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 * $FreeBSD$ 31 */ 32 33 #include "opt_bootp.h" 34 #include "opt_ipfw.h" 35 #include "opt_ipstealth.h" 36 #include "opt_ipsec.h" 37 #include "opt_mac.h" 38 #include "opt_carp.h" 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/callout.h> 43 #include <sys/mbuf.h> 44 #include <sys/malloc.h> 45 #include <sys/domain.h> 46 #include <sys/protosw.h> 47 #include <sys/socket.h> 48 #include <sys/time.h> 49 #include <sys/kernel.h> 50 #include <sys/syslog.h> 51 #include <sys/sysctl.h> 52 53 #include <net/pfil.h> 54 #include <net/if.h> 55 #include <net/if_types.h> 56 #include <net/if_var.h> 57 #include <net/if_dl.h> 58 #include <net/route.h> 59 #include <net/netisr.h> 60 61 #include <netinet/in.h> 62 #include <netinet/in_systm.h> 63 #include <netinet/in_var.h> 64 #include <netinet/ip.h> 65 #include <netinet/in_pcb.h> 66 #include <netinet/ip_var.h> 67 #include <netinet/ip_icmp.h> 68 #include <netinet/ip_options.h> 69 #include <machine/in_cksum.h> 70 #ifdef DEV_CARP 71 #include <netinet/ip_carp.h> 72 #endif 73 #if defined(IPSEC) || defined(FAST_IPSEC) 74 #include <netinet/ip_ipsec.h> 75 #endif /* IPSEC */ 76 77 #include <sys/socketvar.h> 78 79 /* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */ 80 #include <netinet/ip_fw.h> 81 #include <netinet/ip_dummynet.h> 82 83 #include <security/mac/mac_framework.h> 84 85 int rsvp_on = 0; 86 87 int ipforwarding = 0; 88 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 89 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 90 91 static int ipsendredirects = 1; /* XXX */ 92 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 93 &ipsendredirects, 0, "Enable sending IP redirects"); 94 95 int ip_defttl = IPDEFTTL; 96 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 97 &ip_defttl, 0, "Maximum TTL on IP packets"); 98 99 static int ip_keepfaith = 0; 100 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 101 &ip_keepfaith, 0, 102 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 103 104 static int ip_sendsourcequench = 0; 105 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 106 &ip_sendsourcequench, 0, 107 "Enable the transmission of source quench packets"); 108 109 int ip_do_randomid = 0; 110 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 111 &ip_do_randomid, 0, 112 "Assign random ip_id values"); 113 114 /* 115 * XXX - Setting ip_checkinterface mostly implements the receive side of 116 * the Strong ES model described in RFC 1122, but since the routing table 117 * and transmit implementation do not implement the Strong ES model, 118 * setting this to 1 results in an odd hybrid. 119 * 120 * XXX - ip_checkinterface currently must be disabled if you use ipnat 121 * to translate the destination address to another local interface. 122 * 123 * XXX - ip_checkinterface must be disabled if you add IP aliases 124 * to the loopback interface instead of the interface where the 125 * packets for those addresses are received. 126 */ 127 static int ip_checkinterface = 0; 128 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 129 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 130 131 struct pfil_head inet_pfil_hook; /* Packet filter hooks */ 132 133 static struct ifqueue ipintrq; 134 static int ipqmaxlen = IFQ_MAXLEN; 135 136 extern struct domain inetdomain; 137 extern struct protosw inetsw[]; 138 u_char ip_protox[IPPROTO_MAX]; 139 struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 140 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 141 u_long in_ifaddrhmask; /* mask for hash table */ 142 143 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 144 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 145 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 146 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 147 148 struct ipstat ipstat; 149 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 150 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 151 152 /* 153 * IP datagram reassembly. 154 */ 155 #define IPREASS_NHASH_LOG2 6 156 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 157 #define IPREASS_HMASK (IPREASS_NHASH - 1) 158 #define IPREASS_HASH(x,y) \ 159 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 160 161 static uma_zone_t ipq_zone; 162 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 163 static struct mtx ipqlock; 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) 169 170 static void maxnipq_update(void); 171 static void ipq_zone_change(void *); 172 173 static int maxnipq; /* Administrative limit on # reass queues. */ 174 static int nipq = 0; /* Total # of reass queues */ 175 SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, &nipq, 0, 176 "Current number of IPv4 fragment reassembly queue entries"); 177 178 static int maxfragsperpacket; 179 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 180 &maxfragsperpacket, 0, 181 "Maximum number of IPv4 fragments allowed per packet"); 182 183 struct callout ipport_tick_callout; 184 185 #ifdef IPCTL_DEFMTU 186 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 187 &ip_mtu, 0, "Default MTU"); 188 #endif 189 190 #ifdef IPSTEALTH 191 int ipstealth = 0; 192 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 193 &ipstealth, 0, ""); 194 #endif 195 196 /* 197 * ipfw_ether and ipfw_bridge hooks. 198 * XXX: Temporary until those are converted to pfil_hooks as well. 199 */ 200 ip_fw_chk_t *ip_fw_chk_ptr = NULL; 201 ip_dn_io_t *ip_dn_io_ptr = NULL; 202 int fw_one_pass = 1; 203 204 static void ip_freef(struct ipqhead *, struct ipq *); 205 206 /* 207 * IP initialization: fill in IP protocol switch table. 208 * All protocols not implemented in kernel go to raw IP protocol handler. 209 */ 210 void 211 ip_init() 212 { 213 register struct protosw *pr; 214 register int i; 215 216 TAILQ_INIT(&in_ifaddrhead); 217 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 218 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 219 if (pr == NULL) 220 panic("ip_init: PF_INET not found"); 221 222 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 223 for (i = 0; i < IPPROTO_MAX; i++) 224 ip_protox[i] = pr - inetsw; 225 /* 226 * Cycle through IP protocols and put them into the appropriate place 227 * in ip_protox[]. 228 */ 229 for (pr = inetdomain.dom_protosw; 230 pr < inetdomain.dom_protoswNPROTOSW; pr++) 231 if (pr->pr_domain->dom_family == PF_INET && 232 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 233 /* Be careful to only index valid IP protocols. */ 234 if (pr->pr_protocol < IPPROTO_MAX) 235 ip_protox[pr->pr_protocol] = pr - inetsw; 236 } 237 238 /* Initialize packet filter hooks. */ 239 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 240 inet_pfil_hook.ph_af = AF_INET; 241 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 242 printf("%s: WARNING: unable to register pfil hook, " 243 "error %d\n", __func__, i); 244 245 /* Initialize IP reassembly queue. */ 246 IPQ_LOCK_INIT(); 247 for (i = 0; i < IPREASS_NHASH; i++) 248 TAILQ_INIT(&ipq[i]); 249 maxnipq = nmbclusters / 32; 250 maxfragsperpacket = 16; 251 ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 252 NULL, UMA_ALIGN_PTR, 0); 253 maxnipq_update(); 254 255 /* Start ipport_tick. */ 256 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE); 257 ipport_tick(NULL); 258 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, 259 SHUTDOWN_PRI_DEFAULT); 260 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, 261 NULL, EVENTHANDLER_PRI_ANY); 262 263 /* Initialize various other remaining things. */ 264 ip_id = time_second & 0xffff; 265 ipintrq.ifq_maxlen = ipqmaxlen; 266 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF); 267 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE); 268 } 269 270 void ip_fini(xtp) 271 void *xtp; 272 { 273 callout_stop(&ipport_tick_callout); 274 } 275 276 /* 277 * Ip input routine. Checksum and byte swap header. If fragmented 278 * try to reassemble. Process options. Pass to next level. 279 */ 280 void 281 ip_input(struct mbuf *m) 282 { 283 struct ip *ip = NULL; 284 struct in_ifaddr *ia = NULL; 285 struct ifaddr *ifa; 286 int checkif, hlen = 0; 287 u_short sum; 288 int dchg = 0; /* dest changed after fw */ 289 struct in_addr odst; /* original dst address */ 290 291 M_ASSERTPKTHDR(m); 292 293 if (m->m_flags & M_FASTFWD_OURS) { 294 /* 295 * Firewall or NAT changed destination to local. 296 * We expect ip_len and ip_off to be in host byte order. 297 */ 298 m->m_flags &= ~M_FASTFWD_OURS; 299 /* Set up some basics that will be used later. */ 300 ip = mtod(m, struct ip *); 301 hlen = ip->ip_hl << 2; 302 goto ours; 303 } 304 305 ipstat.ips_total++; 306 307 if (m->m_pkthdr.len < sizeof(struct ip)) 308 goto tooshort; 309 310 if (m->m_len < sizeof (struct ip) && 311 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 312 ipstat.ips_toosmall++; 313 return; 314 } 315 ip = mtod(m, struct ip *); 316 317 if (ip->ip_v != IPVERSION) { 318 ipstat.ips_badvers++; 319 goto bad; 320 } 321 322 hlen = ip->ip_hl << 2; 323 if (hlen < sizeof(struct ip)) { /* minimum header length */ 324 ipstat.ips_badhlen++; 325 goto bad; 326 } 327 if (hlen > m->m_len) { 328 if ((m = m_pullup(m, hlen)) == NULL) { 329 ipstat.ips_badhlen++; 330 return; 331 } 332 ip = mtod(m, struct ip *); 333 } 334 335 /* 127/8 must not appear on wire - RFC1122 */ 336 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 337 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 338 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 339 ipstat.ips_badaddr++; 340 goto bad; 341 } 342 } 343 344 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 345 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 346 } else { 347 if (hlen == sizeof(struct ip)) { 348 sum = in_cksum_hdr(ip); 349 } else { 350 sum = in_cksum(m, hlen); 351 } 352 } 353 if (sum) { 354 ipstat.ips_badsum++; 355 goto bad; 356 } 357 358 #ifdef ALTQ 359 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 360 /* packet is dropped by traffic conditioner */ 361 return; 362 #endif 363 364 /* 365 * Convert fields to host representation. 366 */ 367 ip->ip_len = ntohs(ip->ip_len); 368 if (ip->ip_len < hlen) { 369 ipstat.ips_badlen++; 370 goto bad; 371 } 372 ip->ip_off = ntohs(ip->ip_off); 373 374 /* 375 * Check that the amount of data in the buffers 376 * is as at least much as the IP header would have us expect. 377 * Trim mbufs if longer than we expect. 378 * Drop packet if shorter than we expect. 379 */ 380 if (m->m_pkthdr.len < ip->ip_len) { 381 tooshort: 382 ipstat.ips_tooshort++; 383 goto bad; 384 } 385 if (m->m_pkthdr.len > ip->ip_len) { 386 if (m->m_len == m->m_pkthdr.len) { 387 m->m_len = ip->ip_len; 388 m->m_pkthdr.len = ip->ip_len; 389 } else 390 m_adj(m, ip->ip_len - m->m_pkthdr.len); 391 } 392 #if defined(IPSEC) || defined(FAST_IPSEC) 393 /* 394 * Bypass packet filtering for packets from a tunnel (gif). 395 */ 396 if (ip_ipsec_filtergif(m)) 397 goto passin; 398 #endif /* IPSEC */ 399 400 /* 401 * Run through list of hooks for input packets. 402 * 403 * NB: Beware of the destination address changing (e.g. 404 * by NAT rewriting). When this happens, tell 405 * ip_forward to do the right thing. 406 */ 407 408 /* Jump over all PFIL processing if hooks are not active. */ 409 if (!PFIL_HOOKED(&inet_pfil_hook)) 410 goto passin; 411 412 odst = ip->ip_dst; 413 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 414 PFIL_IN, NULL) != 0) 415 return; 416 if (m == NULL) /* consumed by filter */ 417 return; 418 419 ip = mtod(m, struct ip *); 420 dchg = (odst.s_addr != ip->ip_dst.s_addr); 421 422 #ifdef IPFIREWALL_FORWARD 423 if (m->m_flags & M_FASTFWD_OURS) { 424 m->m_flags &= ~M_FASTFWD_OURS; 425 goto ours; 426 } 427 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 428 /* 429 * Directly ship on the packet. This allows to forward packets 430 * that were destined for us to some other directly connected 431 * host. 432 */ 433 ip_forward(m, dchg); 434 return; 435 } 436 #endif /* IPFIREWALL_FORWARD */ 437 438 passin: 439 /* 440 * Process options and, if not destined for us, 441 * ship it on. ip_dooptions returns 1 when an 442 * error was detected (causing an icmp message 443 * to be sent and the original packet to be freed). 444 */ 445 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 446 return; 447 448 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 449 * matter if it is destined to another node, or whether it is 450 * a multicast one, RSVP wants it! and prevents it from being forwarded 451 * anywhere else. Also checks if the rsvp daemon is running before 452 * grabbing the packet. 453 */ 454 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 455 goto ours; 456 457 /* 458 * Check our list of addresses, to see if the packet is for us. 459 * If we don't have any addresses, assume any unicast packet 460 * we receive might be for us (and let the upper layers deal 461 * with it). 462 */ 463 if (TAILQ_EMPTY(&in_ifaddrhead) && 464 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 465 goto ours; 466 467 /* 468 * Enable a consistency check between the destination address 469 * and the arrival interface for a unicast packet (the RFC 1122 470 * strong ES model) if IP forwarding is disabled and the packet 471 * is not locally generated and the packet is not subject to 472 * 'ipfw fwd'. 473 * 474 * XXX - Checking also should be disabled if the destination 475 * address is ipnat'ed to a different interface. 476 * 477 * XXX - Checking is incompatible with IP aliases added 478 * to the loopback interface instead of the interface where 479 * the packets are received. 480 * 481 * XXX - This is the case for carp vhost IPs as well so we 482 * insert a workaround. If the packet got here, we already 483 * checked with carp_iamatch() and carp_forus(). 484 */ 485 checkif = ip_checkinterface && (ipforwarding == 0) && 486 m->m_pkthdr.rcvif != NULL && 487 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 488 #ifdef DEV_CARP 489 !m->m_pkthdr.rcvif->if_carp && 490 #endif 491 (dchg == 0); 492 493 /* 494 * Check for exact addresses in the hash bucket. 495 */ 496 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 497 /* 498 * If the address matches, verify that the packet 499 * arrived via the correct interface if checking is 500 * enabled. 501 */ 502 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 503 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 504 goto ours; 505 } 506 /* 507 * Check for broadcast addresses. 508 * 509 * Only accept broadcast packets that arrive via the matching 510 * interface. Reception of forwarded directed broadcasts would 511 * be handled via ip_forward() and ether_output() with the loopback 512 * into the stack for SIMPLEX interfaces handled by ether_output(). 513 */ 514 if (m->m_pkthdr.rcvif != NULL && 515 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 516 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 517 if (ifa->ifa_addr->sa_family != AF_INET) 518 continue; 519 ia = ifatoia(ifa); 520 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 521 ip->ip_dst.s_addr) 522 goto ours; 523 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) 524 goto ours; 525 #ifdef BOOTP_COMPAT 526 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 527 goto ours; 528 #endif 529 } 530 } 531 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 532 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 533 ipstat.ips_cantforward++; 534 m_freem(m); 535 return; 536 } 537 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 538 struct in_multi *inm; 539 if (ip_mrouter) { 540 /* 541 * If we are acting as a multicast router, all 542 * incoming multicast packets are passed to the 543 * kernel-level multicast forwarding function. 544 * The packet is returned (relatively) intact; if 545 * ip_mforward() returns a non-zero value, the packet 546 * must be discarded, else it may be accepted below. 547 */ 548 if (ip_mforward && 549 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 550 ipstat.ips_cantforward++; 551 m_freem(m); 552 return; 553 } 554 555 /* 556 * The process-level routing daemon needs to receive 557 * all multicast IGMP packets, whether or not this 558 * host belongs to their destination groups. 559 */ 560 if (ip->ip_p == IPPROTO_IGMP) 561 goto ours; 562 ipstat.ips_forward++; 563 } 564 /* 565 * See if we belong to the destination multicast group on the 566 * arrival interface. 567 */ 568 IN_MULTI_LOCK(); 569 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 570 IN_MULTI_UNLOCK(); 571 if (inm == NULL) { 572 ipstat.ips_notmember++; 573 m_freem(m); 574 return; 575 } 576 goto ours; 577 } 578 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 579 goto ours; 580 if (ip->ip_dst.s_addr == INADDR_ANY) 581 goto ours; 582 583 /* 584 * FAITH(Firewall Aided Internet Translator) 585 */ 586 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 587 if (ip_keepfaith) { 588 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 589 goto ours; 590 } 591 m_freem(m); 592 return; 593 } 594 595 /* 596 * Not for us; forward if possible and desirable. 597 */ 598 if (ipforwarding == 0) { 599 ipstat.ips_cantforward++; 600 m_freem(m); 601 } else { 602 #if defined(IPSEC) || defined(FAST_IPSEC) 603 if (ip_ipsec_fwd(m)) 604 goto bad; 605 #endif /* IPSEC */ 606 ip_forward(m, dchg); 607 } 608 return; 609 610 ours: 611 #ifdef IPSTEALTH 612 /* 613 * IPSTEALTH: Process non-routing options only 614 * if the packet is destined for us. 615 */ 616 if (ipstealth && hlen > sizeof (struct ip) && 617 ip_dooptions(m, 1)) 618 return; 619 #endif /* IPSTEALTH */ 620 621 /* Count the packet in the ip address stats */ 622 if (ia != NULL) { 623 ia->ia_ifa.if_ipackets++; 624 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 625 } 626 627 /* 628 * Attempt reassembly; if it succeeds, proceed. 629 * ip_reass() will return a different mbuf. 630 */ 631 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 632 m = ip_reass(m); 633 if (m == NULL) 634 return; 635 ip = mtod(m, struct ip *); 636 /* Get the header length of the reassembled packet */ 637 hlen = ip->ip_hl << 2; 638 } 639 640 /* 641 * Further protocols expect the packet length to be w/o the 642 * IP header. 643 */ 644 ip->ip_len -= hlen; 645 646 #if defined(IPSEC) || defined(FAST_IPSEC) 647 /* 648 * enforce IPsec policy checking if we are seeing last header. 649 * note that we do not visit this with protocols with pcb layer 650 * code - like udp/tcp/raw ip. 651 */ 652 if (ip_ipsec_input(m)) 653 goto bad; 654 #endif /* IPSEC */ 655 656 /* 657 * Switch out to protocol's input routine. 658 */ 659 ipstat.ips_delivered++; 660 661 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 662 return; 663 bad: 664 m_freem(m); 665 } 666 667 /* 668 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 669 * max has slightly different semantics than the sysctl, for historical 670 * reasons. 671 */ 672 static void 673 maxnipq_update(void) 674 { 675 676 /* 677 * -1 for unlimited allocation. 678 */ 679 if (maxnipq < 0) 680 uma_zone_set_max(ipq_zone, 0); 681 /* 682 * Positive number for specific bound. 683 */ 684 if (maxnipq > 0) 685 uma_zone_set_max(ipq_zone, maxnipq); 686 /* 687 * Zero specifies no further fragment queue allocation -- set the 688 * bound very low, but rely on implementation elsewhere to actually 689 * prevent allocation and reclaim current queues. 690 */ 691 if (maxnipq == 0) 692 uma_zone_set_max(ipq_zone, 1); 693 } 694 695 static void 696 ipq_zone_change(void *tag) 697 { 698 699 if (maxnipq > 0 && maxnipq < (nmbclusters / 32)) { 700 maxnipq = nmbclusters / 32; 701 maxnipq_update(); 702 } 703 } 704 705 static int 706 sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 707 { 708 int error, i; 709 710 i = maxnipq; 711 error = sysctl_handle_int(oidp, &i, 0, req); 712 if (error || !req->newptr) 713 return (error); 714 715 /* 716 * XXXRW: Might be a good idea to sanity check the argument and place 717 * an extreme upper bound. 718 */ 719 if (i < -1) 720 return (EINVAL); 721 maxnipq = i; 722 maxnipq_update(); 723 return (0); 724 } 725 726 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 727 NULL, 0, sysctl_maxnipq, "I", 728 "Maximum number of IPv4 fragment reassembly queue entries"); 729 730 /* 731 * Take incoming datagram fragment and try to reassemble it into 732 * whole datagram. If the argument is the first fragment or one 733 * in between the function will return NULL and store the mbuf 734 * in the fragment chain. If the argument is the last fragment 735 * the packet will be reassembled and the pointer to the new 736 * mbuf returned for further processing. Only m_tags attached 737 * to the first packet/fragment are preserved. 738 * The IP header is *NOT* adjusted out of iplen. 739 */ 740 741 struct mbuf * 742 ip_reass(struct mbuf *m) 743 { 744 struct ip *ip; 745 struct mbuf *p, *q, *nq, *t; 746 struct ipq *fp = NULL; 747 struct ipqhead *head; 748 int i, hlen, next; 749 u_int8_t ecn, ecn0; 750 u_short hash; 751 752 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 753 if (maxnipq == 0 || maxfragsperpacket == 0) { 754 ipstat.ips_fragments++; 755 ipstat.ips_fragdropped++; 756 m_freem(m); 757 return (NULL); 758 } 759 760 ip = mtod(m, struct ip *); 761 hlen = ip->ip_hl << 2; 762 763 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 764 head = &ipq[hash]; 765 IPQ_LOCK(); 766 767 /* 768 * Look for queue of fragments 769 * of this datagram. 770 */ 771 TAILQ_FOREACH(fp, head, ipq_list) 772 if (ip->ip_id == fp->ipq_id && 773 ip->ip_src.s_addr == fp->ipq_src.s_addr && 774 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 775 #ifdef MAC 776 mac_fragment_match(m, fp) && 777 #endif 778 ip->ip_p == fp->ipq_p) 779 goto found; 780 781 fp = NULL; 782 783 /* 784 * Attempt to trim the number of allocated fragment queues if it 785 * exceeds the administrative limit. 786 */ 787 if ((nipq > maxnipq) && (maxnipq > 0)) { 788 /* 789 * drop something from the tail of the current queue 790 * before proceeding further 791 */ 792 struct ipq *q = TAILQ_LAST(head, ipqhead); 793 if (q == NULL) { /* gak */ 794 for (i = 0; i < IPREASS_NHASH; i++) { 795 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 796 if (r) { 797 ipstat.ips_fragtimeout += r->ipq_nfrags; 798 ip_freef(&ipq[i], r); 799 break; 800 } 801 } 802 } else { 803 ipstat.ips_fragtimeout += q->ipq_nfrags; 804 ip_freef(head, q); 805 } 806 } 807 808 found: 809 /* 810 * Adjust ip_len to not reflect header, 811 * convert offset of this to bytes. 812 */ 813 ip->ip_len -= hlen; 814 if (ip->ip_off & IP_MF) { 815 /* 816 * Make sure that fragments have a data length 817 * that's a non-zero multiple of 8 bytes. 818 */ 819 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 820 ipstat.ips_toosmall++; /* XXX */ 821 goto dropfrag; 822 } 823 m->m_flags |= M_FRAG; 824 } else 825 m->m_flags &= ~M_FRAG; 826 ip->ip_off <<= 3; 827 828 829 /* 830 * Attempt reassembly; if it succeeds, proceed. 831 * ip_reass() will return a different mbuf. 832 */ 833 ipstat.ips_fragments++; 834 m->m_pkthdr.header = ip; 835 836 /* Previous ip_reass() started here. */ 837 /* 838 * Presence of header sizes in mbufs 839 * would confuse code below. 840 */ 841 m->m_data += hlen; 842 m->m_len -= hlen; 843 844 /* 845 * If first fragment to arrive, create a reassembly queue. 846 */ 847 if (fp == NULL) { 848 fp = uma_zalloc(ipq_zone, M_NOWAIT); 849 if (fp == NULL) 850 goto dropfrag; 851 #ifdef MAC 852 if (mac_init_ipq(fp, M_NOWAIT) != 0) { 853 uma_zfree(ipq_zone, fp); 854 fp = NULL; 855 goto dropfrag; 856 } 857 mac_create_ipq(m, fp); 858 #endif 859 TAILQ_INSERT_HEAD(head, fp, ipq_list); 860 nipq++; 861 fp->ipq_nfrags = 1; 862 fp->ipq_ttl = IPFRAGTTL; 863 fp->ipq_p = ip->ip_p; 864 fp->ipq_id = ip->ip_id; 865 fp->ipq_src = ip->ip_src; 866 fp->ipq_dst = ip->ip_dst; 867 fp->ipq_frags = m; 868 m->m_nextpkt = NULL; 869 goto done; 870 } else { 871 fp->ipq_nfrags++; 872 #ifdef MAC 873 mac_update_ipq(m, fp); 874 #endif 875 } 876 877 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 878 879 /* 880 * Handle ECN by comparing this segment with the first one; 881 * if CE is set, do not lose CE. 882 * drop if CE and not-ECT are mixed for the same packet. 883 */ 884 ecn = ip->ip_tos & IPTOS_ECN_MASK; 885 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 886 if (ecn == IPTOS_ECN_CE) { 887 if (ecn0 == IPTOS_ECN_NOTECT) 888 goto dropfrag; 889 if (ecn0 != IPTOS_ECN_CE) 890 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 891 } 892 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 893 goto dropfrag; 894 895 /* 896 * Find a segment which begins after this one does. 897 */ 898 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 899 if (GETIP(q)->ip_off > ip->ip_off) 900 break; 901 902 /* 903 * If there is a preceding segment, it may provide some of 904 * our data already. If so, drop the data from the incoming 905 * segment. If it provides all of our data, drop us, otherwise 906 * stick new segment in the proper place. 907 * 908 * If some of the data is dropped from the the preceding 909 * segment, then it's checksum is invalidated. 910 */ 911 if (p) { 912 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 913 if (i > 0) { 914 if (i >= ip->ip_len) 915 goto dropfrag; 916 m_adj(m, i); 917 m->m_pkthdr.csum_flags = 0; 918 ip->ip_off += i; 919 ip->ip_len -= i; 920 } 921 m->m_nextpkt = p->m_nextpkt; 922 p->m_nextpkt = m; 923 } else { 924 m->m_nextpkt = fp->ipq_frags; 925 fp->ipq_frags = m; 926 } 927 928 /* 929 * While we overlap succeeding segments trim them or, 930 * if they are completely covered, dequeue them. 931 */ 932 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 933 q = nq) { 934 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 935 if (i < GETIP(q)->ip_len) { 936 GETIP(q)->ip_len -= i; 937 GETIP(q)->ip_off += i; 938 m_adj(q, i); 939 q->m_pkthdr.csum_flags = 0; 940 break; 941 } 942 nq = q->m_nextpkt; 943 m->m_nextpkt = nq; 944 ipstat.ips_fragdropped++; 945 fp->ipq_nfrags--; 946 m_freem(q); 947 } 948 949 /* 950 * Check for complete reassembly and perform frag per packet 951 * limiting. 952 * 953 * Frag limiting is performed here so that the nth frag has 954 * a chance to complete the packet before we drop the packet. 955 * As a result, n+1 frags are actually allowed per packet, but 956 * only n will ever be stored. (n = maxfragsperpacket.) 957 * 958 */ 959 next = 0; 960 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 961 if (GETIP(q)->ip_off != next) { 962 if (fp->ipq_nfrags > maxfragsperpacket) { 963 ipstat.ips_fragdropped += fp->ipq_nfrags; 964 ip_freef(head, fp); 965 } 966 goto done; 967 } 968 next += GETIP(q)->ip_len; 969 } 970 /* Make sure the last packet didn't have the IP_MF flag */ 971 if (p->m_flags & M_FRAG) { 972 if (fp->ipq_nfrags > maxfragsperpacket) { 973 ipstat.ips_fragdropped += fp->ipq_nfrags; 974 ip_freef(head, fp); 975 } 976 goto done; 977 } 978 979 /* 980 * Reassembly is complete. Make sure the packet is a sane size. 981 */ 982 q = fp->ipq_frags; 983 ip = GETIP(q); 984 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 985 ipstat.ips_toolong++; 986 ipstat.ips_fragdropped += fp->ipq_nfrags; 987 ip_freef(head, fp); 988 goto done; 989 } 990 991 /* 992 * Concatenate fragments. 993 */ 994 m = q; 995 t = m->m_next; 996 m->m_next = NULL; 997 m_cat(m, t); 998 nq = q->m_nextpkt; 999 q->m_nextpkt = NULL; 1000 for (q = nq; q != NULL; q = nq) { 1001 nq = q->m_nextpkt; 1002 q->m_nextpkt = NULL; 1003 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1004 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1005 m_cat(m, q); 1006 } 1007 /* 1008 * In order to do checksumming faster we do 'end-around carry' here 1009 * (and not in for{} loop), though it implies we are not going to 1010 * reassemble more than 64k fragments. 1011 */ 1012 m->m_pkthdr.csum_data = 1013 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1014 #ifdef MAC 1015 mac_create_datagram_from_ipq(fp, m); 1016 mac_destroy_ipq(fp); 1017 #endif 1018 1019 /* 1020 * Create header for new ip packet by modifying header of first 1021 * packet; dequeue and discard fragment reassembly header. 1022 * Make header visible. 1023 */ 1024 ip->ip_len = (ip->ip_hl << 2) + next; 1025 ip->ip_src = fp->ipq_src; 1026 ip->ip_dst = fp->ipq_dst; 1027 TAILQ_REMOVE(head, fp, ipq_list); 1028 nipq--; 1029 uma_zfree(ipq_zone, fp); 1030 m->m_len += (ip->ip_hl << 2); 1031 m->m_data -= (ip->ip_hl << 2); 1032 /* some debugging cruft by sklower, below, will go away soon */ 1033 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1034 m_fixhdr(m); 1035 ipstat.ips_reassembled++; 1036 IPQ_UNLOCK(); 1037 return (m); 1038 1039 dropfrag: 1040 ipstat.ips_fragdropped++; 1041 if (fp != NULL) 1042 fp->ipq_nfrags--; 1043 m_freem(m); 1044 done: 1045 IPQ_UNLOCK(); 1046 return (NULL); 1047 1048 #undef GETIP 1049 } 1050 1051 /* 1052 * Free a fragment reassembly header and all 1053 * associated datagrams. 1054 */ 1055 static void 1056 ip_freef(fhp, fp) 1057 struct ipqhead *fhp; 1058 struct ipq *fp; 1059 { 1060 register 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() 1081 { 1082 register 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() 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(dst) 1201 struct in_addr dst; 1202 { 1203 struct route sro; 1204 struct sockaddr_in *sin; 1205 struct in_ifaddr *ifa; 1206 1207 bzero(&sro, sizeof(sro)); 1208 sin = (struct sockaddr_in *)&sro.ro_dst; 1209 sin->sin_family = AF_INET; 1210 sin->sin_len = sizeof(*sin); 1211 sin->sin_addr = dst; 1212 rtalloc_ign(&sro, RTF_CLONING); 1213 1214 if (sro.ro_rt == NULL) 1215 return (NULL); 1216 1217 ifa = ifatoia(sro.ro_rt->rt_ifa); 1218 RTFREE(sro.ro_rt); 1219 return (ifa); 1220 } 1221 1222 u_char inetctlerrmap[PRC_NCMDS] = { 1223 0, 0, 0, 0, 1224 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1225 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1226 EMSGSIZE, EHOSTUNREACH, 0, 0, 1227 0, 0, EHOSTUNREACH, 0, 1228 ENOPROTOOPT, ECONNREFUSED 1229 }; 1230 1231 /* 1232 * Forward a packet. If some error occurs return the sender 1233 * an icmp packet. Note we can't always generate a meaningful 1234 * icmp message because icmp doesn't have a large enough repertoire 1235 * of codes and types. 1236 * 1237 * If not forwarding, just drop the packet. This could be confusing 1238 * if ipforwarding was zero but some routing protocol was advancing 1239 * us as a gateway to somewhere. However, we must let the routing 1240 * protocol deal with that. 1241 * 1242 * The srcrt parameter indicates whether the packet is being forwarded 1243 * via a source route. 1244 */ 1245 void 1246 ip_forward(struct mbuf *m, int srcrt) 1247 { 1248 struct ip *ip = mtod(m, struct ip *); 1249 struct in_ifaddr *ia = NULL; 1250 struct mbuf *mcopy; 1251 struct in_addr dest; 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 if (!srcrt && (ia = ip_rtaddr(ip->ip_dst)) == NULL) { 1272 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1273 return; 1274 } 1275 1276 /* 1277 * Save the IP header and at most 8 bytes of the payload, 1278 * in case we need to generate an ICMP message to the src. 1279 * 1280 * XXX this can be optimized a lot by saving the data in a local 1281 * buffer on the stack (72 bytes at most), and only allocating the 1282 * mbuf if really necessary. The vast majority of the packets 1283 * are forwarded without having to send an ICMP back (either 1284 * because unnecessary, or because rate limited), so we are 1285 * really we are wasting a lot of work here. 1286 * 1287 * We don't use m_copy() because it might return a reference 1288 * to a shared cluster. Both this function and ip_output() 1289 * assume exclusive access to the IP header in `m', so any 1290 * data in a cluster may change before we reach icmp_error(). 1291 */ 1292 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1293 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1294 /* 1295 * It's probably ok if the pkthdr dup fails (because 1296 * the deep copy of the tag chain failed), but for now 1297 * be conservative and just discard the copy since 1298 * code below may some day want the tags. 1299 */ 1300 m_free(mcopy); 1301 mcopy = NULL; 1302 } 1303 if (mcopy != NULL) { 1304 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy)); 1305 mcopy->m_pkthdr.len = mcopy->m_len; 1306 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1307 } 1308 1309 #ifdef IPSTEALTH 1310 if (!ipstealth) { 1311 #endif 1312 ip->ip_ttl -= IPTTLDEC; 1313 #ifdef IPSTEALTH 1314 } 1315 #endif 1316 1317 /* 1318 * If forwarding packet using same interface that it came in on, 1319 * perhaps should send a redirect to sender to shortcut a hop. 1320 * Only send redirect if source is sending directly to us, 1321 * and if packet was not source routed (or has any options). 1322 * Also, don't send redirect if forwarding using a default route 1323 * or a route modified by a redirect. 1324 */ 1325 dest.s_addr = 0; 1326 if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) { 1327 struct sockaddr_in *sin; 1328 struct route ro; 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 rtalloc_ign(&ro, RTF_CLONING); 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 error = ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL); 1361 if (error) 1362 ipstat.ips_cantforward++; 1363 else { 1364 ipstat.ips_forward++; 1365 if (type) 1366 ipstat.ips_redirectsent++; 1367 else { 1368 if (mcopy) 1369 m_freem(mcopy); 1370 return; 1371 } 1372 } 1373 if (mcopy == NULL) 1374 return; 1375 1376 switch (error) { 1377 1378 case 0: /* forwarded, but need redirect */ 1379 /* type, code set above */ 1380 break; 1381 1382 case ENETUNREACH: /* shouldn't happen, checked above */ 1383 case EHOSTUNREACH: 1384 case ENETDOWN: 1385 case EHOSTDOWN: 1386 default: 1387 type = ICMP_UNREACH; 1388 code = ICMP_UNREACH_HOST; 1389 break; 1390 1391 case EMSGSIZE: 1392 type = ICMP_UNREACH; 1393 code = ICMP_UNREACH_NEEDFRAG; 1394 1395 #if defined(IPSEC) || defined(FAST_IPSEC) 1396 mtu = ip_ipsec_mtu(m); 1397 #endif /* IPSEC */ 1398 /* 1399 * If the MTU wasn't set before use the interface mtu or 1400 * fall back to the next smaller mtu step compared to the 1401 * current packet size. 1402 */ 1403 if (mtu == 0) { 1404 if (ia != NULL) 1405 mtu = ia->ia_ifp->if_mtu; 1406 else 1407 mtu = ip_next_mtu(ip->ip_len, 0); 1408 } 1409 ipstat.ips_cantfrag++; 1410 break; 1411 1412 case ENOBUFS: 1413 /* 1414 * A router should not generate ICMP_SOURCEQUENCH as 1415 * required in RFC1812 Requirements for IP Version 4 Routers. 1416 * Source quench could be a big problem under DoS attacks, 1417 * or if the underlying interface is rate-limited. 1418 * Those who need source quench packets may re-enable them 1419 * via the net.inet.ip.sendsourcequench sysctl. 1420 */ 1421 if (ip_sendsourcequench == 0) { 1422 m_freem(mcopy); 1423 return; 1424 } else { 1425 type = ICMP_SOURCEQUENCH; 1426 code = 0; 1427 } 1428 break; 1429 1430 case EACCES: /* ipfw denied packet */ 1431 m_freem(mcopy); 1432 return; 1433 } 1434 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1435 } 1436 1437 void 1438 ip_savecontrol(inp, mp, ip, m) 1439 register struct inpcb *inp; 1440 register struct mbuf **mp; 1441 register struct ip *ip; 1442 register struct mbuf *m; 1443 { 1444 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1445 struct bintime bt; 1446 1447 bintime(&bt); 1448 if (inp->inp_socket->so_options & SO_BINTIME) { 1449 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1450 SCM_BINTIME, SOL_SOCKET); 1451 if (*mp) 1452 mp = &(*mp)->m_next; 1453 } 1454 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1455 struct timeval tv; 1456 1457 bintime2timeval(&bt, &tv); 1458 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1459 SCM_TIMESTAMP, SOL_SOCKET); 1460 if (*mp) 1461 mp = &(*mp)->m_next; 1462 } 1463 } 1464 if (inp->inp_flags & INP_RECVDSTADDR) { 1465 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1466 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1467 if (*mp) 1468 mp = &(*mp)->m_next; 1469 } 1470 if (inp->inp_flags & INP_RECVTTL) { 1471 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1472 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1473 if (*mp) 1474 mp = &(*mp)->m_next; 1475 } 1476 #ifdef notyet 1477 /* XXX 1478 * Moving these out of udp_input() made them even more broken 1479 * than they already were. 1480 */ 1481 /* options were tossed already */ 1482 if (inp->inp_flags & INP_RECVOPTS) { 1483 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1484 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1485 if (*mp) 1486 mp = &(*mp)->m_next; 1487 } 1488 /* ip_srcroute doesn't do what we want here, need to fix */ 1489 if (inp->inp_flags & INP_RECVRETOPTS) { 1490 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1491 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1492 if (*mp) 1493 mp = &(*mp)->m_next; 1494 } 1495 #endif 1496 if (inp->inp_flags & INP_RECVIF) { 1497 struct ifnet *ifp; 1498 struct sdlbuf { 1499 struct sockaddr_dl sdl; 1500 u_char pad[32]; 1501 } sdlbuf; 1502 struct sockaddr_dl *sdp; 1503 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1504 1505 if (((ifp = m->m_pkthdr.rcvif)) 1506 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1507 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1508 /* 1509 * Change our mind and don't try copy. 1510 */ 1511 if ((sdp->sdl_family != AF_LINK) 1512 || (sdp->sdl_len > sizeof(sdlbuf))) { 1513 goto makedummy; 1514 } 1515 bcopy(sdp, sdl2, sdp->sdl_len); 1516 } else { 1517 makedummy: 1518 sdl2->sdl_len 1519 = offsetof(struct sockaddr_dl, sdl_data[0]); 1520 sdl2->sdl_family = AF_LINK; 1521 sdl2->sdl_index = 0; 1522 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1523 } 1524 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1525 IP_RECVIF, IPPROTO_IP); 1526 if (*mp) 1527 mp = &(*mp)->m_next; 1528 } 1529 } 1530 1531 /* 1532 * XXX these routines are called from the upper part of the kernel. 1533 * They need to be locked when we remove Giant. 1534 * 1535 * They could also be moved to ip_mroute.c, since all the RSVP 1536 * handling is done there already. 1537 */ 1538 static int ip_rsvp_on; 1539 struct socket *ip_rsvpd; 1540 int 1541 ip_rsvp_init(struct socket *so) 1542 { 1543 if (so->so_type != SOCK_RAW || 1544 so->so_proto->pr_protocol != IPPROTO_RSVP) 1545 return EOPNOTSUPP; 1546 1547 if (ip_rsvpd != NULL) 1548 return EADDRINUSE; 1549 1550 ip_rsvpd = so; 1551 /* 1552 * This may seem silly, but we need to be sure we don't over-increment 1553 * the RSVP counter, in case something slips up. 1554 */ 1555 if (!ip_rsvp_on) { 1556 ip_rsvp_on = 1; 1557 rsvp_on++; 1558 } 1559 1560 return 0; 1561 } 1562 1563 int 1564 ip_rsvp_done(void) 1565 { 1566 ip_rsvpd = NULL; 1567 /* 1568 * This may seem silly, but we need to be sure we don't over-decrement 1569 * the RSVP counter, in case something slips up. 1570 */ 1571 if (ip_rsvp_on) { 1572 ip_rsvp_on = 0; 1573 rsvp_on--; 1574 } 1575 return 0; 1576 } 1577 1578 void 1579 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1580 { 1581 if (rsvp_input_p) { /* call the real one if loaded */ 1582 rsvp_input_p(m, off); 1583 return; 1584 } 1585 1586 /* Can still get packets with rsvp_on = 0 if there is a local member 1587 * of the group to which the RSVP packet is addressed. But in this 1588 * case we want to throw the packet away. 1589 */ 1590 1591 if (!rsvp_on) { 1592 m_freem(m); 1593 return; 1594 } 1595 1596 if (ip_rsvpd != NULL) { 1597 rip_input(m, off); 1598 return; 1599 } 1600 /* Drop the packet */ 1601 m_freem(m); 1602 } 1603