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 75 #include <sys/socketvar.h> 76 77 /* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */ 78 #include <netinet/ip_fw.h> 79 #include <netinet/ip_dummynet.h> 80 81 #ifdef IPSEC 82 #include <netinet6/ipsec.h> 83 #include <netkey/key.h> 84 #endif 85 86 #ifdef FAST_IPSEC 87 #include <netipsec/ipsec.h> 88 #include <netipsec/key.h> 89 #endif 90 91 int rsvp_on = 0; 92 93 int ipforwarding = 0; 94 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 95 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 96 97 static int ipsendredirects = 1; /* XXX */ 98 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 99 &ipsendredirects, 0, "Enable sending IP redirects"); 100 101 int ip_defttl = IPDEFTTL; 102 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 103 &ip_defttl, 0, "Maximum TTL on IP packets"); 104 105 static int ip_keepfaith = 0; 106 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 107 &ip_keepfaith, 0, 108 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 109 110 static int ip_sendsourcequench = 0; 111 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 112 &ip_sendsourcequench, 0, 113 "Enable the transmission of source quench packets"); 114 115 int ip_do_randomid = 0; 116 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 117 &ip_do_randomid, 0, 118 "Assign random ip_id values"); 119 120 /* 121 * XXX - Setting ip_checkinterface mostly implements the receive side of 122 * the Strong ES model described in RFC 1122, but since the routing table 123 * and transmit implementation do not implement the Strong ES model, 124 * setting this to 1 results in an odd hybrid. 125 * 126 * XXX - ip_checkinterface currently must be disabled if you use ipnat 127 * to translate the destination address to another local interface. 128 * 129 * XXX - ip_checkinterface must be disabled if you add IP aliases 130 * to the loopback interface instead of the interface where the 131 * packets for those addresses are received. 132 */ 133 static int ip_checkinterface = 0; 134 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 135 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 136 137 struct pfil_head inet_pfil_hook; /* Packet filter hooks */ 138 139 static struct ifqueue ipintrq; 140 static int ipqmaxlen = IFQ_MAXLEN; 141 142 extern struct domain inetdomain; 143 extern struct protosw inetsw[]; 144 u_char ip_protox[IPPROTO_MAX]; 145 struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 146 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 147 u_long in_ifaddrhmask; /* mask for hash table */ 148 149 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 150 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 151 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 152 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 153 154 struct ipstat ipstat; 155 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 156 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 157 158 /* 159 * IP datagram reassembly. 160 */ 161 #define IPREASS_NHASH_LOG2 6 162 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 163 #define IPREASS_HMASK (IPREASS_NHASH - 1) 164 #define IPREASS_HASH(x,y) \ 165 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 166 167 static uma_zone_t ipq_zone; 168 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 169 static struct mtx ipqlock; 170 171 #define IPQ_LOCK() mtx_lock(&ipqlock) 172 #define IPQ_UNLOCK() mtx_unlock(&ipqlock) 173 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 174 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 175 176 static void maxnipq_update(void); 177 178 static int maxnipq; /* Administrative limit on # reass queues. */ 179 static int nipq = 0; /* Total # of reass queues */ 180 SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, &nipq, 0, 181 "Current number of IPv4 fragment reassembly queue entries"); 182 183 static int maxfragsperpacket; 184 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 185 &maxfragsperpacket, 0, 186 "Maximum number of IPv4 fragments allowed per packet"); 187 188 struct callout ipport_tick_callout; 189 190 #ifdef IPCTL_DEFMTU 191 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 192 &ip_mtu, 0, "Default MTU"); 193 #endif 194 195 #ifdef IPSTEALTH 196 int ipstealth = 0; 197 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 198 &ipstealth, 0, ""); 199 #endif 200 201 /* 202 * ipfw_ether and ipfw_bridge hooks. 203 * XXX: Temporary until those are converted to pfil_hooks as well. 204 */ 205 ip_fw_chk_t *ip_fw_chk_ptr = NULL; 206 ip_dn_io_t *ip_dn_io_ptr = NULL; 207 int fw_enable = 1; 208 int fw_one_pass = 1; 209 210 static void ip_freef(struct ipqhead *, struct ipq *); 211 212 /* 213 * IP initialization: fill in IP protocol switch table. 214 * All protocols not implemented in kernel go to raw IP protocol handler. 215 */ 216 void 217 ip_init() 218 { 219 register struct protosw *pr; 220 register int i; 221 222 TAILQ_INIT(&in_ifaddrhead); 223 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 224 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 225 if (pr == NULL) 226 panic("ip_init: PF_INET not found"); 227 228 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 229 for (i = 0; i < IPPROTO_MAX; i++) 230 ip_protox[i] = pr - inetsw; 231 /* 232 * Cycle through IP protocols and put them into the appropriate place 233 * in ip_protox[]. 234 */ 235 for (pr = inetdomain.dom_protosw; 236 pr < inetdomain.dom_protoswNPROTOSW; pr++) 237 if (pr->pr_domain->dom_family == PF_INET && 238 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 239 /* Be careful to only index valid IP protocols. */ 240 if (pr->pr_protocol < IPPROTO_MAX) 241 ip_protox[pr->pr_protocol] = pr - inetsw; 242 } 243 244 /* Initialize packet filter hooks. */ 245 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 246 inet_pfil_hook.ph_af = AF_INET; 247 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 248 printf("%s: WARNING: unable to register pfil hook, " 249 "error %d\n", __func__, i); 250 251 /* Initialize IP reassembly queue. */ 252 IPQ_LOCK_INIT(); 253 for (i = 0; i < IPREASS_NHASH; i++) 254 TAILQ_INIT(&ipq[i]); 255 maxnipq = nmbclusters / 32; 256 maxfragsperpacket = 16; 257 ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 258 NULL, UMA_ALIGN_PTR, 0); 259 maxnipq_update(); 260 261 /* Start ipport_tick. */ 262 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE); 263 ipport_tick(NULL); 264 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, 265 SHUTDOWN_PRI_DEFAULT); 266 267 /* Initialize various other remaining things. */ 268 ip_id = time_second & 0xffff; 269 ipintrq.ifq_maxlen = ipqmaxlen; 270 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF); 271 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE); 272 } 273 274 void ip_fini(xtp) 275 void *xtp; 276 { 277 callout_stop(&ipport_tick_callout); 278 } 279 280 /* 281 * Ip input routine. Checksum and byte swap header. If fragmented 282 * try to reassemble. Process options. Pass to next level. 283 */ 284 void 285 ip_input(struct mbuf *m) 286 { 287 struct ip *ip = NULL; 288 struct in_ifaddr *ia = NULL; 289 struct ifaddr *ifa; 290 int checkif, hlen = 0; 291 u_short sum; 292 int dchg = 0; /* dest changed after fw */ 293 struct in_addr odst; /* original dst address */ 294 #ifdef FAST_IPSEC 295 struct m_tag *mtag; 296 struct tdb_ident *tdbi; 297 struct secpolicy *sp; 298 int s, error; 299 #endif /* FAST_IPSEC */ 300 301 M_ASSERTPKTHDR(m); 302 303 if (m->m_flags & M_FASTFWD_OURS) { 304 /* 305 * Firewall or NAT changed destination to local. 306 * We expect ip_len and ip_off to be in host byte order. 307 */ 308 m->m_flags &= ~M_FASTFWD_OURS; 309 /* Set up some basics that will be used later. */ 310 ip = mtod(m, struct ip *); 311 hlen = ip->ip_hl << 2; 312 goto ours; 313 } 314 315 ipstat.ips_total++; 316 317 if (m->m_pkthdr.len < sizeof(struct ip)) 318 goto tooshort; 319 320 if (m->m_len < sizeof (struct ip) && 321 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 322 ipstat.ips_toosmall++; 323 return; 324 } 325 ip = mtod(m, struct ip *); 326 327 if (ip->ip_v != IPVERSION) { 328 ipstat.ips_badvers++; 329 goto bad; 330 } 331 332 hlen = ip->ip_hl << 2; 333 if (hlen < sizeof(struct ip)) { /* minimum header length */ 334 ipstat.ips_badhlen++; 335 goto bad; 336 } 337 if (hlen > m->m_len) { 338 if ((m = m_pullup(m, hlen)) == NULL) { 339 ipstat.ips_badhlen++; 340 return; 341 } 342 ip = mtod(m, struct ip *); 343 } 344 345 /* 127/8 must not appear on wire - RFC1122 */ 346 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 347 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 348 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 349 ipstat.ips_badaddr++; 350 goto bad; 351 } 352 } 353 354 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 355 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 356 } else { 357 if (hlen == sizeof(struct ip)) { 358 sum = in_cksum_hdr(ip); 359 } else { 360 sum = in_cksum(m, hlen); 361 } 362 } 363 if (sum) { 364 ipstat.ips_badsum++; 365 goto bad; 366 } 367 368 #ifdef ALTQ 369 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 370 /* packet is dropped by traffic conditioner */ 371 return; 372 #endif 373 374 /* 375 * Convert fields to host representation. 376 */ 377 ip->ip_len = ntohs(ip->ip_len); 378 if (ip->ip_len < hlen) { 379 ipstat.ips_badlen++; 380 goto bad; 381 } 382 ip->ip_off = ntohs(ip->ip_off); 383 384 /* 385 * Check that the amount of data in the buffers 386 * is as at least much as the IP header would have us expect. 387 * Trim mbufs if longer than we expect. 388 * Drop packet if shorter than we expect. 389 */ 390 if (m->m_pkthdr.len < ip->ip_len) { 391 tooshort: 392 ipstat.ips_tooshort++; 393 goto bad; 394 } 395 if (m->m_pkthdr.len > ip->ip_len) { 396 if (m->m_len == m->m_pkthdr.len) { 397 m->m_len = ip->ip_len; 398 m->m_pkthdr.len = ip->ip_len; 399 } else 400 m_adj(m, ip->ip_len - m->m_pkthdr.len); 401 } 402 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF) 403 /* 404 * Bypass packet filtering for packets from a tunnel (gif). 405 */ 406 if (ipsec_getnhist(m)) 407 goto passin; 408 #endif 409 #if defined(FAST_IPSEC) && !defined(IPSEC_FILTERGIF) 410 /* 411 * Bypass packet filtering for packets from a tunnel (gif). 412 */ 413 if (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL) 414 goto passin; 415 #endif 416 417 /* 418 * Run through list of hooks for input packets. 419 * 420 * NB: Beware of the destination address changing (e.g. 421 * by NAT rewriting). When this happens, tell 422 * ip_forward to do the right thing. 423 */ 424 425 /* Jump over all PFIL processing if hooks are not active. */ 426 if (inet_pfil_hook.ph_busy_count == -1) 427 goto passin; 428 429 odst = ip->ip_dst; 430 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 431 PFIL_IN, NULL) != 0) 432 return; 433 if (m == NULL) /* consumed by filter */ 434 return; 435 436 ip = mtod(m, struct ip *); 437 dchg = (odst.s_addr != ip->ip_dst.s_addr); 438 439 #ifdef IPFIREWALL_FORWARD 440 if (m->m_flags & M_FASTFWD_OURS) { 441 m->m_flags &= ~M_FASTFWD_OURS; 442 goto ours; 443 } 444 #ifndef IPFIREWALL_FORWARD_EXTENDED 445 dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL); 446 #else 447 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 448 /* 449 * Directly ship on the packet. This allows to forward packets 450 * that were destined for us to some other directly connected 451 * host. 452 */ 453 ip_forward(m, dchg); 454 return; 455 } 456 #endif /* IPFIREWALL_FORWARD_EXTENDED */ 457 #endif /* IPFIREWALL_FORWARD */ 458 459 passin: 460 /* 461 * Process options and, if not destined for us, 462 * ship it on. ip_dooptions returns 1 when an 463 * error was detected (causing an icmp message 464 * to be sent and the original packet to be freed). 465 */ 466 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 467 return; 468 469 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 470 * matter if it is destined to another node, or whether it is 471 * a multicast one, RSVP wants it! and prevents it from being forwarded 472 * anywhere else. Also checks if the rsvp daemon is running before 473 * grabbing the packet. 474 */ 475 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 476 goto ours; 477 478 /* 479 * Check our list of addresses, to see if the packet is for us. 480 * If we don't have any addresses, assume any unicast packet 481 * we receive might be for us (and let the upper layers deal 482 * with it). 483 */ 484 if (TAILQ_EMPTY(&in_ifaddrhead) && 485 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 486 goto ours; 487 488 /* 489 * Enable a consistency check between the destination address 490 * and the arrival interface for a unicast packet (the RFC 1122 491 * strong ES model) if IP forwarding is disabled and the packet 492 * is not locally generated and the packet is not subject to 493 * 'ipfw fwd'. 494 * 495 * XXX - Checking also should be disabled if the destination 496 * address is ipnat'ed to a different interface. 497 * 498 * XXX - Checking is incompatible with IP aliases added 499 * to the loopback interface instead of the interface where 500 * the packets are received. 501 * 502 * XXX - This is the case for carp vhost IPs as well so we 503 * insert a workaround. If the packet got here, we already 504 * checked with carp_iamatch() and carp_forus(). 505 */ 506 checkif = ip_checkinterface && (ipforwarding == 0) && 507 m->m_pkthdr.rcvif != NULL && 508 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 509 #ifdef DEV_CARP 510 !m->m_pkthdr.rcvif->if_carp && 511 #endif 512 (dchg == 0); 513 514 /* 515 * Check for exact addresses in the hash bucket. 516 */ 517 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 518 /* 519 * If the address matches, verify that the packet 520 * arrived via the correct interface if checking is 521 * enabled. 522 */ 523 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 524 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 525 goto ours; 526 } 527 /* 528 * Check for broadcast addresses. 529 * 530 * Only accept broadcast packets that arrive via the matching 531 * interface. Reception of forwarded directed broadcasts would 532 * be handled via ip_forward() and ether_output() with the loopback 533 * into the stack for SIMPLEX interfaces handled by ether_output(). 534 */ 535 if (m->m_pkthdr.rcvif != NULL && 536 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 537 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 538 if (ifa->ifa_addr->sa_family != AF_INET) 539 continue; 540 ia = ifatoia(ifa); 541 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 542 ip->ip_dst.s_addr) 543 goto ours; 544 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) 545 goto ours; 546 #ifdef BOOTP_COMPAT 547 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 548 goto ours; 549 #endif 550 } 551 } 552 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 553 struct in_multi *inm; 554 if (ip_mrouter) { 555 /* 556 * If we are acting as a multicast router, all 557 * incoming multicast packets are passed to the 558 * kernel-level multicast forwarding function. 559 * The packet is returned (relatively) intact; if 560 * ip_mforward() returns a non-zero value, the packet 561 * must be discarded, else it may be accepted below. 562 */ 563 if (ip_mforward && 564 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 565 ipstat.ips_cantforward++; 566 m_freem(m); 567 return; 568 } 569 570 /* 571 * The process-level routing daemon needs to receive 572 * all multicast IGMP packets, whether or not this 573 * host belongs to their destination groups. 574 */ 575 if (ip->ip_p == IPPROTO_IGMP) 576 goto ours; 577 ipstat.ips_forward++; 578 } 579 /* 580 * See if we belong to the destination multicast group on the 581 * arrival interface. 582 */ 583 IN_MULTI_LOCK(); 584 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 585 IN_MULTI_UNLOCK(); 586 if (inm == NULL) { 587 ipstat.ips_notmember++; 588 m_freem(m); 589 return; 590 } 591 goto ours; 592 } 593 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 594 goto ours; 595 if (ip->ip_dst.s_addr == INADDR_ANY) 596 goto ours; 597 598 /* 599 * FAITH(Firewall Aided Internet Translator) 600 */ 601 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 602 if (ip_keepfaith) { 603 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 604 goto ours; 605 } 606 m_freem(m); 607 return; 608 } 609 610 /* 611 * Not for us; forward if possible and desirable. 612 */ 613 if (ipforwarding == 0) { 614 ipstat.ips_cantforward++; 615 m_freem(m); 616 } else { 617 #ifdef IPSEC 618 /* 619 * Enforce inbound IPsec SPD. 620 */ 621 if (ipsec4_in_reject(m, NULL)) { 622 ipsecstat.in_polvio++; 623 goto bad; 624 } 625 #endif /* IPSEC */ 626 #ifdef FAST_IPSEC 627 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 628 s = splnet(); 629 if (mtag != NULL) { 630 tdbi = (struct tdb_ident *)(mtag + 1); 631 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 632 } else { 633 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 634 IP_FORWARDING, &error); 635 } 636 if (sp == NULL) { /* NB: can happen if error */ 637 splx(s); 638 /*XXX error stat???*/ 639 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 640 goto bad; 641 } 642 643 /* 644 * Check security policy against packet attributes. 645 */ 646 error = ipsec_in_reject(sp, m); 647 KEY_FREESP(&sp); 648 splx(s); 649 if (error) { 650 ipstat.ips_cantforward++; 651 goto bad; 652 } 653 #endif /* FAST_IPSEC */ 654 ip_forward(m, dchg); 655 } 656 return; 657 658 ours: 659 #ifdef IPSTEALTH 660 /* 661 * IPSTEALTH: Process non-routing options only 662 * if the packet is destined for us. 663 */ 664 if (ipstealth && hlen > sizeof (struct ip) && 665 ip_dooptions(m, 1)) 666 return; 667 #endif /* IPSTEALTH */ 668 669 /* Count the packet in the ip address stats */ 670 if (ia != NULL) { 671 ia->ia_ifa.if_ipackets++; 672 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 673 } 674 675 /* 676 * Attempt reassembly; if it succeeds, proceed. 677 * ip_reass() will return a different mbuf. 678 */ 679 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 680 m = ip_reass(m); 681 if (m == NULL) 682 return; 683 ip = mtod(m, struct ip *); 684 /* Get the header length of the reassembled packet */ 685 hlen = ip->ip_hl << 2; 686 } 687 688 /* 689 * Further protocols expect the packet length to be w/o the 690 * IP header. 691 */ 692 ip->ip_len -= hlen; 693 694 #ifdef IPSEC 695 /* 696 * enforce IPsec policy checking if we are seeing last header. 697 * note that we do not visit this with protocols with pcb layer 698 * code - like udp/tcp/raw ip. 699 */ 700 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 701 ipsec4_in_reject(m, NULL)) { 702 ipsecstat.in_polvio++; 703 goto bad; 704 } 705 #endif 706 #ifdef FAST_IPSEC 707 /* 708 * enforce IPsec policy checking if we are seeing last header. 709 * note that we do not visit this with protocols with pcb layer 710 * code - like udp/tcp/raw ip. 711 */ 712 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 713 /* 714 * Check if the packet has already had IPsec processing 715 * done. If so, then just pass it along. This tag gets 716 * set during AH, ESP, etc. input handling, before the 717 * packet is returned to the ip input queue for delivery. 718 */ 719 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 720 s = splnet(); 721 if (mtag != NULL) { 722 tdbi = (struct tdb_ident *)(mtag + 1); 723 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 724 } else { 725 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 726 IP_FORWARDING, &error); 727 } 728 if (sp != NULL) { 729 /* 730 * Check security policy against packet attributes. 731 */ 732 error = ipsec_in_reject(sp, m); 733 KEY_FREESP(&sp); 734 } else { 735 /* XXX error stat??? */ 736 error = EINVAL; 737 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 738 goto bad; 739 } 740 splx(s); 741 if (error) 742 goto bad; 743 } 744 #endif /* FAST_IPSEC */ 745 746 /* 747 * Switch out to protocol's input routine. 748 */ 749 ipstat.ips_delivered++; 750 751 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 752 return; 753 bad: 754 m_freem(m); 755 } 756 757 /* 758 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 759 * max has slightly different semantics than the sysctl, for historical 760 * reasons. 761 */ 762 static void 763 maxnipq_update(void) 764 { 765 766 /* 767 * -1 for unlimited allocation. 768 */ 769 if (maxnipq < 0) 770 uma_zone_set_max(ipq_zone, 0); 771 /* 772 * Positive number for specific bound. 773 */ 774 if (maxnipq > 0) 775 uma_zone_set_max(ipq_zone, maxnipq); 776 /* 777 * Zero specifies no further fragment queue allocation -- set the 778 * bound very low, but rely on implementation elsewhere to actually 779 * prevent allocation and reclaim current queues. 780 */ 781 if (maxnipq == 0) 782 uma_zone_set_max(ipq_zone, 1); 783 } 784 785 static int 786 sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 787 { 788 int error, i; 789 790 i = maxnipq; 791 error = sysctl_handle_int(oidp, &i, 0, req); 792 if (error || !req->newptr) 793 return (error); 794 795 /* 796 * XXXRW: Might be a good idea to sanity check the argument and place 797 * an extreme upper bound. 798 */ 799 if (i < -1) 800 return (EINVAL); 801 maxnipq = i; 802 maxnipq_update(); 803 return (0); 804 } 805 806 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 807 NULL, 0, sysctl_maxnipq, "I", 808 "Maximum number of IPv4 fragment reassembly queue entries"); 809 810 /* 811 * Take incoming datagram fragment and try to reassemble it into 812 * whole datagram. If the argument is the first fragment or one 813 * in between the function will return NULL and store the mbuf 814 * in the fragment chain. If the argument is the last fragment 815 * the packet will be reassembled and the pointer to the new 816 * mbuf returned for further processing. Only m_tags attached 817 * to the first packet/fragment are preserved. 818 * The IP header is *NOT* adjusted out of iplen. 819 */ 820 821 struct mbuf * 822 ip_reass(struct mbuf *m) 823 { 824 struct ip *ip; 825 struct mbuf *p, *q, *nq, *t; 826 struct ipq *fp = NULL; 827 struct ipqhead *head; 828 int i, hlen, next; 829 u_int8_t ecn, ecn0; 830 u_short hash; 831 832 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 833 if (maxnipq == 0 || maxfragsperpacket == 0) { 834 ipstat.ips_fragments++; 835 ipstat.ips_fragdropped++; 836 m_freem(m); 837 return (NULL); 838 } 839 840 ip = mtod(m, struct ip *); 841 hlen = ip->ip_hl << 2; 842 843 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 844 head = &ipq[hash]; 845 IPQ_LOCK(); 846 847 /* 848 * Look for queue of fragments 849 * of this datagram. 850 */ 851 TAILQ_FOREACH(fp, head, ipq_list) 852 if (ip->ip_id == fp->ipq_id && 853 ip->ip_src.s_addr == fp->ipq_src.s_addr && 854 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 855 #ifdef MAC 856 mac_fragment_match(m, fp) && 857 #endif 858 ip->ip_p == fp->ipq_p) 859 goto found; 860 861 fp = NULL; 862 863 /* 864 * Attempt to trim the number of allocated fragment queues if it 865 * exceeds the administrative limit. 866 */ 867 if ((nipq > maxnipq) && (maxnipq > 0)) { 868 /* 869 * drop something from the tail of the current queue 870 * before proceeding further 871 */ 872 struct ipq *q = TAILQ_LAST(head, ipqhead); 873 if (q == NULL) { /* gak */ 874 for (i = 0; i < IPREASS_NHASH; i++) { 875 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 876 if (r) { 877 ipstat.ips_fragtimeout += r->ipq_nfrags; 878 ip_freef(&ipq[i], r); 879 break; 880 } 881 } 882 } else { 883 ipstat.ips_fragtimeout += q->ipq_nfrags; 884 ip_freef(head, q); 885 } 886 } 887 888 found: 889 /* 890 * Adjust ip_len to not reflect header, 891 * convert offset of this to bytes. 892 */ 893 ip->ip_len -= hlen; 894 if (ip->ip_off & IP_MF) { 895 /* 896 * Make sure that fragments have a data length 897 * that's a non-zero multiple of 8 bytes. 898 */ 899 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 900 ipstat.ips_toosmall++; /* XXX */ 901 goto dropfrag; 902 } 903 m->m_flags |= M_FRAG; 904 } else 905 m->m_flags &= ~M_FRAG; 906 ip->ip_off <<= 3; 907 908 909 /* 910 * Attempt reassembly; if it succeeds, proceed. 911 * ip_reass() will return a different mbuf. 912 */ 913 ipstat.ips_fragments++; 914 m->m_pkthdr.header = ip; 915 916 /* Previous ip_reass() started here. */ 917 /* 918 * Presence of header sizes in mbufs 919 * would confuse code below. 920 */ 921 m->m_data += hlen; 922 m->m_len -= hlen; 923 924 /* 925 * If first fragment to arrive, create a reassembly queue. 926 */ 927 if (fp == NULL) { 928 fp = uma_zalloc(ipq_zone, M_NOWAIT); 929 if (fp == NULL) 930 goto dropfrag; 931 #ifdef MAC 932 if (mac_init_ipq(fp, M_NOWAIT) != 0) { 933 uma_zfree(ipq_zone, fp); 934 goto dropfrag; 935 } 936 mac_create_ipq(m, fp); 937 #endif 938 TAILQ_INSERT_HEAD(head, fp, ipq_list); 939 nipq++; 940 fp->ipq_nfrags = 1; 941 fp->ipq_ttl = IPFRAGTTL; 942 fp->ipq_p = ip->ip_p; 943 fp->ipq_id = ip->ip_id; 944 fp->ipq_src = ip->ip_src; 945 fp->ipq_dst = ip->ip_dst; 946 fp->ipq_frags = m; 947 m->m_nextpkt = NULL; 948 goto done; 949 } else { 950 fp->ipq_nfrags++; 951 #ifdef MAC 952 mac_update_ipq(m, fp); 953 #endif 954 } 955 956 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 957 958 /* 959 * Handle ECN by comparing this segment with the first one; 960 * if CE is set, do not lose CE. 961 * drop if CE and not-ECT are mixed for the same packet. 962 */ 963 ecn = ip->ip_tos & IPTOS_ECN_MASK; 964 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 965 if (ecn == IPTOS_ECN_CE) { 966 if (ecn0 == IPTOS_ECN_NOTECT) 967 goto dropfrag; 968 if (ecn0 != IPTOS_ECN_CE) 969 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 970 } 971 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 972 goto dropfrag; 973 974 /* 975 * Find a segment which begins after this one does. 976 */ 977 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 978 if (GETIP(q)->ip_off > ip->ip_off) 979 break; 980 981 /* 982 * If there is a preceding segment, it may provide some of 983 * our data already. If so, drop the data from the incoming 984 * segment. If it provides all of our data, drop us, otherwise 985 * stick new segment in the proper place. 986 * 987 * If some of the data is dropped from the the preceding 988 * segment, then it's checksum is invalidated. 989 */ 990 if (p) { 991 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 992 if (i > 0) { 993 if (i >= ip->ip_len) 994 goto dropfrag; 995 m_adj(m, i); 996 m->m_pkthdr.csum_flags = 0; 997 ip->ip_off += i; 998 ip->ip_len -= i; 999 } 1000 m->m_nextpkt = p->m_nextpkt; 1001 p->m_nextpkt = m; 1002 } else { 1003 m->m_nextpkt = fp->ipq_frags; 1004 fp->ipq_frags = m; 1005 } 1006 1007 /* 1008 * While we overlap succeeding segments trim them or, 1009 * if they are completely covered, dequeue them. 1010 */ 1011 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1012 q = nq) { 1013 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1014 if (i < GETIP(q)->ip_len) { 1015 GETIP(q)->ip_len -= i; 1016 GETIP(q)->ip_off += i; 1017 m_adj(q, i); 1018 q->m_pkthdr.csum_flags = 0; 1019 break; 1020 } 1021 nq = q->m_nextpkt; 1022 m->m_nextpkt = nq; 1023 ipstat.ips_fragdropped++; 1024 fp->ipq_nfrags--; 1025 m_freem(q); 1026 } 1027 1028 /* 1029 * Check for complete reassembly and perform frag per packet 1030 * limiting. 1031 * 1032 * Frag limiting is performed here so that the nth frag has 1033 * a chance to complete the packet before we drop the packet. 1034 * As a result, n+1 frags are actually allowed per packet, but 1035 * only n will ever be stored. (n = maxfragsperpacket.) 1036 * 1037 */ 1038 next = 0; 1039 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1040 if (GETIP(q)->ip_off != next) { 1041 if (fp->ipq_nfrags > maxfragsperpacket) { 1042 ipstat.ips_fragdropped += fp->ipq_nfrags; 1043 ip_freef(head, fp); 1044 } 1045 goto done; 1046 } 1047 next += GETIP(q)->ip_len; 1048 } 1049 /* Make sure the last packet didn't have the IP_MF flag */ 1050 if (p->m_flags & M_FRAG) { 1051 if (fp->ipq_nfrags > maxfragsperpacket) { 1052 ipstat.ips_fragdropped += fp->ipq_nfrags; 1053 ip_freef(head, fp); 1054 } 1055 goto done; 1056 } 1057 1058 /* 1059 * Reassembly is complete. Make sure the packet is a sane size. 1060 */ 1061 q = fp->ipq_frags; 1062 ip = GETIP(q); 1063 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1064 ipstat.ips_toolong++; 1065 ipstat.ips_fragdropped += fp->ipq_nfrags; 1066 ip_freef(head, fp); 1067 goto done; 1068 } 1069 1070 /* 1071 * Concatenate fragments. 1072 */ 1073 m = q; 1074 t = m->m_next; 1075 m->m_next = NULL; 1076 m_cat(m, t); 1077 nq = q->m_nextpkt; 1078 q->m_nextpkt = NULL; 1079 for (q = nq; q != NULL; q = nq) { 1080 nq = q->m_nextpkt; 1081 q->m_nextpkt = NULL; 1082 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1083 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1084 m_cat(m, q); 1085 } 1086 #ifdef MAC 1087 mac_create_datagram_from_ipq(fp, m); 1088 mac_destroy_ipq(fp); 1089 #endif 1090 1091 /* 1092 * Create header for new ip packet by modifying header of first 1093 * packet; dequeue and discard fragment reassembly header. 1094 * Make header visible. 1095 */ 1096 ip->ip_len = (ip->ip_hl << 2) + next; 1097 ip->ip_src = fp->ipq_src; 1098 ip->ip_dst = fp->ipq_dst; 1099 TAILQ_REMOVE(head, fp, ipq_list); 1100 nipq--; 1101 uma_zfree(ipq_zone, fp); 1102 m->m_len += (ip->ip_hl << 2); 1103 m->m_data -= (ip->ip_hl << 2); 1104 /* some debugging cruft by sklower, below, will go away soon */ 1105 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1106 m_fixhdr(m); 1107 ipstat.ips_reassembled++; 1108 IPQ_UNLOCK(); 1109 return (m); 1110 1111 dropfrag: 1112 ipstat.ips_fragdropped++; 1113 if (fp != NULL) 1114 fp->ipq_nfrags--; 1115 m_freem(m); 1116 done: 1117 IPQ_UNLOCK(); 1118 return (NULL); 1119 1120 #undef GETIP 1121 } 1122 1123 /* 1124 * Free a fragment reassembly header and all 1125 * associated datagrams. 1126 */ 1127 static void 1128 ip_freef(fhp, fp) 1129 struct ipqhead *fhp; 1130 struct ipq *fp; 1131 { 1132 register struct mbuf *q; 1133 1134 IPQ_LOCK_ASSERT(); 1135 1136 while (fp->ipq_frags) { 1137 q = fp->ipq_frags; 1138 fp->ipq_frags = q->m_nextpkt; 1139 m_freem(q); 1140 } 1141 TAILQ_REMOVE(fhp, fp, ipq_list); 1142 uma_zfree(ipq_zone, fp); 1143 nipq--; 1144 } 1145 1146 /* 1147 * IP timer processing; 1148 * if a timer expires on a reassembly 1149 * queue, discard it. 1150 */ 1151 void 1152 ip_slowtimo() 1153 { 1154 register struct ipq *fp; 1155 int i; 1156 1157 IPQ_LOCK(); 1158 for (i = 0; i < IPREASS_NHASH; i++) { 1159 for(fp = TAILQ_FIRST(&ipq[i]); fp;) { 1160 struct ipq *fpp; 1161 1162 fpp = fp; 1163 fp = TAILQ_NEXT(fp, ipq_list); 1164 if(--fpp->ipq_ttl == 0) { 1165 ipstat.ips_fragtimeout += fpp->ipq_nfrags; 1166 ip_freef(&ipq[i], fpp); 1167 } 1168 } 1169 } 1170 /* 1171 * If we are over the maximum number of fragments 1172 * (due to the limit being lowered), drain off 1173 * enough to get down to the new limit. 1174 */ 1175 if (maxnipq >= 0 && nipq > maxnipq) { 1176 for (i = 0; i < IPREASS_NHASH; i++) { 1177 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) { 1178 ipstat.ips_fragdropped += 1179 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1180 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1181 } 1182 } 1183 } 1184 IPQ_UNLOCK(); 1185 } 1186 1187 /* 1188 * Drain off all datagram fragments. 1189 */ 1190 void 1191 ip_drain() 1192 { 1193 int i; 1194 1195 IPQ_LOCK(); 1196 for (i = 0; i < IPREASS_NHASH; i++) { 1197 while(!TAILQ_EMPTY(&ipq[i])) { 1198 ipstat.ips_fragdropped += 1199 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1200 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1201 } 1202 } 1203 IPQ_UNLOCK(); 1204 in_rtqdrain(); 1205 } 1206 1207 /* 1208 * The protocol to be inserted into ip_protox[] must be already registered 1209 * in inetsw[], either statically or through pf_proto_register(). 1210 */ 1211 int 1212 ipproto_register(u_char ipproto) 1213 { 1214 struct protosw *pr; 1215 1216 /* Sanity checks. */ 1217 if (ipproto == 0) 1218 return (EPROTONOSUPPORT); 1219 1220 /* 1221 * The protocol slot must not be occupied by another protocol 1222 * already. An index pointing to IPPROTO_RAW is unused. 1223 */ 1224 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1225 if (pr == NULL) 1226 return (EPFNOSUPPORT); 1227 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1228 return (EEXIST); 1229 1230 /* Find the protocol position in inetsw[] and set the index. */ 1231 for (pr = inetdomain.dom_protosw; 1232 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1233 if (pr->pr_domain->dom_family == PF_INET && 1234 pr->pr_protocol && pr->pr_protocol == ipproto) { 1235 /* Be careful to only index valid IP protocols. */ 1236 if (pr->pr_protocol < IPPROTO_MAX) { 1237 ip_protox[pr->pr_protocol] = pr - inetsw; 1238 return (0); 1239 } else 1240 return (EINVAL); 1241 } 1242 } 1243 return (EPROTONOSUPPORT); 1244 } 1245 1246 int 1247 ipproto_unregister(u_char ipproto) 1248 { 1249 struct protosw *pr; 1250 1251 /* Sanity checks. */ 1252 if (ipproto == 0) 1253 return (EPROTONOSUPPORT); 1254 1255 /* Check if the protocol was indeed registered. */ 1256 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1257 if (pr == NULL) 1258 return (EPFNOSUPPORT); 1259 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1260 return (ENOENT); 1261 1262 /* Reset the protocol slot to IPPROTO_RAW. */ 1263 ip_protox[ipproto] = pr - inetsw; 1264 return (0); 1265 } 1266 1267 /* 1268 * Given address of next destination (final or next hop), 1269 * return internet address info of interface to be used to get there. 1270 */ 1271 struct in_ifaddr * 1272 ip_rtaddr(dst) 1273 struct in_addr dst; 1274 { 1275 struct route sro; 1276 struct sockaddr_in *sin; 1277 struct in_ifaddr *ifa; 1278 1279 bzero(&sro, sizeof(sro)); 1280 sin = (struct sockaddr_in *)&sro.ro_dst; 1281 sin->sin_family = AF_INET; 1282 sin->sin_len = sizeof(*sin); 1283 sin->sin_addr = dst; 1284 rtalloc_ign(&sro, RTF_CLONING); 1285 1286 if (sro.ro_rt == NULL) 1287 return (NULL); 1288 1289 ifa = ifatoia(sro.ro_rt->rt_ifa); 1290 RTFREE(sro.ro_rt); 1291 return (ifa); 1292 } 1293 1294 u_char inetctlerrmap[PRC_NCMDS] = { 1295 0, 0, 0, 0, 1296 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1297 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1298 EMSGSIZE, EHOSTUNREACH, 0, 0, 1299 0, 0, EHOSTUNREACH, 0, 1300 ENOPROTOOPT, ECONNREFUSED 1301 }; 1302 1303 /* 1304 * Forward a packet. If some error occurs return the sender 1305 * an icmp packet. Note we can't always generate a meaningful 1306 * icmp message because icmp doesn't have a large enough repertoire 1307 * of codes and types. 1308 * 1309 * If not forwarding, just drop the packet. This could be confusing 1310 * if ipforwarding was zero but some routing protocol was advancing 1311 * us as a gateway to somewhere. However, we must let the routing 1312 * protocol deal with that. 1313 * 1314 * The srcrt parameter indicates whether the packet is being forwarded 1315 * via a source route. 1316 */ 1317 void 1318 ip_forward(struct mbuf *m, int srcrt) 1319 { 1320 struct ip *ip = mtod(m, struct ip *); 1321 struct in_ifaddr *ia = NULL; 1322 struct mbuf *mcopy; 1323 struct in_addr dest; 1324 int error, type = 0, code = 0, mtu = 0; 1325 1326 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1327 ipstat.ips_cantforward++; 1328 m_freem(m); 1329 return; 1330 } 1331 #ifdef IPSTEALTH 1332 if (!ipstealth) { 1333 #endif 1334 if (ip->ip_ttl <= IPTTLDEC) { 1335 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1336 0, 0); 1337 return; 1338 } 1339 #ifdef IPSTEALTH 1340 } 1341 #endif 1342 1343 if (!srcrt && (ia = ip_rtaddr(ip->ip_dst)) == NULL) { 1344 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1345 return; 1346 } 1347 1348 /* 1349 * Save the IP header and at most 8 bytes of the payload, 1350 * in case we need to generate an ICMP message to the src. 1351 * 1352 * XXX this can be optimized a lot by saving the data in a local 1353 * buffer on the stack (72 bytes at most), and only allocating the 1354 * mbuf if really necessary. The vast majority of the packets 1355 * are forwarded without having to send an ICMP back (either 1356 * because unnecessary, or because rate limited), so we are 1357 * really we are wasting a lot of work here. 1358 * 1359 * We don't use m_copy() because it might return a reference 1360 * to a shared cluster. Both this function and ip_output() 1361 * assume exclusive access to the IP header in `m', so any 1362 * data in a cluster may change before we reach icmp_error(). 1363 */ 1364 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1365 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1366 /* 1367 * It's probably ok if the pkthdr dup fails (because 1368 * the deep copy of the tag chain failed), but for now 1369 * be conservative and just discard the copy since 1370 * code below may some day want the tags. 1371 */ 1372 m_free(mcopy); 1373 mcopy = NULL; 1374 } 1375 if (mcopy != NULL) { 1376 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy)); 1377 mcopy->m_pkthdr.len = mcopy->m_len; 1378 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1379 } 1380 1381 #ifdef IPSTEALTH 1382 if (!ipstealth) { 1383 #endif 1384 ip->ip_ttl -= IPTTLDEC; 1385 #ifdef IPSTEALTH 1386 } 1387 #endif 1388 1389 /* 1390 * If forwarding packet using same interface that it came in on, 1391 * perhaps should send a redirect to sender to shortcut a hop. 1392 * Only send redirect if source is sending directly to us, 1393 * and if packet was not source routed (or has any options). 1394 * Also, don't send redirect if forwarding using a default route 1395 * or a route modified by a redirect. 1396 */ 1397 dest.s_addr = 0; 1398 if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) { 1399 struct sockaddr_in *sin; 1400 struct route ro; 1401 struct rtentry *rt; 1402 1403 bzero(&ro, sizeof(ro)); 1404 sin = (struct sockaddr_in *)&ro.ro_dst; 1405 sin->sin_family = AF_INET; 1406 sin->sin_len = sizeof(*sin); 1407 sin->sin_addr = ip->ip_dst; 1408 rtalloc_ign(&ro, RTF_CLONING); 1409 1410 rt = ro.ro_rt; 1411 1412 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1413 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1414 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1415 u_long src = ntohl(ip->ip_src.s_addr); 1416 1417 if (RTA(rt) && 1418 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1419 if (rt->rt_flags & RTF_GATEWAY) 1420 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1421 else 1422 dest.s_addr = ip->ip_dst.s_addr; 1423 /* Router requirements says to only send host redirects */ 1424 type = ICMP_REDIRECT; 1425 code = ICMP_REDIRECT_HOST; 1426 } 1427 } 1428 if (rt) 1429 RTFREE(rt); 1430 } 1431 1432 error = ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL); 1433 if (error) 1434 ipstat.ips_cantforward++; 1435 else { 1436 ipstat.ips_forward++; 1437 if (type) 1438 ipstat.ips_redirectsent++; 1439 else { 1440 if (mcopy) 1441 m_freem(mcopy); 1442 return; 1443 } 1444 } 1445 if (mcopy == NULL) 1446 return; 1447 1448 switch (error) { 1449 1450 case 0: /* forwarded, but need redirect */ 1451 /* type, code set above */ 1452 break; 1453 1454 case ENETUNREACH: /* shouldn't happen, checked above */ 1455 case EHOSTUNREACH: 1456 case ENETDOWN: 1457 case EHOSTDOWN: 1458 default: 1459 type = ICMP_UNREACH; 1460 code = ICMP_UNREACH_HOST; 1461 break; 1462 1463 case EMSGSIZE: 1464 type = ICMP_UNREACH; 1465 code = ICMP_UNREACH_NEEDFRAG; 1466 #if defined(IPSEC) || defined(FAST_IPSEC) 1467 /* 1468 * If the packet is routed over IPsec tunnel, tell the 1469 * originator the tunnel MTU. 1470 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1471 * XXX quickhack!!! 1472 */ 1473 { 1474 struct secpolicy *sp = NULL; 1475 int ipsecerror; 1476 int ipsechdr; 1477 struct route *ro; 1478 1479 #ifdef IPSEC 1480 sp = ipsec4_getpolicybyaddr(mcopy, 1481 IPSEC_DIR_OUTBOUND, 1482 IP_FORWARDING, 1483 &ipsecerror); 1484 #else /* FAST_IPSEC */ 1485 sp = ipsec_getpolicybyaddr(mcopy, 1486 IPSEC_DIR_OUTBOUND, 1487 IP_FORWARDING, 1488 &ipsecerror); 1489 #endif 1490 if (sp != NULL) { 1491 /* count IPsec header size */ 1492 ipsechdr = ipsec4_hdrsiz(mcopy, 1493 IPSEC_DIR_OUTBOUND, 1494 NULL); 1495 1496 /* 1497 * find the correct route for outer IPv4 1498 * header, compute tunnel MTU. 1499 */ 1500 if (sp->req != NULL 1501 && sp->req->sav != NULL 1502 && sp->req->sav->sah != NULL) { 1503 ro = &sp->req->sav->sah->sa_route; 1504 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1505 mtu = 1506 ro->ro_rt->rt_rmx.rmx_mtu ? 1507 ro->ro_rt->rt_rmx.rmx_mtu : 1508 ro->ro_rt->rt_ifp->if_mtu; 1509 mtu -= ipsechdr; 1510 } 1511 } 1512 1513 #ifdef IPSEC 1514 key_freesp(sp); 1515 #else /* FAST_IPSEC */ 1516 KEY_FREESP(&sp); 1517 #endif 1518 ipstat.ips_cantfrag++; 1519 break; 1520 } 1521 #endif /*IPSEC || FAST_IPSEC*/ 1522 /* 1523 * If the MTU wasn't set before use the interface mtu or 1524 * fall back to the next smaller mtu step compared to the 1525 * current packet size. 1526 */ 1527 if (mtu == 0) { 1528 if (ia != NULL) 1529 mtu = ia->ia_ifp->if_mtu; 1530 else 1531 mtu = ip_next_mtu(ip->ip_len, 0); 1532 } 1533 #if defined(IPSEC) || defined(FAST_IPSEC) 1534 } 1535 #endif /*IPSEC || FAST_IPSEC*/ 1536 ipstat.ips_cantfrag++; 1537 break; 1538 1539 case ENOBUFS: 1540 /* 1541 * A router should not generate ICMP_SOURCEQUENCH as 1542 * required in RFC1812 Requirements for IP Version 4 Routers. 1543 * Source quench could be a big problem under DoS attacks, 1544 * or if the underlying interface is rate-limited. 1545 * Those who need source quench packets may re-enable them 1546 * via the net.inet.ip.sendsourcequench sysctl. 1547 */ 1548 if (ip_sendsourcequench == 0) { 1549 m_freem(mcopy); 1550 return; 1551 } else { 1552 type = ICMP_SOURCEQUENCH; 1553 code = 0; 1554 } 1555 break; 1556 1557 case EACCES: /* ipfw denied packet */ 1558 m_freem(mcopy); 1559 return; 1560 } 1561 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1562 } 1563 1564 void 1565 ip_savecontrol(inp, mp, ip, m) 1566 register struct inpcb *inp; 1567 register struct mbuf **mp; 1568 register struct ip *ip; 1569 register struct mbuf *m; 1570 { 1571 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1572 struct bintime bt; 1573 1574 bintime(&bt); 1575 if (inp->inp_socket->so_options & SO_BINTIME) { 1576 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1577 SCM_BINTIME, SOL_SOCKET); 1578 if (*mp) 1579 mp = &(*mp)->m_next; 1580 } 1581 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1582 struct timeval tv; 1583 1584 bintime2timeval(&bt, &tv); 1585 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1586 SCM_TIMESTAMP, SOL_SOCKET); 1587 if (*mp) 1588 mp = &(*mp)->m_next; 1589 } 1590 } 1591 if (inp->inp_flags & INP_RECVDSTADDR) { 1592 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1593 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1594 if (*mp) 1595 mp = &(*mp)->m_next; 1596 } 1597 if (inp->inp_flags & INP_RECVTTL) { 1598 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1599 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1600 if (*mp) 1601 mp = &(*mp)->m_next; 1602 } 1603 #ifdef notyet 1604 /* XXX 1605 * Moving these out of udp_input() made them even more broken 1606 * than they already were. 1607 */ 1608 /* options were tossed already */ 1609 if (inp->inp_flags & INP_RECVOPTS) { 1610 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1611 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1612 if (*mp) 1613 mp = &(*mp)->m_next; 1614 } 1615 /* ip_srcroute doesn't do what we want here, need to fix */ 1616 if (inp->inp_flags & INP_RECVRETOPTS) { 1617 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1618 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1619 if (*mp) 1620 mp = &(*mp)->m_next; 1621 } 1622 #endif 1623 if (inp->inp_flags & INP_RECVIF) { 1624 struct ifnet *ifp; 1625 struct sdlbuf { 1626 struct sockaddr_dl sdl; 1627 u_char pad[32]; 1628 } sdlbuf; 1629 struct sockaddr_dl *sdp; 1630 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1631 1632 if (((ifp = m->m_pkthdr.rcvif)) 1633 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1634 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1635 /* 1636 * Change our mind and don't try copy. 1637 */ 1638 if ((sdp->sdl_family != AF_LINK) 1639 || (sdp->sdl_len > sizeof(sdlbuf))) { 1640 goto makedummy; 1641 } 1642 bcopy(sdp, sdl2, sdp->sdl_len); 1643 } else { 1644 makedummy: 1645 sdl2->sdl_len 1646 = offsetof(struct sockaddr_dl, sdl_data[0]); 1647 sdl2->sdl_family = AF_LINK; 1648 sdl2->sdl_index = 0; 1649 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1650 } 1651 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1652 IP_RECVIF, IPPROTO_IP); 1653 if (*mp) 1654 mp = &(*mp)->m_next; 1655 } 1656 } 1657 1658 /* 1659 * XXX these routines are called from the upper part of the kernel. 1660 * They need to be locked when we remove Giant. 1661 * 1662 * They could also be moved to ip_mroute.c, since all the RSVP 1663 * handling is done there already. 1664 */ 1665 static int ip_rsvp_on; 1666 struct socket *ip_rsvpd; 1667 int 1668 ip_rsvp_init(struct socket *so) 1669 { 1670 if (so->so_type != SOCK_RAW || 1671 so->so_proto->pr_protocol != IPPROTO_RSVP) 1672 return EOPNOTSUPP; 1673 1674 if (ip_rsvpd != NULL) 1675 return EADDRINUSE; 1676 1677 ip_rsvpd = so; 1678 /* 1679 * This may seem silly, but we need to be sure we don't over-increment 1680 * the RSVP counter, in case something slips up. 1681 */ 1682 if (!ip_rsvp_on) { 1683 ip_rsvp_on = 1; 1684 rsvp_on++; 1685 } 1686 1687 return 0; 1688 } 1689 1690 int 1691 ip_rsvp_done(void) 1692 { 1693 ip_rsvpd = NULL; 1694 /* 1695 * This may seem silly, but we need to be sure we don't over-decrement 1696 * the RSVP counter, in case something slips up. 1697 */ 1698 if (ip_rsvp_on) { 1699 ip_rsvp_on = 0; 1700 rsvp_on--; 1701 } 1702 return 0; 1703 } 1704 1705 void 1706 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1707 { 1708 if (rsvp_input_p) { /* call the real one if loaded */ 1709 rsvp_input_p(m, off); 1710 return; 1711 } 1712 1713 /* Can still get packets with rsvp_on = 0 if there is a local member 1714 * of the group to which the RSVP packet is addressed. But in this 1715 * case we want to throw the packet away. 1716 */ 1717 1718 if (!rsvp_on) { 1719 m_freem(m); 1720 return; 1721 } 1722 1723 if (ip_rsvpd != NULL) { 1724 rip_input(m, off); 1725 return; 1726 } 1727 /* Drop the packet */ 1728 m_freem(m); 1729 } 1730