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> 321da177e4SLinus Torvalds #include <linux/in_route.h> 331da177e4SLinus Torvalds #include <linux/rtnetlink.h> 34c6cffba4SDavid S. Miller #include <linux/rcupdate.h> 351da177e4SLinus Torvalds #include <linux/route.h> 361da177e4SLinus Torvalds #include <linux/ip.h> 371da177e4SLinus Torvalds #include <linux/cache.h> 38beb8d13bSVenkat Yekkirala #include <linux/security.h> 391da177e4SLinus Torvalds 40f87c10a8SHannes Frederic Sowa /* IPv4 datagram length is stored into 16bit field (tot_len) */ 41f87c10a8SHannes Frederic Sowa #define IP_MAX_MTU 0xFFFFU 42f87c10a8SHannes Frederic Sowa 431da177e4SLinus Torvalds #define RTO_ONLINK 0x01 441da177e4SLinus Torvalds 451da177e4SLinus Torvalds #define RT_CONN_FLAGS(sk) (RT_TOS(inet_sk(sk)->tos) | sock_flag(sk, SOCK_LOCALROUTE)) 46aa661581SFrancesco Fusco #define RT_CONN_FLAGS_TOS(sk,tos) (RT_TOS(tos) | sock_flag(sk, SOCK_LOCALROUTE)) 471da177e4SLinus Torvalds 481da177e4SLinus Torvalds struct fib_nh; 4962fa8a84SDavid S. Miller struct fib_info; 505055c371SEric Dumazet struct uncached_list; 51fd2c3ef7SEric Dumazet struct rtable { 521da177e4SLinus Torvalds struct dst_entry dst; 531da177e4SLinus Torvalds 5429e75252SEric Dumazet int rt_genid; 5595c96174SEric Dumazet unsigned int rt_flags; 561da177e4SLinus Torvalds __u16 rt_type; 57155e8336SJulian Anastasov __u8 rt_is_input; 58155e8336SJulian Anastasov __u8 rt_uses_gateway; 591da177e4SLinus Torvalds 601da177e4SLinus Torvalds int rt_iif; 611da177e4SLinus Torvalds 621da177e4SLinus Torvalds /* Info on neighbour */ 63f2c3fe24SAl Viro __be32 rt_gateway; 641da177e4SLinus Torvalds 651da177e4SLinus Torvalds /* Miscellaneous cached information */ 665943634fSDavid S. Miller u32 rt_pmtu; 67caacf05eSDavid S. Miller 68b7503e0cSDavid Ahern u32 rt_table_id; 69b7503e0cSDavid Ahern 70caacf05eSDavid S. Miller struct list_head rt_uncached; 715055c371SEric Dumazet struct uncached_list *rt_uncached_list; 721da177e4SLinus Torvalds }; 731da177e4SLinus Torvalds 74b8400f37SSteffen Klassert static inline bool rt_is_input_route(const struct rtable *rt) 75c7537967SDavid S. Miller { 769917e1e8SDavid S. Miller return rt->rt_is_input != 0; 77c7537967SDavid S. Miller } 78c7537967SDavid S. Miller 79b8400f37SSteffen Klassert static inline bool rt_is_output_route(const struct rtable *rt) 80c7537967SDavid S. Miller { 819917e1e8SDavid S. Miller return rt->rt_is_input == 0; 82c7537967SDavid S. Miller } 83c7537967SDavid S. Miller 84f8126f1dSDavid S. Miller static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr) 85f8126f1dSDavid S. Miller { 86f8126f1dSDavid S. Miller if (rt->rt_gateway) 87f8126f1dSDavid S. Miller return rt->rt_gateway; 88f8126f1dSDavid S. Miller return daddr; 89f8126f1dSDavid S. Miller } 90f8126f1dSDavid S. Miller 91fd2c3ef7SEric Dumazet struct ip_rt_acct { 921da177e4SLinus Torvalds __u32 o_bytes; 931da177e4SLinus Torvalds __u32 o_packets; 941da177e4SLinus Torvalds __u32 i_bytes; 951da177e4SLinus Torvalds __u32 i_packets; 961da177e4SLinus Torvalds }; 971da177e4SLinus Torvalds 98fd2c3ef7SEric Dumazet struct rt_cache_stat { 991da177e4SLinus Torvalds unsigned int in_slow_tot; 1001da177e4SLinus Torvalds unsigned int in_slow_mc; 1011da177e4SLinus Torvalds unsigned int in_no_route; 1021da177e4SLinus Torvalds unsigned int in_brd; 1031da177e4SLinus Torvalds unsigned int in_martian_dst; 1041da177e4SLinus Torvalds unsigned int in_martian_src; 1051da177e4SLinus Torvalds unsigned int out_slow_tot; 1061da177e4SLinus Torvalds unsigned int out_slow_mc; 1071da177e4SLinus Torvalds }; 1081da177e4SLinus Torvalds 1097d720c3eSTejun Heo extern struct ip_rt_acct __percpu *ip_rt_acct; 1101da177e4SLinus Torvalds 1111da177e4SLinus Torvalds struct in_device; 1122702c4bbSJoe Perches 1132702c4bbSJoe Perches int ip_rt_init(void); 1142702c4bbSJoe Perches void rt_cache_flush(struct net *net); 1152702c4bbSJoe Perches void rt_flush_dev(struct net_device *dev); 1163abd1adeSDavid Ahern struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp, 1173abd1adeSDavid Ahern const struct sk_buff *skb); 1183abd1adeSDavid Ahern struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp, 1193abd1adeSDavid Ahern struct fib_result *res, 120bf4e0a3dSNikolay Aleksandrov const struct sk_buff *skb); 12179a13159SPeter Nørlund 12279a13159SPeter Nørlund static inline struct rtable *__ip_route_output_key(struct net *net, 12379a13159SPeter Nørlund struct flowi4 *flp) 12479a13159SPeter Nørlund { 1253abd1adeSDavid Ahern return ip_route_output_key_hash(net, flp, NULL); 12679a13159SPeter Nørlund } 12779a13159SPeter Nørlund 1282702c4bbSJoe Perches struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp, 1296f9c9615SEric Dumazet const struct sock *sk); 1302702c4bbSJoe Perches struct dst_entry *ipv4_blackhole_route(struct net *net, 1312702c4bbSJoe Perches struct dst_entry *dst_orig); 132407eadd9SEric Dumazet 1339d6ec938SDavid S. Miller static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp) 1345bfa787fSDavid S. Miller { 1355bfa787fSDavid S. Miller return ip_route_output_flow(net, flp, NULL); 1365bfa787fSDavid S. Miller } 1375bfa787fSDavid S. Miller 13878fbfd8aSDavid S. Miller static inline struct rtable *ip_route_output(struct net *net, __be32 daddr, 13978fbfd8aSDavid S. Miller __be32 saddr, u8 tos, int oif) 14078fbfd8aSDavid S. Miller { 1419d6ec938SDavid S. Miller struct flowi4 fl4 = { 1429d6ec938SDavid S. Miller .flowi4_oif = oif, 143c5d21c4bSRoland Dreier .flowi4_tos = tos, 1449d6ec938SDavid S. Miller .daddr = daddr, 1459d6ec938SDavid S. Miller .saddr = saddr, 14678fbfd8aSDavid S. Miller }; 1479d6ec938SDavid S. Miller return ip_route_output_key(net, &fl4); 14878fbfd8aSDavid S. Miller } 14978fbfd8aSDavid S. Miller 15031e4543dSDavid S. Miller static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4, 15131e4543dSDavid S. Miller struct sock *sk, 15278fbfd8aSDavid S. Miller __be32 daddr, __be32 saddr, 15378fbfd8aSDavid S. Miller __be16 dport, __be16 sport, 15478fbfd8aSDavid S. Miller __u8 proto, __u8 tos, int oif) 15578fbfd8aSDavid S. Miller { 15631e4543dSDavid S. Miller flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos, 15794b92b88SDavid S. Miller RT_SCOPE_UNIVERSE, proto, 15894b92b88SDavid S. Miller sk ? inet_sk_flowi_flags(sk) : 0, 159e2d118a1SLorenzo Colitti daddr, saddr, dport, sport, sock_net_uid(net, sk)); 16078fbfd8aSDavid S. Miller if (sk) 16131e4543dSDavid S. Miller security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 16231e4543dSDavid S. Miller return ip_route_output_flow(net, fl4, sk); 16378fbfd8aSDavid S. Miller } 16478fbfd8aSDavid S. Miller 165cbb1e85fSDavid S. Miller static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4, 16678fbfd8aSDavid S. Miller __be32 daddr, __be32 saddr, 16778fbfd8aSDavid S. Miller __be32 gre_key, __u8 tos, int oif) 16878fbfd8aSDavid S. Miller { 169cbb1e85fSDavid S. Miller memset(fl4, 0, sizeof(*fl4)); 170cbb1e85fSDavid S. Miller fl4->flowi4_oif = oif; 171cbb1e85fSDavid S. Miller fl4->daddr = daddr; 172cbb1e85fSDavid S. Miller fl4->saddr = saddr; 173cbb1e85fSDavid S. Miller fl4->flowi4_tos = tos; 174cbb1e85fSDavid S. Miller fl4->flowi4_proto = IPPROTO_GRE; 175cbb1e85fSDavid S. Miller fl4->fl4_gre_key = gre_key; 176cbb1e85fSDavid S. Miller return ip_route_output_key(net, fl4); 17778fbfd8aSDavid S. Miller } 178bc044e8dSPaolo Abeni int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr, 179bc044e8dSPaolo Abeni u8 tos, struct net_device *dev, 180bc044e8dSPaolo Abeni struct in_device *in_dev, u32 *itag); 1812702c4bbSJoe Perches int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src, 18238a424e4SDavid Miller u8 tos, struct net_device *devin); 1835510cdf7SDavid Ahern int ip_route_input_rcu(struct sk_buff *skb, __be32 dst, __be32 src, 1845510cdf7SDavid Ahern u8 tos, struct net_device *devin, 1855510cdf7SDavid Ahern struct fib_result *res); 186407eadd9SEric Dumazet 187c6cffba4SDavid S. Miller static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src, 188c6cffba4SDavid S. Miller u8 tos, struct net_device *devin) 189c6cffba4SDavid S. Miller { 190c6cffba4SDavid S. Miller int err; 191c6cffba4SDavid S. Miller 192c6cffba4SDavid S. Miller rcu_read_lock(); 193c6cffba4SDavid S. Miller err = ip_route_input_noref(skb, dst, src, tos, devin); 19464327fc8SStefano Brivio if (!err) { 195222d7dbdSEric Dumazet skb_dst_force(skb); 1969df16efaSWei Wang if (!skb_dst(skb)) 1979df16efaSWei Wang err = -EINVAL; 19864327fc8SStefano Brivio } 199c6cffba4SDavid S. Miller rcu_read_unlock(); 200c6cffba4SDavid S. Miller 201c6cffba4SDavid S. Miller return err; 202c6cffba4SDavid S. Miller } 203c6cffba4SDavid S. Miller 2042702c4bbSJoe Perches void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif, 2052702c4bbSJoe Perches u32 mark, u8 protocol, int flow_flags); 2062702c4bbSJoe Perches void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu); 2072702c4bbSJoe Perches void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark, 2082702c4bbSJoe Perches u8 protocol, int flow_flags); 2092702c4bbSJoe Perches void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk); 2102702c4bbSJoe Perches void ip_rt_send_redirect(struct sk_buff *skb); 2111da177e4SLinus Torvalds 2122702c4bbSJoe Perches unsigned int inet_addr_type(struct net *net, __be32 addr); 2139b8ff518SDavid Ahern unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id); 2142702c4bbSJoe Perches unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev, 2152702c4bbSJoe Perches __be32 addr); 21630bbaa19SDavid Ahern unsigned int inet_addr_type_dev_table(struct net *net, 21730bbaa19SDavid Ahern const struct net_device *dev, 21830bbaa19SDavid Ahern __be32 addr); 2192702c4bbSJoe Perches void ip_rt_multicast_event(struct in_device *); 220ca25c300SAl Viro int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt); 2212702c4bbSJoe Perches void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt); 2229ab179d8SDavid Ahern struct rtable *rt_dst_alloc(struct net_device *dev, 2239ab179d8SDavid Ahern unsigned int flags, u16 type, 2249ab179d8SDavid Ahern bool nopolicy, bool noxfrm, bool will_cache); 2251da177e4SLinus Torvalds 2260ff60a45SJamal Hadi Salim struct in_ifaddr; 2272702c4bbSJoe Perches void fib_add_ifaddr(struct in_ifaddr *); 2282702c4bbSJoe Perches void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *); 2290ff60a45SJamal Hadi Salim 2301da177e4SLinus Torvalds static inline void ip_rt_put(struct rtable *rt) 2311da177e4SLinus Torvalds { 2326da025faSEric Dumazet /* dst_release() accepts a NULL parameter. 2336da025faSEric Dumazet * We rely on dst being first structure in struct rtable 2346da025faSEric Dumazet */ 2356da025faSEric Dumazet BUILD_BUG_ON(offsetof(struct rtable, dst) != 0); 236d8d1f30bSChangli Gao dst_release(&rt->dst); 2371da177e4SLinus Torvalds } 2381da177e4SLinus Torvalds 2391da177e4SLinus Torvalds #define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3) 2401da177e4SLinus Torvalds 2414839c52bSPhilippe De Muyter extern const __u8 ip_tos2prio[16]; 2421da177e4SLinus Torvalds 2431da177e4SLinus Torvalds static inline char rt_tos2priority(u8 tos) 2441da177e4SLinus Torvalds { 2451da177e4SLinus Torvalds return ip_tos2prio[IPTOS_TOS(tos)>>1]; 2461da177e4SLinus Torvalds } 2471da177e4SLinus Torvalds 2482d7192d6SDavid S. Miller /* ip_route_connect() and ip_route_newports() work in tandem whilst 2492d7192d6SDavid S. Miller * binding a socket for a new outgoing connection. 2502d7192d6SDavid S. Miller * 2512d7192d6SDavid S. Miller * In order to use IPSEC properly, we must, in the end, have a 2522d7192d6SDavid S. Miller * route that was looked up using all available keys including source 2532d7192d6SDavid S. Miller * and destination ports. 2542d7192d6SDavid S. Miller * 2552d7192d6SDavid S. Miller * However, if a source port needs to be allocated (the user specified 2562d7192d6SDavid S. Miller * a wildcard source port) we need to obtain addressing information 2572d7192d6SDavid S. Miller * in order to perform that allocation. 2582d7192d6SDavid S. Miller * 2592d7192d6SDavid S. Miller * So ip_route_connect() looks up a route using wildcarded source and 2602d7192d6SDavid S. Miller * destination ports in the key, simply so that we can get a pair of 2612d7192d6SDavid S. Miller * addresses to use for port allocation. 2622d7192d6SDavid S. Miller * 2632d7192d6SDavid S. Miller * Later, once the ports are allocated, ip_route_newports() will make 2642d7192d6SDavid S. Miller * another route lookup if needed to make sure we catch any IPSEC 2652d7192d6SDavid S. Miller * rules keyed on the port information. 2662d7192d6SDavid S. Miller * 2672d7192d6SDavid S. Miller * The callers allocate the flow key on their stack, and must pass in 2682d7192d6SDavid S. Miller * the same flowi4 object to both the ip_route_connect() and the 2692d7192d6SDavid S. Miller * ip_route_newports() calls. 2702d7192d6SDavid S. Miller */ 2712d7192d6SDavid S. Miller 2722d7192d6SDavid S. Miller static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src, 2732d7192d6SDavid S. Miller u32 tos, int oif, u8 protocol, 274b23dd4feSDavid S. Miller __be16 sport, __be16 dport, 2750e0d44abSSteffen Klassert struct sock *sk) 2761da177e4SLinus Torvalds { 2772d7192d6SDavid S. Miller __u8 flow_flags = 0; 27879876874SKOVACS Krisztian 27979876874SKOVACS Krisztian if (inet_sk(sk)->transparent) 28094b92b88SDavid S. Miller flow_flags |= FLOWI_FLAG_ANYSRC; 28194b92b88SDavid S. Miller 2822d7192d6SDavid S. Miller flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE, 283e2d118a1SLorenzo Colitti protocol, flow_flags, dst, src, dport, sport, 284e2d118a1SLorenzo Colitti sk->sk_uid); 2852d7192d6SDavid S. Miller } 2862d7192d6SDavid S. Miller 2872d7192d6SDavid S. Miller static inline struct rtable *ip_route_connect(struct flowi4 *fl4, 2882d7192d6SDavid S. Miller __be32 dst, __be32 src, u32 tos, 2892d7192d6SDavid S. Miller int oif, u8 protocol, 2902d7192d6SDavid S. Miller __be16 sport, __be16 dport, 2910e0d44abSSteffen Klassert struct sock *sk) 2922d7192d6SDavid S. Miller { 2932d7192d6SDavid S. Miller struct net *net = sock_net(sk); 2942d7192d6SDavid S. Miller struct rtable *rt; 2952d7192d6SDavid S. Miller 2962d7192d6SDavid S. Miller ip_route_connect_init(fl4, dst, src, tos, oif, protocol, 2970e0d44abSSteffen Klassert sport, dport, sk); 29879876874SKOVACS Krisztian 2991da177e4SLinus Torvalds if (!dst || !src) { 3002d7192d6SDavid S. Miller rt = __ip_route_output_key(net, fl4); 301b23dd4feSDavid S. Miller if (IS_ERR(rt)) 302b23dd4feSDavid S. Miller return rt; 303b23dd4feSDavid S. Miller ip_rt_put(rt); 304e6b45241SJulian Anastasov flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr); 3051da177e4SLinus Torvalds } 3062d7192d6SDavid S. Miller security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 3072d7192d6SDavid S. Miller return ip_route_output_flow(net, fl4, sk); 3081da177e4SLinus Torvalds } 3091da177e4SLinus Torvalds 3102d7192d6SDavid S. Miller static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt, 3112d7192d6SDavid S. Miller __be16 orig_sport, __be16 orig_dport, 3122d7192d6SDavid S. Miller __be16 sport, __be16 dport, 3132d7192d6SDavid S. Miller struct sock *sk) 3141da177e4SLinus Torvalds { 315dca8b089SDavid S. Miller if (sport != orig_sport || dport != orig_dport) { 3162d7192d6SDavid S. Miller fl4->fl4_dport = dport; 3172d7192d6SDavid S. Miller fl4->fl4_sport = sport; 318b23dd4feSDavid S. Miller ip_rt_put(rt); 319e6b45241SJulian Anastasov flowi4_update_output(fl4, sk->sk_bound_dev_if, 320e6b45241SJulian Anastasov RT_CONN_FLAGS(sk), fl4->daddr, 321e6b45241SJulian Anastasov fl4->saddr); 3222d7192d6SDavid S. Miller security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 3232d7192d6SDavid S. Miller return ip_route_output_flow(sock_net(sk), fl4, sk); 3241da177e4SLinus Torvalds } 325b23dd4feSDavid S. Miller return rt; 3261da177e4SLinus Torvalds } 3271da177e4SLinus Torvalds 3281668e010SKOVACS Krisztian static inline int inet_iif(const struct sk_buff *skb) 3291668e010SKOVACS Krisztian { 3300340d0b9STom Herbert struct rtable *rt = skb_rtable(skb); 33113378cadSDavid S. Miller 3320340d0b9STom Herbert if (rt && rt->rt_iif) 3330340d0b9STom Herbert return rt->rt_iif; 3340340d0b9STom Herbert 33513378cadSDavid S. Miller return skb->skb_iif; 3361668e010SKOVACS Krisztian } 3371668e010SKOVACS Krisztian 338323e126fSDavid S. Miller static inline int ip4_dst_hoplimit(const struct dst_entry *dst) 339323e126fSDavid S. Miller { 340323e126fSDavid S. Miller int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT); 341fa50d974SNikolay Borisov struct net *net = dev_net(dst->dev); 342323e126fSDavid S. Miller 343323e126fSDavid S. Miller if (hoplimit == 0) 344fa50d974SNikolay Borisov hoplimit = net->ipv4.sysctl_ip_default_ttl; 345323e126fSDavid S. Miller return hoplimit; 346323e126fSDavid S. Miller } 347323e126fSDavid S. Miller 3481da177e4SLinus Torvalds #endif /* _ROUTE_H */ 349