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