xref: /openbmc/linux/include/net/route.h (revision 9917e1e8)
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 fib_info;
44 struct rtable {
45 	struct dst_entry	dst;
46 
47 	int			rt_genid;
48 	unsigned int		rt_flags;
49 	__u16			rt_type;
50 	__u16			rt_is_input;
51 
52 	int			rt_iif;
53 
54 	/* Info on neighbour */
55 	__be32			rt_gateway;
56 
57 	/* Miscellaneous cached information */
58 	u32			rt_pmtu;
59 	struct fib_info		*fi; /* for client ref to shared metrics */
60 };
61 
62 static inline bool rt_is_input_route(const struct rtable *rt)
63 {
64 	return rt->rt_is_input != 0;
65 }
66 
67 static inline bool rt_is_output_route(const struct rtable *rt)
68 {
69 	return rt->rt_is_input == 0;
70 }
71 
72 static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr)
73 {
74 	if (rt->rt_gateway)
75 		return rt->rt_gateway;
76 	return daddr;
77 }
78 
79 struct ip_rt_acct {
80 	__u32 	o_bytes;
81 	__u32 	o_packets;
82 	__u32 	i_bytes;
83 	__u32 	i_packets;
84 };
85 
86 struct rt_cache_stat {
87         unsigned int in_hit;
88         unsigned int in_slow_tot;
89         unsigned int in_slow_mc;
90         unsigned int in_no_route;
91         unsigned int in_brd;
92         unsigned int in_martian_dst;
93         unsigned int in_martian_src;
94         unsigned int out_hit;
95         unsigned int out_slow_tot;
96         unsigned int out_slow_mc;
97         unsigned int gc_total;
98         unsigned int gc_ignored;
99         unsigned int gc_goal_miss;
100         unsigned int gc_dst_overflow;
101         unsigned int in_hlist_search;
102         unsigned int out_hlist_search;
103 };
104 
105 extern struct ip_rt_acct __percpu *ip_rt_acct;
106 
107 struct in_device;
108 extern int		ip_rt_init(void);
109 extern void		rt_cache_flush(struct net *net, int how);
110 extern struct rtable *__ip_route_output_key(struct net *, struct flowi4 *flp);
111 extern struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
112 					   struct sock *sk);
113 extern struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig);
114 
115 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
116 {
117 	return ip_route_output_flow(net, flp, NULL);
118 }
119 
120 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
121 					     __be32 saddr, u8 tos, int oif)
122 {
123 	struct flowi4 fl4 = {
124 		.flowi4_oif = oif,
125 		.flowi4_tos = tos,
126 		.daddr = daddr,
127 		.saddr = saddr,
128 	};
129 	return ip_route_output_key(net, &fl4);
130 }
131 
132 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
133 						   struct sock *sk,
134 						   __be32 daddr, __be32 saddr,
135 						   __be16 dport, __be16 sport,
136 						   __u8 proto, __u8 tos, int oif)
137 {
138 	flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos,
139 			   RT_SCOPE_UNIVERSE, proto,
140 			   sk ? inet_sk_flowi_flags(sk) : 0,
141 			   daddr, saddr, dport, sport);
142 	if (sk)
143 		security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
144 	return ip_route_output_flow(net, fl4, sk);
145 }
146 
147 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4,
148 						 __be32 daddr, __be32 saddr,
149 						 __be32 gre_key, __u8 tos, int oif)
150 {
151 	memset(fl4, 0, sizeof(*fl4));
152 	fl4->flowi4_oif = oif;
153 	fl4->daddr = daddr;
154 	fl4->saddr = saddr;
155 	fl4->flowi4_tos = tos;
156 	fl4->flowi4_proto = IPPROTO_GRE;
157 	fl4->fl4_gre_key = gre_key;
158 	return ip_route_output_key(net, fl4);
159 }
160 
161 extern int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
162 			  u8 tos, struct net_device *devin);
163 
164 extern void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu,
165 			     int oif, u32 mark, u8 protocol, int flow_flags);
166 extern void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
167 extern void ipv4_redirect(struct sk_buff *skb, struct net *net,
168 			  int oif, u32 mark, u8 protocol, int flow_flags);
169 extern void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
170 extern void ip_rt_send_redirect(struct sk_buff *skb);
171 
172 extern unsigned int		inet_addr_type(struct net *net, __be32 addr);
173 extern unsigned int		inet_dev_addr_type(struct net *net, const struct net_device *dev, __be32 addr);
174 extern void		ip_rt_multicast_event(struct in_device *);
175 extern int		ip_rt_ioctl(struct net *, unsigned int cmd, void __user *arg);
176 extern void		ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
177 extern int		ip_rt_dump(struct sk_buff *skb,  struct netlink_callback *cb);
178 
179 struct in_ifaddr;
180 extern void fib_add_ifaddr(struct in_ifaddr *);
181 extern void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
182 
183 static inline void ip_rt_put(struct rtable * rt)
184 {
185 	if (rt)
186 		dst_release(&rt->dst);
187 }
188 
189 #define IPTOS_RT_MASK	(IPTOS_TOS_MASK & ~3)
190 
191 extern const __u8 ip_tos2prio[16];
192 
193 static inline char rt_tos2priority(u8 tos)
194 {
195 	return ip_tos2prio[IPTOS_TOS(tos)>>1];
196 }
197 
198 /* ip_route_connect() and ip_route_newports() work in tandem whilst
199  * binding a socket for a new outgoing connection.
200  *
201  * In order to use IPSEC properly, we must, in the end, have a
202  * route that was looked up using all available keys including source
203  * and destination ports.
204  *
205  * However, if a source port needs to be allocated (the user specified
206  * a wildcard source port) we need to obtain addressing information
207  * in order to perform that allocation.
208  *
209  * So ip_route_connect() looks up a route using wildcarded source and
210  * destination ports in the key, simply so that we can get a pair of
211  * addresses to use for port allocation.
212  *
213  * Later, once the ports are allocated, ip_route_newports() will make
214  * another route lookup if needed to make sure we catch any IPSEC
215  * rules keyed on the port information.
216  *
217  * The callers allocate the flow key on their stack, and must pass in
218  * the same flowi4 object to both the ip_route_connect() and the
219  * ip_route_newports() calls.
220  */
221 
222 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src,
223 					 u32 tos, int oif, u8 protocol,
224 					 __be16 sport, __be16 dport,
225 					 struct sock *sk, bool can_sleep)
226 {
227 	__u8 flow_flags = 0;
228 
229 	if (inet_sk(sk)->transparent)
230 		flow_flags |= FLOWI_FLAG_ANYSRC;
231 	if (can_sleep)
232 		flow_flags |= FLOWI_FLAG_CAN_SLEEP;
233 
234 	flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE,
235 			   protocol, flow_flags, dst, src, dport, sport);
236 }
237 
238 static inline struct rtable *ip_route_connect(struct flowi4 *fl4,
239 					      __be32 dst, __be32 src, u32 tos,
240 					      int oif, u8 protocol,
241 					      __be16 sport, __be16 dport,
242 					      struct sock *sk, bool can_sleep)
243 {
244 	struct net *net = sock_net(sk);
245 	struct rtable *rt;
246 
247 	ip_route_connect_init(fl4, dst, src, tos, oif, protocol,
248 			      sport, dport, sk, can_sleep);
249 
250 	if (!dst || !src) {
251 		rt = __ip_route_output_key(net, fl4);
252 		if (IS_ERR(rt))
253 			return rt;
254 		ip_rt_put(rt);
255 		flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr);
256 	}
257 	security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
258 	return ip_route_output_flow(net, fl4, sk);
259 }
260 
261 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
262 					       __be16 orig_sport, __be16 orig_dport,
263 					       __be16 sport, __be16 dport,
264 					       struct sock *sk)
265 {
266 	if (sport != orig_sport || dport != orig_dport) {
267 		fl4->fl4_dport = dport;
268 		fl4->fl4_sport = sport;
269 		ip_rt_put(rt);
270 		flowi4_update_output(fl4, sk->sk_bound_dev_if,
271 				     RT_CONN_FLAGS(sk), fl4->daddr,
272 				     fl4->saddr);
273 		security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
274 		return ip_route_output_flow(sock_net(sk), fl4, sk);
275 	}
276 	return rt;
277 }
278 
279 static inline int inet_iif(const struct sk_buff *skb)
280 {
281 	return skb_rtable(skb)->rt_iif;
282 }
283 
284 extern int sysctl_ip_default_ttl;
285 
286 static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
287 {
288 	int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
289 
290 	if (hoplimit == 0)
291 		hoplimit = sysctl_ip_default_ttl;
292 	return hoplimit;
293 }
294 
295 #endif	/* _ROUTE_H */
296