xref: /linux/include/net/route.h (revision b24413180f5600bcb3bb70fbed5cf186b60864bd)
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_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_hash(struct net *net, struct flowi4 *flp,
117 					const struct sk_buff *skb);
118 struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp,
119 					    struct fib_result *res,
120 					    const struct sk_buff *skb);
121 
122 static inline struct rtable *__ip_route_output_key(struct net *net,
123 						   struct flowi4 *flp)
124 {
125 	return ip_route_output_key_hash(net, flp, NULL);
126 }
127 
128 struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
129 				    const struct sock *sk);
130 struct dst_entry *ipv4_blackhole_route(struct net *net,
131 				       struct dst_entry *dst_orig);
132 
133 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
134 {
135 	return ip_route_output_flow(net, flp, NULL);
136 }
137 
138 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
139 					     __be32 saddr, u8 tos, int oif)
140 {
141 	struct flowi4 fl4 = {
142 		.flowi4_oif = oif,
143 		.flowi4_tos = tos,
144 		.daddr = daddr,
145 		.saddr = saddr,
146 	};
147 	return ip_route_output_key(net, &fl4);
148 }
149 
150 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
151 						   struct sock *sk,
152 						   __be32 daddr, __be32 saddr,
153 						   __be16 dport, __be16 sport,
154 						   __u8 proto, __u8 tos, int oif)
155 {
156 	flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos,
157 			   RT_SCOPE_UNIVERSE, proto,
158 			   sk ? inet_sk_flowi_flags(sk) : 0,
159 			   daddr, saddr, dport, sport, sock_net_uid(net, sk));
160 	if (sk)
161 		security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
162 	return ip_route_output_flow(net, fl4, sk);
163 }
164 
165 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4,
166 						 __be32 daddr, __be32 saddr,
167 						 __be32 gre_key, __u8 tos, int oif)
168 {
169 	memset(fl4, 0, sizeof(*fl4));
170 	fl4->flowi4_oif = oif;
171 	fl4->daddr = daddr;
172 	fl4->saddr = saddr;
173 	fl4->flowi4_tos = tos;
174 	fl4->flowi4_proto = IPPROTO_GRE;
175 	fl4->fl4_gre_key = gre_key;
176 	return ip_route_output_key(net, fl4);
177 }
178 int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr,
179 			  u8 tos, struct net_device *dev,
180 			  struct in_device *in_dev, u32 *itag);
181 int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src,
182 			 u8 tos, struct net_device *devin);
183 int ip_route_input_rcu(struct sk_buff *skb, __be32 dst, __be32 src,
184 		       u8 tos, struct net_device *devin,
185 		       struct fib_result *res);
186 
187 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
188 				 u8 tos, struct net_device *devin)
189 {
190 	int err;
191 
192 	rcu_read_lock();
193 	err = ip_route_input_noref(skb, dst, src, tos, devin);
194 	if (!err) {
195 		skb_dst_force(skb);
196 		if (!skb_dst(skb))
197 			err = -EINVAL;
198 	}
199 	rcu_read_unlock();
200 
201 	return err;
202 }
203 
204 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif,
205 		      u32 mark, u8 protocol, int flow_flags);
206 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
207 void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
208 		   u8 protocol, int flow_flags);
209 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
210 void ip_rt_send_redirect(struct sk_buff *skb);
211 
212 unsigned int inet_addr_type(struct net *net, __be32 addr);
213 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id);
214 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
215 				__be32 addr);
216 unsigned int inet_addr_type_dev_table(struct net *net,
217 				      const struct net_device *dev,
218 				      __be32 addr);
219 void ip_rt_multicast_event(struct in_device *);
220 int ip_rt_ioctl(struct net *, unsigned int cmd, void __user *arg);
221 void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
222 struct rtable *rt_dst_alloc(struct net_device *dev,
223 			     unsigned int flags, u16 type,
224 			     bool nopolicy, bool noxfrm, bool will_cache);
225 
226 struct in_ifaddr;
227 void fib_add_ifaddr(struct in_ifaddr *);
228 void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
229 
230 static inline void ip_rt_put(struct rtable *rt)
231 {
232 	/* dst_release() accepts a NULL parameter.
233 	 * We rely on dst being first structure in struct rtable
234 	 */
235 	BUILD_BUG_ON(offsetof(struct rtable, dst) != 0);
236 	dst_release(&rt->dst);
237 }
238 
239 #define IPTOS_RT_MASK	(IPTOS_TOS_MASK & ~3)
240 
241 extern const __u8 ip_tos2prio[16];
242 
243 static inline char rt_tos2priority(u8 tos)
244 {
245 	return ip_tos2prio[IPTOS_TOS(tos)>>1];
246 }
247 
248 /* ip_route_connect() and ip_route_newports() work in tandem whilst
249  * binding a socket for a new outgoing connection.
250  *
251  * In order to use IPSEC properly, we must, in the end, have a
252  * route that was looked up using all available keys including source
253  * and destination ports.
254  *
255  * However, if a source port needs to be allocated (the user specified
256  * a wildcard source port) we need to obtain addressing information
257  * in order to perform that allocation.
258  *
259  * So ip_route_connect() looks up a route using wildcarded source and
260  * destination ports in the key, simply so that we can get a pair of
261  * addresses to use for port allocation.
262  *
263  * Later, once the ports are allocated, ip_route_newports() will make
264  * another route lookup if needed to make sure we catch any IPSEC
265  * rules keyed on the port information.
266  *
267  * The callers allocate the flow key on their stack, and must pass in
268  * the same flowi4 object to both the ip_route_connect() and the
269  * ip_route_newports() calls.
270  */
271 
272 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src,
273 					 u32 tos, int oif, u8 protocol,
274 					 __be16 sport, __be16 dport,
275 					 struct sock *sk)
276 {
277 	__u8 flow_flags = 0;
278 
279 	if (inet_sk(sk)->transparent)
280 		flow_flags |= FLOWI_FLAG_ANYSRC;
281 
282 	flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE,
283 			   protocol, flow_flags, dst, src, dport, sport,
284 			   sk->sk_uid);
285 }
286 
287 static inline struct rtable *ip_route_connect(struct flowi4 *fl4,
288 					      __be32 dst, __be32 src, u32 tos,
289 					      int oif, u8 protocol,
290 					      __be16 sport, __be16 dport,
291 					      struct sock *sk)
292 {
293 	struct net *net = sock_net(sk);
294 	struct rtable *rt;
295 
296 	ip_route_connect_init(fl4, dst, src, tos, oif, protocol,
297 			      sport, dport, sk);
298 
299 	if (!dst || !src) {
300 		rt = __ip_route_output_key(net, fl4);
301 		if (IS_ERR(rt))
302 			return rt;
303 		ip_rt_put(rt);
304 		flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr);
305 	}
306 	security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
307 	return ip_route_output_flow(net, fl4, sk);
308 }
309 
310 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
311 					       __be16 orig_sport, __be16 orig_dport,
312 					       __be16 sport, __be16 dport,
313 					       struct sock *sk)
314 {
315 	if (sport != orig_sport || dport != orig_dport) {
316 		fl4->fl4_dport = dport;
317 		fl4->fl4_sport = sport;
318 		ip_rt_put(rt);
319 		flowi4_update_output(fl4, sk->sk_bound_dev_if,
320 				     RT_CONN_FLAGS(sk), fl4->daddr,
321 				     fl4->saddr);
322 		security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
323 		return ip_route_output_flow(sock_net(sk), fl4, sk);
324 	}
325 	return rt;
326 }
327 
328 static inline int inet_iif(const struct sk_buff *skb)
329 {
330 	struct rtable *rt = skb_rtable(skb);
331 
332 	if (rt && rt->rt_iif)
333 		return rt->rt_iif;
334 
335 	return skb->skb_iif;
336 }
337 
338 static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
339 {
340 	int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
341 	struct net *net = dev_net(dst->dev);
342 
343 	if (hoplimit == 0)
344 		hoplimit = net->ipv4.sysctl_ip_default_ttl;
345 	return hoplimit;
346 }
347 
348 #endif	/* _ROUTE_H */
349