1 /* SPDX-License-Identifier: GPL-2.0-or-later */
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.
7 * Definitions for the IP router.
9 * Version: @(#)route.h 1.0.4 05/27/93
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
24 #include <net/inetpeer.h>
26 #include <net/inet_sock.h>
27 #include <net/ip_fib.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>
35 #include <linux/cache.h>
36 #include <linux/security.h>
38 #define RTO_ONLINK 0x01
40 static inline __u8 ip_sock_rt_scope(const struct sock *sk)
42 if (sock_flag(sk, SOCK_LOCALROUTE))
45 return RT_SCOPE_UNIVERSE;
48 static inline __u8 ip_sock_rt_tos(const struct sock *sk)
50 return RT_TOS(READ_ONCE(inet_sk(sk)->tos));
53 struct ip_tunnel_info;
61 unsigned int rt_flags;
69 /* Info on neighbour */
72 struct in6_addr rt_gw6;
75 /* Miscellaneous cached information */
80 static inline bool rt_is_input_route(const struct rtable *rt)
82 return rt->rt_is_input != 0;
85 static inline bool rt_is_output_route(const struct rtable *rt)
87 return rt->rt_is_input == 0;
90 static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr)
92 if (rt->rt_gw_family == AF_INET)
104 struct rt_cache_stat {
105 unsigned int in_slow_tot;
106 unsigned int in_slow_mc;
107 unsigned int in_no_route;
109 unsigned int in_martian_dst;
110 unsigned int in_martian_src;
111 unsigned int out_slow_tot;
112 unsigned int out_slow_mc;
115 extern struct ip_rt_acct __percpu *ip_rt_acct;
119 int ip_rt_init(void);
120 void rt_cache_flush(struct net *net);
121 void rt_flush_dev(struct net_device *dev);
122 struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp,
123 const struct sk_buff *skb);
124 struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp,
125 struct fib_result *res,
126 const struct sk_buff *skb);
128 static inline struct rtable *__ip_route_output_key(struct net *net,
131 return ip_route_output_key_hash(net, flp, NULL);
134 struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
135 const struct sock *sk);
136 struct dst_entry *ipv4_blackhole_route(struct net *net,
137 struct dst_entry *dst_orig);
139 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
141 return ip_route_output_flow(net, flp, NULL);
144 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
145 __be32 saddr, u8 tos, int oif)
147 struct flowi4 fl4 = {
153 return ip_route_output_key(net, &fl4);
156 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
157 const struct sock *sk,
158 __be32 daddr, __be32 saddr,
159 __be16 dport, __be16 sport,
160 __u8 proto, __u8 tos, int oif)
162 flowi4_init_output(fl4, oif, sk ? READ_ONCE(sk->sk_mark) : 0, tos,
163 sk ? ip_sock_rt_scope(sk) : RT_SCOPE_UNIVERSE,
164 proto, sk ? inet_sk_flowi_flags(sk) : 0,
165 daddr, saddr, dport, sport, sock_net_uid(net, sk));
167 security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
168 return ip_route_output_flow(net, fl4, sk);
171 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4,
172 __be32 daddr, __be32 saddr,
173 __be32 gre_key, __u8 tos, int oif)
175 memset(fl4, 0, sizeof(*fl4));
176 fl4->flowi4_oif = oif;
179 fl4->flowi4_tos = tos;
180 fl4->flowi4_proto = IPPROTO_GRE;
181 fl4->fl4_gre_key = gre_key;
182 return ip_route_output_key(net, fl4);
184 int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr,
185 u8 tos, struct net_device *dev,
186 struct in_device *in_dev, u32 *itag);
187 int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src,
188 u8 tos, struct net_device *devin);
189 int ip_route_use_hint(struct sk_buff *skb, __be32 dst, __be32 src,
190 u8 tos, struct net_device *devin,
191 const struct sk_buff *hint);
193 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
194 u8 tos, struct net_device *devin)
199 err = ip_route_input_noref(skb, dst, src, tos, devin);
210 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif,
212 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
213 void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol);
214 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
215 void ip_rt_send_redirect(struct sk_buff *skb);
217 unsigned int inet_addr_type(struct net *net, __be32 addr);
218 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id);
219 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
221 unsigned int inet_addr_type_dev_table(struct net *net,
222 const struct net_device *dev,
224 void ip_rt_multicast_event(struct in_device *);
225 int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt);
226 void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
227 struct rtable *rt_dst_alloc(struct net_device *dev,
228 unsigned int flags, u16 type, bool noxfrm);
229 struct rtable *rt_dst_clone(struct net_device *dev, struct rtable *rt);
232 void fib_add_ifaddr(struct in_ifaddr *);
233 void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
234 void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric);
236 void rt_add_uncached_list(struct rtable *rt);
237 void rt_del_uncached_list(struct rtable *rt);
239 int fib_dump_info_fnhe(struct sk_buff *skb, struct netlink_callback *cb,
240 u32 table_id, struct fib_info *fi,
241 int *fa_index, int fa_start, unsigned int flags);
243 static inline void ip_rt_put(struct rtable *rt)
245 /* dst_release() accepts a NULL parameter.
246 * We rely on dst being first structure in struct rtable
248 BUILD_BUG_ON(offsetof(struct rtable, dst) != 0);
249 dst_release(&rt->dst);
252 #define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3)
254 extern const __u8 ip_tos2prio[16];
256 static inline char rt_tos2priority(u8 tos)
258 return ip_tos2prio[IPTOS_TOS(tos)>>1];
261 /* ip_route_connect() and ip_route_newports() work in tandem whilst
262 * binding a socket for a new outgoing connection.
264 * In order to use IPSEC properly, we must, in the end, have a
265 * route that was looked up using all available keys including source
266 * and destination ports.
268 * However, if a source port needs to be allocated (the user specified
269 * a wildcard source port) we need to obtain addressing information
270 * in order to perform that allocation.
272 * So ip_route_connect() looks up a route using wildcarded source and
273 * destination ports in the key, simply so that we can get a pair of
274 * addresses to use for port allocation.
276 * Later, once the ports are allocated, ip_route_newports() will make
277 * another route lookup if needed to make sure we catch any IPSEC
278 * rules keyed on the port information.
280 * The callers allocate the flow key on their stack, and must pass in
281 * the same flowi4 object to both the ip_route_connect() and the
282 * ip_route_newports() calls.
285 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst,
286 __be32 src, int oif, u8 protocol,
287 __be16 sport, __be16 dport,
288 const struct sock *sk)
292 if (inet_test_bit(TRANSPARENT, sk))
293 flow_flags |= FLOWI_FLAG_ANYSRC;
295 flowi4_init_output(fl4, oif, READ_ONCE(sk->sk_mark), ip_sock_rt_tos(sk),
296 ip_sock_rt_scope(sk), protocol, flow_flags, dst,
297 src, dport, sport, sk->sk_uid);
300 static inline struct rtable *ip_route_connect(struct flowi4 *fl4, __be32 dst,
301 __be32 src, int oif, u8 protocol,
302 __be16 sport, __be16 dport,
303 const struct sock *sk)
305 struct net *net = sock_net(sk);
308 ip_route_connect_init(fl4, dst, src, oif, protocol, sport, dport, sk);
311 rt = __ip_route_output_key(net, fl4);
315 flowi4_update_output(fl4, oif, fl4->daddr, fl4->saddr);
317 security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
318 return ip_route_output_flow(net, fl4, sk);
321 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
322 __be16 orig_sport, __be16 orig_dport,
323 __be16 sport, __be16 dport,
324 const struct sock *sk)
326 if (sport != orig_sport || dport != orig_dport) {
327 fl4->fl4_dport = dport;
328 fl4->fl4_sport = sport;
330 flowi4_update_output(fl4, sk->sk_bound_dev_if, fl4->daddr,
332 security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
333 return ip_route_output_flow(sock_net(sk), fl4, sk);
338 static inline int inet_iif(const struct sk_buff *skb)
340 struct rtable *rt = skb_rtable(skb);
342 if (rt && rt->rt_iif)
348 static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
350 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
351 struct net *net = dev_net(dst->dev);
354 hoplimit = READ_ONCE(net->ipv4.sysctl_ip_default_ttl);
358 static inline struct neighbour *ip_neigh_gw4(struct net_device *dev,
361 struct neighbour *neigh;
363 neigh = __ipv4_neigh_lookup_noref(dev, (__force u32)daddr);
364 if (unlikely(!neigh))
365 neigh = __neigh_create(&arp_tbl, &daddr, dev, false);
370 static inline struct neighbour *ip_neigh_for_gw(struct rtable *rt,
374 struct net_device *dev = rt->dst.dev;
375 struct neighbour *neigh;
377 if (likely(rt->rt_gw_family == AF_INET)) {
378 neigh = ip_neigh_gw4(dev, rt->rt_gw4);
379 } else if (rt->rt_gw_family == AF_INET6) {
380 neigh = ip_neigh_gw6(dev, &rt->rt_gw6);
383 neigh = ip_neigh_gw4(dev, ip_hdr(skb)->daddr);
388 #endif /* _ROUTE_H */