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 <net/l3mdev.h> 32 #include <linux/in_route.h> 33 #include <linux/rtnetlink.h> 34 #include <linux/rcupdate.h> 35 #include <linux/route.h> 36 #include <linux/ip.h> 37 #include <linux/cache.h> 38 #include <linux/security.h> 39 40 /* IPv4 datagram length is stored into 16bit field (tot_len) */ 41 #define IP_MAX_MTU 0xFFFFU 42 43 #define RTO_ONLINK 0x01 44 45 #define RT_CONN_FLAGS(sk) (RT_TOS(inet_sk(sk)->tos) | sock_flag(sk, SOCK_LOCALROUTE)) 46 #define RT_CONN_FLAGS_TOS(sk,tos) (RT_TOS(tos) | sock_flag(sk, SOCK_LOCALROUTE)) 47 48 struct fib_nh; 49 struct fib_info; 50 struct uncached_list; 51 struct rtable { 52 struct dst_entry dst; 53 54 int rt_genid; 55 unsigned int rt_flags; 56 __u16 rt_type; 57 __u8 rt_is_input; 58 __u8 rt_uses_gateway; 59 60 int rt_iif; 61 62 /* Info on neighbour */ 63 __be32 rt_gateway; 64 65 /* Miscellaneous cached information */ 66 u32 rt_pmtu; 67 68 u32 rt_table_id; 69 70 struct list_head rt_uncached; 71 struct uncached_list *rt_uncached_list; 72 }; 73 74 static inline bool rt_is_input_route(const struct rtable *rt) 75 { 76 return rt->rt_is_input != 0; 77 } 78 79 static inline bool rt_is_output_route(const struct rtable *rt) 80 { 81 return rt->rt_is_input == 0; 82 } 83 84 static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr) 85 { 86 if (rt->rt_gateway) 87 return rt->rt_gateway; 88 return daddr; 89 } 90 91 struct ip_rt_acct { 92 __u32 o_bytes; 93 __u32 o_packets; 94 __u32 i_bytes; 95 __u32 i_packets; 96 }; 97 98 struct rt_cache_stat { 99 unsigned int in_slow_tot; 100 unsigned int in_slow_mc; 101 unsigned int in_no_route; 102 unsigned int in_brd; 103 unsigned int in_martian_dst; 104 unsigned int in_martian_src; 105 unsigned int out_slow_tot; 106 unsigned int out_slow_mc; 107 }; 108 109 extern struct ip_rt_acct __percpu *ip_rt_acct; 110 111 struct in_device; 112 113 int ip_rt_init(void); 114 void rt_cache_flush(struct net *net); 115 void rt_flush_dev(struct net_device *dev); 116 struct rtable *__ip_route_output_key(struct net *, struct flowi4 *flp); 117 struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp, 118 const struct sock *sk); 119 struct dst_entry *ipv4_blackhole_route(struct net *net, 120 struct dst_entry *dst_orig); 121 122 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp) 123 { 124 return ip_route_output_flow(net, flp, NULL); 125 } 126 127 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr, 128 __be32 saddr, u8 tos, int oif) 129 { 130 struct flowi4 fl4 = { 131 .flowi4_oif = oif, 132 .flowi4_tos = tos, 133 .daddr = daddr, 134 .saddr = saddr, 135 }; 136 return ip_route_output_key(net, &fl4); 137 } 138 139 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4, 140 struct sock *sk, 141 __be32 daddr, __be32 saddr, 142 __be16 dport, __be16 sport, 143 __u8 proto, __u8 tos, int oif) 144 { 145 flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos, 146 RT_SCOPE_UNIVERSE, proto, 147 sk ? inet_sk_flowi_flags(sk) : 0, 148 daddr, saddr, dport, sport); 149 if (sk) 150 security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 151 return ip_route_output_flow(net, fl4, sk); 152 } 153 154 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4, 155 __be32 daddr, __be32 saddr, 156 __be32 gre_key, __u8 tos, int oif) 157 { 158 memset(fl4, 0, sizeof(*fl4)); 159 fl4->flowi4_oif = oif; 160 fl4->daddr = daddr; 161 fl4->saddr = saddr; 162 fl4->flowi4_tos = tos; 163 fl4->flowi4_proto = IPPROTO_GRE; 164 fl4->fl4_gre_key = gre_key; 165 return ip_route_output_key(net, fl4); 166 } 167 168 int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src, 169 u8 tos, struct net_device *devin); 170 171 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src, 172 u8 tos, struct net_device *devin) 173 { 174 int err; 175 176 rcu_read_lock(); 177 err = ip_route_input_noref(skb, dst, src, tos, devin); 178 if (!err) 179 skb_dst_force(skb); 180 rcu_read_unlock(); 181 182 return err; 183 } 184 185 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif, 186 u32 mark, u8 protocol, int flow_flags); 187 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu); 188 void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark, 189 u8 protocol, int flow_flags); 190 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk); 191 void ip_rt_send_redirect(struct sk_buff *skb); 192 193 unsigned int inet_addr_type(struct net *net, __be32 addr); 194 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id); 195 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev, 196 __be32 addr); 197 unsigned int inet_addr_type_dev_table(struct net *net, 198 const struct net_device *dev, 199 __be32 addr); 200 void ip_rt_multicast_event(struct in_device *); 201 int ip_rt_ioctl(struct net *, unsigned int cmd, void __user *arg); 202 void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt); 203 204 struct in_ifaddr; 205 void fib_add_ifaddr(struct in_ifaddr *); 206 void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *); 207 208 static inline void ip_rt_put(struct rtable *rt) 209 { 210 /* dst_release() accepts a NULL parameter. 211 * We rely on dst being first structure in struct rtable 212 */ 213 BUILD_BUG_ON(offsetof(struct rtable, dst) != 0); 214 dst_release(&rt->dst); 215 } 216 217 #define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3) 218 219 extern const __u8 ip_tos2prio[16]; 220 221 static inline char rt_tos2priority(u8 tos) 222 { 223 return ip_tos2prio[IPTOS_TOS(tos)>>1]; 224 } 225 226 /* ip_route_connect() and ip_route_newports() work in tandem whilst 227 * binding a socket for a new outgoing connection. 228 * 229 * In order to use IPSEC properly, we must, in the end, have a 230 * route that was looked up using all available keys including source 231 * and destination ports. 232 * 233 * However, if a source port needs to be allocated (the user specified 234 * a wildcard source port) we need to obtain addressing information 235 * in order to perform that allocation. 236 * 237 * So ip_route_connect() looks up a route using wildcarded source and 238 * destination ports in the key, simply so that we can get a pair of 239 * addresses to use for port allocation. 240 * 241 * Later, once the ports are allocated, ip_route_newports() will make 242 * another route lookup if needed to make sure we catch any IPSEC 243 * rules keyed on the port information. 244 * 245 * The callers allocate the flow key on their stack, and must pass in 246 * the same flowi4 object to both the ip_route_connect() and the 247 * ip_route_newports() calls. 248 */ 249 250 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src, 251 u32 tos, int oif, u8 protocol, 252 __be16 sport, __be16 dport, 253 struct sock *sk) 254 { 255 __u8 flow_flags = 0; 256 257 if (inet_sk(sk)->transparent) 258 flow_flags |= FLOWI_FLAG_ANYSRC; 259 260 if (netif_index_is_l3_master(sock_net(sk), oif)) 261 flow_flags |= FLOWI_FLAG_VRFSRC | FLOWI_FLAG_SKIP_NH_OIF; 262 263 flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE, 264 protocol, flow_flags, dst, src, dport, sport); 265 } 266 267 static inline struct rtable *ip_route_connect(struct flowi4 *fl4, 268 __be32 dst, __be32 src, u32 tos, 269 int oif, u8 protocol, 270 __be16 sport, __be16 dport, 271 struct sock *sk) 272 { 273 struct net *net = sock_net(sk); 274 struct rtable *rt; 275 276 ip_route_connect_init(fl4, dst, src, tos, oif, protocol, 277 sport, dport, sk); 278 279 if (!dst || !src) { 280 rt = __ip_route_output_key(net, fl4); 281 if (IS_ERR(rt)) 282 return rt; 283 ip_rt_put(rt); 284 flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr); 285 } 286 security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 287 return ip_route_output_flow(net, fl4, sk); 288 } 289 290 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt, 291 __be16 orig_sport, __be16 orig_dport, 292 __be16 sport, __be16 dport, 293 struct sock *sk) 294 { 295 if (sport != orig_sport || dport != orig_dport) { 296 fl4->fl4_dport = dport; 297 fl4->fl4_sport = sport; 298 ip_rt_put(rt); 299 flowi4_update_output(fl4, sk->sk_bound_dev_if, 300 RT_CONN_FLAGS(sk), fl4->daddr, 301 fl4->saddr); 302 security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 303 return ip_route_output_flow(sock_net(sk), fl4, sk); 304 } 305 return rt; 306 } 307 308 static inline int inet_iif(const struct sk_buff *skb) 309 { 310 int iif = skb_rtable(skb)->rt_iif; 311 312 if (iif) 313 return iif; 314 return skb->skb_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