net: Refactor path selection in __ip_route_output_key_hash
[deliverable/linux.git] / include / net / route.h
CommitLineData
1da177e4
LT
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 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
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
1da177e4
LT
27#include <net/dst.h>
28#include <net/inetpeer.h>
29#include <net/flow.h>
79876874 30#include <net/inet_sock.h>
79a13159 31#include <net/ip_fib.h>
9478d12d 32#include <net/l3mdev.h>
1da177e4
LT
33#include <linux/in_route.h>
34#include <linux/rtnetlink.h>
c6cffba4 35#include <linux/rcupdate.h>
1da177e4
LT
36#include <linux/route.h>
37#include <linux/ip.h>
38#include <linux/cache.h>
beb8d13b 39#include <linux/security.h>
1da177e4 40
f87c10a8
HFS
41/* IPv4 datagram length is stored into 16bit field (tot_len) */
42#define IP_MAX_MTU 0xFFFFU
43
1da177e4
LT
44#define RTO_ONLINK 0x01
45
1da177e4 46#define RT_CONN_FLAGS(sk) (RT_TOS(inet_sk(sk)->tos) | sock_flag(sk, SOCK_LOCALROUTE))
aa661581 47#define RT_CONN_FLAGS_TOS(sk,tos) (RT_TOS(tos) | sock_flag(sk, SOCK_LOCALROUTE))
1da177e4
LT
48
49struct fib_nh;
62fa8a84 50struct fib_info;
5055c371 51struct uncached_list;
fd2c3ef7 52struct rtable {
d8d1f30b 53 struct dst_entry dst;
1da177e4 54
29e75252 55 int rt_genid;
95c96174 56 unsigned int rt_flags;
1da177e4 57 __u16 rt_type;
155e8336
JA
58 __u8 rt_is_input;
59 __u8 rt_uses_gateway;
1da177e4 60
1da177e4
LT
61 int rt_iif;
62
63 /* Info on neighbour */
f2c3fe24 64 __be32 rt_gateway;
1da177e4 65
1da177e4 66 /* Miscellaneous cached information */
5943634f 67 u32 rt_pmtu;
caacf05e 68
b7503e0c
DA
69 u32 rt_table_id;
70
caacf05e 71 struct list_head rt_uncached;
5055c371 72 struct uncached_list *rt_uncached_list;
1da177e4
LT
73};
74
b8400f37 75static inline bool rt_is_input_route(const struct rtable *rt)
c7537967 76{
9917e1e8 77 return rt->rt_is_input != 0;
c7537967
DM
78}
79
b8400f37 80static inline bool rt_is_output_route(const struct rtable *rt)
c7537967 81{
9917e1e8 82 return rt->rt_is_input == 0;
c7537967
DM
83}
84
f8126f1d
DM
85static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr)
86{
87 if (rt->rt_gateway)
88 return rt->rt_gateway;
89 return daddr;
90}
91
fd2c3ef7 92struct ip_rt_acct {
1da177e4
LT
93 __u32 o_bytes;
94 __u32 o_packets;
95 __u32 i_bytes;
96 __u32 i_packets;
97};
98
fd2c3ef7 99struct rt_cache_stat {
1da177e4
LT
100 unsigned int in_slow_tot;
101 unsigned int in_slow_mc;
102 unsigned int in_no_route;
103 unsigned int in_brd;
104 unsigned int in_martian_dst;
105 unsigned int in_martian_src;
1da177e4
LT
106 unsigned int out_slow_tot;
107 unsigned int out_slow_mc;
1da177e4
LT
108};
109
7d720c3e 110extern struct ip_rt_acct __percpu *ip_rt_acct;
1da177e4
LT
111
112struct in_device;
2702c4bb
JP
113
114int ip_rt_init(void);
115void rt_cache_flush(struct net *net);
116void rt_flush_dev(struct net_device *dev);
79a13159
PN
117struct rtable *__ip_route_output_key_hash(struct net *, struct flowi4 *flp,
118 int mp_hash);
119
120static inline struct rtable *__ip_route_output_key(struct net *net,
121 struct flowi4 *flp)
122{
123 return __ip_route_output_key_hash(net, flp, -1);
124}
125
2702c4bb 126struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
6f9c9615 127 const struct sock *sk);
2702c4bb
JP
128struct dst_entry *ipv4_blackhole_route(struct net *net,
129 struct dst_entry *dst_orig);
407eadd9 130
9d6ec938 131static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
5bfa787f
DM
132{
133 return ip_route_output_flow(net, flp, NULL);
134}
135
78fbfd8a
DM
136static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
137 __be32 saddr, u8 tos, int oif)
138{
9d6ec938
DM
139 struct flowi4 fl4 = {
140 .flowi4_oif = oif,
c5d21c4b 141 .flowi4_tos = tos,
9d6ec938
DM
142 .daddr = daddr,
143 .saddr = saddr,
78fbfd8a 144 };
9d6ec938 145 return ip_route_output_key(net, &fl4);
78fbfd8a
DM
146}
147
31e4543d
DM
148static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
149 struct sock *sk,
78fbfd8a
DM
150 __be32 daddr, __be32 saddr,
151 __be16 dport, __be16 sport,
152 __u8 proto, __u8 tos, int oif)
153{
31e4543d 154 flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos,
94b92b88
DM
155 RT_SCOPE_UNIVERSE, proto,
156 sk ? inet_sk_flowi_flags(sk) : 0,
157 daddr, saddr, dport, sport);
78fbfd8a 158 if (sk)
31e4543d
DM
159 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
160 return ip_route_output_flow(net, fl4, sk);
78fbfd8a
DM
161}
162
cbb1e85f 163static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4,
78fbfd8a
DM
164 __be32 daddr, __be32 saddr,
165 __be32 gre_key, __u8 tos, int oif)
166{
cbb1e85f
DM
167 memset(fl4, 0, sizeof(*fl4));
168 fl4->flowi4_oif = oif;
169 fl4->daddr = daddr;
170 fl4->saddr = saddr;
171 fl4->flowi4_tos = tos;
172 fl4->flowi4_proto = IPPROTO_GRE;
173 fl4->fl4_gre_key = gre_key;
174 return ip_route_output_key(net, fl4);
78fbfd8a
DM
175}
176
2702c4bb
JP
177int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src,
178 u8 tos, struct net_device *devin);
c6cffba4
DM
179
180static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
181 u8 tos, struct net_device *devin)
182{
183 int err;
184
185 rcu_read_lock();
186 err = ip_route_input_noref(skb, dst, src, tos, devin);
187 if (!err)
188 skb_dst_force(skb);
189 rcu_read_unlock();
190
191 return err;
192}
407eadd9 193
2702c4bb
JP
194void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif,
195 u32 mark, u8 protocol, int flow_flags);
196void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
197void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
198 u8 protocol, int flow_flags);
199void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
200void ip_rt_send_redirect(struct sk_buff *skb);
201
202unsigned int inet_addr_type(struct net *net, __be32 addr);
9b8ff518 203unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id);
2702c4bb
JP
204unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
205 __be32 addr);
30bbaa19
DA
206unsigned int inet_addr_type_dev_table(struct net *net,
207 const struct net_device *dev,
208 __be32 addr);
2702c4bb
JP
209void ip_rt_multicast_event(struct in_device *);
210int ip_rt_ioctl(struct net *, unsigned int cmd, void __user *arg);
211void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
1da177e4 212
0ff60a45 213struct in_ifaddr;
2702c4bb
JP
214void fib_add_ifaddr(struct in_ifaddr *);
215void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
0ff60a45 216
6da025fa 217static inline void ip_rt_put(struct rtable *rt)
1da177e4 218{
6da025fa
ED
219 /* dst_release() accepts a NULL parameter.
220 * We rely on dst being first structure in struct rtable
221 */
222 BUILD_BUG_ON(offsetof(struct rtable, dst) != 0);
223 dst_release(&rt->dst);
1da177e4
LT
224}
225
226#define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3)
227
4839c52b 228extern const __u8 ip_tos2prio[16];
1da177e4
LT
229
230static inline char rt_tos2priority(u8 tos)
231{
232 return ip_tos2prio[IPTOS_TOS(tos)>>1];
233}
234
2d7192d6
DM
235/* ip_route_connect() and ip_route_newports() work in tandem whilst
236 * binding a socket for a new outgoing connection.
237 *
238 * In order to use IPSEC properly, we must, in the end, have a
239 * route that was looked up using all available keys including source
240 * and destination ports.
241 *
242 * However, if a source port needs to be allocated (the user specified
243 * a wildcard source port) we need to obtain addressing information
244 * in order to perform that allocation.
245 *
246 * So ip_route_connect() looks up a route using wildcarded source and
247 * destination ports in the key, simply so that we can get a pair of
248 * addresses to use for port allocation.
249 *
250 * Later, once the ports are allocated, ip_route_newports() will make
251 * another route lookup if needed to make sure we catch any IPSEC
252 * rules keyed on the port information.
253 *
254 * The callers allocate the flow key on their stack, and must pass in
255 * the same flowi4 object to both the ip_route_connect() and the
256 * ip_route_newports() calls.
257 */
258
259static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src,
260 u32 tos, int oif, u8 protocol,
261 __be16 sport, __be16 dport,
0e0d44ab 262 struct sock *sk)
1da177e4 263{
2d7192d6 264 __u8 flow_flags = 0;
79876874
KK
265
266 if (inet_sk(sk)->transparent)
94b92b88 267 flow_flags |= FLOWI_FLAG_ANYSRC;
94b92b88 268
007979ea 269 if (netif_index_is_l3_master(sock_net(sk), oif))
6e2895a8 270 flow_flags |= FLOWI_FLAG_L3MDEV_SRC | FLOWI_FLAG_SKIP_NH_OIF;
613d09b3 271
2d7192d6 272 flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE,
94b92b88 273 protocol, flow_flags, dst, src, dport, sport);
2d7192d6
DM
274}
275
276static inline struct rtable *ip_route_connect(struct flowi4 *fl4,
277 __be32 dst, __be32 src, u32 tos,
278 int oif, u8 protocol,
279 __be16 sport, __be16 dport,
0e0d44ab 280 struct sock *sk)
2d7192d6
DM
281{
282 struct net *net = sock_net(sk);
283 struct rtable *rt;
284
285 ip_route_connect_init(fl4, dst, src, tos, oif, protocol,
0e0d44ab 286 sport, dport, sk);
79876874 287
1da177e4 288 if (!dst || !src) {
2d7192d6 289 rt = __ip_route_output_key(net, fl4);
b23dd4fe
DM
290 if (IS_ERR(rt))
291 return rt;
b23dd4fe 292 ip_rt_put(rt);
e6b45241 293 flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr);
1da177e4 294 }
2d7192d6
DM
295 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
296 return ip_route_output_flow(net, fl4, sk);
1da177e4
LT
297}
298
2d7192d6
DM
299static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
300 __be16 orig_sport, __be16 orig_dport,
301 __be16 sport, __be16 dport,
302 struct sock *sk)
1da177e4 303{
dca8b089 304 if (sport != orig_sport || dport != orig_dport) {
2d7192d6
DM
305 fl4->fl4_dport = dport;
306 fl4->fl4_sport = sport;
b23dd4fe 307 ip_rt_put(rt);
e6b45241
JA
308 flowi4_update_output(fl4, sk->sk_bound_dev_if,
309 RT_CONN_FLAGS(sk), fl4->daddr,
310 fl4->saddr);
2d7192d6
DM
311 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
312 return ip_route_output_flow(sock_net(sk), fl4, sk);
1da177e4 313 }
b23dd4fe 314 return rt;
1da177e4
LT
315}
316
1668e010
KK
317static inline int inet_iif(const struct sk_buff *skb)
318{
13378cad
DM
319 int iif = skb_rtable(skb)->rt_iif;
320
321 if (iif)
322 return iif;
323 return skb->skb_iif;
1668e010
KK
324}
325
323e126f
DM
326extern int sysctl_ip_default_ttl;
327
328static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
329{
330 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
331
332 if (hoplimit == 0)
333 hoplimit = sysctl_ip_default_ttl;
334 return hoplimit;
335}
336
1da177e4 337#endif /* _ROUTE_H */
This page took 0.762083 seconds and 5 git commands to generate.