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 RTO_CONN 0 41 /* RTO_CONN is not used (being alias for 0), but preserved not to break 42 * some modules referring to it. */ 43 44 #define RT_CONN_FLAGS(sk) (RT_TOS(inet_sk(sk)->tos) | sock_flag(sk, SOCK_LOCALROUTE)) 45 46 struct fib_nh; 47 struct inet_peer; 48 struct fib_info; 49 struct rtable { 50 struct dst_entry dst; 51 52 /* Lookup key. */ 53 __be32 rt_key_dst; 54 __be32 rt_key_src; 55 56 int rt_genid; 57 unsigned rt_flags; 58 __u16 rt_type; 59 __u8 rt_tos; 60 61 __be32 rt_dst; /* Path destination */ 62 __be32 rt_src; /* Path source */ 63 int rt_route_iif; 64 int rt_iif; 65 int rt_oif; 66 __u32 rt_mark; 67 68 /* Info on neighbour */ 69 __be32 rt_gateway; 70 71 /* Miscellaneous cached information */ 72 __be32 rt_spec_dst; /* RFC1122 specific destination */ 73 u32 rt_peer_genid; 74 struct inet_peer *peer; /* long-living peer info */ 75 struct fib_info *fi; /* for client ref to shared metrics */ 76 }; 77 78 static inline bool rt_is_input_route(struct rtable *rt) 79 { 80 return rt->rt_route_iif != 0; 81 } 82 83 static inline bool rt_is_output_route(struct rtable *rt) 84 { 85 return rt->rt_route_iif == 0; 86 } 87 88 struct ip_rt_acct { 89 __u32 o_bytes; 90 __u32 o_packets; 91 __u32 i_bytes; 92 __u32 i_packets; 93 }; 94 95 struct rt_cache_stat { 96 unsigned int in_hit; 97 unsigned int in_slow_tot; 98 unsigned int in_slow_mc; 99 unsigned int in_no_route; 100 unsigned int in_brd; 101 unsigned int in_martian_dst; 102 unsigned int in_martian_src; 103 unsigned int out_hit; 104 unsigned int out_slow_tot; 105 unsigned int out_slow_mc; 106 unsigned int gc_total; 107 unsigned int gc_ignored; 108 unsigned int gc_goal_miss; 109 unsigned int gc_dst_overflow; 110 unsigned int in_hlist_search; 111 unsigned int out_hlist_search; 112 }; 113 114 extern struct ip_rt_acct __percpu *ip_rt_acct; 115 116 struct in_device; 117 extern int ip_rt_init(void); 118 extern void ip_rt_redirect(__be32 old_gw, __be32 dst, __be32 new_gw, 119 __be32 src, struct net_device *dev); 120 extern void rt_cache_flush(struct net *net, int how); 121 extern void rt_cache_flush_batch(struct net *net); 122 extern struct rtable *__ip_route_output_key(struct net *, const struct flowi4 *flp); 123 extern struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp, 124 struct sock *sk); 125 extern struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig); 126 127 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp) 128 { 129 return ip_route_output_flow(net, flp, NULL); 130 } 131 132 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr, 133 __be32 saddr, u8 tos, int oif) 134 { 135 struct flowi4 fl4 = { 136 .flowi4_oif = oif, 137 .daddr = daddr, 138 .saddr = saddr, 139 .flowi4_tos = tos, 140 }; 141 return ip_route_output_key(net, &fl4); 142 } 143 144 static inline struct rtable *ip_route_output_ports(struct net *net, struct sock *sk, 145 __be32 daddr, __be32 saddr, 146 __be16 dport, __be16 sport, 147 __u8 proto, __u8 tos, int oif) 148 { 149 struct flowi4 fl4; 150 151 flowi4_init_output(&fl4, oif, sk ? sk->sk_mark : 0, tos, 152 RT_SCOPE_UNIVERSE, proto, 153 sk ? inet_sk_flowi_flags(sk) : 0, 154 daddr, saddr, dport, sport); 155 if (sk) 156 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4)); 157 return ip_route_output_flow(net, &fl4, sk); 158 } 159 160 static inline struct rtable *ip_route_output_gre(struct net *net, 161 __be32 daddr, __be32 saddr, 162 __be32 gre_key, __u8 tos, int oif) 163 { 164 struct flowi4 fl4 = { 165 .flowi4_oif = oif, 166 .daddr = daddr, 167 .saddr = saddr, 168 .flowi4_tos = tos, 169 .flowi4_proto = IPPROTO_GRE, 170 .fl4_gre_key = gre_key, 171 }; 172 return ip_route_output_key(net, &fl4); 173 } 174 175 extern int ip_route_input_common(struct sk_buff *skb, __be32 dst, __be32 src, 176 u8 tos, struct net_device *devin, bool noref); 177 178 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src, 179 u8 tos, struct net_device *devin) 180 { 181 return ip_route_input_common(skb, dst, src, tos, devin, false); 182 } 183 184 static inline int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src, 185 u8 tos, struct net_device *devin) 186 { 187 return ip_route_input_common(skb, dst, src, tos, devin, true); 188 } 189 190 extern unsigned short ip_rt_frag_needed(struct net *net, const struct iphdr *iph, 191 unsigned short new_mtu, struct net_device *dev); 192 extern void ip_rt_send_redirect(struct sk_buff *skb); 193 194 extern unsigned inet_addr_type(struct net *net, __be32 addr); 195 extern unsigned inet_dev_addr_type(struct net *net, const struct net_device *dev, __be32 addr); 196 extern void ip_rt_multicast_event(struct in_device *); 197 extern int ip_rt_ioctl(struct net *, unsigned int cmd, void __user *arg); 198 extern void ip_rt_get_source(u8 *src, struct rtable *rt); 199 extern int ip_rt_dump(struct sk_buff *skb, struct netlink_callback *cb); 200 201 struct in_ifaddr; 202 extern void fib_add_ifaddr(struct in_ifaddr *); 203 extern void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *); 204 205 static inline void ip_rt_put(struct rtable * rt) 206 { 207 if (rt) 208 dst_release(&rt->dst); 209 } 210 211 #define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3) 212 213 extern const __u8 ip_tos2prio[16]; 214 215 static inline char rt_tos2priority(u8 tos) 216 { 217 return ip_tos2prio[IPTOS_TOS(tos)>>1]; 218 } 219 220 /* ip_route_connect() and ip_route_newports() work in tandem whilst 221 * binding a socket for a new outgoing connection. 222 * 223 * In order to use IPSEC properly, we must, in the end, have a 224 * route that was looked up using all available keys including source 225 * and destination ports. 226 * 227 * However, if a source port needs to be allocated (the user specified 228 * a wildcard source port) we need to obtain addressing information 229 * in order to perform that allocation. 230 * 231 * So ip_route_connect() looks up a route using wildcarded source and 232 * destination ports in the key, simply so that we can get a pair of 233 * addresses to use for port allocation. 234 * 235 * Later, once the ports are allocated, ip_route_newports() will make 236 * another route lookup if needed to make sure we catch any IPSEC 237 * rules keyed on the port information. 238 * 239 * The callers allocate the flow key on their stack, and must pass in 240 * the same flowi4 object to both the ip_route_connect() and the 241 * ip_route_newports() calls. 242 */ 243 244 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src, 245 u32 tos, int oif, u8 protocol, 246 __be16 sport, __be16 dport, 247 struct sock *sk, bool can_sleep) 248 { 249 __u8 flow_flags = 0; 250 251 if (inet_sk(sk)->transparent) 252 flow_flags |= FLOWI_FLAG_ANYSRC; 253 if (protocol == IPPROTO_TCP) 254 flow_flags |= FLOWI_FLAG_PRECOW_METRICS; 255 if (can_sleep) 256 flow_flags |= FLOWI_FLAG_CAN_SLEEP; 257 258 flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE, 259 protocol, flow_flags, dst, src, dport, sport); 260 } 261 262 static inline struct rtable *ip_route_connect(struct flowi4 *fl4, 263 __be32 dst, __be32 src, u32 tos, 264 int oif, u8 protocol, 265 __be16 sport, __be16 dport, 266 struct sock *sk, bool can_sleep) 267 { 268 struct net *net = sock_net(sk); 269 struct rtable *rt; 270 271 ip_route_connect_init(fl4, dst, src, tos, oif, protocol, 272 sport, dport, sk, can_sleep); 273 274 if (!dst || !src) { 275 rt = __ip_route_output_key(net, fl4); 276 if (IS_ERR(rt)) 277 return rt; 278 fl4->daddr = rt->rt_dst; 279 fl4->saddr = rt->rt_src; 280 ip_rt_put(rt); 281 } 282 security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 283 return ip_route_output_flow(net, fl4, sk); 284 } 285 286 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt, 287 __be16 orig_sport, __be16 orig_dport, 288 __be16 sport, __be16 dport, 289 struct sock *sk) 290 { 291 if (sport != orig_sport || dport != orig_dport) { 292 fl4->fl4_dport = dport; 293 fl4->fl4_sport = sport; 294 ip_rt_put(rt); 295 security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 296 return ip_route_output_flow(sock_net(sk), fl4, sk); 297 } 298 return rt; 299 } 300 301 extern void rt_bind_peer(struct rtable *rt, int create); 302 303 static inline struct inet_peer *rt_get_peer(struct rtable *rt) 304 { 305 if (rt->peer) 306 return rt->peer; 307 308 rt_bind_peer(rt, 0); 309 return rt->peer; 310 } 311 312 static inline int inet_iif(const struct sk_buff *skb) 313 { 314 return skb_rtable(skb)->rt_iif; 315 } 316 317 extern int sysctl_ip_default_ttl; 318 319 static inline int ip4_dst_hoplimit(const struct dst_entry *dst) 320 { 321 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT); 322 323 if (hoplimit == 0) 324 hoplimit = sysctl_ip_default_ttl; 325 return hoplimit; 326 } 327 328 #endif /* _ROUTE_H */ 329