route.h revision c6cffba4ffa26a8ffacd0bb9f3144e34f20da7de
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/rcupdate.h>
34#include <linux/route.h>
35#include <linux/ip.h>
36#include <linux/cache.h>
37#include <linux/security.h>
38
39#define RTO_ONLINK	0x01
40
41#define RT_CONN_FLAGS(sk)   (RT_TOS(inet_sk(sk)->tos) | sock_flag(sk, SOCK_LOCALROUTE))
42
43struct fib_nh;
44struct fib_info;
45struct rtable {
46	struct dst_entry	dst;
47
48	int			rt_genid;
49	unsigned int		rt_flags;
50	__u16			rt_type;
51	__u16			rt_is_input;
52
53	int			rt_iif;
54
55	/* Info on neighbour */
56	__be32			rt_gateway;
57
58	/* Miscellaneous cached information */
59	u32			rt_pmtu;
60};
61
62static inline bool rt_is_input_route(const struct rtable *rt)
63{
64	return rt->rt_is_input != 0;
65}
66
67static inline bool rt_is_output_route(const struct rtable *rt)
68{
69	return rt->rt_is_input == 0;
70}
71
72static 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
79struct ip_rt_acct {
80	__u32 	o_bytes;
81	__u32 	o_packets;
82	__u32 	i_bytes;
83	__u32 	i_packets;
84};
85
86struct 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
105extern struct ip_rt_acct __percpu *ip_rt_acct;
106
107struct in_device;
108extern int		ip_rt_init(void);
109extern void		rt_cache_flush(struct net *net, int how);
110extern struct rtable *__ip_route_output_key(struct net *, struct flowi4 *flp);
111extern struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
112					   struct sock *sk);
113extern struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig);
114
115static 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
120static 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
132static 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
147static 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
161extern int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src,
162				u8 tos, struct net_device *devin);
163
164static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
165				 u8 tos, struct net_device *devin)
166{
167	int err;
168
169	rcu_read_lock();
170	err = ip_route_input_noref(skb, dst, src, tos, devin);
171	if (!err)
172		skb_dst_force(skb);
173	rcu_read_unlock();
174
175	return err;
176}
177
178extern void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu,
179			     int oif, u32 mark, u8 protocol, int flow_flags);
180extern void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
181extern void ipv4_redirect(struct sk_buff *skb, struct net *net,
182			  int oif, u32 mark, u8 protocol, int flow_flags);
183extern void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
184extern void ip_rt_send_redirect(struct sk_buff *skb);
185
186extern unsigned int		inet_addr_type(struct net *net, __be32 addr);
187extern unsigned int		inet_dev_addr_type(struct net *net, const struct net_device *dev, __be32 addr);
188extern void		ip_rt_multicast_event(struct in_device *);
189extern int		ip_rt_ioctl(struct net *, unsigned int cmd, void __user *arg);
190extern void		ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
191extern int		ip_rt_dump(struct sk_buff *skb,  struct netlink_callback *cb);
192
193struct in_ifaddr;
194extern void fib_add_ifaddr(struct in_ifaddr *);
195extern void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
196
197static inline void ip_rt_put(struct rtable * rt)
198{
199	if (rt)
200		dst_release(&rt->dst);
201}
202
203#define IPTOS_RT_MASK	(IPTOS_TOS_MASK & ~3)
204
205extern const __u8 ip_tos2prio[16];
206
207static inline char rt_tos2priority(u8 tos)
208{
209	return ip_tos2prio[IPTOS_TOS(tos)>>1];
210}
211
212/* ip_route_connect() and ip_route_newports() work in tandem whilst
213 * binding a socket for a new outgoing connection.
214 *
215 * In order to use IPSEC properly, we must, in the end, have a
216 * route that was looked up using all available keys including source
217 * and destination ports.
218 *
219 * However, if a source port needs to be allocated (the user specified
220 * a wildcard source port) we need to obtain addressing information
221 * in order to perform that allocation.
222 *
223 * So ip_route_connect() looks up a route using wildcarded source and
224 * destination ports in the key, simply so that we can get a pair of
225 * addresses to use for port allocation.
226 *
227 * Later, once the ports are allocated, ip_route_newports() will make
228 * another route lookup if needed to make sure we catch any IPSEC
229 * rules keyed on the port information.
230 *
231 * The callers allocate the flow key on their stack, and must pass in
232 * the same flowi4 object to both the ip_route_connect() and the
233 * ip_route_newports() calls.
234 */
235
236static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src,
237					 u32 tos, int oif, u8 protocol,
238					 __be16 sport, __be16 dport,
239					 struct sock *sk, bool can_sleep)
240{
241	__u8 flow_flags = 0;
242
243	if (inet_sk(sk)->transparent)
244		flow_flags |= FLOWI_FLAG_ANYSRC;
245	if (can_sleep)
246		flow_flags |= FLOWI_FLAG_CAN_SLEEP;
247
248	flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE,
249			   protocol, flow_flags, dst, src, dport, sport);
250}
251
252static inline struct rtable *ip_route_connect(struct flowi4 *fl4,
253					      __be32 dst, __be32 src, u32 tos,
254					      int oif, u8 protocol,
255					      __be16 sport, __be16 dport,
256					      struct sock *sk, bool can_sleep)
257{
258	struct net *net = sock_net(sk);
259	struct rtable *rt;
260
261	ip_route_connect_init(fl4, dst, src, tos, oif, protocol,
262			      sport, dport, sk, can_sleep);
263
264	if (!dst || !src) {
265		rt = __ip_route_output_key(net, fl4);
266		if (IS_ERR(rt))
267			return rt;
268		ip_rt_put(rt);
269		flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr);
270	}
271	security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
272	return ip_route_output_flow(net, fl4, sk);
273}
274
275static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
276					       __be16 orig_sport, __be16 orig_dport,
277					       __be16 sport, __be16 dport,
278					       struct sock *sk)
279{
280	if (sport != orig_sport || dport != orig_dport) {
281		fl4->fl4_dport = dport;
282		fl4->fl4_sport = sport;
283		ip_rt_put(rt);
284		flowi4_update_output(fl4, sk->sk_bound_dev_if,
285				     RT_CONN_FLAGS(sk), fl4->daddr,
286				     fl4->saddr);
287		security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
288		return ip_route_output_flow(sock_net(sk), fl4, sk);
289	}
290	return rt;
291}
292
293static inline int inet_iif(const struct sk_buff *skb)
294{
295	int iif = skb_rtable(skb)->rt_iif;
296
297	if (iif)
298		return iif;
299	return skb->skb_iif;
300}
301
302extern int sysctl_ip_default_ttl;
303
304static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
305{
306	int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
307
308	if (hoplimit == 0)
309		hoplimit = sysctl_ip_default_ttl;
310	return hoplimit;
311}
312
313#endif	/* _ROUTE_H */
314