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 41 #include <sys/param.h> 42 #include <sys/systm.h> 43 #include <sys/mbuf.h> 44 #include <sys/malloc.h> 45 #include <sys/domain.h> 46 #include <sys/protosw.h> 47 #include <sys/socket.h> 48 #include <sys/time.h> 49 #include <sys/kernel.h> 50 #include <sys/lock.h> 51 #include <sys/rwlock.h> 52 #include <sys/sdt.h> 53 #include <sys/syslog.h> 54 #include <sys/sysctl.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 #include <net/vnet.h> 64 65 #include <netinet/in.h> 66 #include <netinet/in_kdtrace.h> 67 #include <netinet/in_systm.h> 68 #include <netinet/in_var.h> 69 #include <netinet/ip.h> 70 #include <netinet/in_pcb.h> 71 #include <netinet/ip_var.h> 72 #include <netinet/ip_fw.h> 73 #include <netinet/ip_icmp.h> 74 #include <netinet/ip_options.h> 75 #include <machine/in_cksum.h> 76 #include <netinet/ip_carp.h> 77 #ifdef IPSEC 78 #include <netinet/ip_ipsec.h> 79 #endif /* IPSEC */ 80 81 #include <sys/socketvar.h> 82 83 #include <security/mac/mac_framework.h> 84 85 #ifdef CTASSERT 86 CTASSERT(sizeof(struct ip) == 20); 87 #endif 88 89 struct rwlock in_ifaddr_lock; 90 RW_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock"); 91 92 VNET_DEFINE(int, rsvp_on); 93 94 VNET_DEFINE(int, ipforwarding); 95 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 96 &VNET_NAME(ipforwarding), 0, 97 "Enable IP forwarding between interfaces"); 98 99 static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */ 100 #define V_ipsendredirects VNET(ipsendredirects) 101 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 102 &VNET_NAME(ipsendredirects), 0, 103 "Enable sending IP redirects"); 104 105 static VNET_DEFINE(int, ip_keepfaith); 106 #define V_ip_keepfaith VNET(ip_keepfaith) 107 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 108 &VNET_NAME(ip_keepfaith), 0, 109 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 110 111 static VNET_DEFINE(int, ip_sendsourcequench); 112 #define V_ip_sendsourcequench VNET(ip_sendsourcequench) 113 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 114 &VNET_NAME(ip_sendsourcequench), 0, 115 "Enable the transmission of source quench packets"); 116 117 VNET_DEFINE(int, ip_do_randomid); 118 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 119 &VNET_NAME(ip_do_randomid), 0, 120 "Assign random ip_id values"); 121 122 /* 123 * XXX - Setting ip_checkinterface mostly implements the receive side of 124 * the Strong ES model described in RFC 1122, but since the routing table 125 * and transmit implementation do not implement the Strong ES model, 126 * setting this to 1 results in an odd hybrid. 127 * 128 * XXX - ip_checkinterface currently must be disabled if you use ipnat 129 * to translate the destination address to another local interface. 130 * 131 * XXX - ip_checkinterface must be disabled if you add IP aliases 132 * to the loopback interface instead of the interface where the 133 * packets for those addresses are received. 134 */ 135 static VNET_DEFINE(int, ip_checkinterface); 136 #define V_ip_checkinterface VNET(ip_checkinterface) 137 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 138 &VNET_NAME(ip_checkinterface), 0, 139 "Verify packet arrives on correct interface"); 140 141 VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */ 142 143 static struct netisr_handler ip_nh = { 144 .nh_name = "ip", 145 .nh_handler = ip_input, 146 .nh_proto = NETISR_IP, 147 .nh_policy = NETISR_POLICY_FLOW, 148 }; 149 150 extern struct domain inetdomain; 151 extern struct protosw inetsw[]; 152 u_char ip_protox[IPPROTO_MAX]; 153 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */ 154 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */ 155 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */ 156 157 static VNET_DEFINE(uma_zone_t, ipq_zone); 158 static VNET_DEFINE(TAILQ_HEAD(ipqhead, ipq), ipq[IPREASS_NHASH]); 159 static struct mtx ipqlock; 160 161 #define V_ipq_zone VNET(ipq_zone) 162 #define V_ipq VNET(ipq) 163 164 #define IPQ_LOCK() mtx_lock(&ipqlock) 165 #define IPQ_UNLOCK() mtx_unlock(&ipqlock) 166 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 167 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 168 169 static void maxnipq_update(void); 170 static void ipq_zone_change(void *); 171 static void ip_drain_locked(void); 172 173 static VNET_DEFINE(int, maxnipq); /* Administrative limit on # reass queues. */ 174 static VNET_DEFINE(int, nipq); /* Total # of reass queues */ 175 #define V_maxnipq VNET(maxnipq) 176 #define V_nipq VNET(nipq) 177 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, 178 &VNET_NAME(nipq), 0, 179 "Current number of IPv4 fragment reassembly queue entries"); 180 181 static VNET_DEFINE(int, maxfragsperpacket); 182 #define V_maxfragsperpacket VNET(maxfragsperpacket) 183 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 184 &VNET_NAME(maxfragsperpacket), 0, 185 "Maximum number of IPv4 fragments allowed per packet"); 186 187 #ifdef IPCTL_DEFMTU 188 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 189 &ip_mtu, 0, "Default MTU"); 190 #endif 191 192 #ifdef IPSTEALTH 193 VNET_DEFINE(int, ipstealth); 194 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 195 &VNET_NAME(ipstealth), 0, 196 "IP stealth mode, no TTL decrementation on forwarding"); 197 #endif 198 199 static void ip_freef(struct ipqhead *, struct ipq *); 200 201 /* 202 * IP statistics are stored in the "array" of counter(9)s. 203 */ 204 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat); 205 VNET_PCPUSTAT_SYSINIT(ipstat); 206 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat, 207 "IP statistics (struct ipstat, netinet/ip_var.h)"); 208 209 #ifdef VIMAGE 210 VNET_PCPUSTAT_SYSUNINIT(ipstat); 211 #endif /* VIMAGE */ 212 213 /* 214 * Kernel module interface for updating ipstat. The argument is an index 215 * into ipstat treated as an array. 216 */ 217 void 218 kmod_ipstat_inc(int statnum) 219 { 220 221 counter_u64_add(VNET(ipstat)[statnum], 1); 222 } 223 224 void 225 kmod_ipstat_dec(int statnum) 226 { 227 228 counter_u64_add(VNET(ipstat)[statnum], -1); 229 } 230 231 static int 232 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 233 { 234 int error, qlimit; 235 236 netisr_getqlimit(&ip_nh, &qlimit); 237 error = sysctl_handle_int(oidp, &qlimit, 0, req); 238 if (error || !req->newptr) 239 return (error); 240 if (qlimit < 1) 241 return (EINVAL); 242 return (netisr_setqlimit(&ip_nh, qlimit)); 243 } 244 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, 245 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I", 246 "Maximum size of the IP input queue"); 247 248 static int 249 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) 250 { 251 u_int64_t qdrops_long; 252 int error, qdrops; 253 254 netisr_getqdrops(&ip_nh, &qdrops_long); 255 qdrops = qdrops_long; 256 error = sysctl_handle_int(oidp, &qdrops, 0, req); 257 if (error || !req->newptr) 258 return (error); 259 if (qdrops != 0) 260 return (EINVAL); 261 netisr_clearqdrops(&ip_nh); 262 return (0); 263 } 264 265 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, 266 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I", 267 "Number of packets dropped from the IP input queue"); 268 269 /* 270 * IP initialization: fill in IP protocol switch table. 271 * All protocols not implemented in kernel go to raw IP protocol handler. 272 */ 273 void 274 ip_init(void) 275 { 276 struct protosw *pr; 277 int i; 278 279 V_ip_id = time_second & 0xffff; 280 281 TAILQ_INIT(&V_in_ifaddrhead); 282 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 283 284 /* Initialize IP reassembly queue. */ 285 for (i = 0; i < IPREASS_NHASH; i++) 286 TAILQ_INIT(&V_ipq[i]); 287 V_maxnipq = nmbclusters / 32; 288 V_maxfragsperpacket = 16; 289 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 290 NULL, UMA_ALIGN_PTR, 0); 291 maxnipq_update(); 292 293 /* Initialize packet filter hooks. */ 294 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF; 295 V_inet_pfil_hook.ph_af = AF_INET; 296 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0) 297 printf("%s: WARNING: unable to register pfil hook, " 298 "error %d\n", __func__, i); 299 300 /* Skip initialization of globals for non-default instances. */ 301 if (!IS_DEFAULT_VNET(curvnet)) 302 return; 303 304 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 305 if (pr == NULL) 306 panic("ip_init: PF_INET not found"); 307 308 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 309 for (i = 0; i < IPPROTO_MAX; i++) 310 ip_protox[i] = pr - inetsw; 311 /* 312 * Cycle through IP protocols and put them into the appropriate place 313 * in ip_protox[]. 314 */ 315 for (pr = inetdomain.dom_protosw; 316 pr < inetdomain.dom_protoswNPROTOSW; pr++) 317 if (pr->pr_domain->dom_family == PF_INET && 318 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 319 /* Be careful to only index valid IP protocols. */ 320 if (pr->pr_protocol < IPPROTO_MAX) 321 ip_protox[pr->pr_protocol] = pr - inetsw; 322 } 323 324 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, 325 NULL, EVENTHANDLER_PRI_ANY); 326 327 /* Initialize various other remaining things. */ 328 IPQ_LOCK_INIT(); 329 netisr_register(&ip_nh); 330 } 331 332 #ifdef VIMAGE 333 void 334 ip_destroy(void) 335 { 336 int i; 337 338 if ((i = pfil_head_unregister(&V_inet_pfil_hook)) != 0) 339 printf("%s: WARNING: unable to unregister pfil hook, " 340 "error %d\n", __func__, i); 341 342 /* Cleanup in_ifaddr hash table; should be empty. */ 343 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask); 344 345 IPQ_LOCK(); 346 ip_drain_locked(); 347 IPQ_UNLOCK(); 348 349 uma_zdestroy(V_ipq_zone); 350 } 351 #endif 352 353 /* 354 * Ip input routine. Checksum and byte swap header. If fragmented 355 * try to reassemble. Process options. Pass to next level. 356 */ 357 void 358 ip_input(struct mbuf *m) 359 { 360 struct ip *ip = NULL; 361 struct in_ifaddr *ia = NULL; 362 struct ifaddr *ifa; 363 struct ifnet *ifp; 364 int checkif, hlen = 0; 365 uint16_t sum, ip_len; 366 int dchg = 0; /* dest changed after fw */ 367 struct in_addr odst; /* original dst address */ 368 369 M_ASSERTPKTHDR(m); 370 371 if (m->m_flags & M_FASTFWD_OURS) { 372 m->m_flags &= ~M_FASTFWD_OURS; 373 /* Set up some basics that will be used later. */ 374 ip = mtod(m, struct ip *); 375 hlen = ip->ip_hl << 2; 376 ip_len = ntohs(ip->ip_len); 377 goto ours; 378 } 379 380 IPSTAT_INC(ips_total); 381 382 if (m->m_pkthdr.len < sizeof(struct ip)) 383 goto tooshort; 384 385 if (m->m_len < sizeof (struct ip) && 386 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 387 IPSTAT_INC(ips_toosmall); 388 return; 389 } 390 ip = mtod(m, struct ip *); 391 392 if (ip->ip_v != IPVERSION) { 393 IPSTAT_INC(ips_badvers); 394 goto bad; 395 } 396 397 hlen = ip->ip_hl << 2; 398 if (hlen < sizeof(struct ip)) { /* minimum header length */ 399 IPSTAT_INC(ips_badhlen); 400 goto bad; 401 } 402 if (hlen > m->m_len) { 403 if ((m = m_pullup(m, hlen)) == NULL) { 404 IPSTAT_INC(ips_badhlen); 405 return; 406 } 407 ip = mtod(m, struct ip *); 408 } 409 410 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL); 411 412 /* 127/8 must not appear on wire - RFC1122 */ 413 ifp = m->m_pkthdr.rcvif; 414 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 415 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 416 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 417 IPSTAT_INC(ips_badaddr); 418 goto bad; 419 } 420 } 421 422 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 423 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 424 } else { 425 if (hlen == sizeof(struct ip)) { 426 sum = in_cksum_hdr(ip); 427 } else { 428 sum = in_cksum(m, hlen); 429 } 430 } 431 if (sum) { 432 IPSTAT_INC(ips_badsum); 433 goto bad; 434 } 435 436 #ifdef ALTQ 437 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 438 /* packet is dropped by traffic conditioner */ 439 return; 440 #endif 441 442 ip_len = ntohs(ip->ip_len); 443 if (ip_len < hlen) { 444 IPSTAT_INC(ips_badlen); 445 goto bad; 446 } 447 448 /* 449 * Check that the amount of data in the buffers 450 * is as at least much as the IP header would have us expect. 451 * Trim mbufs if longer than we expect. 452 * Drop packet if shorter than we expect. 453 */ 454 if (m->m_pkthdr.len < ip_len) { 455 tooshort: 456 IPSTAT_INC(ips_tooshort); 457 goto bad; 458 } 459 if (m->m_pkthdr.len > ip_len) { 460 if (m->m_len == m->m_pkthdr.len) { 461 m->m_len = ip_len; 462 m->m_pkthdr.len = ip_len; 463 } else 464 m_adj(m, ip_len - m->m_pkthdr.len); 465 } 466 #ifdef IPSEC 467 /* 468 * Bypass packet filtering for packets previously handled by IPsec. 469 */ 470 if (ip_ipsec_filtertunnel(m)) 471 goto passin; 472 #endif /* IPSEC */ 473 474 /* 475 * Run through list of hooks for input packets. 476 * 477 * NB: Beware of the destination address changing (e.g. 478 * by NAT rewriting). When this happens, tell 479 * ip_forward to do the right thing. 480 */ 481 482 /* Jump over all PFIL processing if hooks are not active. */ 483 if (!PFIL_HOOKED(&V_inet_pfil_hook)) 484 goto passin; 485 486 odst = ip->ip_dst; 487 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0) 488 return; 489 if (m == NULL) /* consumed by filter */ 490 return; 491 492 ip = mtod(m, struct ip *); 493 dchg = (odst.s_addr != ip->ip_dst.s_addr); 494 ifp = m->m_pkthdr.rcvif; 495 496 if (m->m_flags & M_FASTFWD_OURS) { 497 m->m_flags &= ~M_FASTFWD_OURS; 498 goto ours; 499 } 500 if (m->m_flags & M_IP_NEXTHOP) { 501 dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL); 502 if (dchg != 0) { 503 /* 504 * Directly ship the packet on. This allows 505 * forwarding packets originally destined to us 506 * to some other directly connected host. 507 */ 508 ip_forward(m, 1); 509 return; 510 } 511 } 512 passin: 513 514 /* 515 * Process options and, if not destined for us, 516 * ship it on. ip_dooptions returns 1 when an 517 * error was detected (causing an icmp message 518 * to be sent and the original packet to be freed). 519 */ 520 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 521 return; 522 523 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 524 * matter if it is destined to another node, or whether it is 525 * a multicast one, RSVP wants it! and prevents it from being forwarded 526 * anywhere else. Also checks if the rsvp daemon is running before 527 * grabbing the packet. 528 */ 529 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 530 goto ours; 531 532 /* 533 * Check our list of addresses, to see if the packet is for us. 534 * If we don't have any addresses, assume any unicast packet 535 * we receive might be for us (and let the upper layers deal 536 * with it). 537 */ 538 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 539 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 540 goto ours; 541 542 /* 543 * Enable a consistency check between the destination address 544 * and the arrival interface for a unicast packet (the RFC 1122 545 * strong ES model) if IP forwarding is disabled and the packet 546 * is not locally generated and the packet is not subject to 547 * 'ipfw fwd'. 548 * 549 * XXX - Checking also should be disabled if the destination 550 * address is ipnat'ed to a different interface. 551 * 552 * XXX - Checking is incompatible with IP aliases added 553 * to the loopback interface instead of the interface where 554 * the packets are received. 555 * 556 * XXX - This is the case for carp vhost IPs as well so we 557 * insert a workaround. If the packet got here, we already 558 * checked with carp_iamatch() and carp_forus(). 559 */ 560 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 561 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && 562 ifp->if_carp == NULL && (dchg == 0); 563 564 /* 565 * Check for exact addresses in the hash bucket. 566 */ 567 /* IN_IFADDR_RLOCK(); */ 568 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 569 /* 570 * If the address matches, verify that the packet 571 * arrived via the correct interface if checking is 572 * enabled. 573 */ 574 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 575 (!checkif || ia->ia_ifp == ifp)) { 576 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 577 counter_u64_add(ia->ia_ifa.ifa_ibytes, 578 m->m_pkthdr.len); 579 /* IN_IFADDR_RUNLOCK(); */ 580 goto ours; 581 } 582 } 583 /* IN_IFADDR_RUNLOCK(); */ 584 585 /* 586 * Check for broadcast addresses. 587 * 588 * Only accept broadcast packets that arrive via the matching 589 * interface. Reception of forwarded directed broadcasts would 590 * be handled via ip_forward() and ether_output() with the loopback 591 * into the stack for SIMPLEX interfaces handled by ether_output(). 592 */ 593 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 594 IF_ADDR_RLOCK(ifp); 595 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 596 if (ifa->ifa_addr->sa_family != AF_INET) 597 continue; 598 ia = ifatoia(ifa); 599 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 600 ip->ip_dst.s_addr) { 601 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 602 counter_u64_add(ia->ia_ifa.ifa_ibytes, 603 m->m_pkthdr.len); 604 IF_ADDR_RUNLOCK(ifp); 605 goto ours; 606 } 607 #ifdef BOOTP_COMPAT 608 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { 609 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 610 counter_u64_add(ia->ia_ifa.ifa_ibytes, 611 m->m_pkthdr.len); 612 IF_ADDR_RUNLOCK(ifp); 613 goto ours; 614 } 615 #endif 616 } 617 IF_ADDR_RUNLOCK(ifp); 618 ia = NULL; 619 } 620 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 621 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 622 IPSTAT_INC(ips_cantforward); 623 m_freem(m); 624 return; 625 } 626 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 627 if (V_ip_mrouter) { 628 /* 629 * If we are acting as a multicast router, all 630 * incoming multicast packets are passed to the 631 * kernel-level multicast forwarding function. 632 * The packet is returned (relatively) intact; if 633 * ip_mforward() returns a non-zero value, the packet 634 * must be discarded, else it may be accepted below. 635 */ 636 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 637 IPSTAT_INC(ips_cantforward); 638 m_freem(m); 639 return; 640 } 641 642 /* 643 * The process-level routing daemon needs to receive 644 * all multicast IGMP packets, whether or not this 645 * host belongs to their destination groups. 646 */ 647 if (ip->ip_p == IPPROTO_IGMP) 648 goto ours; 649 IPSTAT_INC(ips_forward); 650 } 651 /* 652 * Assume the packet is for us, to avoid prematurely taking 653 * a lock on the in_multi hash. Protocols must perform 654 * their own filtering and update statistics accordingly. 655 */ 656 goto ours; 657 } 658 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 659 goto ours; 660 if (ip->ip_dst.s_addr == INADDR_ANY) 661 goto ours; 662 663 /* 664 * FAITH(Firewall Aided Internet Translator) 665 */ 666 if (ifp && ifp->if_type == IFT_FAITH) { 667 if (V_ip_keepfaith) { 668 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 669 goto ours; 670 } 671 m_freem(m); 672 return; 673 } 674 675 /* 676 * Not for us; forward if possible and desirable. 677 */ 678 if (V_ipforwarding == 0) { 679 IPSTAT_INC(ips_cantforward); 680 m_freem(m); 681 } else { 682 #ifdef IPSEC 683 if (ip_ipsec_fwd(m)) 684 goto bad; 685 #endif /* IPSEC */ 686 ip_forward(m, dchg); 687 } 688 return; 689 690 ours: 691 #ifdef IPSTEALTH 692 /* 693 * IPSTEALTH: Process non-routing options only 694 * if the packet is destined for us. 695 */ 696 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) 697 return; 698 #endif /* IPSTEALTH */ 699 700 /* 701 * Attempt reassembly; if it succeeds, proceed. 702 * ip_reass() will return a different mbuf. 703 */ 704 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) { 705 /* XXXGL: shouldn't we save & set m_flags? */ 706 m = ip_reass(m); 707 if (m == NULL) 708 return; 709 ip = mtod(m, struct ip *); 710 /* Get the header length of the reassembled packet */ 711 hlen = ip->ip_hl << 2; 712 } 713 714 #ifdef IPSEC 715 /* 716 * enforce IPsec policy checking if we are seeing last header. 717 * note that we do not visit this with protocols with pcb layer 718 * code - like udp/tcp/raw ip. 719 */ 720 if (ip_ipsec_input(m)) 721 goto bad; 722 #endif /* IPSEC */ 723 724 /* 725 * Switch out to protocol's input routine. 726 */ 727 IPSTAT_INC(ips_delivered); 728 729 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 730 return; 731 bad: 732 m_freem(m); 733 } 734 735 /* 736 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 737 * max has slightly different semantics than the sysctl, for historical 738 * reasons. 739 */ 740 static void 741 maxnipq_update(void) 742 { 743 744 /* 745 * -1 for unlimited allocation. 746 */ 747 if (V_maxnipq < 0) 748 uma_zone_set_max(V_ipq_zone, 0); 749 /* 750 * Positive number for specific bound. 751 */ 752 if (V_maxnipq > 0) 753 uma_zone_set_max(V_ipq_zone, V_maxnipq); 754 /* 755 * Zero specifies no further fragment queue allocation -- set the 756 * bound very low, but rely on implementation elsewhere to actually 757 * prevent allocation and reclaim current queues. 758 */ 759 if (V_maxnipq == 0) 760 uma_zone_set_max(V_ipq_zone, 1); 761 } 762 763 static void 764 ipq_zone_change(void *tag) 765 { 766 767 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) { 768 V_maxnipq = nmbclusters / 32; 769 maxnipq_update(); 770 } 771 } 772 773 static int 774 sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 775 { 776 int error, i; 777 778 i = V_maxnipq; 779 error = sysctl_handle_int(oidp, &i, 0, req); 780 if (error || !req->newptr) 781 return (error); 782 783 /* 784 * XXXRW: Might be a good idea to sanity check the argument and place 785 * an extreme upper bound. 786 */ 787 if (i < -1) 788 return (EINVAL); 789 V_maxnipq = i; 790 maxnipq_update(); 791 return (0); 792 } 793 794 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 795 NULL, 0, sysctl_maxnipq, "I", 796 "Maximum number of IPv4 fragment reassembly queue entries"); 797 798 #define M_IP_FRAG M_PROTO9 799 800 /* 801 * Take incoming datagram fragment and try to reassemble it into 802 * whole datagram. If the argument is the first fragment or one 803 * in between the function will return NULL and store the mbuf 804 * in the fragment chain. If the argument is the last fragment 805 * the packet will be reassembled and the pointer to the new 806 * mbuf returned for further processing. Only m_tags attached 807 * to the first packet/fragment are preserved. 808 * The IP header is *NOT* adjusted out of iplen. 809 */ 810 struct mbuf * 811 ip_reass(struct mbuf *m) 812 { 813 struct ip *ip; 814 struct mbuf *p, *q, *nq, *t; 815 struct ipq *fp = NULL; 816 struct ipqhead *head; 817 int i, hlen, next; 818 u_int8_t ecn, ecn0; 819 u_short hash; 820 821 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 822 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) { 823 IPSTAT_INC(ips_fragments); 824 IPSTAT_INC(ips_fragdropped); 825 m_freem(m); 826 return (NULL); 827 } 828 829 ip = mtod(m, struct ip *); 830 hlen = ip->ip_hl << 2; 831 832 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 833 head = &V_ipq[hash]; 834 IPQ_LOCK(); 835 836 /* 837 * Look for queue of fragments 838 * of this datagram. 839 */ 840 TAILQ_FOREACH(fp, head, ipq_list) 841 if (ip->ip_id == fp->ipq_id && 842 ip->ip_src.s_addr == fp->ipq_src.s_addr && 843 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 844 #ifdef MAC 845 mac_ipq_match(m, fp) && 846 #endif 847 ip->ip_p == fp->ipq_p) 848 goto found; 849 850 fp = NULL; 851 852 /* 853 * Attempt to trim the number of allocated fragment queues if it 854 * exceeds the administrative limit. 855 */ 856 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) { 857 /* 858 * drop something from the tail of the current queue 859 * before proceeding further 860 */ 861 struct ipq *q = TAILQ_LAST(head, ipqhead); 862 if (q == NULL) { /* gak */ 863 for (i = 0; i < IPREASS_NHASH; i++) { 864 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead); 865 if (r) { 866 IPSTAT_ADD(ips_fragtimeout, 867 r->ipq_nfrags); 868 ip_freef(&V_ipq[i], r); 869 break; 870 } 871 } 872 } else { 873 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags); 874 ip_freef(head, q); 875 } 876 } 877 878 found: 879 /* 880 * Adjust ip_len to not reflect header, 881 * convert offset of this to bytes. 882 */ 883 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 884 if (ip->ip_off & htons(IP_MF)) { 885 /* 886 * Make sure that fragments have a data length 887 * that's a non-zero multiple of 8 bytes. 888 */ 889 if (ip->ip_len == htons(0) || (ntohs(ip->ip_len) & 0x7) != 0) { 890 IPSTAT_INC(ips_toosmall); /* XXX */ 891 goto dropfrag; 892 } 893 m->m_flags |= M_IP_FRAG; 894 } else 895 m->m_flags &= ~M_IP_FRAG; 896 ip->ip_off = htons(ntohs(ip->ip_off) << 3); 897 898 /* 899 * Attempt reassembly; if it succeeds, proceed. 900 * ip_reass() will return a different mbuf. 901 */ 902 IPSTAT_INC(ips_fragments); 903 m->m_pkthdr.PH_loc.ptr = ip; 904 905 /* Previous ip_reass() started here. */ 906 /* 907 * Presence of header sizes in mbufs 908 * would confuse code below. 909 */ 910 m->m_data += hlen; 911 m->m_len -= hlen; 912 913 /* 914 * If first fragment to arrive, create a reassembly queue. 915 */ 916 if (fp == NULL) { 917 fp = uma_zalloc(V_ipq_zone, M_NOWAIT); 918 if (fp == NULL) 919 goto dropfrag; 920 #ifdef MAC 921 if (mac_ipq_init(fp, M_NOWAIT) != 0) { 922 uma_zfree(V_ipq_zone, fp); 923 fp = NULL; 924 goto dropfrag; 925 } 926 mac_ipq_create(m, fp); 927 #endif 928 TAILQ_INSERT_HEAD(head, fp, ipq_list); 929 V_nipq++; 930 fp->ipq_nfrags = 1; 931 fp->ipq_ttl = IPFRAGTTL; 932 fp->ipq_p = ip->ip_p; 933 fp->ipq_id = ip->ip_id; 934 fp->ipq_src = ip->ip_src; 935 fp->ipq_dst = ip->ip_dst; 936 fp->ipq_frags = m; 937 m->m_nextpkt = NULL; 938 goto done; 939 } else { 940 fp->ipq_nfrags++; 941 #ifdef MAC 942 mac_ipq_update(m, fp); 943 #endif 944 } 945 946 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.PH_loc.ptr)) 947 948 /* 949 * Handle ECN by comparing this segment with the first one; 950 * if CE is set, do not lose CE. 951 * drop if CE and not-ECT are mixed for the same packet. 952 */ 953 ecn = ip->ip_tos & IPTOS_ECN_MASK; 954 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 955 if (ecn == IPTOS_ECN_CE) { 956 if (ecn0 == IPTOS_ECN_NOTECT) 957 goto dropfrag; 958 if (ecn0 != IPTOS_ECN_CE) 959 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 960 } 961 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 962 goto dropfrag; 963 964 /* 965 * Find a segment which begins after this one does. 966 */ 967 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 968 if (ntohs(GETIP(q)->ip_off) > ntohs(ip->ip_off)) 969 break; 970 971 /* 972 * If there is a preceding segment, it may provide some of 973 * our data already. If so, drop the data from the incoming 974 * segment. If it provides all of our data, drop us, otherwise 975 * stick new segment in the proper place. 976 * 977 * If some of the data is dropped from the preceding 978 * segment, then it's checksum is invalidated. 979 */ 980 if (p) { 981 i = ntohs(GETIP(p)->ip_off) + ntohs(GETIP(p)->ip_len) - 982 ntohs(ip->ip_off); 983 if (i > 0) { 984 if (i >= ntohs(ip->ip_len)) 985 goto dropfrag; 986 m_adj(m, i); 987 m->m_pkthdr.csum_flags = 0; 988 ip->ip_off = htons(ntohs(ip->ip_off) + i); 989 ip->ip_len = htons(ntohs(ip->ip_len) - i); 990 } 991 m->m_nextpkt = p->m_nextpkt; 992 p->m_nextpkt = m; 993 } else { 994 m->m_nextpkt = fp->ipq_frags; 995 fp->ipq_frags = m; 996 } 997 998 /* 999 * While we overlap succeeding segments trim them or, 1000 * if they are completely covered, dequeue them. 1001 */ 1002 for (; q != NULL && ntohs(ip->ip_off) + ntohs(ip->ip_len) > 1003 ntohs(GETIP(q)->ip_off); q = nq) { 1004 i = (ntohs(ip->ip_off) + ntohs(ip->ip_len)) - 1005 ntohs(GETIP(q)->ip_off); 1006 if (i < ntohs(GETIP(q)->ip_len)) { 1007 GETIP(q)->ip_len = htons(ntohs(GETIP(q)->ip_len) - i); 1008 GETIP(q)->ip_off = htons(ntohs(GETIP(q)->ip_off) + i); 1009 m_adj(q, i); 1010 q->m_pkthdr.csum_flags = 0; 1011 break; 1012 } 1013 nq = q->m_nextpkt; 1014 m->m_nextpkt = nq; 1015 IPSTAT_INC(ips_fragdropped); 1016 fp->ipq_nfrags--; 1017 m_freem(q); 1018 } 1019 1020 /* 1021 * Check for complete reassembly and perform frag per packet 1022 * limiting. 1023 * 1024 * Frag limiting is performed here so that the nth frag has 1025 * a chance to complete the packet before we drop the packet. 1026 * As a result, n+1 frags are actually allowed per packet, but 1027 * only n will ever be stored. (n = maxfragsperpacket.) 1028 * 1029 */ 1030 next = 0; 1031 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1032 if (ntohs(GETIP(q)->ip_off) != next) { 1033 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1034 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1035 ip_freef(head, fp); 1036 } 1037 goto done; 1038 } 1039 next += ntohs(GETIP(q)->ip_len); 1040 } 1041 /* Make sure the last packet didn't have the IP_MF flag */ 1042 if (p->m_flags & M_IP_FRAG) { 1043 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1044 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1045 ip_freef(head, fp); 1046 } 1047 goto done; 1048 } 1049 1050 /* 1051 * Reassembly is complete. Make sure the packet is a sane size. 1052 */ 1053 q = fp->ipq_frags; 1054 ip = GETIP(q); 1055 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1056 IPSTAT_INC(ips_toolong); 1057 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1058 ip_freef(head, fp); 1059 goto done; 1060 } 1061 1062 /* 1063 * Concatenate fragments. 1064 */ 1065 m = q; 1066 t = m->m_next; 1067 m->m_next = NULL; 1068 m_cat(m, t); 1069 nq = q->m_nextpkt; 1070 q->m_nextpkt = NULL; 1071 for (q = nq; q != NULL; q = nq) { 1072 nq = q->m_nextpkt; 1073 q->m_nextpkt = NULL; 1074 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1075 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1076 m_cat(m, q); 1077 } 1078 /* 1079 * In order to do checksumming faster we do 'end-around carry' here 1080 * (and not in for{} loop), though it implies we are not going to 1081 * reassemble more than 64k fragments. 1082 */ 1083 m->m_pkthdr.csum_data = 1084 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1085 #ifdef MAC 1086 mac_ipq_reassemble(fp, m); 1087 mac_ipq_destroy(fp); 1088 #endif 1089 1090 /* 1091 * Create header for new ip packet by modifying header of first 1092 * packet; dequeue and discard fragment reassembly header. 1093 * Make header visible. 1094 */ 1095 ip->ip_len = htons((ip->ip_hl << 2) + next); 1096 ip->ip_src = fp->ipq_src; 1097 ip->ip_dst = fp->ipq_dst; 1098 TAILQ_REMOVE(head, fp, ipq_list); 1099 V_nipq--; 1100 uma_zfree(V_ipq_zone, fp); 1101 m->m_len += (ip->ip_hl << 2); 1102 m->m_data -= (ip->ip_hl << 2); 1103 /* some debugging cruft by sklower, below, will go away soon */ 1104 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1105 m_fixhdr(m); 1106 IPSTAT_INC(ips_reassembled); 1107 IPQ_UNLOCK(); 1108 return (m); 1109 1110 dropfrag: 1111 IPSTAT_INC(ips_fragdropped); 1112 if (fp != NULL) 1113 fp->ipq_nfrags--; 1114 m_freem(m); 1115 done: 1116 IPQ_UNLOCK(); 1117 return (NULL); 1118 1119 #undef GETIP 1120 } 1121 1122 /* 1123 * Free a fragment reassembly header and all 1124 * associated datagrams. 1125 */ 1126 static void 1127 ip_freef(struct ipqhead *fhp, struct ipq *fp) 1128 { 1129 struct mbuf *q; 1130 1131 IPQ_LOCK_ASSERT(); 1132 1133 while (fp->ipq_frags) { 1134 q = fp->ipq_frags; 1135 fp->ipq_frags = q->m_nextpkt; 1136 m_freem(q); 1137 } 1138 TAILQ_REMOVE(fhp, fp, ipq_list); 1139 uma_zfree(V_ipq_zone, fp); 1140 V_nipq--; 1141 } 1142 1143 /* 1144 * IP timer processing; 1145 * if a timer expires on a reassembly 1146 * queue, discard it. 1147 */ 1148 void 1149 ip_slowtimo(void) 1150 { 1151 VNET_ITERATOR_DECL(vnet_iter); 1152 struct ipq *fp; 1153 int i; 1154 1155 VNET_LIST_RLOCK_NOSLEEP(); 1156 IPQ_LOCK(); 1157 VNET_FOREACH(vnet_iter) { 1158 CURVNET_SET(vnet_iter); 1159 for (i = 0; i < IPREASS_NHASH; i++) { 1160 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) { 1161 struct ipq *fpp; 1162 1163 fpp = fp; 1164 fp = TAILQ_NEXT(fp, ipq_list); 1165 if(--fpp->ipq_ttl == 0) { 1166 IPSTAT_ADD(ips_fragtimeout, 1167 fpp->ipq_nfrags); 1168 ip_freef(&V_ipq[i], fpp); 1169 } 1170 } 1171 } 1172 /* 1173 * If we are over the maximum number of fragments 1174 * (due to the limit being lowered), drain off 1175 * enough to get down to the new limit. 1176 */ 1177 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) { 1178 for (i = 0; i < IPREASS_NHASH; i++) { 1179 while (V_nipq > V_maxnipq && 1180 !TAILQ_EMPTY(&V_ipq[i])) { 1181 IPSTAT_ADD(ips_fragdropped, 1182 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1183 ip_freef(&V_ipq[i], 1184 TAILQ_FIRST(&V_ipq[i])); 1185 } 1186 } 1187 } 1188 CURVNET_RESTORE(); 1189 } 1190 IPQ_UNLOCK(); 1191 VNET_LIST_RUNLOCK_NOSLEEP(); 1192 } 1193 1194 /* 1195 * Drain off all datagram fragments. 1196 */ 1197 static void 1198 ip_drain_locked(void) 1199 { 1200 int i; 1201 1202 IPQ_LOCK_ASSERT(); 1203 1204 for (i = 0; i < IPREASS_NHASH; i++) { 1205 while(!TAILQ_EMPTY(&V_ipq[i])) { 1206 IPSTAT_ADD(ips_fragdropped, 1207 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1208 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); 1209 } 1210 } 1211 } 1212 1213 void 1214 ip_drain(void) 1215 { 1216 VNET_ITERATOR_DECL(vnet_iter); 1217 1218 VNET_LIST_RLOCK_NOSLEEP(); 1219 IPQ_LOCK(); 1220 VNET_FOREACH(vnet_iter) { 1221 CURVNET_SET(vnet_iter); 1222 ip_drain_locked(); 1223 CURVNET_RESTORE(); 1224 } 1225 IPQ_UNLOCK(); 1226 VNET_LIST_RUNLOCK_NOSLEEP(); 1227 in_rtqdrain(); 1228 } 1229 1230 /* 1231 * The protocol to be inserted into ip_protox[] must be already registered 1232 * in inetsw[], either statically or through pf_proto_register(). 1233 */ 1234 int 1235 ipproto_register(short ipproto) 1236 { 1237 struct protosw *pr; 1238 1239 /* Sanity checks. */ 1240 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 1241 return (EPROTONOSUPPORT); 1242 1243 /* 1244 * The protocol slot must not be occupied by another protocol 1245 * already. An index pointing to IPPROTO_RAW is unused. 1246 */ 1247 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1248 if (pr == NULL) 1249 return (EPFNOSUPPORT); 1250 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1251 return (EEXIST); 1252 1253 /* Find the protocol position in inetsw[] and set the index. */ 1254 for (pr = inetdomain.dom_protosw; 1255 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1256 if (pr->pr_domain->dom_family == PF_INET && 1257 pr->pr_protocol && pr->pr_protocol == ipproto) { 1258 ip_protox[pr->pr_protocol] = pr - inetsw; 1259 return (0); 1260 } 1261 } 1262 return (EPROTONOSUPPORT); 1263 } 1264 1265 int 1266 ipproto_unregister(short ipproto) 1267 { 1268 struct protosw *pr; 1269 1270 /* Sanity checks. */ 1271 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 1272 return (EPROTONOSUPPORT); 1273 1274 /* Check if the protocol was indeed registered. */ 1275 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1276 if (pr == NULL) 1277 return (EPFNOSUPPORT); 1278 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1279 return (ENOENT); 1280 1281 /* Reset the protocol slot to IPPROTO_RAW. */ 1282 ip_protox[ipproto] = pr - inetsw; 1283 return (0); 1284 } 1285 1286 /* 1287 * Given address of next destination (final or next hop), return (referenced) 1288 * internet address info of interface to be used to get there. 1289 */ 1290 struct in_ifaddr * 1291 ip_rtaddr(struct in_addr dst, u_int fibnum) 1292 { 1293 struct route sro; 1294 struct sockaddr_in *sin; 1295 struct in_ifaddr *ia; 1296 1297 bzero(&sro, sizeof(sro)); 1298 sin = (struct sockaddr_in *)&sro.ro_dst; 1299 sin->sin_family = AF_INET; 1300 sin->sin_len = sizeof(*sin); 1301 sin->sin_addr = dst; 1302 in_rtalloc_ign(&sro, 0, fibnum); 1303 1304 if (sro.ro_rt == NULL) 1305 return (NULL); 1306 1307 ia = ifatoia(sro.ro_rt->rt_ifa); 1308 ifa_ref(&ia->ia_ifa); 1309 RTFREE(sro.ro_rt); 1310 return (ia); 1311 } 1312 1313 u_char inetctlerrmap[PRC_NCMDS] = { 1314 0, 0, 0, 0, 1315 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1316 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1317 EMSGSIZE, EHOSTUNREACH, 0, 0, 1318 0, 0, EHOSTUNREACH, 0, 1319 ENOPROTOOPT, ECONNREFUSED 1320 }; 1321 1322 /* 1323 * Forward a packet. If some error occurs return the sender 1324 * an icmp packet. Note we can't always generate a meaningful 1325 * icmp message because icmp doesn't have a large enough repertoire 1326 * of codes and types. 1327 * 1328 * If not forwarding, just drop the packet. This could be confusing 1329 * if ipforwarding was zero but some routing protocol was advancing 1330 * us as a gateway to somewhere. However, we must let the routing 1331 * protocol deal with that. 1332 * 1333 * The srcrt parameter indicates whether the packet is being forwarded 1334 * via a source route. 1335 */ 1336 void 1337 ip_forward(struct mbuf *m, int srcrt) 1338 { 1339 struct ip *ip = mtod(m, struct ip *); 1340 struct in_ifaddr *ia; 1341 struct mbuf *mcopy; 1342 struct in_addr dest; 1343 struct route ro; 1344 int error, type = 0, code = 0, mtu = 0; 1345 1346 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1347 IPSTAT_INC(ips_cantforward); 1348 m_freem(m); 1349 return; 1350 } 1351 #ifdef IPSTEALTH 1352 if (!V_ipstealth) { 1353 #endif 1354 if (ip->ip_ttl <= IPTTLDEC) { 1355 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1356 0, 0); 1357 return; 1358 } 1359 #ifdef IPSTEALTH 1360 } 1361 #endif 1362 1363 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); 1364 #ifndef IPSEC 1365 /* 1366 * 'ia' may be NULL if there is no route for this destination. 1367 * In case of IPsec, Don't discard it just yet, but pass it to 1368 * ip_output in case of outgoing IPsec policy. 1369 */ 1370 if (!srcrt && ia == NULL) { 1371 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1372 return; 1373 } 1374 #endif 1375 1376 /* 1377 * Save the IP header and at most 8 bytes of the payload, 1378 * in case we need to generate an ICMP message to the src. 1379 * 1380 * XXX this can be optimized a lot by saving the data in a local 1381 * buffer on the stack (72 bytes at most), and only allocating the 1382 * mbuf if really necessary. The vast majority of the packets 1383 * are forwarded without having to send an ICMP back (either 1384 * because unnecessary, or because rate limited), so we are 1385 * really we are wasting a lot of work here. 1386 * 1387 * We don't use m_copy() because it might return a reference 1388 * to a shared cluster. Both this function and ip_output() 1389 * assume exclusive access to the IP header in `m', so any 1390 * data in a cluster may change before we reach icmp_error(). 1391 */ 1392 mcopy = m_gethdr(M_NOWAIT, m->m_type); 1393 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) { 1394 /* 1395 * It's probably ok if the pkthdr dup fails (because 1396 * the deep copy of the tag chain failed), but for now 1397 * be conservative and just discard the copy since 1398 * code below may some day want the tags. 1399 */ 1400 m_free(mcopy); 1401 mcopy = NULL; 1402 } 1403 if (mcopy != NULL) { 1404 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy)); 1405 mcopy->m_pkthdr.len = mcopy->m_len; 1406 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1407 } 1408 1409 #ifdef IPSTEALTH 1410 if (!V_ipstealth) { 1411 #endif 1412 ip->ip_ttl -= IPTTLDEC; 1413 #ifdef IPSTEALTH 1414 } 1415 #endif 1416 1417 /* 1418 * If forwarding packet using same interface that it came in on, 1419 * perhaps should send a redirect to sender to shortcut a hop. 1420 * Only send redirect if source is sending directly to us, 1421 * and if packet was not source routed (or has any options). 1422 * Also, don't send redirect if forwarding using a default route 1423 * or a route modified by a redirect. 1424 */ 1425 dest.s_addr = 0; 1426 if (!srcrt && V_ipsendredirects && 1427 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) { 1428 struct sockaddr_in *sin; 1429 struct rtentry *rt; 1430 1431 bzero(&ro, sizeof(ro)); 1432 sin = (struct sockaddr_in *)&ro.ro_dst; 1433 sin->sin_family = AF_INET; 1434 sin->sin_len = sizeof(*sin); 1435 sin->sin_addr = ip->ip_dst; 1436 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 1437 1438 rt = ro.ro_rt; 1439 1440 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1441 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1442 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1443 u_long src = ntohl(ip->ip_src.s_addr); 1444 1445 if (RTA(rt) && 1446 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1447 if (rt->rt_flags & RTF_GATEWAY) 1448 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1449 else 1450 dest.s_addr = ip->ip_dst.s_addr; 1451 /* Router requirements says to only send host redirects */ 1452 type = ICMP_REDIRECT; 1453 code = ICMP_REDIRECT_HOST; 1454 } 1455 } 1456 if (rt) 1457 RTFREE(rt); 1458 } 1459 1460 /* 1461 * Try to cache the route MTU from ip_output so we can consider it for 1462 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. 1463 */ 1464 bzero(&ro, sizeof(ro)); 1465 1466 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1467 1468 if (error == EMSGSIZE && ro.ro_rt) 1469 mtu = ro.ro_rt->rt_mtu; 1470 RO_RTFREE(&ro); 1471 1472 if (error) 1473 IPSTAT_INC(ips_cantforward); 1474 else { 1475 IPSTAT_INC(ips_forward); 1476 if (type) 1477 IPSTAT_INC(ips_redirectsent); 1478 else { 1479 if (mcopy) 1480 m_freem(mcopy); 1481 if (ia != NULL) 1482 ifa_free(&ia->ia_ifa); 1483 return; 1484 } 1485 } 1486 if (mcopy == NULL) { 1487 if (ia != NULL) 1488 ifa_free(&ia->ia_ifa); 1489 return; 1490 } 1491 1492 switch (error) { 1493 1494 case 0: /* forwarded, but need redirect */ 1495 /* type, code set above */ 1496 break; 1497 1498 case ENETUNREACH: 1499 case EHOSTUNREACH: 1500 case ENETDOWN: 1501 case EHOSTDOWN: 1502 default: 1503 type = ICMP_UNREACH; 1504 code = ICMP_UNREACH_HOST; 1505 break; 1506 1507 case EMSGSIZE: 1508 type = ICMP_UNREACH; 1509 code = ICMP_UNREACH_NEEDFRAG; 1510 1511 #ifdef IPSEC 1512 /* 1513 * If IPsec is configured for this path, 1514 * override any possibly mtu value set by ip_output. 1515 */ 1516 mtu = ip_ipsec_mtu(mcopy, mtu); 1517 #endif /* IPSEC */ 1518 /* 1519 * If the MTU was set before make sure we are below the 1520 * interface MTU. 1521 * If the MTU wasn't set before use the interface mtu or 1522 * fall back to the next smaller mtu step compared to the 1523 * current packet size. 1524 */ 1525 if (mtu != 0) { 1526 if (ia != NULL) 1527 mtu = min(mtu, ia->ia_ifp->if_mtu); 1528 } else { 1529 if (ia != NULL) 1530 mtu = ia->ia_ifp->if_mtu; 1531 else 1532 mtu = ip_next_mtu(ntohs(ip->ip_len), 0); 1533 } 1534 IPSTAT_INC(ips_cantfrag); 1535 break; 1536 1537 case ENOBUFS: 1538 /* 1539 * A router should not generate ICMP_SOURCEQUENCH as 1540 * required in RFC1812 Requirements for IP Version 4 Routers. 1541 * Source quench could be a big problem under DoS attacks, 1542 * or if the underlying interface is rate-limited. 1543 * Those who need source quench packets may re-enable them 1544 * via the net.inet.ip.sendsourcequench sysctl. 1545 */ 1546 if (V_ip_sendsourcequench == 0) { 1547 m_freem(mcopy); 1548 if (ia != NULL) 1549 ifa_free(&ia->ia_ifa); 1550 return; 1551 } else { 1552 type = ICMP_SOURCEQUENCH; 1553 code = 0; 1554 } 1555 break; 1556 1557 case EACCES: /* ipfw denied packet */ 1558 m_freem(mcopy); 1559 if (ia != NULL) 1560 ifa_free(&ia->ia_ifa); 1561 return; 1562 } 1563 if (ia != NULL) 1564 ifa_free(&ia->ia_ifa); 1565 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1566 } 1567 1568 void 1569 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1570 struct mbuf *m) 1571 { 1572 1573 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1574 struct bintime bt; 1575 1576 bintime(&bt); 1577 if (inp->inp_socket->so_options & SO_BINTIME) { 1578 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt), 1579 SCM_BINTIME, SOL_SOCKET); 1580 if (*mp) 1581 mp = &(*mp)->m_next; 1582 } 1583 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1584 struct timeval tv; 1585 1586 bintime2timeval(&bt, &tv); 1587 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), 1588 SCM_TIMESTAMP, SOL_SOCKET); 1589 if (*mp) 1590 mp = &(*mp)->m_next; 1591 } 1592 } 1593 if (inp->inp_flags & INP_RECVDSTADDR) { 1594 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst, 1595 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1596 if (*mp) 1597 mp = &(*mp)->m_next; 1598 } 1599 if (inp->inp_flags & INP_RECVTTL) { 1600 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl, 1601 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1602 if (*mp) 1603 mp = &(*mp)->m_next; 1604 } 1605 #ifdef notyet 1606 /* XXX 1607 * Moving these out of udp_input() made them even more broken 1608 * than they already were. 1609 */ 1610 /* options were tossed already */ 1611 if (inp->inp_flags & INP_RECVOPTS) { 1612 *mp = sbcreatecontrol((caddr_t)opts_deleted_above, 1613 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1614 if (*mp) 1615 mp = &(*mp)->m_next; 1616 } 1617 /* ip_srcroute doesn't do what we want here, need to fix */ 1618 if (inp->inp_flags & INP_RECVRETOPTS) { 1619 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m), 1620 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1621 if (*mp) 1622 mp = &(*mp)->m_next; 1623 } 1624 #endif 1625 if (inp->inp_flags & INP_RECVIF) { 1626 struct ifnet *ifp; 1627 struct sdlbuf { 1628 struct sockaddr_dl sdl; 1629 u_char pad[32]; 1630 } sdlbuf; 1631 struct sockaddr_dl *sdp; 1632 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1633 1634 if ((ifp = m->m_pkthdr.rcvif) && 1635 ifp->if_index && ifp->if_index <= V_if_index) { 1636 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1637 /* 1638 * Change our mind and don't try copy. 1639 */ 1640 if (sdp->sdl_family != AF_LINK || 1641 sdp->sdl_len > sizeof(sdlbuf)) { 1642 goto makedummy; 1643 } 1644 bcopy(sdp, sdl2, sdp->sdl_len); 1645 } else { 1646 makedummy: 1647 sdl2->sdl_len = 1648 offsetof(struct sockaddr_dl, sdl_data[0]); 1649 sdl2->sdl_family = AF_LINK; 1650 sdl2->sdl_index = 0; 1651 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1652 } 1653 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len, 1654 IP_RECVIF, IPPROTO_IP); 1655 if (*mp) 1656 mp = &(*mp)->m_next; 1657 } 1658 if (inp->inp_flags & INP_RECVTOS) { 1659 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos, 1660 sizeof(u_char), IP_RECVTOS, IPPROTO_IP); 1661 if (*mp) 1662 mp = &(*mp)->m_next; 1663 } 1664 } 1665 1666 /* 1667 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1668 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1669 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1670 * compiled. 1671 */ 1672 static VNET_DEFINE(int, ip_rsvp_on); 1673 VNET_DEFINE(struct socket *, ip_rsvpd); 1674 1675 #define V_ip_rsvp_on VNET(ip_rsvp_on) 1676 1677 int 1678 ip_rsvp_init(struct socket *so) 1679 { 1680 1681 if (so->so_type != SOCK_RAW || 1682 so->so_proto->pr_protocol != IPPROTO_RSVP) 1683 return EOPNOTSUPP; 1684 1685 if (V_ip_rsvpd != NULL) 1686 return EADDRINUSE; 1687 1688 V_ip_rsvpd = so; 1689 /* 1690 * This may seem silly, but we need to be sure we don't over-increment 1691 * the RSVP counter, in case something slips up. 1692 */ 1693 if (!V_ip_rsvp_on) { 1694 V_ip_rsvp_on = 1; 1695 V_rsvp_on++; 1696 } 1697 1698 return 0; 1699 } 1700 1701 int 1702 ip_rsvp_done(void) 1703 { 1704 1705 V_ip_rsvpd = NULL; 1706 /* 1707 * This may seem silly, but we need to be sure we don't over-decrement 1708 * the RSVP counter, in case something slips up. 1709 */ 1710 if (V_ip_rsvp_on) { 1711 V_ip_rsvp_on = 0; 1712 V_rsvp_on--; 1713 } 1714 return 0; 1715 } 1716 1717 void 1718 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1719 { 1720 1721 if (rsvp_input_p) { /* call the real one if loaded */ 1722 rsvp_input_p(m, off); 1723 return; 1724 } 1725 1726 /* Can still get packets with rsvp_on = 0 if there is a local member 1727 * of the group to which the RSVP packet is addressed. But in this 1728 * case we want to throw the packet away. 1729 */ 1730 1731 if (!V_rsvp_on) { 1732 m_freem(m); 1733 return; 1734 } 1735 1736 if (V_ip_rsvpd != NULL) { 1737 rip_input(m, off); 1738 return; 1739 } 1740 /* Drop the packet */ 1741 m_freem(m); 1742 } 1743