 1596 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1597 SCM_BINTIME, SOL_SOCKET); 1598 if (*mp) 1599 mp = &(*mp)->m_next; 1600 } 1601 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1602 struct timeval tv; 1603 1604 bintime2timeval(&bt, &tv); 1605 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1606 SCM_TIMESTAMP, SOL_SOCKET); 1607 if (*mp) 1608 mp = &(*mp)->m_next; 1609 } 1610 } 1611 if (inp->inp_flags & INP_RECVDSTADDR) { 1612 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1613 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1614 if (*mp) 1615 mp = &(*mp)->m_next; 1616 } 1617 if (inp->inp_flags & INP_RECVTTL) { 1618 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1619 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1620 if (*mp) 1621 mp = &(*mp)->m_next; 1622 } 1623#ifdef notyet 1624 /* XXX 1625 * Moving these out of udp_input() made them even more broken 1626 * than they already were. 1627 */ 1628 /* options were tossed already */ 1629 if (inp->inp_flags & INP_RECVOPTS) { 1630 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1631 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1632 if (*mp) 1633 mp = &(*mp)->m_next; 1634 } 1635 /* ip_srcroute doesn't do what we want here, need to fix */ 1636 if (inp->inp_flags & INP_RECVRETOPTS) { 1637 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1638 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1639 if (*mp) 1640 mp = &(*mp)->m_next; 1641 } 1642#endif 1643 if (inp->inp_flags & INP_RECVIF) { 1644 struct ifnet *ifp; 1645 struct sdlbuf { 1646 struct sockaddr_dl sdl; 1647 u_char pad[32]; 1648 } sdlbuf; 1649 struct sockaddr_dl *sdp; 1650 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1651 1652 if (((ifp = m->m_pkthdr.rcvif)) 1653 && ( ifp->if_index && (ifp->if_index <= V_if_index))) { 1654 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1655 /* 1656 * Change our mind and don't try copy. 1657 */ 1658 if ((sdp->sdl_family != AF_LINK) 1659 || (sdp->sdl_len > sizeof(sdlbuf))) { 1660 goto makedummy; 1661 } 1662 bcopy(sdp, sdl2, sdp->sdl_len); 1663 } else { 1664makedummy: 1665 sdl2->sdl_len 1666 = offsetof(struct sockaddr_dl, sdl_data[0]); 1667 sdl2->sdl_family = AF_LINK; 1668 sdl2->sdl_index = 0; 1669 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1670 } 1671 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1672 IP_RECVIF, IPPROTO_IP); 1673 if (*mp) 1674 mp = &(*mp)->m_next; 1675 } 1676 if (inp->inp_flags & INP_RECVTOS) { 1677 *mp = sbcreatecontrol((caddr_t) &ip->ip_tos, 1678 sizeof(u_char), IP_RECVTOS, IPPROTO_IP); 1679 if (*mp) 1680 mp = &(*mp)->m_next; 1681 } 1682} 1683 1684/* 1685 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1686 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1687 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1688 * compiled. 1689 */ 1690static VNET_DEFINE(int, ip_rsvp_on); 1691VNET_DEFINE(struct socket *, ip_rsvpd); 1692 1693#define V_ip_rsvp_on VNET(ip_rsvp_on) 1694 1695int 1696ip_rsvp_init(struct socket *so) 1697{ 1698 1699 if (so->so_type != SOCK_RAW || 1700 so->so_proto->pr_protocol != IPPROTO_RSVP) 1701 return EOPNOTSUPP; 1702 1703 if (V_ip_rsvpd != NULL) 1704 return EADDRINUSE; 1705 1706 V_ip_rsvpd = so; 1707 /* 1708 * This may seem silly, but we need to be sure we don't over-increment 1709 * the RSVP counter, in case something slips up. 1710 */ 1711 if (!V_ip_rsvp_on) { 1712 V_ip_rsvp_on = 1; 1713 V_rsvp_on++; 1714 } 1715 1716 return 0; 1717} 1718 1719int 1720ip_rsvp_done(void) 1721{ 1722 1723 V_ip_rsvpd = NULL; 1724 /* 1725 * This may seem silly, but we need to be sure we don't over-decrement 1726 * the RSVP counter, in case something slips up. 1727 */ 1728 if (V_ip_rsvp_on) { 1729 V_ip_rsvp_on = 0; 1730 V_rsvp_on--; 1731 } 1732 return 0; 1733} 1734 1735void 1736rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1737{ 1738 1739 if (rsvp_input_p) { /* call the real one if loaded */ 1740 rsvp_input_p(m, off); 1741 return; 1742 } 1743 1744 /* Can still get packets with rsvp_on = 0 if there is a local member 1745 * of the group to which the RSVP packet is addressed. But in this 1746 * case we want to throw the packet away. 1747 */ 1748 1749 if (!V_rsvp_on) { 1750 m_freem(m); 1751 return; 1752 } 1753 1754 if (V_ip_rsvpd != NULL) { 1755 rip_input(m, off); 1756 return; 1757 } 1758 /* Drop the packet */ 1759 m_freem(m); 1760} 1761