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
48struct fib_nh;
49struct fib_info;
50struct uncached_list;
51struct 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
73static inline bool rt_is_input_route(const struct rtable *rt)
74{
75 return rt->rt_is_input != 0;
76}
77
78static inline bool rt_is_output_route(const struct rtable *rt)
79{
80 return rt->rt_is_input == 0;
81}
82
83static 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
90struct ip_rt_acct {
91 __u32 o_bytes;
92 __u32 o_packets;
93 __u32 i_bytes;
94 __u32 i_packets;
95};
96
97struct 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
108extern struct ip_rt_acct __percpu *ip_rt_acct;
109
110struct in_device;
111
112int ip_rt_init(void);
113void rt_cache_flush(struct net *net);
114void rt_flush_dev(struct net_device *dev);
115struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp,
116 const struct sk_buff *skb);
117struct 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
121static 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
127struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
128 const struct sock *sk);
129struct dst_entry *ipv4_blackhole_route(struct net *net,
130 struct dst_entry *dst_orig);
131
132static 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
137static 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
149static 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
164static 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}
177int 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);
180int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src,
181 u8 tos, struct net_device *devin);
182int 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
186static 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
203void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif,
204 u32 mark, u8 protocol, int flow_flags);
205void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
206void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
207 u8 protocol, int flow_flags);
208void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
209void ip_rt_send_redirect(struct sk_buff *skb);
210
211unsigned int inet_addr_type(struct net *net, __be32 addr);
212unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id);
213unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
214 __be32 addr);
215unsigned int inet_addr_type_dev_table(struct net *net,
216 const struct net_device *dev,
217 __be32 addr);
218void ip_rt_multicast_event(struct in_device *);
219int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt);
220void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
221struct rtable *rt_dst_alloc(struct net_device *dev,
222 unsigned int flags, u16 type,
223 bool nopolicy, bool noxfrm, bool will_cache);
224
225struct in_ifaddr;
226void fib_add_ifaddr(struct in_ifaddr *);
227void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
228void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric);
229
230void rt_add_uncached_list(struct rtable *rt);
231void rt_del_uncached_list(struct rtable *rt);
232
233static inline void ip_rt_put(struct rtable *rt)
234{
235 /* dst_release() accepts a NULL parameter.
236 * We rely on dst being first structure in struct rtable
237 */
238 BUILD_BUG_ON(offsetof(struct rtable, dst) != 0);
239 dst_release(&rt->dst);
240}
241
242#define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3)
243
244extern const __u8 ip_tos2prio[16];
245
246static inline char rt_tos2priority(u8 tos)
247{
248 return ip_tos2prio[IPTOS_TOS(tos)>>1];
249}
250
251/* ip_route_connect() and ip_route_newports() work in tandem whilst
252 * binding a socket for a new outgoing connection.
253 *
254 * In order to use IPSEC properly, we must, in the end, have a
255 * route that was looked up using all available keys including source
256 * and destination ports.
257 *
258 * However, if a source port needs to be allocated (the user specified
259 * a wildcard source port) we need to obtain addressing information
260 * in order to perform that allocation.
261 *
262 * So ip_route_connect() looks up a route using wildcarded source and
263 * destination ports in the key, simply so that we can get a pair of
264 * addresses to use for port allocation.
265 *
266 * Later, once the ports are allocated, ip_route_newports() will make
267 * another route lookup if needed to make sure we catch any IPSEC
268 * rules keyed on the port information.
269 *
270 * The callers allocate the flow key on their stack, and must pass in
271 * the same flowi4 object to both the ip_route_connect() and the
272 * ip_route_newports() calls.
273 */
274
275static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src,
276 u32 tos, int oif, u8 protocol,
277 __be16 sport, __be16 dport,
278 struct sock *sk)
279{
280 __u8 flow_flags = 0;
281
282 if (inet_sk(sk)->transparent)
283 flow_flags |= FLOWI_FLAG_ANYSRC;
284
285 flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE,
286 protocol, flow_flags, dst, src, dport, sport,
287 sk->sk_uid);
288}
289
290static inline struct rtable *ip_route_connect(struct flowi4 *fl4,
291 __be32 dst, __be32 src, u32 tos,
292 int oif, u8 protocol,
293 __be16 sport, __be16 dport,
294 struct sock *sk)
295{
296 struct net *net = sock_net(sk);
297 struct rtable *rt;
298
299 ip_route_connect_init(fl4, dst, src, tos, oif, protocol,
300 sport, dport, sk);
301
302 if (!dst || !src) {
303 rt = __ip_route_output_key(net, fl4);
304 if (IS_ERR(rt))
305 return rt;
306 ip_rt_put(rt);
307 flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr);
308 }
309 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
310 return ip_route_output_flow(net, fl4, sk);
311}
312
313static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
314 __be16 orig_sport, __be16 orig_dport,
315 __be16 sport, __be16 dport,
316 struct sock *sk)
317{
318 if (sport != orig_sport || dport != orig_dport) {
319 fl4->fl4_dport = dport;
320 fl4->fl4_sport = sport;
321 ip_rt_put(rt);
322 flowi4_update_output(fl4, sk->sk_bound_dev_if,
323 RT_CONN_FLAGS(sk), fl4->daddr,
324 fl4->saddr);
325 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
326 return ip_route_output_flow(sock_net(sk), fl4, sk);
327 }
328 return rt;
329}
330
331static inline int inet_iif(const struct sk_buff *skb)
332{
333 struct rtable *rt = skb_rtable(skb);
334
335 if (rt && rt->rt_iif)
336 return rt->rt_iif;
337
338 return skb->skb_iif;
339}
340
341static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
342{
343 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
344 struct net *net = dev_net(dst->dev);
345
346 if (hoplimit == 0)
347 hoplimit = net->ipv4.sysctl_ip_default_ttl;
348 return hoplimit;
349}
350
351#endif /* _ROUTE_H */
352