route.h revision 1f42539d257af671d56d4bdbcf13aef31abff6ef
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
42struct fib_nh;
43struct fib_info;
44struct rtable {
45	struct dst_entry	dst;
46
47	/* Lookup key. */
48	__be32			rt_key_dst;
49	__be32			rt_key_src;
50
51	int			rt_genid;
52	unsigned int		rt_flags;
53	__u16			rt_type;
54	__u8			rt_key_tos;
55
56	__be32			rt_dst;	/* Path destination	*/
57	__be32			rt_src;	/* Path source		*/
58	int			rt_route_iif;
59	int			rt_iif;
60	int			rt_oif;
61	__u32			rt_mark;
62
63	/* Info on neighbour */
64	__be32			rt_gateway;
65
66	/* Miscellaneous cached information */
67	u32			rt_pmtu;
68	struct fib_info		*fi; /* for client ref to shared metrics */
69};
70
71static inline bool rt_is_input_route(const struct rtable *rt)
72{
73	return rt->rt_route_iif != 0;
74}
75
76static inline bool rt_is_output_route(const struct rtable *rt)
77{
78	return rt->rt_route_iif == 0;
79}
80
81struct ip_rt_acct {
82	__u32 	o_bytes;
83	__u32 	o_packets;
84	__u32 	i_bytes;
85	__u32 	i_packets;
86};
87
88struct rt_cache_stat {
89        unsigned int in_hit;
90        unsigned int in_slow_tot;
91        unsigned int in_slow_mc;
92        unsigned int in_no_route;
93        unsigned int in_brd;
94        unsigned int in_martian_dst;
95        unsigned int in_martian_src;
96        unsigned int out_hit;
97        unsigned int out_slow_tot;
98        unsigned int out_slow_mc;
99        unsigned int gc_total;
100        unsigned int gc_ignored;
101        unsigned int gc_goal_miss;
102        unsigned int gc_dst_overflow;
103        unsigned int in_hlist_search;
104        unsigned int out_hlist_search;
105};
106
107extern struct ip_rt_acct __percpu *ip_rt_acct;
108
109struct in_device;
110extern int		ip_rt_init(void);
111extern void		rt_cache_flush(struct net *net, int how);
112extern void		rt_cache_flush_batch(struct net *net);
113extern struct rtable *__ip_route_output_key(struct net *, struct flowi4 *flp);
114extern struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
115					   struct sock *sk);
116extern struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig);
117
118static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
119{
120	return ip_route_output_flow(net, flp, NULL);
121}
122
123static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
124					     __be32 saddr, u8 tos, int oif)
125{
126	struct flowi4 fl4 = {
127		.flowi4_oif = oif,
128		.flowi4_tos = tos,
129		.daddr = daddr,
130		.saddr = saddr,
131	};
132	return ip_route_output_key(net, &fl4);
133}
134
135static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
136						   struct sock *sk,
137						   __be32 daddr, __be32 saddr,
138						   __be16 dport, __be16 sport,
139						   __u8 proto, __u8 tos, int oif)
140{
141	flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos,
142			   RT_SCOPE_UNIVERSE, proto,
143			   sk ? inet_sk_flowi_flags(sk) : 0,
144			   daddr, saddr, dport, sport);
145	if (sk)
146		security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
147	return ip_route_output_flow(net, fl4, sk);
148}
149
150static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4,
151						 __be32 daddr, __be32 saddr,
152						 __be32 gre_key, __u8 tos, int oif)
153{
154	memset(fl4, 0, sizeof(*fl4));
155	fl4->flowi4_oif = oif;
156	fl4->daddr = daddr;
157	fl4->saddr = saddr;
158	fl4->flowi4_tos = tos;
159	fl4->flowi4_proto = IPPROTO_GRE;
160	fl4->fl4_gre_key = gre_key;
161	return ip_route_output_key(net, fl4);
162}
163
164extern int ip_route_input_common(struct sk_buff *skb, __be32 dst, __be32 src,
165				 u8 tos, struct net_device *devin, bool noref);
166
167static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
168				 u8 tos, struct net_device *devin)
169{
170	return ip_route_input_common(skb, dst, src, tos, devin, false);
171}
172
173static inline int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src,
174				       u8 tos, struct net_device *devin)
175{
176	return ip_route_input_common(skb, dst, src, tos, devin, true);
177}
178
179extern void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu,
180			     int oif, u32 mark, u8 protocol, int flow_flags);
181extern void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
182extern void ipv4_redirect(struct sk_buff *skb, struct net *net,
183			  int oif, u32 mark, u8 protocol, int flow_flags);
184extern void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
185extern void ip_rt_send_redirect(struct sk_buff *skb);
186
187extern unsigned int		inet_addr_type(struct net *net, __be32 addr);
188extern unsigned int		inet_dev_addr_type(struct net *net, const struct net_device *dev, __be32 addr);
189extern void		ip_rt_multicast_event(struct in_device *);
190extern int		ip_rt_ioctl(struct net *, unsigned int cmd, void __user *arg);
191extern void		ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
192extern int		ip_rt_dump(struct sk_buff *skb,  struct netlink_callback *cb);
193
194struct in_ifaddr;
195extern void fib_add_ifaddr(struct in_ifaddr *);
196extern void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
197
198static inline void ip_rt_put(struct rtable * rt)
199{
200	if (rt)
201		dst_release(&rt->dst);
202}
203
204#define IPTOS_RT_MASK	(IPTOS_TOS_MASK & ~3)
205
206extern const __u8 ip_tos2prio[16];
207
208static inline char rt_tos2priority(u8 tos)
209{
210	return ip_tos2prio[IPTOS_TOS(tos)>>1];
211}
212
213/* ip_route_connect() and ip_route_newports() work in tandem whilst
214 * binding a socket for a new outgoing connection.
215 *
216 * In order to use IPSEC properly, we must, in the end, have a
217 * route that was looked up using all available keys including source
218 * and destination ports.
219 *
220 * However, if a source port needs to be allocated (the user specified
221 * a wildcard source port) we need to obtain addressing information
222 * in order to perform that allocation.
223 *
224 * So ip_route_connect() looks up a route using wildcarded source and
225 * destination ports in the key, simply so that we can get a pair of
226 * addresses to use for port allocation.
227 *
228 * Later, once the ports are allocated, ip_route_newports() will make
229 * another route lookup if needed to make sure we catch any IPSEC
230 * rules keyed on the port information.
231 *
232 * The callers allocate the flow key on their stack, and must pass in
233 * the same flowi4 object to both the ip_route_connect() and the
234 * ip_route_newports() calls.
235 */
236
237static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src,
238					 u32 tos, int oif, u8 protocol,
239					 __be16 sport, __be16 dport,
240					 struct sock *sk, bool can_sleep)
241{
242	__u8 flow_flags = 0;
243
244	if (inet_sk(sk)->transparent)
245		flow_flags |= FLOWI_FLAG_ANYSRC;
246	if (can_sleep)
247		flow_flags |= FLOWI_FLAG_CAN_SLEEP;
248
249	flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE,
250			   protocol, flow_flags, dst, src, dport, sport);
251}
252
253static inline struct rtable *ip_route_connect(struct flowi4 *fl4,
254					      __be32 dst, __be32 src, u32 tos,
255					      int oif, u8 protocol,
256					      __be16 sport, __be16 dport,
257					      struct sock *sk, bool can_sleep)
258{
259	struct net *net = sock_net(sk);
260	struct rtable *rt;
261
262	ip_route_connect_init(fl4, dst, src, tos, oif, protocol,
263			      sport, dport, sk, can_sleep);
264
265	if (!dst || !src) {
266		rt = __ip_route_output_key(net, fl4);
267		if (IS_ERR(rt))
268			return rt;
269		ip_rt_put(rt);
270		flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr);
271	}
272	security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
273	return ip_route_output_flow(net, fl4, sk);
274}
275
276static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
277					       __be16 orig_sport, __be16 orig_dport,
278					       __be16 sport, __be16 dport,
279					       struct sock *sk)
280{
281	if (sport != orig_sport || dport != orig_dport) {
282		fl4->fl4_dport = dport;
283		fl4->fl4_sport = sport;
284		ip_rt_put(rt);
285		flowi4_update_output(fl4, sk->sk_bound_dev_if,
286				     RT_CONN_FLAGS(sk), fl4->daddr,
287				     fl4->saddr);
288		security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
289		return ip_route_output_flow(sock_net(sk), fl4, sk);
290	}
291	return rt;
292}
293
294static inline int inet_iif(const struct sk_buff *skb)
295{
296	return skb_rtable(skb)->rt_iif;
297}
298
299extern int sysctl_ip_default_ttl;
300
301static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
302{
303	int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
304
305	if (hoplimit == 0)
306		hoplimit = sysctl_ip_default_ttl;
307	return hoplimit;
308}
309
310#endif	/* _ROUTE_H */
311