11da177e4SLinus Torvalds /* 21da177e4SLinus Torvalds * INET An implementation of the TCP/IP protocol suite for the LINUX 31da177e4SLinus Torvalds * operating system. INET is implemented using the BSD Socket 41da177e4SLinus Torvalds * interface as the means of communication with the user level. 51da177e4SLinus Torvalds * 61da177e4SLinus Torvalds * The Internet Protocol (IP) module. 71da177e4SLinus Torvalds * 81da177e4SLinus Torvalds * Version: $Id: ip_input.c,v 1.55 2002/01/12 07:39:45 davem Exp $ 91da177e4SLinus Torvalds * 1002c30a84SJesper Juhl * Authors: Ross Biro 111da177e4SLinus Torvalds * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 121da177e4SLinus Torvalds * Donald Becker, <becker@super.org> 131da177e4SLinus Torvalds * Alan Cox, <Alan.Cox@linux.org> 141da177e4SLinus Torvalds * Richard Underwood 151da177e4SLinus Torvalds * Stefan Becker, <stefanb@yello.ping.de> 161da177e4SLinus Torvalds * Jorge Cwik, <jorge@laser.satlink.net> 171da177e4SLinus Torvalds * Arnt Gulbrandsen, <agulbra@nvg.unit.no> 181da177e4SLinus Torvalds * 191da177e4SLinus Torvalds * 201da177e4SLinus Torvalds * Fixes: 211da177e4SLinus Torvalds * Alan Cox : Commented a couple of minor bits of surplus code 221da177e4SLinus Torvalds * Alan Cox : Undefining IP_FORWARD doesn't include the code 231da177e4SLinus Torvalds * (just stops a compiler warning). 241da177e4SLinus Torvalds * Alan Cox : Frames with >=MAX_ROUTE record routes, strict routes or loose routes 251da177e4SLinus Torvalds * are junked rather than corrupting things. 261da177e4SLinus Torvalds * Alan Cox : Frames to bad broadcast subnets are dumped 271da177e4SLinus Torvalds * We used to process them non broadcast and 281da177e4SLinus Torvalds * boy could that cause havoc. 291da177e4SLinus Torvalds * Alan Cox : ip_forward sets the free flag on the 301da177e4SLinus Torvalds * new frame it queues. Still crap because 311da177e4SLinus Torvalds * it copies the frame but at least it 321da177e4SLinus Torvalds * doesn't eat memory too. 331da177e4SLinus Torvalds * Alan Cox : Generic queue code and memory fixes. 341da177e4SLinus Torvalds * Fred Van Kempen : IP fragment support (borrowed from NET2E) 351da177e4SLinus Torvalds * Gerhard Koerting: Forward fragmented frames correctly. 361da177e4SLinus Torvalds * Gerhard Koerting: Fixes to my fix of the above 8-). 371da177e4SLinus Torvalds * Gerhard Koerting: IP interface addressing fix. 381da177e4SLinus Torvalds * Linus Torvalds : More robustness checks 391da177e4SLinus Torvalds * Alan Cox : Even more checks: Still not as robust as it ought to be 401da177e4SLinus Torvalds * Alan Cox : Save IP header pointer for later 411da177e4SLinus Torvalds * Alan Cox : ip option setting 421da177e4SLinus Torvalds * Alan Cox : Use ip_tos/ip_ttl settings 431da177e4SLinus Torvalds * Alan Cox : Fragmentation bogosity removed 441da177e4SLinus Torvalds * (Thanks to Mark.Bush@prg.ox.ac.uk) 451da177e4SLinus Torvalds * Dmitry Gorodchanin : Send of a raw packet crash fix. 461da177e4SLinus Torvalds * Alan Cox : Silly ip bug when an overlength 471da177e4SLinus Torvalds * fragment turns up. Now frees the 481da177e4SLinus Torvalds * queue. 491da177e4SLinus Torvalds * Linus Torvalds/ : Memory leakage on fragmentation 501da177e4SLinus Torvalds * Alan Cox : handling. 511da177e4SLinus Torvalds * Gerhard Koerting: Forwarding uses IP priority hints 521da177e4SLinus Torvalds * Teemu Rantanen : Fragment problems. 531da177e4SLinus Torvalds * Alan Cox : General cleanup, comments and reformat 541da177e4SLinus Torvalds * Alan Cox : SNMP statistics 551da177e4SLinus Torvalds * Alan Cox : BSD address rule semantics. Also see 561da177e4SLinus Torvalds * UDP as there is a nasty checksum issue 571da177e4SLinus Torvalds * if you do things the wrong way. 581da177e4SLinus Torvalds * Alan Cox : Always defrag, moved IP_FORWARD to the config.in file 591da177e4SLinus Torvalds * Alan Cox : IP options adjust sk->priority. 601da177e4SLinus Torvalds * Pedro Roque : Fix mtu/length error in ip_forward. 611da177e4SLinus Torvalds * Alan Cox : Avoid ip_chk_addr when possible. 621da177e4SLinus Torvalds * Richard Underwood : IP multicasting. 631da177e4SLinus Torvalds * Alan Cox : Cleaned up multicast handlers. 641da177e4SLinus Torvalds * Alan Cox : RAW sockets demultiplex in the BSD style. 651da177e4SLinus Torvalds * Gunther Mayer : Fix the SNMP reporting typo 661da177e4SLinus Torvalds * Alan Cox : Always in group 224.0.0.1 671da177e4SLinus Torvalds * Pauline Middelink : Fast ip_checksum update when forwarding 681da177e4SLinus Torvalds * Masquerading support. 691da177e4SLinus Torvalds * Alan Cox : Multicast loopback error for 224.0.0.1 701da177e4SLinus Torvalds * Alan Cox : IP_MULTICAST_LOOP option. 711da177e4SLinus Torvalds * Alan Cox : Use notifiers. 721da177e4SLinus Torvalds * Bjorn Ekwall : Removed ip_csum (from slhc.c too) 731da177e4SLinus Torvalds * Bjorn Ekwall : Moved ip_fast_csum to ip.h (inline!) 741da177e4SLinus Torvalds * Stefan Becker : Send out ICMP HOST REDIRECT 751da177e4SLinus Torvalds * Arnt Gulbrandsen : ip_build_xmit 761da177e4SLinus Torvalds * Alan Cox : Per socket routing cache 771da177e4SLinus Torvalds * Alan Cox : Fixed routing cache, added header cache. 781da177e4SLinus Torvalds * Alan Cox : Loopback didn't work right in original ip_build_xmit - fixed it. 791da177e4SLinus Torvalds * Alan Cox : Only send ICMP_REDIRECT if src/dest are the same net. 801da177e4SLinus Torvalds * Alan Cox : Incoming IP option handling. 811da177e4SLinus Torvalds * Alan Cox : Set saddr on raw output frames as per BSD. 821da177e4SLinus Torvalds * Alan Cox : Stopped broadcast source route explosions. 831da177e4SLinus Torvalds * Alan Cox : Can disable source routing 841da177e4SLinus Torvalds * Takeshi Sone : Masquerading didn't work. 851da177e4SLinus Torvalds * Dave Bonn,Alan Cox : Faster IP forwarding whenever possible. 861da177e4SLinus Torvalds * Alan Cox : Memory leaks, tramples, misc debugging. 871da177e4SLinus Torvalds * Alan Cox : Fixed multicast (by popular demand 8)) 881da177e4SLinus Torvalds * Alan Cox : Fixed forwarding (by even more popular demand 8)) 891da177e4SLinus Torvalds * Alan Cox : Fixed SNMP statistics [I think] 901da177e4SLinus Torvalds * Gerhard Koerting : IP fragmentation forwarding fix 911da177e4SLinus Torvalds * Alan Cox : Device lock against page fault. 921da177e4SLinus Torvalds * Alan Cox : IP_HDRINCL facility. 931da177e4SLinus Torvalds * Werner Almesberger : Zero fragment bug 941da177e4SLinus Torvalds * Alan Cox : RAW IP frame length bug 951da177e4SLinus Torvalds * Alan Cox : Outgoing firewall on build_xmit 961da177e4SLinus Torvalds * A.N.Kuznetsov : IP_OPTIONS support throughout the kernel 971da177e4SLinus Torvalds * Alan Cox : Multicast routing hooks 981da177e4SLinus Torvalds * Jos Vos : Do accounting *before* call_in_firewall 991da177e4SLinus Torvalds * Willy Konynenberg : Transparent proxying support 1001da177e4SLinus Torvalds * 1011da177e4SLinus Torvalds * 1021da177e4SLinus Torvalds * 1031da177e4SLinus Torvalds * To Fix: 1041da177e4SLinus Torvalds * IP fragmentation wants rewriting cleanly. The RFC815 algorithm is much more efficient 1051da177e4SLinus Torvalds * and could be made very efficient with the addition of some virtual memory hacks to permit 1061da177e4SLinus Torvalds * the allocation of a buffer that can then be 'grown' by twiddling page tables. 1071da177e4SLinus Torvalds * Output fragmentation wants updating along with the buffer management to use a single 1081da177e4SLinus Torvalds * interleaved copy algorithm so that fragmenting has a one copy overhead. Actual packet 1091da177e4SLinus Torvalds * output should probably do its own fragmentation at the UDP/RAW layer. TCP shouldn't cause 1101da177e4SLinus Torvalds * fragmentation anyway. 1111da177e4SLinus Torvalds * 1121da177e4SLinus Torvalds * This program is free software; you can redistribute it and/or 1131da177e4SLinus Torvalds * modify it under the terms of the GNU General Public License 1141da177e4SLinus Torvalds * as published by the Free Software Foundation; either version 1151da177e4SLinus Torvalds * 2 of the License, or (at your option) any later version. 1161da177e4SLinus Torvalds */ 1171da177e4SLinus Torvalds 1181da177e4SLinus Torvalds #include <asm/system.h> 1191da177e4SLinus Torvalds #include <linux/module.h> 1201da177e4SLinus Torvalds #include <linux/types.h> 1211da177e4SLinus Torvalds #include <linux/kernel.h> 1221da177e4SLinus Torvalds #include <linux/string.h> 1231da177e4SLinus Torvalds #include <linux/errno.h> 1241da177e4SLinus Torvalds 1251da177e4SLinus Torvalds #include <linux/net.h> 1261da177e4SLinus Torvalds #include <linux/socket.h> 1271da177e4SLinus Torvalds #include <linux/sockios.h> 1281da177e4SLinus Torvalds #include <linux/in.h> 1291da177e4SLinus Torvalds #include <linux/inet.h> 13014c85021SArnaldo Carvalho de Melo #include <linux/inetdevice.h> 1311da177e4SLinus Torvalds #include <linux/netdevice.h> 1321da177e4SLinus Torvalds #include <linux/etherdevice.h> 1331da177e4SLinus Torvalds 1341da177e4SLinus Torvalds #include <net/snmp.h> 1351da177e4SLinus Torvalds #include <net/ip.h> 1361da177e4SLinus Torvalds #include <net/protocol.h> 1371da177e4SLinus Torvalds #include <net/route.h> 1381da177e4SLinus Torvalds #include <linux/skbuff.h> 1391da177e4SLinus Torvalds #include <net/sock.h> 1401da177e4SLinus Torvalds #include <net/arp.h> 1411da177e4SLinus Torvalds #include <net/icmp.h> 1421da177e4SLinus Torvalds #include <net/raw.h> 1431da177e4SLinus Torvalds #include <net/checksum.h> 1441da177e4SLinus Torvalds #include <linux/netfilter_ipv4.h> 1451da177e4SLinus Torvalds #include <net/xfrm.h> 1461da177e4SLinus Torvalds #include <linux/mroute.h> 1471da177e4SLinus Torvalds #include <linux/netlink.h> 1481da177e4SLinus Torvalds 1491da177e4SLinus Torvalds /* 1501da177e4SLinus Torvalds * SNMP management statistics 1511da177e4SLinus Torvalds */ 1521da177e4SLinus Torvalds 153ba89966cSEric Dumazet DEFINE_SNMP_STAT(struct ipstats_mib, ip_statistics) __read_mostly; 1541da177e4SLinus Torvalds 1551da177e4SLinus Torvalds /* 1561da177e4SLinus Torvalds * Process Router Attention IP option 1571da177e4SLinus Torvalds */ 1581da177e4SLinus Torvalds int ip_call_ra_chain(struct sk_buff *skb) 1591da177e4SLinus Torvalds { 1601da177e4SLinus Torvalds struct ip_ra_chain *ra; 161eddc9ec5SArnaldo Carvalho de Melo u8 protocol = ip_hdr(skb)->protocol; 1621da177e4SLinus Torvalds struct sock *last = NULL; 1631da177e4SLinus Torvalds 1641da177e4SLinus Torvalds read_lock(&ip_ra_lock); 1651da177e4SLinus Torvalds for (ra = ip_ra_chain; ra; ra = ra->next) { 1661da177e4SLinus Torvalds struct sock *sk = ra->sk; 1671da177e4SLinus Torvalds 1681da177e4SLinus Torvalds /* If socket is bound to an interface, only report 1691da177e4SLinus Torvalds * the packet if it came from that interface. 1701da177e4SLinus Torvalds */ 1711da177e4SLinus Torvalds if (sk && inet_sk(sk)->num == protocol && 1721da177e4SLinus Torvalds (!sk->sk_bound_dev_if || 1731da177e4SLinus Torvalds sk->sk_bound_dev_if == skb->dev->ifindex)) { 174eddc9ec5SArnaldo Carvalho de Melo if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) { 1751da177e4SLinus Torvalds skb = ip_defrag(skb, IP_DEFRAG_CALL_RA_CHAIN); 1761da177e4SLinus Torvalds if (skb == NULL) { 1771da177e4SLinus Torvalds read_unlock(&ip_ra_lock); 1781da177e4SLinus Torvalds return 1; 1791da177e4SLinus Torvalds } 1801da177e4SLinus Torvalds } 1811da177e4SLinus Torvalds if (last) { 1821da177e4SLinus Torvalds struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 1831da177e4SLinus Torvalds if (skb2) 1841da177e4SLinus Torvalds raw_rcv(last, skb2); 1851da177e4SLinus Torvalds } 1861da177e4SLinus Torvalds last = sk; 1871da177e4SLinus Torvalds } 1881da177e4SLinus Torvalds } 1891da177e4SLinus Torvalds 1901da177e4SLinus Torvalds if (last) { 1911da177e4SLinus Torvalds raw_rcv(last, skb); 1921da177e4SLinus Torvalds read_unlock(&ip_ra_lock); 1931da177e4SLinus Torvalds return 1; 1941da177e4SLinus Torvalds } 1951da177e4SLinus Torvalds read_unlock(&ip_ra_lock); 1961da177e4SLinus Torvalds return 0; 1971da177e4SLinus Torvalds } 1981da177e4SLinus Torvalds 1991da177e4SLinus Torvalds static inline int ip_local_deliver_finish(struct sk_buff *skb) 2001da177e4SLinus Torvalds { 201c9bdd4b5SArnaldo Carvalho de Melo __skb_pull(skb, ip_hdrlen(skb)); 2021da177e4SLinus Torvalds 2031da177e4SLinus Torvalds /* Point into the IP datagram, just past the header. */ 204badff6d0SArnaldo Carvalho de Melo skb_reset_transport_header(skb); 2051da177e4SLinus Torvalds 2061da177e4SLinus Torvalds rcu_read_lock(); 2071da177e4SLinus Torvalds { 2081da177e4SLinus Torvalds /* Note: See raw.c and net/raw.h, RAWV4_HTABLE_SIZE==MAX_INET_PROTOS */ 209eddc9ec5SArnaldo Carvalho de Melo int protocol = ip_hdr(skb)->protocol; 2101da177e4SLinus Torvalds int hash; 2111da177e4SLinus Torvalds struct sock *raw_sk; 2121da177e4SLinus Torvalds struct net_protocol *ipprot; 2131da177e4SLinus Torvalds 2141da177e4SLinus Torvalds resubmit: 2151da177e4SLinus Torvalds hash = protocol & (MAX_INET_PROTOS - 1); 2161da177e4SLinus Torvalds raw_sk = sk_head(&raw_v4_htable[hash]); 2171da177e4SLinus Torvalds 2181da177e4SLinus Torvalds /* If there maybe a raw socket we must check - if not we 2191da177e4SLinus Torvalds * don't care less 2201da177e4SLinus Torvalds */ 221eddc9ec5SArnaldo Carvalho de Melo if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash)) 222d13964f4SPatrick McHardy raw_sk = NULL; 2231da177e4SLinus Torvalds 2241da177e4SLinus Torvalds if ((ipprot = rcu_dereference(inet_protos[hash])) != NULL) { 2251da177e4SLinus Torvalds int ret; 2261da177e4SLinus Torvalds 227b59c2701SPatrick McHardy if (!ipprot->no_policy) { 228b59c2701SPatrick McHardy if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) { 2291da177e4SLinus Torvalds kfree_skb(skb); 2301da177e4SLinus Torvalds goto out; 2311da177e4SLinus Torvalds } 232b59c2701SPatrick McHardy nf_reset(skb); 233b59c2701SPatrick McHardy } 2341da177e4SLinus Torvalds ret = ipprot->handler(skb); 2351da177e4SLinus Torvalds if (ret < 0) { 2361da177e4SLinus Torvalds protocol = -ret; 2371da177e4SLinus Torvalds goto resubmit; 2381da177e4SLinus Torvalds } 2391da177e4SLinus Torvalds IP_INC_STATS_BH(IPSTATS_MIB_INDELIVERS); 2401da177e4SLinus Torvalds } else { 2411da177e4SLinus Torvalds if (!raw_sk) { 2421da177e4SLinus Torvalds if (xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) { 2431da177e4SLinus Torvalds IP_INC_STATS_BH(IPSTATS_MIB_INUNKNOWNPROTOS); 2441da177e4SLinus Torvalds icmp_send(skb, ICMP_DEST_UNREACH, 2451da177e4SLinus Torvalds ICMP_PROT_UNREACH, 0); 2461da177e4SLinus Torvalds } 2471da177e4SLinus Torvalds } else 2481da177e4SLinus Torvalds IP_INC_STATS_BH(IPSTATS_MIB_INDELIVERS); 2491da177e4SLinus Torvalds kfree_skb(skb); 2501da177e4SLinus Torvalds } 2511da177e4SLinus Torvalds } 2521da177e4SLinus Torvalds out: 2531da177e4SLinus Torvalds rcu_read_unlock(); 2541da177e4SLinus Torvalds 2551da177e4SLinus Torvalds return 0; 2561da177e4SLinus Torvalds } 2571da177e4SLinus Torvalds 2581da177e4SLinus Torvalds /* 2591da177e4SLinus Torvalds * Deliver IP Packets to the higher protocol layers. 2601da177e4SLinus Torvalds */ 2611da177e4SLinus Torvalds int ip_local_deliver(struct sk_buff *skb) 2621da177e4SLinus Torvalds { 2631da177e4SLinus Torvalds /* 2641da177e4SLinus Torvalds * Reassemble IP fragments. 2651da177e4SLinus Torvalds */ 2661da177e4SLinus Torvalds 267eddc9ec5SArnaldo Carvalho de Melo if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) { 2681da177e4SLinus Torvalds skb = ip_defrag(skb, IP_DEFRAG_LOCAL_DELIVER); 2691da177e4SLinus Torvalds if (!skb) 2701da177e4SLinus Torvalds return 0; 2711da177e4SLinus Torvalds } 2721da177e4SLinus Torvalds 2731da177e4SLinus Torvalds return NF_HOOK(PF_INET, NF_IP_LOCAL_IN, skb, skb->dev, NULL, 2741da177e4SLinus Torvalds ip_local_deliver_finish); 2751da177e4SLinus Torvalds } 2761da177e4SLinus Torvalds 277d245407eSThomas Graf static inline int ip_rcv_options(struct sk_buff *skb) 278d245407eSThomas Graf { 279d245407eSThomas Graf struct ip_options *opt; 280d245407eSThomas Graf struct iphdr *iph; 281d245407eSThomas Graf struct net_device *dev = skb->dev; 282d245407eSThomas Graf 283d245407eSThomas Graf /* It looks as overkill, because not all 284d245407eSThomas Graf IP options require packet mangling. 285d245407eSThomas Graf But it is the easiest for now, especially taking 286d245407eSThomas Graf into account that combination of IP options 287d245407eSThomas Graf and running sniffer is extremely rare condition. 288d245407eSThomas Graf --ANK (980813) 289d245407eSThomas Graf */ 290d245407eSThomas Graf if (skb_cow(skb, skb_headroom(skb))) { 291d245407eSThomas Graf IP_INC_STATS_BH(IPSTATS_MIB_INDISCARDS); 292d245407eSThomas Graf goto drop; 293d245407eSThomas Graf } 294d245407eSThomas Graf 295eddc9ec5SArnaldo Carvalho de Melo iph = ip_hdr(skb); 296d245407eSThomas Graf 297d245407eSThomas Graf if (ip_options_compile(NULL, skb)) { 298d245407eSThomas Graf IP_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS); 299d245407eSThomas Graf goto drop; 300d245407eSThomas Graf } 301d245407eSThomas Graf 302d245407eSThomas Graf opt = &(IPCB(skb)->opt); 303d245407eSThomas Graf if (unlikely(opt->srr)) { 304d245407eSThomas Graf struct in_device *in_dev = in_dev_get(dev); 305d245407eSThomas Graf if (in_dev) { 306d245407eSThomas Graf if (!IN_DEV_SOURCE_ROUTE(in_dev)) { 307d245407eSThomas Graf if (IN_DEV_LOG_MARTIANS(in_dev) && 308d245407eSThomas Graf net_ratelimit()) 309d245407eSThomas Graf printk(KERN_INFO "source route option " 310d245407eSThomas Graf "%u.%u.%u.%u -> %u.%u.%u.%u\n", 311d245407eSThomas Graf NIPQUAD(iph->saddr), 312d245407eSThomas Graf NIPQUAD(iph->daddr)); 313d245407eSThomas Graf in_dev_put(in_dev); 314d245407eSThomas Graf goto drop; 315d245407eSThomas Graf } 316d245407eSThomas Graf 317d245407eSThomas Graf in_dev_put(in_dev); 318d245407eSThomas Graf } 319d245407eSThomas Graf 320d245407eSThomas Graf if (ip_options_rcv_srr(skb)) 321d245407eSThomas Graf goto drop; 322d245407eSThomas Graf } 323d245407eSThomas Graf 324d245407eSThomas Graf return 0; 325d245407eSThomas Graf drop: 326d245407eSThomas Graf return -1; 327d245407eSThomas Graf } 328d245407eSThomas Graf 3291da177e4SLinus Torvalds static inline int ip_rcv_finish(struct sk_buff *skb) 3301da177e4SLinus Torvalds { 331eddc9ec5SArnaldo Carvalho de Melo const struct iphdr *iph = ip_hdr(skb); 3321da177e4SLinus Torvalds 3331da177e4SLinus Torvalds /* 3341da177e4SLinus Torvalds * Initialise the virtual path cache for the packet. It describes 3351da177e4SLinus Torvalds * how the packet travels inside Linux networking. 3361da177e4SLinus Torvalds */ 3370182bd2bSHua Zhong if (skb->dst == NULL) { 3383e192beaSThomas Graf int err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, 3393e192beaSThomas Graf skb->dev); 3403e192beaSThomas Graf if (unlikely(err)) { 3412c2910a4SDietmar Eggemann if (err == -EHOSTUNREACH) 3422c2910a4SDietmar Eggemann IP_INC_STATS_BH(IPSTATS_MIB_INADDRERRORS); 343e91a47ebSMitsuru Chinen else if (err == -ENETUNREACH) 344e91a47ebSMitsuru Chinen IP_INC_STATS_BH(IPSTATS_MIB_INNOROUTES); 3451da177e4SLinus Torvalds goto drop; 3461da177e4SLinus Torvalds } 3472c2910a4SDietmar Eggemann } 3481da177e4SLinus Torvalds 3491da177e4SLinus Torvalds #ifdef CONFIG_NET_CLS_ROUTE 3503e192beaSThomas Graf if (unlikely(skb->dst->tclassid)) { 3511da177e4SLinus Torvalds struct ip_rt_acct *st = ip_rt_acct + 256*smp_processor_id(); 3521da177e4SLinus Torvalds u32 idx = skb->dst->tclassid; 3531da177e4SLinus Torvalds st[idx&0xFF].o_packets++; 3541da177e4SLinus Torvalds st[idx&0xFF].o_bytes+=skb->len; 3551da177e4SLinus Torvalds st[(idx>>16)&0xFF].i_packets++; 3561da177e4SLinus Torvalds st[(idx>>16)&0xFF].i_bytes+=skb->len; 3571da177e4SLinus Torvalds } 3581da177e4SLinus Torvalds #endif 3591da177e4SLinus Torvalds 360d245407eSThomas Graf if (iph->ihl > 5 && ip_rcv_options(skb)) 3611da177e4SLinus Torvalds goto drop; 3621da177e4SLinus Torvalds 3631da177e4SLinus Torvalds return dst_input(skb); 3641da177e4SLinus Torvalds 3651da177e4SLinus Torvalds drop: 3661da177e4SLinus Torvalds kfree_skb(skb); 3671da177e4SLinus Torvalds return NET_RX_DROP; 3681da177e4SLinus Torvalds } 3691da177e4SLinus Torvalds 3701da177e4SLinus Torvalds /* 3711da177e4SLinus Torvalds * Main IP Receive routine. 3721da177e4SLinus Torvalds */ 373f2ccd8faSDavid S. Miller int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev) 3741da177e4SLinus Torvalds { 3751da177e4SLinus Torvalds struct iphdr *iph; 37658615242SThomas Graf u32 len; 3771da177e4SLinus Torvalds 3781da177e4SLinus Torvalds /* When the interface is in promisc. mode, drop all the crap 3791da177e4SLinus Torvalds * that it receives, do not try to analyse it. 3801da177e4SLinus Torvalds */ 3811da177e4SLinus Torvalds if (skb->pkt_type == PACKET_OTHERHOST) 3821da177e4SLinus Torvalds goto drop; 3831da177e4SLinus Torvalds 3841da177e4SLinus Torvalds IP_INC_STATS_BH(IPSTATS_MIB_INRECEIVES); 3851da177e4SLinus Torvalds 3861da177e4SLinus Torvalds if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL) { 3871da177e4SLinus Torvalds IP_INC_STATS_BH(IPSTATS_MIB_INDISCARDS); 3881da177e4SLinus Torvalds goto out; 3891da177e4SLinus Torvalds } 3901da177e4SLinus Torvalds 3911da177e4SLinus Torvalds if (!pskb_may_pull(skb, sizeof(struct iphdr))) 3921da177e4SLinus Torvalds goto inhdr_error; 3931da177e4SLinus Torvalds 394eddc9ec5SArnaldo Carvalho de Melo iph = ip_hdr(skb); 3951da177e4SLinus Torvalds 3961da177e4SLinus Torvalds /* 3971da177e4SLinus Torvalds * RFC1122: 3.1.2.2 MUST silently discard any IP frame that fails the checksum. 3981da177e4SLinus Torvalds * 3991da177e4SLinus Torvalds * Is the datagram acceptable? 4001da177e4SLinus Torvalds * 4011da177e4SLinus Torvalds * 1. Length at least the size of an ip header 4021da177e4SLinus Torvalds * 2. Version of 4 4031da177e4SLinus Torvalds * 3. Checksums correctly. [Speed optimisation for later, skip loopback checksums] 4041da177e4SLinus Torvalds * 4. Doesn't have a bogus length 4051da177e4SLinus Torvalds */ 4061da177e4SLinus Torvalds 4071da177e4SLinus Torvalds if (iph->ihl < 5 || iph->version != 4) 4081da177e4SLinus Torvalds goto inhdr_error; 4091da177e4SLinus Torvalds 4101da177e4SLinus Torvalds if (!pskb_may_pull(skb, iph->ihl*4)) 4111da177e4SLinus Torvalds goto inhdr_error; 4121da177e4SLinus Torvalds 413eddc9ec5SArnaldo Carvalho de Melo iph = ip_hdr(skb); 4141da177e4SLinus Torvalds 415e9c60422SThomas Graf if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl))) 4161da177e4SLinus Torvalds goto inhdr_error; 4171da177e4SLinus Torvalds 41858615242SThomas Graf len = ntohs(iph->tot_len); 41958615242SThomas Graf if (skb->len < len || len < (iph->ihl*4)) 4201da177e4SLinus Torvalds goto inhdr_error; 4211da177e4SLinus Torvalds 4221da177e4SLinus Torvalds /* Our transport medium may have padded the buffer out. Now we know it 4231da177e4SLinus Torvalds * is IP we can trim to the true length of the frame. 4241da177e4SLinus Torvalds * Note this now means skb->len holds ntohs(iph->tot_len). 4251da177e4SLinus Torvalds */ 4261da177e4SLinus Torvalds if (pskb_trim_rcsum(skb, len)) { 4271da177e4SLinus Torvalds IP_INC_STATS_BH(IPSTATS_MIB_INDISCARDS); 4281da177e4SLinus Torvalds goto drop; 4291da177e4SLinus Torvalds } 4301da177e4SLinus Torvalds 43153602f92SStephen Hemminger /* Remove any debris in the socket control block */ 432d569f1d7SGuillaume Chazarain memset(IPCB(skb), 0, sizeof(struct inet_skb_parm)); 43353602f92SStephen Hemminger 4341da177e4SLinus Torvalds return NF_HOOK(PF_INET, NF_IP_PRE_ROUTING, skb, dev, NULL, 4351da177e4SLinus Torvalds ip_rcv_finish); 4361da177e4SLinus Torvalds 4371da177e4SLinus Torvalds inhdr_error: 4381da177e4SLinus Torvalds IP_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS); 4391da177e4SLinus Torvalds drop: 4401da177e4SLinus Torvalds kfree_skb(skb); 4411da177e4SLinus Torvalds out: 4421da177e4SLinus Torvalds return NET_RX_DROP; 4431da177e4SLinus Torvalds } 4441da177e4SLinus Torvalds 4451da177e4SLinus Torvalds EXPORT_SYMBOL(ip_statistics); 446