xref: /openbmc/linux/include/net/route.h (revision 0c874100)
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 	struct list_head	rt_uncached;
70 	struct uncached_list	*rt_uncached_list;
71 };
72 
73 static inline bool rt_is_input_route(const struct rtable *rt)
74 {
75 	return rt->rt_is_input != 0;
76 }
77 
78 static inline bool rt_is_output_route(const struct rtable *rt)
79 {
80 	return rt->rt_is_input == 0;
81 }
82 
83 static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr)
84 {
85 	if (rt->rt_gateway)
86 		return rt->rt_gateway;
87 	return daddr;
88 }
89 
90 struct ip_rt_acct {
91 	__u32 	o_bytes;
92 	__u32 	o_packets;
93 	__u32 	i_bytes;
94 	__u32 	i_packets;
95 };
96 
97 struct rt_cache_stat {
98         unsigned int in_slow_tot;
99         unsigned int in_slow_mc;
100         unsigned int in_no_route;
101         unsigned int in_brd;
102         unsigned int in_martian_dst;
103         unsigned int in_martian_src;
104         unsigned int out_slow_tot;
105         unsigned int out_slow_mc;
106 };
107 
108 extern struct ip_rt_acct __percpu *ip_rt_acct;
109 
110 struct in_device;
111 
112 int ip_rt_init(void);
113 void rt_cache_flush(struct net *net);
114 void rt_flush_dev(struct net_device *dev);
115 struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp,
116 					const struct sk_buff *skb);
117 struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp,
118 					    struct fib_result *res,
119 					    const struct sk_buff *skb);
120 
121 static inline struct rtable *__ip_route_output_key(struct net *net,
122 						   struct flowi4 *flp)
123 {
124 	return ip_route_output_key_hash(net, flp, NULL);
125 }
126 
127 struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
128 				    const struct sock *sk);
129 struct dst_entry *ipv4_blackhole_route(struct net *net,
130 				       struct dst_entry *dst_orig);
131 
132 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
133 {
134 	return ip_route_output_flow(net, flp, NULL);
135 }
136 
137 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
138 					     __be32 saddr, u8 tos, int oif)
139 {
140 	struct flowi4 fl4 = {
141 		.flowi4_oif = oif,
142 		.flowi4_tos = tos,
143 		.daddr = daddr,
144 		.saddr = saddr,
145 	};
146 	return ip_route_output_key(net, &fl4);
147 }
148 
149 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
150 						   struct sock *sk,
151 						   __be32 daddr, __be32 saddr,
152 						   __be16 dport, __be16 sport,
153 						   __u8 proto, __u8 tos, int oif)
154 {
155 	flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos,
156 			   RT_SCOPE_UNIVERSE, proto,
157 			   sk ? inet_sk_flowi_flags(sk) : 0,
158 			   daddr, saddr, dport, sport, sock_net_uid(net, sk));
159 	if (sk)
160 		security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
161 	return ip_route_output_flow(net, fl4, sk);
162 }
163 
164 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4,
165 						 __be32 daddr, __be32 saddr,
166 						 __be32 gre_key, __u8 tos, int oif)
167 {
168 	memset(fl4, 0, sizeof(*fl4));
169 	fl4->flowi4_oif = oif;
170 	fl4->daddr = daddr;
171 	fl4->saddr = saddr;
172 	fl4->flowi4_tos = tos;
173 	fl4->flowi4_proto = IPPROTO_GRE;
174 	fl4->fl4_gre_key = gre_key;
175 	return ip_route_output_key(net, fl4);
176 }
177 int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr,
178 			  u8 tos, struct net_device *dev,
179 			  struct in_device *in_dev, u32 *itag);
180 int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src,
181 			 u8 tos, struct net_device *devin);
182 int ip_route_input_rcu(struct sk_buff *skb, __be32 dst, __be32 src,
183 		       u8 tos, struct net_device *devin,
184 		       struct fib_result *res);
185 
186 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
187 				 u8 tos, struct net_device *devin)
188 {
189 	int err;
190 
191 	rcu_read_lock();
192 	err = ip_route_input_noref(skb, dst, src, tos, devin);
193 	if (!err) {
194 		skb_dst_force(skb);
195 		if (!skb_dst(skb))
196 			err = -EINVAL;
197 	}
198 	rcu_read_unlock();
199 
200 	return err;
201 }
202 
203 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif,
204 		      u8 protocol);
205 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
206 void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol);
207 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
208 void ip_rt_send_redirect(struct sk_buff *skb);
209 
210 unsigned int inet_addr_type(struct net *net, __be32 addr);
211 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id);
212 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
213 				__be32 addr);
214 unsigned int inet_addr_type_dev_table(struct net *net,
215 				      const struct net_device *dev,
216 				      __be32 addr);
217 void ip_rt_multicast_event(struct in_device *);
218 int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt);
219 void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
220 struct rtable *rt_dst_alloc(struct net_device *dev,
221 			     unsigned int flags, u16 type,
222 			     bool nopolicy, bool noxfrm, bool will_cache);
223 
224 struct in_ifaddr;
225 void fib_add_ifaddr(struct in_ifaddr *);
226 void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
227 void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric);
228 
229 void rt_add_uncached_list(struct rtable *rt);
230 void rt_del_uncached_list(struct rtable *rt);
231 
232 static inline void ip_rt_put(struct rtable *rt)
233 {
234 	/* dst_release() accepts a NULL parameter.
235 	 * We rely on dst being first structure in struct rtable
236 	 */
237 	BUILD_BUG_ON(offsetof(struct rtable, dst) != 0);
238 	dst_release(&rt->dst);
239 }
240 
241 #define IPTOS_RT_MASK	(IPTOS_TOS_MASK & ~3)
242 
243 extern const __u8 ip_tos2prio[16];
244 
245 static inline char rt_tos2priority(u8 tos)
246 {
247 	return ip_tos2prio[IPTOS_TOS(tos)>>1];
248 }
249 
250 /* ip_route_connect() and ip_route_newports() work in tandem whilst
251  * binding a socket for a new outgoing connection.
252  *
253  * In order to use IPSEC properly, we must, in the end, have a
254  * route that was looked up using all available keys including source
255  * and destination ports.
256  *
257  * However, if a source port needs to be allocated (the user specified
258  * a wildcard source port) we need to obtain addressing information
259  * in order to perform that allocation.
260  *
261  * So ip_route_connect() looks up a route using wildcarded source and
262  * destination ports in the key, simply so that we can get a pair of
263  * addresses to use for port allocation.
264  *
265  * Later, once the ports are allocated, ip_route_newports() will make
266  * another route lookup if needed to make sure we catch any IPSEC
267  * rules keyed on the port information.
268  *
269  * The callers allocate the flow key on their stack, and must pass in
270  * the same flowi4 object to both the ip_route_connect() and the
271  * ip_route_newports() calls.
272  */
273 
274 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src,
275 					 u32 tos, int oif, u8 protocol,
276 					 __be16 sport, __be16 dport,
277 					 struct sock *sk)
278 {
279 	__u8 flow_flags = 0;
280 
281 	if (inet_sk(sk)->transparent)
282 		flow_flags |= FLOWI_FLAG_ANYSRC;
283 
284 	flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE,
285 			   protocol, flow_flags, dst, src, dport, sport,
286 			   sk->sk_uid);
287 }
288 
289 static inline struct rtable *ip_route_connect(struct flowi4 *fl4,
290 					      __be32 dst, __be32 src, u32 tos,
291 					      int oif, u8 protocol,
292 					      __be16 sport, __be16 dport,
293 					      struct sock *sk)
294 {
295 	struct net *net = sock_net(sk);
296 	struct rtable *rt;
297 
298 	ip_route_connect_init(fl4, dst, src, tos, oif, protocol,
299 			      sport, dport, sk);
300 
301 	if (!dst || !src) {
302 		rt = __ip_route_output_key(net, fl4);
303 		if (IS_ERR(rt))
304 			return rt;
305 		ip_rt_put(rt);
306 		flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr);
307 	}
308 	security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
309 	return ip_route_output_flow(net, fl4, sk);
310 }
311 
312 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
313 					       __be16 orig_sport, __be16 orig_dport,
314 					       __be16 sport, __be16 dport,
315 					       struct sock *sk)
316 {
317 	if (sport != orig_sport || dport != orig_dport) {
318 		fl4->fl4_dport = dport;
319 		fl4->fl4_sport = sport;
320 		ip_rt_put(rt);
321 		flowi4_update_output(fl4, sk->sk_bound_dev_if,
322 				     RT_CONN_FLAGS(sk), fl4->daddr,
323 				     fl4->saddr);
324 		security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
325 		return ip_route_output_flow(sock_net(sk), fl4, sk);
326 	}
327 	return rt;
328 }
329 
330 static inline int inet_iif(const struct sk_buff *skb)
331 {
332 	struct rtable *rt = skb_rtable(skb);
333 
334 	if (rt && rt->rt_iif)
335 		return rt->rt_iif;
336 
337 	return skb->skb_iif;
338 }
339 
340 static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
341 {
342 	int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
343 	struct net *net = dev_net(dst->dev);
344 
345 	if (hoplimit == 0)
346 		hoplimit = net->ipv4.sysctl_ip_default_ttl;
347 	return hoplimit;
348 }
349 
350 #endif	/* _ROUTE_H */
351