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> 3179a13159SPeter Nørlund #include <net/ip_fib.h> 329478d12dSDavid Ahern #include <net/l3mdev.h> 331da177e4SLinus Torvalds #include <linux/in_route.h> 341da177e4SLinus Torvalds #include <linux/rtnetlink.h> 35c6cffba4SDavid S. Miller #include <linux/rcupdate.h> 361da177e4SLinus Torvalds #include <linux/route.h> 371da177e4SLinus Torvalds #include <linux/ip.h> 381da177e4SLinus Torvalds #include <linux/cache.h> 39beb8d13bSVenkat Yekkirala #include <linux/security.h> 401da177e4SLinus Torvalds 41f87c10a8SHannes Frederic Sowa /* IPv4 datagram length is stored into 16bit field (tot_len) */ 42f87c10a8SHannes Frederic Sowa #define IP_MAX_MTU 0xFFFFU 43f87c10a8SHannes Frederic Sowa 441da177e4SLinus Torvalds #define RTO_ONLINK 0x01 451da177e4SLinus Torvalds 461da177e4SLinus Torvalds #define RT_CONN_FLAGS(sk) (RT_TOS(inet_sk(sk)->tos) | sock_flag(sk, SOCK_LOCALROUTE)) 47aa661581SFrancesco Fusco #define RT_CONN_FLAGS_TOS(sk,tos) (RT_TOS(tos) | sock_flag(sk, SOCK_LOCALROUTE)) 481da177e4SLinus Torvalds 491da177e4SLinus Torvalds struct fib_nh; 5062fa8a84SDavid S. Miller struct fib_info; 515055c371SEric Dumazet struct uncached_list; 52fd2c3ef7SEric Dumazet struct rtable { 531da177e4SLinus Torvalds struct dst_entry dst; 541da177e4SLinus Torvalds 5529e75252SEric Dumazet int rt_genid; 5695c96174SEric Dumazet unsigned int rt_flags; 571da177e4SLinus Torvalds __u16 rt_type; 58155e8336SJulian Anastasov __u8 rt_is_input; 59155e8336SJulian Anastasov __u8 rt_uses_gateway; 601da177e4SLinus Torvalds 611da177e4SLinus Torvalds int rt_iif; 621da177e4SLinus Torvalds 631da177e4SLinus Torvalds /* Info on neighbour */ 64f2c3fe24SAl Viro __be32 rt_gateway; 651da177e4SLinus Torvalds 661da177e4SLinus Torvalds /* Miscellaneous cached information */ 675943634fSDavid S. Miller u32 rt_pmtu; 68caacf05eSDavid S. Miller 69b7503e0cSDavid Ahern u32 rt_table_id; 70b7503e0cSDavid Ahern 71caacf05eSDavid S. Miller struct list_head rt_uncached; 725055c371SEric Dumazet struct uncached_list *rt_uncached_list; 731da177e4SLinus Torvalds }; 741da177e4SLinus Torvalds 75b8400f37SSteffen Klassert static inline bool rt_is_input_route(const struct rtable *rt) 76c7537967SDavid S. Miller { 779917e1e8SDavid S. Miller return rt->rt_is_input != 0; 78c7537967SDavid S. Miller } 79c7537967SDavid S. Miller 80b8400f37SSteffen Klassert static inline bool rt_is_output_route(const struct rtable *rt) 81c7537967SDavid S. Miller { 829917e1e8SDavid S. Miller return rt->rt_is_input == 0; 83c7537967SDavid S. Miller } 84c7537967SDavid S. Miller 85f8126f1dSDavid S. Miller static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr) 86f8126f1dSDavid S. Miller { 87f8126f1dSDavid S. Miller if (rt->rt_gateway) 88f8126f1dSDavid S. Miller return rt->rt_gateway; 89f8126f1dSDavid S. Miller return daddr; 90f8126f1dSDavid S. Miller } 91f8126f1dSDavid S. Miller 92fd2c3ef7SEric Dumazet struct ip_rt_acct { 931da177e4SLinus Torvalds __u32 o_bytes; 941da177e4SLinus Torvalds __u32 o_packets; 951da177e4SLinus Torvalds __u32 i_bytes; 961da177e4SLinus Torvalds __u32 i_packets; 971da177e4SLinus Torvalds }; 981da177e4SLinus Torvalds 99fd2c3ef7SEric Dumazet struct rt_cache_stat { 1001da177e4SLinus Torvalds unsigned int in_slow_tot; 1011da177e4SLinus Torvalds unsigned int in_slow_mc; 1021da177e4SLinus Torvalds unsigned int in_no_route; 1031da177e4SLinus Torvalds unsigned int in_brd; 1041da177e4SLinus Torvalds unsigned int in_martian_dst; 1051da177e4SLinus Torvalds unsigned int in_martian_src; 1061da177e4SLinus Torvalds unsigned int out_slow_tot; 1071da177e4SLinus Torvalds unsigned int out_slow_mc; 1081da177e4SLinus Torvalds }; 1091da177e4SLinus Torvalds 1107d720c3eSTejun Heo extern struct ip_rt_acct __percpu *ip_rt_acct; 1111da177e4SLinus Torvalds 1121da177e4SLinus Torvalds struct in_device; 1132702c4bbSJoe Perches 1142702c4bbSJoe Perches int ip_rt_init(void); 1152702c4bbSJoe Perches void rt_cache_flush(struct net *net); 1162702c4bbSJoe Perches void rt_flush_dev(struct net_device *dev); 11779a13159SPeter Nørlund struct rtable *__ip_route_output_key_hash(struct net *, struct flowi4 *flp, 11879a13159SPeter Nørlund int mp_hash); 11979a13159SPeter Nørlund 12079a13159SPeter Nørlund static inline struct rtable *__ip_route_output_key(struct net *net, 12179a13159SPeter Nørlund struct flowi4 *flp) 12279a13159SPeter Nørlund { 12379a13159SPeter Nørlund return __ip_route_output_key_hash(net, flp, -1); 12479a13159SPeter Nørlund } 12579a13159SPeter Nørlund 1262702c4bbSJoe Perches struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp, 1276f9c9615SEric Dumazet const struct sock *sk); 1282702c4bbSJoe Perches struct dst_entry *ipv4_blackhole_route(struct net *net, 1292702c4bbSJoe Perches struct dst_entry *dst_orig); 130407eadd9SEric Dumazet 1319d6ec938SDavid S. Miller static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp) 1325bfa787fSDavid S. Miller { 1335bfa787fSDavid S. Miller return ip_route_output_flow(net, flp, NULL); 1345bfa787fSDavid S. Miller } 1355bfa787fSDavid S. Miller 13678fbfd8aSDavid S. Miller static inline struct rtable *ip_route_output(struct net *net, __be32 daddr, 13778fbfd8aSDavid S. Miller __be32 saddr, u8 tos, int oif) 13878fbfd8aSDavid S. Miller { 1399d6ec938SDavid S. Miller struct flowi4 fl4 = { 1409d6ec938SDavid S. Miller .flowi4_oif = oif, 141c5d21c4bSRoland Dreier .flowi4_tos = tos, 1429d6ec938SDavid S. Miller .daddr = daddr, 1439d6ec938SDavid S. Miller .saddr = saddr, 14478fbfd8aSDavid S. Miller }; 1459d6ec938SDavid S. Miller return ip_route_output_key(net, &fl4); 14678fbfd8aSDavid S. Miller } 14778fbfd8aSDavid S. Miller 14831e4543dSDavid S. Miller static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4, 14931e4543dSDavid S. Miller struct sock *sk, 15078fbfd8aSDavid S. Miller __be32 daddr, __be32 saddr, 15178fbfd8aSDavid S. Miller __be16 dport, __be16 sport, 15278fbfd8aSDavid S. Miller __u8 proto, __u8 tos, int oif) 15378fbfd8aSDavid S. Miller { 15431e4543dSDavid S. Miller flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos, 15594b92b88SDavid S. Miller RT_SCOPE_UNIVERSE, proto, 15694b92b88SDavid S. Miller sk ? inet_sk_flowi_flags(sk) : 0, 15794b92b88SDavid S. Miller daddr, saddr, dport, sport); 15878fbfd8aSDavid S. Miller if (sk) 15931e4543dSDavid S. Miller security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 16031e4543dSDavid S. Miller return ip_route_output_flow(net, fl4, sk); 16178fbfd8aSDavid S. Miller } 16278fbfd8aSDavid S. Miller 163cbb1e85fSDavid S. Miller static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4, 16478fbfd8aSDavid S. Miller __be32 daddr, __be32 saddr, 16578fbfd8aSDavid S. Miller __be32 gre_key, __u8 tos, int oif) 16678fbfd8aSDavid S. Miller { 167cbb1e85fSDavid S. Miller memset(fl4, 0, sizeof(*fl4)); 168cbb1e85fSDavid S. Miller fl4->flowi4_oif = oif; 169cbb1e85fSDavid S. Miller fl4->daddr = daddr; 170cbb1e85fSDavid S. Miller fl4->saddr = saddr; 171cbb1e85fSDavid S. Miller fl4->flowi4_tos = tos; 172cbb1e85fSDavid S. Miller fl4->flowi4_proto = IPPROTO_GRE; 173cbb1e85fSDavid S. Miller fl4->fl4_gre_key = gre_key; 174cbb1e85fSDavid S. Miller return ip_route_output_key(net, fl4); 17578fbfd8aSDavid S. Miller } 17678fbfd8aSDavid S. Miller 1772702c4bbSJoe Perches int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src, 17838a424e4SDavid Miller u8 tos, struct net_device *devin); 179407eadd9SEric Dumazet 180c6cffba4SDavid S. Miller static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src, 181c6cffba4SDavid S. Miller u8 tos, struct net_device *devin) 182c6cffba4SDavid S. Miller { 183c6cffba4SDavid S. Miller int err; 184c6cffba4SDavid S. Miller 185c6cffba4SDavid S. Miller rcu_read_lock(); 186c6cffba4SDavid S. Miller err = ip_route_input_noref(skb, dst, src, tos, devin); 187c6cffba4SDavid S. Miller if (!err) 188c6cffba4SDavid S. Miller skb_dst_force(skb); 189c6cffba4SDavid S. Miller rcu_read_unlock(); 190c6cffba4SDavid S. Miller 191c6cffba4SDavid S. Miller return err; 192c6cffba4SDavid S. Miller } 193c6cffba4SDavid S. Miller 1942702c4bbSJoe Perches void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif, 1952702c4bbSJoe Perches u32 mark, u8 protocol, int flow_flags); 1962702c4bbSJoe Perches void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu); 1972702c4bbSJoe Perches void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark, 1982702c4bbSJoe Perches u8 protocol, int flow_flags); 1992702c4bbSJoe Perches void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk); 2002702c4bbSJoe Perches void ip_rt_send_redirect(struct sk_buff *skb); 2011da177e4SLinus Torvalds 2022702c4bbSJoe Perches unsigned int inet_addr_type(struct net *net, __be32 addr); 2039b8ff518SDavid Ahern unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id); 2042702c4bbSJoe Perches unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev, 2052702c4bbSJoe Perches __be32 addr); 20630bbaa19SDavid Ahern unsigned int inet_addr_type_dev_table(struct net *net, 20730bbaa19SDavid Ahern const struct net_device *dev, 20830bbaa19SDavid Ahern __be32 addr); 2092702c4bbSJoe Perches void ip_rt_multicast_event(struct in_device *); 2102702c4bbSJoe Perches int ip_rt_ioctl(struct net *, unsigned int cmd, void __user *arg); 2112702c4bbSJoe Perches void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt); 2121da177e4SLinus Torvalds 2130ff60a45SJamal Hadi Salim struct in_ifaddr; 2142702c4bbSJoe Perches void fib_add_ifaddr(struct in_ifaddr *); 2152702c4bbSJoe Perches void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *); 2160ff60a45SJamal Hadi Salim 2171da177e4SLinus Torvalds static inline void ip_rt_put(struct rtable *rt) 2181da177e4SLinus Torvalds { 2196da025faSEric Dumazet /* dst_release() accepts a NULL parameter. 2206da025faSEric Dumazet * We rely on dst being first structure in struct rtable 2216da025faSEric Dumazet */ 2226da025faSEric Dumazet BUILD_BUG_ON(offsetof(struct rtable, dst) != 0); 223d8d1f30bSChangli Gao dst_release(&rt->dst); 2241da177e4SLinus Torvalds } 2251da177e4SLinus Torvalds 2261da177e4SLinus Torvalds #define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3) 2271da177e4SLinus Torvalds 2284839c52bSPhilippe De Muyter extern const __u8 ip_tos2prio[16]; 2291da177e4SLinus Torvalds 2301da177e4SLinus Torvalds static inline char rt_tos2priority(u8 tos) 2311da177e4SLinus Torvalds { 2321da177e4SLinus Torvalds return ip_tos2prio[IPTOS_TOS(tos)>>1]; 2331da177e4SLinus Torvalds } 2341da177e4SLinus Torvalds 2352d7192d6SDavid S. Miller /* ip_route_connect() and ip_route_newports() work in tandem whilst 2362d7192d6SDavid S. Miller * binding a socket for a new outgoing connection. 2372d7192d6SDavid S. Miller * 2382d7192d6SDavid S. Miller * In order to use IPSEC properly, we must, in the end, have a 2392d7192d6SDavid S. Miller * route that was looked up using all available keys including source 2402d7192d6SDavid S. Miller * and destination ports. 2412d7192d6SDavid S. Miller * 2422d7192d6SDavid S. Miller * However, if a source port needs to be allocated (the user specified 2432d7192d6SDavid S. Miller * a wildcard source port) we need to obtain addressing information 2442d7192d6SDavid S. Miller * in order to perform that allocation. 2452d7192d6SDavid S. Miller * 2462d7192d6SDavid S. Miller * So ip_route_connect() looks up a route using wildcarded source and 2472d7192d6SDavid S. Miller * destination ports in the key, simply so that we can get a pair of 2482d7192d6SDavid S. Miller * addresses to use for port allocation. 2492d7192d6SDavid S. Miller * 2502d7192d6SDavid S. Miller * Later, once the ports are allocated, ip_route_newports() will make 2512d7192d6SDavid S. Miller * another route lookup if needed to make sure we catch any IPSEC 2522d7192d6SDavid S. Miller * rules keyed on the port information. 2532d7192d6SDavid S. Miller * 2542d7192d6SDavid S. Miller * The callers allocate the flow key on their stack, and must pass in 2552d7192d6SDavid S. Miller * the same flowi4 object to both the ip_route_connect() and the 2562d7192d6SDavid S. Miller * ip_route_newports() calls. 2572d7192d6SDavid S. Miller */ 2582d7192d6SDavid S. Miller 2592d7192d6SDavid S. Miller static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src, 2602d7192d6SDavid S. Miller u32 tos, int oif, u8 protocol, 261b23dd4feSDavid S. Miller __be16 sport, __be16 dport, 2620e0d44abSSteffen Klassert struct sock *sk) 2631da177e4SLinus Torvalds { 2642d7192d6SDavid S. Miller __u8 flow_flags = 0; 26579876874SKOVACS Krisztian 26679876874SKOVACS Krisztian if (inet_sk(sk)->transparent) 26794b92b88SDavid S. Miller flow_flags |= FLOWI_FLAG_ANYSRC; 26894b92b88SDavid S. Miller 269007979eaSDavid Ahern if (netif_index_is_l3_master(sock_net(sk), oif)) 27058189ca7SDavid Ahern flow_flags |= FLOWI_FLAG_VRFSRC | FLOWI_FLAG_SKIP_NH_OIF; 271613d09b3SDavid Ahern 2722d7192d6SDavid S. Miller flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE, 27394b92b88SDavid S. Miller protocol, flow_flags, dst, src, dport, sport); 2742d7192d6SDavid S. Miller } 2752d7192d6SDavid S. Miller 2762d7192d6SDavid S. Miller static inline struct rtable *ip_route_connect(struct flowi4 *fl4, 2772d7192d6SDavid S. Miller __be32 dst, __be32 src, u32 tos, 2782d7192d6SDavid S. Miller int oif, u8 protocol, 2792d7192d6SDavid S. Miller __be16 sport, __be16 dport, 2800e0d44abSSteffen Klassert struct sock *sk) 2812d7192d6SDavid S. Miller { 2822d7192d6SDavid S. Miller struct net *net = sock_net(sk); 2832d7192d6SDavid S. Miller struct rtable *rt; 2842d7192d6SDavid S. Miller 2852d7192d6SDavid S. Miller ip_route_connect_init(fl4, dst, src, tos, oif, protocol, 2860e0d44abSSteffen Klassert sport, dport, sk); 28779876874SKOVACS Krisztian 2881da177e4SLinus Torvalds if (!dst || !src) { 2892d7192d6SDavid S. Miller rt = __ip_route_output_key(net, fl4); 290b23dd4feSDavid S. Miller if (IS_ERR(rt)) 291b23dd4feSDavid S. Miller return rt; 292b23dd4feSDavid S. Miller ip_rt_put(rt); 293e6b45241SJulian Anastasov flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr); 2941da177e4SLinus Torvalds } 2952d7192d6SDavid S. Miller security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 2962d7192d6SDavid S. Miller return ip_route_output_flow(net, fl4, sk); 2971da177e4SLinus Torvalds } 2981da177e4SLinus Torvalds 2992d7192d6SDavid S. Miller static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt, 3002d7192d6SDavid S. Miller __be16 orig_sport, __be16 orig_dport, 3012d7192d6SDavid S. Miller __be16 sport, __be16 dport, 3022d7192d6SDavid S. Miller struct sock *sk) 3031da177e4SLinus Torvalds { 304dca8b089SDavid S. Miller if (sport != orig_sport || dport != orig_dport) { 3052d7192d6SDavid S. Miller fl4->fl4_dport = dport; 3062d7192d6SDavid S. Miller fl4->fl4_sport = sport; 307b23dd4feSDavid S. Miller ip_rt_put(rt); 308e6b45241SJulian Anastasov flowi4_update_output(fl4, sk->sk_bound_dev_if, 309e6b45241SJulian Anastasov RT_CONN_FLAGS(sk), fl4->daddr, 310e6b45241SJulian Anastasov fl4->saddr); 3112d7192d6SDavid S. Miller security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 3122d7192d6SDavid S. Miller return ip_route_output_flow(sock_net(sk), fl4, sk); 3131da177e4SLinus Torvalds } 314b23dd4feSDavid S. Miller return rt; 3151da177e4SLinus Torvalds } 3161da177e4SLinus Torvalds 3171668e010SKOVACS Krisztian static inline int inet_iif(const struct sk_buff *skb) 3181668e010SKOVACS Krisztian { 31913378cadSDavid S. Miller int iif = skb_rtable(skb)->rt_iif; 32013378cadSDavid S. Miller 32113378cadSDavid S. Miller if (iif) 32213378cadSDavid S. Miller return iif; 32313378cadSDavid S. Miller return skb->skb_iif; 3241668e010SKOVACS Krisztian } 3251668e010SKOVACS Krisztian 326323e126fSDavid S. Miller extern int sysctl_ip_default_ttl; 327323e126fSDavid S. Miller 328323e126fSDavid S. Miller static inline int ip4_dst_hoplimit(const struct dst_entry *dst) 329323e126fSDavid S. Miller { 330323e126fSDavid S. Miller int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT); 331323e126fSDavid S. Miller 332323e126fSDavid S. Miller if (hoplimit == 0) 333323e126fSDavid S. Miller hoplimit = sysctl_ip_default_ttl; 334323e126fSDavid S. Miller return hoplimit; 335323e126fSDavid S. Miller } 336323e126fSDavid S. Miller 3371da177e4SLinus Torvalds #endif /* _ROUTE_H */ 338