ip_input.c revision 262743
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: stable/10/sys/netinet/ip_input.c 262743 2014-03-04 15:14:47Z 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_kdtrace.h" 40#include "opt_route.h" 41 42#include <sys/param.h> 43#include <sys/systm.h> 44#include <sys/mbuf.h> 45#include <sys/malloc.h> 46#include <sys/domain.h> 47#include <sys/protosw.h> 48#include <sys/socket.h> 49#include <sys/time.h> 50#include <sys/kernel.h> 51#include <sys/lock.h> 52#include <sys/rwlock.h> 53#include <sys/sdt.h> 54#include <sys/syslog.h> 55#include <sys/sysctl.h> 56 57#include <net/pfil.h> 58#include <net/if.h> 59#include <net/if_types.h> 60#include <net/if_var.h> 61#include <net/if_dl.h> 62#include <net/route.h> 63#include <net/netisr.h> 64#include <net/vnet.h> 65 66#include <netinet/in.h> 67#include <netinet/in_kdtrace.h> 68#include <netinet/in_systm.h> 69#include <netinet/in_var.h> 70#include <netinet/ip.h> 71#include <netinet/in_pcb.h> 72#include <netinet/ip_var.h> 73#include <netinet/ip_fw.h> 74#include <netinet/ip_icmp.h> 75#include <netinet/ip_options.h> 76#include <machine/in_cksum.h> 77#include <netinet/ip_carp.h> 78#ifdef IPSEC 79#include <netinet/ip_ipsec.h> 80#endif /* IPSEC */ 81 82#include <sys/socketvar.h> 83 84#include <security/mac/mac_framework.h> 85 86#ifdef CTASSERT 87CTASSERT(sizeof(struct ip) == 20); 88#endif 89 90struct rwlock in_ifaddr_lock; 91RW_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock"); 92 93VNET_DEFINE(int, rsvp_on); 94 95VNET_DEFINE(int, ipforwarding); 96SYSCTL_VNET_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 97 &VNET_NAME(ipforwarding), 0, 98 "Enable IP forwarding between interfaces"); 99 100static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */ 101#define V_ipsendredirects VNET(ipsendredirects) 102SYSCTL_VNET_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 103 &VNET_NAME(ipsendredirects), 0, 104 "Enable sending IP redirects"); 105 106static VNET_DEFINE(int, ip_keepfaith); 107#define V_ip_keepfaith VNET(ip_keepfaith) 108SYSCTL_VNET_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 109 &VNET_NAME(ip_keepfaith), 0, 110 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 111 112static VNET_DEFINE(int, ip_sendsourcequench); 113#define V_ip_sendsourcequench VNET(ip_sendsourcequench) 114SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 115 &VNET_NAME(ip_sendsourcequench), 0, 116 "Enable the transmission of source quench packets"); 117 118VNET_DEFINE(int, ip_do_randomid); 119SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 120 &VNET_NAME(ip_do_randomid), 0, 121 "Assign random ip_id values"); 122 123/* 124 * XXX - Setting ip_checkinterface mostly implements the receive side of 125 * the Strong ES model described in RFC 1122, but since the routing table 126 * and transmit implementation do not implement the Strong ES model, 127 * setting this to 1 results in an odd hybrid. 128 * 129 * XXX - ip_checkinterface currently must be disabled if you use ipnat 130 * to translate the destination address to another local interface. 131 * 132 * XXX - ip_checkinterface must be disabled if you add IP aliases 133 * to the loopback interface instead of the interface where the 134 * packets for those addresses are received. 135 */ 136static VNET_DEFINE(int, ip_checkinterface); 137#define V_ip_checkinterface VNET(ip_checkinterface) 138SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 139 &VNET_NAME(ip_checkinterface), 0, 140 "Verify packet arrives on correct interface"); 141 142VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */ 143 144static struct netisr_handler ip_nh = { 145 .nh_name = "ip", 146 .nh_handler = ip_input, 147 .nh_proto = NETISR_IP, 148 .nh_policy = NETISR_POLICY_FLOW, 149}; 150 151extern struct domain inetdomain; 152extern struct protosw inetsw[]; 153u_char ip_protox[IPPROTO_MAX]; 154VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */ 155VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */ 156VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */ 157 158static VNET_DEFINE(uma_zone_t, ipq_zone); 159static VNET_DEFINE(TAILQ_HEAD(ipqhead, ipq), ipq[IPREASS_NHASH]); 160static struct mtx ipqlock; 161 162#define V_ipq_zone VNET(ipq_zone) 163#define V_ipq VNET(ipq) 164 165#define IPQ_LOCK() mtx_lock(&ipqlock) 166#define IPQ_UNLOCK() mtx_unlock(&ipqlock) 167#define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 168#define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 169 170static void maxnipq_update(void); 171static void ipq_zone_change(void *); 172static void ip_drain_locked(void); 173 174static VNET_DEFINE(int, maxnipq); /* Administrative limit on # reass queues. */ 175static VNET_DEFINE(int, nipq); /* Total # of reass queues */ 176#define V_maxnipq VNET(maxnipq) 177#define V_nipq VNET(nipq) 178SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, 179 &VNET_NAME(nipq), 0, 180 "Current number of IPv4 fragment reassembly queue entries"); 181 182static VNET_DEFINE(int, maxfragsperpacket); 183#define V_maxfragsperpacket VNET(maxfragsperpacket) 184SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 185 &VNET_NAME(maxfragsperpacket), 0, 186 "Maximum number of IPv4 fragments allowed per packet"); 187 188#ifdef IPCTL_DEFMTU 189SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 190 &ip_mtu, 0, "Default MTU"); 191#endif 192 193#ifdef IPSTEALTH 194VNET_DEFINE(int, ipstealth); 195SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 196 &VNET_NAME(ipstealth), 0, 197 "IP stealth mode, no TTL decrementation on forwarding"); 198#endif 199 200static void ip_freef(struct ipqhead *, struct ipq *); 201 202/* 203 * IP statistics are stored in the "array" of counter(9)s. 204 */ 205VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat); 206VNET_PCPUSTAT_SYSINIT(ipstat); 207SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat, 208 "IP statistics (struct ipstat, netinet/ip_var.h)"); 209 210#ifdef VIMAGE 211VNET_PCPUSTAT_SYSUNINIT(ipstat); 212#endif /* VIMAGE */ 213 214/* 215 * Kernel module interface for updating ipstat. The argument is an index 216 * into ipstat treated as an array. 217 */ 218void 219kmod_ipstat_inc(int statnum) 220{ 221 222 counter_u64_add(VNET(ipstat)[statnum], 1); 223} 224 225void 226kmod_ipstat_dec(int statnum) 227{ 228 229 counter_u64_add(VNET(ipstat)[statnum], -1); 230} 231 232static int 233sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 234{ 235 int error, qlimit; 236 237 netisr_getqlimit(&ip_nh, &qlimit); 238 error = sysctl_handle_int(oidp, &qlimit, 0, req); 239 if (error || !req->newptr) 240 return (error); 241 if (qlimit < 1) 242 return (EINVAL); 243 return (netisr_setqlimit(&ip_nh, qlimit)); 244} 245SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, 246 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I", 247 "Maximum size of the IP input queue"); 248 249static int 250sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) 251{ 252 u_int64_t qdrops_long; 253 int error, qdrops; 254 255 netisr_getqdrops(&ip_nh, &qdrops_long); 256 qdrops = qdrops_long; 257 error = sysctl_handle_int(oidp, &qdrops, 0, req); 258 if (error || !req->newptr) 259 return (error); 260 if (qdrops != 0) 261 return (EINVAL); 262 netisr_clearqdrops(&ip_nh); 263 return (0); 264} 265 266SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, 267 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I", 268 "Number of packets dropped from the IP input queue"); 269 270/* 271 * IP initialization: fill in IP protocol switch table. 272 * All protocols not implemented in kernel go to raw IP protocol handler. 273 */ 274void 275ip_init(void) 276{ 277 struct protosw *pr; 278 int i; 279 280 V_ip_id = time_second & 0xffff; 281 282 TAILQ_INIT(&V_in_ifaddrhead); 283 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 284 285 /* Initialize IP reassembly queue. */ 286 for (i = 0; i < IPREASS_NHASH; i++) 287 TAILQ_INIT(&V_ipq[i]); 288 V_maxnipq = nmbclusters / 32; 289 V_maxfragsperpacket = 16; 290 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 291 NULL, UMA_ALIGN_PTR, 0); 292 maxnipq_update(); 293 294 /* Initialize packet filter hooks. */ 295 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF; 296 V_inet_pfil_hook.ph_af = AF_INET; 297 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0) 298 printf("%s: WARNING: unable to register pfil hook, " 299 "error %d\n", __func__, i); 300 301 /* Skip initialization of globals for non-default instances. */ 302 if (!IS_DEFAULT_VNET(curvnet)) 303 return; 304 305 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 306 if (pr == NULL) 307 panic("ip_init: PF_INET not found"); 308 309 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 310 for (i = 0; i < IPPROTO_MAX; i++) 311 ip_protox[i] = pr - inetsw; 312 /* 313 * Cycle through IP protocols and put them into the appropriate place 314 * in ip_protox[]. 315 */ 316 for (pr = inetdomain.dom_protosw; 317 pr < inetdomain.dom_protoswNPROTOSW; pr++) 318 if (pr->pr_domain->dom_family == PF_INET && 319 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 320 /* Be careful to only index valid IP protocols. */ 321 if (pr->pr_protocol < IPPROTO_MAX) 322 ip_protox[pr->pr_protocol] = pr - inetsw; 323 } 324 325 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, 326 NULL, EVENTHANDLER_PRI_ANY); 327 328 /* Initialize various other remaining things. */ 329 IPQ_LOCK_INIT(); 330 netisr_register(&ip_nh); 331} 332 333#ifdef VIMAGE 334void 335ip_destroy(void) 336{ 337 int i; 338 339 if ((i = pfil_head_unregister(&V_inet_pfil_hook)) != 0) 340 printf("%s: WARNING: unable to unregister pfil hook, " 341 "error %d\n", __func__, i); 342 343 /* Cleanup in_ifaddr hash table; should be empty. */ 344 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask); 345 346 IPQ_LOCK(); 347 ip_drain_locked(); 348 IPQ_UNLOCK(); 349 350 uma_zdestroy(V_ipq_zone); 351} 352#endif 353 354/* 355 * Ip input routine. Checksum and byte swap header. If fragmented 356 * try to reassemble. Process options. Pass to next level. 357 */ 358void 359ip_input(struct mbuf *m) 360{ 361 struct ip *ip = NULL; 362 struct in_ifaddr *ia = NULL; 363 struct ifaddr *ifa; 364 struct ifnet *ifp; 365 int checkif, hlen = 0; 366 uint16_t sum, ip_len; 367 int dchg = 0; /* dest changed after fw */ 368 struct in_addr odst; /* original dst address */ 369 370 M_ASSERTPKTHDR(m); 371 372 if (m->m_flags & M_FASTFWD_OURS) { 373 m->m_flags &= ~M_FASTFWD_OURS; 374 /* Set up some basics that will be used later. */ 375 ip = mtod(m, struct ip *); 376 hlen = ip->ip_hl << 2; 377 ip_len = ntohs(ip->ip_len); 378 goto ours; 379 } 380 381 IPSTAT_INC(ips_total); 382 383 if (m->m_pkthdr.len < sizeof(struct ip)) 384 goto tooshort; 385 386 if (m->m_len < sizeof (struct ip) && 387 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 388 IPSTAT_INC(ips_toosmall); 389 return; 390 } 391 ip = mtod(m, struct ip *); 392 393 if (ip->ip_v != IPVERSION) { 394 IPSTAT_INC(ips_badvers); 395 goto bad; 396 } 397 398 hlen = ip->ip_hl << 2; 399 if (hlen < sizeof(struct ip)) { /* minimum header length */ 400 IPSTAT_INC(ips_badhlen); 401 goto bad; 402 } 403 if (hlen > m->m_len) { 404 if ((m = m_pullup(m, hlen)) == NULL) { 405 IPSTAT_INC(ips_badhlen); 406 return; 407 } 408 ip = mtod(m, struct ip *); 409 } 410 411 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL); 412 413 /* 127/8 must not appear on wire - RFC1122 */ 414 ifp = m->m_pkthdr.rcvif; 415 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 416 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 417 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 418 IPSTAT_INC(ips_badaddr); 419 goto bad; 420 } 421 } 422 423 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 424 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 425 } else { 426 if (hlen == sizeof(struct ip)) { 427 sum = in_cksum_hdr(ip); 428 } else { 429 sum = in_cksum(m, hlen); 430 } 431 } 432 if (sum) { 433 IPSTAT_INC(ips_badsum); 434 goto bad; 435 } 436 437#ifdef ALTQ 438 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 439 /* packet is dropped by traffic conditioner */ 440 return; 441#endif 442 443 ip_len = ntohs(ip->ip_len); 444 if (ip_len < hlen) { 445 IPSTAT_INC(ips_badlen); 446 goto bad; 447 } 448 449 /* 450 * Check that the amount of data in the buffers 451 * is as at least much as the IP header would have us expect. 452 * Trim mbufs if longer than we expect. 453 * Drop packet if shorter than we expect. 454 */ 455 if (m->m_pkthdr.len < ip_len) { 456tooshort: 457 IPSTAT_INC(ips_tooshort); 458 goto bad; 459 } 460 if (m->m_pkthdr.len > ip_len) { 461 if (m->m_len == m->m_pkthdr.len) { 462 m->m_len = ip_len; 463 m->m_pkthdr.len = ip_len; 464 } else 465 m_adj(m, ip_len - m->m_pkthdr.len); 466 } 467#ifdef IPSEC 468 /* 469 * Bypass packet filtering for packets previously handled by IPsec. 470 */ 471 if (ip_ipsec_filtertunnel(m)) 472 goto passin; 473#endif /* IPSEC */ 474 475 /* 476 * Run through list of hooks for input packets. 477 * 478 * NB: Beware of the destination address changing (e.g. 479 * by NAT rewriting). When this happens, tell 480 * ip_forward to do the right thing. 481 */ 482 483 /* Jump over all PFIL processing if hooks are not active. */ 484 if (!PFIL_HOOKED(&V_inet_pfil_hook)) 485 goto passin; 486 487 odst = ip->ip_dst; 488 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0) 489 return; 490 if (m == NULL) /* consumed by filter */ 491 return; 492 493 ip = mtod(m, struct ip *); 494 dchg = (odst.s_addr != ip->ip_dst.s_addr); 495 ifp = m->m_pkthdr.rcvif; 496 497 if (m->m_flags & M_FASTFWD_OURS) { 498 m->m_flags &= ~M_FASTFWD_OURS; 499 goto ours; 500 } 501 if (m->m_flags & M_IP_NEXTHOP) { 502 dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL); 503 if (dchg != 0) { 504 /* 505 * Directly ship the packet on. This allows 506 * forwarding packets originally destined to us 507 * to some other directly connected host. 508 */ 509 ip_forward(m, 1); 510 return; 511 } 512 } 513passin: 514 515 /* 516 * Process options and, if not destined for us, 517 * ship it on. ip_dooptions returns 1 when an 518 * error was detected (causing an icmp message 519 * to be sent and the original packet to be freed). 520 */ 521 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 522 return; 523 524 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 525 * matter if it is destined to another node, or whether it is 526 * a multicast one, RSVP wants it! and prevents it from being forwarded 527 * anywhere else. Also checks if the rsvp daemon is running before 528 * grabbing the packet. 529 */ 530 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 531 goto ours; 532 533 /* 534 * Check our list of addresses, to see if the packet is for us. 535 * If we don't have any addresses, assume any unicast packet 536 * we receive might be for us (and let the upper layers deal 537 * with it). 538 */ 539 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 540 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 541 goto ours; 542 543 /* 544 * Enable a consistency check between the destination address 545 * and the arrival interface for a unicast packet (the RFC 1122 546 * strong ES model) if IP forwarding is disabled and the packet 547 * is not locally generated and the packet is not subject to 548 * 'ipfw fwd'. 549 * 550 * XXX - Checking also should be disabled if the destination 551 * address is ipnat'ed to a different interface. 552 * 553 * XXX - Checking is incompatible with IP aliases added 554 * to the loopback interface instead of the interface where 555 * the packets are received. 556 * 557 * XXX - This is the case for carp vhost IPs as well so we 558 * insert a workaround. If the packet got here, we already 559 * checked with carp_iamatch() and carp_forus(). 560 */ 561 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 562 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && 563 ifp->if_carp == NULL && (dchg == 0); 564 565 /* 566 * Check for exact addresses in the hash bucket. 567 */ 568 /* IN_IFADDR_RLOCK(); */ 569 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 570 /* 571 * If the address matches, verify that the packet 572 * arrived via the correct interface if checking is 573 * enabled. 574 */ 575 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 576 (!checkif || ia->ia_ifp == ifp)) { 577 ifa_ref(&ia->ia_ifa); 578 /* IN_IFADDR_RUNLOCK(); */ 579 goto ours; 580 } 581 } 582 /* IN_IFADDR_RUNLOCK(); */ 583 584 /* 585 * Check for broadcast addresses. 586 * 587 * Only accept broadcast packets that arrive via the matching 588 * interface. Reception of forwarded directed broadcasts would 589 * be handled via ip_forward() and ether_output() with the loopback 590 * into the stack for SIMPLEX interfaces handled by ether_output(). 591 */ 592 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 593 IF_ADDR_RLOCK(ifp); 594 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 595 if (ifa->ifa_addr->sa_family != AF_INET) 596 continue; 597 ia = ifatoia(ifa); 598 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 599 ip->ip_dst.s_addr) { 600 ifa_ref(ifa); 601 IF_ADDR_RUNLOCK(ifp); 602 goto ours; 603 } 604#ifdef BOOTP_COMPAT 605 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { 606 ifa_ref(ifa); 607 IF_ADDR_RUNLOCK(ifp); 608 goto ours; 609 } 610#endif 611 } 612 IF_ADDR_RUNLOCK(ifp); 613 ia = NULL; 614 } 615 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 616 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 617 IPSTAT_INC(ips_cantforward); 618 m_freem(m); 619 return; 620 } 621 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 622 if (V_ip_mrouter) { 623 /* 624 * If we are acting as a multicast router, all 625 * incoming multicast packets are passed to the 626 * kernel-level multicast forwarding function. 627 * The packet is returned (relatively) intact; if 628 * ip_mforward() returns a non-zero value, the packet 629 * must be discarded, else it may be accepted below. 630 */ 631 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 632 IPSTAT_INC(ips_cantforward); 633 m_freem(m); 634 return; 635 } 636 637 /* 638 * The process-level routing daemon needs to receive 639 * all multicast IGMP packets, whether or not this 640 * host belongs to their destination groups. 641 */ 642 if (ip->ip_p == IPPROTO_IGMP) 643 goto ours; 644 IPSTAT_INC(ips_forward); 645 } 646 /* 647 * Assume the packet is for us, to avoid prematurely taking 648 * a lock on the in_multi hash. Protocols must perform 649 * their own filtering and update statistics accordingly. 650 */ 651 goto ours; 652 } 653 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 654 goto ours; 655 if (ip->ip_dst.s_addr == INADDR_ANY) 656 goto ours; 657 658 /* 659 * FAITH(Firewall Aided Internet Translator) 660 */ 661 if (ifp && ifp->if_type == IFT_FAITH) { 662 if (V_ip_keepfaith) { 663 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 664 goto ours; 665 } 666 m_freem(m); 667 return; 668 } 669 670 /* 671 * Not for us; forward if possible and desirable. 672 */ 673 if (V_ipforwarding == 0) { 674 IPSTAT_INC(ips_cantforward); 675 m_freem(m); 676 } else { 677#ifdef IPSEC 678 if (ip_ipsec_fwd(m)) 679 goto bad; 680#endif /* IPSEC */ 681 ip_forward(m, dchg); 682 } 683 return; 684 685ours: 686#ifdef IPSTEALTH 687 /* 688 * IPSTEALTH: Process non-routing options only 689 * if the packet is destined for us. 690 */ 691 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) { 692 if (ia != NULL) 693 ifa_free(&ia->ia_ifa); 694 return; 695 } 696#endif /* IPSTEALTH */ 697 698 /* Count the packet in the ip address stats */ 699 if (ia != NULL) { 700 ia->ia_ifa.if_ipackets++; 701 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 702 ifa_free(&ia->ia_ifa); 703 } 704 705 /* 706 * Attempt reassembly; if it succeeds, proceed. 707 * ip_reass() will return a different mbuf. 708 */ 709 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) { 710 m = ip_reass(m); 711 if (m == NULL) 712 return; 713 ip = mtod(m, struct ip *); 714 /* Get the header length of the reassembled packet */ 715 hlen = ip->ip_hl << 2; 716 } 717 718#ifdef IPSEC 719 /* 720 * enforce IPsec policy checking if we are seeing last header. 721 * note that we do not visit this with protocols with pcb layer 722 * code - like udp/tcp/raw ip. 723 */ 724 if (ip_ipsec_input(m)) 725 goto bad; 726#endif /* IPSEC */ 727 728 /* 729 * Switch out to protocol's input routine. 730 */ 731 IPSTAT_INC(ips_delivered); 732 733 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 734 return; 735bad: 736 m_freem(m); 737} 738 739/* 740 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 741 * max has slightly different semantics than the sysctl, for historical 742 * reasons. 743 */ 744static void 745maxnipq_update(void) 746{ 747 748 /* 749 * -1 for unlimited allocation. 750 */ 751 if (V_maxnipq < 0) 752 uma_zone_set_max(V_ipq_zone, 0); 753 /* 754 * Positive number for specific bound. 755 */ 756 if (V_maxnipq > 0) 757 uma_zone_set_max(V_ipq_zone, V_maxnipq); 758 /* 759 * Zero specifies no further fragment queue allocation -- set the 760 * bound very low, but rely on implementation elsewhere to actually 761 * prevent allocation and reclaim current queues. 762 */ 763 if (V_maxnipq == 0) 764 uma_zone_set_max(V_ipq_zone, 1); 765} 766 767static void 768ipq_zone_change(void *tag) 769{ 770 771 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) { 772 V_maxnipq = nmbclusters / 32; 773 maxnipq_update(); 774 } 775} 776 777static int 778sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 779{ 780 int error, i; 781 782 i = V_maxnipq; 783 error = sysctl_handle_int(oidp, &i, 0, req); 784 if (error || !req->newptr) 785 return (error); 786 787 /* 788 * XXXRW: Might be a good idea to sanity check the argument and place 789 * an extreme upper bound. 790 */ 791 if (i < -1) 792 return (EINVAL); 793 V_maxnipq = i; 794 maxnipq_update(); 795 return (0); 796} 797 798SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 799 NULL, 0, sysctl_maxnipq, "I", 800 "Maximum number of IPv4 fragment reassembly queue entries"); 801 802/* 803 * Take incoming datagram fragment and try to reassemble it into 804 * whole datagram. If the argument is the first fragment or one 805 * in between the function will return NULL and store the mbuf 806 * in the fragment chain. If the argument is the last fragment 807 * the packet will be reassembled and the pointer to the new 808 * mbuf returned for further processing. Only m_tags attached 809 * to the first packet/fragment are preserved. 810 * The IP header is *NOT* adjusted out of iplen. 811 */ 812struct mbuf * 813ip_reass(struct mbuf *m) 814{ 815 struct ip *ip; 816 struct mbuf *p, *q, *nq, *t; 817 struct ipq *fp = NULL; 818 struct ipqhead *head; 819 int i, hlen, next; 820 u_int8_t ecn, ecn0; 821 u_short hash; 822 823 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 824 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) { 825 IPSTAT_INC(ips_fragments); 826 IPSTAT_INC(ips_fragdropped); 827 m_freem(m); 828 return (NULL); 829 } 830 831 ip = mtod(m, struct ip *); 832 hlen = ip->ip_hl << 2; 833 834 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 835 head = &V_ipq[hash]; 836 IPQ_LOCK(); 837 838 /* 839 * Look for queue of fragments 840 * of this datagram. 841 */ 842 TAILQ_FOREACH(fp, head, ipq_list) 843 if (ip->ip_id == fp->ipq_id && 844 ip->ip_src.s_addr == fp->ipq_src.s_addr && 845 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 846#ifdef MAC 847 mac_ipq_match(m, fp) && 848#endif 849 ip->ip_p == fp->ipq_p) 850 goto found; 851 852 fp = NULL; 853 854 /* 855 * Attempt to trim the number of allocated fragment queues if it 856 * exceeds the administrative limit. 857 */ 858 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) { 859 /* 860 * drop something from the tail of the current queue 861 * before proceeding further 862 */ 863 struct ipq *q = TAILQ_LAST(head, ipqhead); 864 if (q == NULL) { /* gak */ 865 for (i = 0; i < IPREASS_NHASH; i++) { 866 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead); 867 if (r) { 868 IPSTAT_ADD(ips_fragtimeout, 869 r->ipq_nfrags); 870 ip_freef(&V_ipq[i], r); 871 break; 872 } 873 } 874 } else { 875 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags); 876 ip_freef(head, q); 877 } 878 } 879 880found: 881 /* 882 * Adjust ip_len to not reflect header, 883 * convert offset of this to bytes. 884 */ 885 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 886 if (ip->ip_off & htons(IP_MF)) { 887 /* 888 * Make sure that fragments have a data length 889 * that's a non-zero multiple of 8 bytes. 890 */ 891 if (ip->ip_len == htons(0) || (ntohs(ip->ip_len) & 0x7) != 0) { 892 IPSTAT_INC(ips_toosmall); /* XXX */ 893 goto dropfrag; 894 } 895 m->m_flags |= M_IP_FRAG; 896 } else 897 m->m_flags &= ~M_IP_FRAG; 898 ip->ip_off = htons(ntohs(ip->ip_off) << 3); 899 900 /* 901 * Attempt reassembly; if it succeeds, proceed. 902 * ip_reass() will return a different mbuf. 903 */ 904 IPSTAT_INC(ips_fragments); 905 m->m_pkthdr.PH_loc.ptr = ip; 906 907 /* Previous ip_reass() started here. */ 908 /* 909 * Presence of header sizes in mbufs 910 * would confuse code below. 911 */ 912 m->m_data += hlen; 913 m->m_len -= hlen; 914 915 /* 916 * If first fragment to arrive, create a reassembly queue. 917 */ 918 if (fp == NULL) { 919 fp = uma_zalloc(V_ipq_zone, M_NOWAIT); 920 if (fp == NULL) 921 goto dropfrag; 922#ifdef MAC 923 if (mac_ipq_init(fp, M_NOWAIT) != 0) { 924 uma_zfree(V_ipq_zone, fp); 925 fp = NULL; 926 goto dropfrag; 927 } 928 mac_ipq_create(m, fp); 929#endif 930 TAILQ_INSERT_HEAD(head, fp, ipq_list); 931 V_nipq++; 932 fp->ipq_nfrags = 1; 933 fp->ipq_ttl = IPFRAGTTL; 934 fp->ipq_p = ip->ip_p; 935 fp->ipq_id = ip->ip_id; 936 fp->ipq_src = ip->ip_src; 937 fp->ipq_dst = ip->ip_dst; 938 fp->ipq_frags = m; 939 m->m_nextpkt = NULL; 940 goto done; 941 } else { 942 fp->ipq_nfrags++; 943#ifdef MAC 944 mac_ipq_update(m, fp); 945#endif 946 } 947 948#define GETIP(m) ((struct ip*)((m)->m_pkthdr.PH_loc.ptr)) 949 950 /* 951 * Handle ECN by comparing this segment with the first one; 952 * if CE is set, do not lose CE. 953 * drop if CE and not-ECT are mixed for the same packet. 954 */ 955 ecn = ip->ip_tos & IPTOS_ECN_MASK; 956 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 957 if (ecn == IPTOS_ECN_CE) { 958 if (ecn0 == IPTOS_ECN_NOTECT) 959 goto dropfrag; 960 if (ecn0 != IPTOS_ECN_CE) 961 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 962 } 963 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 964 goto dropfrag; 965 966 /* 967 * Find a segment which begins after this one does. 968 */ 969 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 970 if (ntohs(GETIP(q)->ip_off) > ntohs(ip->ip_off)) 971 break; 972 973 /* 974 * If there is a preceding segment, it may provide some of 975 * our data already. If so, drop the data from the incoming 976 * segment. If it provides all of our data, drop us, otherwise 977 * stick new segment in the proper place. 978 * 979 * If some of the data is dropped from the preceding 980 * segment, then it's checksum is invalidated. 981 */ 982 if (p) { 983 i = ntohs(GETIP(p)->ip_off) + ntohs(GETIP(p)->ip_len) - 984 ntohs(ip->ip_off); 985 if (i > 0) { 986 if (i >= ntohs(ip->ip_len)) 987 goto dropfrag; 988 m_adj(m, i); 989 m->m_pkthdr.csum_flags = 0; 990 ip->ip_off = htons(ntohs(ip->ip_off) + i); 991 ip->ip_len = htons(ntohs(ip->ip_len) - i); 992 } 993 m->m_nextpkt = p->m_nextpkt; 994 p->m_nextpkt = m; 995 } else { 996 m->m_nextpkt = fp->ipq_frags; 997 fp->ipq_frags = m; 998 } 999 1000 /* 1001 * While we overlap succeeding segments trim them or, 1002 * if they are completely covered, dequeue them. 1003 */ 1004 for (; q != NULL && ntohs(ip->ip_off) + ntohs(ip->ip_len) > 1005 ntohs(GETIP(q)->ip_off); q = nq) { 1006 i = (ntohs(ip->ip_off) + ntohs(ip->ip_len)) - 1007 ntohs(GETIP(q)->ip_off); 1008 if (i < ntohs(GETIP(q)->ip_len)) { 1009 GETIP(q)->ip_len = htons(ntohs(GETIP(q)->ip_len) - i); 1010 GETIP(q)->ip_off = htons(ntohs(GETIP(q)->ip_off) + i); 1011 m_adj(q, i); 1012 q->m_pkthdr.csum_flags = 0; 1013 break; 1014 } 1015 nq = q->m_nextpkt; 1016 m->m_nextpkt = nq; 1017 IPSTAT_INC(ips_fragdropped); 1018 fp->ipq_nfrags--; 1019 m_freem(q); 1020 } 1021 1022 /* 1023 * Check for complete reassembly and perform frag per packet 1024 * limiting. 1025 * 1026 * Frag limiting is performed here so that the nth frag has 1027 * a chance to complete the packet before we drop the packet. 1028 * As a result, n+1 frags are actually allowed per packet, but 1029 * only n will ever be stored. (n = maxfragsperpacket.) 1030 * 1031 */ 1032 next = 0; 1033 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1034 if (ntohs(GETIP(q)->ip_off) != next) { 1035 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1036 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1037 ip_freef(head, fp); 1038 } 1039 goto done; 1040 } 1041 next += ntohs(GETIP(q)->ip_len); 1042 } 1043 /* Make sure the last packet didn't have the IP_MF flag */ 1044 if (p->m_flags & M_IP_FRAG) { 1045 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1046 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1047 ip_freef(head, fp); 1048 } 1049 goto done; 1050 } 1051 1052 /* 1053 * Reassembly is complete. Make sure the packet is a sane size. 1054 */ 1055 q = fp->ipq_frags; 1056 ip = GETIP(q); 1057 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1058 IPSTAT_INC(ips_toolong); 1059 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1060 ip_freef(head, fp); 1061 goto done; 1062 } 1063 1064 /* 1065 * Concatenate fragments. 1066 */ 1067 m = q; 1068 t = m->m_next; 1069 m->m_next = NULL; 1070 m_cat(m, t); 1071 nq = q->m_nextpkt; 1072 q->m_nextpkt = NULL; 1073 for (q = nq; q != NULL; q = nq) { 1074 nq = q->m_nextpkt; 1075 q->m_nextpkt = NULL; 1076 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1077 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1078 m_cat(m, q); 1079 } 1080 /* 1081 * In order to do checksumming faster we do 'end-around carry' here 1082 * (and not in for{} loop), though it implies we are not going to 1083 * reassemble more than 64k fragments. 1084 */ 1085 m->m_pkthdr.csum_data = 1086 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1087#ifdef MAC 1088 mac_ipq_reassemble(fp, m); 1089 mac_ipq_destroy(fp); 1090#endif 1091 1092 /* 1093 * Create header for new ip packet by modifying header of first 1094 * packet; dequeue and discard fragment reassembly header. 1095 * Make header visible. 1096 */ 1097 ip->ip_len = htons((ip->ip_hl << 2) + next); 1098 ip->ip_src = fp->ipq_src; 1099 ip->ip_dst = fp->ipq_dst; 1100 TAILQ_REMOVE(head, fp, ipq_list); 1101 V_nipq--; 1102 uma_zfree(V_ipq_zone, fp); 1103 m->m_len += (ip->ip_hl << 2); 1104 m->m_data -= (ip->ip_hl << 2); 1105 /* some debugging cruft by sklower, below, will go away soon */ 1106 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1107 m_fixhdr(m); 1108 IPSTAT_INC(ips_reassembled); 1109 IPQ_UNLOCK(); 1110 return (m); 1111 1112dropfrag: 1113 IPSTAT_INC(ips_fragdropped); 1114 if (fp != NULL) 1115 fp->ipq_nfrags--; 1116 m_freem(m); 1117done: 1118 IPQ_UNLOCK(); 1119 return (NULL); 1120 1121#undef GETIP 1122} 1123 1124/* 1125 * Free a fragment reassembly header and all 1126 * associated datagrams. 1127 */ 1128static void 1129ip_freef(struct ipqhead *fhp, struct ipq *fp) 1130{ 1131 struct mbuf *q; 1132 1133 IPQ_LOCK_ASSERT(); 1134 1135 while (fp->ipq_frags) { 1136 q = fp->ipq_frags; 1137 fp->ipq_frags = q->m_nextpkt; 1138 m_freem(q); 1139 } 1140 TAILQ_REMOVE(fhp, fp, ipq_list); 1141 uma_zfree(V_ipq_zone, fp); 1142 V_nipq--; 1143} 1144 1145/* 1146 * IP timer processing; 1147 * if a timer expires on a reassembly 1148 * queue, discard it. 1149 */ 1150void 1151ip_slowtimo(void) 1152{ 1153 VNET_ITERATOR_DECL(vnet_iter); 1154 struct ipq *fp; 1155 int i; 1156 1157 VNET_LIST_RLOCK_NOSLEEP(); 1158 IPQ_LOCK(); 1159 VNET_FOREACH(vnet_iter) { 1160 CURVNET_SET(vnet_iter); 1161 for (i = 0; i < IPREASS_NHASH; i++) { 1162 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) { 1163 struct ipq *fpp; 1164 1165 fpp = fp; 1166 fp = TAILQ_NEXT(fp, ipq_list); 1167 if(--fpp->ipq_ttl == 0) { 1168 IPSTAT_ADD(ips_fragtimeout, 1169 fpp->ipq_nfrags); 1170 ip_freef(&V_ipq[i], fpp); 1171 } 1172 } 1173 } 1174 /* 1175 * If we are over the maximum number of fragments 1176 * (due to the limit being lowered), drain off 1177 * enough to get down to the new limit. 1178 */ 1179 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) { 1180 for (i = 0; i < IPREASS_NHASH; i++) { 1181 while (V_nipq > V_maxnipq && 1182 !TAILQ_EMPTY(&V_ipq[i])) { 1183 IPSTAT_ADD(ips_fragdropped, 1184 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1185 ip_freef(&V_ipq[i], 1186 TAILQ_FIRST(&V_ipq[i])); 1187 } 1188 } 1189 } 1190 CURVNET_RESTORE(); 1191 } 1192 IPQ_UNLOCK(); 1193 VNET_LIST_RUNLOCK_NOSLEEP(); 1194} 1195 1196/* 1197 * Drain off all datagram fragments. 1198 */ 1199static void 1200ip_drain_locked(void) 1201{ 1202 int i; 1203 1204 IPQ_LOCK_ASSERT(); 1205 1206 for (i = 0; i < IPREASS_NHASH; i++) { 1207 while(!TAILQ_EMPTY(&V_ipq[i])) { 1208 IPSTAT_ADD(ips_fragdropped, 1209 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1210 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); 1211 } 1212 } 1213} 1214 1215void 1216ip_drain(void) 1217{ 1218 VNET_ITERATOR_DECL(vnet_iter); 1219 1220 VNET_LIST_RLOCK_NOSLEEP(); 1221 IPQ_LOCK(); 1222 VNET_FOREACH(vnet_iter) { 1223 CURVNET_SET(vnet_iter); 1224 ip_drain_locked(); 1225 CURVNET_RESTORE(); 1226 } 1227 IPQ_UNLOCK(); 1228 VNET_LIST_RUNLOCK_NOSLEEP(); 1229 in_rtqdrain(); 1230} 1231 1232/* 1233 * The protocol to be inserted into ip_protox[] must be already registered 1234 * in inetsw[], either statically or through pf_proto_register(). 1235 */ 1236int 1237ipproto_register(short ipproto) 1238{ 1239 struct protosw *pr; 1240 1241 /* Sanity checks. */ 1242 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 1243 return (EPROTONOSUPPORT); 1244 1245 /* 1246 * The protocol slot must not be occupied by another protocol 1247 * already. An index pointing to IPPROTO_RAW is unused. 1248 */ 1249 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1250 if (pr == NULL) 1251 return (EPFNOSUPPORT); 1252 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1253 return (EEXIST); 1254 1255 /* Find the protocol position in inetsw[] and set the index. */ 1256 for (pr = inetdomain.dom_protosw; 1257 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1258 if (pr->pr_domain->dom_family == PF_INET && 1259 pr->pr_protocol && pr->pr_protocol == ipproto) { 1260 ip_protox[pr->pr_protocol] = pr - inetsw; 1261 return (0); 1262 } 1263 } 1264 return (EPROTONOSUPPORT); 1265} 1266 1267int 1268ipproto_unregister(short ipproto) 1269{ 1270 struct protosw *pr; 1271 1272 /* Sanity checks. */ 1273 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 1274 return (EPROTONOSUPPORT); 1275 1276 /* Check if the protocol was indeed registered. */ 1277 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1278 if (pr == NULL) 1279 return (EPFNOSUPPORT); 1280 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1281 return (ENOENT); 1282 1283 /* Reset the protocol slot to IPPROTO_RAW. */ 1284 ip_protox[ipproto] = pr - inetsw; 1285 return (0); 1286} 1287 1288/* 1289 * Given address of next destination (final or next hop), return (referenced) 1290 * internet address info of interface to be used to get there. 1291 */ 1292struct in_ifaddr * 1293ip_rtaddr(struct in_addr dst, u_int fibnum) 1294{ 1295 struct route sro; 1296 struct sockaddr_in *sin; 1297 struct in_ifaddr *ia; 1298 1299 bzero(&sro, sizeof(sro)); 1300 sin = (struct sockaddr_in *)&sro.ro_dst; 1301 sin->sin_family = AF_INET; 1302 sin->sin_len = sizeof(*sin); 1303 sin->sin_addr = dst; 1304 in_rtalloc_ign(&sro, 0, fibnum); 1305 1306 if (sro.ro_rt == NULL) 1307 return (NULL); 1308 1309 ia = ifatoia(sro.ro_rt->rt_ifa); 1310 ifa_ref(&ia->ia_ifa); 1311 RTFREE(sro.ro_rt); 1312 return (ia); 1313} 1314 1315u_char inetctlerrmap[PRC_NCMDS] = { 1316 0, 0, 0, 0, 1317 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1318 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1319 EMSGSIZE, EHOSTUNREACH, 0, 0, 1320 0, 0, EHOSTUNREACH, 0, 1321 ENOPROTOOPT, ECONNREFUSED 1322}; 1323 1324/* 1325 * Forward a packet. If some error occurs return the sender 1326 * an icmp packet. Note we can't always generate a meaningful 1327 * icmp message because icmp doesn't have a large enough repertoire 1328 * of codes and types. 1329 * 1330 * If not forwarding, just drop the packet. This could be confusing 1331 * if ipforwarding was zero but some routing protocol was advancing 1332 * us as a gateway to somewhere. However, we must let the routing 1333 * protocol deal with that. 1334 * 1335 * The srcrt parameter indicates whether the packet is being forwarded 1336 * via a source route. 1337 */ 1338void 1339ip_forward(struct mbuf *m, int srcrt) 1340{ 1341 struct ip *ip = mtod(m, struct ip *); 1342 struct in_ifaddr *ia; 1343 struct mbuf *mcopy; 1344 struct in_addr dest; 1345 struct route ro; 1346 int error, type = 0, code = 0, mtu = 0; 1347 1348 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1349 IPSTAT_INC(ips_cantforward); 1350 m_freem(m); 1351 return; 1352 } 1353#ifdef IPSTEALTH 1354 if (!V_ipstealth) { 1355#endif 1356 if (ip->ip_ttl <= IPTTLDEC) { 1357 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1358 0, 0); 1359 return; 1360 } 1361#ifdef IPSTEALTH 1362 } 1363#endif 1364 1365 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); 1366#ifndef IPSEC 1367 /* 1368 * 'ia' may be NULL if there is no route for this destination. 1369 * In case of IPsec, Don't discard it just yet, but pass it to 1370 * ip_output in case of outgoing IPsec policy. 1371 */ 1372 if (!srcrt && ia == NULL) { 1373 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1374 return; 1375 } 1376#endif 1377 1378 /* 1379 * Save the IP header and at most 8 bytes of the payload, 1380 * in case we need to generate an ICMP message to the src. 1381 * 1382 * XXX this can be optimized a lot by saving the data in a local 1383 * buffer on the stack (72 bytes at most), and only allocating the 1384 * mbuf if really necessary. The vast majority of the packets 1385 * are forwarded without having to send an ICMP back (either 1386 * because unnecessary, or because rate limited), so we are 1387 * really we are wasting a lot of work here. 1388 * 1389 * We don't use m_copy() because it might return a reference 1390 * to a shared cluster. Both this function and ip_output() 1391 * assume exclusive access to the IP header in `m', so any 1392 * data in a cluster may change before we reach icmp_error(). 1393 */ 1394 mcopy = m_gethdr(M_NOWAIT, m->m_type); 1395 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) { 1396 /* 1397 * It's probably ok if the pkthdr dup fails (because 1398 * the deep copy of the tag chain failed), but for now 1399 * be conservative and just discard the copy since 1400 * code below may some day want the tags. 1401 */ 1402 m_free(mcopy); 1403 mcopy = NULL; 1404 } 1405 if (mcopy != NULL) { 1406 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy)); 1407 mcopy->m_pkthdr.len = mcopy->m_len; 1408 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1409 } 1410 1411#ifdef IPSTEALTH 1412 if (!V_ipstealth) { 1413#endif 1414 ip->ip_ttl -= IPTTLDEC; 1415#ifdef IPSTEALTH 1416 } 1417#endif 1418 1419 /* 1420 * If forwarding packet using same interface that it came in on, 1421 * perhaps should send a redirect to sender to shortcut a hop. 1422 * Only send redirect if source is sending directly to us, 1423 * and if packet was not source routed (or has any options). 1424 * Also, don't send redirect if forwarding using a default route 1425 * or a route modified by a redirect. 1426 */ 1427 dest.s_addr = 0; 1428 if (!srcrt && V_ipsendredirects && 1429 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) { 1430 struct sockaddr_in *sin; 1431 struct rtentry *rt; 1432 1433 bzero(&ro, sizeof(ro)); 1434 sin = (struct sockaddr_in *)&ro.ro_dst; 1435 sin->sin_family = AF_INET; 1436 sin->sin_len = sizeof(*sin); 1437 sin->sin_addr = ip->ip_dst; 1438 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 1439 1440 rt = ro.ro_rt; 1441 1442 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1443 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1444#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1445 u_long src = ntohl(ip->ip_src.s_addr); 1446 1447 if (RTA(rt) && 1448 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1449 if (rt->rt_flags & RTF_GATEWAY) 1450 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1451 else 1452 dest.s_addr = ip->ip_dst.s_addr; 1453 /* Router requirements says to only send host redirects */ 1454 type = ICMP_REDIRECT; 1455 code = ICMP_REDIRECT_HOST; 1456 } 1457 } 1458 if (rt) 1459 RTFREE(rt); 1460 } 1461 1462 /* 1463 * Try to cache the route MTU from ip_output so we can consider it for 1464 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. 1465 */ 1466 bzero(&ro, sizeof(ro)); 1467 1468 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1469 1470 if (error == EMSGSIZE && ro.ro_rt) 1471 mtu = ro.ro_rt->rt_rmx.rmx_mtu; 1472 RO_RTFREE(&ro); 1473 1474 if (error) 1475 IPSTAT_INC(ips_cantforward); 1476 else { 1477 IPSTAT_INC(ips_forward); 1478 if (type) 1479 IPSTAT_INC(ips_redirectsent); 1480 else { 1481 if (mcopy) 1482 m_freem(mcopy); 1483 if (ia != NULL) 1484 ifa_free(&ia->ia_ifa); 1485 return; 1486 } 1487 } 1488 if (mcopy == NULL) { 1489 if (ia != NULL) 1490 ifa_free(&ia->ia_ifa); 1491 return; 1492 } 1493 1494 switch (error) { 1495 1496 case 0: /* forwarded, but need redirect */ 1497 /* type, code set above */ 1498 break; 1499 1500 case ENETUNREACH: 1501 case EHOSTUNREACH: 1502 case ENETDOWN: 1503 case EHOSTDOWN: 1504 default: 1505 type = ICMP_UNREACH; 1506 code = ICMP_UNREACH_HOST; 1507 break; 1508 1509 case EMSGSIZE: 1510 type = ICMP_UNREACH; 1511 code = ICMP_UNREACH_NEEDFRAG; 1512 1513#ifdef IPSEC 1514 /* 1515 * If IPsec is configured for this path, 1516 * override any possibly mtu value set by ip_output. 1517 */ 1518 mtu = ip_ipsec_mtu(mcopy, mtu); 1519#endif /* IPSEC */ 1520 /* 1521 * If the MTU was set before make sure we are below the 1522 * interface MTU. 1523 * If the MTU wasn't set before use the interface mtu or 1524 * fall back to the next smaller mtu step compared to the 1525 * current packet size. 1526 */ 1527 if (mtu != 0) { 1528 if (ia != NULL) 1529 mtu = min(mtu, ia->ia_ifp->if_mtu); 1530 } else { 1531 if (ia != NULL) 1532 mtu = ia->ia_ifp->if_mtu; 1533 else 1534 mtu = ip_next_mtu(ntohs(ip->ip_len), 0); 1535 } 1536 IPSTAT_INC(ips_cantfrag); 1537 break; 1538 1539 case ENOBUFS: 1540 /* 1541 * A router should not generate ICMP_SOURCEQUENCH as 1542 * required in RFC1812 Requirements for IP Version 4 Routers. 1543 * Source quench could be a big problem under DoS attacks, 1544 * or if the underlying interface is rate-limited. 1545 * Those who need source quench packets may re-enable them 1546 * via the net.inet.ip.sendsourcequench sysctl. 1547 */ 1548 if (V_ip_sendsourcequench == 0) { 1549 m_freem(mcopy); 1550 if (ia != NULL) 1551 ifa_free(&ia->ia_ifa); 1552 return; 1553 } else { 1554 type = ICMP_SOURCEQUENCH; 1555 code = 0; 1556 } 1557 break; 1558 1559 case EACCES: /* ipfw denied packet */ 1560 m_freem(mcopy); 1561 if (ia != NULL) 1562 ifa_free(&ia->ia_ifa); 1563 return; 1564 } 1565 if (ia != NULL) 1566 ifa_free(&ia->ia_ifa); 1567 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1568} 1569 1570void 1571ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1572 struct mbuf *m) 1573{ 1574 1575 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1576 struct bintime bt; 1577 1578 bintime(&bt); 1579 if (inp->inp_socket->so_options & SO_BINTIME) { 1580 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt), 1581 SCM_BINTIME, SOL_SOCKET); 1582 if (*mp) 1583 mp = &(*mp)->m_next; 1584 } 1585 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1586 struct timeval tv; 1587 1588 bintime2timeval(&bt, &tv); 1589 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), 1590 SCM_TIMESTAMP, SOL_SOCKET); 1591 if (*mp) 1592 mp = &(*mp)->m_next; 1593 } 1594 } 1595 if (inp->inp_flags & INP_RECVDSTADDR) { 1596 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst, 1597 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1598 if (*mp) 1599 mp = &(*mp)->m_next; 1600 } 1601 if (inp->inp_flags & INP_RECVTTL) { 1602 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl, 1603 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1604 if (*mp) 1605 mp = &(*mp)->m_next; 1606 } 1607#ifdef notyet 1608 /* XXX 1609 * Moving these out of udp_input() made them even more broken 1610 * than they already were. 1611 */ 1612 /* options were tossed already */ 1613 if (inp->inp_flags & INP_RECVOPTS) { 1614 *mp = sbcreatecontrol((caddr_t)opts_deleted_above, 1615 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1616 if (*mp) 1617 mp = &(*mp)->m_next; 1618 } 1619 /* ip_srcroute doesn't do what we want here, need to fix */ 1620 if (inp->inp_flags & INP_RECVRETOPTS) { 1621 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m), 1622 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1623 if (*mp) 1624 mp = &(*mp)->m_next; 1625 } 1626#endif 1627 if (inp->inp_flags & INP_RECVIF) { 1628 struct ifnet *ifp; 1629 struct sdlbuf { 1630 struct sockaddr_dl sdl; 1631 u_char pad[32]; 1632 } sdlbuf; 1633 struct sockaddr_dl *sdp; 1634 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1635 1636 if ((ifp = m->m_pkthdr.rcvif) && 1637 ifp->if_index && ifp->if_index <= V_if_index) { 1638 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1639 /* 1640 * Change our mind and don't try copy. 1641 */ 1642 if (sdp->sdl_family != AF_LINK || 1643 sdp->sdl_len > sizeof(sdlbuf)) { 1644 goto makedummy; 1645 } 1646 bcopy(sdp, sdl2, sdp->sdl_len); 1647 } else { 1648makedummy: 1649 sdl2->sdl_len = 1650 offsetof(struct sockaddr_dl, sdl_data[0]); 1651 sdl2->sdl_family = AF_LINK; 1652 sdl2->sdl_index = 0; 1653 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1654 } 1655 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len, 1656 IP_RECVIF, IPPROTO_IP); 1657 if (*mp) 1658 mp = &(*mp)->m_next; 1659 } 1660 if (inp->inp_flags & INP_RECVTOS) { 1661 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos, 1662 sizeof(u_char), IP_RECVTOS, IPPROTO_IP); 1663 if (*mp) 1664 mp = &(*mp)->m_next; 1665 } 1666} 1667 1668/* 1669 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1670 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1671 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1672 * compiled. 1673 */ 1674static VNET_DEFINE(int, ip_rsvp_on); 1675VNET_DEFINE(struct socket *, ip_rsvpd); 1676 1677#define V_ip_rsvp_on VNET(ip_rsvp_on) 1678 1679int 1680ip_rsvp_init(struct socket *so) 1681{ 1682 1683 if (so->so_type != SOCK_RAW || 1684 so->so_proto->pr_protocol != IPPROTO_RSVP) 1685 return EOPNOTSUPP; 1686 1687 if (V_ip_rsvpd != NULL) 1688 return EADDRINUSE; 1689 1690 V_ip_rsvpd = so; 1691 /* 1692 * This may seem silly, but we need to be sure we don't over-increment 1693 * the RSVP counter, in case something slips up. 1694 */ 1695 if (!V_ip_rsvp_on) { 1696 V_ip_rsvp_on = 1; 1697 V_rsvp_on++; 1698 } 1699 1700 return 0; 1701} 1702 1703int 1704ip_rsvp_done(void) 1705{ 1706 1707 V_ip_rsvpd = NULL; 1708 /* 1709 * This may seem silly, but we need to be sure we don't over-decrement 1710 * the RSVP counter, in case something slips up. 1711 */ 1712 if (V_ip_rsvp_on) { 1713 V_ip_rsvp_on = 0; 1714 V_rsvp_on--; 1715 } 1716 return 0; 1717} 1718 1719void 1720rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1721{ 1722 1723 if (rsvp_input_p) { /* call the real one if loaded */ 1724 rsvp_input_p(m, off); 1725 return; 1726 } 1727 1728 /* Can still get packets with rsvp_on = 0 if there is a local member 1729 * of the group to which the RSVP packet is addressed. But in this 1730 * case we want to throw the packet away. 1731 */ 1732 1733 if (!V_rsvp_on) { 1734 m_freem(m); 1735 return; 1736 } 1737 1738 if (V_ip_rsvpd != NULL) { 1739 rip_input(m, off); 1740 return; 1741 } 1742 /* Drop the packet */ 1743 m_freem(m); 1744} 1745