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/mac.h> 44 #include <sys/mbuf.h> 45 #include <sys/malloc.h> 46 #include <sys/domain.h> 47 #include <sys/protosw.h> 48 #include <sys/socket.h> 49 #include <sys/time.h> 50 #include <sys/kernel.h> 51 #include <sys/syslog.h> 52 #include <sys/sysctl.h> 53 54 #include <net/pfil.h> 55 #include <net/if.h> 56 #include <net/if_types.h> 57 #include <net/if_var.h> 58 #include <net/if_dl.h> 59 #include <net/route.h> 60 #include <net/netisr.h> 61 62 #include <netinet/in.h> 63 #include <netinet/in_systm.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 #if defined(IPSEC) || defined(FAST_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 int rsvp_on = 0; 85 86 int ipforwarding = 0; 87 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 88 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 89 90 static int ipsendredirects = 1; /* XXX */ 91 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 92 &ipsendredirects, 0, "Enable sending IP redirects"); 93 94 int ip_defttl = IPDEFTTL; 95 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 96 &ip_defttl, 0, "Maximum TTL on IP packets"); 97 98 static int ip_keepfaith = 0; 99 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 100 &ip_keepfaith, 0, 101 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 102 103 static int ip_sendsourcequench = 0; 104 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 105 &ip_sendsourcequench, 0, 106 "Enable the transmission of source quench packets"); 107 108 int ip_do_randomid = 0; 109 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 110 &ip_do_randomid, 0, 111 "Assign random ip_id values"); 112 113 /* 114 * XXX - Setting ip_checkinterface mostly implements the receive side of 115 * the Strong ES model described in RFC 1122, but since the routing table 116 * and transmit implementation do not implement the Strong ES model, 117 * setting this to 1 results in an odd hybrid. 118 * 119 * XXX - ip_checkinterface currently must be disabled if you use ipnat 120 * to translate the destination address to another local interface. 121 * 122 * XXX - ip_checkinterface must be disabled if you add IP aliases 123 * to the loopback interface instead of the interface where the 124 * packets for those addresses are received. 125 */ 126 static int ip_checkinterface = 0; 127 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 128 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 129 130 struct pfil_head inet_pfil_hook; /* Packet filter hooks */ 131 132 static struct ifqueue ipintrq; 133 static int ipqmaxlen = IFQ_MAXLEN; 134 135 extern struct domain inetdomain; 136 extern struct protosw inetsw[]; 137 u_char ip_protox[IPPROTO_MAX]; 138 struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 139 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 140 u_long in_ifaddrhmask; /* mask for hash table */ 141 142 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 143 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 144 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 145 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 146 147 struct ipstat ipstat; 148 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 149 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 150 151 /* 152 * IP datagram reassembly. 153 */ 154 #define IPREASS_NHASH_LOG2 6 155 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 156 #define IPREASS_HMASK (IPREASS_NHASH - 1) 157 #define IPREASS_HASH(x,y) \ 158 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 159 160 static uma_zone_t ipq_zone; 161 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 162 static struct mtx ipqlock; 163 164 #define IPQ_LOCK() mtx_lock(&ipqlock) 165 #define IPQ_UNLOCK() mtx_unlock(&ipqlock) 166 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 167 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 168 169 static void maxnipq_update(void); 170 static void ipq_zone_change(void *); 171 172 static int maxnipq; /* Administrative limit on # reass queues. */ 173 static int nipq = 0; /* Total # of reass queues */ 174 SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, &nipq, 0, 175 "Current number of IPv4 fragment reassembly queue entries"); 176 177 static int maxfragsperpacket; 178 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 179 &maxfragsperpacket, 0, 180 "Maximum number of IPv4 fragments allowed per packet"); 181 182 struct callout ipport_tick_callout; 183 184 #ifdef IPCTL_DEFMTU 185 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 186 &ip_mtu, 0, "Default MTU"); 187 #endif 188 189 #ifdef IPSTEALTH 190 int ipstealth = 0; 191 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 192 &ipstealth, 0, ""); 193 #endif 194 195 /* 196 * ipfw_ether and ipfw_bridge hooks. 197 * XXX: Temporary until those are converted to pfil_hooks as well. 198 */ 199 ip_fw_chk_t *ip_fw_chk_ptr = NULL; 200 ip_dn_io_t *ip_dn_io_ptr = NULL; 201 int fw_enable = 1; 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 #ifndef IPFIREWALL_FORWARD_EXTENDED 428 dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL); 429 #else 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_EXTENDED */ 440 #endif /* IPFIREWALL_FORWARD */ 441 442 passin: 443 /* 444 * Process options and, if not destined for us, 445 * ship it on. ip_dooptions returns 1 when an 446 * error was detected (causing an icmp message 447 * to be sent and the original packet to be freed). 448 */ 449 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 450 return; 451 452 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 453 * matter if it is destined to another node, or whether it is 454 * a multicast one, RSVP wants it! and prevents it from being forwarded 455 * anywhere else. Also checks if the rsvp daemon is running before 456 * grabbing the packet. 457 */ 458 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 459 goto ours; 460 461 /* 462 * Check our list of addresses, to see if the packet is for us. 463 * If we don't have any addresses, assume any unicast packet 464 * we receive might be for us (and let the upper layers deal 465 * with it). 466 */ 467 if (TAILQ_EMPTY(&in_ifaddrhead) && 468 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 469 goto ours; 470 471 /* 472 * Enable a consistency check between the destination address 473 * and the arrival interface for a unicast packet (the RFC 1122 474 * strong ES model) if IP forwarding is disabled and the packet 475 * is not locally generated and the packet is not subject to 476 * 'ipfw fwd'. 477 * 478 * XXX - Checking also should be disabled if the destination 479 * address is ipnat'ed to a different interface. 480 * 481 * XXX - Checking is incompatible with IP aliases added 482 * to the loopback interface instead of the interface where 483 * the packets are received. 484 * 485 * XXX - This is the case for carp vhost IPs as well so we 486 * insert a workaround. If the packet got here, we already 487 * checked with carp_iamatch() and carp_forus(). 488 */ 489 checkif = ip_checkinterface && (ipforwarding == 0) && 490 m->m_pkthdr.rcvif != NULL && 491 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 492 #ifdef DEV_CARP 493 !m->m_pkthdr.rcvif->if_carp && 494 #endif 495 (dchg == 0); 496 497 /* 498 * Check for exact addresses in the hash bucket. 499 */ 500 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 501 /* 502 * If the address matches, verify that the packet 503 * arrived via the correct interface if checking is 504 * enabled. 505 */ 506 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 507 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 508 goto ours; 509 } 510 /* 511 * Check for broadcast addresses. 512 * 513 * Only accept broadcast packets that arrive via the matching 514 * interface. Reception of forwarded directed broadcasts would 515 * be handled via ip_forward() and ether_output() with the loopback 516 * into the stack for SIMPLEX interfaces handled by ether_output(). 517 */ 518 if (m->m_pkthdr.rcvif != NULL && 519 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 520 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 521 if (ifa->ifa_addr->sa_family != AF_INET) 522 continue; 523 ia = ifatoia(ifa); 524 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 525 ip->ip_dst.s_addr) 526 goto ours; 527 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) 528 goto ours; 529 #ifdef BOOTP_COMPAT 530 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 531 goto ours; 532 #endif 533 } 534 } 535 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 536 struct in_multi *inm; 537 if (ip_mrouter) { 538 /* 539 * If we are acting as a multicast router, all 540 * incoming multicast packets are passed to the 541 * kernel-level multicast forwarding function. 542 * The packet is returned (relatively) intact; if 543 * ip_mforward() returns a non-zero value, the packet 544 * must be discarded, else it may be accepted below. 545 */ 546 if (ip_mforward && 547 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 548 ipstat.ips_cantforward++; 549 m_freem(m); 550 return; 551 } 552 553 /* 554 * The process-level routing daemon needs to receive 555 * all multicast IGMP packets, whether or not this 556 * host belongs to their destination groups. 557 */ 558 if (ip->ip_p == IPPROTO_IGMP) 559 goto ours; 560 ipstat.ips_forward++; 561 } 562 /* 563 * See if we belong to the destination multicast group on the 564 * arrival interface. 565 */ 566 IN_MULTI_LOCK(); 567 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 568 IN_MULTI_UNLOCK(); 569 if (inm == NULL) { 570 ipstat.ips_notmember++; 571 m_freem(m); 572 return; 573 } 574 goto ours; 575 } 576 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 577 goto ours; 578 if (ip->ip_dst.s_addr == INADDR_ANY) 579 goto ours; 580 581 /* 582 * FAITH(Firewall Aided Internet Translator) 583 */ 584 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 585 if (ip_keepfaith) { 586 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 587 goto ours; 588 } 589 m_freem(m); 590 return; 591 } 592 593 /* 594 * Not for us; forward if possible and desirable. 595 */ 596 if (ipforwarding == 0) { 597 ipstat.ips_cantforward++; 598 m_freem(m); 599 } else { 600 #if defined(IPSEC) || defined(FAST_IPSEC) 601 if (ip_ipsec_fwd(m)) 602 goto bad; 603 #endif /* IPSEC */ 604 ip_forward(m, dchg); 605 } 606 return; 607 608 ours: 609 #ifdef IPSTEALTH 610 /* 611 * IPSTEALTH: Process non-routing options only 612 * if the packet is destined for us. 613 */ 614 if (ipstealth && hlen > sizeof (struct ip) && 615 ip_dooptions(m, 1)) 616 return; 617 #endif /* IPSTEALTH */ 618 619 /* Count the packet in the ip address stats */ 620 if (ia != NULL) { 621 ia->ia_ifa.if_ipackets++; 622 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 623 } 624 625 /* 626 * Attempt reassembly; if it succeeds, proceed. 627 * ip_reass() will return a different mbuf. 628 */ 629 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 630 m = ip_reass(m); 631 if (m == NULL) 632 return; 633 ip = mtod(m, struct ip *); 634 /* Get the header length of the reassembled packet */ 635 hlen = ip->ip_hl << 2; 636 } 637 638 /* 639 * Further protocols expect the packet length to be w/o the 640 * IP header. 641 */ 642 ip->ip_len -= hlen; 643 644 #if defined(IPSEC) || defined(FAST_IPSEC) 645 /* 646 * enforce IPsec policy checking if we are seeing last header. 647 * note that we do not visit this with protocols with pcb layer 648 * code - like udp/tcp/raw ip. 649 */ 650 if (ip_ipsec_input(m)) 651 goto bad; 652 #endif /* IPSEC */ 653 654 /* 655 * Switch out to protocol's input routine. 656 */ 657 ipstat.ips_delivered++; 658 659 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 660 return; 661 bad: 662 m_freem(m); 663 } 664 665 /* 666 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 667 * max has slightly different semantics than the sysctl, for historical 668 * reasons. 669 */ 670 static void 671 maxnipq_update(void) 672 { 673 674 /* 675 * -1 for unlimited allocation. 676 */ 677 if (maxnipq < 0) 678 uma_zone_set_max(ipq_zone, 0); 679 /* 680 * Positive number for specific bound. 681 */ 682 if (maxnipq > 0) 683 uma_zone_set_max(ipq_zone, maxnipq); 684 /* 685 * Zero specifies no further fragment queue allocation -- set the 686 * bound very low, but rely on implementation elsewhere to actually 687 * prevent allocation and reclaim current queues. 688 */ 689 if (maxnipq == 0) 690 uma_zone_set_max(ipq_zone, 1); 691 } 692 693 static void 694 ipq_zone_change(void *tag) 695 { 696 697 if (maxnipq > 0 && maxnipq < (nmbclusters / 32)) { 698 maxnipq = nmbclusters / 32; 699 maxnipq_update(); 700 } 701 } 702 703 static int 704 sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 705 { 706 int error, i; 707 708 i = maxnipq; 709 error = sysctl_handle_int(oidp, &i, 0, req); 710 if (error || !req->newptr) 711 return (error); 712 713 /* 714 * XXXRW: Might be a good idea to sanity check the argument and place 715 * an extreme upper bound. 716 */ 717 if (i < -1) 718 return (EINVAL); 719 maxnipq = i; 720 maxnipq_update(); 721 return (0); 722 } 723 724 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 725 NULL, 0, sysctl_maxnipq, "I", 726 "Maximum number of IPv4 fragment reassembly queue entries"); 727 728 /* 729 * Take incoming datagram fragment and try to reassemble it into 730 * whole datagram. If the argument is the first fragment or one 731 * in between the function will return NULL and store the mbuf 732 * in the fragment chain. If the argument is the last fragment 733 * the packet will be reassembled and the pointer to the new 734 * mbuf returned for further processing. Only m_tags attached 735 * to the first packet/fragment are preserved. 736 * The IP header is *NOT* adjusted out of iplen. 737 */ 738 739 struct mbuf * 740 ip_reass(struct mbuf *m) 741 { 742 struct ip *ip; 743 struct mbuf *p, *q, *nq, *t; 744 struct ipq *fp = NULL; 745 struct ipqhead *head; 746 int i, hlen, next; 747 u_int8_t ecn, ecn0; 748 u_short hash; 749 750 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 751 if (maxnipq == 0 || maxfragsperpacket == 0) { 752 ipstat.ips_fragments++; 753 ipstat.ips_fragdropped++; 754 m_freem(m); 755 return (NULL); 756 } 757 758 ip = mtod(m, struct ip *); 759 hlen = ip->ip_hl << 2; 760 761 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 762 head = &ipq[hash]; 763 IPQ_LOCK(); 764 765 /* 766 * Look for queue of fragments 767 * of this datagram. 768 */ 769 TAILQ_FOREACH(fp, head, ipq_list) 770 if (ip->ip_id == fp->ipq_id && 771 ip->ip_src.s_addr == fp->ipq_src.s_addr && 772 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 773 #ifdef MAC 774 mac_fragment_match(m, fp) && 775 #endif 776 ip->ip_p == fp->ipq_p) 777 goto found; 778 779 fp = NULL; 780 781 /* 782 * Attempt to trim the number of allocated fragment queues if it 783 * exceeds the administrative limit. 784 */ 785 if ((nipq > maxnipq) && (maxnipq > 0)) { 786 /* 787 * drop something from the tail of the current queue 788 * before proceeding further 789 */ 790 struct ipq *q = TAILQ_LAST(head, ipqhead); 791 if (q == NULL) { /* gak */ 792 for (i = 0; i < IPREASS_NHASH; i++) { 793 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 794 if (r) { 795 ipstat.ips_fragtimeout += r->ipq_nfrags; 796 ip_freef(&ipq[i], r); 797 break; 798 } 799 } 800 } else { 801 ipstat.ips_fragtimeout += q->ipq_nfrags; 802 ip_freef(head, q); 803 } 804 } 805 806 found: 807 /* 808 * Adjust ip_len to not reflect header, 809 * convert offset of this to bytes. 810 */ 811 ip->ip_len -= hlen; 812 if (ip->ip_off & IP_MF) { 813 /* 814 * Make sure that fragments have a data length 815 * that's a non-zero multiple of 8 bytes. 816 */ 817 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 818 ipstat.ips_toosmall++; /* XXX */ 819 goto dropfrag; 820 } 821 m->m_flags |= M_FRAG; 822 } else 823 m->m_flags &= ~M_FRAG; 824 ip->ip_off <<= 3; 825 826 827 /* 828 * Attempt reassembly; if it succeeds, proceed. 829 * ip_reass() will return a different mbuf. 830 */ 831 ipstat.ips_fragments++; 832 m->m_pkthdr.header = ip; 833 834 /* Previous ip_reass() started here. */ 835 /* 836 * Presence of header sizes in mbufs 837 * would confuse code below. 838 */ 839 m->m_data += hlen; 840 m->m_len -= hlen; 841 842 /* 843 * If first fragment to arrive, create a reassembly queue. 844 */ 845 if (fp == NULL) { 846 fp = uma_zalloc(ipq_zone, M_NOWAIT); 847 if (fp == NULL) 848 goto dropfrag; 849 #ifdef MAC 850 if (mac_init_ipq(fp, M_NOWAIT) != 0) { 851 uma_zfree(ipq_zone, fp); 852 fp = NULL; 853 goto dropfrag; 854 } 855 mac_create_ipq(m, fp); 856 #endif 857 TAILQ_INSERT_HEAD(head, fp, ipq_list); 858 nipq++; 859 fp->ipq_nfrags = 1; 860 fp->ipq_ttl = IPFRAGTTL; 861 fp->ipq_p = ip->ip_p; 862 fp->ipq_id = ip->ip_id; 863 fp->ipq_src = ip->ip_src; 864 fp->ipq_dst = ip->ip_dst; 865 fp->ipq_frags = m; 866 m->m_nextpkt = NULL; 867 goto done; 868 } else { 869 fp->ipq_nfrags++; 870 #ifdef MAC 871 mac_update_ipq(m, fp); 872 #endif 873 } 874 875 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 876 877 /* 878 * Handle ECN by comparing this segment with the first one; 879 * if CE is set, do not lose CE. 880 * drop if CE and not-ECT are mixed for the same packet. 881 */ 882 ecn = ip->ip_tos & IPTOS_ECN_MASK; 883 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 884 if (ecn == IPTOS_ECN_CE) { 885 if (ecn0 == IPTOS_ECN_NOTECT) 886 goto dropfrag; 887 if (ecn0 != IPTOS_ECN_CE) 888 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 889 } 890 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 891 goto dropfrag; 892 893 /* 894 * Find a segment which begins after this one does. 895 */ 896 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 897 if (GETIP(q)->ip_off > ip->ip_off) 898 break; 899 900 /* 901 * If there is a preceding segment, it may provide some of 902 * our data already. If so, drop the data from the incoming 903 * segment. If it provides all of our data, drop us, otherwise 904 * stick new segment in the proper place. 905 * 906 * If some of the data is dropped from the the preceding 907 * segment, then it's checksum is invalidated. 908 */ 909 if (p) { 910 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 911 if (i > 0) { 912 if (i >= ip->ip_len) 913 goto dropfrag; 914 m_adj(m, i); 915 m->m_pkthdr.csum_flags = 0; 916 ip->ip_off += i; 917 ip->ip_len -= i; 918 } 919 m->m_nextpkt = p->m_nextpkt; 920 p->m_nextpkt = m; 921 } else { 922 m->m_nextpkt = fp->ipq_frags; 923 fp->ipq_frags = m; 924 } 925 926 /* 927 * While we overlap succeeding segments trim them or, 928 * if they are completely covered, dequeue them. 929 */ 930 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 931 q = nq) { 932 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 933 if (i < GETIP(q)->ip_len) { 934 GETIP(q)->ip_len -= i; 935 GETIP(q)->ip_off += i; 936 m_adj(q, i); 937 q->m_pkthdr.csum_flags = 0; 938 break; 939 } 940 nq = q->m_nextpkt; 941 m->m_nextpkt = nq; 942 ipstat.ips_fragdropped++; 943 fp->ipq_nfrags--; 944 m_freem(q); 945 } 946 947 /* 948 * Check for complete reassembly and perform frag per packet 949 * limiting. 950 * 951 * Frag limiting is performed here so that the nth frag has 952 * a chance to complete the packet before we drop the packet. 953 * As a result, n+1 frags are actually allowed per packet, but 954 * only n will ever be stored. (n = maxfragsperpacket.) 955 * 956 */ 957 next = 0; 958 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 959 if (GETIP(q)->ip_off != next) { 960 if (fp->ipq_nfrags > maxfragsperpacket) { 961 ipstat.ips_fragdropped += fp->ipq_nfrags; 962 ip_freef(head, fp); 963 } 964 goto done; 965 } 966 next += GETIP(q)->ip_len; 967 } 968 /* Make sure the last packet didn't have the IP_MF flag */ 969 if (p->m_flags & M_FRAG) { 970 if (fp->ipq_nfrags > maxfragsperpacket) { 971 ipstat.ips_fragdropped += fp->ipq_nfrags; 972 ip_freef(head, fp); 973 } 974 goto done; 975 } 976 977 /* 978 * Reassembly is complete. Make sure the packet is a sane size. 979 */ 980 q = fp->ipq_frags; 981 ip = GETIP(q); 982 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 983 ipstat.ips_toolong++; 984 ipstat.ips_fragdropped += fp->ipq_nfrags; 985 ip_freef(head, fp); 986 goto done; 987 } 988 989 /* 990 * Concatenate fragments. 991 */ 992 m = q; 993 t = m->m_next; 994 m->m_next = NULL; 995 m_cat(m, t); 996 nq = q->m_nextpkt; 997 q->m_nextpkt = NULL; 998 for (q = nq; q != NULL; q = nq) { 999 nq = q->m_nextpkt; 1000 q->m_nextpkt = NULL; 1001 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1002 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1003 m_cat(m, q); 1004 } 1005 /* 1006 * In order to do checksumming faster we do 'end-around carry' here 1007 * (and not in for{} loop), though it implies we are not going to 1008 * reassemble more than 64k fragments. 1009 */ 1010 m->m_pkthdr.csum_data = 1011 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1012 #ifdef MAC 1013 mac_create_datagram_from_ipq(fp, m); 1014 mac_destroy_ipq(fp); 1015 #endif 1016 1017 /* 1018 * Create header for new ip packet by modifying header of first 1019 * packet; dequeue and discard fragment reassembly header. 1020 * Make header visible. 1021 */ 1022 ip->ip_len = (ip->ip_hl << 2) + next; 1023 ip->ip_src = fp->ipq_src; 1024 ip->ip_dst = fp->ipq_dst; 1025 TAILQ_REMOVE(head, fp, ipq_list); 1026 nipq--; 1027 uma_zfree(ipq_zone, fp); 1028 m->m_len += (ip->ip_hl << 2); 1029 m->m_data -= (ip->ip_hl << 2); 1030 /* some debugging cruft by sklower, below, will go away soon */ 1031 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1032 m_fixhdr(m); 1033 ipstat.ips_reassembled++; 1034 IPQ_UNLOCK(); 1035 return (m); 1036 1037 dropfrag: 1038 ipstat.ips_fragdropped++; 1039 if (fp != NULL) 1040 fp->ipq_nfrags--; 1041 m_freem(m); 1042 done: 1043 IPQ_UNLOCK(); 1044 return (NULL); 1045 1046 #undef GETIP 1047 } 1048 1049 /* 1050 * Free a fragment reassembly header and all 1051 * associated datagrams. 1052 */ 1053 static void 1054 ip_freef(fhp, fp) 1055 struct ipqhead *fhp; 1056 struct ipq *fp; 1057 { 1058 register struct mbuf *q; 1059 1060 IPQ_LOCK_ASSERT(); 1061 1062 while (fp->ipq_frags) { 1063 q = fp->ipq_frags; 1064 fp->ipq_frags = q->m_nextpkt; 1065 m_freem(q); 1066 } 1067 TAILQ_REMOVE(fhp, fp, ipq_list); 1068 uma_zfree(ipq_zone, fp); 1069 nipq--; 1070 } 1071 1072 /* 1073 * IP timer processing; 1074 * if a timer expires on a reassembly 1075 * queue, discard it. 1076 */ 1077 void 1078 ip_slowtimo() 1079 { 1080 register struct ipq *fp; 1081 int i; 1082 1083 IPQ_LOCK(); 1084 for (i = 0; i < IPREASS_NHASH; i++) { 1085 for(fp = TAILQ_FIRST(&ipq[i]); fp;) { 1086 struct ipq *fpp; 1087 1088 fpp = fp; 1089 fp = TAILQ_NEXT(fp, ipq_list); 1090 if(--fpp->ipq_ttl == 0) { 1091 ipstat.ips_fragtimeout += fpp->ipq_nfrags; 1092 ip_freef(&ipq[i], fpp); 1093 } 1094 } 1095 } 1096 /* 1097 * If we are over the maximum number of fragments 1098 * (due to the limit being lowered), drain off 1099 * enough to get down to the new limit. 1100 */ 1101 if (maxnipq >= 0 && nipq > maxnipq) { 1102 for (i = 0; i < IPREASS_NHASH; i++) { 1103 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) { 1104 ipstat.ips_fragdropped += 1105 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1106 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1107 } 1108 } 1109 } 1110 IPQ_UNLOCK(); 1111 } 1112 1113 /* 1114 * Drain off all datagram fragments. 1115 */ 1116 void 1117 ip_drain() 1118 { 1119 int i; 1120 1121 IPQ_LOCK(); 1122 for (i = 0; i < IPREASS_NHASH; i++) { 1123 while(!TAILQ_EMPTY(&ipq[i])) { 1124 ipstat.ips_fragdropped += 1125 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1126 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1127 } 1128 } 1129 IPQ_UNLOCK(); 1130 in_rtqdrain(); 1131 } 1132 1133 /* 1134 * The protocol to be inserted into ip_protox[] must be already registered 1135 * in inetsw[], either statically or through pf_proto_register(). 1136 */ 1137 int 1138 ipproto_register(u_char ipproto) 1139 { 1140 struct protosw *pr; 1141 1142 /* Sanity checks. */ 1143 if (ipproto == 0) 1144 return (EPROTONOSUPPORT); 1145 1146 /* 1147 * The protocol slot must not be occupied by another protocol 1148 * already. An index pointing to IPPROTO_RAW is unused. 1149 */ 1150 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1151 if (pr == NULL) 1152 return (EPFNOSUPPORT); 1153 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1154 return (EEXIST); 1155 1156 /* Find the protocol position in inetsw[] and set the index. */ 1157 for (pr = inetdomain.dom_protosw; 1158 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1159 if (pr->pr_domain->dom_family == PF_INET && 1160 pr->pr_protocol && pr->pr_protocol == ipproto) { 1161 /* Be careful to only index valid IP protocols. */ 1162 if (pr->pr_protocol < IPPROTO_MAX) { 1163 ip_protox[pr->pr_protocol] = pr - inetsw; 1164 return (0); 1165 } else 1166 return (EINVAL); 1167 } 1168 } 1169 return (EPROTONOSUPPORT); 1170 } 1171 1172 int 1173 ipproto_unregister(u_char ipproto) 1174 { 1175 struct protosw *pr; 1176 1177 /* Sanity checks. */ 1178 if (ipproto == 0) 1179 return (EPROTONOSUPPORT); 1180 1181 /* Check if the protocol was indeed registered. */ 1182 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1183 if (pr == NULL) 1184 return (EPFNOSUPPORT); 1185 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1186 return (ENOENT); 1187 1188 /* Reset the protocol slot to IPPROTO_RAW. */ 1189 ip_protox[ipproto] = pr - inetsw; 1190 return (0); 1191 } 1192 1193 /* 1194 * Given address of next destination (final or next hop), 1195 * return internet address info of interface to be used to get there. 1196 */ 1197 struct in_ifaddr * 1198 ip_rtaddr(dst) 1199 struct in_addr dst; 1200 { 1201 struct route sro; 1202 struct sockaddr_in *sin; 1203 struct in_ifaddr *ifa; 1204 1205 bzero(&sro, sizeof(sro)); 1206 sin = (struct sockaddr_in *)&sro.ro_dst; 1207 sin->sin_family = AF_INET; 1208 sin->sin_len = sizeof(*sin); 1209 sin->sin_addr = dst; 1210 rtalloc_ign(&sro, RTF_CLONING); 1211 1212 if (sro.ro_rt == NULL) 1213 return (NULL); 1214 1215 ifa = ifatoia(sro.ro_rt->rt_ifa); 1216 RTFREE(sro.ro_rt); 1217 return (ifa); 1218 } 1219 1220 u_char inetctlerrmap[PRC_NCMDS] = { 1221 0, 0, 0, 0, 1222 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1223 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1224 EMSGSIZE, EHOSTUNREACH, 0, 0, 1225 0, 0, EHOSTUNREACH, 0, 1226 ENOPROTOOPT, ECONNREFUSED 1227 }; 1228 1229 /* 1230 * Forward a packet. If some error occurs return the sender 1231 * an icmp packet. Note we can't always generate a meaningful 1232 * icmp message because icmp doesn't have a large enough repertoire 1233 * of codes and types. 1234 * 1235 * If not forwarding, just drop the packet. This could be confusing 1236 * if ipforwarding was zero but some routing protocol was advancing 1237 * us as a gateway to somewhere. However, we must let the routing 1238 * protocol deal with that. 1239 * 1240 * The srcrt parameter indicates whether the packet is being forwarded 1241 * via a source route. 1242 */ 1243 void 1244 ip_forward(struct mbuf *m, int srcrt) 1245 { 1246 struct ip *ip = mtod(m, struct ip *); 1247 struct in_ifaddr *ia = NULL; 1248 struct mbuf *mcopy; 1249 struct in_addr dest; 1250 int error, type = 0, code = 0, mtu = 0; 1251 1252 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1253 ipstat.ips_cantforward++; 1254 m_freem(m); 1255 return; 1256 } 1257 #ifdef IPSTEALTH 1258 if (!ipstealth) { 1259 #endif 1260 if (ip->ip_ttl <= IPTTLDEC) { 1261 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1262 0, 0); 1263 return; 1264 } 1265 #ifdef IPSTEALTH 1266 } 1267 #endif 1268 1269 if (!srcrt && (ia = ip_rtaddr(ip->ip_dst)) == NULL) { 1270 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1271 return; 1272 } 1273 1274 /* 1275 * Save the IP header and at most 8 bytes of the payload, 1276 * in case we need to generate an ICMP message to the src. 1277 * 1278 * XXX this can be optimized a lot by saving the data in a local 1279 * buffer on the stack (72 bytes at most), and only allocating the 1280 * mbuf if really necessary. The vast majority of the packets 1281 * are forwarded without having to send an ICMP back (either 1282 * because unnecessary, or because rate limited), so we are 1283 * really we are wasting a lot of work here. 1284 * 1285 * We don't use m_copy() because it might return a reference 1286 * to a shared cluster. Both this function and ip_output() 1287 * assume exclusive access to the IP header in `m', so any 1288 * data in a cluster may change before we reach icmp_error(). 1289 */ 1290 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1291 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1292 /* 1293 * It's probably ok if the pkthdr dup fails (because 1294 * the deep copy of the tag chain failed), but for now 1295 * be conservative and just discard the copy since 1296 * code below may some day want the tags. 1297 */ 1298 m_free(mcopy); 1299 mcopy = NULL; 1300 } 1301 if (mcopy != NULL) { 1302 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy)); 1303 mcopy->m_pkthdr.len = mcopy->m_len; 1304 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1305 } 1306 1307 #ifdef IPSTEALTH 1308 if (!ipstealth) { 1309 #endif 1310 ip->ip_ttl -= IPTTLDEC; 1311 #ifdef IPSTEALTH 1312 } 1313 #endif 1314 1315 /* 1316 * If forwarding packet using same interface that it came in on, 1317 * perhaps should send a redirect to sender to shortcut a hop. 1318 * Only send redirect if source is sending directly to us, 1319 * and if packet was not source routed (or has any options). 1320 * Also, don't send redirect if forwarding using a default route 1321 * or a route modified by a redirect. 1322 */ 1323 dest.s_addr = 0; 1324 if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) { 1325 struct sockaddr_in *sin; 1326 struct route ro; 1327 struct rtentry *rt; 1328 1329 bzero(&ro, sizeof(ro)); 1330 sin = (struct sockaddr_in *)&ro.ro_dst; 1331 sin->sin_family = AF_INET; 1332 sin->sin_len = sizeof(*sin); 1333 sin->sin_addr = ip->ip_dst; 1334 rtalloc_ign(&ro, RTF_CLONING); 1335 1336 rt = ro.ro_rt; 1337 1338 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1339 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1340 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1341 u_long src = ntohl(ip->ip_src.s_addr); 1342 1343 if (RTA(rt) && 1344 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1345 if (rt->rt_flags & RTF_GATEWAY) 1346 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1347 else 1348 dest.s_addr = ip->ip_dst.s_addr; 1349 /* Router requirements says to only send host redirects */ 1350 type = ICMP_REDIRECT; 1351 code = ICMP_REDIRECT_HOST; 1352 } 1353 } 1354 if (rt) 1355 RTFREE(rt); 1356 } 1357 1358 error = ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL); 1359 if (error) 1360 ipstat.ips_cantforward++; 1361 else { 1362 ipstat.ips_forward++; 1363 if (type) 1364 ipstat.ips_redirectsent++; 1365 else { 1366 if (mcopy) 1367 m_freem(mcopy); 1368 return; 1369 } 1370 } 1371 if (mcopy == NULL) 1372 return; 1373 1374 switch (error) { 1375 1376 case 0: /* forwarded, but need redirect */ 1377 /* type, code set above */ 1378 break; 1379 1380 case ENETUNREACH: /* shouldn't happen, checked above */ 1381 case EHOSTUNREACH: 1382 case ENETDOWN: 1383 case EHOSTDOWN: 1384 default: 1385 type = ICMP_UNREACH; 1386 code = ICMP_UNREACH_HOST; 1387 break; 1388 1389 case EMSGSIZE: 1390 type = ICMP_UNREACH; 1391 code = ICMP_UNREACH_NEEDFRAG; 1392 1393 #if defined(IPSEC) || defined(FAST_IPSEC) 1394 mtu = ip_ipsec_mtu(m); 1395 #endif /* IPSEC */ 1396 /* 1397 * If the MTU wasn't set before use the interface mtu or 1398 * fall back to the next smaller mtu step compared to the 1399 * current packet size. 1400 */ 1401 if (mtu == 0) { 1402 if (ia != NULL) 1403 mtu = ia->ia_ifp->if_mtu; 1404 else 1405 mtu = ip_next_mtu(ip->ip_len, 0); 1406 } 1407 ipstat.ips_cantfrag++; 1408 break; 1409 1410 case ENOBUFS: 1411 /* 1412 * A router should not generate ICMP_SOURCEQUENCH as 1413 * required in RFC1812 Requirements for IP Version 4 Routers. 1414 * Source quench could be a big problem under DoS attacks, 1415 * or if the underlying interface is rate-limited. 1416 * Those who need source quench packets may re-enable them 1417 * via the net.inet.ip.sendsourcequench sysctl. 1418 */ 1419 if (ip_sendsourcequench == 0) { 1420 m_freem(mcopy); 1421 return; 1422 } else { 1423 type = ICMP_SOURCEQUENCH; 1424 code = 0; 1425 } 1426 break; 1427 1428 case EACCES: /* ipfw denied packet */ 1429 m_freem(mcopy); 1430 return; 1431 } 1432 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1433 } 1434 1435 void 1436 ip_savecontrol(inp, mp, ip, m) 1437 register struct inpcb *inp; 1438 register struct mbuf **mp; 1439 register struct ip *ip; 1440 register struct mbuf *m; 1441 { 1442 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1443 struct bintime bt; 1444 1445 bintime(&bt); 1446 if (inp->inp_socket->so_options & SO_BINTIME) { 1447 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1448 SCM_BINTIME, SOL_SOCKET); 1449 if (*mp) 1450 mp = &(*mp)->m_next; 1451 } 1452 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1453 struct timeval tv; 1454 1455 bintime2timeval(&bt, &tv); 1456 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1457 SCM_TIMESTAMP, SOL_SOCKET); 1458 if (*mp) 1459 mp = &(*mp)->m_next; 1460 } 1461 } 1462 if (inp->inp_flags & INP_RECVDSTADDR) { 1463 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1464 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1465 if (*mp) 1466 mp = &(*mp)->m_next; 1467 } 1468 if (inp->inp_flags & INP_RECVTTL) { 1469 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1470 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1471 if (*mp) 1472 mp = &(*mp)->m_next; 1473 } 1474 #ifdef notyet 1475 /* XXX 1476 * Moving these out of udp_input() made them even more broken 1477 * than they already were. 1478 */ 1479 /* options were tossed already */ 1480 if (inp->inp_flags & INP_RECVOPTS) { 1481 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1482 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1483 if (*mp) 1484 mp = &(*mp)->m_next; 1485 } 1486 /* ip_srcroute doesn't do what we want here, need to fix */ 1487 if (inp->inp_flags & INP_RECVRETOPTS) { 1488 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1489 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1490 if (*mp) 1491 mp = &(*mp)->m_next; 1492 } 1493 #endif 1494 if (inp->inp_flags & INP_RECVIF) { 1495 struct ifnet *ifp; 1496 struct sdlbuf { 1497 struct sockaddr_dl sdl; 1498 u_char pad[32]; 1499 } sdlbuf; 1500 struct sockaddr_dl *sdp; 1501 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1502 1503 if (((ifp = m->m_pkthdr.rcvif)) 1504 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1505 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1506 /* 1507 * Change our mind and don't try copy. 1508 */ 1509 if ((sdp->sdl_family != AF_LINK) 1510 || (sdp->sdl_len > sizeof(sdlbuf))) { 1511 goto makedummy; 1512 } 1513 bcopy(sdp, sdl2, sdp->sdl_len); 1514 } else { 1515 makedummy: 1516 sdl2->sdl_len 1517 = offsetof(struct sockaddr_dl, sdl_data[0]); 1518 sdl2->sdl_family = AF_LINK; 1519 sdl2->sdl_index = 0; 1520 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1521 } 1522 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1523 IP_RECVIF, IPPROTO_IP); 1524 if (*mp) 1525 mp = &(*mp)->m_next; 1526 } 1527 } 1528 1529 /* 1530 * XXX these routines are called from the upper part of the kernel. 1531 * They need to be locked when we remove Giant. 1532 * 1533 * They could also be moved to ip_mroute.c, since all the RSVP 1534 * handling is done there already. 1535 */ 1536 static int ip_rsvp_on; 1537 struct socket *ip_rsvpd; 1538 int 1539 ip_rsvp_init(struct socket *so) 1540 { 1541 if (so->so_type != SOCK_RAW || 1542 so->so_proto->pr_protocol != IPPROTO_RSVP) 1543 return EOPNOTSUPP; 1544 1545 if (ip_rsvpd != NULL) 1546 return EADDRINUSE; 1547 1548 ip_rsvpd = so; 1549 /* 1550 * This may seem silly, but we need to be sure we don't over-increment 1551 * the RSVP counter, in case something slips up. 1552 */ 1553 if (!ip_rsvp_on) { 1554 ip_rsvp_on = 1; 1555 rsvp_on++; 1556 } 1557 1558 return 0; 1559 } 1560 1561 int 1562 ip_rsvp_done(void) 1563 { 1564 ip_rsvpd = NULL; 1565 /* 1566 * This may seem silly, but we need to be sure we don't over-decrement 1567 * the RSVP counter, in case something slips up. 1568 */ 1569 if (ip_rsvp_on) { 1570 ip_rsvp_on = 0; 1571 rsvp_on--; 1572 } 1573 return 0; 1574 } 1575 1576 void 1577 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1578 { 1579 if (rsvp_input_p) { /* call the real one if loaded */ 1580 rsvp_input_p(m, off); 1581 return; 1582 } 1583 1584 /* Can still get packets with rsvp_on = 0 if there is a local member 1585 * of the group to which the RSVP packet is addressed. But in this 1586 * case we want to throw the packet away. 1587 */ 1588 1589 if (!rsvp_on) { 1590 m_freem(m); 1591 return; 1592 } 1593 1594 if (ip_rsvpd != NULL) { 1595 rip_input(m, off); 1596 return; 1597 } 1598 /* Drop the packet */ 1599 m_freem(m); 1600 } 1601