1 /*- 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include "opt_bootp.h" 36 #include "opt_ipfw.h" 37 #include "opt_ipstealth.h" 38 #include "opt_ipsec.h" 39 #include "opt_mac.h" 40 #include "opt_carp.h" 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/callout.h> 45 #include <sys/mbuf.h> 46 #include <sys/malloc.h> 47 #include <sys/domain.h> 48 #include <sys/protosw.h> 49 #include <sys/socket.h> 50 #include <sys/time.h> 51 #include <sys/kernel.h> 52 #include <sys/syslog.h> 53 #include <sys/sysctl.h> 54 #include <sys/vimage.h> 55 56 #include <net/pfil.h> 57 #include <net/if.h> 58 #include <net/if_types.h> 59 #include <net/if_var.h> 60 #include <net/if_dl.h> 61 #include <net/route.h> 62 #include <net/netisr.h> 63 64 #include <netinet/in.h> 65 #include <netinet/in_systm.h> 66 #include <netinet/in_var.h> 67 #include <netinet/ip.h> 68 #include <netinet/in_pcb.h> 69 #include <netinet/ip_var.h> 70 #include <netinet/ip_icmp.h> 71 #include <netinet/ip_options.h> 72 #include <machine/in_cksum.h> 73 #ifdef DEV_CARP 74 #include <netinet/ip_carp.h> 75 #endif 76 #ifdef IPSEC 77 #include <netinet/ip_ipsec.h> 78 #endif /* IPSEC */ 79 80 #include <sys/socketvar.h> 81 82 /* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */ 83 #include <netinet/ip_fw.h> 84 #include <netinet/ip_dummynet.h> 85 86 #include <security/mac/mac_framework.h> 87 88 #ifdef CTASSERT 89 CTASSERT(sizeof(struct ip) == 20); 90 #endif 91 92 int rsvp_on = 0; 93 94 int ipforwarding = 0; 95 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 96 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 97 98 static int ipsendredirects = 1; /* XXX */ 99 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 100 &ipsendredirects, 0, "Enable sending IP redirects"); 101 102 int ip_defttl = IPDEFTTL; 103 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 104 &ip_defttl, 0, "Maximum TTL on IP packets"); 105 106 static int ip_keepfaith = 0; 107 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 108 &ip_keepfaith, 0, 109 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 110 111 static int ip_sendsourcequench = 0; 112 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 113 &ip_sendsourcequench, 0, 114 "Enable the transmission of source quench packets"); 115 116 int ip_do_randomid = 0; 117 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 118 &ip_do_randomid, 0, 119 "Assign random ip_id values"); 120 121 /* 122 * XXX - Setting ip_checkinterface mostly implements the receive side of 123 * the Strong ES model described in RFC 1122, but since the routing table 124 * and transmit implementation do not implement the Strong ES model, 125 * setting this to 1 results in an odd hybrid. 126 * 127 * XXX - ip_checkinterface currently must be disabled if you use ipnat 128 * to translate the destination address to another local interface. 129 * 130 * XXX - ip_checkinterface must be disabled if you add IP aliases 131 * to the loopback interface instead of the interface where the 132 * packets for those addresses are received. 133 */ 134 static int ip_checkinterface = 0; 135 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 136 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 137 138 struct pfil_head inet_pfil_hook; /* Packet filter hooks */ 139 140 static struct ifqueue ipintrq; 141 static int ipqmaxlen = IFQ_MAXLEN; 142 143 extern struct domain inetdomain; 144 extern struct protosw inetsw[]; 145 u_char ip_protox[IPPROTO_MAX]; 146 struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 147 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 148 u_long in_ifaddrhmask; /* mask for hash table */ 149 150 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 151 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 152 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 153 &ipintrq.ifq_drops, 0, 154 "Number of packets dropped from the IP input queue"); 155 156 struct ipstat ipstat; 157 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 158 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 159 160 /* 161 * IP datagram reassembly. 162 */ 163 #define IPREASS_NHASH_LOG2 6 164 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 165 #define IPREASS_HMASK (IPREASS_NHASH - 1) 166 #define IPREASS_HASH(x,y) \ 167 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 168 169 static uma_zone_t ipq_zone; 170 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 171 static struct mtx ipqlock; 172 173 #define IPQ_LOCK() mtx_lock(&ipqlock) 174 #define IPQ_UNLOCK() mtx_unlock(&ipqlock) 175 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 176 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 177 178 static void maxnipq_update(void); 179 static void ipq_zone_change(void *); 180 181 static int maxnipq; /* Administrative limit on # reass queues. */ 182 static int nipq = 0; /* Total # of reass queues */ 183 SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, 184 &nipq, 0, "Current number of IPv4 fragment reassembly queue entries"); 185 186 static int maxfragsperpacket; 187 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 188 &maxfragsperpacket, 0, 189 "Maximum number of IPv4 fragments allowed per packet"); 190 191 struct callout ipport_tick_callout; 192 193 #ifdef IPCTL_DEFMTU 194 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 195 &ip_mtu, 0, "Default MTU"); 196 #endif 197 198 #ifdef IPSTEALTH 199 int ipstealth = 0; 200 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 201 &ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding"); 202 #endif 203 204 /* 205 * ipfw_ether and ipfw_bridge hooks. 206 * XXX: Temporary until those are converted to pfil_hooks as well. 207 */ 208 ip_fw_chk_t *ip_fw_chk_ptr = NULL; 209 ip_dn_io_t *ip_dn_io_ptr = NULL; 210 int fw_one_pass = 1; 211 212 static void ip_freef(struct ipqhead *, struct ipq *); 213 214 /* 215 * IP initialization: fill in IP protocol switch table. 216 * All protocols not implemented in kernel go to raw IP protocol handler. 217 */ 218 void 219 ip_init(void) 220 { 221 struct protosw *pr; 222 int i; 223 224 TAILQ_INIT(&V_in_ifaddrhead); 225 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 226 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 227 if (pr == NULL) 228 panic("ip_init: PF_INET not found"); 229 230 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 231 for (i = 0; i < IPPROTO_MAX; i++) 232 ip_protox[i] = pr - inetsw; 233 /* 234 * Cycle through IP protocols and put them into the appropriate place 235 * in ip_protox[]. 236 */ 237 for (pr = inetdomain.dom_protosw; 238 pr < inetdomain.dom_protoswNPROTOSW; pr++) 239 if (pr->pr_domain->dom_family == PF_INET && 240 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 241 /* Be careful to only index valid IP protocols. */ 242 if (pr->pr_protocol < IPPROTO_MAX) 243 ip_protox[pr->pr_protocol] = pr - inetsw; 244 } 245 246 /* Initialize packet filter hooks. */ 247 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 248 inet_pfil_hook.ph_af = AF_INET; 249 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 250 printf("%s: WARNING: unable to register pfil hook, " 251 "error %d\n", __func__, i); 252 253 /* Initialize IP reassembly queue. */ 254 IPQ_LOCK_INIT(); 255 for (i = 0; i < IPREASS_NHASH; i++) 256 TAILQ_INIT(&V_ipq[i]); 257 V_maxnipq = nmbclusters / 32; 258 V_maxfragsperpacket = 16; 259 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 260 NULL, UMA_ALIGN_PTR, 0); 261 maxnipq_update(); 262 263 /* Start ipport_tick. */ 264 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE); 265 ipport_tick(NULL); 266 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, 267 SHUTDOWN_PRI_DEFAULT); 268 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, 269 NULL, EVENTHANDLER_PRI_ANY); 270 271 /* Initialize various other remaining things. */ 272 V_ip_id = time_second & 0xffff; 273 ipintrq.ifq_maxlen = ipqmaxlen; 274 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF); 275 netisr_register(NETISR_IP, ip_input, &ipintrq, 0); 276 } 277 278 void 279 ip_fini(void *xtp) 280 { 281 282 callout_stop(&ipport_tick_callout); 283 } 284 285 /* 286 * Ip input routine. Checksum and byte swap header. If fragmented 287 * try to reassemble. Process options. Pass to next level. 288 */ 289 void 290 ip_input(struct mbuf *m) 291 { 292 struct ip *ip = NULL; 293 struct in_ifaddr *ia = NULL; 294 struct ifaddr *ifa; 295 int checkif, hlen = 0; 296 u_short sum; 297 int dchg = 0; /* dest changed after fw */ 298 struct in_addr odst; /* original dst address */ 299 300 M_ASSERTPKTHDR(m); 301 302 if (m->m_flags & M_FASTFWD_OURS) { 303 /* 304 * Firewall or NAT changed destination to local. 305 * We expect ip_len and ip_off to be in host byte order. 306 */ 307 m->m_flags &= ~M_FASTFWD_OURS; 308 /* Set up some basics that will be used later. */ 309 ip = mtod(m, struct ip *); 310 hlen = ip->ip_hl << 2; 311 goto ours; 312 } 313 314 V_ipstat.ips_total++; 315 316 if (m->m_pkthdr.len < sizeof(struct ip)) 317 goto tooshort; 318 319 if (m->m_len < sizeof (struct ip) && 320 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 321 V_ipstat.ips_toosmall++; 322 return; 323 } 324 ip = mtod(m, struct ip *); 325 326 if (ip->ip_v != IPVERSION) { 327 V_ipstat.ips_badvers++; 328 goto bad; 329 } 330 331 hlen = ip->ip_hl << 2; 332 if (hlen < sizeof(struct ip)) { /* minimum header length */ 333 V_ipstat.ips_badhlen++; 334 goto bad; 335 } 336 if (hlen > m->m_len) { 337 if ((m = m_pullup(m, hlen)) == NULL) { 338 V_ipstat.ips_badhlen++; 339 return; 340 } 341 ip = mtod(m, struct ip *); 342 } 343 344 /* 127/8 must not appear on wire - RFC1122 */ 345 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 346 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 347 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 348 V_ipstat.ips_badaddr++; 349 goto bad; 350 } 351 } 352 353 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 354 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 355 } else { 356 if (hlen == sizeof(struct ip)) { 357 sum = in_cksum_hdr(ip); 358 } else { 359 sum = in_cksum(m, hlen); 360 } 361 } 362 if (sum) { 363 V_ipstat.ips_badsum++; 364 goto bad; 365 } 366 367 #ifdef ALTQ 368 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 369 /* packet is dropped by traffic conditioner */ 370 return; 371 #endif 372 373 /* 374 * Convert fields to host representation. 375 */ 376 ip->ip_len = ntohs(ip->ip_len); 377 if (ip->ip_len < hlen) { 378 V_ipstat.ips_badlen++; 379 goto bad; 380 } 381 ip->ip_off = ntohs(ip->ip_off); 382 383 /* 384 * Check that the amount of data in the buffers 385 * is as at least much as the IP header would have us expect. 386 * Trim mbufs if longer than we expect. 387 * Drop packet if shorter than we expect. 388 */ 389 if (m->m_pkthdr.len < ip->ip_len) { 390 tooshort: 391 V_ipstat.ips_tooshort++; 392 goto bad; 393 } 394 if (m->m_pkthdr.len > ip->ip_len) { 395 if (m->m_len == m->m_pkthdr.len) { 396 m->m_len = ip->ip_len; 397 m->m_pkthdr.len = ip->ip_len; 398 } else 399 m_adj(m, ip->ip_len - m->m_pkthdr.len); 400 } 401 #ifdef IPSEC 402 /* 403 * Bypass packet filtering for packets from a tunnel (gif). 404 */ 405 if (ip_ipsec_filtertunnel(m)) 406 goto passin; 407 #endif /* IPSEC */ 408 409 /* 410 * Run through list of hooks for input packets. 411 * 412 * NB: Beware of the destination address changing (e.g. 413 * by NAT rewriting). When this happens, tell 414 * ip_forward to do the right thing. 415 */ 416 417 /* Jump over all PFIL processing if hooks are not active. */ 418 if (!PFIL_HOOKED(&inet_pfil_hook)) 419 goto passin; 420 421 odst = ip->ip_dst; 422 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 423 PFIL_IN, NULL) != 0) 424 return; 425 if (m == NULL) /* consumed by filter */ 426 return; 427 428 ip = mtod(m, struct ip *); 429 dchg = (odst.s_addr != ip->ip_dst.s_addr); 430 431 #ifdef IPFIREWALL_FORWARD 432 if (m->m_flags & M_FASTFWD_OURS) { 433 m->m_flags &= ~M_FASTFWD_OURS; 434 goto ours; 435 } 436 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 437 /* 438 * Directly ship on the packet. This allows to forward packets 439 * that were destined for us to some other directly connected 440 * host. 441 */ 442 ip_forward(m, dchg); 443 return; 444 } 445 #endif /* IPFIREWALL_FORWARD */ 446 447 passin: 448 /* 449 * Process options and, if not destined for us, 450 * ship it on. ip_dooptions returns 1 when an 451 * error was detected (causing an icmp message 452 * to be sent and the original packet to be freed). 453 */ 454 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 455 return; 456 457 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 458 * matter if it is destined to another node, or whether it is 459 * a multicast one, RSVP wants it! and prevents it from being forwarded 460 * anywhere else. Also checks if the rsvp daemon is running before 461 * grabbing the packet. 462 */ 463 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 464 goto ours; 465 466 /* 467 * Check our list of addresses, to see if the packet is for us. 468 * If we don't have any addresses, assume any unicast packet 469 * we receive might be for us (and let the upper layers deal 470 * with it). 471 */ 472 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 473 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 474 goto ours; 475 476 /* 477 * Enable a consistency check between the destination address 478 * and the arrival interface for a unicast packet (the RFC 1122 479 * strong ES model) if IP forwarding is disabled and the packet 480 * is not locally generated and the packet is not subject to 481 * 'ipfw fwd'. 482 * 483 * XXX - Checking also should be disabled if the destination 484 * address is ipnat'ed to a different interface. 485 * 486 * XXX - Checking is incompatible with IP aliases added 487 * to the loopback interface instead of the interface where 488 * the packets are received. 489 * 490 * XXX - This is the case for carp vhost IPs as well so we 491 * insert a workaround. If the packet got here, we already 492 * checked with carp_iamatch() and carp_forus(). 493 */ 494 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 495 m->m_pkthdr.rcvif != NULL && 496 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 497 #ifdef DEV_CARP 498 !m->m_pkthdr.rcvif->if_carp && 499 #endif 500 (dchg == 0); 501 502 /* 503 * Check for exact addresses in the hash bucket. 504 */ 505 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 506 /* 507 * If the address matches, verify that the packet 508 * arrived via the correct interface if checking is 509 * enabled. 510 */ 511 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 512 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 513 goto ours; 514 } 515 /* 516 * Check for broadcast addresses. 517 * 518 * Only accept broadcast packets that arrive via the matching 519 * interface. Reception of forwarded directed broadcasts would 520 * be handled via ip_forward() and ether_output() with the loopback 521 * into the stack for SIMPLEX interfaces handled by ether_output(). 522 */ 523 if (m->m_pkthdr.rcvif != NULL && 524 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 525 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 526 if (ifa->ifa_addr->sa_family != AF_INET) 527 continue; 528 ia = ifatoia(ifa); 529 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 530 ip->ip_dst.s_addr) 531 goto ours; 532 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) 533 goto ours; 534 #ifdef BOOTP_COMPAT 535 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 536 goto ours; 537 #endif 538 } 539 } 540 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 541 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 542 V_ipstat.ips_cantforward++; 543 m_freem(m); 544 return; 545 } 546 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 547 struct in_multi *inm; 548 if (V_ip_mrouter) { 549 /* 550 * If we are acting as a multicast router, all 551 * incoming multicast packets are passed to the 552 * kernel-level multicast forwarding function. 553 * The packet is returned (relatively) intact; if 554 * ip_mforward() returns a non-zero value, the packet 555 * must be discarded, else it may be accepted below. 556 */ 557 if (ip_mforward && 558 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 559 V_ipstat.ips_cantforward++; 560 m_freem(m); 561 return; 562 } 563 564 /* 565 * The process-level routing daemon needs to receive 566 * all multicast IGMP packets, whether or not this 567 * host belongs to their destination groups. 568 */ 569 if (ip->ip_p == IPPROTO_IGMP) 570 goto ours; 571 V_ipstat.ips_forward++; 572 } 573 /* 574 * See if we belong to the destination multicast group on the 575 * arrival interface. 576 */ 577 IN_MULTI_LOCK(); 578 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 579 IN_MULTI_UNLOCK(); 580 if (inm == NULL) { 581 V_ipstat.ips_notmember++; 582 m_freem(m); 583 return; 584 } 585 goto ours; 586 } 587 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 588 goto ours; 589 if (ip->ip_dst.s_addr == INADDR_ANY) 590 goto ours; 591 592 /* 593 * FAITH(Firewall Aided Internet Translator) 594 */ 595 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 596 if (V_ip_keepfaith) { 597 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 598 goto ours; 599 } 600 m_freem(m); 601 return; 602 } 603 604 /* 605 * Not for us; forward if possible and desirable. 606 */ 607 if (V_ipforwarding == 0) { 608 V_ipstat.ips_cantforward++; 609 m_freem(m); 610 } else { 611 #ifdef IPSEC 612 if (ip_ipsec_fwd(m)) 613 goto bad; 614 #endif /* IPSEC */ 615 ip_forward(m, dchg); 616 } 617 return; 618 619 ours: 620 #ifdef IPSTEALTH 621 /* 622 * IPSTEALTH: Process non-routing options only 623 * if the packet is destined for us. 624 */ 625 if (V_ipstealth && hlen > sizeof (struct ip) && 626 ip_dooptions(m, 1)) 627 return; 628 #endif /* IPSTEALTH */ 629 630 /* Count the packet in the ip address stats */ 631 if (ia != NULL) { 632 ia->ia_ifa.if_ipackets++; 633 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 634 } 635 636 /* 637 * Attempt reassembly; if it succeeds, proceed. 638 * ip_reass() will return a different mbuf. 639 */ 640 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 641 m = ip_reass(m); 642 if (m == NULL) 643 return; 644 ip = mtod(m, struct ip *); 645 /* Get the header length of the reassembled packet */ 646 hlen = ip->ip_hl << 2; 647 } 648 649 /* 650 * Further protocols expect the packet length to be w/o the 651 * IP header. 652 */ 653 ip->ip_len -= hlen; 654 655 #ifdef IPSEC 656 /* 657 * enforce IPsec policy checking if we are seeing last header. 658 * note that we do not visit this with protocols with pcb layer 659 * code - like udp/tcp/raw ip. 660 */ 661 if (ip_ipsec_input(m)) 662 goto bad; 663 #endif /* IPSEC */ 664 665 /* 666 * Switch out to protocol's input routine. 667 */ 668 V_ipstat.ips_delivered++; 669 670 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 671 return; 672 bad: 673 m_freem(m); 674 } 675 676 /* 677 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 678 * max has slightly different semantics than the sysctl, for historical 679 * reasons. 680 */ 681 static void 682 maxnipq_update(void) 683 { 684 685 /* 686 * -1 for unlimited allocation. 687 */ 688 if (V_maxnipq < 0) 689 uma_zone_set_max(V_ipq_zone, 0); 690 /* 691 * Positive number for specific bound. 692 */ 693 if (V_maxnipq > 0) 694 uma_zone_set_max(V_ipq_zone, V_maxnipq); 695 /* 696 * Zero specifies no further fragment queue allocation -- set the 697 * bound very low, but rely on implementation elsewhere to actually 698 * prevent allocation and reclaim current queues. 699 */ 700 if (V_maxnipq == 0) 701 uma_zone_set_max(V_ipq_zone, 1); 702 } 703 704 static void 705 ipq_zone_change(void *tag) 706 { 707 708 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) { 709 V_maxnipq = nmbclusters / 32; 710 maxnipq_update(); 711 } 712 } 713 714 static int 715 sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 716 { 717 int error, i; 718 719 i = V_maxnipq; 720 error = sysctl_handle_int(oidp, &i, 0, req); 721 if (error || !req->newptr) 722 return (error); 723 724 /* 725 * XXXRW: Might be a good idea to sanity check the argument and place 726 * an extreme upper bound. 727 */ 728 if (i < -1) 729 return (EINVAL); 730 V_maxnipq = i; 731 maxnipq_update(); 732 return (0); 733 } 734 735 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 736 NULL, 0, sysctl_maxnipq, "I", 737 "Maximum number of IPv4 fragment reassembly queue entries"); 738 739 /* 740 * Take incoming datagram fragment and try to reassemble it into 741 * whole datagram. If the argument is the first fragment or one 742 * in between the function will return NULL and store the mbuf 743 * in the fragment chain. If the argument is the last fragment 744 * the packet will be reassembled and the pointer to the new 745 * mbuf returned for further processing. Only m_tags attached 746 * to the first packet/fragment are preserved. 747 * The IP header is *NOT* adjusted out of iplen. 748 */ 749 struct mbuf * 750 ip_reass(struct mbuf *m) 751 { 752 struct ip *ip; 753 struct mbuf *p, *q, *nq, *t; 754 struct ipq *fp = NULL; 755 struct ipqhead *head; 756 int i, hlen, next; 757 u_int8_t ecn, ecn0; 758 u_short hash; 759 760 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 761 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) { 762 V_ipstat.ips_fragments++; 763 V_ipstat.ips_fragdropped++; 764 m_freem(m); 765 return (NULL); 766 } 767 768 ip = mtod(m, struct ip *); 769 hlen = ip->ip_hl << 2; 770 771 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 772 head = &V_ipq[hash]; 773 IPQ_LOCK(); 774 775 /* 776 * Look for queue of fragments 777 * of this datagram. 778 */ 779 TAILQ_FOREACH(fp, head, ipq_list) 780 if (ip->ip_id == fp->ipq_id && 781 ip->ip_src.s_addr == fp->ipq_src.s_addr && 782 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 783 #ifdef MAC 784 mac_ipq_match(m, fp) && 785 #endif 786 ip->ip_p == fp->ipq_p) 787 goto found; 788 789 fp = NULL; 790 791 /* 792 * Attempt to trim the number of allocated fragment queues if it 793 * exceeds the administrative limit. 794 */ 795 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) { 796 /* 797 * drop something from the tail of the current queue 798 * before proceeding further 799 */ 800 struct ipq *q = TAILQ_LAST(head, ipqhead); 801 if (q == NULL) { /* gak */ 802 for (i = 0; i < IPREASS_NHASH; i++) { 803 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead); 804 if (r) { 805 V_ipstat.ips_fragtimeout += 806 r->ipq_nfrags; 807 ip_freef(&V_ipq[i], r); 808 break; 809 } 810 } 811 } else { 812 V_ipstat.ips_fragtimeout += q->ipq_nfrags; 813 ip_freef(head, q); 814 } 815 } 816 817 found: 818 /* 819 * Adjust ip_len to not reflect header, 820 * convert offset of this to bytes. 821 */ 822 ip->ip_len -= hlen; 823 if (ip->ip_off & IP_MF) { 824 /* 825 * Make sure that fragments have a data length 826 * that's a non-zero multiple of 8 bytes. 827 */ 828 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 829 V_ipstat.ips_toosmall++; /* XXX */ 830 goto dropfrag; 831 } 832 m->m_flags |= M_FRAG; 833 } else 834 m->m_flags &= ~M_FRAG; 835 ip->ip_off <<= 3; 836 837 838 /* 839 * Attempt reassembly; if it succeeds, proceed. 840 * ip_reass() will return a different mbuf. 841 */ 842 V_ipstat.ips_fragments++; 843 m->m_pkthdr.header = ip; 844 845 /* Previous ip_reass() started here. */ 846 /* 847 * Presence of header sizes in mbufs 848 * would confuse code below. 849 */ 850 m->m_data += hlen; 851 m->m_len -= hlen; 852 853 /* 854 * If first fragment to arrive, create a reassembly queue. 855 */ 856 if (fp == NULL) { 857 fp = uma_zalloc(V_ipq_zone, M_NOWAIT); 858 if (fp == NULL) 859 goto dropfrag; 860 #ifdef MAC 861 if (mac_ipq_init(fp, M_NOWAIT) != 0) { 862 uma_zfree(V_ipq_zone, fp); 863 fp = NULL; 864 goto dropfrag; 865 } 866 mac_ipq_create(m, fp); 867 #endif 868 TAILQ_INSERT_HEAD(head, fp, ipq_list); 869 V_nipq++; 870 fp->ipq_nfrags = 1; 871 fp->ipq_ttl = IPFRAGTTL; 872 fp->ipq_p = ip->ip_p; 873 fp->ipq_id = ip->ip_id; 874 fp->ipq_src = ip->ip_src; 875 fp->ipq_dst = ip->ip_dst; 876 fp->ipq_frags = m; 877 m->m_nextpkt = NULL; 878 goto done; 879 } else { 880 fp->ipq_nfrags++; 881 #ifdef MAC 882 mac_ipq_update(m, fp); 883 #endif 884 } 885 886 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 887 888 /* 889 * Handle ECN by comparing this segment with the first one; 890 * if CE is set, do not lose CE. 891 * drop if CE and not-ECT are mixed for the same packet. 892 */ 893 ecn = ip->ip_tos & IPTOS_ECN_MASK; 894 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 895 if (ecn == IPTOS_ECN_CE) { 896 if (ecn0 == IPTOS_ECN_NOTECT) 897 goto dropfrag; 898 if (ecn0 != IPTOS_ECN_CE) 899 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 900 } 901 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 902 goto dropfrag; 903 904 /* 905 * Find a segment which begins after this one does. 906 */ 907 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 908 if (GETIP(q)->ip_off > ip->ip_off) 909 break; 910 911 /* 912 * If there is a preceding segment, it may provide some of 913 * our data already. If so, drop the data from the incoming 914 * segment. If it provides all of our data, drop us, otherwise 915 * stick new segment in the proper place. 916 * 917 * If some of the data is dropped from the the preceding 918 * segment, then it's checksum is invalidated. 919 */ 920 if (p) { 921 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 922 if (i > 0) { 923 if (i >= ip->ip_len) 924 goto dropfrag; 925 m_adj(m, i); 926 m->m_pkthdr.csum_flags = 0; 927 ip->ip_off += i; 928 ip->ip_len -= i; 929 } 930 m->m_nextpkt = p->m_nextpkt; 931 p->m_nextpkt = m; 932 } else { 933 m->m_nextpkt = fp->ipq_frags; 934 fp->ipq_frags = m; 935 } 936 937 /* 938 * While we overlap succeeding segments trim them or, 939 * if they are completely covered, dequeue them. 940 */ 941 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 942 q = nq) { 943 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 944 if (i < GETIP(q)->ip_len) { 945 GETIP(q)->ip_len -= i; 946 GETIP(q)->ip_off += i; 947 m_adj(q, i); 948 q->m_pkthdr.csum_flags = 0; 949 break; 950 } 951 nq = q->m_nextpkt; 952 m->m_nextpkt = nq; 953 V_ipstat.ips_fragdropped++; 954 fp->ipq_nfrags--; 955 m_freem(q); 956 } 957 958 /* 959 * Check for complete reassembly and perform frag per packet 960 * limiting. 961 * 962 * Frag limiting is performed here so that the nth frag has 963 * a chance to complete the packet before we drop the packet. 964 * As a result, n+1 frags are actually allowed per packet, but 965 * only n will ever be stored. (n = maxfragsperpacket.) 966 * 967 */ 968 next = 0; 969 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 970 if (GETIP(q)->ip_off != next) { 971 if (fp->ipq_nfrags > V_maxfragsperpacket) { 972 V_ipstat.ips_fragdropped += fp->ipq_nfrags; 973 ip_freef(head, fp); 974 } 975 goto done; 976 } 977 next += GETIP(q)->ip_len; 978 } 979 /* Make sure the last packet didn't have the IP_MF flag */ 980 if (p->m_flags & M_FRAG) { 981 if (fp->ipq_nfrags > V_maxfragsperpacket) { 982 V_ipstat.ips_fragdropped += fp->ipq_nfrags; 983 ip_freef(head, fp); 984 } 985 goto done; 986 } 987 988 /* 989 * Reassembly is complete. Make sure the packet is a sane size. 990 */ 991 q = fp->ipq_frags; 992 ip = GETIP(q); 993 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 994 V_ipstat.ips_toolong++; 995 V_ipstat.ips_fragdropped += fp->ipq_nfrags; 996 ip_freef(head, fp); 997 goto done; 998 } 999 1000 /* 1001 * Concatenate fragments. 1002 */ 1003 m = q; 1004 t = m->m_next; 1005 m->m_next = NULL; 1006 m_cat(m, t); 1007 nq = q->m_nextpkt; 1008 q->m_nextpkt = NULL; 1009 for (q = nq; q != NULL; q = nq) { 1010 nq = q->m_nextpkt; 1011 q->m_nextpkt = NULL; 1012 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1013 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1014 m_cat(m, q); 1015 } 1016 /* 1017 * In order to do checksumming faster we do 'end-around carry' here 1018 * (and not in for{} loop), though it implies we are not going to 1019 * reassemble more than 64k fragments. 1020 */ 1021 m->m_pkthdr.csum_data = 1022 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1023 #ifdef MAC 1024 mac_ipq_reassemble(fp, m); 1025 mac_ipq_destroy(fp); 1026 #endif 1027 1028 /* 1029 * Create header for new ip packet by modifying header of first 1030 * packet; dequeue and discard fragment reassembly header. 1031 * Make header visible. 1032 */ 1033 ip->ip_len = (ip->ip_hl << 2) + next; 1034 ip->ip_src = fp->ipq_src; 1035 ip->ip_dst = fp->ipq_dst; 1036 TAILQ_REMOVE(head, fp, ipq_list); 1037 V_nipq--; 1038 uma_zfree(V_ipq_zone, fp); 1039 m->m_len += (ip->ip_hl << 2); 1040 m->m_data -= (ip->ip_hl << 2); 1041 /* some debugging cruft by sklower, below, will go away soon */ 1042 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1043 m_fixhdr(m); 1044 V_ipstat.ips_reassembled++; 1045 IPQ_UNLOCK(); 1046 return (m); 1047 1048 dropfrag: 1049 V_ipstat.ips_fragdropped++; 1050 if (fp != NULL) 1051 fp->ipq_nfrags--; 1052 m_freem(m); 1053 done: 1054 IPQ_UNLOCK(); 1055 return (NULL); 1056 1057 #undef GETIP 1058 } 1059 1060 /* 1061 * Free a fragment reassembly header and all 1062 * associated datagrams. 1063 */ 1064 static void 1065 ip_freef(struct ipqhead *fhp, struct ipq *fp) 1066 { 1067 struct mbuf *q; 1068 1069 IPQ_LOCK_ASSERT(); 1070 1071 while (fp->ipq_frags) { 1072 q = fp->ipq_frags; 1073 fp->ipq_frags = q->m_nextpkt; 1074 m_freem(q); 1075 } 1076 TAILQ_REMOVE(fhp, fp, ipq_list); 1077 uma_zfree(V_ipq_zone, fp); 1078 V_nipq--; 1079 } 1080 1081 /* 1082 * IP timer processing; 1083 * if a timer expires on a reassembly 1084 * queue, discard it. 1085 */ 1086 void 1087 ip_slowtimo(void) 1088 { 1089 struct ipq *fp; 1090 int i; 1091 1092 IPQ_LOCK(); 1093 for (i = 0; i < IPREASS_NHASH; i++) { 1094 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) { 1095 struct ipq *fpp; 1096 1097 fpp = fp; 1098 fp = TAILQ_NEXT(fp, ipq_list); 1099 if(--fpp->ipq_ttl == 0) { 1100 V_ipstat.ips_fragtimeout += fpp->ipq_nfrags; 1101 ip_freef(&V_ipq[i], fpp); 1102 } 1103 } 1104 } 1105 /* 1106 * If we are over the maximum number of fragments 1107 * (due to the limit being lowered), drain off 1108 * enough to get down to the new limit. 1109 */ 1110 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) { 1111 for (i = 0; i < IPREASS_NHASH; i++) { 1112 while (V_nipq > V_maxnipq && !TAILQ_EMPTY(&V_ipq[i])) { 1113 V_ipstat.ips_fragdropped += 1114 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags; 1115 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); 1116 } 1117 } 1118 } 1119 IPQ_UNLOCK(); 1120 } 1121 1122 /* 1123 * Drain off all datagram fragments. 1124 */ 1125 void 1126 ip_drain(void) 1127 { 1128 int i; 1129 1130 IPQ_LOCK(); 1131 for (i = 0; i < IPREASS_NHASH; i++) { 1132 while(!TAILQ_EMPTY(&V_ipq[i])) { 1133 V_ipstat.ips_fragdropped += 1134 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags; 1135 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); 1136 } 1137 } 1138 IPQ_UNLOCK(); 1139 in_rtqdrain(); 1140 } 1141 1142 /* 1143 * The protocol to be inserted into ip_protox[] must be already registered 1144 * in inetsw[], either statically or through pf_proto_register(). 1145 */ 1146 int 1147 ipproto_register(u_char ipproto) 1148 { 1149 struct protosw *pr; 1150 1151 /* Sanity checks. */ 1152 if (ipproto == 0) 1153 return (EPROTONOSUPPORT); 1154 1155 /* 1156 * The protocol slot must not be occupied by another protocol 1157 * already. An index pointing to IPPROTO_RAW is unused. 1158 */ 1159 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1160 if (pr == NULL) 1161 return (EPFNOSUPPORT); 1162 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1163 return (EEXIST); 1164 1165 /* Find the protocol position in inetsw[] and set the index. */ 1166 for (pr = inetdomain.dom_protosw; 1167 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1168 if (pr->pr_domain->dom_family == PF_INET && 1169 pr->pr_protocol && pr->pr_protocol == ipproto) { 1170 /* Be careful to only index valid IP protocols. */ 1171 if (pr->pr_protocol < IPPROTO_MAX) { 1172 ip_protox[pr->pr_protocol] = pr - inetsw; 1173 return (0); 1174 } else 1175 return (EINVAL); 1176 } 1177 } 1178 return (EPROTONOSUPPORT); 1179 } 1180 1181 int 1182 ipproto_unregister(u_char ipproto) 1183 { 1184 struct protosw *pr; 1185 1186 /* Sanity checks. */ 1187 if (ipproto == 0) 1188 return (EPROTONOSUPPORT); 1189 1190 /* Check if the protocol was indeed registered. */ 1191 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1192 if (pr == NULL) 1193 return (EPFNOSUPPORT); 1194 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1195 return (ENOENT); 1196 1197 /* Reset the protocol slot to IPPROTO_RAW. */ 1198 ip_protox[ipproto] = pr - inetsw; 1199 return (0); 1200 } 1201 1202 /* 1203 * Given address of next destination (final or next hop), 1204 * return internet address info of interface to be used to get there. 1205 */ 1206 struct in_ifaddr * 1207 ip_rtaddr(struct in_addr dst, u_int fibnum) 1208 { 1209 struct route sro; 1210 struct sockaddr_in *sin; 1211 struct in_ifaddr *ifa; 1212 1213 bzero(&sro, sizeof(sro)); 1214 sin = (struct sockaddr_in *)&sro.ro_dst; 1215 sin->sin_family = AF_INET; 1216 sin->sin_len = sizeof(*sin); 1217 sin->sin_addr = dst; 1218 in_rtalloc_ign(&sro, RTF_CLONING, fibnum); 1219 1220 if (sro.ro_rt == NULL) 1221 return (NULL); 1222 1223 ifa = ifatoia(sro.ro_rt->rt_ifa); 1224 RTFREE(sro.ro_rt); 1225 return (ifa); 1226 } 1227 1228 u_char inetctlerrmap[PRC_NCMDS] = { 1229 0, 0, 0, 0, 1230 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1231 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1232 EMSGSIZE, EHOSTUNREACH, 0, 0, 1233 0, 0, EHOSTUNREACH, 0, 1234 ENOPROTOOPT, ECONNREFUSED 1235 }; 1236 1237 /* 1238 * Forward a packet. If some error occurs return the sender 1239 * an icmp packet. Note we can't always generate a meaningful 1240 * icmp message because icmp doesn't have a large enough repertoire 1241 * of codes and types. 1242 * 1243 * If not forwarding, just drop the packet. This could be confusing 1244 * if ipforwarding was zero but some routing protocol was advancing 1245 * us as a gateway to somewhere. However, we must let the routing 1246 * protocol deal with that. 1247 * 1248 * The srcrt parameter indicates whether the packet is being forwarded 1249 * via a source route. 1250 */ 1251 void 1252 ip_forward(struct mbuf *m, int srcrt) 1253 { 1254 struct ip *ip = mtod(m, struct ip *); 1255 struct in_ifaddr *ia = NULL; 1256 struct mbuf *mcopy; 1257 struct in_addr dest; 1258 struct route ro; 1259 int error, type = 0, code = 0, mtu = 0; 1260 1261 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1262 V_ipstat.ips_cantforward++; 1263 m_freem(m); 1264 return; 1265 } 1266 #ifdef IPSTEALTH 1267 if (!V_ipstealth) { 1268 #endif 1269 if (ip->ip_ttl <= IPTTLDEC) { 1270 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1271 0, 0); 1272 return; 1273 } 1274 #ifdef IPSTEALTH 1275 } 1276 #endif 1277 1278 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); 1279 if (!srcrt && ia == NULL) { 1280 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1281 return; 1282 } 1283 1284 /* 1285 * Save the IP header and at most 8 bytes of the payload, 1286 * in case we need to generate an ICMP message to the src. 1287 * 1288 * XXX this can be optimized a lot by saving the data in a local 1289 * buffer on the stack (72 bytes at most), and only allocating the 1290 * mbuf if really necessary. The vast majority of the packets 1291 * are forwarded without having to send an ICMP back (either 1292 * because unnecessary, or because rate limited), so we are 1293 * really we are wasting a lot of work here. 1294 * 1295 * We don't use m_copy() because it might return a reference 1296 * to a shared cluster. Both this function and ip_output() 1297 * assume exclusive access to the IP header in `m', so any 1298 * data in a cluster may change before we reach icmp_error(). 1299 */ 1300 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1301 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1302 /* 1303 * It's probably ok if the pkthdr dup fails (because 1304 * the deep copy of the tag chain failed), but for now 1305 * be conservative and just discard the copy since 1306 * code below may some day want the tags. 1307 */ 1308 m_free(mcopy); 1309 mcopy = NULL; 1310 } 1311 if (mcopy != NULL) { 1312 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy)); 1313 mcopy->m_pkthdr.len = mcopy->m_len; 1314 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1315 } 1316 1317 #ifdef IPSTEALTH 1318 if (!V_ipstealth) { 1319 #endif 1320 ip->ip_ttl -= IPTTLDEC; 1321 #ifdef IPSTEALTH 1322 } 1323 #endif 1324 1325 /* 1326 * If forwarding packet using same interface that it came in on, 1327 * perhaps should send a redirect to sender to shortcut a hop. 1328 * Only send redirect if source is sending directly to us, 1329 * and if packet was not source routed (or has any options). 1330 * Also, don't send redirect if forwarding using a default route 1331 * or a route modified by a redirect. 1332 */ 1333 dest.s_addr = 0; 1334 if (!srcrt && V_ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) { 1335 struct sockaddr_in *sin; 1336 struct rtentry *rt; 1337 1338 bzero(&ro, sizeof(ro)); 1339 sin = (struct sockaddr_in *)&ro.ro_dst; 1340 sin->sin_family = AF_INET; 1341 sin->sin_len = sizeof(*sin); 1342 sin->sin_addr = ip->ip_dst; 1343 in_rtalloc_ign(&ro, RTF_CLONING, M_GETFIB(m)); 1344 1345 rt = ro.ro_rt; 1346 1347 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1348 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1349 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1350 u_long src = ntohl(ip->ip_src.s_addr); 1351 1352 if (RTA(rt) && 1353 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1354 if (rt->rt_flags & RTF_GATEWAY) 1355 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1356 else 1357 dest.s_addr = ip->ip_dst.s_addr; 1358 /* Router requirements says to only send host redirects */ 1359 type = ICMP_REDIRECT; 1360 code = ICMP_REDIRECT_HOST; 1361 } 1362 } 1363 if (rt) 1364 RTFREE(rt); 1365 } 1366 1367 /* 1368 * Try to cache the route MTU from ip_output so we can consider it for 1369 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. 1370 */ 1371 bzero(&ro, sizeof(ro)); 1372 1373 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1374 1375 if (error == EMSGSIZE && ro.ro_rt) 1376 mtu = ro.ro_rt->rt_rmx.rmx_mtu; 1377 if (ro.ro_rt) 1378 RTFREE(ro.ro_rt); 1379 1380 if (error) 1381 V_ipstat.ips_cantforward++; 1382 else { 1383 V_ipstat.ips_forward++; 1384 if (type) 1385 V_ipstat.ips_redirectsent++; 1386 else { 1387 if (mcopy) 1388 m_freem(mcopy); 1389 return; 1390 } 1391 } 1392 if (mcopy == NULL) 1393 return; 1394 1395 switch (error) { 1396 1397 case 0: /* forwarded, but need redirect */ 1398 /* type, code set above */ 1399 break; 1400 1401 case ENETUNREACH: /* shouldn't happen, checked above */ 1402 case EHOSTUNREACH: 1403 case ENETDOWN: 1404 case EHOSTDOWN: 1405 default: 1406 type = ICMP_UNREACH; 1407 code = ICMP_UNREACH_HOST; 1408 break; 1409 1410 case EMSGSIZE: 1411 type = ICMP_UNREACH; 1412 code = ICMP_UNREACH_NEEDFRAG; 1413 1414 #ifdef IPSEC 1415 /* 1416 * If IPsec is configured for this path, 1417 * override any possibly mtu value set by ip_output. 1418 */ 1419 mtu = ip_ipsec_mtu(m, mtu); 1420 #endif /* IPSEC */ 1421 /* 1422 * If the MTU was set before make sure we are below the 1423 * interface MTU. 1424 * If the MTU wasn't set before use the interface mtu or 1425 * fall back to the next smaller mtu step compared to the 1426 * current packet size. 1427 */ 1428 if (mtu != 0) { 1429 if (ia != NULL) 1430 mtu = min(mtu, ia->ia_ifp->if_mtu); 1431 } else { 1432 if (ia != NULL) 1433 mtu = ia->ia_ifp->if_mtu; 1434 else 1435 mtu = ip_next_mtu(ip->ip_len, 0); 1436 } 1437 V_ipstat.ips_cantfrag++; 1438 break; 1439 1440 case ENOBUFS: 1441 /* 1442 * A router should not generate ICMP_SOURCEQUENCH as 1443 * required in RFC1812 Requirements for IP Version 4 Routers. 1444 * Source quench could be a big problem under DoS attacks, 1445 * or if the underlying interface is rate-limited. 1446 * Those who need source quench packets may re-enable them 1447 * via the net.inet.ip.sendsourcequench sysctl. 1448 */ 1449 if (V_ip_sendsourcequench == 0) { 1450 m_freem(mcopy); 1451 return; 1452 } else { 1453 type = ICMP_SOURCEQUENCH; 1454 code = 0; 1455 } 1456 break; 1457 1458 case EACCES: /* ipfw denied packet */ 1459 m_freem(mcopy); 1460 return; 1461 } 1462 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1463 } 1464 1465 void 1466 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1467 struct mbuf *m) 1468 { 1469 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1470 struct bintime bt; 1471 1472 bintime(&bt); 1473 if (inp->inp_socket->so_options & SO_BINTIME) { 1474 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1475 SCM_BINTIME, SOL_SOCKET); 1476 if (*mp) 1477 mp = &(*mp)->m_next; 1478 } 1479 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1480 struct timeval tv; 1481 1482 bintime2timeval(&bt, &tv); 1483 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1484 SCM_TIMESTAMP, SOL_SOCKET); 1485 if (*mp) 1486 mp = &(*mp)->m_next; 1487 } 1488 } 1489 if (inp->inp_flags & INP_RECVDSTADDR) { 1490 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1491 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1492 if (*mp) 1493 mp = &(*mp)->m_next; 1494 } 1495 if (inp->inp_flags & INP_RECVTTL) { 1496 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1497 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1498 if (*mp) 1499 mp = &(*mp)->m_next; 1500 } 1501 #ifdef notyet 1502 /* XXX 1503 * Moving these out of udp_input() made them even more broken 1504 * than they already were. 1505 */ 1506 /* options were tossed already */ 1507 if (inp->inp_flags & INP_RECVOPTS) { 1508 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1509 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1510 if (*mp) 1511 mp = &(*mp)->m_next; 1512 } 1513 /* ip_srcroute doesn't do what we want here, need to fix */ 1514 if (inp->inp_flags & INP_RECVRETOPTS) { 1515 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1516 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1517 if (*mp) 1518 mp = &(*mp)->m_next; 1519 } 1520 #endif 1521 if (inp->inp_flags & INP_RECVIF) { 1522 struct ifnet *ifp; 1523 struct sdlbuf { 1524 struct sockaddr_dl sdl; 1525 u_char pad[32]; 1526 } sdlbuf; 1527 struct sockaddr_dl *sdp; 1528 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1529 1530 if (((ifp = m->m_pkthdr.rcvif)) 1531 && ( ifp->if_index && (ifp->if_index <= V_if_index))) { 1532 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1533 /* 1534 * Change our mind and don't try copy. 1535 */ 1536 if ((sdp->sdl_family != AF_LINK) 1537 || (sdp->sdl_len > sizeof(sdlbuf))) { 1538 goto makedummy; 1539 } 1540 bcopy(sdp, sdl2, sdp->sdl_len); 1541 } else { 1542 makedummy: 1543 sdl2->sdl_len 1544 = offsetof(struct sockaddr_dl, sdl_data[0]); 1545 sdl2->sdl_family = AF_LINK; 1546 sdl2->sdl_index = 0; 1547 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1548 } 1549 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1550 IP_RECVIF, IPPROTO_IP); 1551 if (*mp) 1552 mp = &(*mp)->m_next; 1553 } 1554 } 1555 1556 /* 1557 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1558 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1559 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1560 * compiled. 1561 */ 1562 static int ip_rsvp_on; 1563 struct socket *ip_rsvpd; 1564 int 1565 ip_rsvp_init(struct socket *so) 1566 { 1567 if (so->so_type != SOCK_RAW || 1568 so->so_proto->pr_protocol != IPPROTO_RSVP) 1569 return EOPNOTSUPP; 1570 1571 if (V_ip_rsvpd != NULL) 1572 return EADDRINUSE; 1573 1574 V_ip_rsvpd = so; 1575 /* 1576 * This may seem silly, but we need to be sure we don't over-increment 1577 * the RSVP counter, in case something slips up. 1578 */ 1579 if (!V_ip_rsvp_on) { 1580 V_ip_rsvp_on = 1; 1581 V_rsvp_on++; 1582 } 1583 1584 return 0; 1585 } 1586 1587 int 1588 ip_rsvp_done(void) 1589 { 1590 V_ip_rsvpd = NULL; 1591 /* 1592 * This may seem silly, but we need to be sure we don't over-decrement 1593 * the RSVP counter, in case something slips up. 1594 */ 1595 if (V_ip_rsvp_on) { 1596 V_ip_rsvp_on = 0; 1597 V_rsvp_on--; 1598 } 1599 return 0; 1600 } 1601 1602 void 1603 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1604 { 1605 if (rsvp_input_p) { /* call the real one if loaded */ 1606 rsvp_input_p(m, off); 1607 return; 1608 } 1609 1610 /* Can still get packets with rsvp_on = 0 if there is a local member 1611 * of the group to which the RSVP packet is addressed. But in this 1612 * case we want to throw the packet away. 1613 */ 1614 1615 if (!V_rsvp_on) { 1616 m_freem(m); 1617 return; 1618 } 1619 1620 if (V_ip_rsvpd != NULL) { 1621 rip_input(m, off); 1622 return; 1623 } 1624 /* Drop the packet */ 1625 m_freem(m); 1626 } 1627