xref: /linux/include/net/route.h (revision 300a90b2cb5d442879e6398920c49aebbd5c8e40)
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 static inline __u8 ip_sock_rt_scope(const struct sock *sk)
39 {
40 	if (sock_flag(sk, SOCK_LOCALROUTE))
41 		return RT_SCOPE_LINK;
42 
43 	return RT_SCOPE_UNIVERSE;
44 }
45 
46 static inline __u8 ip_sock_rt_tos(const struct sock *sk)
47 {
48 	return RT_TOS(READ_ONCE(inet_sk(sk)->tos));
49 }
50 
51 struct ip_tunnel_info;
52 struct fib_nh;
53 struct fib_info;
54 struct uncached_list;
55 struct rtable {
56 	struct dst_entry	dst;
57 
58 	int			rt_genid;
59 	unsigned int		rt_flags;
60 	__u16			rt_type;
61 	__u8			rt_is_input;
62 	__u8			rt_uses_gateway;
63 
64 	int			rt_iif;
65 
66 	u8			rt_gw_family;
67 	/* Info on neighbour */
68 	union {
69 		__be32		rt_gw4;
70 		struct in6_addr	rt_gw6;
71 	};
72 
73 	/* Miscellaneous cached information */
74 	u32			rt_mtu_locked:1,
75 				rt_pmtu:31;
76 };
77 
78 #define dst_rtable(_ptr) container_of_const(_ptr, struct rtable, dst)
79 
80 /**
81  * skb_rtable - Returns the skb &rtable
82  * @skb: buffer
83  */
84 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
85 {
86 	return dst_rtable(skb_dst(skb));
87 }
88 
89 static inline bool rt_is_input_route(const struct rtable *rt)
90 {
91 	return rt->rt_is_input != 0;
92 }
93 
94 static inline bool rt_is_output_route(const struct rtable *rt)
95 {
96 	return rt->rt_is_input == 0;
97 }
98 
99 static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr)
100 {
101 	if (rt->rt_gw_family == AF_INET)
102 		return rt->rt_gw4;
103 	return daddr;
104 }
105 
106 struct ip_rt_acct {
107 	__u32 	o_bytes;
108 	__u32 	o_packets;
109 	__u32 	i_bytes;
110 	__u32 	i_packets;
111 };
112 
113 struct rt_cache_stat {
114         unsigned int in_slow_tot;
115         unsigned int in_slow_mc;
116         unsigned int in_no_route;
117         unsigned int in_brd;
118         unsigned int in_martian_dst;
119         unsigned int in_martian_src;
120         unsigned int out_slow_tot;
121         unsigned int out_slow_mc;
122 };
123 
124 extern struct ip_rt_acct __percpu *ip_rt_acct;
125 
126 struct in_device;
127 
128 int ip_rt_init(void);
129 void rt_cache_flush(struct net *net);
130 void rt_flush_dev(struct net_device *dev);
131 struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp,
132 					const struct sk_buff *skb);
133 struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp,
134 					    struct fib_result *res,
135 					    const struct sk_buff *skb);
136 
137 static inline struct rtable *__ip_route_output_key(struct net *net,
138 						   struct flowi4 *flp)
139 {
140 	return ip_route_output_key_hash(net, flp, NULL);
141 }
142 
143 struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
144 				    const struct sock *sk);
145 struct dst_entry *ipv4_blackhole_route(struct net *net,
146 				       struct dst_entry *dst_orig);
147 
148 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
149 {
150 	return ip_route_output_flow(net, flp, NULL);
151 }
152 
153 /* Simplistic IPv4 route lookup function.
154  * This is only suitable for some particular use cases: since the flowi4
155  * structure is only partially set, it may bypass some fib-rules.
156  */
157 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
158 					     __be32 saddr, u8 tos, int oif,
159 					     __u8 scope)
160 {
161 	struct flowi4 fl4 = {
162 		.flowi4_oif = oif,
163 		.flowi4_tos = tos,
164 		.flowi4_scope = scope,
165 		.daddr = daddr,
166 		.saddr = saddr,
167 	};
168 
169 	return ip_route_output_key(net, &fl4);
170 }
171 
172 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
173 						   const struct sock *sk,
174 						   __be32 daddr, __be32 saddr,
175 						   __be16 dport, __be16 sport,
176 						   __u8 proto, __u8 tos, int oif)
177 {
178 	flowi4_init_output(fl4, oif, sk ? READ_ONCE(sk->sk_mark) : 0, tos,
179 			   sk ? ip_sock_rt_scope(sk) : RT_SCOPE_UNIVERSE,
180 			   proto, sk ? inet_sk_flowi_flags(sk) : 0,
181 			   daddr, saddr, dport, sport, sock_net_uid(net, sk));
182 	if (sk)
183 		security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
184 	return ip_route_output_flow(net, fl4, sk);
185 }
186 
187 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4,
188 						 __be32 daddr, __be32 saddr,
189 						 __be32 gre_key, __u8 tos, int oif)
190 {
191 	memset(fl4, 0, sizeof(*fl4));
192 	fl4->flowi4_oif = oif;
193 	fl4->daddr = daddr;
194 	fl4->saddr = saddr;
195 	fl4->flowi4_tos = tos;
196 	fl4->flowi4_proto = IPPROTO_GRE;
197 	fl4->fl4_gre_key = gre_key;
198 	return ip_route_output_key(net, fl4);
199 }
200 int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr,
201 			  u8 tos, struct net_device *dev,
202 			  struct in_device *in_dev, u32 *itag);
203 int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src,
204 			 u8 tos, struct net_device *devin);
205 int ip_route_use_hint(struct sk_buff *skb, __be32 dst, __be32 src,
206 		      u8 tos, struct net_device *devin,
207 		      const struct sk_buff *hint);
208 
209 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
210 				 u8 tos, struct net_device *devin)
211 {
212 	int err;
213 
214 	rcu_read_lock();
215 	err = ip_route_input_noref(skb, dst, src, tos, devin);
216 	if (!err) {
217 		skb_dst_force(skb);
218 		if (!skb_dst(skb))
219 			err = -EINVAL;
220 	}
221 	rcu_read_unlock();
222 
223 	return err;
224 }
225 
226 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif,
227 		      u8 protocol);
228 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
229 void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol);
230 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
231 void ip_rt_send_redirect(struct sk_buff *skb);
232 
233 unsigned int inet_addr_type(struct net *net, __be32 addr);
234 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id);
235 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
236 				__be32 addr);
237 unsigned int inet_addr_type_dev_table(struct net *net,
238 				      const struct net_device *dev,
239 				      __be32 addr);
240 void ip_rt_multicast_event(struct in_device *);
241 int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt);
242 void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
243 struct rtable *rt_dst_alloc(struct net_device *dev,
244 			    unsigned int flags, u16 type, bool noxfrm);
245 struct rtable *rt_dst_clone(struct net_device *dev, struct rtable *rt);
246 
247 struct in_ifaddr;
248 void fib_add_ifaddr(struct in_ifaddr *);
249 void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
250 void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric);
251 
252 void rt_add_uncached_list(struct rtable *rt);
253 void rt_del_uncached_list(struct rtable *rt);
254 
255 int fib_dump_info_fnhe(struct sk_buff *skb, struct netlink_callback *cb,
256 		       u32 table_id, struct fib_info *fi,
257 		       int *fa_index, int fa_start, unsigned int flags);
258 
259 static inline void ip_rt_put(struct rtable *rt)
260 {
261 	/* dst_release() accepts a NULL parameter.
262 	 * We rely on dst being first structure in struct rtable
263 	 */
264 	BUILD_BUG_ON(offsetof(struct rtable, dst) != 0);
265 	dst_release(&rt->dst);
266 }
267 
268 #define IPTOS_RT_MASK	(IPTOS_TOS_MASK & ~3)
269 
270 extern const __u8 ip_tos2prio[16];
271 
272 static inline char rt_tos2priority(u8 tos)
273 {
274 	return ip_tos2prio[IPTOS_TOS(tos)>>1];
275 }
276 
277 /* ip_route_connect() and ip_route_newports() work in tandem whilst
278  * binding a socket for a new outgoing connection.
279  *
280  * In order to use IPSEC properly, we must, in the end, have a
281  * route that was looked up using all available keys including source
282  * and destination ports.
283  *
284  * However, if a source port needs to be allocated (the user specified
285  * a wildcard source port) we need to obtain addressing information
286  * in order to perform that allocation.
287  *
288  * So ip_route_connect() looks up a route using wildcarded source and
289  * destination ports in the key, simply so that we can get a pair of
290  * addresses to use for port allocation.
291  *
292  * Later, once the ports are allocated, ip_route_newports() will make
293  * another route lookup if needed to make sure we catch any IPSEC
294  * rules keyed on the port information.
295  *
296  * The callers allocate the flow key on their stack, and must pass in
297  * the same flowi4 object to both the ip_route_connect() and the
298  * ip_route_newports() calls.
299  */
300 
301 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst,
302 					 __be32 src, int oif, u8 protocol,
303 					 __be16 sport, __be16 dport,
304 					 const struct sock *sk)
305 {
306 	__u8 flow_flags = 0;
307 
308 	if (inet_test_bit(TRANSPARENT, sk))
309 		flow_flags |= FLOWI_FLAG_ANYSRC;
310 
311 	flowi4_init_output(fl4, oif, READ_ONCE(sk->sk_mark), ip_sock_rt_tos(sk),
312 			   ip_sock_rt_scope(sk), protocol, flow_flags, dst,
313 			   src, dport, sport, sk->sk_uid);
314 }
315 
316 static inline struct rtable *ip_route_connect(struct flowi4 *fl4, __be32 dst,
317 					      __be32 src, int oif, u8 protocol,
318 					      __be16 sport, __be16 dport,
319 					      const struct sock *sk)
320 {
321 	struct net *net = sock_net(sk);
322 	struct rtable *rt;
323 
324 	ip_route_connect_init(fl4, dst, src, oif, protocol, sport, dport, sk);
325 
326 	if (!dst || !src) {
327 		rt = __ip_route_output_key(net, fl4);
328 		if (IS_ERR(rt))
329 			return rt;
330 		ip_rt_put(rt);
331 		flowi4_update_output(fl4, oif, fl4->daddr, fl4->saddr);
332 	}
333 	security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
334 	return ip_route_output_flow(net, fl4, sk);
335 }
336 
337 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
338 					       __be16 orig_sport, __be16 orig_dport,
339 					       __be16 sport, __be16 dport,
340 					       const struct sock *sk)
341 {
342 	if (sport != orig_sport || dport != orig_dport) {
343 		fl4->fl4_dport = dport;
344 		fl4->fl4_sport = sport;
345 		ip_rt_put(rt);
346 		flowi4_update_output(fl4, sk->sk_bound_dev_if, fl4->daddr,
347 				     fl4->saddr);
348 		security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
349 		return ip_route_output_flow(sock_net(sk), fl4, sk);
350 	}
351 	return rt;
352 }
353 
354 static inline int inet_iif(const struct sk_buff *skb)
355 {
356 	struct rtable *rt = skb_rtable(skb);
357 
358 	if (rt && rt->rt_iif)
359 		return rt->rt_iif;
360 
361 	return skb->skb_iif;
362 }
363 
364 static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
365 {
366 	int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
367 	struct net *net = dev_net(dst->dev);
368 
369 	if (hoplimit == 0)
370 		hoplimit = READ_ONCE(net->ipv4.sysctl_ip_default_ttl);
371 	return hoplimit;
372 }
373 
374 static inline struct neighbour *ip_neigh_gw4(struct net_device *dev,
375 					     __be32 daddr)
376 {
377 	struct neighbour *neigh;
378 
379 	neigh = __ipv4_neigh_lookup_noref(dev, (__force u32)daddr);
380 	if (unlikely(!neigh))
381 		neigh = __neigh_create(&arp_tbl, &daddr, dev, false);
382 
383 	return neigh;
384 }
385 
386 static inline struct neighbour *ip_neigh_for_gw(struct rtable *rt,
387 						struct sk_buff *skb,
388 						bool *is_v6gw)
389 {
390 	struct net_device *dev = rt->dst.dev;
391 	struct neighbour *neigh;
392 
393 	if (likely(rt->rt_gw_family == AF_INET)) {
394 		neigh = ip_neigh_gw4(dev, rt->rt_gw4);
395 	} else if (rt->rt_gw_family == AF_INET6) {
396 		neigh = ip_neigh_gw6(dev, &rt->rt_gw6);
397 		*is_v6gw = true;
398 	} else {
399 		neigh = ip_neigh_gw4(dev, ip_hdr(skb)->daddr);
400 	}
401 	return neigh;
402 }
403 
404 #endif	/* _ROUTE_H */
405