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/rcupdate.h> 34 #include <linux/route.h> 35 #include <linux/ip.h> 36 #include <linux/cache.h> 37 #include <linux/security.h> 38 39 #define RTO_ONLINK 0x01 40 41 #define RT_CONN_FLAGS(sk) (RT_TOS(inet_sk(sk)->tos) | sock_flag(sk, SOCK_LOCALROUTE)) 42 43 struct fib_nh; 44 struct fib_info; 45 struct rtable { 46 struct dst_entry dst; 47 48 int rt_genid; 49 unsigned int rt_flags; 50 __u16 rt_type; 51 __u8 rt_is_input; 52 __u8 rt_uses_gateway; 53 54 int rt_iif; 55 56 /* Info on neighbour */ 57 __be32 rt_gateway; 58 59 /* Miscellaneous cached information */ 60 u32 rt_pmtu; 61 62 struct list_head rt_uncached; 63 }; 64 65 static inline bool rt_is_input_route(const struct rtable *rt) 66 { 67 return rt->rt_is_input != 0; 68 } 69 70 static inline bool rt_is_output_route(const struct rtable *rt) 71 { 72 return rt->rt_is_input == 0; 73 } 74 75 static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr) 76 { 77 if (rt->rt_gateway) 78 return rt->rt_gateway; 79 return daddr; 80 } 81 82 struct ip_rt_acct { 83 __u32 o_bytes; 84 __u32 o_packets; 85 __u32 i_bytes; 86 __u32 i_packets; 87 }; 88 89 struct rt_cache_stat { 90 unsigned int in_hit; 91 unsigned int in_slow_tot; 92 unsigned int in_slow_mc; 93 unsigned int in_no_route; 94 unsigned int in_brd; 95 unsigned int in_martian_dst; 96 unsigned int in_martian_src; 97 unsigned int out_hit; 98 unsigned int out_slow_tot; 99 unsigned int out_slow_mc; 100 unsigned int gc_total; 101 unsigned int gc_ignored; 102 unsigned int gc_goal_miss; 103 unsigned int gc_dst_overflow; 104 unsigned int in_hlist_search; 105 unsigned int out_hlist_search; 106 }; 107 108 extern struct ip_rt_acct __percpu *ip_rt_acct; 109 110 struct in_device; 111 112 int ip_rt_init(void); 113 void rt_cache_flush(struct net *net); 114 void rt_flush_dev(struct net_device *dev); 115 struct rtable *__ip_route_output_key(struct net *, struct flowi4 *flp); 116 struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp, 117 struct sock *sk); 118 struct dst_entry *ipv4_blackhole_route(struct net *net, 119 struct dst_entry *dst_orig); 120 121 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp) 122 { 123 return ip_route_output_flow(net, flp, NULL); 124 } 125 126 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr, 127 __be32 saddr, u8 tos, int oif) 128 { 129 struct flowi4 fl4 = { 130 .flowi4_oif = oif, 131 .flowi4_tos = tos, 132 .daddr = daddr, 133 .saddr = saddr, 134 }; 135 return ip_route_output_key(net, &fl4); 136 } 137 138 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4, 139 struct sock *sk, 140 __be32 daddr, __be32 saddr, 141 __be16 dport, __be16 sport, 142 __u8 proto, __u8 tos, int oif) 143 { 144 flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos, 145 RT_SCOPE_UNIVERSE, proto, 146 sk ? inet_sk_flowi_flags(sk) : 0, 147 daddr, saddr, dport, sport); 148 if (sk) 149 security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 150 return ip_route_output_flow(net, fl4, sk); 151 } 152 153 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4, 154 __be32 daddr, __be32 saddr, 155 __be32 gre_key, __u8 tos, int oif) 156 { 157 memset(fl4, 0, sizeof(*fl4)); 158 fl4->flowi4_oif = oif; 159 fl4->daddr = daddr; 160 fl4->saddr = saddr; 161 fl4->flowi4_tos = tos; 162 fl4->flowi4_proto = IPPROTO_GRE; 163 fl4->fl4_gre_key = gre_key; 164 return ip_route_output_key(net, fl4); 165 } 166 167 int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src, 168 u8 tos, struct net_device *devin); 169 170 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src, 171 u8 tos, struct net_device *devin) 172 { 173 int err; 174 175 rcu_read_lock(); 176 err = ip_route_input_noref(skb, dst, src, tos, devin); 177 if (!err) 178 skb_dst_force(skb); 179 rcu_read_unlock(); 180 181 return err; 182 } 183 184 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif, 185 u32 mark, u8 protocol, int flow_flags); 186 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu); 187 void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark, 188 u8 protocol, int flow_flags); 189 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk); 190 void ip_rt_send_redirect(struct sk_buff *skb); 191 192 unsigned int inet_addr_type(struct net *net, __be32 addr); 193 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev, 194 __be32 addr); 195 void ip_rt_multicast_event(struct in_device *); 196 int ip_rt_ioctl(struct net *, unsigned int cmd, void __user *arg); 197 void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt); 198 int ip_rt_dump(struct sk_buff *skb, struct netlink_callback *cb); 199 200 struct in_ifaddr; 201 void fib_add_ifaddr(struct in_ifaddr *); 202 void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *); 203 204 static inline void ip_rt_put(struct rtable *rt) 205 { 206 /* dst_release() accepts a NULL parameter. 207 * We rely on dst being first structure in struct rtable 208 */ 209 BUILD_BUG_ON(offsetof(struct rtable, dst) != 0); 210 dst_release(&rt->dst); 211 } 212 213 #define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3) 214 215 extern const __u8 ip_tos2prio[16]; 216 217 static inline char rt_tos2priority(u8 tos) 218 { 219 return ip_tos2prio[IPTOS_TOS(tos)>>1]; 220 } 221 222 /* ip_route_connect() and ip_route_newports() work in tandem whilst 223 * binding a socket for a new outgoing connection. 224 * 225 * In order to use IPSEC properly, we must, in the end, have a 226 * route that was looked up using all available keys including source 227 * and destination ports. 228 * 229 * However, if a source port needs to be allocated (the user specified 230 * a wildcard source port) we need to obtain addressing information 231 * in order to perform that allocation. 232 * 233 * So ip_route_connect() looks up a route using wildcarded source and 234 * destination ports in the key, simply so that we can get a pair of 235 * addresses to use for port allocation. 236 * 237 * Later, once the ports are allocated, ip_route_newports() will make 238 * another route lookup if needed to make sure we catch any IPSEC 239 * rules keyed on the port information. 240 * 241 * The callers allocate the flow key on their stack, and must pass in 242 * the same flowi4 object to both the ip_route_connect() and the 243 * ip_route_newports() calls. 244 */ 245 246 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src, 247 u32 tos, int oif, u8 protocol, 248 __be16 sport, __be16 dport, 249 struct sock *sk, bool can_sleep) 250 { 251 __u8 flow_flags = 0; 252 253 if (inet_sk(sk)->transparent) 254 flow_flags |= FLOWI_FLAG_ANYSRC; 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 ip_rt_put(rt); 279 flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr); 280 } 281 security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 282 return ip_route_output_flow(net, fl4, sk); 283 } 284 285 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt, 286 __be16 orig_sport, __be16 orig_dport, 287 __be16 sport, __be16 dport, 288 struct sock *sk) 289 { 290 if (sport != orig_sport || dport != orig_dport) { 291 fl4->fl4_dport = dport; 292 fl4->fl4_sport = sport; 293 ip_rt_put(rt); 294 flowi4_update_output(fl4, sk->sk_bound_dev_if, 295 RT_CONN_FLAGS(sk), fl4->daddr, 296 fl4->saddr); 297 security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 298 return ip_route_output_flow(sock_net(sk), fl4, sk); 299 } 300 return rt; 301 } 302 303 static inline int inet_iif(const struct sk_buff *skb) 304 { 305 int iif = skb_rtable(skb)->rt_iif; 306 307 if (iif) 308 return iif; 309 return skb->skb_iif; 310 } 311 312 extern int sysctl_ip_default_ttl; 313 314 static inline int ip4_dst_hoplimit(const struct dst_entry *dst) 315 { 316 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT); 317 318 if (hoplimit == 0) 319 hoplimit = sysctl_ip_default_ttl; 320 return hoplimit; 321 } 322 323 static inline int ip_skb_dst_mtu(struct sk_buff *skb) 324 { 325 struct inet_sock *inet = skb->sk ? inet_sk(skb->sk) : NULL; 326 327 return (inet && inet->pmtudisc == IP_PMTUDISC_PROBE) ? 328 skb_dst(skb)->dev->mtu : dst_mtu(skb_dst(skb)); 329 } 330 331 #endif /* _ROUTE_H */ 332