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 * Definitions for the IP router. 71da177e4SLinus Torvalds * 81da177e4SLinus Torvalds * Version: @(#)route.h 1.0.4 05/27/93 91da177e4SLinus Torvalds * 1002c30a84SJesper Juhl * Authors: Ross Biro 111da177e4SLinus Torvalds * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 121da177e4SLinus Torvalds * Fixes: 131da177e4SLinus Torvalds * Alan Cox : Reformatted. Added ip_rt_local() 141da177e4SLinus Torvalds * Alan Cox : Support for TCP parameters. 151da177e4SLinus Torvalds * Alexey Kuznetsov: Major changes for new routing code. 161da177e4SLinus Torvalds * Mike McLagan : Routing by source 171da177e4SLinus Torvalds * Robert Olsson : Added rt_cache statistics 181da177e4SLinus Torvalds * 191da177e4SLinus Torvalds * This program is free software; you can redistribute it and/or 201da177e4SLinus Torvalds * modify it under the terms of the GNU General Public License 211da177e4SLinus Torvalds * as published by the Free Software Foundation; either version 221da177e4SLinus Torvalds * 2 of the License, or (at your option) any later version. 231da177e4SLinus Torvalds */ 241da177e4SLinus Torvalds #ifndef _ROUTE_H 251da177e4SLinus Torvalds #define _ROUTE_H 261da177e4SLinus Torvalds 271da177e4SLinus Torvalds #include <net/dst.h> 281da177e4SLinus Torvalds #include <net/inetpeer.h> 291da177e4SLinus Torvalds #include <net/flow.h> 3079876874SKOVACS Krisztian #include <net/inet_sock.h> 311da177e4SLinus Torvalds #include <linux/in_route.h> 321da177e4SLinus Torvalds #include <linux/rtnetlink.h> 331da177e4SLinus Torvalds #include <linux/route.h> 341da177e4SLinus Torvalds #include <linux/ip.h> 351da177e4SLinus Torvalds #include <linux/cache.h> 36beb8d13bSVenkat Yekkirala #include <linux/security.h> 371da177e4SLinus Torvalds 381da177e4SLinus Torvalds #define RTO_ONLINK 0x01 391da177e4SLinus Torvalds 401da177e4SLinus Torvalds #define RT_CONN_FLAGS(sk) (RT_TOS(inet_sk(sk)->tos) | sock_flag(sk, SOCK_LOCALROUTE)) 411da177e4SLinus Torvalds 421da177e4SLinus Torvalds struct fib_nh; 4362fa8a84SDavid S. Miller struct fib_info; 44fd2c3ef7SEric Dumazet struct rtable { 451da177e4SLinus Torvalds struct dst_entry dst; 461da177e4SLinus Torvalds 4729e75252SEric Dumazet int rt_genid; 4895c96174SEric Dumazet unsigned int rt_flags; 491da177e4SLinus Torvalds __u16 rt_type; 501da177e4SLinus Torvalds 511b86a58fSOGAWA Hirofumi int rt_route_iif; 521da177e4SLinus Torvalds int rt_iif; 535e2b61f7SDavid S. Miller int rt_oif; 541da177e4SLinus Torvalds 551da177e4SLinus Torvalds /* Info on neighbour */ 56f2c3fe24SAl Viro __be32 rt_gateway; 571da177e4SLinus Torvalds 581da177e4SLinus Torvalds /* Miscellaneous cached information */ 595943634fSDavid S. Miller u32 rt_pmtu; 6062fa8a84SDavid S. Miller struct fib_info *fi; /* for client ref to shared metrics */ 611da177e4SLinus Torvalds }; 621da177e4SLinus Torvalds 63b8400f37SSteffen Klassert static inline bool rt_is_input_route(const struct rtable *rt) 64c7537967SDavid S. Miller { 651b86a58fSOGAWA Hirofumi return rt->rt_route_iif != 0; 66c7537967SDavid S. Miller } 67c7537967SDavid S. Miller 68b8400f37SSteffen Klassert static inline bool rt_is_output_route(const struct rtable *rt) 69c7537967SDavid S. Miller { 701b86a58fSOGAWA Hirofumi return rt->rt_route_iif == 0; 71c7537967SDavid S. Miller } 72c7537967SDavid S. Miller 73f8126f1dSDavid S. Miller static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr) 74f8126f1dSDavid S. Miller { 75f8126f1dSDavid S. Miller if (rt->rt_gateway) 76f8126f1dSDavid S. Miller return rt->rt_gateway; 77f8126f1dSDavid S. Miller return daddr; 78f8126f1dSDavid S. Miller } 79f8126f1dSDavid S. Miller 80fd2c3ef7SEric Dumazet struct ip_rt_acct { 811da177e4SLinus Torvalds __u32 o_bytes; 821da177e4SLinus Torvalds __u32 o_packets; 831da177e4SLinus Torvalds __u32 i_bytes; 841da177e4SLinus Torvalds __u32 i_packets; 851da177e4SLinus Torvalds }; 861da177e4SLinus Torvalds 87fd2c3ef7SEric Dumazet struct rt_cache_stat { 881da177e4SLinus Torvalds unsigned int in_hit; 891da177e4SLinus Torvalds unsigned int in_slow_tot; 901da177e4SLinus Torvalds unsigned int in_slow_mc; 911da177e4SLinus Torvalds unsigned int in_no_route; 921da177e4SLinus Torvalds unsigned int in_brd; 931da177e4SLinus Torvalds unsigned int in_martian_dst; 941da177e4SLinus Torvalds unsigned int in_martian_src; 951da177e4SLinus Torvalds unsigned int out_hit; 961da177e4SLinus Torvalds unsigned int out_slow_tot; 971da177e4SLinus Torvalds unsigned int out_slow_mc; 981da177e4SLinus Torvalds unsigned int gc_total; 991da177e4SLinus Torvalds unsigned int gc_ignored; 1001da177e4SLinus Torvalds unsigned int gc_goal_miss; 1011da177e4SLinus Torvalds unsigned int gc_dst_overflow; 1021da177e4SLinus Torvalds unsigned int in_hlist_search; 1031da177e4SLinus Torvalds unsigned int out_hlist_search; 1041da177e4SLinus Torvalds }; 1051da177e4SLinus Torvalds 1067d720c3eSTejun Heo extern struct ip_rt_acct __percpu *ip_rt_acct; 1071da177e4SLinus Torvalds 1081da177e4SLinus Torvalds struct in_device; 1091da177e4SLinus Torvalds extern int ip_rt_init(void); 11076e6ebfbSDenis V. Lunev extern void rt_cache_flush(struct net *net, int how); 111813b3b5dSDavid S. Miller extern struct rtable *__ip_route_output_key(struct net *, struct flowi4 *flp); 1129d6ec938SDavid S. Miller extern struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp, 113b23dd4feSDavid S. Miller struct sock *sk); 1142774c131SDavid S. Miller extern struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig); 115407eadd9SEric Dumazet 1169d6ec938SDavid S. Miller static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp) 1175bfa787fSDavid S. Miller { 1185bfa787fSDavid S. Miller return ip_route_output_flow(net, flp, NULL); 1195bfa787fSDavid S. Miller } 1205bfa787fSDavid S. Miller 12178fbfd8aSDavid S. Miller static inline struct rtable *ip_route_output(struct net *net, __be32 daddr, 12278fbfd8aSDavid S. Miller __be32 saddr, u8 tos, int oif) 12378fbfd8aSDavid S. Miller { 1249d6ec938SDavid S. Miller struct flowi4 fl4 = { 1259d6ec938SDavid S. Miller .flowi4_oif = oif, 126c5d21c4bSRoland Dreier .flowi4_tos = tos, 1279d6ec938SDavid S. Miller .daddr = daddr, 1289d6ec938SDavid S. Miller .saddr = saddr, 12978fbfd8aSDavid S. Miller }; 1309d6ec938SDavid S. Miller return ip_route_output_key(net, &fl4); 13178fbfd8aSDavid S. Miller } 13278fbfd8aSDavid S. Miller 13331e4543dSDavid S. Miller static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4, 13431e4543dSDavid S. Miller struct sock *sk, 13578fbfd8aSDavid S. Miller __be32 daddr, __be32 saddr, 13678fbfd8aSDavid S. Miller __be16 dport, __be16 sport, 13778fbfd8aSDavid S. Miller __u8 proto, __u8 tos, int oif) 13878fbfd8aSDavid S. Miller { 13931e4543dSDavid S. Miller flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos, 14094b92b88SDavid S. Miller RT_SCOPE_UNIVERSE, proto, 14194b92b88SDavid S. Miller sk ? inet_sk_flowi_flags(sk) : 0, 14294b92b88SDavid S. Miller daddr, saddr, dport, sport); 14378fbfd8aSDavid S. Miller if (sk) 14431e4543dSDavid S. Miller security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 14531e4543dSDavid S. Miller return ip_route_output_flow(net, fl4, sk); 14678fbfd8aSDavid S. Miller } 14778fbfd8aSDavid S. Miller 148cbb1e85fSDavid S. Miller static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4, 14978fbfd8aSDavid S. Miller __be32 daddr, __be32 saddr, 15078fbfd8aSDavid S. Miller __be32 gre_key, __u8 tos, int oif) 15178fbfd8aSDavid S. Miller { 152cbb1e85fSDavid S. Miller memset(fl4, 0, sizeof(*fl4)); 153cbb1e85fSDavid S. Miller fl4->flowi4_oif = oif; 154cbb1e85fSDavid S. Miller fl4->daddr = daddr; 155cbb1e85fSDavid S. Miller fl4->saddr = saddr; 156cbb1e85fSDavid S. Miller fl4->flowi4_tos = tos; 157cbb1e85fSDavid S. Miller fl4->flowi4_proto = IPPROTO_GRE; 158cbb1e85fSDavid S. Miller fl4->fl4_gre_key = gre_key; 159cbb1e85fSDavid S. Miller return ip_route_output_key(net, fl4); 16078fbfd8aSDavid S. Miller } 16178fbfd8aSDavid S. Miller 16238a424e4SDavid Miller extern int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src, 16338a424e4SDavid Miller u8 tos, struct net_device *devin); 164407eadd9SEric Dumazet 16536393395SDavid S. Miller extern void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, 16636393395SDavid S. Miller int oif, u32 mark, u8 protocol, int flow_flags); 16736393395SDavid S. Miller extern void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu); 168b42597e2SDavid S. Miller extern void ipv4_redirect(struct sk_buff *skb, struct net *net, 169b42597e2SDavid S. Miller int oif, u32 mark, u8 protocol, int flow_flags); 170b42597e2SDavid S. Miller extern void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk); 1711da177e4SLinus Torvalds extern void ip_rt_send_redirect(struct sk_buff *skb); 1721da177e4SLinus Torvalds 17395c96174SEric Dumazet extern unsigned int inet_addr_type(struct net *net, __be32 addr); 17495c96174SEric Dumazet extern unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev, __be32 addr); 1751da177e4SLinus Torvalds extern void ip_rt_multicast_event(struct in_device *); 1761bad118aSDenis V. Lunev extern int ip_rt_ioctl(struct net *, unsigned int cmd, void __user *arg); 1778e36360aSDavid S. Miller extern void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt); 1781da177e4SLinus Torvalds extern int ip_rt_dump(struct sk_buff *skb, struct netlink_callback *cb); 1791da177e4SLinus Torvalds 1800ff60a45SJamal Hadi Salim struct in_ifaddr; 1810ff60a45SJamal Hadi Salim extern void fib_add_ifaddr(struct in_ifaddr *); 182e6abbaa2SJulian Anastasov extern void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *); 1830ff60a45SJamal Hadi Salim 1841da177e4SLinus Torvalds static inline void ip_rt_put(struct rtable * rt) 1851da177e4SLinus Torvalds { 1861da177e4SLinus Torvalds if (rt) 187d8d1f30bSChangli Gao dst_release(&rt->dst); 1881da177e4SLinus Torvalds } 1891da177e4SLinus Torvalds 1901da177e4SLinus Torvalds #define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3) 1911da177e4SLinus Torvalds 1924839c52bSPhilippe De Muyter extern const __u8 ip_tos2prio[16]; 1931da177e4SLinus Torvalds 1941da177e4SLinus Torvalds static inline char rt_tos2priority(u8 tos) 1951da177e4SLinus Torvalds { 1961da177e4SLinus Torvalds return ip_tos2prio[IPTOS_TOS(tos)>>1]; 1971da177e4SLinus Torvalds } 1981da177e4SLinus Torvalds 1992d7192d6SDavid S. Miller /* ip_route_connect() and ip_route_newports() work in tandem whilst 2002d7192d6SDavid S. Miller * binding a socket for a new outgoing connection. 2012d7192d6SDavid S. Miller * 2022d7192d6SDavid S. Miller * In order to use IPSEC properly, we must, in the end, have a 2032d7192d6SDavid S. Miller * route that was looked up using all available keys including source 2042d7192d6SDavid S. Miller * and destination ports. 2052d7192d6SDavid S. Miller * 2062d7192d6SDavid S. Miller * However, if a source port needs to be allocated (the user specified 2072d7192d6SDavid S. Miller * a wildcard source port) we need to obtain addressing information 2082d7192d6SDavid S. Miller * in order to perform that allocation. 2092d7192d6SDavid S. Miller * 2102d7192d6SDavid S. Miller * So ip_route_connect() looks up a route using wildcarded source and 2112d7192d6SDavid S. Miller * destination ports in the key, simply so that we can get a pair of 2122d7192d6SDavid S. Miller * addresses to use for port allocation. 2132d7192d6SDavid S. Miller * 2142d7192d6SDavid S. Miller * Later, once the ports are allocated, ip_route_newports() will make 2152d7192d6SDavid S. Miller * another route lookup if needed to make sure we catch any IPSEC 2162d7192d6SDavid S. Miller * rules keyed on the port information. 2172d7192d6SDavid S. Miller * 2182d7192d6SDavid S. Miller * The callers allocate the flow key on their stack, and must pass in 2192d7192d6SDavid S. Miller * the same flowi4 object to both the ip_route_connect() and the 2202d7192d6SDavid S. Miller * ip_route_newports() calls. 2212d7192d6SDavid S. Miller */ 2222d7192d6SDavid S. Miller 2232d7192d6SDavid S. Miller static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src, 2242d7192d6SDavid S. Miller u32 tos, int oif, u8 protocol, 225b23dd4feSDavid S. Miller __be16 sport, __be16 dport, 226b23dd4feSDavid S. Miller struct sock *sk, bool can_sleep) 2271da177e4SLinus Torvalds { 2282d7192d6SDavid S. Miller __u8 flow_flags = 0; 22979876874SKOVACS Krisztian 23079876874SKOVACS Krisztian if (inet_sk(sk)->transparent) 23194b92b88SDavid S. Miller flow_flags |= FLOWI_FLAG_ANYSRC; 2325df65e55SDavid S. Miller if (can_sleep) 23394b92b88SDavid S. Miller flow_flags |= FLOWI_FLAG_CAN_SLEEP; 23494b92b88SDavid S. Miller 2352d7192d6SDavid S. Miller flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE, 23694b92b88SDavid S. Miller protocol, flow_flags, dst, src, dport, sport); 2372d7192d6SDavid S. Miller } 2382d7192d6SDavid S. Miller 2392d7192d6SDavid S. Miller static inline struct rtable *ip_route_connect(struct flowi4 *fl4, 2402d7192d6SDavid S. Miller __be32 dst, __be32 src, u32 tos, 2412d7192d6SDavid S. Miller int oif, u8 protocol, 2422d7192d6SDavid S. Miller __be16 sport, __be16 dport, 2432d7192d6SDavid S. Miller struct sock *sk, bool can_sleep) 2442d7192d6SDavid S. Miller { 2452d7192d6SDavid S. Miller struct net *net = sock_net(sk); 2462d7192d6SDavid S. Miller struct rtable *rt; 2472d7192d6SDavid S. Miller 2482d7192d6SDavid S. Miller ip_route_connect_init(fl4, dst, src, tos, oif, protocol, 2492d7192d6SDavid S. Miller sport, dport, sk, can_sleep); 25079876874SKOVACS Krisztian 2511da177e4SLinus Torvalds if (!dst || !src) { 2522d7192d6SDavid S. Miller rt = __ip_route_output_key(net, fl4); 253b23dd4feSDavid S. Miller if (IS_ERR(rt)) 254b23dd4feSDavid S. Miller return rt; 255b23dd4feSDavid S. Miller ip_rt_put(rt); 256e6b45241SJulian Anastasov flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr); 2571da177e4SLinus Torvalds } 2582d7192d6SDavid S. Miller security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 2592d7192d6SDavid S. Miller return ip_route_output_flow(net, fl4, sk); 2601da177e4SLinus Torvalds } 2611da177e4SLinus Torvalds 2622d7192d6SDavid S. Miller static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt, 2632d7192d6SDavid S. Miller __be16 orig_sport, __be16 orig_dport, 2642d7192d6SDavid S. Miller __be16 sport, __be16 dport, 2652d7192d6SDavid S. Miller struct sock *sk) 2661da177e4SLinus Torvalds { 267dca8b089SDavid S. Miller if (sport != orig_sport || dport != orig_dport) { 2682d7192d6SDavid S. Miller fl4->fl4_dport = dport; 2692d7192d6SDavid S. Miller fl4->fl4_sport = sport; 270b23dd4feSDavid S. Miller ip_rt_put(rt); 271e6b45241SJulian Anastasov flowi4_update_output(fl4, sk->sk_bound_dev_if, 272e6b45241SJulian Anastasov RT_CONN_FLAGS(sk), fl4->daddr, 273e6b45241SJulian Anastasov fl4->saddr); 2742d7192d6SDavid S. Miller security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 2752d7192d6SDavid S. Miller return ip_route_output_flow(sock_net(sk), fl4, sk); 2761da177e4SLinus Torvalds } 277b23dd4feSDavid S. Miller return rt; 2781da177e4SLinus Torvalds } 2791da177e4SLinus Torvalds 2801668e010SKOVACS Krisztian static inline int inet_iif(const struct sk_buff *skb) 2811668e010SKOVACS Krisztian { 282511c3f92SEric Dumazet return skb_rtable(skb)->rt_iif; 2831668e010SKOVACS Krisztian } 2841668e010SKOVACS Krisztian 285323e126fSDavid S. Miller extern int sysctl_ip_default_ttl; 286323e126fSDavid S. Miller 287323e126fSDavid S. Miller static inline int ip4_dst_hoplimit(const struct dst_entry *dst) 288323e126fSDavid S. Miller { 289323e126fSDavid S. Miller int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT); 290323e126fSDavid S. Miller 291323e126fSDavid S. Miller if (hoplimit == 0) 292323e126fSDavid S. Miller hoplimit = sysctl_ip_default_ttl; 293323e126fSDavid S. Miller return hoplimit; 294323e126fSDavid S. Miller } 295323e126fSDavid S. Miller 2961da177e4SLinus Torvalds #endif /* _ROUTE_H */ 297