xref: /linux/include/net/route.h (revision cca0d69baf950e5a82c21d09917b4cc654c83fe9)
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 <net/inet_dscp.h>
31 #include <linux/in_route.h>
32 #include <linux/rtnetlink.h>
33 #include <linux/rcupdate.h>
34 #include <linux/route.h>
35 #include <linux/ip.h>
36 #include <linux/cache.h>
37 #include <linux/security.h>
38 
39 static inline __u8 ip_sock_rt_scope(const struct sock *sk)
40 {
41 	if (sock_flag(sk, SOCK_LOCALROUTE))
42 		return RT_SCOPE_LINK;
43 
44 	return RT_SCOPE_UNIVERSE;
45 }
46 
47 static inline __u8 ip_sock_rt_tos(const struct sock *sk)
48 {
49 	return READ_ONCE(inet_sk(sk)->tos) & INET_DSCP_MASK;
50 }
51 
52 struct ip_tunnel_info;
53 struct fib_nh;
54 struct fib_info;
55 struct uncached_list;
56 struct rtable {
57 	struct dst_entry	dst;
58 
59 	int			rt_genid;
60 	unsigned int		rt_flags;
61 	__u16			rt_type;
62 	__u8			rt_is_input;
63 	__u8			rt_uses_gateway;
64 
65 	int			rt_iif;
66 
67 	u8			rt_gw_family;
68 	/* Info on neighbour */
69 	union {
70 		__be32		rt_gw4;
71 		struct in6_addr	rt_gw6;
72 	};
73 
74 	/* Miscellaneous cached information */
75 	u32			rt_mtu_locked:1,
76 				rt_pmtu:31;
77 };
78 
79 #define dst_rtable(_ptr) container_of_const(_ptr, struct rtable, dst)
80 
81 /**
82  * skb_rtable - Returns the skb &rtable
83  * @skb: buffer
84  */
85 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
86 {
87 	return dst_rtable(skb_dst(skb));
88 }
89 
90 static inline bool rt_is_input_route(const struct rtable *rt)
91 {
92 	return rt->rt_is_input != 0;
93 }
94 
95 static inline bool rt_is_output_route(const struct rtable *rt)
96 {
97 	return rt->rt_is_input == 0;
98 }
99 
100 static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr)
101 {
102 	if (rt->rt_gw_family == AF_INET)
103 		return rt->rt_gw4;
104 	return daddr;
105 }
106 
107 struct ip_rt_acct {
108 	__u32 	o_bytes;
109 	__u32 	o_packets;
110 	__u32 	i_bytes;
111 	__u32 	i_packets;
112 };
113 
114 struct rt_cache_stat {
115         unsigned int in_slow_tot;
116         unsigned int in_slow_mc;
117         unsigned int in_no_route;
118         unsigned int in_brd;
119         unsigned int in_martian_dst;
120         unsigned int in_martian_src;
121         unsigned int out_slow_tot;
122         unsigned int out_slow_mc;
123 };
124 
125 extern struct ip_rt_acct __percpu *ip_rt_acct;
126 
127 struct in_device;
128 
129 int ip_rt_init(void);
130 void rt_cache_flush(struct net *net);
131 void rt_flush_dev(struct net_device *dev);
132 struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp,
133 					const struct sk_buff *skb);
134 struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp,
135 					    struct fib_result *res,
136 					    const struct sk_buff *skb);
137 
138 static inline struct rtable *__ip_route_output_key(struct net *net,
139 						   struct flowi4 *flp)
140 {
141 	return ip_route_output_key_hash(net, flp, NULL);
142 }
143 
144 struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
145 				    const struct sock *sk);
146 struct dst_entry *ipv4_blackhole_route(struct net *net,
147 				       struct dst_entry *dst_orig);
148 
149 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
150 {
151 	return ip_route_output_flow(net, flp, NULL);
152 }
153 
154 /* Simplistic IPv4 route lookup function.
155  * This is only suitable for some particular use cases: since the flowi4
156  * structure is only partially set, it may bypass some fib-rules.
157  */
158 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
159 					     __be32 saddr, u8 tos, int oif,
160 					     __u8 scope)
161 {
162 	struct flowi4 fl4 = {
163 		.flowi4_oif = oif,
164 		.flowi4_tos = tos,
165 		.flowi4_scope = scope,
166 		.daddr = daddr,
167 		.saddr = saddr,
168 	};
169 
170 	return ip_route_output_key(net, &fl4);
171 }
172 
173 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
174 						   const struct sock *sk,
175 						   __be32 daddr, __be32 saddr,
176 						   __be16 dport, __be16 sport,
177 						   __u8 proto, __u8 tos, int oif)
178 {
179 	flowi4_init_output(fl4, oif, sk ? READ_ONCE(sk->sk_mark) : 0, tos,
180 			   sk ? ip_sock_rt_scope(sk) : RT_SCOPE_UNIVERSE,
181 			   proto, sk ? inet_sk_flowi_flags(sk) : 0,
182 			   daddr, saddr, dport, sport, sock_net_uid(net, sk));
183 	if (sk)
184 		security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
185 	return ip_route_output_flow(net, fl4, sk);
186 }
187 
188 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4,
189 						 __be32 daddr, __be32 saddr,
190 						 __be32 gre_key, __u8 tos, int oif)
191 {
192 	memset(fl4, 0, sizeof(*fl4));
193 	fl4->flowi4_oif = oif;
194 	fl4->daddr = daddr;
195 	fl4->saddr = saddr;
196 	fl4->flowi4_tos = tos;
197 	fl4->flowi4_proto = IPPROTO_GRE;
198 	fl4->fl4_gre_key = gre_key;
199 	return ip_route_output_key(net, fl4);
200 }
201 int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr,
202 			  u8 tos, struct net_device *dev,
203 			  struct in_device *in_dev, u32 *itag);
204 int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src,
205 			 u8 tos, struct net_device *devin);
206 int ip_route_use_hint(struct sk_buff *skb, __be32 dst, __be32 src,
207 		      u8 tos, struct net_device *devin,
208 		      const struct sk_buff *hint);
209 
210 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
211 				 u8 tos, struct net_device *devin)
212 {
213 	int err;
214 
215 	rcu_read_lock();
216 	err = ip_route_input_noref(skb, dst, src, tos, devin);
217 	if (!err) {
218 		skb_dst_force(skb);
219 		if (!skb_dst(skb))
220 			err = -EINVAL;
221 	}
222 	rcu_read_unlock();
223 
224 	return err;
225 }
226 
227 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif,
228 		      u8 protocol);
229 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
230 void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol);
231 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
232 void ip_rt_send_redirect(struct sk_buff *skb);
233 
234 unsigned int inet_addr_type(struct net *net, __be32 addr);
235 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id);
236 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
237 				__be32 addr);
238 unsigned int inet_addr_type_dev_table(struct net *net,
239 				      const struct net_device *dev,
240 				      __be32 addr);
241 void ip_rt_multicast_event(struct in_device *);
242 int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt);
243 void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
244 struct rtable *rt_dst_alloc(struct net_device *dev,
245 			    unsigned int flags, u16 type, bool noxfrm);
246 struct rtable *rt_dst_clone(struct net_device *dev, struct rtable *rt);
247 
248 struct in_ifaddr;
249 void fib_add_ifaddr(struct in_ifaddr *);
250 void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
251 void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric);
252 
253 void rt_add_uncached_list(struct rtable *rt);
254 void rt_del_uncached_list(struct rtable *rt);
255 
256 int fib_dump_info_fnhe(struct sk_buff *skb, struct netlink_callback *cb,
257 		       u32 table_id, struct fib_info *fi,
258 		       int *fa_index, int fa_start, unsigned int flags);
259 
260 static inline void ip_rt_put(struct rtable *rt)
261 {
262 	/* dst_release() accepts a NULL parameter.
263 	 * We rely on dst being first structure in struct rtable
264 	 */
265 	BUILD_BUG_ON(offsetof(struct rtable, dst) != 0);
266 	dst_release(&rt->dst);
267 }
268 
269 extern const __u8 ip_tos2prio[16];
270 
271 static inline char rt_tos2priority(u8 tos)
272 {
273 	return ip_tos2prio[IPTOS_TOS(tos)>>1];
274 }
275 
276 /* ip_route_connect() and ip_route_newports() work in tandem whilst
277  * binding a socket for a new outgoing connection.
278  *
279  * In order to use IPSEC properly, we must, in the end, have a
280  * route that was looked up using all available keys including source
281  * and destination ports.
282  *
283  * However, if a source port needs to be allocated (the user specified
284  * a wildcard source port) we need to obtain addressing information
285  * in order to perform that allocation.
286  *
287  * So ip_route_connect() looks up a route using wildcarded source and
288  * destination ports in the key, simply so that we can get a pair of
289  * addresses to use for port allocation.
290  *
291  * Later, once the ports are allocated, ip_route_newports() will make
292  * another route lookup if needed to make sure we catch any IPSEC
293  * rules keyed on the port information.
294  *
295  * The callers allocate the flow key on their stack, and must pass in
296  * the same flowi4 object to both the ip_route_connect() and the
297  * ip_route_newports() calls.
298  */
299 
300 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst,
301 					 __be32 src, int oif, u8 protocol,
302 					 __be16 sport, __be16 dport,
303 					 const struct sock *sk)
304 {
305 	__u8 flow_flags = 0;
306 
307 	if (inet_test_bit(TRANSPARENT, sk))
308 		flow_flags |= FLOWI_FLAG_ANYSRC;
309 
310 	flowi4_init_output(fl4, oif, READ_ONCE(sk->sk_mark), ip_sock_rt_tos(sk),
311 			   ip_sock_rt_scope(sk), protocol, flow_flags, dst,
312 			   src, dport, sport, sk->sk_uid);
313 }
314 
315 static inline struct rtable *ip_route_connect(struct flowi4 *fl4, __be32 dst,
316 					      __be32 src, int oif, u8 protocol,
317 					      __be16 sport, __be16 dport,
318 					      const struct sock *sk)
319 {
320 	struct net *net = sock_net(sk);
321 	struct rtable *rt;
322 
323 	ip_route_connect_init(fl4, dst, src, oif, protocol, sport, dport, sk);
324 
325 	if (!dst || !src) {
326 		rt = __ip_route_output_key(net, fl4);
327 		if (IS_ERR(rt))
328 			return rt;
329 		ip_rt_put(rt);
330 		flowi4_update_output(fl4, oif, fl4->daddr, fl4->saddr);
331 	}
332 	security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
333 	return ip_route_output_flow(net, fl4, sk);
334 }
335 
336 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
337 					       __be16 orig_sport, __be16 orig_dport,
338 					       __be16 sport, __be16 dport,
339 					       const struct sock *sk)
340 {
341 	if (sport != orig_sport || dport != orig_dport) {
342 		fl4->fl4_dport = dport;
343 		fl4->fl4_sport = sport;
344 		ip_rt_put(rt);
345 		flowi4_update_output(fl4, sk->sk_bound_dev_if, fl4->daddr,
346 				     fl4->saddr);
347 		security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
348 		return ip_route_output_flow(sock_net(sk), fl4, sk);
349 	}
350 	return rt;
351 }
352 
353 static inline int inet_iif(const struct sk_buff *skb)
354 {
355 	struct rtable *rt = skb_rtable(skb);
356 
357 	if (rt && rt->rt_iif)
358 		return rt->rt_iif;
359 
360 	return skb->skb_iif;
361 }
362 
363 static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
364 {
365 	int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
366 	struct net *net = dev_net(dst->dev);
367 
368 	if (hoplimit == 0)
369 		hoplimit = READ_ONCE(net->ipv4.sysctl_ip_default_ttl);
370 	return hoplimit;
371 }
372 
373 static inline struct neighbour *ip_neigh_gw4(struct net_device *dev,
374 					     __be32 daddr)
375 {
376 	struct neighbour *neigh;
377 
378 	neigh = __ipv4_neigh_lookup_noref(dev, (__force u32)daddr);
379 	if (unlikely(!neigh))
380 		neigh = __neigh_create(&arp_tbl, &daddr, dev, false);
381 
382 	return neigh;
383 }
384 
385 static inline struct neighbour *ip_neigh_for_gw(struct rtable *rt,
386 						struct sk_buff *skb,
387 						bool *is_v6gw)
388 {
389 	struct net_device *dev = rt->dst.dev;
390 	struct neighbour *neigh;
391 
392 	if (likely(rt->rt_gw_family == AF_INET)) {
393 		neigh = ip_neigh_gw4(dev, rt->rt_gw4);
394 	} else if (rt->rt_gw_family == AF_INET6) {
395 		neigh = ip_neigh_gw6(dev, &rt->rt_gw6);
396 		*is_v6gw = true;
397 	} else {
398 		neigh = ip_neigh_gw4(dev, ip_hdr(skb)->daddr);
399 	}
400 	return neigh;
401 }
402 
403 #endif	/* _ROUTE_H */
404