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 inet_peer;
44 struct fib_info;
45 struct rtable {
46 	struct dst_entry	dst;
47 
48 	/* Lookup key. */
49 	__be32			rt_key_dst;
50 	__be32			rt_key_src;
51 
52 	int			rt_genid;
53 	unsigned		rt_flags;
54 	__u16			rt_type;
55 	__u8			rt_key_tos;
56 
57 	__be32			rt_dst;	/* Path destination	*/
58 	__be32			rt_src;	/* Path source		*/
59 	int			rt_route_iif;
60 	int			rt_iif;
61 	int			rt_oif;
62 	__u32			rt_mark;
63 
64 	/* Info on neighbour */
65 	__be32			rt_gateway;
66 
67 	/* Miscellaneous cached information */
68 	__be32			rt_spec_dst; /* RFC1122 specific destination */
69 	u32			rt_peer_genid;
70 	struct inet_peer	*peer; /* long-living peer info */
71 	struct fib_info		*fi; /* for client ref to shared metrics */
72 };
73 
rt_is_input_route(const struct rtable * rt)74 static inline bool rt_is_input_route(const struct rtable *rt)
75 {
76 	return rt->rt_route_iif != 0;
77 }
78 
rt_is_output_route(const struct rtable * rt)79 static inline bool rt_is_output_route(const struct rtable *rt)
80 {
81 	return rt->rt_route_iif == 0;
82 }
83 
84 struct ip_rt_acct {
85 	__u32 	o_bytes;
86 	__u32 	o_packets;
87 	__u32 	i_bytes;
88 	__u32 	i_packets;
89 };
90 
91 struct rt_cache_stat {
92         unsigned int in_hit;
93         unsigned int in_slow_tot;
94         unsigned int in_slow_mc;
95         unsigned int in_no_route;
96         unsigned int in_brd;
97         unsigned int in_martian_dst;
98         unsigned int in_martian_src;
99         unsigned int out_hit;
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;
108 };
109 
110 extern struct ip_rt_acct __percpu *ip_rt_acct;
111 
112 struct in_device;
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,
120 					   struct sock *sk);
121 extern struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig);
122 
ip_route_output_key(struct net * net,struct flowi4 * flp)123 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
124 {
125 	return ip_route_output_flow(net, flp, NULL);
126 }
127 
ip_route_output(struct net * net,__be32 daddr,__be32 saddr,u8 tos,int oif)128 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
129 					     __be32 saddr, u8 tos, int oif)
130 {
131 	struct flowi4 fl4 = {
132 		.flowi4_oif = oif,
133 		.daddr = daddr,
134 		.saddr = saddr,
135 		.flowi4_tos = tos,
136 	};
137 	return ip_route_output_key(net, &fl4);
138 }
139 
ip_route_output_ports(struct net * net,struct flowi4 * fl4,struct sock * sk,__be32 daddr,__be32 saddr,__be16 dport,__be16 sport,__u8 proto,__u8 tos,int oif)140 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
141 						   struct sock *sk,
142 						   __be32 daddr, __be32 saddr,
143 						   __be16 dport, __be16 sport,
144 						   __u8 proto, __u8 tos, int oif)
145 {
146 	flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos,
147 			   RT_SCOPE_UNIVERSE, proto,
148 			   sk ? inet_sk_flowi_flags(sk) : 0,
149 			   daddr, saddr, dport, sport);
150 	if (sk)
151 		security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
152 	return ip_route_output_flow(net, fl4, sk);
153 }
154 
ip_route_output_gre(struct net * net,struct flowi4 * fl4,__be32 daddr,__be32 saddr,__be32 gre_key,__u8 tos,int oif)155 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4,
156 						 __be32 daddr, __be32 saddr,
157 						 __be32 gre_key, __u8 tos, int oif)
158 {
159 	memset(fl4, 0, sizeof(*fl4));
160 	fl4->flowi4_oif = oif;
161 	fl4->daddr = daddr;
162 	fl4->saddr = saddr;
163 	fl4->flowi4_tos = tos;
164 	fl4->flowi4_proto = IPPROTO_GRE;
165 	fl4->fl4_gre_key = gre_key;
166 	return ip_route_output_key(net, fl4);
167 }
168 
169 extern int ip_route_input_common(struct sk_buff *skb, __be32 dst, __be32 src,
170 				 u8 tos, struct net_device *devin, bool noref);
171 
ip_route_input(struct sk_buff * skb,__be32 dst,__be32 src,u8 tos,struct net_device * devin)172 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
173 				 u8 tos, struct net_device *devin)
174 {
175 	return ip_route_input_common(skb, dst, src, tos, devin, false);
176 }
177 
ip_route_input_noref(struct sk_buff * skb,__be32 dst,__be32 src,u8 tos,struct net_device * devin)178 static inline int ip_route_input_noref(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, true);
182 }
183 
184 extern unsigned short	ip_rt_frag_needed(struct net *net, const struct iphdr *iph,
185 					  unsigned short new_mtu, struct net_device *dev);
186 extern void		ip_rt_send_redirect(struct sk_buff *skb);
187 
188 extern unsigned		inet_addr_type(struct net *net, __be32 addr);
189 extern unsigned		inet_dev_addr_type(struct net *net, const struct net_device *dev, __be32 addr);
190 extern void		ip_rt_multicast_event(struct in_device *);
191 extern int		ip_rt_ioctl(struct net *, unsigned int cmd, void __user *arg);
192 extern void		ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
193 extern int		ip_rt_dump(struct sk_buff *skb,  struct netlink_callback *cb);
194 
195 struct in_ifaddr;
196 extern void fib_add_ifaddr(struct in_ifaddr *);
197 extern void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
198 
ip_rt_put(struct rtable * rt)199 static inline void ip_rt_put(struct rtable * rt)
200 {
201 	if (rt)
202 		dst_release(&rt->dst);
203 }
204 
205 #define IPTOS_RT_MASK	(IPTOS_TOS_MASK & ~3)
206 
207 extern const __u8 ip_tos2prio[16];
208 
rt_tos2priority(u8 tos)209 static inline char rt_tos2priority(u8 tos)
210 {
211 	return ip_tos2prio[IPTOS_TOS(tos)>>1];
212 }
213 
214 /* ip_route_connect() and ip_route_newports() work in tandem whilst
215  * binding a socket for a new outgoing connection.
216  *
217  * In order to use IPSEC properly, we must, in the end, have a
218  * route that was looked up using all available keys including source
219  * and destination ports.
220  *
221  * However, if a source port needs to be allocated (the user specified
222  * a wildcard source port) we need to obtain addressing information
223  * in order to perform that allocation.
224  *
225  * So ip_route_connect() looks up a route using wildcarded source and
226  * destination ports in the key, simply so that we can get a pair of
227  * addresses to use for port allocation.
228  *
229  * Later, once the ports are allocated, ip_route_newports() will make
230  * another route lookup if needed to make sure we catch any IPSEC
231  * rules keyed on the port information.
232  *
233  * The callers allocate the flow key on their stack, and must pass in
234  * the same flowi4 object to both the ip_route_connect() and the
235  * ip_route_newports() calls.
236  */
237 
ip_route_connect_init(struct flowi4 * fl4,__be32 dst,__be32 src,u32 tos,int oif,u8 protocol,__be16 sport,__be16 dport,struct sock * sk,bool can_sleep)238 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src,
239 					 u32 tos, int oif, u8 protocol,
240 					 __be16 sport, __be16 dport,
241 					 struct sock *sk, bool can_sleep)
242 {
243 	__u8 flow_flags = 0;
244 
245 	if (inet_sk(sk)->transparent)
246 		flow_flags |= FLOWI_FLAG_ANYSRC;
247 	if (protocol == IPPROTO_TCP)
248 		flow_flags |= FLOWI_FLAG_PRECOW_METRICS;
249 	if (can_sleep)
250 		flow_flags |= FLOWI_FLAG_CAN_SLEEP;
251 
252 	flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE,
253 			   protocol, flow_flags, dst, src, dport, sport);
254 }
255 
ip_route_connect(struct flowi4 * fl4,__be32 dst,__be32 src,u32 tos,int oif,u8 protocol,__be16 sport,__be16 dport,struct sock * sk,bool can_sleep)256 static inline struct rtable *ip_route_connect(struct flowi4 *fl4,
257 					      __be32 dst, __be32 src, u32 tos,
258 					      int oif, u8 protocol,
259 					      __be16 sport, __be16 dport,
260 					      struct sock *sk, bool can_sleep)
261 {
262 	struct net *net = sock_net(sk);
263 	struct rtable *rt;
264 
265 	ip_route_connect_init(fl4, dst, src, tos, oif, protocol,
266 			      sport, dport, sk, can_sleep);
267 
268 	if (!dst || !src) {
269 		rt = __ip_route_output_key(net, fl4);
270 		if (IS_ERR(rt))
271 			return rt;
272 		ip_rt_put(rt);
273 		flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr);
274 	}
275 	security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
276 	return ip_route_output_flow(net, fl4, sk);
277 }
278 
ip_route_newports(struct flowi4 * fl4,struct rtable * rt,__be16 orig_sport,__be16 orig_dport,__be16 sport,__be16 dport,struct sock * sk)279 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
280 					       __be16 orig_sport, __be16 orig_dport,
281 					       __be16 sport, __be16 dport,
282 					       struct sock *sk)
283 {
284 	if (sport != orig_sport || dport != orig_dport) {
285 		fl4->fl4_dport = dport;
286 		fl4->fl4_sport = sport;
287 		ip_rt_put(rt);
288 		flowi4_update_output(fl4, sk->sk_bound_dev_if,
289 				     RT_CONN_FLAGS(sk), fl4->daddr,
290 				     fl4->saddr);
291 		security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
292 		return ip_route_output_flow(sock_net(sk), fl4, sk);
293 	}
294 	return rt;
295 }
296 
297 extern void rt_bind_peer(struct rtable *rt, __be32 daddr, int create);
298 
rt_get_peer(struct rtable * rt,__be32 daddr)299 static inline struct inet_peer *rt_get_peer(struct rtable *rt, __be32 daddr)
300 {
301 	if (rt->peer)
302 		return rt->peer;
303 
304 	rt_bind_peer(rt, daddr, 0);
305 	return rt->peer;
306 }
307 
inet_iif(const struct sk_buff * skb)308 static inline int inet_iif(const struct sk_buff *skb)
309 {
310 	return skb_rtable(skb)->rt_iif;
311 }
312 
313 extern int sysctl_ip_default_ttl;
314 
ip4_dst_hoplimit(const struct dst_entry * dst)315 static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
316 {
317 	int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
318 
319 	if (hoplimit == 0)
320 		hoplimit = sysctl_ip_default_ttl;
321 	return hoplimit;
322 }
323 
324 #endif	/* _ROUTE_H */
325