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 * Definitions for the IP router. 7 * 8 * Version: @(#)route.h 1.0.4 05/27/93 9 * 10 * Authors: Ross Biro 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 12 * Fixes: 13 * Alan Cox : Reformatted. Added ip_rt_local() 14 * Alan Cox : Support for TCP parameters. 15 * Alexey Kuznetsov: Major changes for new routing code. 16 * Mike McLagan : Routing by source 17 * Robert Olsson : Added rt_cache statistics 18 * 19 * This program is free software; you can redistribute it and/or 20 * modify it under the terms of the GNU General Public License 21 * as published by the Free Software Foundation; either version 22 * 2 of the License, or (at your option) any later version. 23 */ 24 #ifndef _ROUTE_H 25 #define _ROUTE_H 26 27 #include <net/dst.h> 28 #include <net/inetpeer.h> 29 #include <net/flow.h> 30 #include <net/inet_sock.h> 31 #include <linux/in_route.h> 32 #include <linux/rtnetlink.h> 33 #include <linux/route.h> 34 #include <linux/ip.h> 35 #include <linux/cache.h> 36 #include <linux/security.h> 37 38 #define RTO_ONLINK 0x01 39 40 #define RT_CONN_FLAGS(sk) (RT_TOS(inet_sk(sk)->tos) | sock_flag(sk, SOCK_LOCALROUTE)) 41 42 struct fib_nh; 43 struct fib_info; 44 struct rtable { 45 struct dst_entry dst; 46 47 int rt_genid; 48 unsigned int rt_flags; 49 __u16 rt_type; 50 __u16 rt_is_input; 51 52 int rt_iif; 53 54 /* Info on neighbour */ 55 __be32 rt_gateway; 56 57 /* Miscellaneous cached information */ 58 u32 rt_pmtu; 59 }; 60 61 static inline bool rt_is_input_route(const struct rtable *rt) 62 { 63 return rt->rt_is_input != 0; 64 } 65 66 static inline bool rt_is_output_route(const struct rtable *rt) 67 { 68 return rt->rt_is_input == 0; 69 } 70 71 static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr) 72 { 73 if (rt->rt_gateway) 74 return rt->rt_gateway; 75 return daddr; 76 } 77 78 struct ip_rt_acct { 79 __u32 o_bytes; 80 __u32 o_packets; 81 __u32 i_bytes; 82 __u32 i_packets; 83 }; 84 85 struct rt_cache_stat { 86 unsigned int in_hit; 87 unsigned int in_slow_tot; 88 unsigned int in_slow_mc; 89 unsigned int in_no_route; 90 unsigned int in_brd; 91 unsigned int in_martian_dst; 92 unsigned int in_martian_src; 93 unsigned int out_hit; 94 unsigned int out_slow_tot; 95 unsigned int out_slow_mc; 96 unsigned int gc_total; 97 unsigned int gc_ignored; 98 unsigned int gc_goal_miss; 99 unsigned int gc_dst_overflow; 100 unsigned int in_hlist_search; 101 unsigned int out_hlist_search; 102 }; 103 104 extern struct ip_rt_acct __percpu *ip_rt_acct; 105 106 struct in_device; 107 extern int ip_rt_init(void); 108 extern void rt_cache_flush(struct net *net, int how); 109 extern struct rtable *__ip_route_output_key(struct net *, struct flowi4 *flp); 110 extern struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp, 111 struct sock *sk); 112 extern struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig); 113 114 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp) 115 { 116 return ip_route_output_flow(net, flp, NULL); 117 } 118 119 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr, 120 __be32 saddr, u8 tos, int oif) 121 { 122 struct flowi4 fl4 = { 123 .flowi4_oif = oif, 124 .flowi4_tos = tos, 125 .daddr = daddr, 126 .saddr = saddr, 127 }; 128 return ip_route_output_key(net, &fl4); 129 } 130 131 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4, 132 struct sock *sk, 133 __be32 daddr, __be32 saddr, 134 __be16 dport, __be16 sport, 135 __u8 proto, __u8 tos, int oif) 136 { 137 flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos, 138 RT_SCOPE_UNIVERSE, proto, 139 sk ? inet_sk_flowi_flags(sk) : 0, 140 daddr, saddr, dport, sport); 141 if (sk) 142 security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 143 return ip_route_output_flow(net, fl4, sk); 144 } 145 146 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4, 147 __be32 daddr, __be32 saddr, 148 __be32 gre_key, __u8 tos, int oif) 149 { 150 memset(fl4, 0, sizeof(*fl4)); 151 fl4->flowi4_oif = oif; 152 fl4->daddr = daddr; 153 fl4->saddr = saddr; 154 fl4->flowi4_tos = tos; 155 fl4->flowi4_proto = IPPROTO_GRE; 156 fl4->fl4_gre_key = gre_key; 157 return ip_route_output_key(net, fl4); 158 } 159 160 extern int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src, 161 u8 tos, struct net_device *devin); 162 163 extern void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, 164 int oif, u32 mark, u8 protocol, int flow_flags); 165 extern void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu); 166 extern void ipv4_redirect(struct sk_buff *skb, struct net *net, 167 int oif, u32 mark, u8 protocol, int flow_flags); 168 extern void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk); 169 extern void ip_rt_send_redirect(struct sk_buff *skb); 170 171 extern unsigned int inet_addr_type(struct net *net, __be32 addr); 172 extern unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev, __be32 addr); 173 extern void ip_rt_multicast_event(struct in_device *); 174 extern int ip_rt_ioctl(struct net *, unsigned int cmd, void __user *arg); 175 extern void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt); 176 extern int ip_rt_dump(struct sk_buff *skb, struct netlink_callback *cb); 177 178 struct in_ifaddr; 179 extern void fib_add_ifaddr(struct in_ifaddr *); 180 extern void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *); 181 182 static inline void ip_rt_put(struct rtable * rt) 183 { 184 if (rt) 185 dst_release(&rt->dst); 186 } 187 188 #define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3) 189 190 extern const __u8 ip_tos2prio[16]; 191 192 static inline char rt_tos2priority(u8 tos) 193 { 194 return ip_tos2prio[IPTOS_TOS(tos)>>1]; 195 } 196 197 /* ip_route_connect() and ip_route_newports() work in tandem whilst 198 * binding a socket for a new outgoing connection. 199 * 200 * In order to use IPSEC properly, we must, in the end, have a 201 * route that was looked up using all available keys including source 202 * and destination ports. 203 * 204 * However, if a source port needs to be allocated (the user specified 205 * a wildcard source port) we need to obtain addressing information 206 * in order to perform that allocation. 207 * 208 * So ip_route_connect() looks up a route using wildcarded source and 209 * destination ports in the key, simply so that we can get a pair of 210 * addresses to use for port allocation. 211 * 212 * Later, once the ports are allocated, ip_route_newports() will make 213 * another route lookup if needed to make sure we catch any IPSEC 214 * rules keyed on the port information. 215 * 216 * The callers allocate the flow key on their stack, and must pass in 217 * the same flowi4 object to both the ip_route_connect() and the 218 * ip_route_newports() calls. 219 */ 220 221 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src, 222 u32 tos, int oif, u8 protocol, 223 __be16 sport, __be16 dport, 224 struct sock *sk, bool can_sleep) 225 { 226 __u8 flow_flags = 0; 227 228 if (inet_sk(sk)->transparent) 229 flow_flags |= FLOWI_FLAG_ANYSRC; 230 if (can_sleep) 231 flow_flags |= FLOWI_FLAG_CAN_SLEEP; 232 233 flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE, 234 protocol, flow_flags, dst, src, dport, sport); 235 } 236 237 static inline struct rtable *ip_route_connect(struct flowi4 *fl4, 238 __be32 dst, __be32 src, u32 tos, 239 int oif, u8 protocol, 240 __be16 sport, __be16 dport, 241 struct sock *sk, bool can_sleep) 242 { 243 struct net *net = sock_net(sk); 244 struct rtable *rt; 245 246 ip_route_connect_init(fl4, dst, src, tos, oif, protocol, 247 sport, dport, sk, can_sleep); 248 249 if (!dst || !src) { 250 rt = __ip_route_output_key(net, fl4); 251 if (IS_ERR(rt)) 252 return rt; 253 ip_rt_put(rt); 254 flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr); 255 } 256 security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 257 return ip_route_output_flow(net, fl4, sk); 258 } 259 260 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt, 261 __be16 orig_sport, __be16 orig_dport, 262 __be16 sport, __be16 dport, 263 struct sock *sk) 264 { 265 if (sport != orig_sport || dport != orig_dport) { 266 fl4->fl4_dport = dport; 267 fl4->fl4_sport = sport; 268 ip_rt_put(rt); 269 flowi4_update_output(fl4, sk->sk_bound_dev_if, 270 RT_CONN_FLAGS(sk), fl4->daddr, 271 fl4->saddr); 272 security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 273 return ip_route_output_flow(sock_net(sk), fl4, sk); 274 } 275 return rt; 276 } 277 278 static inline int inet_iif(const struct sk_buff *skb) 279 { 280 return skb_rtable(skb)->rt_iif; 281 } 282 283 extern int sysctl_ip_default_ttl; 284 285 static inline int ip4_dst_hoplimit(const struct dst_entry *dst) 286 { 287 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT); 288 289 if (hoplimit == 0) 290 hoplimit = sysctl_ip_default_ttl; 291 return hoplimit; 292 } 293 294 #endif /* _ROUTE_H */ 295