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.
6 * Definitions for the IP router.
8 * Version: @(#)route.h 1.0.4 05/27/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
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
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.
28 #include <net/inetpeer.h>
30 #include <net/inet_sock.h>
31 #include <linux/in_route.h>
32 #include <linux/rtnetlink.h>
33 #include <linux/route.h>
35 #include <linux/cache.h>
36 #include <linux/security.h>
38 #define RTO_ONLINK 0x01
40 #define RT_CONN_FLAGS(sk) (RT_TOS(inet_sk(sk)->tos) | sock_flag(sk, SOCK_LOCALROUTE))
57 __be32 rt_dst; /* Path destination */
58 __be32 rt_src; /* Path source */
64 /* Info on neighbour */
67 /* Miscellaneous cached information */
68 __be32 rt_spec_dst; /* RFC1122 specific destination */
70 struct inet_peer *peer; /* long-living peer info */
71 struct fib_info *fi; /* for client ref to shared metrics */
74 static inline bool rt_is_input_route(struct rtable *rt)
76 return rt->rt_route_iif != 0;
79 static inline bool rt_is_output_route(struct rtable *rt)
81 return rt->rt_route_iif == 0;
91 struct rt_cache_stat {
93 unsigned int in_slow_tot;
94 unsigned int in_slow_mc;
95 unsigned int in_no_route;
97 unsigned int in_martian_dst;
98 unsigned int in_martian_src;
100 unsigned int out_slow_tot;
101 unsigned int out_slow_mc;
102 unsigned int gc_total;
103 unsigned int gc_ignored;
104 unsigned int gc_goal_miss;
105 unsigned int gc_dst_overflow;
106 unsigned int in_hlist_search;
107 unsigned int out_hlist_search;
110 extern struct ip_rt_acct __percpu *ip_rt_acct;
113 extern int ip_rt_init(void);
114 extern void ip_rt_redirect(__be32 old_gw, __be32 dst, __be32 new_gw,
115 __be32 src, struct net_device *dev);
116 extern void rt_cache_flush(struct net *net, int how);
117 extern void rt_cache_flush_batch(struct net *net);
118 extern struct rtable *__ip_route_output_key(struct net *, struct flowi4 *flp);
119 extern struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
121 extern struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig);
123 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
125 return ip_route_output_flow(net, flp, NULL);
128 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
129 __be32 saddr, u8 tos, int oif)
131 struct flowi4 fl4 = {
137 return ip_route_output_key(net, &fl4);
140 static inline struct rtable *ip_route_output_ports(struct net *net, struct sock *sk,
141 __be32 daddr, __be32 saddr,
142 __be16 dport, __be16 sport,
143 __u8 proto, __u8 tos, int oif)
147 flowi4_init_output(&fl4, oif, sk ? sk->sk_mark : 0, tos,
148 RT_SCOPE_UNIVERSE, proto,
149 sk ? inet_sk_flowi_flags(sk) : 0,
150 daddr, saddr, dport, sport);
152 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
153 return ip_route_output_flow(net, &fl4, sk);
156 static inline struct rtable *ip_route_output_gre(struct net *net,
157 __be32 daddr, __be32 saddr,
158 __be32 gre_key, __u8 tos, int oif)
160 struct flowi4 fl4 = {
165 .flowi4_proto = IPPROTO_GRE,
166 .fl4_gre_key = gre_key,
168 return ip_route_output_key(net, &fl4);
171 extern int ip_route_input_common(struct sk_buff *skb, __be32 dst, __be32 src,
172 u8 tos, struct net_device *devin, bool noref);
174 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
175 u8 tos, struct net_device *devin)
177 return ip_route_input_common(skb, dst, src, tos, devin, false);
180 static inline int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src,
181 u8 tos, struct net_device *devin)
183 return ip_route_input_common(skb, dst, src, tos, devin, true);
186 extern unsigned short ip_rt_frag_needed(struct net *net, const struct iphdr *iph,
187 unsigned short new_mtu, struct net_device *dev);
188 extern void ip_rt_send_redirect(struct sk_buff *skb);
190 extern unsigned inet_addr_type(struct net *net, __be32 addr);
191 extern unsigned inet_dev_addr_type(struct net *net, const struct net_device *dev, __be32 addr);
192 extern void ip_rt_multicast_event(struct in_device *);
193 extern int ip_rt_ioctl(struct net *, unsigned int cmd, void __user *arg);
194 extern void ip_rt_get_source(u8 *src, struct rtable *rt);
195 extern int ip_rt_dump(struct sk_buff *skb, struct netlink_callback *cb);
198 extern void fib_add_ifaddr(struct in_ifaddr *);
199 extern void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
201 static inline void ip_rt_put(struct rtable * rt)
204 dst_release(&rt->dst);
207 #define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3)
209 extern const __u8 ip_tos2prio[16];
211 static inline char rt_tos2priority(u8 tos)
213 return ip_tos2prio[IPTOS_TOS(tos)>>1];
216 /* ip_route_connect() and ip_route_newports() work in tandem whilst
217 * binding a socket for a new outgoing connection.
219 * In order to use IPSEC properly, we must, in the end, have a
220 * route that was looked up using all available keys including source
221 * and destination ports.
223 * However, if a source port needs to be allocated (the user specified
224 * a wildcard source port) we need to obtain addressing information
225 * in order to perform that allocation.
227 * So ip_route_connect() looks up a route using wildcarded source and
228 * destination ports in the key, simply so that we can get a pair of
229 * addresses to use for port allocation.
231 * Later, once the ports are allocated, ip_route_newports() will make
232 * another route lookup if needed to make sure we catch any IPSEC
233 * rules keyed on the port information.
235 * The callers allocate the flow key on their stack, and must pass in
236 * the same flowi4 object to both the ip_route_connect() and the
237 * ip_route_newports() calls.
240 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src,
241 u32 tos, int oif, u8 protocol,
242 __be16 sport, __be16 dport,
243 struct sock *sk, bool can_sleep)
247 if (inet_sk(sk)->transparent)
248 flow_flags |= FLOWI_FLAG_ANYSRC;
249 if (protocol == IPPROTO_TCP)
250 flow_flags |= FLOWI_FLAG_PRECOW_METRICS;
252 flow_flags |= FLOWI_FLAG_CAN_SLEEP;
254 flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE,
255 protocol, flow_flags, dst, src, dport, sport);
258 static inline struct rtable *ip_route_connect(struct flowi4 *fl4,
259 __be32 dst, __be32 src, u32 tos,
260 int oif, u8 protocol,
261 __be16 sport, __be16 dport,
262 struct sock *sk, bool can_sleep)
264 struct net *net = sock_net(sk);
267 ip_route_connect_init(fl4, dst, src, tos, oif, protocol,
268 sport, dport, sk, can_sleep);
271 rt = __ip_route_output_key(net, fl4);
274 fl4->daddr = rt->rt_dst;
275 fl4->saddr = rt->rt_src;
278 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
279 return ip_route_output_flow(net, fl4, sk);
282 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
283 __be16 orig_sport, __be16 orig_dport,
284 __be16 sport, __be16 dport,
287 if (sport != orig_sport || dport != orig_dport) {
288 fl4->fl4_dport = dport;
289 fl4->fl4_sport = sport;
291 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
292 return ip_route_output_flow(sock_net(sk), fl4, sk);
297 extern void rt_bind_peer(struct rtable *rt, int create);
299 static inline struct inet_peer *rt_get_peer(struct rtable *rt)
308 static inline int inet_iif(const struct sk_buff *skb)
310 return skb_rtable(skb)->rt_iif;
313 extern int sysctl_ip_default_ttl;
315 static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
317 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
320 hoplimit = sysctl_ip_default_ttl;
324 #endif /* _ROUTE_H */