Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
tjh.dev
kernel
os
linux
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the IP router.
8 *
9 * Version: @(#)route.h 1.0.4 05/27/93
10 *
11 * Authors: Ross Biro
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Fixes:
14 * Alan Cox : Reformatted. Added ip_rt_local()
15 * Alan Cox : Support for TCP parameters.
16 * Alexey Kuznetsov: Major changes for new routing code.
17 * Mike McLagan : Routing by source
18 * Robert Olsson : Added rt_cache statistics
19 */
20#ifndef _ROUTE_H
21#define _ROUTE_H
22
23#include <net/dst.h>
24#include <net/inetpeer.h>
25#include <net/flow.h>
26#include <net/inet_sock.h>
27#include <net/ip_fib.h>
28#include <net/arp.h>
29#include <net/ndisc.h>
30#include <linux/in_route.h>
31#include <linux/rtnetlink.h>
32#include <linux/rcupdate.h>
33#include <linux/route.h>
34#include <linux/ip.h>
35#include <linux/cache.h>
36#include <linux/security.h>
37
38/* IPv4 datagram length is stored into 16bit field (tot_len) */
39#define IP_MAX_MTU 0xFFFFU
40
41#define RTO_ONLINK 0x01
42
43#define RT_CONN_FLAGS(sk) (RT_TOS(inet_sk(sk)->tos) | sock_flag(sk, SOCK_LOCALROUTE))
44#define RT_CONN_FLAGS_TOS(sk,tos) (RT_TOS(tos) | sock_flag(sk, SOCK_LOCALROUTE))
45
46struct fib_nh;
47struct fib_info;
48struct uncached_list;
49struct rtable {
50 struct dst_entry dst;
51
52 int rt_genid;
53 unsigned int rt_flags;
54 __u16 rt_type;
55 __u8 rt_is_input;
56 u8 rt_gw_family;
57
58 int rt_iif;
59
60 /* Info on neighbour */
61 union {
62 __be32 rt_gw4;
63 struct in6_addr rt_gw6;
64 };
65
66 /* Miscellaneous cached information */
67 u32 rt_mtu_locked:1,
68 rt_pmtu:31;
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_gw_family == AF_INET)
87 return rt->rt_gw4;
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 *net, struct flowi4 *flp,
117 const struct sk_buff *skb);
118struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp,
119 struct fib_result *res,
120 const struct sk_buff *skb);
121
122static inline struct rtable *__ip_route_output_key(struct net *net,
123 struct flowi4 *flp)
124{
125 return ip_route_output_key_hash(net, flp, NULL);
126}
127
128struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
129 const struct sock *sk);
130struct dst_entry *ipv4_blackhole_route(struct net *net,
131 struct dst_entry *dst_orig);
132
133static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
134{
135 return ip_route_output_flow(net, flp, NULL);
136}
137
138static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
139 __be32 saddr, u8 tos, int oif)
140{
141 struct flowi4 fl4 = {
142 .flowi4_oif = oif,
143 .flowi4_tos = tos,
144 .daddr = daddr,
145 .saddr = saddr,
146 };
147 return ip_route_output_key(net, &fl4);
148}
149
150static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
151 struct sock *sk,
152 __be32 daddr, __be32 saddr,
153 __be16 dport, __be16 sport,
154 __u8 proto, __u8 tos, int oif)
155{
156 flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos,
157 RT_SCOPE_UNIVERSE, proto,
158 sk ? inet_sk_flowi_flags(sk) : 0,
159 daddr, saddr, dport, sport, sock_net_uid(net, sk));
160 if (sk)
161 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
162 return ip_route_output_flow(net, fl4, sk);
163}
164
165static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4,
166 __be32 daddr, __be32 saddr,
167 __be32 gre_key, __u8 tos, int oif)
168{
169 memset(fl4, 0, sizeof(*fl4));
170 fl4->flowi4_oif = oif;
171 fl4->daddr = daddr;
172 fl4->saddr = saddr;
173 fl4->flowi4_tos = tos;
174 fl4->flowi4_proto = IPPROTO_GRE;
175 fl4->fl4_gre_key = gre_key;
176 return ip_route_output_key(net, fl4);
177}
178int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr,
179 u8 tos, struct net_device *dev,
180 struct in_device *in_dev, u32 *itag);
181int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src,
182 u8 tos, struct net_device *devin);
183int ip_route_input_rcu(struct sk_buff *skb, __be32 dst, __be32 src,
184 u8 tos, struct net_device *devin,
185 struct fib_result *res);
186
187static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
188 u8 tos, struct net_device *devin)
189{
190 int err;
191
192 rcu_read_lock();
193 err = ip_route_input_noref(skb, dst, src, tos, devin);
194 if (!err) {
195 skb_dst_force(skb);
196 if (!skb_dst(skb))
197 err = -EINVAL;
198 }
199 rcu_read_unlock();
200
201 return err;
202}
203
204void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif,
205 u8 protocol);
206void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
207void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol);
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);
224struct rtable *rt_dst_clone(struct net_device *dev, struct rtable *rt);
225
226struct in_ifaddr;
227void fib_add_ifaddr(struct in_ifaddr *);
228void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
229void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric);
230
231void rt_add_uncached_list(struct rtable *rt);
232void rt_del_uncached_list(struct rtable *rt);
233
234static inline void ip_rt_put(struct rtable *rt)
235{
236 /* dst_release() accepts a NULL parameter.
237 * We rely on dst being first structure in struct rtable
238 */
239 BUILD_BUG_ON(offsetof(struct rtable, dst) != 0);
240 dst_release(&rt->dst);
241}
242
243#define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3)
244
245extern const __u8 ip_tos2prio[16];
246
247static inline char rt_tos2priority(u8 tos)
248{
249 return ip_tos2prio[IPTOS_TOS(tos)>>1];
250}
251
252/* ip_route_connect() and ip_route_newports() work in tandem whilst
253 * binding a socket for a new outgoing connection.
254 *
255 * In order to use IPSEC properly, we must, in the end, have a
256 * route that was looked up using all available keys including source
257 * and destination ports.
258 *
259 * However, if a source port needs to be allocated (the user specified
260 * a wildcard source port) we need to obtain addressing information
261 * in order to perform that allocation.
262 *
263 * So ip_route_connect() looks up a route using wildcarded source and
264 * destination ports in the key, simply so that we can get a pair of
265 * addresses to use for port allocation.
266 *
267 * Later, once the ports are allocated, ip_route_newports() will make
268 * another route lookup if needed to make sure we catch any IPSEC
269 * rules keyed on the port information.
270 *
271 * The callers allocate the flow key on their stack, and must pass in
272 * the same flowi4 object to both the ip_route_connect() and the
273 * ip_route_newports() calls.
274 */
275
276static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src,
277 u32 tos, int oif, u8 protocol,
278 __be16 sport, __be16 dport,
279 struct sock *sk)
280{
281 __u8 flow_flags = 0;
282
283 if (inet_sk(sk)->transparent)
284 flow_flags |= FLOWI_FLAG_ANYSRC;
285
286 flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE,
287 protocol, flow_flags, dst, src, dport, sport,
288 sk->sk_uid);
289}
290
291static inline struct rtable *ip_route_connect(struct flowi4 *fl4,
292 __be32 dst, __be32 src, u32 tos,
293 int oif, u8 protocol,
294 __be16 sport, __be16 dport,
295 struct sock *sk)
296{
297 struct net *net = sock_net(sk);
298 struct rtable *rt;
299
300 ip_route_connect_init(fl4, dst, src, tos, oif, protocol,
301 sport, dport, sk);
302
303 if (!dst || !src) {
304 rt = __ip_route_output_key(net, fl4);
305 if (IS_ERR(rt))
306 return rt;
307 ip_rt_put(rt);
308 flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr);
309 }
310 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
311 return ip_route_output_flow(net, fl4, sk);
312}
313
314static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
315 __be16 orig_sport, __be16 orig_dport,
316 __be16 sport, __be16 dport,
317 struct sock *sk)
318{
319 if (sport != orig_sport || dport != orig_dport) {
320 fl4->fl4_dport = dport;
321 fl4->fl4_sport = sport;
322 ip_rt_put(rt);
323 flowi4_update_output(fl4, sk->sk_bound_dev_if,
324 RT_CONN_FLAGS(sk), fl4->daddr,
325 fl4->saddr);
326 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
327 return ip_route_output_flow(sock_net(sk), fl4, sk);
328 }
329 return rt;
330}
331
332static inline int inet_iif(const struct sk_buff *skb)
333{
334 struct rtable *rt = skb_rtable(skb);
335
336 if (rt && rt->rt_iif)
337 return rt->rt_iif;
338
339 return skb->skb_iif;
340}
341
342static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
343{
344 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
345 struct net *net = dev_net(dst->dev);
346
347 if (hoplimit == 0)
348 hoplimit = net->ipv4.sysctl_ip_default_ttl;
349 return hoplimit;
350}
351
352static inline struct neighbour *ip_neigh_gw4(struct net_device *dev,
353 __be32 daddr)
354{
355 struct neighbour *neigh;
356
357 neigh = __ipv4_neigh_lookup_noref(dev, daddr);
358 if (unlikely(!neigh))
359 neigh = __neigh_create(&arp_tbl, &daddr, dev, false);
360
361 return neigh;
362}
363
364static inline struct neighbour *ip_neigh_for_gw(struct rtable *rt,
365 struct sk_buff *skb,
366 bool *is_v6gw)
367{
368 struct net_device *dev = rt->dst.dev;
369 struct neighbour *neigh;
370
371 if (likely(rt->rt_gw_family == AF_INET)) {
372 neigh = ip_neigh_gw4(dev, rt->rt_gw4);
373 } else if (rt->rt_gw_family == AF_INET6) {
374 neigh = ip_neigh_gw6(dev, &rt->rt_gw6);
375 *is_v6gw = true;
376 } else {
377 neigh = ip_neigh_gw4(dev, ip_hdr(skb)->daddr);
378 }
379 return neigh;
380}
381
382#endif /* _ROUTE_H */