xref: /linux/include/net/route.h (revision b0d5c81e872ed21de1e56feb0fa6e4161da7be61)
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/ip_fib.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_mtu_locked:1,
67 				rt_pmtu:31;
68 
69 	u32			rt_table_id;
70 
71 	struct list_head	rt_uncached;
72 	struct uncached_list	*rt_uncached_list;
73 };
74 
75 static inline bool rt_is_input_route(const struct rtable *rt)
76 {
77 	return rt->rt_is_input != 0;
78 }
79 
80 static inline bool rt_is_output_route(const struct rtable *rt)
81 {
82 	return rt->rt_is_input == 0;
83 }
84 
85 static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr)
86 {
87 	if (rt->rt_gateway)
88 		return rt->rt_gateway;
89 	return daddr;
90 }
91 
92 struct ip_rt_acct {
93 	__u32 	o_bytes;
94 	__u32 	o_packets;
95 	__u32 	i_bytes;
96 	__u32 	i_packets;
97 };
98 
99 struct rt_cache_stat {
100         unsigned int in_slow_tot;
101         unsigned int in_slow_mc;
102         unsigned int in_no_route;
103         unsigned int in_brd;
104         unsigned int in_martian_dst;
105         unsigned int in_martian_src;
106         unsigned int out_slow_tot;
107         unsigned int out_slow_mc;
108 };
109 
110 extern struct ip_rt_acct __percpu *ip_rt_acct;
111 
112 struct in_device;
113 
114 int ip_rt_init(void);
115 void rt_cache_flush(struct net *net);
116 void rt_flush_dev(struct net_device *dev);
117 struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp,
118 					const struct sk_buff *skb);
119 struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp,
120 					    struct fib_result *res,
121 					    const struct sk_buff *skb);
122 
123 static inline struct rtable *__ip_route_output_key(struct net *net,
124 						   struct flowi4 *flp)
125 {
126 	return ip_route_output_key_hash(net, flp, NULL);
127 }
128 
129 struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
130 				    const struct sock *sk);
131 struct dst_entry *ipv4_blackhole_route(struct net *net,
132 				       struct dst_entry *dst_orig);
133 
134 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
135 {
136 	return ip_route_output_flow(net, flp, NULL);
137 }
138 
139 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
140 					     __be32 saddr, u8 tos, int oif)
141 {
142 	struct flowi4 fl4 = {
143 		.flowi4_oif = oif,
144 		.flowi4_tos = tos,
145 		.daddr = daddr,
146 		.saddr = saddr,
147 	};
148 	return ip_route_output_key(net, &fl4);
149 }
150 
151 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
152 						   struct sock *sk,
153 						   __be32 daddr, __be32 saddr,
154 						   __be16 dport, __be16 sport,
155 						   __u8 proto, __u8 tos, int oif)
156 {
157 	flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos,
158 			   RT_SCOPE_UNIVERSE, proto,
159 			   sk ? inet_sk_flowi_flags(sk) : 0,
160 			   daddr, saddr, dport, sport, sock_net_uid(net, sk));
161 	if (sk)
162 		security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
163 	return ip_route_output_flow(net, fl4, sk);
164 }
165 
166 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4,
167 						 __be32 daddr, __be32 saddr,
168 						 __be32 gre_key, __u8 tos, int oif)
169 {
170 	memset(fl4, 0, sizeof(*fl4));
171 	fl4->flowi4_oif = oif;
172 	fl4->daddr = daddr;
173 	fl4->saddr = saddr;
174 	fl4->flowi4_tos = tos;
175 	fl4->flowi4_proto = IPPROTO_GRE;
176 	fl4->fl4_gre_key = gre_key;
177 	return ip_route_output_key(net, fl4);
178 }
179 int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr,
180 			  u8 tos, struct net_device *dev,
181 			  struct in_device *in_dev, u32 *itag);
182 int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src,
183 			 u8 tos, struct net_device *devin);
184 int ip_route_input_rcu(struct sk_buff *skb, __be32 dst, __be32 src,
185 		       u8 tos, struct net_device *devin,
186 		       struct fib_result *res);
187 
188 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
189 				 u8 tos, struct net_device *devin)
190 {
191 	int err;
192 
193 	rcu_read_lock();
194 	err = ip_route_input_noref(skb, dst, src, tos, devin);
195 	if (!err) {
196 		skb_dst_force(skb);
197 		if (!skb_dst(skb))
198 			err = -EINVAL;
199 	}
200 	rcu_read_unlock();
201 
202 	return err;
203 }
204 
205 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif,
206 		      u32 mark, u8 protocol, int flow_flags);
207 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
208 void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
209 		   u8 protocol, int flow_flags);
210 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
211 void ip_rt_send_redirect(struct sk_buff *skb);
212 
213 unsigned int inet_addr_type(struct net *net, __be32 addr);
214 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id);
215 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
216 				__be32 addr);
217 unsigned int inet_addr_type_dev_table(struct net *net,
218 				      const struct net_device *dev,
219 				      __be32 addr);
220 void ip_rt_multicast_event(struct in_device *);
221 int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt);
222 void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
223 struct rtable *rt_dst_alloc(struct net_device *dev,
224 			     unsigned int flags, u16 type,
225 			     bool nopolicy, bool noxfrm, bool will_cache);
226 
227 struct in_ifaddr;
228 void fib_add_ifaddr(struct in_ifaddr *);
229 void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
230 
231 void rt_add_uncached_list(struct rtable *rt);
232 void rt_del_uncached_list(struct rtable *rt);
233 
234 static inline void ip_rt_put(struct rtable *rt)
235 {
236 	/* dst_release() accepts a NULL parameter.
237 	 * We rely on dst being first structure in struct rtable
238 	 */
239 	BUILD_BUG_ON(offsetof(struct rtable, dst) != 0);
240 	dst_release(&rt->dst);
241 }
242 
243 #define IPTOS_RT_MASK	(IPTOS_TOS_MASK & ~3)
244 
245 extern const __u8 ip_tos2prio[16];
246 
247 static inline char rt_tos2priority(u8 tos)
248 {
249 	return ip_tos2prio[IPTOS_TOS(tos)>>1];
250 }
251 
252 /* ip_route_connect() and ip_route_newports() work in tandem whilst
253  * binding a socket for a new outgoing connection.
254  *
255  * In order to use IPSEC properly, we must, in the end, have a
256  * route that was looked up using all available keys including source
257  * and destination ports.
258  *
259  * However, if a source port needs to be allocated (the user specified
260  * a wildcard source port) we need to obtain addressing information
261  * in order to perform that allocation.
262  *
263  * So ip_route_connect() looks up a route using wildcarded source and
264  * destination ports in the key, simply so that we can get a pair of
265  * addresses to use for port allocation.
266  *
267  * Later, once the ports are allocated, ip_route_newports() will make
268  * another route lookup if needed to make sure we catch any IPSEC
269  * rules keyed on the port information.
270  *
271  * The callers allocate the flow key on their stack, and must pass in
272  * the same flowi4 object to both the ip_route_connect() and the
273  * ip_route_newports() calls.
274  */
275 
276 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src,
277 					 u32 tos, int oif, u8 protocol,
278 					 __be16 sport, __be16 dport,
279 					 struct sock *sk)
280 {
281 	__u8 flow_flags = 0;
282 
283 	if (inet_sk(sk)->transparent)
284 		flow_flags |= FLOWI_FLAG_ANYSRC;
285 
286 	flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE,
287 			   protocol, flow_flags, dst, src, dport, sport,
288 			   sk->sk_uid);
289 }
290 
291 static inline struct rtable *ip_route_connect(struct flowi4 *fl4,
292 					      __be32 dst, __be32 src, u32 tos,
293 					      int oif, u8 protocol,
294 					      __be16 sport, __be16 dport,
295 					      struct sock *sk)
296 {
297 	struct net *net = sock_net(sk);
298 	struct rtable *rt;
299 
300 	ip_route_connect_init(fl4, dst, src, tos, oif, protocol,
301 			      sport, dport, sk);
302 
303 	if (!dst || !src) {
304 		rt = __ip_route_output_key(net, fl4);
305 		if (IS_ERR(rt))
306 			return rt;
307 		ip_rt_put(rt);
308 		flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr);
309 	}
310 	security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
311 	return ip_route_output_flow(net, fl4, sk);
312 }
313 
314 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
315 					       __be16 orig_sport, __be16 orig_dport,
316 					       __be16 sport, __be16 dport,
317 					       struct sock *sk)
318 {
319 	if (sport != orig_sport || dport != orig_dport) {
320 		fl4->fl4_dport = dport;
321 		fl4->fl4_sport = sport;
322 		ip_rt_put(rt);
323 		flowi4_update_output(fl4, sk->sk_bound_dev_if,
324 				     RT_CONN_FLAGS(sk), fl4->daddr,
325 				     fl4->saddr);
326 		security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
327 		return ip_route_output_flow(sock_net(sk), fl4, sk);
328 	}
329 	return rt;
330 }
331 
332 static inline int inet_iif(const struct sk_buff *skb)
333 {
334 	struct rtable *rt = skb_rtable(skb);
335 
336 	if (rt && rt->rt_iif)
337 		return rt->rt_iif;
338 
339 	return skb->skb_iif;
340 }
341 
342 static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
343 {
344 	int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
345 	struct net *net = dev_net(dst->dev);
346 
347 	if (hoplimit == 0)
348 		hoplimit = net->ipv4.sysctl_ip_default_ttl;
349 	return hoplimit;
350 }
351 
352 #endif	/* _ROUTE_H */
353