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/callout.h> 44 #include <sys/mbuf.h> 45 #include <sys/malloc.h> 46 #include <sys/domain.h> 47 #include <sys/protosw.h> 48 #include <sys/socket.h> 49 #include <sys/time.h> 50 #include <sys/kernel.h> 51 #include <sys/lock.h> 52 #include <sys/rwlock.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 #include <net/flowtable.h> 65 66 #include <netinet/in.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 VNET_DEFINE(int, ip_defttl) = IPDEFTTL; 106 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 107 &VNET_NAME(ip_defttl), 0, 108 "Maximum TTL on IP packets"); 109 110 static VNET_DEFINE(int, ip_keepfaith); 111 #define V_ip_keepfaith VNET(ip_keepfaith) 112 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 113 &VNET_NAME(ip_keepfaith), 0, 114 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 115 116 static VNET_DEFINE(int, ip_sendsourcequench); 117 #define V_ip_sendsourcequench VNET(ip_sendsourcequench) 118 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 119 &VNET_NAME(ip_sendsourcequench), 0, 120 "Enable the transmission of source quench packets"); 121 122 VNET_DEFINE(int, ip_do_randomid); 123 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 124 &VNET_NAME(ip_do_randomid), 0, 125 "Assign random ip_id values"); 126 127 /* 128 * XXX - Setting ip_checkinterface mostly implements the receive side of 129 * the Strong ES model described in RFC 1122, but since the routing table 130 * and transmit implementation do not implement the Strong ES model, 131 * setting this to 1 results in an odd hybrid. 132 * 133 * XXX - ip_checkinterface currently must be disabled if you use ipnat 134 * to translate the destination address to another local interface. 135 * 136 * XXX - ip_checkinterface must be disabled if you add IP aliases 137 * to the loopback interface instead of the interface where the 138 * packets for those addresses are received. 139 */ 140 static VNET_DEFINE(int, ip_checkinterface); 141 #define V_ip_checkinterface VNET(ip_checkinterface) 142 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 143 &VNET_NAME(ip_checkinterface), 0, 144 "Verify packet arrives on correct interface"); 145 146 VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */ 147 148 static struct netisr_handler ip_nh = { 149 .nh_name = "ip", 150 .nh_handler = ip_input, 151 .nh_proto = NETISR_IP, 152 .nh_policy = NETISR_POLICY_FLOW, 153 }; 154 155 extern struct domain inetdomain; 156 extern struct protosw inetsw[]; 157 u_char ip_protox[IPPROTO_MAX]; 158 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */ 159 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */ 160 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */ 161 162 VNET_DEFINE(struct ipstat, ipstat); 163 SYSCTL_VNET_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 164 &VNET_NAME(ipstat), ipstat, 165 "IP statistics (struct ipstat, netinet/ip_var.h)"); 166 167 static VNET_DEFINE(uma_zone_t, ipq_zone); 168 static VNET_DEFINE(TAILQ_HEAD(ipqhead, ipq), ipq[IPREASS_NHASH]); 169 static struct mtx ipqlock; 170 171 #define V_ipq_zone VNET(ipq_zone) 172 #define V_ipq VNET(ipq) 173 174 #define IPQ_LOCK() mtx_lock(&ipqlock) 175 #define IPQ_UNLOCK() mtx_unlock(&ipqlock) 176 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 177 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 178 179 static void maxnipq_update(void); 180 static void ipq_zone_change(void *); 181 static void ip_drain_locked(void); 182 183 static VNET_DEFINE(int, maxnipq); /* Administrative limit on # reass queues. */ 184 static VNET_DEFINE(int, nipq); /* Total # of reass queues */ 185 #define V_maxnipq VNET(maxnipq) 186 #define V_nipq VNET(nipq) 187 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, 188 &VNET_NAME(nipq), 0, 189 "Current number of IPv4 fragment reassembly queue entries"); 190 191 static VNET_DEFINE(int, maxfragsperpacket); 192 #define V_maxfragsperpacket VNET(maxfragsperpacket) 193 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 194 &VNET_NAME(maxfragsperpacket), 0, 195 "Maximum number of IPv4 fragments allowed per packet"); 196 197 struct callout ipport_tick_callout; 198 199 #ifdef IPCTL_DEFMTU 200 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 201 &ip_mtu, 0, "Default MTU"); 202 #endif 203 204 #ifdef IPSTEALTH 205 VNET_DEFINE(int, ipstealth); 206 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 207 &VNET_NAME(ipstealth), 0, 208 "IP stealth mode, no TTL decrementation on forwarding"); 209 #endif 210 211 #ifdef FLOWTABLE 212 static VNET_DEFINE(int, ip_output_flowtable_size) = 2048; 213 VNET_DEFINE(struct flowtable *, ip_ft); 214 #define V_ip_output_flowtable_size VNET(ip_output_flowtable_size) 215 216 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, output_flowtable_size, CTLFLAG_RDTUN, 217 &VNET_NAME(ip_output_flowtable_size), 2048, 218 "number of entries in the per-cpu output flow caches"); 219 #endif 220 221 VNET_DEFINE(int, fw_one_pass) = 1; 222 223 static void ip_freef(struct ipqhead *, struct ipq *); 224 225 /* 226 * Kernel module interface for updating ipstat. The argument is an index 227 * into ipstat treated as an array of u_long. While this encodes the general 228 * layout of ipstat into the caller, it doesn't encode its location, so that 229 * future changes to add, for example, per-CPU stats support won't cause 230 * binary compatibility problems for kernel modules. 231 */ 232 void 233 kmod_ipstat_inc(int statnum) 234 { 235 236 (*((u_long *)&V_ipstat + statnum))++; 237 } 238 239 void 240 kmod_ipstat_dec(int statnum) 241 { 242 243 (*((u_long *)&V_ipstat + statnum))--; 244 } 245 246 static int 247 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 248 { 249 int error, qlimit; 250 251 netisr_getqlimit(&ip_nh, &qlimit); 252 error = sysctl_handle_int(oidp, &qlimit, 0, req); 253 if (error || !req->newptr) 254 return (error); 255 if (qlimit < 1) 256 return (EINVAL); 257 return (netisr_setqlimit(&ip_nh, qlimit)); 258 } 259 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, 260 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I", 261 "Maximum size of the IP input queue"); 262 263 static int 264 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) 265 { 266 u_int64_t qdrops_long; 267 int error, qdrops; 268 269 netisr_getqdrops(&ip_nh, &qdrops_long); 270 qdrops = qdrops_long; 271 error = sysctl_handle_int(oidp, &qdrops, 0, req); 272 if (error || !req->newptr) 273 return (error); 274 if (qdrops != 0) 275 return (EINVAL); 276 netisr_clearqdrops(&ip_nh); 277 return (0); 278 } 279 280 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, 281 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I", 282 "Number of packets dropped from the IP input queue"); 283 284 /* 285 * IP initialization: fill in IP protocol switch table. 286 * All protocols not implemented in kernel go to raw IP protocol handler. 287 */ 288 void 289 ip_init(void) 290 { 291 struct protosw *pr; 292 int i; 293 294 V_ip_id = time_second & 0xffff; 295 296 TAILQ_INIT(&V_in_ifaddrhead); 297 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 298 299 /* Initialize IP reassembly queue. */ 300 for (i = 0; i < IPREASS_NHASH; i++) 301 TAILQ_INIT(&V_ipq[i]); 302 V_maxnipq = nmbclusters / 32; 303 V_maxfragsperpacket = 16; 304 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 305 NULL, UMA_ALIGN_PTR, 0); 306 maxnipq_update(); 307 308 /* Initialize packet filter hooks. */ 309 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF; 310 V_inet_pfil_hook.ph_af = AF_INET; 311 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0) 312 printf("%s: WARNING: unable to register pfil hook, " 313 "error %d\n", __func__, i); 314 315 #ifdef FLOWTABLE 316 if (TUNABLE_INT_FETCH("net.inet.ip.output_flowtable_size", 317 &V_ip_output_flowtable_size)) { 318 if (V_ip_output_flowtable_size < 256) 319 V_ip_output_flowtable_size = 256; 320 if (!powerof2(V_ip_output_flowtable_size)) { 321 printf("flowtable must be power of 2 size\n"); 322 V_ip_output_flowtable_size = 2048; 323 } 324 } else { 325 /* 326 * round up to the next power of 2 327 */ 328 V_ip_output_flowtable_size = 1 << fls((1024 + maxusers * 64)-1); 329 } 330 V_ip_ft = flowtable_alloc("ipv4", V_ip_output_flowtable_size, FL_PCPU); 331 #endif 332 333 /* Skip initialization of globals for non-default instances. */ 334 if (!IS_DEFAULT_VNET(curvnet)) 335 return; 336 337 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 338 if (pr == NULL) 339 panic("ip_init: PF_INET not found"); 340 341 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 342 for (i = 0; i < IPPROTO_MAX; i++) 343 ip_protox[i] = pr - inetsw; 344 /* 345 * Cycle through IP protocols and put them into the appropriate place 346 * in ip_protox[]. 347 */ 348 for (pr = inetdomain.dom_protosw; 349 pr < inetdomain.dom_protoswNPROTOSW; pr++) 350 if (pr->pr_domain->dom_family == PF_INET && 351 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 352 /* Be careful to only index valid IP protocols. */ 353 if (pr->pr_protocol < IPPROTO_MAX) 354 ip_protox[pr->pr_protocol] = pr - inetsw; 355 } 356 357 /* Start ipport_tick. */ 358 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE); 359 callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL); 360 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, 361 SHUTDOWN_PRI_DEFAULT); 362 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, 363 NULL, EVENTHANDLER_PRI_ANY); 364 365 /* Initialize various other remaining things. */ 366 IPQ_LOCK_INIT(); 367 netisr_register(&ip_nh); 368 } 369 370 #ifdef VIMAGE 371 void 372 ip_destroy(void) 373 { 374 375 /* Cleanup in_ifaddr hash table; should be empty. */ 376 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask); 377 378 IPQ_LOCK(); 379 ip_drain_locked(); 380 IPQ_UNLOCK(); 381 382 uma_zdestroy(V_ipq_zone); 383 } 384 #endif 385 386 void 387 ip_fini(void *xtp) 388 { 389 390 callout_stop(&ipport_tick_callout); 391 } 392 393 /* 394 * Ip input routine. Checksum and byte swap header. If fragmented 395 * try to reassemble. Process options. Pass to next level. 396 */ 397 void 398 ip_input(struct mbuf *m) 399 { 400 struct ip *ip = NULL; 401 struct in_ifaddr *ia = NULL; 402 struct ifaddr *ifa; 403 struct ifnet *ifp; 404 int checkif, hlen = 0; 405 u_short sum; 406 int dchg = 0; /* dest changed after fw */ 407 struct in_addr odst; /* original dst address */ 408 409 M_ASSERTPKTHDR(m); 410 411 if (m->m_flags & M_FASTFWD_OURS) { 412 /* 413 * Firewall or NAT changed destination to local. 414 * We expect ip_len and ip_off to be in host byte order. 415 */ 416 m->m_flags &= ~M_FASTFWD_OURS; 417 /* Set up some basics that will be used later. */ 418 ip = mtod(m, struct ip *); 419 hlen = ip->ip_hl << 2; 420 goto ours; 421 } 422 423 IPSTAT_INC(ips_total); 424 425 if (m->m_pkthdr.len < sizeof(struct ip)) 426 goto tooshort; 427 428 if (m->m_len < sizeof (struct ip) && 429 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 430 IPSTAT_INC(ips_toosmall); 431 return; 432 } 433 ip = mtod(m, struct ip *); 434 435 if (ip->ip_v != IPVERSION) { 436 IPSTAT_INC(ips_badvers); 437 goto bad; 438 } 439 440 hlen = ip->ip_hl << 2; 441 if (hlen < sizeof(struct ip)) { /* minimum header length */ 442 IPSTAT_INC(ips_badhlen); 443 goto bad; 444 } 445 if (hlen > m->m_len) { 446 if ((m = m_pullup(m, hlen)) == NULL) { 447 IPSTAT_INC(ips_badhlen); 448 return; 449 } 450 ip = mtod(m, struct ip *); 451 } 452 453 /* 127/8 must not appear on wire - RFC1122 */ 454 ifp = m->m_pkthdr.rcvif; 455 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 456 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 457 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 458 IPSTAT_INC(ips_badaddr); 459 goto bad; 460 } 461 } 462 463 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 464 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 465 } else { 466 if (hlen == sizeof(struct ip)) { 467 sum = in_cksum_hdr(ip); 468 } else { 469 sum = in_cksum(m, hlen); 470 } 471 } 472 if (sum) { 473 IPSTAT_INC(ips_badsum); 474 goto bad; 475 } 476 477 #ifdef ALTQ 478 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 479 /* packet is dropped by traffic conditioner */ 480 return; 481 #endif 482 483 /* 484 * Convert fields to host representation. 485 */ 486 ip->ip_len = ntohs(ip->ip_len); 487 if (ip->ip_len < hlen) { 488 IPSTAT_INC(ips_badlen); 489 goto bad; 490 } 491 ip->ip_off = ntohs(ip->ip_off); 492 493 /* 494 * Check that the amount of data in the buffers 495 * is as at least much as the IP header would have us expect. 496 * Trim mbufs if longer than we expect. 497 * Drop packet if shorter than we expect. 498 */ 499 if (m->m_pkthdr.len < ip->ip_len) { 500 tooshort: 501 IPSTAT_INC(ips_tooshort); 502 goto bad; 503 } 504 if (m->m_pkthdr.len > ip->ip_len) { 505 if (m->m_len == m->m_pkthdr.len) { 506 m->m_len = ip->ip_len; 507 m->m_pkthdr.len = ip->ip_len; 508 } else 509 m_adj(m, ip->ip_len - m->m_pkthdr.len); 510 } 511 #ifdef IPSEC 512 /* 513 * Bypass packet filtering for packets from a tunnel (gif). 514 */ 515 if (ip_ipsec_filtertunnel(m)) 516 goto passin; 517 #endif /* IPSEC */ 518 519 /* 520 * Run through list of hooks for input packets. 521 * 522 * NB: Beware of the destination address changing (e.g. 523 * by NAT rewriting). When this happens, tell 524 * ip_forward to do the right thing. 525 */ 526 527 /* Jump over all PFIL processing if hooks are not active. */ 528 if (!PFIL_HOOKED(&V_inet_pfil_hook)) 529 goto passin; 530 531 odst = ip->ip_dst; 532 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0) 533 return; 534 if (m == NULL) /* consumed by filter */ 535 return; 536 537 ip = mtod(m, struct ip *); 538 dchg = (odst.s_addr != ip->ip_dst.s_addr); 539 ifp = m->m_pkthdr.rcvif; 540 541 #ifdef IPFIREWALL_FORWARD 542 if (m->m_flags & M_FASTFWD_OURS) { 543 m->m_flags &= ~M_FASTFWD_OURS; 544 goto ours; 545 } 546 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 547 /* 548 * Directly ship the packet on. This allows forwarding 549 * packets originally destined to us to some other directly 550 * connected host. 551 */ 552 ip_forward(m, dchg); 553 return; 554 } 555 #endif /* IPFIREWALL_FORWARD */ 556 557 passin: 558 /* 559 * Process options and, if not destined for us, 560 * ship it on. ip_dooptions returns 1 when an 561 * error was detected (causing an icmp message 562 * to be sent and the original packet to be freed). 563 */ 564 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 565 return; 566 567 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 568 * matter if it is destined to another node, or whether it is 569 * a multicast one, RSVP wants it! and prevents it from being forwarded 570 * anywhere else. Also checks if the rsvp daemon is running before 571 * grabbing the packet. 572 */ 573 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 574 goto ours; 575 576 /* 577 * Check our list of addresses, to see if the packet is for us. 578 * If we don't have any addresses, assume any unicast packet 579 * we receive might be for us (and let the upper layers deal 580 * with it). 581 */ 582 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 583 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 584 goto ours; 585 586 /* 587 * Enable a consistency check between the destination address 588 * and the arrival interface for a unicast packet (the RFC 1122 589 * strong ES model) if IP forwarding is disabled and the packet 590 * is not locally generated and the packet is not subject to 591 * 'ipfw fwd'. 592 * 593 * XXX - Checking also should be disabled if the destination 594 * address is ipnat'ed to a different interface. 595 * 596 * XXX - Checking is incompatible with IP aliases added 597 * to the loopback interface instead of the interface where 598 * the packets are received. 599 * 600 * XXX - This is the case for carp vhost IPs as well so we 601 * insert a workaround. If the packet got here, we already 602 * checked with carp_iamatch() and carp_forus(). 603 */ 604 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 605 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && 606 ifp->if_carp == NULL && (dchg == 0); 607 608 /* 609 * Check for exact addresses in the hash bucket. 610 */ 611 /* IN_IFADDR_RLOCK(); */ 612 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 613 /* 614 * If the address matches, verify that the packet 615 * arrived via the correct interface if checking is 616 * enabled. 617 */ 618 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 619 (!checkif || ia->ia_ifp == ifp)) { 620 ifa_ref(&ia->ia_ifa); 621 /* IN_IFADDR_RUNLOCK(); */ 622 goto ours; 623 } 624 } 625 /* IN_IFADDR_RUNLOCK(); */ 626 627 /* 628 * Check for broadcast addresses. 629 * 630 * Only accept broadcast packets that arrive via the matching 631 * interface. Reception of forwarded directed broadcasts would 632 * be handled via ip_forward() and ether_output() with the loopback 633 * into the stack for SIMPLEX interfaces handled by ether_output(). 634 */ 635 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 636 IF_ADDR_LOCK(ifp); 637 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 638 if (ifa->ifa_addr->sa_family != AF_INET) 639 continue; 640 ia = ifatoia(ifa); 641 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 642 ip->ip_dst.s_addr) { 643 ifa_ref(ifa); 644 IF_ADDR_UNLOCK(ifp); 645 goto ours; 646 } 647 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) { 648 ifa_ref(ifa); 649 IF_ADDR_UNLOCK(ifp); 650 goto ours; 651 } 652 #ifdef BOOTP_COMPAT 653 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { 654 ifa_ref(ifa); 655 IF_ADDR_UNLOCK(ifp); 656 goto ours; 657 } 658 #endif 659 } 660 IF_ADDR_UNLOCK(ifp); 661 ia = NULL; 662 } 663 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 664 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 665 IPSTAT_INC(ips_cantforward); 666 m_freem(m); 667 return; 668 } 669 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 670 if (V_ip_mrouter) { 671 /* 672 * If we are acting as a multicast router, all 673 * incoming multicast packets are passed to the 674 * kernel-level multicast forwarding function. 675 * The packet is returned (relatively) intact; if 676 * ip_mforward() returns a non-zero value, the packet 677 * must be discarded, else it may be accepted below. 678 */ 679 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 680 IPSTAT_INC(ips_cantforward); 681 m_freem(m); 682 return; 683 } 684 685 /* 686 * The process-level routing daemon needs to receive 687 * all multicast IGMP packets, whether or not this 688 * host belongs to their destination groups. 689 */ 690 if (ip->ip_p == IPPROTO_IGMP) 691 goto ours; 692 IPSTAT_INC(ips_forward); 693 } 694 /* 695 * Assume the packet is for us, to avoid prematurely taking 696 * a lock on the in_multi hash. Protocols must perform 697 * their own filtering and update statistics accordingly. 698 */ 699 goto ours; 700 } 701 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 702 goto ours; 703 if (ip->ip_dst.s_addr == INADDR_ANY) 704 goto ours; 705 706 /* 707 * FAITH(Firewall Aided Internet Translator) 708 */ 709 if (ifp && ifp->if_type == IFT_FAITH) { 710 if (V_ip_keepfaith) { 711 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 712 goto ours; 713 } 714 m_freem(m); 715 return; 716 } 717 718 /* 719 * Not for us; forward if possible and desirable. 720 */ 721 if (V_ipforwarding == 0) { 722 IPSTAT_INC(ips_cantforward); 723 m_freem(m); 724 } else { 725 #ifdef IPSEC 726 if (ip_ipsec_fwd(m)) 727 goto bad; 728 #endif /* IPSEC */ 729 ip_forward(m, dchg); 730 } 731 return; 732 733 ours: 734 #ifdef IPSTEALTH 735 /* 736 * IPSTEALTH: Process non-routing options only 737 * if the packet is destined for us. 738 */ 739 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) { 740 if (ia != NULL) 741 ifa_free(&ia->ia_ifa); 742 return; 743 } 744 #endif /* IPSTEALTH */ 745 746 /* Count the packet in the ip address stats */ 747 if (ia != NULL) { 748 ia->ia_ifa.if_ipackets++; 749 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 750 ifa_free(&ia->ia_ifa); 751 } 752 753 /* 754 * Attempt reassembly; if it succeeds, proceed. 755 * ip_reass() will return a different mbuf. 756 */ 757 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 758 m = ip_reass(m); 759 if (m == NULL) 760 return; 761 ip = mtod(m, struct ip *); 762 /* Get the header length of the reassembled packet */ 763 hlen = ip->ip_hl << 2; 764 } 765 766 /* 767 * Further protocols expect the packet length to be w/o the 768 * IP header. 769 */ 770 ip->ip_len -= hlen; 771 772 #ifdef IPSEC 773 /* 774 * enforce IPsec policy checking if we are seeing last header. 775 * note that we do not visit this with protocols with pcb layer 776 * code - like udp/tcp/raw ip. 777 */ 778 if (ip_ipsec_input(m)) 779 goto bad; 780 #endif /* IPSEC */ 781 782 /* 783 * Switch out to protocol's input routine. 784 */ 785 IPSTAT_INC(ips_delivered); 786 787 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 788 return; 789 bad: 790 m_freem(m); 791 } 792 793 /* 794 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 795 * max has slightly different semantics than the sysctl, for historical 796 * reasons. 797 */ 798 static void 799 maxnipq_update(void) 800 { 801 802 /* 803 * -1 for unlimited allocation. 804 */ 805 if (V_maxnipq < 0) 806 uma_zone_set_max(V_ipq_zone, 0); 807 /* 808 * Positive number for specific bound. 809 */ 810 if (V_maxnipq > 0) 811 uma_zone_set_max(V_ipq_zone, V_maxnipq); 812 /* 813 * Zero specifies no further fragment queue allocation -- set the 814 * bound very low, but rely on implementation elsewhere to actually 815 * prevent allocation and reclaim current queues. 816 */ 817 if (V_maxnipq == 0) 818 uma_zone_set_max(V_ipq_zone, 1); 819 } 820 821 static void 822 ipq_zone_change(void *tag) 823 { 824 825 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) { 826 V_maxnipq = nmbclusters / 32; 827 maxnipq_update(); 828 } 829 } 830 831 static int 832 sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 833 { 834 int error, i; 835 836 i = V_maxnipq; 837 error = sysctl_handle_int(oidp, &i, 0, req); 838 if (error || !req->newptr) 839 return (error); 840 841 /* 842 * XXXRW: Might be a good idea to sanity check the argument and place 843 * an extreme upper bound. 844 */ 845 if (i < -1) 846 return (EINVAL); 847 V_maxnipq = i; 848 maxnipq_update(); 849 return (0); 850 } 851 852 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 853 NULL, 0, sysctl_maxnipq, "I", 854 "Maximum number of IPv4 fragment reassembly queue entries"); 855 856 /* 857 * Take incoming datagram fragment and try to reassemble it into 858 * whole datagram. If the argument is the first fragment or one 859 * in between the function will return NULL and store the mbuf 860 * in the fragment chain. If the argument is the last fragment 861 * the packet will be reassembled and the pointer to the new 862 * mbuf returned for further processing. Only m_tags attached 863 * to the first packet/fragment are preserved. 864 * The IP header is *NOT* adjusted out of iplen. 865 */ 866 struct mbuf * 867 ip_reass(struct mbuf *m) 868 { 869 struct ip *ip; 870 struct mbuf *p, *q, *nq, *t; 871 struct ipq *fp = NULL; 872 struct ipqhead *head; 873 int i, hlen, next; 874 u_int8_t ecn, ecn0; 875 u_short hash; 876 877 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 878 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) { 879 IPSTAT_INC(ips_fragments); 880 IPSTAT_INC(ips_fragdropped); 881 m_freem(m); 882 return (NULL); 883 } 884 885 ip = mtod(m, struct ip *); 886 hlen = ip->ip_hl << 2; 887 888 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 889 head = &V_ipq[hash]; 890 IPQ_LOCK(); 891 892 /* 893 * Look for queue of fragments 894 * of this datagram. 895 */ 896 TAILQ_FOREACH(fp, head, ipq_list) 897 if (ip->ip_id == fp->ipq_id && 898 ip->ip_src.s_addr == fp->ipq_src.s_addr && 899 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 900 #ifdef MAC 901 mac_ipq_match(m, fp) && 902 #endif 903 ip->ip_p == fp->ipq_p) 904 goto found; 905 906 fp = NULL; 907 908 /* 909 * Attempt to trim the number of allocated fragment queues if it 910 * exceeds the administrative limit. 911 */ 912 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) { 913 /* 914 * drop something from the tail of the current queue 915 * before proceeding further 916 */ 917 struct ipq *q = TAILQ_LAST(head, ipqhead); 918 if (q == NULL) { /* gak */ 919 for (i = 0; i < IPREASS_NHASH; i++) { 920 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead); 921 if (r) { 922 IPSTAT_ADD(ips_fragtimeout, 923 r->ipq_nfrags); 924 ip_freef(&V_ipq[i], r); 925 break; 926 } 927 } 928 } else { 929 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags); 930 ip_freef(head, q); 931 } 932 } 933 934 found: 935 /* 936 * Adjust ip_len to not reflect header, 937 * convert offset of this to bytes. 938 */ 939 ip->ip_len -= hlen; 940 if (ip->ip_off & IP_MF) { 941 /* 942 * Make sure that fragments have a data length 943 * that's a non-zero multiple of 8 bytes. 944 */ 945 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 946 IPSTAT_INC(ips_toosmall); /* XXX */ 947 goto dropfrag; 948 } 949 m->m_flags |= M_FRAG; 950 } else 951 m->m_flags &= ~M_FRAG; 952 ip->ip_off <<= 3; 953 954 955 /* 956 * Attempt reassembly; if it succeeds, proceed. 957 * ip_reass() will return a different mbuf. 958 */ 959 IPSTAT_INC(ips_fragments); 960 m->m_pkthdr.header = ip; 961 962 /* Previous ip_reass() started here. */ 963 /* 964 * Presence of header sizes in mbufs 965 * would confuse code below. 966 */ 967 m->m_data += hlen; 968 m->m_len -= hlen; 969 970 /* 971 * If first fragment to arrive, create a reassembly queue. 972 */ 973 if (fp == NULL) { 974 fp = uma_zalloc(V_ipq_zone, M_NOWAIT); 975 if (fp == NULL) 976 goto dropfrag; 977 #ifdef MAC 978 if (mac_ipq_init(fp, M_NOWAIT) != 0) { 979 uma_zfree(V_ipq_zone, fp); 980 fp = NULL; 981 goto dropfrag; 982 } 983 mac_ipq_create(m, fp); 984 #endif 985 TAILQ_INSERT_HEAD(head, fp, ipq_list); 986 V_nipq++; 987 fp->ipq_nfrags = 1; 988 fp->ipq_ttl = IPFRAGTTL; 989 fp->ipq_p = ip->ip_p; 990 fp->ipq_id = ip->ip_id; 991 fp->ipq_src = ip->ip_src; 992 fp->ipq_dst = ip->ip_dst; 993 fp->ipq_frags = m; 994 m->m_nextpkt = NULL; 995 goto done; 996 } else { 997 fp->ipq_nfrags++; 998 #ifdef MAC 999 mac_ipq_update(m, fp); 1000 #endif 1001 } 1002 1003 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1004 1005 /* 1006 * Handle ECN by comparing this segment with the first one; 1007 * if CE is set, do not lose CE. 1008 * drop if CE and not-ECT are mixed for the same packet. 1009 */ 1010 ecn = ip->ip_tos & IPTOS_ECN_MASK; 1011 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 1012 if (ecn == IPTOS_ECN_CE) { 1013 if (ecn0 == IPTOS_ECN_NOTECT) 1014 goto dropfrag; 1015 if (ecn0 != IPTOS_ECN_CE) 1016 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 1017 } 1018 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 1019 goto dropfrag; 1020 1021 /* 1022 * Find a segment which begins after this one does. 1023 */ 1024 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1025 if (GETIP(q)->ip_off > ip->ip_off) 1026 break; 1027 1028 /* 1029 * If there is a preceding segment, it may provide some of 1030 * our data already. If so, drop the data from the incoming 1031 * segment. If it provides all of our data, drop us, otherwise 1032 * stick new segment in the proper place. 1033 * 1034 * If some of the data is dropped from the preceding 1035 * segment, then it's checksum is invalidated. 1036 */ 1037 if (p) { 1038 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1039 if (i > 0) { 1040 if (i >= ip->ip_len) 1041 goto dropfrag; 1042 m_adj(m, i); 1043 m->m_pkthdr.csum_flags = 0; 1044 ip->ip_off += i; 1045 ip->ip_len -= i; 1046 } 1047 m->m_nextpkt = p->m_nextpkt; 1048 p->m_nextpkt = m; 1049 } else { 1050 m->m_nextpkt = fp->ipq_frags; 1051 fp->ipq_frags = m; 1052 } 1053 1054 /* 1055 * While we overlap succeeding segments trim them or, 1056 * if they are completely covered, dequeue them. 1057 */ 1058 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1059 q = nq) { 1060 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1061 if (i < GETIP(q)->ip_len) { 1062 GETIP(q)->ip_len -= i; 1063 GETIP(q)->ip_off += i; 1064 m_adj(q, i); 1065 q->m_pkthdr.csum_flags = 0; 1066 break; 1067 } 1068 nq = q->m_nextpkt; 1069 m->m_nextpkt = nq; 1070 IPSTAT_INC(ips_fragdropped); 1071 fp->ipq_nfrags--; 1072 m_freem(q); 1073 } 1074 1075 /* 1076 * Check for complete reassembly and perform frag per packet 1077 * limiting. 1078 * 1079 * Frag limiting is performed here so that the nth frag has 1080 * a chance to complete the packet before we drop the packet. 1081 * As a result, n+1 frags are actually allowed per packet, but 1082 * only n will ever be stored. (n = maxfragsperpacket.) 1083 * 1084 */ 1085 next = 0; 1086 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1087 if (GETIP(q)->ip_off != next) { 1088 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1089 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1090 ip_freef(head, fp); 1091 } 1092 goto done; 1093 } 1094 next += GETIP(q)->ip_len; 1095 } 1096 /* Make sure the last packet didn't have the IP_MF flag */ 1097 if (p->m_flags & M_FRAG) { 1098 if (fp->ipq_nfrags > V_maxfragsperpacket) { 1099 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1100 ip_freef(head, fp); 1101 } 1102 goto done; 1103 } 1104 1105 /* 1106 * Reassembly is complete. Make sure the packet is a sane size. 1107 */ 1108 q = fp->ipq_frags; 1109 ip = GETIP(q); 1110 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1111 IPSTAT_INC(ips_toolong); 1112 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); 1113 ip_freef(head, fp); 1114 goto done; 1115 } 1116 1117 /* 1118 * Concatenate fragments. 1119 */ 1120 m = q; 1121 t = m->m_next; 1122 m->m_next = NULL; 1123 m_cat(m, t); 1124 nq = q->m_nextpkt; 1125 q->m_nextpkt = NULL; 1126 for (q = nq; q != NULL; q = nq) { 1127 nq = q->m_nextpkt; 1128 q->m_nextpkt = NULL; 1129 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1130 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1131 m_cat(m, q); 1132 } 1133 /* 1134 * In order to do checksumming faster we do 'end-around carry' here 1135 * (and not in for{} loop), though it implies we are not going to 1136 * reassemble more than 64k fragments. 1137 */ 1138 m->m_pkthdr.csum_data = 1139 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 1140 #ifdef MAC 1141 mac_ipq_reassemble(fp, m); 1142 mac_ipq_destroy(fp); 1143 #endif 1144 1145 /* 1146 * Create header for new ip packet by modifying header of first 1147 * packet; dequeue and discard fragment reassembly header. 1148 * Make header visible. 1149 */ 1150 ip->ip_len = (ip->ip_hl << 2) + next; 1151 ip->ip_src = fp->ipq_src; 1152 ip->ip_dst = fp->ipq_dst; 1153 TAILQ_REMOVE(head, fp, ipq_list); 1154 V_nipq--; 1155 uma_zfree(V_ipq_zone, fp); 1156 m->m_len += (ip->ip_hl << 2); 1157 m->m_data -= (ip->ip_hl << 2); 1158 /* some debugging cruft by sklower, below, will go away soon */ 1159 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1160 m_fixhdr(m); 1161 IPSTAT_INC(ips_reassembled); 1162 IPQ_UNLOCK(); 1163 return (m); 1164 1165 dropfrag: 1166 IPSTAT_INC(ips_fragdropped); 1167 if (fp != NULL) 1168 fp->ipq_nfrags--; 1169 m_freem(m); 1170 done: 1171 IPQ_UNLOCK(); 1172 return (NULL); 1173 1174 #undef GETIP 1175 } 1176 1177 /* 1178 * Free a fragment reassembly header and all 1179 * associated datagrams. 1180 */ 1181 static void 1182 ip_freef(struct ipqhead *fhp, struct ipq *fp) 1183 { 1184 struct mbuf *q; 1185 1186 IPQ_LOCK_ASSERT(); 1187 1188 while (fp->ipq_frags) { 1189 q = fp->ipq_frags; 1190 fp->ipq_frags = q->m_nextpkt; 1191 m_freem(q); 1192 } 1193 TAILQ_REMOVE(fhp, fp, ipq_list); 1194 uma_zfree(V_ipq_zone, fp); 1195 V_nipq--; 1196 } 1197 1198 /* 1199 * IP timer processing; 1200 * if a timer expires on a reassembly 1201 * queue, discard it. 1202 */ 1203 void 1204 ip_slowtimo(void) 1205 { 1206 VNET_ITERATOR_DECL(vnet_iter); 1207 struct ipq *fp; 1208 int i; 1209 1210 VNET_LIST_RLOCK_NOSLEEP(); 1211 IPQ_LOCK(); 1212 VNET_FOREACH(vnet_iter) { 1213 CURVNET_SET(vnet_iter); 1214 for (i = 0; i < IPREASS_NHASH; i++) { 1215 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) { 1216 struct ipq *fpp; 1217 1218 fpp = fp; 1219 fp = TAILQ_NEXT(fp, ipq_list); 1220 if(--fpp->ipq_ttl == 0) { 1221 IPSTAT_ADD(ips_fragtimeout, 1222 fpp->ipq_nfrags); 1223 ip_freef(&V_ipq[i], fpp); 1224 } 1225 } 1226 } 1227 /* 1228 * If we are over the maximum number of fragments 1229 * (due to the limit being lowered), drain off 1230 * enough to get down to the new limit. 1231 */ 1232 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) { 1233 for (i = 0; i < IPREASS_NHASH; i++) { 1234 while (V_nipq > V_maxnipq && 1235 !TAILQ_EMPTY(&V_ipq[i])) { 1236 IPSTAT_ADD(ips_fragdropped, 1237 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1238 ip_freef(&V_ipq[i], 1239 TAILQ_FIRST(&V_ipq[i])); 1240 } 1241 } 1242 } 1243 CURVNET_RESTORE(); 1244 } 1245 IPQ_UNLOCK(); 1246 VNET_LIST_RUNLOCK_NOSLEEP(); 1247 } 1248 1249 /* 1250 * Drain off all datagram fragments. 1251 */ 1252 static void 1253 ip_drain_locked(void) 1254 { 1255 int i; 1256 1257 IPQ_LOCK_ASSERT(); 1258 1259 for (i = 0; i < IPREASS_NHASH; i++) { 1260 while(!TAILQ_EMPTY(&V_ipq[i])) { 1261 IPSTAT_ADD(ips_fragdropped, 1262 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); 1263 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); 1264 } 1265 } 1266 } 1267 1268 void 1269 ip_drain(void) 1270 { 1271 VNET_ITERATOR_DECL(vnet_iter); 1272 1273 VNET_LIST_RLOCK_NOSLEEP(); 1274 IPQ_LOCK(); 1275 VNET_FOREACH(vnet_iter) { 1276 CURVNET_SET(vnet_iter); 1277 ip_drain_locked(); 1278 CURVNET_RESTORE(); 1279 } 1280 IPQ_UNLOCK(); 1281 VNET_LIST_RUNLOCK_NOSLEEP(); 1282 in_rtqdrain(); 1283 } 1284 1285 /* 1286 * The protocol to be inserted into ip_protox[] must be already registered 1287 * in inetsw[], either statically or through pf_proto_register(). 1288 */ 1289 int 1290 ipproto_register(short ipproto) 1291 { 1292 struct protosw *pr; 1293 1294 /* Sanity checks. */ 1295 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 1296 return (EPROTONOSUPPORT); 1297 1298 /* 1299 * The protocol slot must not be occupied by another protocol 1300 * already. An index pointing to IPPROTO_RAW is unused. 1301 */ 1302 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1303 if (pr == NULL) 1304 return (EPFNOSUPPORT); 1305 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1306 return (EEXIST); 1307 1308 /* Find the protocol position in inetsw[] and set the index. */ 1309 for (pr = inetdomain.dom_protosw; 1310 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1311 if (pr->pr_domain->dom_family == PF_INET && 1312 pr->pr_protocol && pr->pr_protocol == ipproto) { 1313 ip_protox[pr->pr_protocol] = pr - inetsw; 1314 return (0); 1315 } 1316 } 1317 return (EPROTONOSUPPORT); 1318 } 1319 1320 int 1321 ipproto_unregister(short ipproto) 1322 { 1323 struct protosw *pr; 1324 1325 /* Sanity checks. */ 1326 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 1327 return (EPROTONOSUPPORT); 1328 1329 /* Check if the protocol was indeed registered. */ 1330 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1331 if (pr == NULL) 1332 return (EPFNOSUPPORT); 1333 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1334 return (ENOENT); 1335 1336 /* Reset the protocol slot to IPPROTO_RAW. */ 1337 ip_protox[ipproto] = pr - inetsw; 1338 return (0); 1339 } 1340 1341 /* 1342 * Given address of next destination (final or next hop), return (referenced) 1343 * internet address info of interface to be used to get there. 1344 */ 1345 struct in_ifaddr * 1346 ip_rtaddr(struct in_addr dst, u_int fibnum) 1347 { 1348 struct route sro; 1349 struct sockaddr_in *sin; 1350 struct in_ifaddr *ia; 1351 1352 bzero(&sro, sizeof(sro)); 1353 sin = (struct sockaddr_in *)&sro.ro_dst; 1354 sin->sin_family = AF_INET; 1355 sin->sin_len = sizeof(*sin); 1356 sin->sin_addr = dst; 1357 in_rtalloc_ign(&sro, 0, fibnum); 1358 1359 if (sro.ro_rt == NULL) 1360 return (NULL); 1361 1362 ia = ifatoia(sro.ro_rt->rt_ifa); 1363 ifa_ref(&ia->ia_ifa); 1364 RTFREE(sro.ro_rt); 1365 return (ia); 1366 } 1367 1368 u_char inetctlerrmap[PRC_NCMDS] = { 1369 0, 0, 0, 0, 1370 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1371 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1372 EMSGSIZE, EHOSTUNREACH, 0, 0, 1373 0, 0, EHOSTUNREACH, 0, 1374 ENOPROTOOPT, ECONNREFUSED 1375 }; 1376 1377 /* 1378 * Forward a packet. If some error occurs return the sender 1379 * an icmp packet. Note we can't always generate a meaningful 1380 * icmp message because icmp doesn't have a large enough repertoire 1381 * of codes and types. 1382 * 1383 * If not forwarding, just drop the packet. This could be confusing 1384 * if ipforwarding was zero but some routing protocol was advancing 1385 * us as a gateway to somewhere. However, we must let the routing 1386 * protocol deal with that. 1387 * 1388 * The srcrt parameter indicates whether the packet is being forwarded 1389 * via a source route. 1390 */ 1391 void 1392 ip_forward(struct mbuf *m, int srcrt) 1393 { 1394 struct ip *ip = mtod(m, struct ip *); 1395 struct in_ifaddr *ia; 1396 struct mbuf *mcopy; 1397 struct in_addr dest; 1398 struct route ro; 1399 int error, type = 0, code = 0, mtu = 0; 1400 1401 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1402 IPSTAT_INC(ips_cantforward); 1403 m_freem(m); 1404 return; 1405 } 1406 #ifdef IPSTEALTH 1407 if (!V_ipstealth) { 1408 #endif 1409 if (ip->ip_ttl <= IPTTLDEC) { 1410 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1411 0, 0); 1412 return; 1413 } 1414 #ifdef IPSTEALTH 1415 } 1416 #endif 1417 1418 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); 1419 #ifndef IPSEC 1420 /* 1421 * 'ia' may be NULL if there is no route for this destination. 1422 * In case of IPsec, Don't discard it just yet, but pass it to 1423 * ip_output in case of outgoing IPsec policy. 1424 */ 1425 if (!srcrt && ia == NULL) { 1426 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1427 return; 1428 } 1429 #endif 1430 1431 /* 1432 * Save the IP header and at most 8 bytes of the payload, 1433 * in case we need to generate an ICMP message to the src. 1434 * 1435 * XXX this can be optimized a lot by saving the data in a local 1436 * buffer on the stack (72 bytes at most), and only allocating the 1437 * mbuf if really necessary. The vast majority of the packets 1438 * are forwarded without having to send an ICMP back (either 1439 * because unnecessary, or because rate limited), so we are 1440 * really we are wasting a lot of work here. 1441 * 1442 * We don't use m_copy() because it might return a reference 1443 * to a shared cluster. Both this function and ip_output() 1444 * assume exclusive access to the IP header in `m', so any 1445 * data in a cluster may change before we reach icmp_error(). 1446 */ 1447 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1448 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1449 /* 1450 * It's probably ok if the pkthdr dup fails (because 1451 * the deep copy of the tag chain failed), but for now 1452 * be conservative and just discard the copy since 1453 * code below may some day want the tags. 1454 */ 1455 m_free(mcopy); 1456 mcopy = NULL; 1457 } 1458 if (mcopy != NULL) { 1459 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy)); 1460 mcopy->m_pkthdr.len = mcopy->m_len; 1461 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1462 } 1463 1464 #ifdef IPSTEALTH 1465 if (!V_ipstealth) { 1466 #endif 1467 ip->ip_ttl -= IPTTLDEC; 1468 #ifdef IPSTEALTH 1469 } 1470 #endif 1471 1472 /* 1473 * If forwarding packet using same interface that it came in on, 1474 * perhaps should send a redirect to sender to shortcut a hop. 1475 * Only send redirect if source is sending directly to us, 1476 * and if packet was not source routed (or has any options). 1477 * Also, don't send redirect if forwarding using a default route 1478 * or a route modified by a redirect. 1479 */ 1480 dest.s_addr = 0; 1481 if (!srcrt && V_ipsendredirects && 1482 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) { 1483 struct sockaddr_in *sin; 1484 struct rtentry *rt; 1485 1486 bzero(&ro, sizeof(ro)); 1487 sin = (struct sockaddr_in *)&ro.ro_dst; 1488 sin->sin_family = AF_INET; 1489 sin->sin_len = sizeof(*sin); 1490 sin->sin_addr = ip->ip_dst; 1491 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 1492 1493 rt = ro.ro_rt; 1494 1495 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1496 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1497 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1498 u_long src = ntohl(ip->ip_src.s_addr); 1499 1500 if (RTA(rt) && 1501 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1502 if (rt->rt_flags & RTF_GATEWAY) 1503 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1504 else 1505 dest.s_addr = ip->ip_dst.s_addr; 1506 /* Router requirements says to only send host redirects */ 1507 type = ICMP_REDIRECT; 1508 code = ICMP_REDIRECT_HOST; 1509 } 1510 } 1511 if (rt) 1512 RTFREE(rt); 1513 } 1514 1515 /* 1516 * Try to cache the route MTU from ip_output so we can consider it for 1517 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. 1518 */ 1519 bzero(&ro, sizeof(ro)); 1520 1521 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1522 1523 if (error == EMSGSIZE && ro.ro_rt) 1524 mtu = ro.ro_rt->rt_rmx.rmx_mtu; 1525 if (ro.ro_rt) 1526 RTFREE(ro.ro_rt); 1527 1528 if (error) 1529 IPSTAT_INC(ips_cantforward); 1530 else { 1531 IPSTAT_INC(ips_forward); 1532 if (type) 1533 IPSTAT_INC(ips_redirectsent); 1534 else { 1535 if (mcopy) 1536 m_freem(mcopy); 1537 if (ia != NULL) 1538 ifa_free(&ia->ia_ifa); 1539 return; 1540 } 1541 } 1542 if (mcopy == NULL) { 1543 if (ia != NULL) 1544 ifa_free(&ia->ia_ifa); 1545 return; 1546 } 1547 1548 switch (error) { 1549 1550 case 0: /* forwarded, but need redirect */ 1551 /* type, code set above */ 1552 break; 1553 1554 case ENETUNREACH: 1555 case EHOSTUNREACH: 1556 case ENETDOWN: 1557 case EHOSTDOWN: 1558 default: 1559 type = ICMP_UNREACH; 1560 code = ICMP_UNREACH_HOST; 1561 break; 1562 1563 case EMSGSIZE: 1564 type = ICMP_UNREACH; 1565 code = ICMP_UNREACH_NEEDFRAG; 1566 1567 #ifdef IPSEC 1568 /* 1569 * If IPsec is configured for this path, 1570 * override any possibly mtu value set by ip_output. 1571 */ 1572 mtu = ip_ipsec_mtu(mcopy, mtu); 1573 #endif /* IPSEC */ 1574 /* 1575 * If the MTU was set before make sure we are below the 1576 * interface MTU. 1577 * If the MTU wasn't set before use the interface mtu or 1578 * fall back to the next smaller mtu step compared to the 1579 * current packet size. 1580 */ 1581 if (mtu != 0) { 1582 if (ia != NULL) 1583 mtu = min(mtu, ia->ia_ifp->if_mtu); 1584 } else { 1585 if (ia != NULL) 1586 mtu = ia->ia_ifp->if_mtu; 1587 else 1588 mtu = ip_next_mtu(ip->ip_len, 0); 1589 } 1590 IPSTAT_INC(ips_cantfrag); 1591 break; 1592 1593 case ENOBUFS: 1594 /* 1595 * A router should not generate ICMP_SOURCEQUENCH as 1596 * required in RFC1812 Requirements for IP Version 4 Routers. 1597 * Source quench could be a big problem under DoS attacks, 1598 * or if the underlying interface is rate-limited. 1599 * Those who need source quench packets may re-enable them 1600 * via the net.inet.ip.sendsourcequench sysctl. 1601 */ 1602 if (V_ip_sendsourcequench == 0) { 1603 m_freem(mcopy); 1604 if (ia != NULL) 1605 ifa_free(&ia->ia_ifa); 1606 return; 1607 } else { 1608 type = ICMP_SOURCEQUENCH; 1609 code = 0; 1610 } 1611 break; 1612 1613 case EACCES: /* ipfw denied packet */ 1614 m_freem(mcopy); 1615 if (ia != NULL) 1616 ifa_free(&ia->ia_ifa); 1617 return; 1618 } 1619 if (ia != NULL) 1620 ifa_free(&ia->ia_ifa); 1621 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1622 } 1623 1624 void 1625 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1626 struct mbuf *m) 1627 { 1628 1629 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1630 struct bintime bt; 1631 1632 bintime(&bt); 1633 if (inp->inp_socket->so_options & SO_BINTIME) { 1634 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1635 SCM_BINTIME, SOL_SOCKET); 1636 if (*mp) 1637 mp = &(*mp)->m_next; 1638 } 1639 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1640 struct timeval tv; 1641 1642 bintime2timeval(&bt, &tv); 1643 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1644 SCM_TIMESTAMP, SOL_SOCKET); 1645 if (*mp) 1646 mp = &(*mp)->m_next; 1647 } 1648 } 1649 if (inp->inp_flags & INP_RECVDSTADDR) { 1650 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1651 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1652 if (*mp) 1653 mp = &(*mp)->m_next; 1654 } 1655 if (inp->inp_flags & INP_RECVTTL) { 1656 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1657 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1658 if (*mp) 1659 mp = &(*mp)->m_next; 1660 } 1661 #ifdef notyet 1662 /* XXX 1663 * Moving these out of udp_input() made them even more broken 1664 * than they already were. 1665 */ 1666 /* options were tossed already */ 1667 if (inp->inp_flags & INP_RECVOPTS) { 1668 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1669 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1670 if (*mp) 1671 mp = &(*mp)->m_next; 1672 } 1673 /* ip_srcroute doesn't do what we want here, need to fix */ 1674 if (inp->inp_flags & INP_RECVRETOPTS) { 1675 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1676 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1677 if (*mp) 1678 mp = &(*mp)->m_next; 1679 } 1680 #endif 1681 if (inp->inp_flags & INP_RECVIF) { 1682 struct ifnet *ifp; 1683 struct sdlbuf { 1684 struct sockaddr_dl sdl; 1685 u_char pad[32]; 1686 } sdlbuf; 1687 struct sockaddr_dl *sdp; 1688 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1689 1690 if (((ifp = m->m_pkthdr.rcvif)) 1691 && ( ifp->if_index && (ifp->if_index <= V_if_index))) { 1692 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1693 /* 1694 * Change our mind and don't try copy. 1695 */ 1696 if ((sdp->sdl_family != AF_LINK) 1697 || (sdp->sdl_len > sizeof(sdlbuf))) { 1698 goto makedummy; 1699 } 1700 bcopy(sdp, sdl2, sdp->sdl_len); 1701 } else { 1702 makedummy: 1703 sdl2->sdl_len 1704 = offsetof(struct sockaddr_dl, sdl_data[0]); 1705 sdl2->sdl_family = AF_LINK; 1706 sdl2->sdl_index = 0; 1707 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1708 } 1709 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1710 IP_RECVIF, IPPROTO_IP); 1711 if (*mp) 1712 mp = &(*mp)->m_next; 1713 } 1714 } 1715 1716 /* 1717 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1718 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1719 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1720 * compiled. 1721 */ 1722 static VNET_DEFINE(int, ip_rsvp_on); 1723 VNET_DEFINE(struct socket *, ip_rsvpd); 1724 1725 #define V_ip_rsvp_on VNET(ip_rsvp_on) 1726 1727 int 1728 ip_rsvp_init(struct socket *so) 1729 { 1730 1731 if (so->so_type != SOCK_RAW || 1732 so->so_proto->pr_protocol != IPPROTO_RSVP) 1733 return EOPNOTSUPP; 1734 1735 if (V_ip_rsvpd != NULL) 1736 return EADDRINUSE; 1737 1738 V_ip_rsvpd = so; 1739 /* 1740 * This may seem silly, but we need to be sure we don't over-increment 1741 * the RSVP counter, in case something slips up. 1742 */ 1743 if (!V_ip_rsvp_on) { 1744 V_ip_rsvp_on = 1; 1745 V_rsvp_on++; 1746 } 1747 1748 return 0; 1749 } 1750 1751 int 1752 ip_rsvp_done(void) 1753 { 1754 1755 V_ip_rsvpd = NULL; 1756 /* 1757 * This may seem silly, but we need to be sure we don't over-decrement 1758 * the RSVP counter, in case something slips up. 1759 */ 1760 if (V_ip_rsvp_on) { 1761 V_ip_rsvp_on = 0; 1762 V_rsvp_on--; 1763 } 1764 return 0; 1765 } 1766 1767 void 1768 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1769 { 1770 1771 if (rsvp_input_p) { /* call the real one if loaded */ 1772 rsvp_input_p(m, off); 1773 return; 1774 } 1775 1776 /* Can still get packets with rsvp_on = 0 if there is a local member 1777 * of the group to which the RSVP packet is addressed. But in this 1778 * case we want to throw the packet away. 1779 */ 1780 1781 if (!V_rsvp_on) { 1782 m_freem(m); 1783 return; 1784 } 1785 1786 if (V_ip_rsvpd != NULL) { 1787 rip_input(m, off); 1788 return; 1789 } 1790 /* Drop the packet */ 1791 m_freem(m); 1792 } 1793