1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * The Internet Protocol (IP) module. 7 * 8 * Authors: Ross Biro 9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 10 * Donald Becker, <becker@super.org> 11 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 12 * Richard Underwood 13 * Stefan Becker, <stefanb@yello.ping.de> 14 * Jorge Cwik, <jorge@laser.satlink.net> 15 * Arnt Gulbrandsen, <agulbra@nvg.unit.no> 16 * 17 * 18 * Fixes: 19 * Alan Cox : Commented a couple of minor bits of surplus code 20 * Alan Cox : Undefining IP_FORWARD doesn't include the code 21 * (just stops a compiler warning). 22 * Alan Cox : Frames with >=MAX_ROUTE record routes, strict routes or loose routes 23 * are junked rather than corrupting things. 24 * Alan Cox : Frames to bad broadcast subnets are dumped 25 * We used to process them non broadcast and 26 * boy could that cause havoc. 27 * Alan Cox : ip_forward sets the free flag on the 28 * new frame it queues. Still crap because 29 * it copies the frame but at least it 30 * doesn't eat memory too. 31 * Alan Cox : Generic queue code and memory fixes. 32 * Fred Van Kempen : IP fragment support (borrowed from NET2E) 33 * Gerhard Koerting: Forward fragmented frames correctly. 34 * Gerhard Koerting: Fixes to my fix of the above 8-). 35 * Gerhard Koerting: IP interface addressing fix. 36 * Linus Torvalds : More robustness checks 37 * Alan Cox : Even more checks: Still not as robust as it ought to be 38 * Alan Cox : Save IP header pointer for later 39 * Alan Cox : ip option setting 40 * Alan Cox : Use ip_tos/ip_ttl settings 41 * Alan Cox : Fragmentation bogosity removed 42 * (Thanks to Mark.Bush@prg.ox.ac.uk) 43 * Dmitry Gorodchanin : Send of a raw packet crash fix. 44 * Alan Cox : Silly ip bug when an overlength 45 * fragment turns up. Now frees the 46 * queue. 47 * Linus Torvalds/ : Memory leakage on fragmentation 48 * Alan Cox : handling. 49 * Gerhard Koerting: Forwarding uses IP priority hints 50 * Teemu Rantanen : Fragment problems. 51 * Alan Cox : General cleanup, comments and reformat 52 * Alan Cox : SNMP statistics 53 * Alan Cox : BSD address rule semantics. Also see 54 * UDP as there is a nasty checksum issue 55 * if you do things the wrong way. 56 * Alan Cox : Always defrag, moved IP_FORWARD to the config.in file 57 * Alan Cox : IP options adjust sk->priority. 58 * Pedro Roque : Fix mtu/length error in ip_forward. 59 * Alan Cox : Avoid ip_chk_addr when possible. 60 * Richard Underwood : IP multicasting. 61 * Alan Cox : Cleaned up multicast handlers. 62 * Alan Cox : RAW sockets demultiplex in the BSD style. 63 * Gunther Mayer : Fix the SNMP reporting typo 64 * Alan Cox : Always in group 224.0.0.1 65 * Pauline Middelink : Fast ip_checksum update when forwarding 66 * Masquerading support. 67 * Alan Cox : Multicast loopback error for 224.0.0.1 68 * Alan Cox : IP_MULTICAST_LOOP option. 69 * Alan Cox : Use notifiers. 70 * Bjorn Ekwall : Removed ip_csum (from slhc.c too) 71 * Bjorn Ekwall : Moved ip_fast_csum to ip.h (inline!) 72 * Stefan Becker : Send out ICMP HOST REDIRECT 73 * Arnt Gulbrandsen : ip_build_xmit 74 * Alan Cox : Per socket routing cache 75 * Alan Cox : Fixed routing cache, added header cache. 76 * Alan Cox : Loopback didn't work right in original ip_build_xmit - fixed it. 77 * Alan Cox : Only send ICMP_REDIRECT if src/dest are the same net. 78 * Alan Cox : Incoming IP option handling. 79 * Alan Cox : Set saddr on raw output frames as per BSD. 80 * Alan Cox : Stopped broadcast source route explosions. 81 * Alan Cox : Can disable source routing 82 * Takeshi Sone : Masquerading didn't work. 83 * Dave Bonn,Alan Cox : Faster IP forwarding whenever possible. 84 * Alan Cox : Memory leaks, tramples, misc debugging. 85 * Alan Cox : Fixed multicast (by popular demand 8)) 86 * Alan Cox : Fixed forwarding (by even more popular demand 8)) 87 * Alan Cox : Fixed SNMP statistics [I think] 88 * Gerhard Koerting : IP fragmentation forwarding fix 89 * Alan Cox : Device lock against page fault. 90 * Alan Cox : IP_HDRINCL facility. 91 * Werner Almesberger : Zero fragment bug 92 * Alan Cox : RAW IP frame length bug 93 * Alan Cox : Outgoing firewall on build_xmit 94 * A.N.Kuznetsov : IP_OPTIONS support throughout the kernel 95 * Alan Cox : Multicast routing hooks 96 * Jos Vos : Do accounting *before* call_in_firewall 97 * Willy Konynenberg : Transparent proxying support 98 * 99 * 100 * 101 * To Fix: 102 * IP fragmentation wants rewriting cleanly. The RFC815 algorithm is much more efficient 103 * and could be made very efficient with the addition of some virtual memory hacks to permit 104 * the allocation of a buffer that can then be 'grown' by twiddling page tables. 105 * Output fragmentation wants updating along with the buffer management to use a single 106 * interleaved copy algorithm so that fragmenting has a one copy overhead. Actual packet 107 * output should probably do its own fragmentation at the UDP/RAW layer. TCP shouldn't cause 108 * fragmentation anyway. 109 * 110 * This program is free software; you can redistribute it and/or 111 * modify it under the terms of the GNU General Public License 112 * as published by the Free Software Foundation; either version 113 * 2 of the License, or (at your option) any later version. 114 */ 115 116 #define pr_fmt(fmt) "IPv4: " fmt 117 118 #include <linux/module.h> 119 #include <linux/types.h> 120 #include <linux/kernel.h> 121 #include <linux/string.h> 122 #include <linux/errno.h> 123 #include <linux/slab.h> 124 125 #include <linux/net.h> 126 #include <linux/socket.h> 127 #include <linux/sockios.h> 128 #include <linux/in.h> 129 #include <linux/inet.h> 130 #include <linux/inetdevice.h> 131 #include <linux/netdevice.h> 132 #include <linux/etherdevice.h> 133 134 #include <net/snmp.h> 135 #include <net/ip.h> 136 #include <net/protocol.h> 137 #include <net/route.h> 138 #include <linux/skbuff.h> 139 #include <net/sock.h> 140 #include <net/arp.h> 141 #include <net/icmp.h> 142 #include <net/raw.h> 143 #include <net/checksum.h> 144 #include <net/inet_ecn.h> 145 #include <linux/netfilter_ipv4.h> 146 #include <net/xfrm.h> 147 #include <linux/mroute.h> 148 #include <linux/netlink.h> 149 #include <net/dst_metadata.h> 150 151 /* 152 * Process Router Attention IP option (RFC 2113) 153 */ 154 bool ip_call_ra_chain(struct sk_buff *skb) 155 { 156 struct ip_ra_chain *ra; 157 u8 protocol = ip_hdr(skb)->protocol; 158 struct sock *last = NULL; 159 struct net_device *dev = skb->dev; 160 struct net *net = dev_net(dev); 161 162 for (ra = rcu_dereference(net->ipv4.ra_chain); ra; ra = rcu_dereference(ra->next)) { 163 struct sock *sk = ra->sk; 164 165 /* If socket is bound to an interface, only report 166 * the packet if it came from that interface. 167 */ 168 if (sk && inet_sk(sk)->inet_num == protocol && 169 (!sk->sk_bound_dev_if || 170 sk->sk_bound_dev_if == dev->ifindex)) { 171 if (ip_is_fragment(ip_hdr(skb))) { 172 if (ip_defrag(net, skb, IP_DEFRAG_CALL_RA_CHAIN)) 173 return true; 174 } 175 if (last) { 176 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 177 if (skb2) 178 raw_rcv(last, skb2); 179 } 180 last = sk; 181 } 182 } 183 184 if (last) { 185 raw_rcv(last, skb); 186 return true; 187 } 188 return false; 189 } 190 191 void ip_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int protocol) 192 { 193 const struct net_protocol *ipprot; 194 int raw, ret; 195 196 resubmit: 197 raw = raw_local_deliver(skb, protocol); 198 199 ipprot = rcu_dereference(inet_protos[protocol]); 200 if (ipprot) { 201 if (!ipprot->no_policy) { 202 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) { 203 kfree_skb(skb); 204 return; 205 } 206 nf_reset(skb); 207 } 208 ret = ipprot->handler(skb); 209 if (ret < 0) { 210 protocol = -ret; 211 goto resubmit; 212 } 213 __IP_INC_STATS(net, IPSTATS_MIB_INDELIVERS); 214 } else { 215 if (!raw) { 216 if (xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) { 217 __IP_INC_STATS(net, IPSTATS_MIB_INUNKNOWNPROTOS); 218 icmp_send(skb, ICMP_DEST_UNREACH, 219 ICMP_PROT_UNREACH, 0); 220 } 221 kfree_skb(skb); 222 } else { 223 __IP_INC_STATS(net, IPSTATS_MIB_INDELIVERS); 224 consume_skb(skb); 225 } 226 } 227 } 228 229 static int ip_local_deliver_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 230 { 231 __skb_pull(skb, skb_network_header_len(skb)); 232 233 rcu_read_lock(); 234 ip_protocol_deliver_rcu(net, skb, ip_hdr(skb)->protocol); 235 rcu_read_unlock(); 236 237 return 0; 238 } 239 240 /* 241 * Deliver IP Packets to the higher protocol layers. 242 */ 243 int ip_local_deliver(struct sk_buff *skb) 244 { 245 /* 246 * Reassemble IP fragments. 247 */ 248 struct net *net = dev_net(skb->dev); 249 250 if (ip_is_fragment(ip_hdr(skb))) { 251 if (ip_defrag(net, skb, IP_DEFRAG_LOCAL_DELIVER)) 252 return 0; 253 } 254 255 return NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_IN, 256 net, NULL, skb, skb->dev, NULL, 257 ip_local_deliver_finish); 258 } 259 260 static inline bool ip_rcv_options(struct sk_buff *skb) 261 { 262 struct ip_options *opt; 263 const struct iphdr *iph; 264 struct net_device *dev = skb->dev; 265 266 /* It looks as overkill, because not all 267 IP options require packet mangling. 268 But it is the easiest for now, especially taking 269 into account that combination of IP options 270 and running sniffer is extremely rare condition. 271 --ANK (980813) 272 */ 273 if (skb_cow(skb, skb_headroom(skb))) { 274 __IP_INC_STATS(dev_net(dev), IPSTATS_MIB_INDISCARDS); 275 goto drop; 276 } 277 278 iph = ip_hdr(skb); 279 opt = &(IPCB(skb)->opt); 280 opt->optlen = iph->ihl*4 - sizeof(struct iphdr); 281 282 if (ip_options_compile(dev_net(dev), opt, skb)) { 283 __IP_INC_STATS(dev_net(dev), IPSTATS_MIB_INHDRERRORS); 284 goto drop; 285 } 286 287 if (unlikely(opt->srr)) { 288 struct in_device *in_dev = __in_dev_get_rcu(dev); 289 290 if (in_dev) { 291 if (!IN_DEV_SOURCE_ROUTE(in_dev)) { 292 if (IN_DEV_LOG_MARTIANS(in_dev)) 293 net_info_ratelimited("source route option %pI4 -> %pI4\n", 294 &iph->saddr, 295 &iph->daddr); 296 goto drop; 297 } 298 } 299 300 if (ip_options_rcv_srr(skb)) 301 goto drop; 302 } 303 304 return false; 305 drop: 306 return true; 307 } 308 309 static int ip_rcv_finish_core(struct net *net, struct sock *sk, 310 struct sk_buff *skb) 311 { 312 const struct iphdr *iph = ip_hdr(skb); 313 int (*edemux)(struct sk_buff *skb); 314 struct net_device *dev = skb->dev; 315 struct rtable *rt; 316 int err; 317 318 if (net->ipv4.sysctl_ip_early_demux && 319 !skb_dst(skb) && 320 !skb->sk && 321 !ip_is_fragment(iph)) { 322 const struct net_protocol *ipprot; 323 int protocol = iph->protocol; 324 325 ipprot = rcu_dereference(inet_protos[protocol]); 326 if (ipprot && (edemux = READ_ONCE(ipprot->early_demux))) { 327 err = edemux(skb); 328 if (unlikely(err)) 329 goto drop_error; 330 /* must reload iph, skb->head might have changed */ 331 iph = ip_hdr(skb); 332 } 333 } 334 335 /* 336 * Initialise the virtual path cache for the packet. It describes 337 * how the packet travels inside Linux networking. 338 */ 339 if (!skb_valid_dst(skb)) { 340 err = ip_route_input_noref(skb, iph->daddr, iph->saddr, 341 iph->tos, dev); 342 if (unlikely(err)) 343 goto drop_error; 344 } 345 346 #ifdef CONFIG_IP_ROUTE_CLASSID 347 if (unlikely(skb_dst(skb)->tclassid)) { 348 struct ip_rt_acct *st = this_cpu_ptr(ip_rt_acct); 349 u32 idx = skb_dst(skb)->tclassid; 350 st[idx&0xFF].o_packets++; 351 st[idx&0xFF].o_bytes += skb->len; 352 st[(idx>>16)&0xFF].i_packets++; 353 st[(idx>>16)&0xFF].i_bytes += skb->len; 354 } 355 #endif 356 357 if (iph->ihl > 5 && ip_rcv_options(skb)) 358 goto drop; 359 360 rt = skb_rtable(skb); 361 if (rt->rt_type == RTN_MULTICAST) { 362 __IP_UPD_PO_STATS(net, IPSTATS_MIB_INMCAST, skb->len); 363 } else if (rt->rt_type == RTN_BROADCAST) { 364 __IP_UPD_PO_STATS(net, IPSTATS_MIB_INBCAST, skb->len); 365 } else if (skb->pkt_type == PACKET_BROADCAST || 366 skb->pkt_type == PACKET_MULTICAST) { 367 struct in_device *in_dev = __in_dev_get_rcu(dev); 368 369 /* RFC 1122 3.3.6: 370 * 371 * When a host sends a datagram to a link-layer broadcast 372 * address, the IP destination address MUST be a legal IP 373 * broadcast or IP multicast address. 374 * 375 * A host SHOULD silently discard a datagram that is received 376 * via a link-layer broadcast (see Section 2.4) but does not 377 * specify an IP multicast or broadcast destination address. 378 * 379 * This doesn't explicitly say L2 *broadcast*, but broadcast is 380 * in a way a form of multicast and the most common use case for 381 * this is 802.11 protecting against cross-station spoofing (the 382 * so-called "hole-196" attack) so do it for both. 383 */ 384 if (in_dev && 385 IN_DEV_ORCONF(in_dev, DROP_UNICAST_IN_L2_MULTICAST)) 386 goto drop; 387 } 388 389 return NET_RX_SUCCESS; 390 391 drop: 392 kfree_skb(skb); 393 return NET_RX_DROP; 394 395 drop_error: 396 if (err == -EXDEV) 397 __NET_INC_STATS(net, LINUX_MIB_IPRPFILTER); 398 goto drop; 399 } 400 401 static int ip_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 402 { 403 int ret; 404 405 /* if ingress device is enslaved to an L3 master device pass the 406 * skb to its handler for processing 407 */ 408 skb = l3mdev_ip_rcv(skb); 409 if (!skb) 410 return NET_RX_SUCCESS; 411 412 ret = ip_rcv_finish_core(net, sk, skb); 413 if (ret != NET_RX_DROP) 414 ret = dst_input(skb); 415 return ret; 416 } 417 418 /* 419 * Main IP Receive routine. 420 */ 421 static struct sk_buff *ip_rcv_core(struct sk_buff *skb, struct net *net) 422 { 423 const struct iphdr *iph; 424 u32 len; 425 426 /* When the interface is in promisc. mode, drop all the crap 427 * that it receives, do not try to analyse it. 428 */ 429 if (skb->pkt_type == PACKET_OTHERHOST) 430 goto drop; 431 432 433 __IP_UPD_PO_STATS(net, IPSTATS_MIB_IN, skb->len); 434 435 skb = skb_share_check(skb, GFP_ATOMIC); 436 if (!skb) { 437 __IP_INC_STATS(net, IPSTATS_MIB_INDISCARDS); 438 goto out; 439 } 440 441 if (!pskb_may_pull(skb, sizeof(struct iphdr))) 442 goto inhdr_error; 443 444 iph = ip_hdr(skb); 445 446 /* 447 * RFC1122: 3.2.1.2 MUST silently discard any IP frame that fails the checksum. 448 * 449 * Is the datagram acceptable? 450 * 451 * 1. Length at least the size of an ip header 452 * 2. Version of 4 453 * 3. Checksums correctly. [Speed optimisation for later, skip loopback checksums] 454 * 4. Doesn't have a bogus length 455 */ 456 457 if (iph->ihl < 5 || iph->version != 4) 458 goto inhdr_error; 459 460 BUILD_BUG_ON(IPSTATS_MIB_ECT1PKTS != IPSTATS_MIB_NOECTPKTS + INET_ECN_ECT_1); 461 BUILD_BUG_ON(IPSTATS_MIB_ECT0PKTS != IPSTATS_MIB_NOECTPKTS + INET_ECN_ECT_0); 462 BUILD_BUG_ON(IPSTATS_MIB_CEPKTS != IPSTATS_MIB_NOECTPKTS + INET_ECN_CE); 463 __IP_ADD_STATS(net, 464 IPSTATS_MIB_NOECTPKTS + (iph->tos & INET_ECN_MASK), 465 max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs)); 466 467 if (!pskb_may_pull(skb, iph->ihl*4)) 468 goto inhdr_error; 469 470 iph = ip_hdr(skb); 471 472 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl))) 473 goto csum_error; 474 475 len = ntohs(iph->tot_len); 476 if (skb->len < len) { 477 __IP_INC_STATS(net, IPSTATS_MIB_INTRUNCATEDPKTS); 478 goto drop; 479 } else if (len < (iph->ihl*4)) 480 goto inhdr_error; 481 482 /* Our transport medium may have padded the buffer out. Now we know it 483 * is IP we can trim to the true length of the frame. 484 * Note this now means skb->len holds ntohs(iph->tot_len). 485 */ 486 if (pskb_trim_rcsum(skb, len)) { 487 __IP_INC_STATS(net, IPSTATS_MIB_INDISCARDS); 488 goto drop; 489 } 490 491 iph = ip_hdr(skb); 492 skb->transport_header = skb->network_header + iph->ihl*4; 493 494 /* Remove any debris in the socket control block */ 495 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm)); 496 IPCB(skb)->iif = skb->skb_iif; 497 498 /* Must drop socket now because of tproxy. */ 499 skb_orphan(skb); 500 501 return skb; 502 503 csum_error: 504 __IP_INC_STATS(net, IPSTATS_MIB_CSUMERRORS); 505 inhdr_error: 506 __IP_INC_STATS(net, IPSTATS_MIB_INHDRERRORS); 507 drop: 508 kfree_skb(skb); 509 out: 510 return NULL; 511 } 512 513 /* 514 * IP receive entry point 515 */ 516 int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, 517 struct net_device *orig_dev) 518 { 519 struct net *net = dev_net(dev); 520 521 skb = ip_rcv_core(skb, net); 522 if (skb == NULL) 523 return NET_RX_DROP; 524 return NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, 525 net, NULL, skb, dev, NULL, 526 ip_rcv_finish); 527 } 528 529 static void ip_sublist_rcv_finish(struct list_head *head) 530 { 531 struct sk_buff *skb, *next; 532 533 list_for_each_entry_safe(skb, next, head, list) { 534 skb_list_del_init(skb); 535 dst_input(skb); 536 } 537 } 538 539 static void ip_list_rcv_finish(struct net *net, struct sock *sk, 540 struct list_head *head) 541 { 542 struct dst_entry *curr_dst = NULL; 543 struct sk_buff *skb, *next; 544 struct list_head sublist; 545 546 INIT_LIST_HEAD(&sublist); 547 list_for_each_entry_safe(skb, next, head, list) { 548 struct dst_entry *dst; 549 550 skb_list_del_init(skb); 551 /* if ingress device is enslaved to an L3 master device pass the 552 * skb to its handler for processing 553 */ 554 skb = l3mdev_ip_rcv(skb); 555 if (!skb) 556 continue; 557 if (ip_rcv_finish_core(net, sk, skb) == NET_RX_DROP) 558 continue; 559 560 dst = skb_dst(skb); 561 if (curr_dst != dst) { 562 /* dispatch old sublist */ 563 if (!list_empty(&sublist)) 564 ip_sublist_rcv_finish(&sublist); 565 /* start new sublist */ 566 INIT_LIST_HEAD(&sublist); 567 curr_dst = dst; 568 } 569 list_add_tail(&skb->list, &sublist); 570 } 571 /* dispatch final sublist */ 572 ip_sublist_rcv_finish(&sublist); 573 } 574 575 static void ip_sublist_rcv(struct list_head *head, struct net_device *dev, 576 struct net *net) 577 { 578 NF_HOOK_LIST(NFPROTO_IPV4, NF_INET_PRE_ROUTING, net, NULL, 579 head, dev, NULL, ip_rcv_finish); 580 ip_list_rcv_finish(net, NULL, head); 581 } 582 583 /* Receive a list of IP packets */ 584 void ip_list_rcv(struct list_head *head, struct packet_type *pt, 585 struct net_device *orig_dev) 586 { 587 struct net_device *curr_dev = NULL; 588 struct net *curr_net = NULL; 589 struct sk_buff *skb, *next; 590 struct list_head sublist; 591 592 INIT_LIST_HEAD(&sublist); 593 list_for_each_entry_safe(skb, next, head, list) { 594 struct net_device *dev = skb->dev; 595 struct net *net = dev_net(dev); 596 597 skb_list_del_init(skb); 598 skb = ip_rcv_core(skb, net); 599 if (skb == NULL) 600 continue; 601 602 if (curr_dev != dev || curr_net != net) { 603 /* dispatch old sublist */ 604 if (!list_empty(&sublist)) 605 ip_sublist_rcv(&sublist, curr_dev, curr_net); 606 /* start new sublist */ 607 INIT_LIST_HEAD(&sublist); 608 curr_dev = dev; 609 curr_net = net; 610 } 611 list_add_tail(&skb->list, &sublist); 612 } 613 /* dispatch final sublist */ 614 ip_sublist_rcv(&sublist, curr_dev, curr_net); 615 } 616