ip_input.c revision 241245
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 241245 2012-10-06 10:02:11Z glebius $"); 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 ip->ip_len = ntohs(ip->ip_len); 394 ip->ip_off = ntohs(ip->ip_off); 395 hlen = ip->ip_hl << 2; 396 goto ours; 397 } 398 399 IPSTAT_INC(ips_total); 400 401 if (m->m_pkthdr.len < sizeof(struct ip)) 402 goto tooshort; 403 404 if (m->m_len < sizeof (struct ip) && 405 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 406 IPSTAT_INC(ips_toosmall); 407 return; 408 } 409 ip = mtod(m, struct ip *); 410 411 if (ip->ip_v != IPVERSION) { 412 IPSTAT_INC(ips_badvers); 413 goto bad; 414 } 415 416 hlen = ip->ip_hl << 2; 417 if (hlen < sizeof(struct ip)) { /* minimum header length */ 418 IPSTAT_INC(ips_badhlen); 419 goto bad; 420 } 421 if (hlen > m->m_len) { 422 if ((m = m_pullup(m, hlen)) == NULL) { 423 IPSTAT_INC(ips_badhlen); 424 return; 425 } 426 ip = mtod(m, struct ip *); 427 } 428 429 /* 127/8 must not appear on wire - RFC1122 */ 430 ifp = m->m_pkthdr.rcvif; 431 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 432 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 433 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 434 IPSTAT_INC(ips_badaddr); 435 goto bad; 436 } 437 } 438 439 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 440 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 441 } else { 442 if (hlen == sizeof(struct ip)) { 443 sum = in_cksum_hdr(ip); 444 } else { 445 sum = in_cksum(m, hlen); 446 } 447 } 448 if (sum) { 449 IPSTAT_INC(ips_badsum); 450 goto bad; 451 } 452 453#ifdef ALTQ 454 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 455 /* packet is dropped by traffic conditioner */ 456 return; 457#endif 458 459 ip_len = ntohs(ip->ip_len); 460 if (ip_len < hlen) { 461 IPSTAT_INC(ips_badlen); 462 goto bad; 463 } 464 465 /* 466 * Check that the amount of data in the buffers 467 * is as at least much as the IP header would have us expect. 468 * Trim mbufs if longer than we expect. 469 * Drop packet if shorter than we expect. 470 */ 471 if (m->m_pkthdr.len < ip_len) { 472tooshort: 473 IPSTAT_INC(ips_tooshort); 474 goto bad; 475 } 476 if (m->m_pkthdr.len > ip_len) { 477 if (m->m_len == m->m_pkthdr.len) { 478 m->m_len = ip_len; 479 m->m_pkthdr.len = ip_len; 480 } else 481 m_adj(m, ip_len - m->m_pkthdr.len); 482 } 483#ifdef IPSEC 484 /* 485 * Bypass packet filtering for packets previously handled by IPsec. 486 */ 487 if (ip_ipsec_filtertunnel(m)) 488 goto passin; 489#endif /* IPSEC */ 490 491 /* 492 * Run through list of hooks for input packets. 493 * 494 * NB: Beware of the destination address changing (e.g. 495 * by NAT rewriting). When this happens, tell 496 * ip_forward to do the right thing. 497 */ 498 499 /* Jump over all PFIL processing if hooks are not active. */ 500 if (!PFIL_HOOKED(&V_inet_pfil_hook)) 501 goto passin; 502 503 odst = ip->ip_dst; 504 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0) 505 return; 506 if (m == NULL) /* consumed by filter */ 507 return; 508 509 ip = mtod(m, struct ip *); 510 dchg = (odst.s_addr != ip->ip_dst.s_addr); 511 ifp = m->m_pkthdr.rcvif; 512 513#ifdef IPFIREWALL_FORWARD 514 if (m->m_flags & M_FASTFWD_OURS) { 515 m->m_flags &= ~M_FASTFWD_OURS; 516 ip->ip_len = ntohs(ip->ip_len); 517 ip->ip_off = ntohs(ip->ip_off); 518 goto ours; 519 } 520 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 521 /* 522 * Directly ship the packet on. This allows forwarding 523 * packets originally destined to us to some other directly 524 * connected host. 525 */ 526 ip->ip_len = ntohs(ip->ip_len); 527 ip->ip_off = ntohs(ip->ip_off); 528 ip_forward(m, dchg); 529 return; 530 } 531#endif /* IPFIREWALL_FORWARD */ 532 533passin: 534 /* 535 * From now and up to output pfil(9) processing in ip_output() 536 * the header is in host byte order. 537 */ 538 ip->ip_len = ntohs(ip->ip_len); 539 ip->ip_off = ntohs(ip->ip_off); 540 541 /* 542 * Process options and, if not destined for us, 543 * ship it on. ip_dooptions returns 1 when an 544 * error was detected (causing an icmp message 545 * to be sent and the original packet to be freed). 546 */ 547 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 548 return; 549 550 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 551 * matter if it is destined to another node, or whether it is 552 * a multicast one, RSVP wants it! and prevents it from being forwarded 553 * anywhere else. Also checks if the rsvp daemon is running before 554 * grabbing the packet. 555 */ 556 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 557 goto ours; 558 559 /* 560 * Check our list of addresses, to see if the packet is for us. 561 * If we don't have any addresses, assume any unicast packet 562 * we receive might be for us (and let the upper layers deal 563 * with it). 564 */ 565 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 566 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 567 goto ours; 568 569 /* 570 * Enable a consistency check between the destination address 571 * and the arrival interface for a unicast packet (the RFC 1122 572 * strong ES model) if IP forwarding is disabled and the packet 573 * is not locally generated and the packet is not subject to 574 * 'ipfw fwd'. 575 * 576 * XXX - Checking also should be disabled if the destination 577 * address is ipnat'ed to a different interface. 578 * 579 * XXX - Checking is incompatible with IP aliases added 580 * to the loopback interface instead of the interface where 581 * the packets are received. 582 * 583 * XXX - This is the case for carp vhost IPs as well so we 584 * insert a workaround. If the packet got here, we already 585 * checked with carp_iamatch() and carp_forus(). 586 */ 587 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 588 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && 589 ifp->if_carp == NULL && (dchg == 0); 590 591 /* 592 * Check for exact addresses in the hash bucket. 593 */ 594 /* IN_IFADDR_RLOCK(); */ 595 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 596 /* 597 * If the address matches, verify that the packet 598 * arrived via the correct interface if checking is 599 * enabled. 600 */ 601 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 602 (!checkif || ia->ia_ifp == ifp)) { 603 ifa_ref(&ia->ia_ifa); 604 /* IN_IFADDR_RUNLOCK(); */ 605 goto ours; 606 } 607 } 608 /* IN_IFADDR_RUNLOCK(); */ 609 610 /* 611 * Check for broadcast addresses. 612 * 613 * Only accept broadcast packets that arrive via the matching 614 * interface. Reception of forwarded directed broadcasts would 615 * be handled via ip_forward() and ether_output() with the loopback 616 * into the stack for SIMPLEX interfaces handled by ether_output(). 617 */ 618 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 619 IF_ADDR_RLOCK(ifp); 620 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 621 if (ifa->ifa_addr->sa_family != AF_INET) 622 continue; 623 ia = ifatoia(ifa); 624 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 625 ip->ip_dst.s_addr) { 626 ifa_ref(ifa); 627 IF_ADDR_RUNLOCK(ifp); 628 goto ours; 629 } 630#ifdef BOOTP_COMPAT 631 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { 632 ifa_ref(ifa); 633 IF_ADDR_RUNLOCK(ifp); 634 goto ours; 635 } 636#endif 637 } 638 IF_ADDR_RUNLOCK(ifp); 639 ia = NULL; 640 } 641 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 642 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 643 IPSTAT_INC(ips_cantforward); 644 m_freem(m); 645 return; 646 } 647 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 648 if (V_ip_mrouter) { 649 /* 650 * If we are acting as a multicast router, all 651 * incoming multicast packets are passed to the 652 * kernel-level multicast forwarding function. 653 * The packet is returned (relatively) intact; if 654 * ip_mforward() returns a non-zero value, the packet 655 * must be discarded, else it may be accepted below. 656 */ 657 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 658 IPSTAT_INC(ips_cantforward); 659 m_freem(m); 660 return; 661 } 662 663 /* 664 * The process-level routing daemon needs to receive 665 * all multicast IGMP packets, whether or not this 666 * host belongs to their destination groups. 667 */ 668 if (ip->ip_p == IPPROTO_IGMP) 669 goto ours; 670 IPSTAT_INC(ips_forward); 671 } 672 /* 673 * Assume the packet is for us, to avoid prematurely taking 674 * a lock on the in_multi hash. Protocols must perform 675 * their own filtering and update statistics accordingly. 676 */ 677 goto ours; 678 } 679 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 680 goto ours; 681 if (ip->ip_dst.s_addr == INADDR_ANY) 682 goto ours; 683 684 /* 685 * FAITH(Firewall Aided Internet Translator) 686 */ 687 if (ifp && ifp->if_type == IFT_FAITH) { 688 if (V_ip_keepfaith) { 689 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 690 goto ours; 691 } 692 m_freem(m); 693 return; 694 } 695 696 /* 697 * Not for us; forward if possible and desirable. 698 */ 699 if (V_ipforwarding == 0) { 700 IPSTAT_INC(ips_cantforward); 701 m_freem(m); 702 } else { 703#ifdef IPSEC 704 if (ip_ipsec_fwd(m)) 705 goto bad; 706#endif /* IPSEC */ 707 ip_forward(m, dchg); 708 } 709 return; 710 711ours: 712#ifdef IPSTEALTH 713 /* 714 * IPSTEALTH: Process non-routing options only 715 * if the packet is destined for us. 716 */ 717 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) { 718 if (ia != NULL) 719 ifa_free(&ia->ia_ifa); 720 return; 721 } 722#endif /* IPSTEALTH */ 723 724 /* Count the packet in the ip address stats */ 725 if (ia != NULL) { 726 ia->ia_ifa.if_ipackets++; 727 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 728 ifa_free(&ia->ia_ifa); 729 } 730 731 /* 732 * Attempt reassembly; if it succeeds, proceed. 733 * ip_reass() will return a different mbuf. 734 */ 735 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 736 m = ip_reass(m); 737 if (m == NULL) 738 return; 739 ip = mtod(m, struct ip *); 740 /* Get the header length of the reassembled packet */ 741 hlen = ip->ip_hl << 2; 742 } 743 744 /* 745 * Further protocols expect the packet length to be w/o the 746 * IP header. 747 */ 748 ip->ip_len -= hlen; 749 750#ifdef IPSEC 751 /* 752 * enforce IPsec policy checking if we are seeing last header. 753 * note that we do not visit this with protocols with pcb layer 754 * code - like udp/tcp/raw ip. 755 */ 756 if (ip_ipsec_input(m)) 757 goto bad; 758#endif /* IPSEC */ 759 760 /* 761 * Switch out to protocol's input routine. 762 */ 763 IPSTAT_INC(ips_delivered); 764 765 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 766 return; 767bad: 768 m_freem(m); 769} 770 771/* 772 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 773 * max has slightly different semantics than the sysctl, for historical 774 * reasons. 775 */ 776static void 777maxnipq_update(void) 778{ 779 780 /* 781 * -1 for unlimited allocation. 782 */ 783 if (V_maxnipq < 0) 784 uma_zone_set_max(V_ipq_zone, 0); 785 /* 786 * Positive number for specific bound. 787 */ 788 if (V_maxnipq > 0) 789 uma_zone_set_max(V_ipq_zone, V_maxnipq); 790 /* 791 * Zero specifies no further fragment queue allocation -- set the 792 * bound very low, but rely on implementation elsewhere to actually 793 * prevent allocation and reclaim current queues. 794 */ 795 if (V_maxnipq == 0) 796 uma_zone_set_max(V_ipq_zone, 1); 797} 798 799static void 800ipq_zone_change(void *tag) 801{ 802 803 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) { 804 V_maxnipq = nmbclusters / 32; 805 maxnipq_update(); 806 } 807} 808 809static int 810sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 811{ 812 int error, i; 813 814 i = V_maxnipq; 815 error = sysctl_handle_int(oidp, &i, 0, req); 816 if (error || !req->newptr) 817 return (error); 818 819 /* 820 * XXXRW: Might be a good idea to sanity check the argument and place 821 * an extreme upper bound. 822 */ 823 if (i < -1) 824 return (EINVAL); 825 V_maxnipq = i; 826 maxnipq_update(); 827 return (0); 828} 829 830SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 831 NULL, 0, sysctl_maxnipq, "I", 832 "Maximum number of IPv4 fragment reassembly queue entries"); 833 834/* 835 * Take incoming datagram fragment and try to reassemble it into 836 * whole datagram. If the argument is the first fragment or one 837 * in between the function will return NULL and store the mbuf 838 * in the fragment chain. If the argument is the last fragment 839 * the packet will be reassembled and the pointer to the new 840 * mbuf returned for further processing. Only m_tags attached 841 * to the first packet/fragment are preserved. 842 * The IP header is *NOT* adjusted out of iplen. 843 */ 844struct mbuf * 845ip_reass(struct mbuf *m) 846{ 847 struct ip *ip; 848 struct mbuf *p, *q, *nq, *t; 849 struct ipq *fp = NULL; 850 struct ipqhead *head; 851 int i, hlen, next; 852 u_int8_t ecn, ecn0; 853 u_short hash; 854 855 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 856 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) { 857 IPSTAT_INC(ips_fragments); 858 IPSTAT_INC(ips_fragdropped); 859 m_freem(m); 860 return (NULL); 861 } 862 863 ip = mtod(m, struct ip *); 864 hlen = ip->ip_hl << 2; 865 866 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 867 head = &V_ipq[hash]; 868 IPQ_LOCK(); 869 870 /* 871 * Look for queue of fragments 872 * of this datagram. 873 */ 874 TAILQ_FOREACH(fp, head, ipq_list) 875 if (ip->ip_id == fp->ipq_id && 876 ip->ip_src.s_addr == fp->ipq_src.s_addr && 877 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 878#ifdef MAC 879 mac_ipq_match(m, fp) && 880#endif 881 ip->ip_p == fp->ipq_p) 882 goto found; 883 884 fp = NULL; 885 886 /* 887 * Attempt to trim the number of allocated fragment queues if it 888 * exceeds the administrative limit. 889 */ 890 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) { 891 /* 892 * drop something from the tail of the current queue 893 * before proceeding further 894 */ 895 struct ipq *q = TAILQ_LAST(head, ipqhead); 896 if (q == NULL) { /* gak */ 897 for (i = 0; i < IPREASS_NHASH; i++) { 898 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead); 899 if (r) { 900 IPSTAT_ADD(ips_fragtimeout, 901 r->ipq_nfrags); 902 ip_freef(&V_ipq[i], r); 903 break; 904 } 905 } 906 } else { 907 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags); 908 ip_freef(head, q); 909 } 910 } 911 912found: 913 /* 914 * Adjust ip_len to not reflect header, 915 * convert offset of this to bytes. 916 */ 917 ip->ip_len -= hlen; 918 if (ip->ip_off & IP_MF) { 919 /* 920 * Make sure that fragments have a data length 921 * that's a non-zero multiple of 8 bytes. 922 */ 923 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 924 IPSTAT_INC(ips_toosmall); /* XXX */ 925 goto dropfrag; 926 } 927 m->m_flags |= M_FRAG; 928 } else 929 m->m_flags &= ~M_FRAG; 930 ip->ip_off <<= 3; 931 932 933 /* 934 * Attempt reassembly; if it succeeds, proceed. 935 * ip_reass() will return a different mbuf. 936 */ 937 IPSTAT_INC(ips_fragments); 938 m->m_pkthdr.header = ip; 939 940 /* Previous ip_reass() started here. */ 941 /* 942 * Presence of header sizes in mbufs 943 * would confuse code below. 944 */ 945 m->m_data += hlen; 946 m->m_len -= hlen; 947 948 /* 949 * If first fragment to arrive, create a reassembly queue. 950 */ 951 if (fp == NULL) { 952 fp = uma_zalloc(V_ipq_zone, M_NOWAIT); 953 if (fp == NULL) 954 goto dropfrag; 955#ifdef MAC 956 if (mac_ipq_init(fp, M_NOWAIT) != 0) { 957 uma_zfree(V_ipq_zone, fp); 958 fp = NULL; 959 goto dropfrag; 960 } 961 mac_ipq_create(m, fp); 962#endif 963 TAILQ_INSERT_HEAD(head, fp, ipq_list); 964 V_nipq++; 965 fp->ipq_nfrags = 1; 966 fp->ipq_ttl = IPFRAGTTL; 967 fp->ipq_p = ip->ip_p; 968 fp->ipq_id = ip->ip_id; 969 fp->ipq_src = ip->ip_src; 970 fp->ipq_dst = ip->ip_dst; 971 fp->ipq_frags = m; 972 m->m_nextpkt = NULL; 973 goto done; 974 } else { 975 fp->ipq_nfrags++; 976#ifdef MAC 977 mac_ipq_update(m, fp); 978#endif 979 } 980 981#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 982 983 /* 984 * Handle ECN by comparing this segment with the first one; 985 * if CE is set, do not lose CE. 986 * drop if CE and not-ECT are mixed for the same packet. 987 */ 988 ecn = ip->ip_tos & IPTOS_ECN_MASK; 989 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 990 if (ecn == IPTOS_ECN_CE) { 991 if (ecn0 == IPTOS_ECN_NOTECT) 992 goto dropfrag; 993 if (ecn0 != IPTOS_ECN_CE) 994 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 995 } 996 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 997 goto dropfrag; 998 999 /* 1000 * Find a segment which begins after this one does. 1001 */ 1002 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1003 if (GETIP(q)->ip_off > ip->ip_off) 1004 break; 1005 1006 /* 1007 * If there is a preceding segment, it may provide some of 1008 * our data already. If so, drop the data from the incoming 1009 * segment. If it provides all of our data, drop us, otherwise 1010 * stick new segment in the proper place. 1011 * 1012 * If some of the data is dropped from the preceding 1013 * segment, then it's checksum is invalidated. 1014 */ 1015 if (p) { 1016 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1017 if (i > 0) { 1018 if (i >= ip->ip_len) 1019 goto dropfrag; 1020 m_adj(m, i); 1021 m->m_pkthdr.csum_flags = 0; 1022 ip->ip_off += i; 1023 ip->ip_len -= i; 1024 } 1025 m->m_nextpkt = p->m_nextpkt; 1026 p->m_nextpkt = m; 1027 } else { 1028 m->m_nextpkt = fp->ipq_frags; 1029 fp->ipq_frags = m; 1030 } 1031 1032 /* 1033 * While we overlap succeeding segments trim them or, 1034 * if they are completely covered, dequeue them. 1035 */ 1036 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1037 q = nq) { 1038 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1039 if (i < GETIP(q)->ip_len) { 1040 GETIP(q)->ip_len -= i; 1041 GETIP(q)->ip_off += i; 1042 m_adj(q, i); 1043 q->m_pkthdr.csum_flags = 0; 1044 break; 1045 } 1046 nq = q->m_nextpkt; 1047 m->m_nextpkt = nq; 1048 IPSTAT_INC(ips_fragdropped); 1049 fp->ipq_nfrags--; 1050 m_freem(q); 1051 } 1052 1053 /* 1054 * Check for complete reassembly and perform frag per packet 1055 * limiting. 1056 * 1057 * Frag limiting is performed here so that the nth frag has 1058 * a chance to complete the packet before we drop the packet. 1059 * As a result, n+1 frags are actually allowed per packet, but 1060 * only n will ever be stored. (n = maxfragsperpacket.) 1061 * 1062 */ 1063 next = 0; 1064 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1065 if (GETIP(q)->ip_off != next) { 1066 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1067 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1068 ip_freef(head, fp); 1069 } 1070 goto done; 1071 } 1072 next += GETIP(q)->ip_len; 1073 } 1074 /* Make sure the last packet didn't have the IP_MF flag */ 1075 if (p->m_flags & M_FRAG) { 1076 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1077 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1078 ip_freef(head, fp); 1079 } 1080 goto done; 1081 } 1082 1083 /* 1084 * Reassembly is complete. Make sure the packet is a sane size. 1085 */ 1086 q = fp->ipq_frags; 1087 ip = GETIP(q); 1088 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1089 IPSTAT_INC(ips_toolong); 1090 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1091 ip_freef(head, fp); 1092 goto done; 1093 } 1094 1095 /* 1096 * Concatenate fragments. 1097 */ 1098 m = q; 1099 t = m->m_next; 1100 m->m_next = NULL; 1101 m_cat(m, t); 1102 nq = q->m_nextpkt; 1103 q->m_nextpkt = NULL; 1104 for (q = nq; q != NULL; q = nq) { 1105 nq = q->m_nextpkt; 1106 q->m_nextpkt = NULL; 1107 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1108 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1109 m_cat(m, q); 1110 } 1111 /* 1112 * In order to do checksumming faster we do 'end-around carry' here 1113 * (and not in for{} loop), though it implies we are not going to 1114 * reassemble more than 64k fragments. 1115 */ 1116 m->m_pkthdr.csum_data = 1117 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1118#ifdef MAC 1119 mac_ipq_reassemble(fp, m); 1120 mac_ipq_destroy(fp); 1121#endif 1122 1123 /* 1124 * Create header for new ip packet by modifying header of first 1125 * packet; dequeue and discard fragment reassembly header. 1126 * Make header visible. 1127 */ 1128 ip->ip_len = (ip->ip_hl << 2) + next; 1129 ip->ip_src = fp->ipq_src; 1130 ip->ip_dst = fp->ipq_dst; 1131 TAILQ_REMOVE(head, fp, ipq_list); 1132 V_nipq--; 1133 uma_zfree(V_ipq_zone, fp); 1134 m->m_len += (ip->ip_hl << 2); 1135 m->m_data -= (ip->ip_hl << 2); 1136 /* some debugging cruft by sklower, below, will go away soon */ 1137 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1138 m_fixhdr(m); 1139 IPSTAT_INC(ips_reassembled); 1140 IPQ_UNLOCK(); 1141 return (m); 1142 1143dropfrag: 1144 IPSTAT_INC(ips_fragdropped); 1145 if (fp != NULL) 1146 fp->ipq_nfrags--; 1147 m_freem(m); 1148done: 1149 IPQ_UNLOCK(); 1150 return (NULL); 1151 1152#undef GETIP 1153} 1154 1155/* 1156 * Free a fragment reassembly header and all 1157 * associated datagrams. 1158 */ 1159static void 1160ip_freef(struct ipqhead *fhp, struct ipq *fp) 1161{ 1162 struct mbuf *q; 1163 1164 IPQ_LOCK_ASSERT(); 1165 1166 while (fp->ipq_frags) { 1167 q = fp->ipq_frags; 1168 fp->ipq_frags = q->m_nextpkt; 1169 m_freem(q); 1170 } 1171 TAILQ_REMOVE(fhp, fp, ipq_list); 1172 uma_zfree(V_ipq_zone, fp); 1173 V_nipq--; 1174} 1175 1176/* 1177 * IP timer processing; 1178 * if a timer expires on a reassembly 1179 * queue, discard it. 1180 */ 1181void 1182ip_slowtimo(void) 1183{ 1184 VNET_ITERATOR_DECL(vnet_iter); 1185 struct ipq *fp; 1186 int i; 1187 1188 VNET_LIST_RLOCK_NOSLEEP(); 1189 IPQ_LOCK(); 1190 VNET_FOREACH(vnet_iter) { 1191 CURVNET_SET(vnet_iter); 1192 for (i = 0; i < IPREASS_NHASH; i++) { 1193 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) { 1194 struct ipq *fpp; 1195 1196 fpp = fp; 1197 fp = TAILQ_NEXT(fp, ipq_list); 1198 if(--fpp->ipq_ttl == 0) { 1199 IPSTAT_ADD(ips_fragtimeout, 1200 fpp->ipq_nfrags); 1201 ip_freef(&V_ipq[i], fpp); 1202 } 1203 } 1204 } 1205 /* 1206 * If we are over the maximum number of fragments 1207 * (due to the limit being lowered), drain off 1208 * enough to get down to the new limit. 1209 */ 1210 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) { 1211 for (i = 0; i < IPREASS_NHASH; i++) { 1212 while (V_nipq > V_maxnipq && 1213 !TAILQ_EMPTY(&V_ipq[i])) { 1214 IPSTAT_ADD(ips_fragdropped, 1215 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1216 ip_freef(&V_ipq[i], 1217 TAILQ_FIRST(&V_ipq[i])); 1218 } 1219 } 1220 } 1221 CURVNET_RESTORE(); 1222 } 1223 IPQ_UNLOCK(); 1224 VNET_LIST_RUNLOCK_NOSLEEP(); 1225} 1226 1227/* 1228 * Drain off all datagram fragments. 1229 */ 1230static void 1231ip_drain_locked(void) 1232{ 1233 int i; 1234 1235 IPQ_LOCK_ASSERT(); 1236 1237 for (i = 0; i < IPREASS_NHASH; i++) { 1238 while(!TAILQ_EMPTY(&V_ipq[i])) { 1239 IPSTAT_ADD(ips_fragdropped, 1240 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1241 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); 1242 } 1243 } 1244} 1245 1246void 1247ip_drain(void) 1248{ 1249 VNET_ITERATOR_DECL(vnet_iter); 1250 1251 VNET_LIST_RLOCK_NOSLEEP(); 1252 IPQ_LOCK(); 1253 VNET_FOREACH(vnet_iter) { 1254 CURVNET_SET(vnet_iter); 1255 ip_drain_locked(); 1256 CURVNET_RESTORE(); 1257 } 1258 IPQ_UNLOCK(); 1259 VNET_LIST_RUNLOCK_NOSLEEP(); 1260 in_rtqdrain(); 1261} 1262 1263/* 1264 * The protocol to be inserted into ip_protox[] must be already registered 1265 * in inetsw[], either statically or through pf_proto_register(). 1266 */ 1267int 1268ipproto_register(short ipproto) 1269{ 1270 struct protosw *pr; 1271 1272 /* Sanity checks. */ 1273 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 1274 return (EPROTONOSUPPORT); 1275 1276 /* 1277 * The protocol slot must not be occupied by another protocol 1278 * already. An index pointing to IPPROTO_RAW is unused. 1279 */ 1280 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1281 if (pr == NULL) 1282 return (EPFNOSUPPORT); 1283 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1284 return (EEXIST); 1285 1286 /* Find the protocol position in inetsw[] and set the index. */ 1287 for (pr = inetdomain.dom_protosw; 1288 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1289 if (pr->pr_domain->dom_family == PF_INET && 1290 pr->pr_protocol && pr->pr_protocol == ipproto) { 1291 ip_protox[pr->pr_protocol] = pr - inetsw; 1292 return (0); 1293 } 1294 } 1295 return (EPROTONOSUPPORT); 1296} 1297 1298int 1299ipproto_unregister(short ipproto) 1300{ 1301 struct protosw *pr; 1302 1303 /* Sanity checks. */ 1304 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 1305 return (EPROTONOSUPPORT); 1306 1307 /* Check if the protocol was indeed registered. */ 1308 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1309 if (pr == NULL) 1310 return (EPFNOSUPPORT); 1311 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1312 return (ENOENT); 1313 1314 /* Reset the protocol slot to IPPROTO_RAW. */ 1315 ip_protox[ipproto] = pr - inetsw; 1316 return (0); 1317} 1318 1319/* 1320 * Given address of next destination (final or next hop), return (referenced) 1321 * internet address info of interface to be used to get there. 1322 */ 1323struct in_ifaddr * 1324ip_rtaddr(struct in_addr dst, u_int fibnum) 1325{ 1326 struct route sro; 1327 struct sockaddr_in *sin; 1328 struct in_ifaddr *ia; 1329 1330 bzero(&sro, sizeof(sro)); 1331 sin = (struct sockaddr_in *)&sro.ro_dst; 1332 sin->sin_family = AF_INET; 1333 sin->sin_len = sizeof(*sin); 1334 sin->sin_addr = dst; 1335 in_rtalloc_ign(&sro, 0, fibnum); 1336 1337 if (sro.ro_rt == NULL) 1338 return (NULL); 1339 1340 ia = ifatoia(sro.ro_rt->rt_ifa); 1341 ifa_ref(&ia->ia_ifa); 1342 RTFREE(sro.ro_rt); 1343 return (ia); 1344} 1345 1346u_char inetctlerrmap[PRC_NCMDS] = { 1347 0, 0, 0, 0, 1348 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1349 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1350 EMSGSIZE, EHOSTUNREACH, 0, 0, 1351 0, 0, EHOSTUNREACH, 0, 1352 ENOPROTOOPT, ECONNREFUSED 1353}; 1354 1355/* 1356 * Forward a packet. If some error occurs return the sender 1357 * an icmp packet. Note we can't always generate a meaningful 1358 * icmp message because icmp doesn't have a large enough repertoire 1359 * of codes and types. 1360 * 1361 * If not forwarding, just drop the packet. This could be confusing 1362 * if ipforwarding was zero but some routing protocol was advancing 1363 * us as a gateway to somewhere. However, we must let the routing 1364 * protocol deal with that. 1365 * 1366 * The srcrt parameter indicates whether the packet is being forwarded 1367 * via a source route. 1368 * 1369 * IP header in host byte order. 1370 */ 1371void 1372ip_forward(struct mbuf *m, int srcrt) 1373{ 1374 struct ip *ip = mtod(m, struct ip *); 1375 struct in_ifaddr *ia; 1376 struct mbuf *mcopy; 1377 struct in_addr dest; 1378 struct route ro; 1379 int error, type = 0, code = 0, mtu = 0; 1380 1381 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1382 IPSTAT_INC(ips_cantforward); 1383 m_freem(m); 1384 return; 1385 } 1386#ifdef IPSTEALTH 1387 if (!V_ipstealth) { 1388#endif 1389 if (ip->ip_ttl <= IPTTLDEC) { 1390 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1391 0, 0); 1392 return; 1393 } 1394#ifdef IPSTEALTH 1395 } 1396#endif 1397 1398 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); 1399#ifndef IPSEC 1400 /* 1401 * 'ia' may be NULL if there is no route for this destination. 1402 * In case of IPsec, Don't discard it just yet, but pass it to 1403 * ip_output in case of outgoing IPsec policy. 1404 */ 1405 if (!srcrt && ia == NULL) { 1406 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1407 return; 1408 } 1409#endif 1410 1411 /* 1412 * Save the IP header and at most 8 bytes of the payload, 1413 * in case we need to generate an ICMP message to the src. 1414 * 1415 * XXX this can be optimized a lot by saving the data in a local 1416 * buffer on the stack (72 bytes at most), and only allocating the 1417 * mbuf if really necessary. The vast majority of the packets 1418 * are forwarded without having to send an ICMP back (either 1419 * because unnecessary, or because rate limited), so we are 1420 * really we are wasting a lot of work here. 1421 * 1422 * We don't use m_copy() because it might return a reference 1423 * to a shared cluster. Both this function and ip_output() 1424 * assume exclusive access to the IP header in `m', so any 1425 * data in a cluster may change before we reach icmp_error(). 1426 */ 1427 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1428 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1429 /* 1430 * It's probably ok if the pkthdr dup fails (because 1431 * the deep copy of the tag chain failed), but for now 1432 * be conservative and just discard the copy since 1433 * code below may some day want the tags. 1434 */ 1435 m_free(mcopy); 1436 mcopy = NULL; 1437 } 1438 if (mcopy != NULL) { 1439 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy)); 1440 mcopy->m_pkthdr.len = mcopy->m_len; 1441 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1442 } 1443 1444#ifdef IPSTEALTH 1445 if (!V_ipstealth) { 1446#endif 1447 ip->ip_ttl -= IPTTLDEC; 1448#ifdef IPSTEALTH 1449 } 1450#endif 1451 1452 /* 1453 * If forwarding packet using same interface that it came in on, 1454 * perhaps should send a redirect to sender to shortcut a hop. 1455 * Only send redirect if source is sending directly to us, 1456 * and if packet was not source routed (or has any options). 1457 * Also, don't send redirect if forwarding using a default route 1458 * or a route modified by a redirect. 1459 */ 1460 dest.s_addr = 0; 1461 if (!srcrt && V_ipsendredirects && 1462 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) { 1463 struct sockaddr_in *sin; 1464 struct rtentry *rt; 1465 1466 bzero(&ro, sizeof(ro)); 1467 sin = (struct sockaddr_in *)&ro.ro_dst; 1468 sin->sin_family = AF_INET; 1469 sin->sin_len = sizeof(*sin); 1470 sin->sin_addr = ip->ip_dst; 1471 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 1472 1473 rt = ro.ro_rt; 1474 1475 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1476 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1477#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1478 u_long src = ntohl(ip->ip_src.s_addr); 1479 1480 if (RTA(rt) && 1481 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1482 if (rt->rt_flags & RTF_GATEWAY) 1483 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1484 else 1485 dest.s_addr = ip->ip_dst.s_addr; 1486 /* Router requirements says to only send host redirects */ 1487 type = ICMP_REDIRECT; 1488 code = ICMP_REDIRECT_HOST; 1489 } 1490 } 1491 if (rt) 1492 RTFREE(rt); 1493 } 1494 1495 /* 1496 * Try to cache the route MTU from ip_output so we can consider it for 1497 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. 1498 */ 1499 bzero(&ro, sizeof(ro)); 1500 1501 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1502 1503 if (error == EMSGSIZE && ro.ro_rt) 1504 mtu = ro.ro_rt->rt_rmx.rmx_mtu; 1505 RO_RTFREE(&ro); 1506 1507 if (error) 1508 IPSTAT_INC(ips_cantforward); 1509 else { 1510 IPSTAT_INC(ips_forward); 1511 if (type) 1512 IPSTAT_INC(ips_redirectsent); 1513 else { 1514 if (mcopy) 1515 m_freem(mcopy); 1516 if (ia != NULL) 1517 ifa_free(&ia->ia_ifa); 1518 return; 1519 } 1520 } 1521 if (mcopy == NULL) { 1522 if (ia != NULL) 1523 ifa_free(&ia->ia_ifa); 1524 return; 1525 } 1526 1527 switch (error) { 1528 1529 case 0: /* forwarded, but need redirect */ 1530 /* type, code set above */ 1531 break; 1532 1533 case ENETUNREACH: 1534 case EHOSTUNREACH: 1535 case ENETDOWN: 1536 case EHOSTDOWN: 1537 default: 1538 type = ICMP_UNREACH; 1539 code = ICMP_UNREACH_HOST; 1540 break; 1541 1542 case EMSGSIZE: 1543 type = ICMP_UNREACH; 1544 code = ICMP_UNREACH_NEEDFRAG; 1545 1546#ifdef IPSEC 1547 /* 1548 * If IPsec is configured for this path, 1549 * override any possibly mtu value set by ip_output. 1550 */ 1551 mtu = ip_ipsec_mtu(mcopy, mtu); 1552#endif /* IPSEC */ 1553 /* 1554 * If the MTU was set before make sure we are below the 1555 * interface MTU. 1556 * If the MTU wasn't set before use the interface mtu or 1557 * fall back to the next smaller mtu step compared to the 1558 * current packet size. 1559 */ 1560 if (mtu != 0) { 1561 if (ia != NULL) 1562 mtu = min(mtu, ia->ia_ifp->if_mtu); 1563 } else { 1564 if (ia != NULL) 1565 mtu = ia->ia_ifp->if_mtu; 1566 else 1567 mtu = ip_next_mtu(ip->ip_len, 0); 1568 } 1569 IPSTAT_INC(ips_cantfrag); 1570 break; 1571 1572 case ENOBUFS: 1573 /* 1574 * A router should not generate ICMP_SOURCEQUENCH as 1575 * required in RFC1812 Requirements for IP Version 4 Routers. 1576 * Source quench could be a big problem under DoS attacks, 1577 * or if the underlying interface is rate-limited. 1578 * Those who need source quench packets may re-enable them 1579 * via the net.inet.ip.sendsourcequench sysctl. 1580 */ 1581 if (V_ip_sendsourcequench == 0) { 1582 m_freem(mcopy); 1583 if (ia != NULL) 1584 ifa_free(&ia->ia_ifa); 1585 return; 1586 } else { 1587 type = ICMP_SOURCEQUENCH; 1588 code = 0; 1589 } 1590 break; 1591 1592 case EACCES: /* ipfw denied packet */ 1593 m_freem(mcopy); 1594 if (ia != NULL) 1595 ifa_free(&ia->ia_ifa); 1596 return; 1597 } 1598 if (ia != NULL) 1599 ifa_free(&ia->ia_ifa); 1600 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1601} 1602 1603void 1604ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1605 struct mbuf *m) 1606{ 1607 1608 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1609 struct bintime bt; 1610 1611 bintime(&bt); 1612 if (inp->inp_socket->so_options & SO_BINTIME) { 1613 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1614 SCM_BINTIME, SOL_SOCKET); 1615 if (*mp) 1616 mp = &(*mp)->m_next; 1617 } 1618 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1619 struct timeval tv; 1620 1621 bintime2timeval(&bt, &tv); 1622 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1623 SCM_TIMESTAMP, SOL_SOCKET); 1624 if (*mp) 1625 mp = &(*mp)->m_next; 1626 } 1627 } 1628 if (inp->inp_flags & INP_RECVDSTADDR) { 1629 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1630 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1631 if (*mp) 1632 mp = &(*mp)->m_next; 1633 } 1634 if (inp->inp_flags & INP_RECVTTL) { 1635 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1636 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1637 if (*mp) 1638 mp = &(*mp)->m_next; 1639 } 1640#ifdef notyet 1641 /* XXX 1642 * Moving these out of udp_input() made them even more broken 1643 * than they already were. 1644 */ 1645 /* options were tossed already */ 1646 if (inp->inp_flags & INP_RECVOPTS) { 1647 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1648 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1649 if (*mp) 1650 mp = &(*mp)->m_next; 1651 } 1652 /* ip_srcroute doesn't do what we want here, need to fix */ 1653 if (inp->inp_flags & INP_RECVRETOPTS) { 1654 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1655 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1656 if (*mp) 1657 mp = &(*mp)->m_next; 1658 } 1659#endif 1660 if (inp->inp_flags & INP_RECVIF) { 1661 struct ifnet *ifp; 1662 struct sdlbuf { 1663 struct sockaddr_dl sdl; 1664 u_char pad[32]; 1665 } sdlbuf; 1666 struct sockaddr_dl *sdp; 1667 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1668 1669 if (((ifp = m->m_pkthdr.rcvif)) 1670 && ( ifp->if_index && (ifp->if_index <= V_if_index))) { 1671 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1672 /* 1673 * Change our mind and don't try copy. 1674 */ 1675 if ((sdp->sdl_family != AF_LINK) 1676 || (sdp->sdl_len > sizeof(sdlbuf))) { 1677 goto makedummy; 1678 } 1679 bcopy(sdp, sdl2, sdp->sdl_len); 1680 } else { 1681makedummy: 1682 sdl2->sdl_len 1683 = offsetof(struct sockaddr_dl, sdl_data[0]); 1684 sdl2->sdl_family = AF_LINK; 1685 sdl2->sdl_index = 0; 1686 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1687 } 1688 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1689 IP_RECVIF, IPPROTO_IP); 1690 if (*mp) 1691 mp = &(*mp)->m_next; 1692 } 1693 if (inp->inp_flags & INP_RECVTOS) { 1694 *mp = sbcreatecontrol((caddr_t) &ip->ip_tos, 1695 sizeof(u_char), IP_RECVTOS, IPPROTO_IP); 1696 if (*mp) 1697 mp = &(*mp)->m_next; 1698 } 1699} 1700 1701/* 1702 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1703 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1704 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1705 * compiled. 1706 */ 1707static VNET_DEFINE(int, ip_rsvp_on); 1708VNET_DEFINE(struct socket *, ip_rsvpd); 1709 1710#define V_ip_rsvp_on VNET(ip_rsvp_on) 1711 1712int 1713ip_rsvp_init(struct socket *so) 1714{ 1715 1716 if (so->so_type != SOCK_RAW || 1717 so->so_proto->pr_protocol != IPPROTO_RSVP) 1718 return EOPNOTSUPP; 1719 1720 if (V_ip_rsvpd != NULL) 1721 return EADDRINUSE; 1722 1723 V_ip_rsvpd = so; 1724 /* 1725 * This may seem silly, but we need to be sure we don't over-increment 1726 * the RSVP counter, in case something slips up. 1727 */ 1728 if (!V_ip_rsvp_on) { 1729 V_ip_rsvp_on = 1; 1730 V_rsvp_on++; 1731 } 1732 1733 return 0; 1734} 1735 1736int 1737ip_rsvp_done(void) 1738{ 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 1752void 1753rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1754{ 1755 1756 if (rsvp_input_p) { /* call the real one if loaded */ 1757 rsvp_input_p(m, off); 1758 return; 1759 } 1760 1761 /* Can still get packets with rsvp_on = 0 if there is a local member 1762 * of the group to which the RSVP packet is addressed. But in this 1763 * case we want to throw the packet away. 1764 */ 1765 1766 if (!V_rsvp_on) { 1767 m_freem(m); 1768 return; 1769 } 1770 1771 if (V_ip_rsvpd != NULL) { 1772 rip_input(m, off); 1773 return; 1774 } 1775 /* Drop the packet */ 1776 m_freem(m); 1777} 1778