ip_input.c revision 163545
1/*- 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 30 * $FreeBSD: head/sys/netinet/ip_input.c 163545 2006-10-20 19:32:08Z julian $ 31 */ 32 33#include "opt_bootp.h" 34#include "opt_ipfw.h" 35#include "opt_ipstealth.h" 36#include "opt_ipsec.h" 37#include "opt_mac.h" 38#include "opt_carp.h" 39 40#include <sys/param.h> 41#include <sys/systm.h> 42#include <sys/callout.h> 43#include <sys/mac.h> 44#include <sys/mbuf.h> 45#include <sys/malloc.h> 46#include <sys/domain.h> 47#include <sys/protosw.h> 48#include <sys/socket.h> 49#include <sys/time.h> 50#include <sys/kernel.h> 51#include <sys/syslog.h> 52#include <sys/sysctl.h> 53 54#include <net/pfil.h> 55#include <net/if.h> 56#include <net/if_types.h> 57#include <net/if_var.h> 58#include <net/if_dl.h> 59#include <net/route.h> 60#include <net/netisr.h> 61 62#include <netinet/in.h> 63#include <netinet/in_systm.h> 64#include <netinet/in_var.h> 65#include <netinet/ip.h> 66#include <netinet/in_pcb.h> 67#include <netinet/ip_var.h> 68#include <netinet/ip_icmp.h> 69#include <netinet/ip_options.h> 70#include <machine/in_cksum.h> 71#ifdef DEV_CARP 72#include <netinet/ip_carp.h> 73#endif 74#if defined(IPSEC) || defined(FAST_IPSEC) 75#include <netinet/ip_ipsec.h> 76#endif /* IPSEC */ 77 78#include <sys/socketvar.h> 79 80int rsvp_on = 0; 81 82int ipforwarding = 0; 83SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 84 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 85 86static int ipsendredirects = 1; /* XXX */ 87SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 88 &ipsendredirects, 0, "Enable sending IP redirects"); 89 90int ip_defttl = IPDEFTTL; 91SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 92 &ip_defttl, 0, "Maximum TTL on IP packets"); 93 94static int ip_keepfaith = 0; 95SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 96 &ip_keepfaith, 0, 97 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 98 99static int ip_sendsourcequench = 0; 100SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 101 &ip_sendsourcequench, 0, 102 "Enable the transmission of source quench packets"); 103 104int ip_do_randomid = 0; 105SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 106 &ip_do_randomid, 0, 107 "Assign random ip_id values"); 108 109/* 110 * XXX - Setting ip_checkinterface mostly implements the receive side of 111 * the Strong ES model described in RFC 1122, but since the routing table 112 * and transmit implementation do not implement the Strong ES model, 113 * setting this to 1 results in an odd hybrid. 114 * 115 * XXX - ip_checkinterface currently must be disabled if you use ipnat 116 * to translate the destination address to another local interface. 117 * 118 * XXX - ip_checkinterface must be disabled if you add IP aliases 119 * to the loopback interface instead of the interface where the 120 * packets for those addresses are received. 121 */ 122static int ip_checkinterface = 0; 123SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 124 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 125 126struct pfil_head inet_pfil_hook; /* Packet filter hooks */ 127 128static struct ifqueue ipintrq; 129static int ipqmaxlen = IFQ_MAXLEN; 130 131extern struct domain inetdomain; 132extern struct protosw inetsw[]; 133u_char ip_protox[IPPROTO_MAX]; 134struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 135struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 136u_long in_ifaddrhmask; /* mask for hash table */ 137 138SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 139 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 140SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 141 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 142 143struct ipstat ipstat; 144SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 145 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 146 147/* 148 * IP datagram reassembly. 149 */ 150#define IPREASS_NHASH_LOG2 6 151#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 152#define IPREASS_HMASK (IPREASS_NHASH - 1) 153#define IPREASS_HASH(x,y) \ 154 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 155 156static uma_zone_t ipq_zone; 157static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 158static struct mtx ipqlock; 159 160#define IPQ_LOCK() mtx_lock(&ipqlock) 161#define IPQ_UNLOCK() mtx_unlock(&ipqlock) 162#define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 163#define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 164 165static void maxnipq_update(void); 166static void ipq_zone_change(void *); 167 168static int maxnipq; /* Administrative limit on # reass queues. */ 169static int nipq = 0; /* Total # of reass queues */ 170SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, &nipq, 0, 171 "Current number of IPv4 fragment reassembly queue entries"); 172 173static int maxfragsperpacket; 174SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 175 &maxfragsperpacket, 0, 176 "Maximum number of IPv4 fragments allowed per packet"); 177 178struct callout ipport_tick_callout; 179 180#ifdef IPCTL_DEFMTU 181SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 182 &ip_mtu, 0, "Default MTU"); 183#endif 184 185#ifdef IPSTEALTH 186int ipstealth = 0; 187SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 188 &ipstealth, 0, ""); 189#endif 190 191static void ip_freef(struct ipqhead *, struct ipq *); 192 193/* 194 * IP initialization: fill in IP protocol switch table. 195 * All protocols not implemented in kernel go to raw IP protocol handler. 196 */ 197void 198ip_init() 199{ 200 register struct protosw *pr; 201 register int i; 202 203 TAILQ_INIT(&in_ifaddrhead); 204 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 205 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 206 if (pr == NULL) 207 panic("ip_init: PF_INET not found"); 208 209 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 210 for (i = 0; i < IPPROTO_MAX; i++) 211 ip_protox[i] = pr - inetsw; 212 /* 213 * Cycle through IP protocols and put them into the appropriate place 214 * in ip_protox[]. 215 */ 216 for (pr = inetdomain.dom_protosw; 217 pr < inetdomain.dom_protoswNPROTOSW; pr++) 218 if (pr->pr_domain->dom_family == PF_INET && 219 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 220 /* Be careful to only index valid IP protocols. */ 221 if (pr->pr_protocol < IPPROTO_MAX) 222 ip_protox[pr->pr_protocol] = pr - inetsw; 223 } 224 225 /* Initialize packet filter hooks. */ 226 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 227 inet_pfil_hook.ph_af = AF_INET; 228 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 229 printf("%s: WARNING: unable to register pfil hook, " 230 "error %d\n", __func__, i); 231 232 /* Initialize IP reassembly queue. */ 233 IPQ_LOCK_INIT(); 234 for (i = 0; i < IPREASS_NHASH; i++) 235 TAILQ_INIT(&ipq[i]); 236 maxnipq = nmbclusters / 32; 237 maxfragsperpacket = 16; 238 ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, 239 NULL, UMA_ALIGN_PTR, 0); 240 maxnipq_update(); 241 242 /* Start ipport_tick. */ 243 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE); 244 ipport_tick(NULL); 245 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, 246 SHUTDOWN_PRI_DEFAULT); 247 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, 248 NULL, EVENTHANDLER_PRI_ANY); 249 250 /* Initialize various other remaining things. */ 251 ip_id = time_second & 0xffff; 252 ipintrq.ifq_maxlen = ipqmaxlen; 253 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF); 254 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE); 255} 256 257void ip_fini(xtp) 258 void *xtp; 259{ 260 callout_stop(&ipport_tick_callout); 261} 262 263/* 264 * Ip input routine. Checksum and byte swap header. If fragmented 265 * try to reassemble. Process options. Pass to next level. 266 */ 267void 268ip_input(struct mbuf *m) 269{ 270 struct ip *ip = NULL; 271 struct in_ifaddr *ia = NULL; 272 struct ifaddr *ifa; 273 int checkif, hlen = 0; 274 u_short sum; 275 int dchg = 0; /* dest changed after fw */ 276 struct in_addr odst; /* original dst address */ 277 278 M_ASSERTPKTHDR(m); 279 280 if (m->m_flags & M_FASTFWD_OURS) { 281 /* 282 * Firewall or NAT changed destination to local. 283 * We expect ip_len and ip_off to be in host byte order. 284 */ 285 m->m_flags &= ~M_FASTFWD_OURS; 286 /* Set up some basics that will be used later. */ 287 ip = mtod(m, struct ip *); 288 hlen = ip->ip_hl << 2; 289 goto ours; 290 } 291 292 ipstat.ips_total++; 293 294 if (m->m_pkthdr.len < sizeof(struct ip)) 295 goto tooshort; 296 297 if (m->m_len < sizeof (struct ip) && 298 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 299 ipstat.ips_toosmall++; 300 return; 301 } 302 ip = mtod(m, struct ip *); 303 304 if (ip->ip_v != IPVERSION) { 305 ipstat.ips_badvers++; 306 goto bad; 307 } 308 309 hlen = ip->ip_hl << 2; 310 if (hlen < sizeof(struct ip)) { /* minimum header length */ 311 ipstat.ips_badhlen++; 312 goto bad; 313 } 314 if (hlen > m->m_len) { 315 if ((m = m_pullup(m, hlen)) == NULL) { 316 ipstat.ips_badhlen++; 317 return; 318 } 319 ip = mtod(m, struct ip *); 320 } 321 322 /* 127/8 must not appear on wire - RFC1122 */ 323 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 324 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 325 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 326 ipstat.ips_badaddr++; 327 goto bad; 328 } 329 } 330 331 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 332 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 333 } else { 334 if (hlen == sizeof(struct ip)) { 335 sum = in_cksum_hdr(ip); 336 } else { 337 sum = in_cksum(m, hlen); 338 } 339 } 340 if (sum) { 341 ipstat.ips_badsum++; 342 goto bad; 343 } 344 345#ifdef ALTQ 346 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 347 /* packet is dropped by traffic conditioner */ 348 return; 349#endif 350 351 /* 352 * Convert fields to host representation. 353 */ 354 ip->ip_len = ntohs(ip->ip_len); 355 if (ip->ip_len < hlen) { 356 ipstat.ips_badlen++; 357 goto bad; 358 } 359 ip->ip_off = ntohs(ip->ip_off); 360 361 /* 362 * Check that the amount of data in the buffers 363 * is as at least much as the IP header would have us expect. 364 * Trim mbufs if longer than we expect. 365 * Drop packet if shorter than we expect. 366 */ 367 if (m->m_pkthdr.len < ip->ip_len) { 368tooshort: 369 ipstat.ips_tooshort++; 370 goto bad; 371 } 372 if (m->m_pkthdr.len > ip->ip_len) { 373 if (m->m_len == m->m_pkthdr.len) { 374 m->m_len = ip->ip_len; 375 m->m_pkthdr.len = ip->ip_len; 376 } else 377 m_adj(m, ip->ip_len - m->m_pkthdr.len); 378 } 379#if defined(IPSEC) || defined(FAST_IPSEC) 380 /* 381 * Bypass packet filtering for packets from a tunnel (gif). 382 */ 383 if (ip_ipsec_filtergif(m)) 384 goto passin; 385#endif /* IPSEC */ 386 387 /* 388 * Run through list of hooks for input packets. 389 * 390 * NB: Beware of the destination address changing (e.g. 391 * by NAT rewriting). When this happens, tell 392 * ip_forward to do the right thing. 393 */ 394 395 /* Jump over all PFIL processing if hooks are not active. */ 396 if (!PFIL_HOOKED(&inet_pfil_hook)) 397 goto passin; 398 399 odst = ip->ip_dst; 400 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 401 PFIL_IN, NULL) != 0) 402 return; 403 if (m == NULL) /* consumed by filter */ 404 return; 405 406 ip = mtod(m, struct ip *); 407 dchg = (odst.s_addr != ip->ip_dst.s_addr); 408 409#ifdef IPFIREWALL_FORWARD 410 if (m->m_flags & M_FASTFWD_OURS) { 411 m->m_flags &= ~M_FASTFWD_OURS; 412 goto ours; 413 } 414 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) { 415 /* 416 * Directly ship on the packet. This allows to forward packets 417 * that were destined for us to some other directly connected 418 * host. 419 */ 420 ip_forward(m, dchg); 421 return; 422 } 423#endif /* IPFIREWALL_FORWARD */ 424 425passin: 426 /* 427 * Process options and, if not destined for us, 428 * ship it on. ip_dooptions returns 1 when an 429 * error was detected (causing an icmp message 430 * to be sent and the original packet to be freed). 431 */ 432 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 433 return; 434 435 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 436 * matter if it is destined to another node, or whether it is 437 * a multicast one, RSVP wants it! and prevents it from being forwarded 438 * anywhere else. Also checks if the rsvp daemon is running before 439 * grabbing the packet. 440 */ 441 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 442 goto ours; 443 444 /* 445 * Check our list of addresses, to see if the packet is for us. 446 * If we don't have any addresses, assume any unicast packet 447 * we receive might be for us (and let the upper layers deal 448 * with it). 449 */ 450 if (TAILQ_EMPTY(&in_ifaddrhead) && 451 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 452 goto ours; 453 454 /* 455 * Enable a consistency check between the destination address 456 * and the arrival interface for a unicast packet (the RFC 1122 457 * strong ES model) if IP forwarding is disabled and the packet 458 * is not locally generated and the packet is not subject to 459 * 'ipfw fwd'. 460 * 461 * XXX - Checking also should be disabled if the destination 462 * address is ipnat'ed to a different interface. 463 * 464 * XXX - Checking is incompatible with IP aliases added 465 * to the loopback interface instead of the interface where 466 * the packets are received. 467 * 468 * XXX - This is the case for carp vhost IPs as well so we 469 * insert a workaround. If the packet got here, we already 470 * checked with carp_iamatch() and carp_forus(). 471 */ 472 checkif = ip_checkinterface && (ipforwarding == 0) && 473 m->m_pkthdr.rcvif != NULL && 474 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 475#ifdef DEV_CARP 476 !m->m_pkthdr.rcvif->if_carp && 477#endif 478 (dchg == 0); 479 480 /* 481 * Check for exact addresses in the hash bucket. 482 */ 483 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 484 /* 485 * If the address matches, verify that the packet 486 * arrived via the correct interface if checking is 487 * enabled. 488 */ 489 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 490 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 491 goto ours; 492 } 493 /* 494 * Check for broadcast addresses. 495 * 496 * Only accept broadcast packets that arrive via the matching 497 * interface. Reception of forwarded directed broadcasts would 498 * be handled via ip_forward() and ether_output() with the loopback 499 * into the stack for SIMPLEX interfaces handled by ether_output(). 500 */ 501 if (m->m_pkthdr.rcvif != NULL && 502 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 503 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 504 if (ifa->ifa_addr->sa_family != AF_INET) 505 continue; 506 ia = ifatoia(ifa); 507 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 508 ip->ip_dst.s_addr) 509 goto ours; 510 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) 511 goto ours; 512#ifdef BOOTP_COMPAT 513 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 514 goto ours; 515#endif 516 } 517 } 518 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 519 struct in_multi *inm; 520 if (ip_mrouter) { 521 /* 522 * If we are acting as a multicast router, all 523 * incoming multicast packets are passed to the 524 * kernel-level multicast forwarding function. 525 * The packet is returned (relatively) intact; if 526 * ip_mforward() returns a non-zero value, the packet 527 * must be discarded, else it may be accepted below. 528 */ 529 if (ip_mforward && 530 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 531 ipstat.ips_cantforward++; 532 m_freem(m); 533 return; 534 } 535 536 /* 537 * The process-level routing daemon needs to receive 538 * all multicast IGMP packets, whether or not this 539 * host belongs to their destination groups. 540 */ 541 if (ip->ip_p == IPPROTO_IGMP) 542 goto ours; 543 ipstat.ips_forward++; 544 } 545 /* 546 * See if we belong to the destination multicast group on the 547 * arrival interface. 548 */ 549 IN_MULTI_LOCK(); 550 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 551 IN_MULTI_UNLOCK(); 552 if (inm == NULL) { 553 ipstat.ips_notmember++; 554 m_freem(m); 555 return; 556 } 557 goto ours; 558 } 559 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 560 goto ours; 561 if (ip->ip_dst.s_addr == INADDR_ANY) 562 goto ours; 563 564 /* 565 * FAITH(Firewall Aided Internet Translator) 566 */ 567 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 568 if (ip_keepfaith) { 569 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 570 goto ours; 571 } 572 m_freem(m); 573 return; 574 } 575 576 /* 577 * Not for us; forward if possible and desirable. 578 */ 579 if (ipforwarding == 0) { 580 ipstat.ips_cantforward++; 581 m_freem(m); 582 } else { 583#if defined(IPSEC) || defined(FAST_IPSEC) 584 if (ip_ipsec_fwd(m)) 585 goto bad; 586#endif /* IPSEC */ 587 ip_forward(m, dchg); 588 } 589 return; 590 591ours: 592#ifdef IPSTEALTH 593 /* 594 * IPSTEALTH: Process non-routing options only 595 * if the packet is destined for us. 596 */ 597 if (ipstealth && hlen > sizeof (struct ip) && 598 ip_dooptions(m, 1)) 599 return; 600#endif /* IPSTEALTH */ 601 602 /* Count the packet in the ip address stats */ 603 if (ia != NULL) { 604 ia->ia_ifa.if_ipackets++; 605 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 606 } 607 608 /* 609 * Attempt reassembly; if it succeeds, proceed. 610 * ip_reass() will return a different mbuf. 611 */ 612 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 613 m = ip_reass(m); 614 if (m == NULL) 615 return; 616 ip = mtod(m, struct ip *); 617 /* Get the header length of the reassembled packet */ 618 hlen = ip->ip_hl << 2; 619 } 620 621 /* 622 * Further protocols expect the packet length to be w/o the 623 * IP header. 624 */ 625 ip->ip_len -= hlen; 626 627#if defined(IPSEC) || defined(FAST_IPSEC) 628 /* 629 * enforce IPsec policy checking if we are seeing last header. 630 * note that we do not visit this with protocols with pcb layer 631 * code - like udp/tcp/raw ip. 632 */ 633 if (ip_ipsec_input(m)) 634 goto bad; 635#endif /* IPSEC */ 636 637 /* 638 * Switch out to protocol's input routine. 639 */ 640 ipstat.ips_delivered++; 641 642 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 643 return; 644bad: 645 m_freem(m); 646} 647 648/* 649 * After maxnipq has been updated, propagate the change to UMA. The UMA zone 650 * max has slightly different semantics than the sysctl, for historical 651 * reasons. 652 */ 653static void 654maxnipq_update(void) 655{ 656 657 /* 658 * -1 for unlimited allocation. 659 */ 660 if (maxnipq < 0) 661 uma_zone_set_max(ipq_zone, 0); 662 /* 663 * Positive number for specific bound. 664 */ 665 if (maxnipq > 0) 666 uma_zone_set_max(ipq_zone, maxnipq); 667 /* 668 * Zero specifies no further fragment queue allocation -- set the 669 * bound very low, but rely on implementation elsewhere to actually 670 * prevent allocation and reclaim current queues. 671 */ 672 if (maxnipq == 0) 673 uma_zone_set_max(ipq_zone, 1); 674} 675 676static void 677ipq_zone_change(void *tag) 678{ 679 680 if (maxnipq > 0 && maxnipq < (nmbclusters / 32)) { 681 maxnipq = nmbclusters / 32; 682 maxnipq_update(); 683 } 684} 685 686static int 687sysctl_maxnipq(SYSCTL_HANDLER_ARGS) 688{ 689 int error, i; 690 691 i = maxnipq; 692 error = sysctl_handle_int(oidp, &i, 0, req); 693 if (error || !req->newptr) 694 return (error); 695 696 /* 697 * XXXRW: Might be a good idea to sanity check the argument and place 698 * an extreme upper bound. 699 */ 700 if (i < -1) 701 return (EINVAL); 702 maxnipq = i; 703 maxnipq_update(); 704 return (0); 705} 706 707SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, 708 NULL, 0, sysctl_maxnipq, "I", 709 "Maximum number of IPv4 fragment reassembly queue entries"); 710 711/* 712 * Take incoming datagram fragment and try to reassemble it into 713 * whole datagram. If the argument is the first fragment or one 714 * in between the function will return NULL and store the mbuf 715 * in the fragment chain. If the argument is the last fragment 716 * the packet will be reassembled and the pointer to the new 717 * mbuf returned for further processing. Only m_tags attached 718 * to the first packet/fragment are preserved. 719 * The IP header is *NOT* adjusted out of iplen. 720 */ 721 722struct mbuf * 723ip_reass(struct mbuf *m) 724{ 725 struct ip *ip; 726 struct mbuf *p, *q, *nq, *t; 727 struct ipq *fp = NULL; 728 struct ipqhead *head; 729 int i, hlen, next; 730 u_int8_t ecn, ecn0; 731 u_short hash; 732 733 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 734 if (maxnipq == 0 || maxfragsperpacket == 0) { 735 ipstat.ips_fragments++; 736 ipstat.ips_fragdropped++; 737 m_freem(m); 738 return (NULL); 739 } 740 741 ip = mtod(m, struct ip *); 742 hlen = ip->ip_hl << 2; 743 744 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 745 head = &ipq[hash]; 746 IPQ_LOCK(); 747 748 /* 749 * Look for queue of fragments 750 * of this datagram. 751 */ 752 TAILQ_FOREACH(fp, head, ipq_list) 753 if (ip->ip_id == fp->ipq_id && 754 ip->ip_src.s_addr == fp->ipq_src.s_addr && 755 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 756#ifdef MAC 757 mac_fragment_match(m, fp) && 758#endif 759 ip->ip_p == fp->ipq_p) 760 goto found; 761 762 fp = NULL; 763 764 /* 765 * Attempt to trim the number of allocated fragment queues if it 766 * exceeds the administrative limit. 767 */ 768 if ((nipq > maxnipq) && (maxnipq > 0)) { 769 /* 770 * drop something from the tail of the current queue 771 * before proceeding further 772 */ 773 struct ipq *q = TAILQ_LAST(head, ipqhead); 774 if (q == NULL) { /* gak */ 775 for (i = 0; i < IPREASS_NHASH; i++) { 776 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 777 if (r) { 778 ipstat.ips_fragtimeout += r->ipq_nfrags; 779 ip_freef(&ipq[i], r); 780 break; 781 } 782 } 783 } else { 784 ipstat.ips_fragtimeout += q->ipq_nfrags; 785 ip_freef(head, q); 786 } 787 } 788 789found: 790 /* 791 * Adjust ip_len to not reflect header, 792 * convert offset of this to bytes. 793 */ 794 ip->ip_len -= hlen; 795 if (ip->ip_off & IP_MF) { 796 /* 797 * Make sure that fragments have a data length 798 * that's a non-zero multiple of 8 bytes. 799 */ 800 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 801 ipstat.ips_toosmall++; /* XXX */ 802 goto dropfrag; 803 } 804 m->m_flags |= M_FRAG; 805 } else 806 m->m_flags &= ~M_FRAG; 807 ip->ip_off <<= 3; 808 809 810 /* 811 * Attempt reassembly; if it succeeds, proceed. 812 * ip_reass() will return a different mbuf. 813 */ 814 ipstat.ips_fragments++; 815 m->m_pkthdr.header = ip; 816 817 /* Previous ip_reass() started here. */ 818 /* 819 * Presence of header sizes in mbufs 820 * would confuse code below. 821 */ 822 m->m_data += hlen; 823 m->m_len -= hlen; 824 825 /* 826 * If first fragment to arrive, create a reassembly queue. 827 */ 828 if (fp == NULL) { 829 fp = uma_zalloc(ipq_zone, M_NOWAIT); 830 if (fp == NULL) 831 goto dropfrag; 832#ifdef MAC 833 if (mac_init_ipq(fp, M_NOWAIT) != 0) { 834 uma_zfree(ipq_zone, fp); 835 fp = NULL; 836 goto dropfrag; 837 } 838 mac_create_ipq(m, fp); 839#endif 840 TAILQ_INSERT_HEAD(head, fp, ipq_list); 841 nipq++; 842 fp->ipq_nfrags = 1; 843 fp->ipq_ttl = IPFRAGTTL; 844 fp->ipq_p = ip->ip_p; 845 fp->ipq_id = ip->ip_id; 846 fp->ipq_src = ip->ip_src; 847 fp->ipq_dst = ip->ip_dst; 848 fp->ipq_frags = m; 849 m->m_nextpkt = NULL; 850 goto done; 851 } else { 852 fp->ipq_nfrags++; 853#ifdef MAC 854 mac_update_ipq(m, fp); 855#endif 856 } 857 858#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 859 860 /* 861 * Handle ECN by comparing this segment with the first one; 862 * if CE is set, do not lose CE. 863 * drop if CE and not-ECT are mixed for the same packet. 864 */ 865 ecn = ip->ip_tos & IPTOS_ECN_MASK; 866 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 867 if (ecn == IPTOS_ECN_CE) { 868 if (ecn0 == IPTOS_ECN_NOTECT) 869 goto dropfrag; 870 if (ecn0 != IPTOS_ECN_CE) 871 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 872 } 873 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 874 goto dropfrag; 875 876 /* 877 * Find a segment which begins after this one does. 878 */ 879 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 880 if (GETIP(q)->ip_off > ip->ip_off) 881 break; 882 883 /* 884 * If there is a preceding segment, it may provide some of 885 * our data already. If so, drop the data from the incoming 886 * segment. If it provides all of our data, drop us, otherwise 887 * stick new segment in the proper place. 888 * 889 * If some of the data is dropped from the the preceding 890 * segment, then it's checksum is invalidated. 891 */ 892 if (p) { 893 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 894 if (i > 0) { 895 if (i >= ip->ip_len) 896 goto dropfrag; 897 m_adj(m, i); 898 m->m_pkthdr.csum_flags = 0; 899 ip->ip_off += i; 900 ip->ip_len -= i; 901 } 902 m->m_nextpkt = p->m_nextpkt; 903 p->m_nextpkt = m; 904 } else { 905 m->m_nextpkt = fp->ipq_frags; 906 fp->ipq_frags = m; 907 } 908 909 /* 910 * While we overlap succeeding segments trim them or, 911 * if they are completely covered, dequeue them. 912 */ 913 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 914 q = nq) { 915 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 916 if (i < GETIP(q)->ip_len) { 917 GETIP(q)->ip_len -= i; 918 GETIP(q)->ip_off += i; 919 m_adj(q, i); 920 q->m_pkthdr.csum_flags = 0; 921 break; 922 } 923 nq = q->m_nextpkt; 924 m->m_nextpkt = nq; 925 ipstat.ips_fragdropped++; 926 fp->ipq_nfrags--; 927 m_freem(q); 928 } 929 930 /* 931 * Check for complete reassembly and perform frag per packet 932 * limiting. 933 * 934 * Frag limiting is performed here so that the nth frag has 935 * a chance to complete the packet before we drop the packet. 936 * As a result, n+1 frags are actually allowed per packet, but 937 * only n will ever be stored. (n = maxfragsperpacket.) 938 * 939 */ 940 next = 0; 941 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 942 if (GETIP(q)->ip_off != next) { 943 if (fp->ipq_nfrags > maxfragsperpacket) { 944 ipstat.ips_fragdropped += fp->ipq_nfrags; 945 ip_freef(head, fp); 946 } 947 goto done; 948 } 949 next += GETIP(q)->ip_len; 950 } 951 /* Make sure the last packet didn't have the IP_MF flag */ 952 if (p->m_flags & M_FRAG) { 953 if (fp->ipq_nfrags > maxfragsperpacket) { 954 ipstat.ips_fragdropped += fp->ipq_nfrags; 955 ip_freef(head, fp); 956 } 957 goto done; 958 } 959 960 /* 961 * Reassembly is complete. Make sure the packet is a sane size. 962 */ 963 q = fp->ipq_frags; 964 ip = GETIP(q); 965 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 966 ipstat.ips_toolong++; 967 ipstat.ips_fragdropped += fp->ipq_nfrags; 968 ip_freef(head, fp); 969 goto done; 970 } 971 972 /* 973 * Concatenate fragments. 974 */ 975 m = q; 976 t = m->m_next; 977 m->m_next = NULL; 978 m_cat(m, t); 979 nq = q->m_nextpkt; 980 q->m_nextpkt = NULL; 981 for (q = nq; q != NULL; q = nq) { 982 nq = q->m_nextpkt; 983 q->m_nextpkt = NULL; 984 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 985 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 986 m_cat(m, q); 987 } 988 /* 989 * In order to do checksumming faster we do 'end-around carry' here 990 * (and not in for{} loop), though it implies we are not going to 991 * reassemble more than 64k fragments. 992 */ 993 m->m_pkthdr.csum_data = 994 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); 995#ifdef MAC 996 mac_create_datagram_from_ipq(fp, m); 997 mac_destroy_ipq(fp); 998#endif 999 1000 /* 1001 * Create header for new ip packet by modifying header of first 1002 * packet; dequeue and discard fragment reassembly header. 1003 * Make header visible. 1004 */ 1005 ip->ip_len = (ip->ip_hl << 2) + next; 1006 ip->ip_src = fp->ipq_src; 1007 ip->ip_dst = fp->ipq_dst; 1008 TAILQ_REMOVE(head, fp, ipq_list); 1009 nipq--; 1010 uma_zfree(ipq_zone, fp); 1011 m->m_len += (ip->ip_hl << 2); 1012 m->m_data -= (ip->ip_hl << 2); 1013 /* some debugging cruft by sklower, below, will go away soon */ 1014 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1015 m_fixhdr(m); 1016 ipstat.ips_reassembled++; 1017 IPQ_UNLOCK(); 1018 return (m); 1019 1020dropfrag: 1021 ipstat.ips_fragdropped++; 1022 if (fp != NULL) 1023 fp->ipq_nfrags--; 1024 m_freem(m); 1025done: 1026 IPQ_UNLOCK(); 1027 return (NULL); 1028 1029#undef GETIP 1030} 1031 1032/* 1033 * Free a fragment reassembly header and all 1034 * associated datagrams. 1035 */ 1036static void 1037ip_freef(fhp, fp) 1038 struct ipqhead *fhp; 1039 struct ipq *fp; 1040{ 1041 register struct mbuf *q; 1042 1043 IPQ_LOCK_ASSERT(); 1044 1045 while (fp->ipq_frags) { 1046 q = fp->ipq_frags; 1047 fp->ipq_frags = q->m_nextpkt; 1048 m_freem(q); 1049 } 1050 TAILQ_REMOVE(fhp, fp, ipq_list); 1051 uma_zfree(ipq_zone, fp); 1052 nipq--; 1053} 1054 1055/* 1056 * IP timer processing; 1057 * if a timer expires on a reassembly 1058 * queue, discard it. 1059 */ 1060void 1061ip_slowtimo() 1062{ 1063 register struct ipq *fp; 1064 int i; 1065 1066 IPQ_LOCK(); 1067 for (i = 0; i < IPREASS_NHASH; i++) { 1068 for(fp = TAILQ_FIRST(&ipq[i]); fp;) { 1069 struct ipq *fpp; 1070 1071 fpp = fp; 1072 fp = TAILQ_NEXT(fp, ipq_list); 1073 if(--fpp->ipq_ttl == 0) { 1074 ipstat.ips_fragtimeout += fpp->ipq_nfrags; 1075 ip_freef(&ipq[i], fpp); 1076 } 1077 } 1078 } 1079 /* 1080 * If we are over the maximum number of fragments 1081 * (due to the limit being lowered), drain off 1082 * enough to get down to the new limit. 1083 */ 1084 if (maxnipq >= 0 && nipq > maxnipq) { 1085 for (i = 0; i < IPREASS_NHASH; i++) { 1086 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) { 1087 ipstat.ips_fragdropped += 1088 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1089 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1090 } 1091 } 1092 } 1093 IPQ_UNLOCK(); 1094} 1095 1096/* 1097 * Drain off all datagram fragments. 1098 */ 1099void 1100ip_drain() 1101{ 1102 int i; 1103 1104 IPQ_LOCK(); 1105 for (i = 0; i < IPREASS_NHASH; i++) { 1106 while(!TAILQ_EMPTY(&ipq[i])) { 1107 ipstat.ips_fragdropped += 1108 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1109 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1110 } 1111 } 1112 IPQ_UNLOCK(); 1113 in_rtqdrain(); 1114} 1115 1116/* 1117 * The protocol to be inserted into ip_protox[] must be already registered 1118 * in inetsw[], either statically or through pf_proto_register(). 1119 */ 1120int 1121ipproto_register(u_char ipproto) 1122{ 1123 struct protosw *pr; 1124 1125 /* Sanity checks. */ 1126 if (ipproto == 0) 1127 return (EPROTONOSUPPORT); 1128 1129 /* 1130 * The protocol slot must not be occupied by another protocol 1131 * already. An index pointing to IPPROTO_RAW is unused. 1132 */ 1133 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1134 if (pr == NULL) 1135 return (EPFNOSUPPORT); 1136 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 1137 return (EEXIST); 1138 1139 /* Find the protocol position in inetsw[] and set the index. */ 1140 for (pr = inetdomain.dom_protosw; 1141 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 1142 if (pr->pr_domain->dom_family == PF_INET && 1143 pr->pr_protocol && pr->pr_protocol == ipproto) { 1144 /* Be careful to only index valid IP protocols. */ 1145 if (pr->pr_protocol < IPPROTO_MAX) { 1146 ip_protox[pr->pr_protocol] = pr - inetsw; 1147 return (0); 1148 } else 1149 return (EINVAL); 1150 } 1151 } 1152 return (EPROTONOSUPPORT); 1153} 1154 1155int 1156ipproto_unregister(u_char ipproto) 1157{ 1158 struct protosw *pr; 1159 1160 /* Sanity checks. */ 1161 if (ipproto == 0) 1162 return (EPROTONOSUPPORT); 1163 1164 /* Check if the protocol was indeed registered. */ 1165 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 1166 if (pr == NULL) 1167 return (EPFNOSUPPORT); 1168 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 1169 return (ENOENT); 1170 1171 /* Reset the protocol slot to IPPROTO_RAW. */ 1172 ip_protox[ipproto] = pr - inetsw; 1173 return (0); 1174} 1175 1176/* 1177 * Given address of next destination (final or next hop), 1178 * return internet address info of interface to be used to get there. 1179 */ 1180struct in_ifaddr * 1181ip_rtaddr(dst) 1182 struct in_addr dst; 1183{ 1184 struct route sro; 1185 struct sockaddr_in *sin; 1186 struct in_ifaddr *ifa; 1187 1188 bzero(&sro, sizeof(sro)); 1189 sin = (struct sockaddr_in *)&sro.ro_dst; 1190 sin->sin_family = AF_INET; 1191 sin->sin_len = sizeof(*sin); 1192 sin->sin_addr = dst; 1193 rtalloc_ign(&sro, RTF_CLONING); 1194 1195 if (sro.ro_rt == NULL) 1196 return (NULL); 1197 1198 ifa = ifatoia(sro.ro_rt->rt_ifa); 1199 RTFREE(sro.ro_rt); 1200 return (ifa); 1201} 1202 1203u_char inetctlerrmap[PRC_NCMDS] = { 1204 0, 0, 0, 0, 1205 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1206 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1207 EMSGSIZE, EHOSTUNREACH, 0, 0, 1208 0, 0, EHOSTUNREACH, 0, 1209 ENOPROTOOPT, ECONNREFUSED 1210}; 1211 1212/* 1213 * Forward a packet. If some error occurs return the sender 1214 * an icmp packet. Note we can't always generate a meaningful 1215 * icmp message because icmp doesn't have a large enough repertoire 1216 * of codes and types. 1217 * 1218 * If not forwarding, just drop the packet. This could be confusing 1219 * if ipforwarding was zero but some routing protocol was advancing 1220 * us as a gateway to somewhere. However, we must let the routing 1221 * protocol deal with that. 1222 * 1223 * The srcrt parameter indicates whether the packet is being forwarded 1224 * via a source route. 1225 */ 1226void 1227ip_forward(struct mbuf *m, int srcrt) 1228{ 1229 struct ip *ip = mtod(m, struct ip *); 1230 struct in_ifaddr *ia = NULL; 1231 struct mbuf *mcopy; 1232 struct in_addr dest; 1233 int error, type = 0, code = 0, mtu = 0; 1234 1235 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1236 ipstat.ips_cantforward++; 1237 m_freem(m); 1238 return; 1239 } 1240#ifdef IPSTEALTH 1241 if (!ipstealth) { 1242#endif 1243 if (ip->ip_ttl <= IPTTLDEC) { 1244 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1245 0, 0); 1246 return; 1247 } 1248#ifdef IPSTEALTH 1249 } 1250#endif 1251 1252 if (!srcrt && (ia = ip_rtaddr(ip->ip_dst)) == NULL) { 1253 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1254 return; 1255 } 1256 1257 /* 1258 * Save the IP header and at most 8 bytes of the payload, 1259 * in case we need to generate an ICMP message to the src. 1260 * 1261 * XXX this can be optimized a lot by saving the data in a local 1262 * buffer on the stack (72 bytes at most), and only allocating the 1263 * mbuf if really necessary. The vast majority of the packets 1264 * are forwarded without having to send an ICMP back (either 1265 * because unnecessary, or because rate limited), so we are 1266 * really we are wasting a lot of work here. 1267 * 1268 * We don't use m_copy() because it might return a reference 1269 * to a shared cluster. Both this function and ip_output() 1270 * assume exclusive access to the IP header in `m', so any 1271 * data in a cluster may change before we reach icmp_error(). 1272 */ 1273 MGETHDR(mcopy, M_DONTWAIT, m->m_type); 1274 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1275 /* 1276 * It's probably ok if the pkthdr dup fails (because 1277 * the deep copy of the tag chain failed), but for now 1278 * be conservative and just discard the copy since 1279 * code below may some day want the tags. 1280 */ 1281 m_free(mcopy); 1282 mcopy = NULL; 1283 } 1284 if (mcopy != NULL) { 1285 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy)); 1286 mcopy->m_pkthdr.len = mcopy->m_len; 1287 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1288 } 1289 1290#ifdef IPSTEALTH 1291 if (!ipstealth) { 1292#endif 1293 ip->ip_ttl -= IPTTLDEC; 1294#ifdef IPSTEALTH 1295 } 1296#endif 1297 1298 /* 1299 * If forwarding packet using same interface that it came in on, 1300 * perhaps should send a redirect to sender to shortcut a hop. 1301 * Only send redirect if source is sending directly to us, 1302 * and if packet was not source routed (or has any options). 1303 * Also, don't send redirect if forwarding using a default route 1304 * or a route modified by a redirect. 1305 */ 1306 dest.s_addr = 0; 1307 if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) { 1308 struct sockaddr_in *sin; 1309 struct route ro; 1310 struct rtentry *rt; 1311 1312 bzero(&ro, sizeof(ro)); 1313 sin = (struct sockaddr_in *)&ro.ro_dst; 1314 sin->sin_family = AF_INET; 1315 sin->sin_len = sizeof(*sin); 1316 sin->sin_addr = ip->ip_dst; 1317 rtalloc_ign(&ro, RTF_CLONING); 1318 1319 rt = ro.ro_rt; 1320 1321 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1322 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1323#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1324 u_long src = ntohl(ip->ip_src.s_addr); 1325 1326 if (RTA(rt) && 1327 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1328 if (rt->rt_flags & RTF_GATEWAY) 1329 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1330 else 1331 dest.s_addr = ip->ip_dst.s_addr; 1332 /* Router requirements says to only send host redirects */ 1333 type = ICMP_REDIRECT; 1334 code = ICMP_REDIRECT_HOST; 1335 } 1336 } 1337 if (rt) 1338 RTFREE(rt); 1339 } 1340 1341 error = ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL); 1342 if (error) 1343 ipstat.ips_cantforward++; 1344 else { 1345 ipstat.ips_forward++; 1346 if (type) 1347 ipstat.ips_redirectsent++; 1348 else { 1349 if (mcopy) 1350 m_freem(mcopy); 1351 return; 1352 } 1353 } 1354 if (mcopy == NULL) 1355 return; 1356 1357 switch (error) { 1358 1359 case 0: /* forwarded, but need redirect */ 1360 /* type, code set above */ 1361 break; 1362 1363 case ENETUNREACH: /* shouldn't happen, checked above */ 1364 case EHOSTUNREACH: 1365 case ENETDOWN: 1366 case EHOSTDOWN: 1367 default: 1368 type = ICMP_UNREACH; 1369 code = ICMP_UNREACH_HOST; 1370 break; 1371 1372 case EMSGSIZE: 1373 type = ICMP_UNREACH; 1374 code = ICMP_UNREACH_NEEDFRAG; 1375 1376#if defined(IPSEC) || defined(FAST_IPSEC) 1377 mtu = ip_ipsec_mtu(m); 1378#endif /* IPSEC */ 1379 /* 1380 * If the MTU wasn't set before use the interface mtu or 1381 * fall back to the next smaller mtu step compared to the 1382 * current packet size. 1383 */ 1384 if (mtu == 0) { 1385 if (ia != NULL) 1386 mtu = ia->ia_ifp->if_mtu; 1387 else 1388 mtu = ip_next_mtu(ip->ip_len, 0); 1389 } 1390 ipstat.ips_cantfrag++; 1391 break; 1392 1393 case ENOBUFS: 1394 /* 1395 * A router should not generate ICMP_SOURCEQUENCH as 1396 * required in RFC1812 Requirements for IP Version 4 Routers. 1397 * Source quench could be a big problem under DoS attacks, 1398 * or if the underlying interface is rate-limited. 1399 * Those who need source quench packets may re-enable them 1400 * via the net.inet.ip.sendsourcequench sysctl. 1401 */ 1402 if (ip_sendsourcequench == 0) { 1403 m_freem(mcopy); 1404 return; 1405 } else { 1406 type = ICMP_SOURCEQUENCH; 1407 code = 0; 1408 } 1409 break; 1410 1411 case EACCES: /* ipfw denied packet */ 1412 m_freem(mcopy); 1413 return; 1414 } 1415 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1416} 1417 1418void 1419ip_savecontrol(inp, mp, ip, m) 1420 register struct inpcb *inp; 1421 register struct mbuf **mp; 1422 register struct ip *ip; 1423 register struct mbuf *m; 1424{ 1425 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1426 struct bintime bt; 1427 1428 bintime(&bt); 1429 if (inp->inp_socket->so_options & SO_BINTIME) { 1430 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 1431 SCM_BINTIME, SOL_SOCKET); 1432 if (*mp) 1433 mp = &(*mp)->m_next; 1434 } 1435 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1436 struct timeval tv; 1437 1438 bintime2timeval(&bt, &tv); 1439 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1440 SCM_TIMESTAMP, SOL_SOCKET); 1441 if (*mp) 1442 mp = &(*mp)->m_next; 1443 } 1444 } 1445 if (inp->inp_flags & INP_RECVDSTADDR) { 1446 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1447 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1448 if (*mp) 1449 mp = &(*mp)->m_next; 1450 } 1451 if (inp->inp_flags & INP_RECVTTL) { 1452 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1453 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1454 if (*mp) 1455 mp = &(*mp)->m_next; 1456 } 1457#ifdef notyet 1458 /* XXX 1459 * Moving these out of udp_input() made them even more broken 1460 * than they already were. 1461 */ 1462 /* options were tossed already */ 1463 if (inp->inp_flags & INP_RECVOPTS) { 1464 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1465 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1466 if (*mp) 1467 mp = &(*mp)->m_next; 1468 } 1469 /* ip_srcroute doesn't do what we want here, need to fix */ 1470 if (inp->inp_flags & INP_RECVRETOPTS) { 1471 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1472 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1473 if (*mp) 1474 mp = &(*mp)->m_next; 1475 } 1476#endif 1477 if (inp->inp_flags & INP_RECVIF) { 1478 struct ifnet *ifp; 1479 struct sdlbuf { 1480 struct sockaddr_dl sdl; 1481 u_char pad[32]; 1482 } sdlbuf; 1483 struct sockaddr_dl *sdp; 1484 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1485 1486 if (((ifp = m->m_pkthdr.rcvif)) 1487 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1488 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1489 /* 1490 * Change our mind and don't try copy. 1491 */ 1492 if ((sdp->sdl_family != AF_LINK) 1493 || (sdp->sdl_len > sizeof(sdlbuf))) { 1494 goto makedummy; 1495 } 1496 bcopy(sdp, sdl2, sdp->sdl_len); 1497 } else { 1498makedummy: 1499 sdl2->sdl_len 1500 = offsetof(struct sockaddr_dl, sdl_data[0]); 1501 sdl2->sdl_family = AF_LINK; 1502 sdl2->sdl_index = 0; 1503 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1504 } 1505 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1506 IP_RECVIF, IPPROTO_IP); 1507 if (*mp) 1508 mp = &(*mp)->m_next; 1509 } 1510} 1511 1512/* 1513 * XXX these routines are called from the upper part of the kernel. 1514 * They need to be locked when we remove Giant. 1515 * 1516 * They could also be moved to ip_mroute.c, since all the RSVP 1517 * handling is done there already. 1518 */ 1519static int ip_rsvp_on; 1520struct socket *ip_rsvpd; 1521int 1522ip_rsvp_init(struct socket *so) 1523{ 1524 if (so->so_type != SOCK_RAW || 1525 so->so_proto->pr_protocol != IPPROTO_RSVP) 1526 return EOPNOTSUPP; 1527 1528 if (ip_rsvpd != NULL) 1529 return EADDRINUSE; 1530 1531 ip_rsvpd = so; 1532 /* 1533 * This may seem silly, but we need to be sure we don't over-increment 1534 * the RSVP counter, in case something slips up. 1535 */ 1536 if (!ip_rsvp_on) { 1537 ip_rsvp_on = 1; 1538 rsvp_on++; 1539 } 1540 1541 return 0; 1542} 1543 1544int 1545ip_rsvp_done(void) 1546{ 1547 ip_rsvpd = NULL; 1548 /* 1549 * This may seem silly, but we need to be sure we don't over-decrement 1550 * the RSVP counter, in case something slips up. 1551 */ 1552 if (ip_rsvp_on) { 1553 ip_rsvp_on = 0; 1554 rsvp_on--; 1555 } 1556 return 0; 1557} 1558 1559void 1560rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 1561{ 1562 if (rsvp_input_p) { /* call the real one if loaded */ 1563 rsvp_input_p(m, off); 1564 return; 1565 } 1566 1567 /* Can still get packets with rsvp_on = 0 if there is a local member 1568 * of the group to which the RSVP packet is addressed. But in this 1569 * case we want to throw the packet away. 1570 */ 1571 1572 if (!rsvp_on) { 1573 m_freem(m); 1574 return; 1575 } 1576 1577 if (ip_rsvpd != NULL) { 1578 rip_input(m, off); 1579 return; 1580 } 1581 /* Drop the packet */ 1582 m_freem(m); 1583} 1584