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_route.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/lock.h> 53 #include <sys/rwlock.h> 54 #include <sys/syslog.h> 55 #include <sys/sysctl.h> 56 #include <sys/vimage.h> 57 58 #include <net/pfil.h> 59 #include <net/if.h> 60 #include <net/if_types.h> 61 #include <net/if_var.h> 62 #include <net/if_dl.h> 63 #include <net/route.h> 64 #include <net/netisr.h> 65 #include <net/vnet.h> 66 #include <net/flowtable.h> 67 68 #include <netinet/in.h> 69 #include <netinet/in_systm.h> 70 #include <netinet/in_var.h> 71 #include <netinet/ip.h> 72 #include <netinet/in_pcb.h> 73 #include <netinet/ip_var.h> 74 #include <netinet/ip_fw.h> 75 #include <netinet/ip_icmp.h> 76 #include <netinet/ip_options.h> 77 #include <machine/in_cksum.h> 78 #ifdef DEV_CARP 79 #include <netinet/ip_carp.h> 80 #endif 81 #ifdef IPSEC 82 #include <netinet/ip_ipsec.h> 83 #endif /* IPSEC */ 84 85 #include <sys/socketvar.h> 86 87 #include <security/mac/mac_framework.h> 88 89 #ifdef CTASSERT 90 CTASSERT(sizeof(struct ip) == 20); 91 #endif 92 93 static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */ 94 static VNET_DEFINE(int, ip_checkinterface); 95 static VNET_DEFINE(int, ip_keepfaith); 96 static VNET_DEFINE(int, ip_sendsourcequench); 97 98 #define V_ipsendredirects VNET(ipsendredirects) 99 #define V_ip_checkinterface VNET(ip_checkinterface) 100 #define V_ip_keepfaith VNET(ip_keepfaith) 101 #define V_ip_sendsourcequench VNET(ip_sendsourcequench) 102 103 VNET_DEFINE(int, ip_defttl) = IPDEFTTL; 104 VNET_DEFINE(int, ip_do_randomid); 105 VNET_DEFINE(int, ipforwarding); 106 107 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */ 108 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */ 109 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */ 110 VNET_DEFINE(struct ipstat, ipstat); 111 112 static VNET_DEFINE(int, ip_rsvp_on); 113 VNET_DEFINE(struct socket *, ip_rsvpd); 114 VNET_DEFINE(int, rsvp_on); 115 116 #define V_ip_rsvp_on VNET(ip_rsvp_on) 117 118 static VNET_DEFINE(TAILQ_HEAD(ipqhead, ipq), ipq[IPREASS_NHASH]); 119 static VNET_DEFINE(int, maxnipq); /* Administrative limit on # reass queues. */ 120 static VNET_DEFINE(int, maxfragsperpacket); 121 static VNET_DEFINE(int, nipq); /* Total # of reass queues */ 122 123 #define V_ipq VNET(ipq) 124 #define V_maxnipq VNET(maxnipq) 125 #define V_maxfragsperpacket VNET(maxfragsperpacket) 126 #define V_nipq VNET(nipq) 127 128 VNET_DEFINE(int, ipstealth); 129 130 struct rwlock in_ifaddr_lock; 131 RW_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock"); 132 133 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 134 &VNET_NAME(ipforwarding), 0, 135 "Enable IP forwarding between interfaces"); 136 137 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 138 &VNET_NAME(ipsendredirects), 0, 139 "Enable sending IP redirects"); 140 141 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 142 &VNET_NAME(ip_defttl), 0, 143 "Maximum TTL on IP packets"); 144 145 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 146 &VNET_NAME(ip_keepfaith), 0, 147 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 148 149 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 150 &VNET_NAME(ip_sendsourcequench), 0, 151 "Enable the transmission of source quench packets"); 152 153 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 154 &VNET_NAME(ip_do_randomid), 0, 155 "Assign random ip_id values"); 156 157 /* 158 * XXX - Setting ip_checkinterface mostly implements the receive side of 159 * the Strong ES model described in RFC 1122, but since the routing table 160 * and transmit implementation do not implement the Strong ES model, 161 * setting this to 1 results in an odd hybrid. 162 * 163 * XXX - ip_checkinterface currently must be disabled if you use ipnat 164 * to translate the destination address to another local interface. 165 * 166 * XXX - ip_checkinterface must be disabled if you add IP aliases 167 * to the loopback interface instead of the interface where the 168 * packets for those addresses are received. 169 */ 170 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 171 &VNET_NAME(ip_checkinterface), 0, 172 "Verify packet arrives on correct interface"); 173 174 struct pfil_head inet_pfil_hook; /* Packet filter hooks */ 175 176 static struct netisr_handler ip_nh = { 177 .nh_name = "ip", 178 .nh_handler = ip_input, 179 .nh_proto = NETISR_IP, 180 .nh_policy = NETISR_POLICY_FLOW, 181 }; 182 183 extern struct domain inetdomain; 184 extern struct protosw inetsw[]; 185 u_char ip_protox[IPPROTO_MAX]; 186 187 SYSCTL_VNET_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 188 &VNET_NAME(ipstat), ipstat, 189 "IP statistics (struct ipstat, netinet/ip_var.h)"); 190 191 static VNET_DEFINE(uma_zone_t, ipq_zone); 192 #define V_ipq_zone VNET(ipq_zone) 193 194 static struct mtx ipqlock; 195 196 #define IPQ_LOCK() mtx_lock(&ipqlock) 197 #define IPQ_UNLOCK() mtx_unlock(&ipqlock) 198 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 199 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 200 201 static void maxnipq_update(void); 202 static void ipq_zone_change(void *); 203 204 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, 205 &VNET_NAME(nipq), 0, 206 "Current number of IPv4 fragment reassembly queue entries"); 207 208 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 209 &VNET_NAME(maxfragsperpacket), 0, 210 "Maximum number of IPv4 fragments allowed per packet"); 211 212 struct callout ipport_tick_callout; 213 214 #ifdef IPCTL_DEFMTU 215 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 216 &ip_mtu, 0, "Default MTU"); 217 #endif 218 219 #ifdef IPSTEALTH 220 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 221 &VNET_NAME(ipstealth), 0, 222 "IP stealth mode, no TTL decrementation on forwarding"); 223 #endif 224 225 #ifdef FLOWTABLE 226 static VNET_DEFINE(int, ip_output_flowtable_size) = 2048; 227 VNET_DEFINE(struct flowtable *, ip_ft); 228 #define V_ip_output_flowtable_size VNET(ip_output_flowtable_size) 229 230 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, output_flowtable_size, CTLFLAG_RDTUN, 231 &VNET_NAME(ip_output_flowtable_size), 2048, 232 "number of entries in the per-cpu output flow caches"); 233 #endif 234 235 VNET_DEFINE(int, fw_one_pass) = 1; 236 237 static void ip_freef(struct ipqhead *, struct ipq *); 238 239 static int 240 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 241 { 242 int error, qlimit; 243 244 netisr_getqlimit(&ip_nh, &qlimit); 245 error = sysctl_handle_int(oidp, &qlimit, 0, req); 246 if (error || !req->newptr) 247 return (error); 248 if (qlimit < 1) 249 return (EINVAL); 250 return (netisr_setqlimit(&ip_nh, qlimit)); 251 } 252 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, 253 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I", 254 "Maximum size of the IP input queue"); 255 256 static int 257 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) 258 { 259 u_int64_t qdrops_long; 260 int error, qdrops; 261 262 netisr_getqdrops(&ip_nh, &qdrops_long); 263 qdrops = qdrops_long; 264 error = sysctl_handle_int(oidp, &qdrops, 0, req); 265 if (error || !req->newptr) 266 return (error); 267 if (qdrops != 0) 268 return (EINVAL); 269 netisr_clearqdrops(&ip_nh); 270 return (0); 271 } 272 273 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, 274 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I", 275 "Number of packets dropped from the IP input queue"); 276 277 /* 278 * IP initialization: fill in IP protocol switch table. 279 * All protocols not implemented in kernel go to raw IP protocol handler. 280 */ 281 void 282 ip_init(void) 283 { 284 struct protosw *pr; 285 int i; 286 287 V_ip_id = time_second & 0xffff; 288 289 TAILQ_INIT(&V_in_ifaddrhead); 290 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 291 292 /* Initialize IP reassembly queue. */ 293 for (i = 0; i < IPREASS_NHASH; i++) 294 TAILQ_INIT(&V_ipq[i]); 295 V_maxnipq = nmbclusters / 32; 296 V_maxfragsperpacket = 16; 297 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 298 NULL, UMA_ALIGN_PTR, 0); 299 maxnipq_update(); 300 301 #ifdef FLOWTABLE 302 TUNABLE_INT_FETCH("net.inet.ip.output_flowtable_size", 303 &V_ip_output_flowtable_size); 304 V_ip_ft = flowtable_alloc(V_ip_output_flowtable_size, FL_PCPU); 305 #endif 306 307 /* Skip initialization of globals for non-default instances. */ 308 if (!IS_DEFAULT_VNET(curvnet)) 309 return; 310 311 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 312 if (pr == NULL) 313 panic("ip_init: PF_INET not found"); 314 315 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 316 for (i = 0; i < IPPROTO_MAX; i++) 317 ip_protox[i] = pr - inetsw; 318 /* 319 * Cycle through IP protocols and put them into the appropriate place 320 * in ip_protox[]. 321 */ 322 for (pr = inetdomain.dom_protosw; 323 pr < inetdomain.dom_protoswNPROTOSW; pr++) 324 if (pr->pr_domain->dom_family == PF_INET && 325 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 326 /* Be careful to only index valid IP protocols. */ 327 if (pr->pr_protocol < IPPROTO_MAX) 328 ip_protox[pr->pr_protocol] = pr - inetsw; 329 } 330 331 /* Initialize packet filter hooks. */ 332 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 333 inet_pfil_hook.ph_af = AF_INET; 334 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 335 printf("%s: WARNING: unable to register pfil hook, " 336 "error %d\n", __func__, i); 337 338 /* Start ipport_tick. */ 339 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE); 340 callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL); 341 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, 342 SHUTDOWN_PRI_DEFAULT); 343 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, 344 NULL, EVENTHANDLER_PRI_ANY); 345 346 /* Initialize various other remaining things. */ 347 IPQ_LOCK_INIT(); 348 netisr_register(&ip_nh); 349 } 350 351 void 352 ip_fini(void *xtp) 353 { 354 355 callout_stop(&ipport_tick_callout); 356 } 357 358 /* 359 * Ip input routine. Checksum and byte swap header. If fragmented 360 * try to reassemble. Process options. Pass to next level. 361 */ 362 void 363 ip_input(struct mbuf *m) 364 { 365 struct ip *ip = NULL; 366 struct in_ifaddr *ia = NULL; 367 struct ifaddr *ifa; 368 struct ifnet *ifp; 369 int checkif, hlen = 0; 370 u_short sum; 371 int dchg = 0; /* dest changed after fw */ 372 struct in_addr odst; /* original dst address */ 373 374 M_ASSERTPKTHDR(m); 375 376 if (m->m_flags & M_FASTFWD_OURS) { 377 /* 378 * Firewall or NAT changed destination to local. 379 * We expect ip_len and ip_off to be in host byte order. 380 */ 381 m->m_flags &= ~M_FASTFWD_OURS; 382 /* Set up some basics that will be used later. */ 383 ip = mtod(m, struct ip *); 384 hlen = ip->ip_hl << 2; 385 goto ours; 386 } 387 388 IPSTAT_INC(ips_total); 389 390 if (m->m_pkthdr.len < sizeof(struct ip)) 391 goto tooshort; 392 393 if (m->m_len < sizeof (struct ip) && 394 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 395 IPSTAT_INC(ips_toosmall); 396 return; 397 } 398 ip = mtod(m, struct ip *); 399 400 if (ip->ip_v != IPVERSION) { 401 IPSTAT_INC(ips_badvers); 402 goto bad; 403 } 404 405 hlen = ip->ip_hl << 2; 406 if (hlen < sizeof(struct ip)) { /* minimum header length */ 407 IPSTAT_INC(ips_badhlen); 408 goto bad; 409 } 410 if (hlen > m->m_len) { 411 if ((m = m_pullup(m, hlen)) == NULL) { 412 IPSTAT_INC(ips_badhlen); 413 return; 414 } 415 ip = mtod(m, struct ip *); 416 } 417 418 /* 127/8 must not appear on wire - RFC1122 */ 419 ifp = m->m_pkthdr.rcvif; 420 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 421 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 422 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 423 IPSTAT_INC(ips_badaddr); 424 goto bad; 425 } 426 } 427 428 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 429 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 430 } else { 431 if (hlen == sizeof(struct ip)) { 432 sum = in_cksum_hdr(ip); 433 } else { 434 sum = in_cksum(m, hlen); 435 } 436 } 437 if (sum) { 438 IPSTAT_INC(ips_badsum); 439 goto bad; 440 } 441 442 #ifdef ALTQ 443 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 444 /* packet is dropped by traffic conditioner */ 445 return; 446 #endif 447 448 /* 449 * Convert fields to host representation. 450 */ 451 ip->ip_len = ntohs(ip->ip_len); 452 if (ip->ip_len < hlen) { 453 IPSTAT_INC(ips_badlen); 454 goto bad; 455 } 456 ip->ip_off = ntohs(ip->ip_off); 457 458 /* 459 * Check that the amount of data in the buffers 460 * is as at least much as the IP header would have us expect. 461 * Trim mbufs if longer than we expect. 462 * Drop packet if shorter than we expect. 463 */ 464 if (m->m_pkthdr.len < ip->ip_len) { 465 tooshort: 466 IPSTAT_INC(ips_tooshort); 467 goto bad; 468 } 469 if (m->m_pkthdr.len > ip->ip_len) { 470 if (m->m_len == m->m_pkthdr.len) { 471 m->m_len = ip->ip_len; 472 m->m_pkthdr.len = ip->ip_len; 473 } else 474 m_adj(m, ip->ip_len - m->m_pkthdr.len); 475 } 476 #ifdef IPSEC 477 /* 478 * Bypass packet filtering for packets from a tunnel (gif). 479 */ 480 if (ip_ipsec_filtertunnel(m)) 481 goto passin; 482 #endif /* IPSEC */ 483 484 /* 485 * Run through list of hooks for input packets. 486 * 487 * NB: Beware of the destination address changing (e.g. 488 * by NAT rewriting). When this happens, tell 489 * ip_forward to do the right thing. 490 */ 491 492 /* Jump over all PFIL processing if hooks are not active. */ 493 if (!PFIL_HOOKED(&inet_pfil_hook)) 494 goto passin; 495 496 odst = ip->ip_dst; 497 if (pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0) 498 return; 499 if (m == NULL) /* consumed by filter */ 500 return; 501 502 ip = mtod(m, struct ip *); 503 dchg = (odst.s_addr != ip->ip_dst.s_addr); 504 ifp = m->m_pkthdr.rcvif; 505 506 #ifdef IPFIREWALL_FORWARD 507 if (m->m_flags & M_FASTFWD_OURS) { 508 m->m_flags &= ~M_FASTFWD_OURS; 509 goto ours; 510 } 511 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 512 /* 513 * Directly ship on the packet. This allows to forward packets 514 * that were destined for us to some other directly connected 515 * host. 516 */ 517 ip_forward(m, dchg); 518 return; 519 } 520 #endif /* IPFIREWALL_FORWARD */ 521 522 passin: 523 /* 524 * Process options and, if not destined for us, 525 * ship it on. ip_dooptions returns 1 when an 526 * error was detected (causing an icmp message 527 * to be sent and the original packet to be freed). 528 */ 529 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 530 return; 531 532 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 533 * matter if it is destined to another node, or whether it is 534 * a multicast one, RSVP wants it! and prevents it from being forwarded 535 * anywhere else. Also checks if the rsvp daemon is running before 536 * grabbing the packet. 537 */ 538 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 539 goto ours; 540 541 /* 542 * Check our list of addresses, to see if the packet is for us. 543 * If we don't have any addresses, assume any unicast packet 544 * we receive might be for us (and let the upper layers deal 545 * with it). 546 */ 547 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 548 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 549 goto ours; 550 551 /* 552 * Enable a consistency check between the destination address 553 * and the arrival interface for a unicast packet (the RFC 1122 554 * strong ES model) if IP forwarding is disabled and the packet 555 * is not locally generated and the packet is not subject to 556 * 'ipfw fwd'. 557 * 558 * XXX - Checking also should be disabled if the destination 559 * address is ipnat'ed to a different interface. 560 * 561 * XXX - Checking is incompatible with IP aliases added 562 * to the loopback interface instead of the interface where 563 * the packets are received. 564 * 565 * XXX - This is the case for carp vhost IPs as well so we 566 * insert a workaround. If the packet got here, we already 567 * checked with carp_iamatch() and carp_forus(). 568 */ 569 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 570 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && 571 #ifdef DEV_CARP 572 !ifp->if_carp && 573 #endif 574 (dchg == 0); 575 576 /* 577 * Check for exact addresses in the hash bucket. 578 */ 579 /* IN_IFADDR_RLOCK(); */ 580 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 581 /* 582 * If the address matches, verify that the packet 583 * arrived via the correct interface if checking is 584 * enabled. 585 */ 586 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 587 (!checkif || ia->ia_ifp == ifp)) { 588 ifa_ref(&ia->ia_ifa); 589 /* IN_IFADDR_RUNLOCK(); */ 590 goto ours; 591 } 592 } 593 /* IN_IFADDR_RUNLOCK(); */ 594 595 /* 596 * Check for broadcast addresses. 597 * 598 * Only accept broadcast packets that arrive via the matching 599 * interface. Reception of forwarded directed broadcasts would 600 * be handled via ip_forward() and ether_output() with the loopback 601 * into the stack for SIMPLEX interfaces handled by ether_output(). 602 */ 603 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 604 IF_ADDR_LOCK(ifp); 605 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 606 if (ifa->ifa_addr->sa_family != AF_INET) 607 continue; 608 ia = ifatoia(ifa); 609 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 610 ip->ip_dst.s_addr) { 611 ifa_ref(ifa); 612 IF_ADDR_UNLOCK(ifp); 613 goto ours; 614 } 615 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) { 616 ifa_ref(ifa); 617 IF_ADDR_UNLOCK(ifp); 618 goto ours; 619 } 620 #ifdef BOOTP_COMPAT 621 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { 622 ifa_ref(ifa); 623 IF_ADDR_UNLOCK(ifp); 624 goto ours; 625 } 626 #endif 627 } 628 IF_ADDR_UNLOCK(ifp); 629 ia = NULL; 630 } 631 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 632 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 633 IPSTAT_INC(ips_cantforward); 634 m_freem(m); 635 return; 636 } 637 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 638 if (V_ip_mrouter) { 639 /* 640 * If we are acting as a multicast router, all 641 * incoming multicast packets are passed to the 642 * kernel-level multicast forwarding function. 643 * The packet is returned (relatively) intact; if 644 * ip_mforward() returns a non-zero value, the packet 645 * must be discarded, else it may be accepted below. 646 */ 647 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 648 IPSTAT_INC(ips_cantforward); 649 m_freem(m); 650 return; 651 } 652 653 /* 654 * The process-level routing daemon needs to receive 655 * all multicast IGMP packets, whether or not this 656 * host belongs to their destination groups. 657 */ 658 if (ip->ip_p == IPPROTO_IGMP) 659 goto ours; 660 IPSTAT_INC(ips_forward); 661 } 662 /* 663 * Assume the packet is for us, to avoid prematurely taking 664 * a lock on the in_multi hash. Protocols must perform 665 * their own filtering and update statistics accordingly. 666 */ 667 goto ours; 668 } 669 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 670 goto ours; 671 if (ip->ip_dst.s_addr == INADDR_ANY) 672 goto ours; 673 674 /* 675 * FAITH(Firewall Aided Internet Translator) 676 */ 677 if (ifp && ifp->if_type == IFT_FAITH) { 678 if (V_ip_keepfaith) { 679 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 680 goto ours; 681 } 682 m_freem(m); 683 return; 684 } 685 686 /* 687 * Not for us; forward if possible and desirable. 688 */ 689 if (V_ipforwarding == 0) { 690 IPSTAT_INC(ips_cantforward); 691 m_freem(m); 692 } else { 693 #ifdef IPSEC 694 if (ip_ipsec_fwd(m)) 695 goto bad; 696 #endif /* IPSEC */ 697 ip_forward(m, dchg); 698 } 699 return; 700 701 ours: 702 #ifdef IPSTEALTH 703 /* 704 * IPSTEALTH: Process non-routing options only 705 * if the packet is destined for us. 706 */ 707 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) { 708 if (ia != NULL) 709 ifa_free(&ia->ia_ifa); 710 return; 711 } 712 #endif /* IPSTEALTH */ 713 714 /* Count the packet in the ip address stats */ 715 if (ia != NULL) { 716 ia->ia_ifa.if_ipackets++; 717 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 718 ifa_free(&ia->ia_ifa); 719 } 720 721 /* 722 * Attempt reassembly; if it succeeds, proceed. 723 * ip_reass() will return a different mbuf. 724 */ 725 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 726 m = ip_reass(m); 727 if (m == NULL) 728 return; 729 ip = mtod(m, struct ip *); 730 /* Get the header length of the reassembled packet */ 731 hlen = ip->ip_hl << 2; 732 } 733 734 /* 735 * Further protocols expect the packet length to be w/o the 736 * IP header. 737 */ 738 ip->ip_len -= hlen; 739 740 #ifdef IPSEC 741 /* 742 * enforce IPsec policy checking if we are seeing last header. 743 * note that we do not visit this with protocols with pcb layer 744 * code - like udp/tcp/raw ip. 745 */ 746 if (ip_ipsec_input(m)) 747 goto bad; 748 #endif /* IPSEC */ 749 750 /* 751 * Switch out to protocol's input routine. 752 */ 753 IPSTAT_INC(ips_delivered); 754 755 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 756 return; 757 bad: 758 m_freem(m); 759 } 760 761 /* 762 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 763 * max has slightly different semantics than the sysctl, for historical 764 * reasons. 765 */ 766 static void 767 maxnipq_update(void) 768 { 769 770 /* 771 * -1 for unlimited allocation. 772 */ 773 if (V_maxnipq < 0) 774 uma_zone_set_max(V_ipq_zone, 0); 775 /* 776 * Positive number for specific bound. 777 */ 778 if (V_maxnipq > 0) 779 uma_zone_set_max(V_ipq_zone, V_maxnipq); 780 /* 781 * Zero specifies no further fragment queue allocation -- set the 782 * bound very low, but rely on implementation elsewhere to actually 783 * prevent allocation and reclaim current queues. 784 */ 785 if (V_maxnipq == 0) 786 uma_zone_set_max(V_ipq_zone, 1); 787 } 788 789 static void 790 ipq_zone_change(void *tag) 791 { 792 793 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) { 794 V_maxnipq = nmbclusters / 32; 795 maxnipq_update(); 796 } 797 } 798 799 static int 800 sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 801 { 802 int error, i; 803 804 i = V_maxnipq; 805 error = sysctl_handle_int(oidp, &i, 0, req); 806 if (error || !req->newptr) 807 return (error); 808 809 /* 810 * XXXRW: Might be a good idea to sanity check the argument and place 811 * an extreme upper bound. 812 */ 813 if (i < -1) 814 return (EINVAL); 815 V_maxnipq = i; 816 maxnipq_update(); 817 return (0); 818 } 819 820 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 821 NULL, 0, sysctl_maxnipq, "I", 822 "Maximum number of IPv4 fragment reassembly queue entries"); 823 824 /* 825 * Take incoming datagram fragment and try to reassemble it into 826 * whole datagram. If the argument is the first fragment or one 827 * in between the function will return NULL and store the mbuf 828 * in the fragment chain. If the argument is the last fragment 829 * the packet will be reassembled and the pointer to the new 830 * mbuf returned for further processing. Only m_tags attached 831 * to the first packet/fragment are preserved. 832 * The IP header is *NOT* adjusted out of iplen. 833 */ 834 struct mbuf * 835 ip_reass(struct mbuf *m) 836 { 837 struct ip *ip; 838 struct mbuf *p, *q, *nq, *t; 839 struct ipq *fp = NULL; 840 struct ipqhead *head; 841 int i, hlen, next; 842 u_int8_t ecn, ecn0; 843 u_short hash; 844 845 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 846 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) { 847 IPSTAT_INC(ips_fragments); 848 IPSTAT_INC(ips_fragdropped); 849 m_freem(m); 850 return (NULL); 851 } 852 853 ip = mtod(m, struct ip *); 854 hlen = ip->ip_hl << 2; 855 856 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 857 head = &V_ipq[hash]; 858 IPQ_LOCK(); 859 860 /* 861 * Look for queue of fragments 862 * of this datagram. 863 */ 864 TAILQ_FOREACH(fp, head, ipq_list) 865 if (ip->ip_id == fp->ipq_id && 866 ip->ip_src.s_addr == fp->ipq_src.s_addr && 867 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 868 #ifdef MAC 869 mac_ipq_match(m, fp) && 870 #endif 871 ip->ip_p == fp->ipq_p) 872 goto found; 873 874 fp = NULL; 875 876 /* 877 * Attempt to trim the number of allocated fragment queues if it 878 * exceeds the administrative limit. 879 */ 880 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) { 881 /* 882 * drop something from the tail of the current queue 883 * before proceeding further 884 */ 885 struct ipq *q = TAILQ_LAST(head, ipqhead); 886 if (q == NULL) { /* gak */ 887 for (i = 0; i < IPREASS_NHASH; i++) { 888 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead); 889 if (r) { 890 IPSTAT_ADD(ips_fragtimeout, 891 r->ipq_nfrags); 892 ip_freef(&V_ipq[i], r); 893 break; 894 } 895 } 896 } else { 897 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags); 898 ip_freef(head, q); 899 } 900 } 901 902 found: 903 /* 904 * Adjust ip_len to not reflect header, 905 * convert offset of this to bytes. 906 */ 907 ip->ip_len -= hlen; 908 if (ip->ip_off & IP_MF) { 909 /* 910 * Make sure that fragments have a data length 911 * that's a non-zero multiple of 8 bytes. 912 */ 913 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 914 IPSTAT_INC(ips_toosmall); /* XXX */ 915 goto dropfrag; 916 } 917 m->m_flags |= M_FRAG; 918 } else 919 m->m_flags &= ~M_FRAG; 920 ip->ip_off <<= 3; 921 922 923 /* 924 * Attempt reassembly; if it succeeds, proceed. 925 * ip_reass() will return a different mbuf. 926 */ 927 IPSTAT_INC(ips_fragments); 928 m->m_pkthdr.header = ip; 929 930 /* Previous ip_reass() started here. */ 931 /* 932 * Presence of header sizes in mbufs 933 * would confuse code below. 934 */ 935 m->m_data += hlen; 936 m->m_len -= hlen; 937 938 /* 939 * If first fragment to arrive, create a reassembly queue. 940 */ 941 if (fp == NULL) { 942 fp = uma_zalloc(V_ipq_zone, M_NOWAIT); 943 if (fp == NULL) 944 goto dropfrag; 945 #ifdef MAC 946 if (mac_ipq_init(fp, M_NOWAIT) != 0) { 947 uma_zfree(V_ipq_zone, fp); 948 fp = NULL; 949 goto dropfrag; 950 } 951 mac_ipq_create(m, fp); 952 #endif 953 TAILQ_INSERT_HEAD(head, fp, ipq_list); 954 V_nipq++; 955 fp->ipq_nfrags = 1; 956 fp->ipq_ttl = IPFRAGTTL; 957 fp->ipq_p = ip->ip_p; 958 fp->ipq_id = ip->ip_id; 959 fp->ipq_src = ip->ip_src; 960 fp->ipq_dst = ip->ip_dst; 961 fp->ipq_frags = m; 962 m->m_nextpkt = NULL; 963 goto done; 964 } else { 965 fp->ipq_nfrags++; 966 #ifdef MAC 967 mac_ipq_update(m, fp); 968 #endif 969 } 970 971 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 972 973 /* 974 * Handle ECN by comparing this segment with the first one; 975 * if CE is set, do not lose CE. 976 * drop if CE and not-ECT are mixed for the same packet. 977 */ 978 ecn = ip->ip_tos & IPTOS_ECN_MASK; 979 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 980 if (ecn == IPTOS_ECN_CE) { 981 if (ecn0 == IPTOS_ECN_NOTECT) 982 goto dropfrag; 983 if (ecn0 != IPTOS_ECN_CE) 984 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 985 } 986 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 987 goto dropfrag; 988 989 /* 990 * Find a segment which begins after this one does. 991 */ 992 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 993 if (GETIP(q)->ip_off > ip->ip_off) 994 break; 995 996 /* 997 * If there is a preceding segment, it may provide some of 998 * our data already. If so, drop the data from the incoming 999 * segment. If it provides all of our data, drop us, otherwise 1000 * stick new segment in the proper place. 1001 * 1002 * If some of the data is dropped from the the preceding 1003 * segment, then it's checksum is invalidated. 1004 */ 1005 if (p) { 1006 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1007 if (i > 0) { 1008 if (i >= ip->ip_len) 1009 goto dropfrag; 1010 m_adj(m, i); 1011 m->m_pkthdr.csum_flags = 0; 1012 ip->ip_off += i; 1013 ip->ip_len -= i; 1014 } 1015 m->m_nextpkt = p->m_nextpkt; 1016 p->m_nextpkt = m; 1017 } else { 1018 m->m_nextpkt = fp->ipq_frags; 1019 fp->ipq_frags = m; 1020 } 1021 1022 /* 1023 * While we overlap succeeding segments trim them or, 1024 * if they are completely covered, dequeue them. 1025 */ 1026 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1027 q = nq) { 1028 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1029 if (i < GETIP(q)->ip_len) { 1030 GETIP(q)->ip_len -= i; 1031 GETIP(q)->ip_off += i; 1032 m_adj(q, i); 1033 q->m_pkthdr.csum_flags = 0; 1034 break; 1035 } 1036 nq = q->m_nextpkt; 1037 m->m_nextpkt = nq; 1038 IPSTAT_INC(ips_fragdropped); 1039 fp->ipq_nfrags--; 1040 m_freem(q); 1041 } 1042 1043 /* 1044 * Check for complete reassembly and perform frag per packet 1045 * limiting. 1046 * 1047 * Frag limiting is performed here so that the nth frag has 1048 * a chance to complete the packet before we drop the packet. 1049 * As a result, n+1 frags are actually allowed per packet, but 1050 * only n will ever be stored. (n = maxfragsperpacket.) 1051 * 1052 */ 1053 next = 0; 1054 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1055 if (GETIP(q)->ip_off != next) { 1056 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1057 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1058 ip_freef(head, fp); 1059 } 1060 goto done; 1061 } 1062 next += GETIP(q)->ip_len; 1063 } 1064 /* Make sure the last packet didn't have the IP_MF flag */ 1065 if (p->m_flags & M_FRAG) { 1066 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1067 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1068 ip_freef(head, fp); 1069 } 1070 goto done; 1071 } 1072 1073 /* 1074 * Reassembly is complete. Make sure the packet is a sane size. 1075 */ 1076 q = fp->ipq_frags; 1077 ip = GETIP(q); 1078 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1079 IPSTAT_INC(ips_toolong); 1080 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1081 ip_freef(head, fp); 1082 goto done; 1083 } 1084 1085 /* 1086 * Concatenate fragments. 1087 */ 1088 m = q; 1089 t = m->m_next; 1090 m->m_next = NULL; 1091 m_cat(m, t); 1092 nq = q->m_nextpkt; 1093 q->m_nextpkt = NULL; 1094 for (q = nq; q != NULL; q = nq) { 1095 nq = q->m_nextpkt; 1096 q->m_nextpkt = NULL; 1097 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1098 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1099 m_cat(m, q); 1100 } 1101 /* 1102 * In order to do checksumming faster we do 'end-around carry' here 1103 * (and not in for{} loop), though it implies we are not going to 1104 * reassemble more than 64k fragments. 1105 */ 1106 m->m_pkthdr.csum_data = 1107 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1108 #ifdef MAC 1109 mac_ipq_reassemble(fp, m); 1110 mac_ipq_destroy(fp); 1111 #endif 1112 1113 /* 1114 * Create header for new ip packet by modifying header of first 1115 * packet; dequeue and discard fragment reassembly header. 1116 * Make header visible. 1117 */ 1118 ip->ip_len = (ip->ip_hl << 2) + next; 1119 ip->ip_src = fp->ipq_src; 1120 ip->ip_dst = fp->ipq_dst; 1121 TAILQ_REMOVE(head, fp, ipq_list); 1122 V_nipq--; 1123 uma_zfree(V_ipq_zone, fp); 1124 m->m_len += (ip->ip_hl << 2); 1125 m->m_data -= (ip->ip_hl << 2); 1126 /* some debugging cruft by sklower, below, will go away soon */ 1127 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1128 m_fixhdr(m); 1129 IPSTAT_INC(ips_reassembled); 1130 IPQ_UNLOCK(); 1131 return (m); 1132 1133 dropfrag: 1134 IPSTAT_INC(ips_fragdropped); 1135 if (fp != NULL) 1136 fp->ipq_nfrags--; 1137 m_freem(m); 1138 done: 1139 IPQ_UNLOCK(); 1140 return (NULL); 1141 1142 #undef GETIP 1143 } 1144 1145 /* 1146 * Free a fragment reassembly header and all 1147 * associated datagrams. 1148 */ 1149 static void 1150 ip_freef(struct ipqhead *fhp, struct ipq *fp) 1151 { 1152 struct mbuf *q; 1153 1154 IPQ_LOCK_ASSERT(); 1155 1156 while (fp->ipq_frags) { 1157 q = fp->ipq_frags; 1158 fp->ipq_frags = q->m_nextpkt; 1159 m_freem(q); 1160 } 1161 TAILQ_REMOVE(fhp, fp, ipq_list); 1162 uma_zfree(V_ipq_zone, fp); 1163 V_nipq--; 1164 } 1165 1166 /* 1167 * IP timer processing; 1168 * if a timer expires on a reassembly 1169 * queue, discard it. 1170 */ 1171 void 1172 ip_slowtimo(void) 1173 { 1174 VNET_ITERATOR_DECL(vnet_iter); 1175 struct ipq *fp; 1176 int i; 1177 1178 VNET_LIST_RLOCK_NOSLEEP(); 1179 IPQ_LOCK(); 1180 VNET_FOREACH(vnet_iter) { 1181 CURVNET_SET(vnet_iter); 1182 for (i = 0; i < IPREASS_NHASH; i++) { 1183 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) { 1184 struct ipq *fpp; 1185 1186 fpp = fp; 1187 fp = TAILQ_NEXT(fp, ipq_list); 1188 if(--fpp->ipq_ttl == 0) { 1189 IPSTAT_ADD(ips_fragtimeout, 1190 fpp->ipq_nfrags); 1191 ip_freef(&V_ipq[i], fpp); 1192 } 1193 } 1194 } 1195 /* 1196 * If we are over the maximum number of fragments 1197 * (due to the limit being lowered), drain off 1198 * enough to get down to the new limit. 1199 */ 1200 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) { 1201 for (i = 0; i < IPREASS_NHASH; i++) { 1202 while (V_nipq > V_maxnipq && 1203 !TAILQ_EMPTY(&V_ipq[i])) { 1204 IPSTAT_ADD(ips_fragdropped, 1205 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1206 ip_freef(&V_ipq[i], 1207 TAILQ_FIRST(&V_ipq[i])); 1208 } 1209 } 1210 } 1211 CURVNET_RESTORE(); 1212 } 1213 IPQ_UNLOCK(); 1214 VNET_LIST_RUNLOCK_NOSLEEP(); 1215 } 1216 1217 /* 1218 * Drain off all datagram fragments. 1219 */ 1220 void 1221 ip_drain(void) 1222 { 1223 VNET_ITERATOR_DECL(vnet_iter); 1224 int i; 1225 1226 VNET_LIST_RLOCK_NOSLEEP(); 1227 IPQ_LOCK(); 1228 VNET_FOREACH(vnet_iter) { 1229 CURVNET_SET(vnet_iter); 1230 for (i = 0; i < IPREASS_NHASH; i++) { 1231 while(!TAILQ_EMPTY(&V_ipq[i])) { 1232 IPSTAT_ADD(ips_fragdropped, 1233 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1234 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); 1235 } 1236 } 1237 CURVNET_RESTORE(); 1238 } 1239 IPQ_UNLOCK(); 1240 VNET_LIST_RUNLOCK_NOSLEEP(); 1241 in_rtqdrain(); 1242 } 1243 1244 /* 1245 * The protocol to be inserted into ip_protox[] must be already registered 1246 * in inetsw[], either statically or through pf_proto_register(). 1247 */ 1248 int 1249 ipproto_register(u_char ipproto) 1250 { 1251 struct protosw *pr; 1252 1253 /* Sanity checks. */ 1254 if (ipproto == 0) 1255 return (EPROTONOSUPPORT); 1256 1257 /* 1258 * The protocol slot must not be occupied by another protocol 1259 * already. An index pointing to IPPROTO_RAW is unused. 1260 */ 1261 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1262 if (pr == NULL) 1263 return (EPFNOSUPPORT); 1264 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1265 return (EEXIST); 1266 1267 /* Find the protocol position in inetsw[] and set the index. */ 1268 for (pr = inetdomain.dom_protosw; 1269 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1270 if (pr->pr_domain->dom_family == PF_INET && 1271 pr->pr_protocol && pr->pr_protocol == ipproto) { 1272 /* Be careful to only index valid IP protocols. */ 1273 if (pr->pr_protocol < IPPROTO_MAX) { 1274 ip_protox[pr->pr_protocol] = pr - inetsw; 1275 return (0); 1276 } else 1277 return (EINVAL); 1278 } 1279 } 1280 return (EPROTONOSUPPORT); 1281 } 1282 1283 int 1284 ipproto_unregister(u_char ipproto) 1285 { 1286 struct protosw *pr; 1287 1288 /* Sanity checks. */ 1289 if (ipproto == 0) 1290 return (EPROTONOSUPPORT); 1291 1292 /* Check if the protocol was indeed registered. */ 1293 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1294 if (pr == NULL) 1295 return (EPFNOSUPPORT); 1296 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1297 return (ENOENT); 1298 1299 /* Reset the protocol slot to IPPROTO_RAW. */ 1300 ip_protox[ipproto] = pr - inetsw; 1301 return (0); 1302 } 1303 1304 /* 1305 * Given address of next destination (final or next hop), return (referenced) 1306 * internet address info of interface to be used to get there. 1307 */ 1308 struct in_ifaddr * 1309 ip_rtaddr(struct in_addr dst, u_int fibnum) 1310 { 1311 struct route sro; 1312 struct sockaddr_in *sin; 1313 struct in_ifaddr *ia; 1314 1315 bzero(&sro, sizeof(sro)); 1316 sin = (struct sockaddr_in *)&sro.ro_dst; 1317 sin->sin_family = AF_INET; 1318 sin->sin_len = sizeof(*sin); 1319 sin->sin_addr = dst; 1320 in_rtalloc_ign(&sro, 0, fibnum); 1321 1322 if (sro.ro_rt == NULL) 1323 return (NULL); 1324 1325 ia = ifatoia(sro.ro_rt->rt_ifa); 1326 ifa_ref(&ia->ia_ifa); 1327 RTFREE(sro.ro_rt); 1328 return (ia); 1329 } 1330 1331 u_char inetctlerrmap[PRC_NCMDS] = { 1332 0, 0, 0, 0, 1333 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1334 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1335 EMSGSIZE, EHOSTUNREACH, 0, 0, 1336 0, 0, EHOSTUNREACH, 0, 1337 ENOPROTOOPT, ECONNREFUSED 1338 }; 1339 1340 /* 1341 * Forward a packet. If some error occurs return the sender 1342 * an icmp packet. Note we can't always generate a meaningful 1343 * icmp message because icmp doesn't have a large enough repertoire 1344 * of codes and types. 1345 * 1346 * If not forwarding, just drop the packet. This could be confusing 1347 * if ipforwarding was zero but some routing protocol was advancing 1348 * us as a gateway to somewhere. However, we must let the routing 1349 * protocol deal with that. 1350 * 1351 * The srcrt parameter indicates whether the packet is being forwarded 1352 * via a source route. 1353 */ 1354 void 1355 ip_forward(struct mbuf *m, int srcrt) 1356 { 1357 struct ip *ip = mtod(m, struct ip *); 1358 struct in_ifaddr *ia; 1359 struct mbuf *mcopy; 1360 struct in_addr dest; 1361 struct route ro; 1362 int error, type = 0, code = 0, mtu = 0; 1363 1364 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1365 IPSTAT_INC(ips_cantforward); 1366 m_freem(m); 1367 return; 1368 } 1369 #ifdef IPSTEALTH 1370 if (!V_ipstealth) { 1371 #endif 1372 if (ip->ip_ttl <= IPTTLDEC) { 1373 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1374 0, 0); 1375 return; 1376 } 1377 #ifdef IPSTEALTH 1378 } 1379 #endif 1380 1381 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); 1382 #ifndef IPSEC 1383 /* 1384 * 'ia' may be NULL if there is no route for this destination. 1385 * In case of IPsec, Don't discard it just yet, but pass it to 1386 * ip_output in case of outgoing IPsec policy. 1387 */ 1388 if (!srcrt && ia == NULL) { 1389 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1390 return; 1391 } 1392 #endif 1393 1394 /* 1395 * Save the IP header and at most 8 bytes of the payload, 1396 * in case we need to generate an ICMP message to the src. 1397 * 1398 * XXX this can be optimized a lot by saving the data in a local 1399 * buffer on the stack (72 bytes at most), and only allocating the 1400 * mbuf if really necessary. The vast majority of the packets 1401 * are forwarded without having to send an ICMP back (either 1402 * because unnecessary, or because rate limited), so we are 1403 * really we are wasting a lot of work here. 1404 * 1405 * We don't use m_copy() because it might return a reference 1406 * to a shared cluster. Both this function and ip_output() 1407 * assume exclusive access to the IP header in `m', so any 1408 * data in a cluster may change before we reach icmp_error(). 1409 */ 1410 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1411 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1412 /* 1413 * It's probably ok if the pkthdr dup fails (because 1414 * the deep copy of the tag chain failed), but for now 1415 * be conservative and just discard the copy since 1416 * code below may some day want the tags. 1417 */ 1418 m_free(mcopy); 1419 mcopy = NULL; 1420 } 1421 if (mcopy != NULL) { 1422 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy)); 1423 mcopy->m_pkthdr.len = mcopy->m_len; 1424 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1425 } 1426 1427 #ifdef IPSTEALTH 1428 if (!V_ipstealth) { 1429 #endif 1430 ip->ip_ttl -= IPTTLDEC; 1431 #ifdef IPSTEALTH 1432 } 1433 #endif 1434 1435 /* 1436 * If forwarding packet using same interface that it came in on, 1437 * perhaps should send a redirect to sender to shortcut a hop. 1438 * Only send redirect if source is sending directly to us, 1439 * and if packet was not source routed (or has any options). 1440 * Also, don't send redirect if forwarding using a default route 1441 * or a route modified by a redirect. 1442 */ 1443 dest.s_addr = 0; 1444 if (!srcrt && V_ipsendredirects && 1445 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) { 1446 struct sockaddr_in *sin; 1447 struct rtentry *rt; 1448 1449 bzero(&ro, sizeof(ro)); 1450 sin = (struct sockaddr_in *)&ro.ro_dst; 1451 sin->sin_family = AF_INET; 1452 sin->sin_len = sizeof(*sin); 1453 sin->sin_addr = ip->ip_dst; 1454 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 1455 1456 rt = ro.ro_rt; 1457 1458 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1459 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1460 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1461 u_long src = ntohl(ip->ip_src.s_addr); 1462 1463 if (RTA(rt) && 1464 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1465 if (rt->rt_flags & RTF_GATEWAY) 1466 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1467 else 1468 dest.s_addr = ip->ip_dst.s_addr; 1469 /* Router requirements says to only send host redirects */ 1470 type = ICMP_REDIRECT; 1471 code = ICMP_REDIRECT_HOST; 1472 } 1473 } 1474 if (rt) 1475 RTFREE(rt); 1476 } 1477 1478 /* 1479 * Try to cache the route MTU from ip_output so we can consider it for 1480 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. 1481 */ 1482 bzero(&ro, sizeof(ro)); 1483 1484 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1485 1486 if (error == EMSGSIZE && ro.ro_rt) 1487 mtu = ro.ro_rt->rt_rmx.rmx_mtu; 1488 if (ro.ro_rt) 1489 RTFREE(ro.ro_rt); 1490 1491 if (error) 1492 IPSTAT_INC(ips_cantforward); 1493 else { 1494 IPSTAT_INC(ips_forward); 1495 if (type) 1496 IPSTAT_INC(ips_redirectsent); 1497 else { 1498 if (mcopy) 1499 m_freem(mcopy); 1500 if (ia != NULL) 1501 ifa_free(&ia->ia_ifa); 1502 return; 1503 } 1504 } 1505 if (mcopy == NULL) { 1506 if (ia != NULL) 1507 ifa_free(&ia->ia_ifa); 1508 return; 1509 } 1510 1511 switch (error) { 1512 1513 case 0: /* forwarded, but need redirect */ 1514 /* type, code set above */ 1515 break; 1516 1517 case ENETUNREACH: 1518 case EHOSTUNREACH: 1519 case ENETDOWN: 1520 case EHOSTDOWN: 1521 default: 1522 type = ICMP_UNREACH; 1523 code = ICMP_UNREACH_HOST; 1524 break; 1525 1526 case EMSGSIZE: 1527 type = ICMP_UNREACH; 1528 code = ICMP_UNREACH_NEEDFRAG; 1529 1530 #ifdef IPSEC 1531 /* 1532 * If IPsec is configured for this path, 1533 * override any possibly mtu value set by ip_output. 1534 */ 1535 mtu = ip_ipsec_mtu(m, mtu); 1536 #endif /* IPSEC */ 1537 /* 1538 * If the MTU was set before make sure we are below the 1539 * interface MTU. 1540 * If the MTU wasn't set before use the interface mtu or 1541 * fall back to the next smaller mtu step compared to the 1542 * current packet size. 1543 */ 1544 if (mtu != 0) { 1545 if (ia != NULL) 1546 mtu = min(mtu, ia->ia_ifp->if_mtu); 1547 } else { 1548 if (ia != NULL) 1549 mtu = ia->ia_ifp->if_mtu; 1550 else 1551 mtu = ip_next_mtu(ip->ip_len, 0); 1552 } 1553 IPSTAT_INC(ips_cantfrag); 1554 break; 1555 1556 case ENOBUFS: 1557 /* 1558 * A router should not generate ICMP_SOURCEQUENCH as 1559 * required in RFC1812 Requirements for IP Version 4 Routers. 1560 * Source quench could be a big problem under DoS attacks, 1561 * or if the underlying interface is rate-limited. 1562 * Those who need source quench packets may re-enable them 1563 * via the net.inet.ip.sendsourcequench sysctl. 1564 */ 1565 if (V_ip_sendsourcequench == 0) { 1566 m_freem(mcopy); 1567 if (ia != NULL) 1568 ifa_free(&ia->ia_ifa); 1569 return; 1570 } else { 1571 type = ICMP_SOURCEQUENCH; 1572 code = 0; 1573 } 1574 break; 1575 1576 case EACCES: /* ipfw denied packet */ 1577 m_freem(mcopy); 1578 if (ia != NULL) 1579 ifa_free(&ia->ia_ifa); 1580 return; 1581 } 1582 if (ia != NULL) 1583 ifa_free(&ia->ia_ifa); 1584 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1585 } 1586 1587 void 1588 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1589 struct mbuf *m) 1590 { 1591 1592 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1593 struct bintime bt; 1594 1595 bintime(&bt); 1596 if (inp->inp_socket->so_options & SO_BINTIME) { 1597 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1598 SCM_BINTIME, SOL_SOCKET); 1599 if (*mp) 1600 mp = &(*mp)->m_next; 1601 } 1602 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1603 struct timeval tv; 1604 1605 bintime2timeval(&bt, &tv); 1606 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1607 SCM_TIMESTAMP, SOL_SOCKET); 1608 if (*mp) 1609 mp = &(*mp)->m_next; 1610 } 1611 } 1612 if (inp->inp_flags & INP_RECVDSTADDR) { 1613 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1614 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1615 if (*mp) 1616 mp = &(*mp)->m_next; 1617 } 1618 if (inp->inp_flags & INP_RECVTTL) { 1619 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1620 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1621 if (*mp) 1622 mp = &(*mp)->m_next; 1623 } 1624 #ifdef notyet 1625 /* XXX 1626 * Moving these out of udp_input() made them even more broken 1627 * than they already were. 1628 */ 1629 /* options were tossed already */ 1630 if (inp->inp_flags & INP_RECVOPTS) { 1631 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1632 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1633 if (*mp) 1634 mp = &(*mp)->m_next; 1635 } 1636 /* ip_srcroute doesn't do what we want here, need to fix */ 1637 if (inp->inp_flags & INP_RECVRETOPTS) { 1638 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1639 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1640 if (*mp) 1641 mp = &(*mp)->m_next; 1642 } 1643 #endif 1644 if (inp->inp_flags & INP_RECVIF) { 1645 struct ifnet *ifp; 1646 struct sdlbuf { 1647 struct sockaddr_dl sdl; 1648 u_char pad[32]; 1649 } sdlbuf; 1650 struct sockaddr_dl *sdp; 1651 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1652 1653 if (((ifp = m->m_pkthdr.rcvif)) 1654 && ( ifp->if_index && (ifp->if_index <= V_if_index))) { 1655 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1656 /* 1657 * Change our mind and don't try copy. 1658 */ 1659 if ((sdp->sdl_family != AF_LINK) 1660 || (sdp->sdl_len > sizeof(sdlbuf))) { 1661 goto makedummy; 1662 } 1663 bcopy(sdp, sdl2, sdp->sdl_len); 1664 } else { 1665 makedummy: 1666 sdl2->sdl_len 1667 = offsetof(struct sockaddr_dl, sdl_data[0]); 1668 sdl2->sdl_family = AF_LINK; 1669 sdl2->sdl_index = 0; 1670 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1671 } 1672 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1673 IP_RECVIF, IPPROTO_IP); 1674 if (*mp) 1675 mp = &(*mp)->m_next; 1676 } 1677 } 1678 1679 /* 1680 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1681 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1682 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1683 * compiled. 1684 */ 1685 int 1686 ip_rsvp_init(struct socket *so) 1687 { 1688 1689 if (so->so_type != SOCK_RAW || 1690 so->so_proto->pr_protocol != IPPROTO_RSVP) 1691 return EOPNOTSUPP; 1692 1693 if (V_ip_rsvpd != NULL) 1694 return EADDRINUSE; 1695 1696 V_ip_rsvpd = so; 1697 /* 1698 * This may seem silly, but we need to be sure we don't over-increment 1699 * the RSVP counter, in case something slips up. 1700 */ 1701 if (!V_ip_rsvp_on) { 1702 V_ip_rsvp_on = 1; 1703 V_rsvp_on++; 1704 } 1705 1706 return 0; 1707 } 1708 1709 int 1710 ip_rsvp_done(void) 1711 { 1712 1713 V_ip_rsvpd = NULL; 1714 /* 1715 * This may seem silly, but we need to be sure we don't over-decrement 1716 * the RSVP counter, in case something slips up. 1717 */ 1718 if (V_ip_rsvp_on) { 1719 V_ip_rsvp_on = 0; 1720 V_rsvp_on--; 1721 } 1722 return 0; 1723 } 1724 1725 void 1726 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1727 { 1728 1729 if (rsvp_input_p) { /* call the real one if loaded */ 1730 rsvp_input_p(m, off); 1731 return; 1732 } 1733 1734 /* Can still get packets with rsvp_on = 0 if there is a local member 1735 * of the group to which the RSVP packet is addressed. But in this 1736 * case we want to throw the packet away. 1737 */ 1738 1739 if (!V_rsvp_on) { 1740 m_freem(m); 1741 return; 1742 } 1743 1744 if (V_ip_rsvpd != NULL) { 1745 rip_input(m, off); 1746 return; 1747 } 1748 /* Drop the packet */ 1749 m_freem(m); 1750 } 1751