Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]
FreeBSD/Linux Kernel Cross Reference
sys/netinet/ip_input.c
Version:
- FREEBSD - FREEBSD-13-STABLE - FREEBSD-13-0 - FREEBSD-12-STABLE - FREEBSD-12-0 - FREEBSD-11-STABLE - FREEBSD-11-0 - FREEBSD-10-STABLE - FREEBSD-10-0 - FREEBSD-9-STABLE - FREEBSD-9-0 - FREEBSD-8-STABLE - FREEBSD-8-0 - FREEBSD-7-STABLE - FREEBSD-7-0 - FREEBSD-6-STABLE - FREEBSD-6-0 - FREEBSD-5-STABLE - FREEBSD-5-0 - FREEBSD-4-STABLE - FREEBSD-3-STABLE - FREEBSD22 - l41 - OPENBSD - linux-2.6 - MK84 - PLAN9 - xnu-8792
SearchContext: - none - 3 - 10
SearchContext: - none - 3 - 10
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: releng/7.3/sys/netinet/ip_input.c 188589 2009-02-13 18:31:35Z luigi $"); 34 35 #include "opt_bootp.h" 36 #include "opt_ipfw.h" 37 #include "opt_ipstealth.h" 38 #include "opt_ipsec.h" 39 #include "opt_mac.h" 40 #include "opt_carp.h" 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/callout.h> 45 #include <sys/mbuf.h> 46 #include <sys/malloc.h> 47 #include <sys/domain.h> 48 #include <sys/protosw.h> 49 #include <sys/socket.h> 50 #include <sys/time.h> 51 #include <sys/kernel.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 63 #include <netinet/in.h> 64 #include <netinet/in_var.h> 65 #include <netinet/ip.h> 66 #include <netinet/in_pcb.h> 67 #include <netinet/ip_var.h> 68 #include <netinet/ip_icmp.h> 69 #include <netinet/ip_options.h> 70 #include <machine/in_cksum.h> 71 #ifdef DEV_CARP 72 #include <netinet/ip_carp.h> 73 #endif 74 #ifdef IPSEC 75 #include <netinet/ip_ipsec.h> 76 #endif /* IPSEC */ 77 78 #include <sys/socketvar.h> 79 80 /* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */ 81 #include <netinet/ip_fw.h> 82 #include <netinet/ip_dummynet.h> 83 84 #include <security/mac/mac_framework.h> 85 86 int rsvp_on = 0; 87 88 int ipforwarding = 0; 89 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 90 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 91 92 static int ipsendredirects = 1; /* XXX */ 93 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 94 &ipsendredirects, 0, "Enable sending IP redirects"); 95 96 int ip_defttl = IPDEFTTL; 97 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 98 &ip_defttl, 0, "Maximum TTL on IP packets"); 99 100 static int ip_keepfaith = 0; 101 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 102 &ip_keepfaith, 0, 103 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 104 105 static int ip_sendsourcequench = 0; 106 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 107 &ip_sendsourcequench, 0, 108 "Enable the transmission of source quench packets"); 109 110 int ip_do_randomid = 0; 111 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 112 &ip_do_randomid, 0, 113 "Assign random ip_id values"); 114 115 /* 116 * XXX - Setting ip_checkinterface mostly implements the receive side of 117 * the Strong ES model described in RFC 1122, but since the routing table 118 * and transmit implementation do not implement the Strong ES model, 119 * setting this to 1 results in an odd hybrid. 120 * 121 * XXX - ip_checkinterface currently must be disabled if you use ipnat 122 * to translate the destination address to another local interface. 123 * 124 * XXX - ip_checkinterface must be disabled if you add IP aliases 125 * to the loopback interface instead of the interface where the 126 * packets for those addresses are received. 127 */ 128 static int ip_checkinterface = 0; 129 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 130 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 131 132 struct pfil_head inet_pfil_hook; /* Packet filter hooks */ 133 134 static struct ifqueue ipintrq; 135 static int ipqmaxlen = IFQ_MAXLEN; 136 137 extern struct domain inetdomain; 138 extern struct protosw inetsw[]; 139 u_char ip_protox[IPPROTO_MAX]; 140 struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 141 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 142 u_long in_ifaddrhmask; /* mask for hash table */ 143 144 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 145 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 146 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 147 &ipintrq.ifq_drops, 0, 148 "Number of packets dropped from the IP input queue"); 149 150 struct ipstat ipstat; 151 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 152 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 153 154 /* 155 * IP datagram reassembly. 156 */ 157 #define IPREASS_NHASH_LOG2 6 158 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 159 #define IPREASS_HMASK (IPREASS_NHASH - 1) 160 #define IPREASS_HASH(x,y) \ 161 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 162 163 static uma_zone_t ipq_zone; 164 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 165 static struct mtx ipqlock; 166 167 #define IPQ_LOCK() mtx_lock(&ipqlock) 168 #define IPQ_UNLOCK() mtx_unlock(&ipqlock) 169 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 170 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 171 172 static void maxnipq_update(void); 173 static void ipq_zone_change(void *); 174 175 static int maxnipq; /* Administrative limit on # reass queues. */ 176 static int nipq = 0; /* Total # of reass queues */ 177 SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, 178 &nipq, 0, "Current number of IPv4 fragment reassembly queue entries"); 179 180 static int maxfragsperpacket; 181 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 182 &maxfragsperpacket, 0, 183 "Maximum number of IPv4 fragments allowed per packet"); 184 185 struct callout ipport_tick_callout; 186 187 #ifdef IPCTL_DEFMTU 188 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 189 &ip_mtu, 0, "Default MTU"); 190 #endif 191 192 #ifdef IPSTEALTH 193 int ipstealth = 0; 194 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 195 &ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding"); 196 #endif 197 198 /* 199 * ipfw_ether and ipfw_bridge hooks. 200 * XXX: Temporary until those are converted to pfil_hooks as well. 201 */ 202 ip_fw_chk_t *ip_fw_chk_ptr = NULL; 203 ip_dn_io_t *ip_dn_io_ptr = NULL; 204 int fw_one_pass = 1; 205 206 static void ip_freef(struct ipqhead *, struct ipq *); 207 208 /* 209 * IP initialization: fill in IP protocol switch table. 210 * All protocols not implemented in kernel go to raw IP protocol handler. 211 */ 212 void 213 ip_init(void) 214 { 215 struct protosw *pr; 216 int i; 217 218 TAILQ_INIT(&in_ifaddrhead); 219 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 220 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 221 if (pr == NULL) 222 panic("ip_init: PF_INET not found"); 223 224 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 225 for (i = 0; i < IPPROTO_MAX; i++) 226 ip_protox[i] = pr - inetsw; 227 /* 228 * Cycle through IP protocols and put them into the appropriate place 229 * in ip_protox[]. 230 */ 231 for (pr = inetdomain.dom_protosw; 232 pr < inetdomain.dom_protoswNPROTOSW; pr++) 233 if (pr->pr_domain->dom_family == PF_INET && 234 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 235 /* Be careful to only index valid IP protocols. */ 236 if (pr->pr_protocol < IPPROTO_MAX) 237 ip_protox[pr->pr_protocol] = pr - inetsw; 238 } 239 240 /* Initialize packet filter hooks. */ 241 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 242 inet_pfil_hook.ph_af = AF_INET; 243 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 244 printf("%s: WARNING: unable to register pfil hook, " 245 "error %d\n", __func__, i); 246 247 /* Initialize IP reassembly queue. */ 248 IPQ_LOCK_INIT(); 249 for (i = 0; i < IPREASS_NHASH; i++) 250 TAILQ_INIT(&ipq[i]); 251 maxnipq = nmbclusters / 32; 252 maxfragsperpacket = 16; 253 ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 254 NULL, UMA_ALIGN_PTR, 0); 255 maxnipq_update(); 256 257 /* Start ipport_tick. */ 258 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE); 259 ipport_tick(NULL); 260 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, 261 SHUTDOWN_PRI_DEFAULT); 262 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, 263 NULL, EVENTHANDLER_PRI_ANY); 264 265 /* Initialize various other remaining things. */ 266 ip_id = time_second & 0xffff; 267 ipintrq.ifq_maxlen = ipqmaxlen; 268 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF); 269 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE); 270 } 271 272 void 273 ip_fini(void *xtp) 274 { 275 276 callout_stop(&ipport_tick_callout); 277 } 278 279 /* 280 * Ip input routine. Checksum and byte swap header. If fragmented 281 * try to reassemble. Process options. Pass to next level. 282 */ 283 void 284 ip_input(struct mbuf *m) 285 { 286 struct ip *ip = NULL; 287 struct in_ifaddr *ia = NULL; 288 struct ifaddr *ifa; 289 int checkif, hlen = 0; 290 u_short sum; 291 int dchg = 0; /* dest changed after fw */ 292 struct in_addr odst; /* original dst address */ 293 294 M_ASSERTPKTHDR(m); 295 296 if (m->m_flags & M_FASTFWD_OURS) { 297 /* 298 * Firewall or NAT changed destination to local. 299 * We expect ip_len and ip_off to be in host byte order. 300 */ 301 m->m_flags &= ~M_FASTFWD_OURS; 302 /* Set up some basics that will be used later. */ 303 ip = mtod(m, struct ip *); 304 hlen = ip->ip_hl << 2; 305 goto ours; 306 } 307 308 ipstat.ips_total++; 309 310 if (m->m_pkthdr.len < sizeof(struct ip)) 311 goto tooshort; 312 313 if (m->m_len < sizeof (struct ip) && 314 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 315 ipstat.ips_toosmall++; 316 return; 317 } 318 ip = mtod(m, struct ip *); 319 320 if (ip->ip_v != IPVERSION) { 321 ipstat.ips_badvers++; 322 goto bad; 323 } 324 325 hlen = ip->ip_hl << 2; 326 if (hlen < sizeof(struct ip)) { /* minimum header length */ 327 ipstat.ips_badhlen++; 328 goto bad; 329 } 330 if (hlen > m->m_len) { 331 if ((m = m_pullup(m, hlen)) == NULL) { 332 ipstat.ips_badhlen++; 333 return; 334 } 335 ip = mtod(m, struct ip *); 336 } 337 338 /* 127/8 must not appear on wire - RFC1122 */ 339 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 340 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 341 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 342 ipstat.ips_badaddr++; 343 goto bad; 344 } 345 } 346 347 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 348 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 349 } else { 350 if (hlen == sizeof(struct ip)) { 351 sum = in_cksum_hdr(ip); 352 } else { 353 sum = in_cksum(m, hlen); 354 } 355 } 356 if (sum) { 357 ipstat.ips_badsum++; 358 goto bad; 359 } 360 361 #ifdef ALTQ 362 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 363 /* packet is dropped by traffic conditioner */ 364 return; 365 #endif 366 367 /* 368 * Convert fields to host representation. 369 */ 370 ip->ip_len = ntohs(ip->ip_len); 371 if (ip->ip_len < hlen) { 372 ipstat.ips_badlen++; 373 goto bad; 374 } 375 ip->ip_off = ntohs(ip->ip_off); 376 377 /* 378 * Check that the amount of data in the buffers 379 * is as at least much as the IP header would have us expect. 380 * Trim mbufs if longer than we expect. 381 * Drop packet if shorter than we expect. 382 */ 383 if (m->m_pkthdr.len < ip->ip_len) { 384 tooshort: 385 ipstat.ips_tooshort++; 386 goto bad; 387 } 388 if (m->m_pkthdr.len > ip->ip_len) { 389 if (m->m_len == m->m_pkthdr.len) { 390 m->m_len = ip->ip_len; 391 m->m_pkthdr.len = ip->ip_len; 392 } else 393 m_adj(m, ip->ip_len - m->m_pkthdr.len); 394 } 395 #ifdef IPSEC 396 /* 397 * Bypass packet filtering for packets from a tunnel (gif). 398 */ 399 if (ip_ipsec_filtertunnel(m)) 400 goto passin; 401 #endif /* IPSEC */ 402 403 /* 404 * Run through list of hooks for input packets. 405 * 406 * NB: Beware of the destination address changing (e.g. 407 * by NAT rewriting). When this happens, tell 408 * ip_forward to do the right thing. 409 */ 410 411 /* Jump over all PFIL processing if hooks are not active. */ 412 if (!PFIL_HOOKED(&inet_pfil_hook)) 413 goto passin; 414 415 odst = ip->ip_dst; 416 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 417 PFIL_IN, NULL) != 0) 418 return; 419 if (m == NULL) /* consumed by filter */ 420 return; 421 422 ip = mtod(m, struct ip *); 423 dchg = (odst.s_addr != ip->ip_dst.s_addr); 424 425 #ifdef IPFIREWALL_FORWARD 426 if (m->m_flags & M_FASTFWD_OURS) { 427 m->m_flags &= ~M_FASTFWD_OURS; 428 goto ours; 429 } 430 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 431 /* 432 * Directly ship on the packet. This allows to forward packets 433 * that were destined for us to some other directly connected 434 * host. 435 */ 436 ip_forward(m, dchg); 437 return; 438 } 439 #endif /* IPFIREWALL_FORWARD */ 440 441 passin: 442 /* 443 * Process options and, if not destined for us, 444 * ship it on. ip_dooptions returns 1 when an 445 * error was detected (causing an icmp message 446 * to be sent and the original packet to be freed). 447 */ 448 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 449 return; 450 451 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 452 * matter if it is destined to another node, or whether it is 453 * a multicast one, RSVP wants it! and prevents it from being forwarded 454 * anywhere else. Also checks if the rsvp daemon is running before 455 * grabbing the packet. 456 */ 457 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 458 goto ours; 459 460 /* 461 * Check our list of addresses, to see if the packet is for us. 462 * If we don't have any addresses, assume any unicast packet 463 * we receive might be for us (and let the upper layers deal 464 * with it). 465 */ 466 if (TAILQ_EMPTY(&in_ifaddrhead) && 467 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 468 goto ours; 469 470 /* 471 * Enable a consistency check between the destination address 472 * and the arrival interface for a unicast packet (the RFC 1122 473 * strong ES model) if IP forwarding is disabled and the packet 474 * is not locally generated and the packet is not subject to 475 * 'ipfw fwd'. 476 * 477 * XXX - Checking also should be disabled if the destination 478 * address is ipnat'ed to a different interface. 479 * 480 * XXX - Checking is incompatible with IP aliases added 481 * to the loopback interface instead of the interface where 482 * the packets are received. 483 * 484 * XXX - This is the case for carp vhost IPs as well so we 485 * insert a workaround. If the packet got here, we already 486 * checked with carp_iamatch() and carp_forus(). 487 */ 488 checkif = ip_checkinterface && (ipforwarding == 0) && 489 m->m_pkthdr.rcvif != NULL && 490 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 491 #ifdef DEV_CARP 492 !m->m_pkthdr.rcvif->if_carp && 493 #endif 494 (dchg == 0); 495 496 /* 497 * Check for exact addresses in the hash bucket. 498 */ 499 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 500 /* 501 * If the address matches, verify that the packet 502 * arrived via the correct interface if checking is 503 * enabled. 504 */ 505 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 506 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 507 goto ours; 508 } 509 /* 510 * Check for broadcast addresses. 511 * 512 * Only accept broadcast packets that arrive via the matching 513 * interface. Reception of forwarded directed broadcasts would 514 * be handled via ip_forward() and ether_output() with the loopback 515 * into the stack for SIMPLEX interfaces handled by ether_output(). 516 */ 517 if (m->m_pkthdr.rcvif != NULL && 518 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 519 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 520 if (ifa->ifa_addr->sa_family != AF_INET) 521 continue; 522 ia = ifatoia(ifa); 523 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 524 ip->ip_dst.s_addr) 525 goto ours; 526 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) 527 goto ours; 528 #ifdef BOOTP_COMPAT 529 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 530 goto ours; 531 #endif 532 } 533 } 534 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 535 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 536 ipstat.ips_cantforward++; 537 m_freem(m); 538 return; 539 } 540 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 541 struct in_multi *inm; 542 if (ip_mrouter) { 543 /* 544 * If we are acting as a multicast router, all 545 * incoming multicast packets are passed to the 546 * kernel-level multicast forwarding function. 547 * The packet is returned (relatively) intact; if 548 * ip_mforward() returns a non-zero value, the packet 549 * must be discarded, else it may be accepted below. 550 */ 551 if (ip_mforward && 552 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 553 ipstat.ips_cantforward++; 554 m_freem(m); 555 return; 556 } 557 558 /* 559 * The process-level routing daemon needs to receive 560 * all multicast IGMP packets, whether or not this 561 * host belongs to their destination groups. 562 */ 563 if (ip->ip_p == IPPROTO_IGMP) 564 goto ours; 565 ipstat.ips_forward++; 566 } 567 /* 568 * See if we belong to the destination multicast group on the 569 * arrival interface. 570 */ 571 IN_MULTI_LOCK(); 572 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 573 IN_MULTI_UNLOCK(); 574 if (inm == NULL) { 575 ipstat.ips_notmember++; 576 m_freem(m); 577 return; 578 } 579 goto ours; 580 } 581 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 582 goto ours; 583 if (ip->ip_dst.s_addr == INADDR_ANY) 584 goto ours; 585 586 /* 587 * FAITH(Firewall Aided Internet Translator) 588 */ 589 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 590 if (ip_keepfaith) { 591 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 592 goto ours; 593 } 594 m_freem(m); 595 return; 596 } 597 598 /* 599 * Not for us; forward if possible and desirable. 600 */ 601 if (ipforwarding == 0) { 602 ipstat.ips_cantforward++; 603 m_freem(m); 604 } else { 605 #ifdef IPSEC 606 if (ip_ipsec_fwd(m)) 607 goto bad; 608 #endif /* IPSEC */ 609 ip_forward(m, dchg); 610 } 611 return; 612 613 ours: 614 #ifdef IPSTEALTH 615 /* 616 * IPSTEALTH: Process non-routing options only 617 * if the packet is destined for us. 618 */ 619 if (ipstealth && hlen > sizeof (struct ip) && 620 ip_dooptions(m, 1)) 621 return; 622 #endif /* IPSTEALTH */ 623 624 /* Count the packet in the ip address stats */ 625 if (ia != NULL) { 626 ia->ia_ifa.if_ipackets++; 627 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 628 } 629 630 /* 631 * Attempt reassembly; if it succeeds, proceed. 632 * ip_reass() will return a different mbuf. 633 */ 634 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 635 m = ip_reass(m); 636 if (m == NULL) 637 return; 638 ip = mtod(m, struct ip *); 639 /* Get the header length of the reassembled packet */ 640 hlen = ip->ip_hl << 2; 641 } 642 643 /* 644 * Further protocols expect the packet length to be w/o the 645 * IP header. 646 */ 647 ip->ip_len -= hlen; 648 649 #ifdef IPSEC 650 /* 651 * enforce IPsec policy checking if we are seeing last header. 652 * note that we do not visit this with protocols with pcb layer 653 * code - like udp/tcp/raw ip. 654 */ 655 if (ip_ipsec_input(m)) 656 goto bad; 657 #endif /* IPSEC */ 658 659 /* 660 * Switch out to protocol's input routine. 661 */ 662 ipstat.ips_delivered++; 663 664 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 665 return; 666 bad: 667 m_freem(m); 668 } 669 670 /* 671 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 672 * max has slightly different semantics than the sysctl, for historical 673 * reasons. 674 */ 675 static void 676 maxnipq_update(void) 677 { 678 679 /* 680 * -1 for unlimited allocation. 681 */ 682 if (maxnipq < 0) 683 uma_zone_set_max(ipq_zone, 0); 684 /* 685 * Positive number for specific bound. 686 */ 687 if (maxnipq > 0) 688 uma_zone_set_max(ipq_zone, maxnipq); 689 /* 690 * Zero specifies no further fragment queue allocation -- set the 691 * bound very low, but rely on implementation elsewhere to actually 692 * prevent allocation and reclaim current queues. 693 */ 694 if (maxnipq == 0) 695 uma_zone_set_max(ipq_zone, 1); 696 } 697 698 static void 699 ipq_zone_change(void *tag) 700 { 701 702 if (maxnipq > 0 && maxnipq < (nmbclusters / 32)) { 703 maxnipq = nmbclusters / 32; 704 maxnipq_update(); 705 } 706 } 707 708 static int 709 sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 710 { 711 int error, i; 712 713 i = maxnipq; 714 error = sysctl_handle_int(oidp, &i, 0, req); 715 if (error || !req->newptr) 716 return (error); 717 718 /* 719 * XXXRW: Might be a good idea to sanity check the argument and place 720 * an extreme upper bound. 721 */ 722 if (i < -1) 723 return (EINVAL); 724 maxnipq = i; 725 maxnipq_update(); 726 return (0); 727 } 728 729 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 730 NULL, 0, sysctl_maxnipq, "I", 731 "Maximum number of IPv4 fragment reassembly queue entries"); 732 733 /* 734 * Take incoming datagram fragment and try to reassemble it into 735 * whole datagram. If the argument is the first fragment or one 736 * in between the function will return NULL and store the mbuf 737 * in the fragment chain. If the argument is the last fragment 738 * the packet will be reassembled and the pointer to the new 739 * mbuf returned for further processing. Only m_tags attached 740 * to the first packet/fragment are preserved. 741 * The IP header is *NOT* adjusted out of iplen. 742 */ 743 struct mbuf * 744 ip_reass(struct mbuf *m) 745 { 746 struct ip *ip; 747 struct mbuf *p, *q, *nq, *t; 748 struct ipq *fp = NULL; 749 struct ipqhead *head; 750 int i, hlen, next; 751 u_int8_t ecn, ecn0; 752 u_short hash; 753 754 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 755 if (maxnipq == 0 || maxfragsperpacket == 0) { 756 ipstat.ips_fragments++; 757 ipstat.ips_fragdropped++; 758 m_freem(m); 759 return (NULL); 760 } 761 762 ip = mtod(m, struct ip *); 763 hlen = ip->ip_hl << 2; 764 765 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 766 head = &ipq[hash]; 767 IPQ_LOCK(); 768 769 /* 770 * Look for queue of fragments 771 * of this datagram. 772 */ 773 TAILQ_FOREACH(fp, head, ipq_list) 774 if (ip->ip_id == fp->ipq_id && 775 ip->ip_src.s_addr == fp->ipq_src.s_addr && 776 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 777 #ifdef MAC 778 mac_fragment_match(m, fp) && 779 #endif 780 ip->ip_p == fp->ipq_p) 781 goto found; 782 783 fp = NULL; 784 785 /* 786 * Attempt to trim the number of allocated fragment queues if it 787 * exceeds the administrative limit. 788 */ 789 if ((nipq > maxnipq) && (maxnipq > 0)) { 790 /* 791 * drop something from the tail of the current queue 792 * before proceeding further 793 */ 794 struct ipq *q = TAILQ_LAST(head, ipqhead); 795 if (q == NULL) { /* gak */ 796 for (i = 0; i < IPREASS_NHASH; i++) { 797 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 798 if (r) { 799 ipstat.ips_fragtimeout += r->ipq_nfrags; 800 ip_freef(&ipq[i], r); 801 break; 802 } 803 } 804 } else { 805 ipstat.ips_fragtimeout += q->ipq_nfrags; 806 ip_freef(head, q); 807 } 808 } 809 810 found: 811 /* 812 * Adjust ip_len to not reflect header, 813 * convert offset of this to bytes. 814 */ 815 ip->ip_len -= hlen; 816 if (ip->ip_off & IP_MF) { 817 /* 818 * Make sure that fragments have a data length 819 * that's a non-zero multiple of 8 bytes. 820 */ 821 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 822 ipstat.ips_toosmall++; /* XXX */ 823 goto dropfrag; 824 } 825 m->m_flags |= M_FRAG; 826 } else 827 m->m_flags &= ~M_FRAG; 828 ip->ip_off <<= 3; 829 830 831 /* 832 * Attempt reassembly; if it succeeds, proceed. 833 * ip_reass() will return a different mbuf. 834 */ 835 ipstat.ips_fragments++; 836 m->m_pkthdr.header = ip; 837 838 /* Previous ip_reass() started here. */ 839 /* 840 * Presence of header sizes in mbufs 841 * would confuse code below. 842 */ 843 m->m_data += hlen; 844 m->m_len -= hlen; 845 846 /* 847 * If first fragment to arrive, create a reassembly queue. 848 */ 849 if (fp == NULL) { 850 fp = uma_zalloc(ipq_zone, M_NOWAIT); 851 if (fp == NULL) 852 goto dropfrag; 853 #ifdef MAC 854 if (mac_init_ipq(fp, M_NOWAIT) != 0) { 855 uma_zfree(ipq_zone, fp); 856 fp = NULL; 857 goto dropfrag; 858 } 859 mac_create_ipq(m, fp); 860 #endif 861 TAILQ_INSERT_HEAD(head, fp, ipq_list); 862 nipq++; 863 fp->ipq_nfrags = 1; 864 fp->ipq_ttl = IPFRAGTTL; 865 fp->ipq_p = ip->ip_p; 866 fp->ipq_id = ip->ip_id; 867 fp->ipq_src = ip->ip_src; 868 fp->ipq_dst = ip->ip_dst; 869 fp->ipq_frags = m; 870 m->m_nextpkt = NULL; 871 goto done; 872 } else { 873 fp->ipq_nfrags++; 874 #ifdef MAC 875 mac_update_ipq(m, fp); 876 #endif 877 } 878 879 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 880 881 /* 882 * Handle ECN by comparing this segment with the first one; 883 * if CE is set, do not lose CE. 884 * drop if CE and not-ECT are mixed for the same packet. 885 */ 886 ecn = ip->ip_tos & IPTOS_ECN_MASK; 887 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 888 if (ecn == IPTOS_ECN_CE) { 889 if (ecn0 == IPTOS_ECN_NOTECT) 890 goto dropfrag; 891 if (ecn0 != IPTOS_ECN_CE) 892 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 893 } 894 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 895 goto dropfrag; 896 897 /* 898 * Find a segment which begins after this one does. 899 */ 900 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 901 if (GETIP(q)->ip_off > ip->ip_off) 902 break; 903 904 /* 905 * If there is a preceding segment, it may provide some of 906 * our data already. If so, drop the data from the incoming 907 * segment. If it provides all of our data, drop us, otherwise 908 * stick new segment in the proper place. 909 * 910 * If some of the data is dropped from the the preceding 911 * segment, then it's checksum is invalidated. 912 */ 913 if (p) { 914 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 915 if (i > 0) { 916 if (i >= ip->ip_len) 917 goto dropfrag; 918 m_adj(m, i); 919 m->m_pkthdr.csum_flags = 0; 920 ip->ip_off += i; 921 ip->ip_len -= i; 922 } 923 m->m_nextpkt = p->m_nextpkt; 924 p->m_nextpkt = m; 925 } else { 926 m->m_nextpkt = fp->ipq_frags; 927 fp->ipq_frags = m; 928 } 929 930 /* 931 * While we overlap succeeding segments trim them or, 932 * if they are completely covered, dequeue them. 933 */ 934 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 935 q = nq) { 936 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 937 if (i < GETIP(q)->ip_len) { 938 GETIP(q)->ip_len -= i; 939 GETIP(q)->ip_off += i; 940 m_adj(q, i); 941 q->m_pkthdr.csum_flags = 0; 942 break; 943 } 944 nq = q->m_nextpkt; 945 m->m_nextpkt = nq; 946 ipstat.ips_fragdropped++; 947 fp->ipq_nfrags--; 948 m_freem(q); 949 } 950 951 /* 952 * Check for complete reassembly and perform frag per packet 953 * limiting. 954 * 955 * Frag limiting is performed here so that the nth frag has 956 * a chance to complete the packet before we drop the packet. 957 * As a result, n+1 frags are actually allowed per packet, but 958 * only n will ever be stored. (n = maxfragsperpacket.) 959 * 960 */ 961 next = 0; 962 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 963 if (GETIP(q)->ip_off != next) { 964 if (fp->ipq_nfrags > maxfragsperpacket) { 965 ipstat.ips_fragdropped += fp->ipq_nfrags; 966 ip_freef(head, fp); 967 } 968 goto done; 969 } 970 next += GETIP(q)->ip_len; 971 } 972 /* Make sure the last packet didn't have the IP_MF flag */ 973 if (p->m_flags & M_FRAG) { 974 if (fp->ipq_nfrags > maxfragsperpacket) { 975 ipstat.ips_fragdropped += fp->ipq_nfrags; 976 ip_freef(head, fp); 977 } 978 goto done; 979 } 980 981 /* 982 * Reassembly is complete. Make sure the packet is a sane size. 983 */ 984 q = fp->ipq_frags; 985 ip = GETIP(q); 986 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 987 ipstat.ips_toolong++; 988 ipstat.ips_fragdropped += fp->ipq_nfrags; 989 ip_freef(head, fp); 990 goto done; 991 } 992 993 /* 994 * Concatenate fragments. 995 */ 996 m = q; 997 t = m->m_next; 998 m->m_next = NULL; 999 m_cat(m, t); 1000 nq = q->m_nextpkt; 1001 q->m_nextpkt = NULL; 1002 for (q = nq; q != NULL; q = nq) { 1003 nq = q->m_nextpkt; 1004 q->m_nextpkt = NULL; 1005 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1006 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1007 m_cat(m, q); 1008 } 1009 /* 1010 * In order to do checksumming faster we do 'end-around carry' here 1011 * (and not in for{} loop), though it implies we are not going to 1012 * reassemble more than 64k fragments. 1013 */ 1014 m->m_pkthdr.csum_data = 1015 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1016 #ifdef MAC 1017 mac_create_datagram_from_ipq(fp, m); 1018 mac_destroy_ipq(fp); 1019 #endif 1020 1021 /* 1022 * Create header for new ip packet by modifying header of first 1023 * packet; dequeue and discard fragment reassembly header. 1024 * Make header visible. 1025 */ 1026 ip->ip_len = (ip->ip_hl << 2) + next; 1027 ip->ip_src = fp->ipq_src; 1028 ip->ip_dst = fp->ipq_dst; 1029 TAILQ_REMOVE(head, fp, ipq_list); 1030 nipq--; 1031 uma_zfree(ipq_zone, fp); 1032 m->m_len += (ip->ip_hl << 2); 1033 m->m_data -= (ip->ip_hl << 2); 1034 /* some debugging cruft by sklower, below, will go away soon */ 1035 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1036 m_fixhdr(m); 1037 ipstat.ips_reassembled++; 1038 IPQ_UNLOCK(); 1039 return (m); 1040 1041 dropfrag: 1042 ipstat.ips_fragdropped++; 1043 if (fp != NULL) 1044 fp->ipq_nfrags--; 1045 m_freem(m); 1046 done: 1047 IPQ_UNLOCK(); 1048 return (NULL); 1049 1050 #undef GETIP 1051 } 1052 1053 /* 1054 * Free a fragment reassembly header and all 1055 * associated datagrams. 1056 */ 1057 static void 1058 ip_freef(struct ipqhead *fhp, struct ipq *fp) 1059 { 1060 struct mbuf *q; 1061 1062 IPQ_LOCK_ASSERT(); 1063 1064 while (fp->ipq_frags) { 1065 q = fp->ipq_frags; 1066 fp->ipq_frags = q->m_nextpkt; 1067 m_freem(q); 1068 } 1069 TAILQ_REMOVE(fhp, fp, ipq_list); 1070 uma_zfree(ipq_zone, fp); 1071 nipq--; 1072 } 1073 1074 /* 1075 * IP timer processing; 1076 * if a timer expires on a reassembly 1077 * queue, discard it. 1078 */ 1079 void 1080 ip_slowtimo(void) 1081 { 1082 struct ipq *fp; 1083 int i; 1084 1085 IPQ_LOCK(); 1086 for (i = 0; i < IPREASS_NHASH; i++) { 1087 for(fp = TAILQ_FIRST(&ipq[i]); fp;) { 1088 struct ipq *fpp; 1089 1090 fpp = fp; 1091 fp = TAILQ_NEXT(fp, ipq_list); 1092 if(--fpp->ipq_ttl == 0) { 1093 ipstat.ips_fragtimeout += fpp->ipq_nfrags; 1094 ip_freef(&ipq[i], fpp); 1095 } 1096 } 1097 } 1098 /* 1099 * If we are over the maximum number of fragments 1100 * (due to the limit being lowered), drain off 1101 * enough to get down to the new limit. 1102 */ 1103 if (maxnipq >= 0 && nipq > maxnipq) { 1104 for (i = 0; i < IPREASS_NHASH; i++) { 1105 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) { 1106 ipstat.ips_fragdropped += 1107 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1108 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1109 } 1110 } 1111 } 1112 IPQ_UNLOCK(); 1113 } 1114 1115 /* 1116 * Drain off all datagram fragments. 1117 */ 1118 void 1119 ip_drain(void) 1120 { 1121 int i; 1122 1123 IPQ_LOCK(); 1124 for (i = 0; i < IPREASS_NHASH; i++) { 1125 while(!TAILQ_EMPTY(&ipq[i])) { 1126 ipstat.ips_fragdropped += 1127 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1128 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1129 } 1130 } 1131 IPQ_UNLOCK(); 1132 in_rtqdrain(); 1133 } 1134 1135 /* 1136 * The protocol to be inserted into ip_protox[] must be already registered 1137 * in inetsw[], either statically or through pf_proto_register(). 1138 */ 1139 int 1140 ipproto_register(u_char ipproto) 1141 { 1142 struct protosw *pr; 1143 1144 /* Sanity checks. */ 1145 if (ipproto == 0) 1146 return (EPROTONOSUPPORT); 1147 1148 /* 1149 * The protocol slot must not be occupied by another protocol 1150 * already. An index pointing to IPPROTO_RAW is unused. 1151 */ 1152 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1153 if (pr == NULL) 1154 return (EPFNOSUPPORT); 1155 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1156 return (EEXIST); 1157 1158 /* Find the protocol position in inetsw[] and set the index. */ 1159 for (pr = inetdomain.dom_protosw; 1160 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1161 if (pr->pr_domain->dom_family == PF_INET && 1162 pr->pr_protocol && pr->pr_protocol == ipproto) { 1163 /* Be careful to only index valid IP protocols. */ 1164 if (pr->pr_protocol < IPPROTO_MAX) { 1165 ip_protox[pr->pr_protocol] = pr - inetsw; 1166 return (0); 1167 } else 1168 return (EINVAL); 1169 } 1170 } 1171 return (EPROTONOSUPPORT); 1172 } 1173 1174 int 1175 ipproto_unregister(u_char ipproto) 1176 { 1177 struct protosw *pr; 1178 1179 /* Sanity checks. */ 1180 if (ipproto == 0) 1181 return (EPROTONOSUPPORT); 1182 1183 /* Check if the protocol was indeed registered. */ 1184 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1185 if (pr == NULL) 1186 return (EPFNOSUPPORT); 1187 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1188 return (ENOENT); 1189 1190 /* Reset the protocol slot to IPPROTO_RAW. */ 1191 ip_protox[ipproto] = pr - inetsw; 1192 return (0); 1193 } 1194 1195 /* 1196 * Given address of next destination (final or next hop), 1197 * return internet address info of interface to be used to get there. 1198 */ 1199 struct in_ifaddr * 1200 ip_rtaddr(struct in_addr dst, u_int fibnum) 1201 { 1202 struct route sro; 1203 struct sockaddr_in *sin; 1204 struct in_ifaddr *ifa; 1205 1206 bzero(&sro, sizeof(sro)); 1207 sin = (struct sockaddr_in *)&sro.ro_dst; 1208 sin->sin_family = AF_INET; 1209 sin->sin_len = sizeof(*sin); 1210 sin->sin_addr = dst; 1211 in_rtalloc_ign(&sro, RTF_CLONING, fibnum); 1212 1213 if (sro.ro_rt == NULL) 1214 return (NULL); 1215 1216 ifa = ifatoia(sro.ro_rt->rt_ifa); 1217 RTFREE(sro.ro_rt); 1218 return (ifa); 1219 } 1220 1221 u_char inetctlerrmap[PRC_NCMDS] = { 1222 0, 0, 0, 0, 1223 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1224 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1225 EMSGSIZE, EHOSTUNREACH, 0, 0, 1226 0, 0, EHOSTUNREACH, 0, 1227 ENOPROTOOPT, ECONNREFUSED 1228 }; 1229 1230 /* 1231 * Forward a packet. If some error occurs return the sender 1232 * an icmp packet. Note we can't always generate a meaningful 1233 * icmp message because icmp doesn't have a large enough repertoire 1234 * of codes and types. 1235 * 1236 * If not forwarding, just drop the packet. This could be confusing 1237 * if ipforwarding was zero but some routing protocol was advancing 1238 * us as a gateway to somewhere. However, we must let the routing 1239 * protocol deal with that. 1240 * 1241 * The srcrt parameter indicates whether the packet is being forwarded 1242 * via a source route. 1243 */ 1244 void 1245 ip_forward(struct mbuf *m, int srcrt) 1246 { 1247 struct ip *ip = mtod(m, struct ip *); 1248 struct in_ifaddr *ia = NULL; 1249 struct mbuf *mcopy; 1250 struct in_addr dest; 1251 struct route ro; 1252 int error, type = 0, code = 0, mtu = 0; 1253 1254 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1255 ipstat.ips_cantforward++; 1256 m_freem(m); 1257 return; 1258 } 1259 #ifdef IPSTEALTH 1260 if (!ipstealth) { 1261 #endif 1262 if (ip->ip_ttl <= IPTTLDEC) { 1263 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1264 0, 0); 1265 return; 1266 } 1267 #ifdef IPSTEALTH 1268 } 1269 #endif 1270 1271 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); 1272 if (!srcrt && ia == NULL) { 1273 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1274 return; 1275 } 1276 1277 /* 1278 * Save the IP header and at most 8 bytes of the payload, 1279 * in case we need to generate an ICMP message to the src. 1280 * 1281 * XXX this can be optimized a lot by saving the data in a local 1282 * buffer on the stack (72 bytes at most), and only allocating the 1283 * mbuf if really necessary. The vast majority of the packets 1284 * are forwarded without having to send an ICMP back (either 1285 * because unnecessary, or because rate limited), so we are 1286 * really we are wasting a lot of work here. 1287 * 1288 * We don't use m_copy() because it might return a reference 1289 * to a shared cluster. Both this function and ip_output() 1290 * assume exclusive access to the IP header in `m', so any 1291 * data in a cluster may change before we reach icmp_error(). 1292 */ 1293 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1294 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1295 /* 1296 * It's probably ok if the pkthdr dup fails (because 1297 * the deep copy of the tag chain failed), but for now 1298 * be conservative and just discard the copy since 1299 * code below may some day want the tags. 1300 */ 1301 m_free(mcopy); 1302 mcopy = NULL; 1303 } 1304 if (mcopy != NULL) { 1305 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy)); 1306 mcopy->m_pkthdr.len = mcopy->m_len; 1307 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1308 } 1309 1310 #ifdef IPSTEALTH 1311 if (!ipstealth) { 1312 #endif 1313 ip->ip_ttl -= IPTTLDEC; 1314 #ifdef IPSTEALTH 1315 } 1316 #endif 1317 1318 /* 1319 * If forwarding packet using same interface that it came in on, 1320 * perhaps should send a redirect to sender to shortcut a hop. 1321 * Only send redirect if source is sending directly to us, 1322 * and if packet was not source routed (or has any options). 1323 * Also, don't send redirect if forwarding using a default route 1324 * or a route modified by a redirect. 1325 */ 1326 dest.s_addr = 0; 1327 if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) { 1328 struct sockaddr_in *sin; 1329 struct rtentry *rt; 1330 1331 bzero(&ro, sizeof(ro)); 1332 sin = (struct sockaddr_in *)&ro.ro_dst; 1333 sin->sin_family = AF_INET; 1334 sin->sin_len = sizeof(*sin); 1335 sin->sin_addr = ip->ip_dst; 1336 in_rtalloc_ign(&ro, RTF_CLONING, M_GETFIB(m)); 1337 1338 rt = ro.ro_rt; 1339 1340 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1341 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1342 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1343 u_long src = ntohl(ip->ip_src.s_addr); 1344 1345 if (RTA(rt) && 1346 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1347 if (rt->rt_flags & RTF_GATEWAY) 1348 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1349 else 1350 dest.s_addr = ip->ip_dst.s_addr; 1351 /* Router requirements says to only send host redirects */ 1352 type = ICMP_REDIRECT; 1353 code = ICMP_REDIRECT_HOST; 1354 } 1355 } 1356 if (rt) 1357 RTFREE(rt); 1358 } 1359 1360 /* 1361 * Try to cache the route MTU from ip_output so we can consider it for 1362 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. 1363 */ 1364 bzero(&ro, sizeof(ro)); 1365 1366 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1367 1368 if (error == EMSGSIZE && ro.ro_rt) 1369 mtu = ro.ro_rt->rt_rmx.rmx_mtu; 1370 if (ro.ro_rt) 1371 RTFREE(ro.ro_rt); 1372 1373 if (error) 1374 ipstat.ips_cantforward++; 1375 else { 1376 ipstat.ips_forward++; 1377 if (type) 1378 ipstat.ips_redirectsent++; 1379 else { 1380 if (mcopy) 1381 m_freem(mcopy); 1382 return; 1383 } 1384 } 1385 if (mcopy == NULL) 1386 return; 1387 1388 switch (error) { 1389 1390 case 0: /* forwarded, but need redirect */ 1391 /* type, code set above */ 1392 break; 1393 1394 case ENETUNREACH: /* shouldn't happen, checked above */ 1395 case EHOSTUNREACH: 1396 case ENETDOWN: 1397 case EHOSTDOWN: 1398 default: 1399 type = ICMP_UNREACH; 1400 code = ICMP_UNREACH_HOST; 1401 break; 1402 1403 case EMSGSIZE: 1404 type = ICMP_UNREACH; 1405 code = ICMP_UNREACH_NEEDFRAG; 1406 1407 #ifdef IPSEC 1408 /* 1409 * If IPsec is configured for this path, 1410 * override any possibly mtu value set by ip_output. 1411 */ 1412 mtu = ip_ipsec_mtu(m, mtu); 1413 #endif /* IPSEC */ 1414 /* 1415 * If the MTU was set before make sure we are below the 1416 * interface MTU. 1417 * If the MTU wasn't set before use the interface mtu or 1418 * fall back to the next smaller mtu step compared to the 1419 * current packet size. 1420 */ 1421 if (mtu != 0) { 1422 if (ia != NULL) 1423 mtu = min(mtu, ia->ia_ifp->if_mtu); 1424 } else { 1425 if (ia != NULL) 1426 mtu = ia->ia_ifp->if_mtu; 1427 else 1428 mtu = ip_next_mtu(ip->ip_len, 0); 1429 } 1430 ipstat.ips_cantfrag++; 1431 break; 1432 1433 case ENOBUFS: 1434 /* 1435 * A router should not generate ICMP_SOURCEQUENCH as 1436 * required in RFC1812 Requirements for IP Version 4 Routers. 1437 * Source quench could be a big problem under DoS attacks, 1438 * or if the underlying interface is rate-limited. 1439 * Those who need source quench packets may re-enable them 1440 * via the net.inet.ip.sendsourcequench sysctl. 1441 */ 1442 if (ip_sendsourcequench == 0) { 1443 m_freem(mcopy); 1444 return; 1445 } else { 1446 type = ICMP_SOURCEQUENCH; 1447 code = 0; 1448 } 1449 break; 1450 1451 case EACCES: /* ipfw denied packet */ 1452 m_freem(mcopy); 1453 return; 1454 } 1455 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1456 } 1457 1458 void 1459 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1460 struct mbuf *m) 1461 { 1462 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1463 struct bintime bt; 1464 1465 bintime(&bt); 1466 if (inp->inp_socket->so_options & SO_BINTIME) { 1467 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1468 SCM_BINTIME, SOL_SOCKET); 1469 if (*mp) 1470 mp = &(*mp)->m_next; 1471 } 1472 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1473 struct timeval tv; 1474 1475 bintime2timeval(&bt, &tv); 1476 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1477 SCM_TIMESTAMP, SOL_SOCKET); 1478 if (*mp) 1479 mp = &(*mp)->m_next; 1480 } 1481 } 1482 if (inp->inp_flags & INP_RECVDSTADDR) { 1483 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1484 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1485 if (*mp) 1486 mp = &(*mp)->m_next; 1487 } 1488 if (inp->inp_flags & INP_RECVTTL) { 1489 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1490 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1491 if (*mp) 1492 mp = &(*mp)->m_next; 1493 } 1494 #ifdef notyet 1495 /* XXX 1496 * Moving these out of udp_input() made them even more broken 1497 * than they already were. 1498 */ 1499 /* options were tossed already */ 1500 if (inp->inp_flags & INP_RECVOPTS) { 1501 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1502 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1503 if (*mp) 1504 mp = &(*mp)->m_next; 1505 } 1506 /* ip_srcroute doesn't do what we want here, need to fix */ 1507 if (inp->inp_flags & INP_RECVRETOPTS) { 1508 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1509 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1510 if (*mp) 1511 mp = &(*mp)->m_next; 1512 } 1513 #endif 1514 if (inp->inp_flags & INP_RECVIF) { 1515 struct ifnet *ifp; 1516 struct sdlbuf { 1517 struct sockaddr_dl sdl; 1518 u_char pad[32]; 1519 } sdlbuf; 1520 struct sockaddr_dl *sdp; 1521 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1522 1523 if (((ifp = m->m_pkthdr.rcvif)) 1524 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1525 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1526 /* 1527 * Change our mind and don't try copy. 1528 */ 1529 if ((sdp->sdl_family != AF_LINK) 1530 || (sdp->sdl_len > sizeof(sdlbuf))) { 1531 goto makedummy; 1532 } 1533 bcopy(sdp, sdl2, sdp->sdl_len); 1534 } else { 1535 makedummy: 1536 sdl2->sdl_len 1537 = offsetof(struct sockaddr_dl, sdl_data[0]); 1538 sdl2->sdl_family = AF_LINK; 1539 sdl2->sdl_index = 0; 1540 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1541 } 1542 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1543 IP_RECVIF, IPPROTO_IP); 1544 if (*mp) 1545 mp = &(*mp)->m_next; 1546 } 1547 } 1548 1549 /* 1550 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1551 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1552 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1553 * compiled. 1554 */ 1555 static int ip_rsvp_on; 1556 struct socket *ip_rsvpd; 1557 int 1558 ip_rsvp_init(struct socket *so) 1559 { 1560 if (so->so_type != SOCK_RAW || 1561 so->so_proto->pr_protocol != IPPROTO_RSVP) 1562 return EOPNOTSUPP; 1563 1564 if (ip_rsvpd != NULL) 1565 return EADDRINUSE; 1566 1567 ip_rsvpd = so; 1568 /* 1569 * This may seem silly, but we need to be sure we don't over-increment 1570 * the RSVP counter, in case something slips up. 1571 */ 1572 if (!ip_rsvp_on) { 1573 ip_rsvp_on = 1; 1574 rsvp_on++; 1575 } 1576 1577 return 0; 1578 } 1579 1580 int 1581 ip_rsvp_done(void) 1582 { 1583 ip_rsvpd = NULL; 1584 /* 1585 * This may seem silly, but we need to be sure we don't over-decrement 1586 * the RSVP counter, in case something slips up. 1587 */ 1588 if (ip_rsvp_on) { 1589 ip_rsvp_on = 0; 1590 rsvp_on--; 1591 } 1592 return 0; 1593 } 1594 1595 void 1596 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1597 { 1598 if (rsvp_input_p) { /* call the real one if loaded */ 1599 rsvp_input_p(m, off); 1600 return; 1601 } 1602 1603 /* Can still get packets with rsvp_on = 0 if there is a local member 1604 * of the group to which the RSVP packet is addressed. But in this 1605 * case we want to throw the packet away. 1606 */ 1607 1608 if (!rsvp_on) { 1609 m_freem(m); 1610 return; 1611 } 1612 1613 if (ip_rsvpd != NULL) { 1614 rip_input(m, off); 1615 return; 1616 } 1617 /* Drop the packet */ 1618 m_freem(m); 1619 }
Cache object: 7291aa12a0d02d4e72d021423848be0c
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]