tjh.dev
Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
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_pmtu;
67
68 u32 rt_table_id;
69
70 struct list_head rt_uncached;
71 struct uncached_list *rt_uncached_list;
72};
73
74static inline bool rt_is_input_route(const struct rtable *rt)
75{
76 return rt->rt_is_input != 0;
77}
78
79static inline bool rt_is_output_route(const struct rtable *rt)
80{
81 return rt->rt_is_input == 0;
82}
83
84static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr)
85{
86 if (rt->rt_gateway)
87 return rt->rt_gateway;
88 return daddr;
89}
90
91struct ip_rt_acct {
92 __u32 o_bytes;
93 __u32 o_packets;
94 __u32 i_bytes;
95 __u32 i_packets;
96};
97
98struct rt_cache_stat {
99 unsigned int in_slow_tot;
100 unsigned int in_slow_mc;
101 unsigned int in_no_route;
102 unsigned int in_brd;
103 unsigned int in_martian_dst;
104 unsigned int in_martian_src;
105 unsigned int out_slow_tot;
106 unsigned int out_slow_mc;
107};
108
109extern struct ip_rt_acct __percpu *ip_rt_acct;
110
111struct in_device;
112
113int ip_rt_init(void);
114void rt_cache_flush(struct net *net);
115void rt_flush_dev(struct net_device *dev);
116struct rtable *__ip_route_output_key_hash(struct net *, struct flowi4 *flp,
117 int mp_hash);
118
119static inline struct rtable *__ip_route_output_key(struct net *net,
120 struct flowi4 *flp)
121{
122 return __ip_route_output_key_hash(net, flp, -1);
123}
124
125struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
126 const struct sock *sk);
127struct dst_entry *ipv4_blackhole_route(struct net *net,
128 struct dst_entry *dst_orig);
129
130static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
131{
132 return ip_route_output_flow(net, flp, NULL);
133}
134
135static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
136 __be32 saddr, u8 tos, int oif)
137{
138 struct flowi4 fl4 = {
139 .flowi4_oif = oif,
140 .flowi4_tos = tos,
141 .daddr = daddr,
142 .saddr = saddr,
143 };
144 return ip_route_output_key(net, &fl4);
145}
146
147static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
148 struct sock *sk,
149 __be32 daddr, __be32 saddr,
150 __be16 dport, __be16 sport,
151 __u8 proto, __u8 tos, int oif)
152{
153 flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos,
154 RT_SCOPE_UNIVERSE, proto,
155 sk ? inet_sk_flowi_flags(sk) : 0,
156 daddr, saddr, dport, sport);
157 if (sk)
158 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
159 return ip_route_output_flow(net, fl4, sk);
160}
161
162static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4,
163 __be32 daddr, __be32 saddr,
164 __be32 gre_key, __u8 tos, int oif)
165{
166 memset(fl4, 0, sizeof(*fl4));
167 fl4->flowi4_oif = oif;
168 fl4->daddr = daddr;
169 fl4->saddr = saddr;
170 fl4->flowi4_tos = tos;
171 fl4->flowi4_proto = IPPROTO_GRE;
172 fl4->fl4_gre_key = gre_key;
173 return ip_route_output_key(net, fl4);
174}
175
176int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src,
177 u8 tos, struct net_device *devin);
178
179static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
180 u8 tos, struct net_device *devin)
181{
182 int err;
183
184 rcu_read_lock();
185 err = ip_route_input_noref(skb, dst, src, tos, devin);
186 if (!err)
187 skb_dst_force(skb);
188 rcu_read_unlock();
189
190 return err;
191}
192
193void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif,
194 u32 mark, u8 protocol, int flow_flags);
195void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
196void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
197 u8 protocol, int flow_flags);
198void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
199void ip_rt_send_redirect(struct sk_buff *skb);
200
201unsigned int inet_addr_type(struct net *net, __be32 addr);
202unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id);
203unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
204 __be32 addr);
205unsigned int inet_addr_type_dev_table(struct net *net,
206 const struct net_device *dev,
207 __be32 addr);
208void ip_rt_multicast_event(struct in_device *);
209int ip_rt_ioctl(struct net *, unsigned int cmd, void __user *arg);
210void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
211struct rtable *rt_dst_alloc(struct net_device *dev,
212 unsigned int flags, u16 type,
213 bool nopolicy, bool noxfrm, bool will_cache);
214
215struct in_ifaddr;
216void fib_add_ifaddr(struct in_ifaddr *);
217void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
218
219static inline void ip_rt_put(struct rtable *rt)
220{
221 /* dst_release() accepts a NULL parameter.
222 * We rely on dst being first structure in struct rtable
223 */
224 BUILD_BUG_ON(offsetof(struct rtable, dst) != 0);
225 dst_release(&rt->dst);
226}
227
228#define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3)
229
230extern const __u8 ip_tos2prio[16];
231
232static inline char rt_tos2priority(u8 tos)
233{
234 return ip_tos2prio[IPTOS_TOS(tos)>>1];
235}
236
237/* ip_route_connect() and ip_route_newports() work in tandem whilst
238 * binding a socket for a new outgoing connection.
239 *
240 * In order to use IPSEC properly, we must, in the end, have a
241 * route that was looked up using all available keys including source
242 * and destination ports.
243 *
244 * However, if a source port needs to be allocated (the user specified
245 * a wildcard source port) we need to obtain addressing information
246 * in order to perform that allocation.
247 *
248 * So ip_route_connect() looks up a route using wildcarded source and
249 * destination ports in the key, simply so that we can get a pair of
250 * addresses to use for port allocation.
251 *
252 * Later, once the ports are allocated, ip_route_newports() will make
253 * another route lookup if needed to make sure we catch any IPSEC
254 * rules keyed on the port information.
255 *
256 * The callers allocate the flow key on their stack, and must pass in
257 * the same flowi4 object to both the ip_route_connect() and the
258 * ip_route_newports() calls.
259 */
260
261static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src,
262 u32 tos, int oif, u8 protocol,
263 __be16 sport, __be16 dport,
264 struct sock *sk)
265{
266 __u8 flow_flags = 0;
267
268 if (inet_sk(sk)->transparent)
269 flow_flags |= FLOWI_FLAG_ANYSRC;
270
271 flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE,
272 protocol, flow_flags, dst, src, dport, sport);
273}
274
275static inline struct rtable *ip_route_connect(struct flowi4 *fl4,
276 __be32 dst, __be32 src, u32 tos,
277 int oif, u8 protocol,
278 __be16 sport, __be16 dport,
279 struct sock *sk)
280{
281 struct net *net = sock_net(sk);
282 struct rtable *rt;
283
284 ip_route_connect_init(fl4, dst, src, tos, oif, protocol,
285 sport, dport, sk);
286
287 if (!dst || !src) {
288 rt = __ip_route_output_key(net, fl4);
289 if (IS_ERR(rt))
290 return rt;
291 ip_rt_put(rt);
292 flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr);
293 }
294 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
295 return ip_route_output_flow(net, fl4, sk);
296}
297
298static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
299 __be16 orig_sport, __be16 orig_dport,
300 __be16 sport, __be16 dport,
301 struct sock *sk)
302{
303 if (sport != orig_sport || dport != orig_dport) {
304 fl4->fl4_dport = dport;
305 fl4->fl4_sport = sport;
306 ip_rt_put(rt);
307 flowi4_update_output(fl4, sk->sk_bound_dev_if,
308 RT_CONN_FLAGS(sk), fl4->daddr,
309 fl4->saddr);
310 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
311 return ip_route_output_flow(sock_net(sk), fl4, sk);
312 }
313 return rt;
314}
315
316static inline int inet_iif(const struct sk_buff *skb)
317{
318 struct rtable *rt = skb_rtable(skb);
319
320 if (rt && rt->rt_iif)
321 return rt->rt_iif;
322
323 return skb->skb_iif;
324}
325
326static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
327{
328 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
329 struct net *net = dev_net(dst->dev);
330
331 if (hoplimit == 0)
332 hoplimit = net->ipv4.sysctl_ip_default_ttl;
333 return hoplimit;
334}
335
336#endif /* _ROUTE_H */