xref: /openbmc/linux/include/net/route.h (revision 1edd0337)
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 #define RTO_ONLINK	0x01
39 
40 #define RT_CONN_FLAGS(sk)   (RT_TOS(inet_sk(sk)->tos) | sock_flag(sk, SOCK_LOCALROUTE))
41 #define RT_CONN_FLAGS_TOS(sk,tos)   (RT_TOS(tos) | sock_flag(sk, SOCK_LOCALROUTE))
42 
43 static inline __u8 ip_sock_rt_scope(const struct sock *sk)
44 {
45 	if (sock_flag(sk, SOCK_LOCALROUTE))
46 		return RT_SCOPE_LINK;
47 
48 	return RT_SCOPE_UNIVERSE;
49 }
50 
51 static inline __u8 ip_sock_rt_tos(const struct sock *sk)
52 {
53 	return RT_TOS(inet_sk(sk)->tos);
54 }
55 
56 struct ip_tunnel_info;
57 struct fib_nh;
58 struct fib_info;
59 struct uncached_list;
60 struct rtable {
61 	struct dst_entry	dst;
62 
63 	int			rt_genid;
64 	unsigned int		rt_flags;
65 	__u16			rt_type;
66 	__u8			rt_is_input;
67 	__u8			rt_uses_gateway;
68 
69 	int			rt_iif;
70 
71 	u8			rt_gw_family;
72 	/* Info on neighbour */
73 	union {
74 		__be32		rt_gw4;
75 		struct in6_addr	rt_gw6;
76 	};
77 
78 	/* Miscellaneous cached information */
79 	u32			rt_mtu_locked:1,
80 				rt_pmtu:31;
81 
82 	struct list_head	rt_uncached;
83 	struct uncached_list	*rt_uncached_list;
84 };
85 
86 static inline bool rt_is_input_route(const struct rtable *rt)
87 {
88 	return rt->rt_is_input != 0;
89 }
90 
91 static inline bool rt_is_output_route(const struct rtable *rt)
92 {
93 	return rt->rt_is_input == 0;
94 }
95 
96 static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr)
97 {
98 	if (rt->rt_gw_family == AF_INET)
99 		return rt->rt_gw4;
100 	return daddr;
101 }
102 
103 struct ip_rt_acct {
104 	__u32 	o_bytes;
105 	__u32 	o_packets;
106 	__u32 	i_bytes;
107 	__u32 	i_packets;
108 };
109 
110 struct rt_cache_stat {
111         unsigned int in_slow_tot;
112         unsigned int in_slow_mc;
113         unsigned int in_no_route;
114         unsigned int in_brd;
115         unsigned int in_martian_dst;
116         unsigned int in_martian_src;
117         unsigned int out_slow_tot;
118         unsigned int out_slow_mc;
119 };
120 
121 extern struct ip_rt_acct __percpu *ip_rt_acct;
122 
123 struct in_device;
124 
125 int ip_rt_init(void);
126 void rt_cache_flush(struct net *net);
127 void rt_flush_dev(struct net_device *dev);
128 struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp,
129 					const struct sk_buff *skb);
130 struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp,
131 					    struct fib_result *res,
132 					    const struct sk_buff *skb);
133 
134 static inline struct rtable *__ip_route_output_key(struct net *net,
135 						   struct flowi4 *flp)
136 {
137 	return ip_route_output_key_hash(net, flp, NULL);
138 }
139 
140 struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
141 				    const struct sock *sk);
142 struct rtable *ip_route_output_tunnel(struct sk_buff *skb,
143 				      struct net_device *dev,
144 				      struct net *net, __be32 *saddr,
145 				      const struct ip_tunnel_info *info,
146 				      u8 protocol, bool use_cache);
147 
148 struct dst_entry *ipv4_blackhole_route(struct net *net,
149 				       struct dst_entry *dst_orig);
150 
151 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
152 {
153 	return ip_route_output_flow(net, flp, NULL);
154 }
155 
156 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
157 					     __be32 saddr, u8 tos, int oif)
158 {
159 	struct flowi4 fl4 = {
160 		.flowi4_oif = oif,
161 		.flowi4_tos = tos,
162 		.daddr = daddr,
163 		.saddr = saddr,
164 	};
165 	return ip_route_output_key(net, &fl4);
166 }
167 
168 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
169 						   struct sock *sk,
170 						   __be32 daddr, __be32 saddr,
171 						   __be16 dport, __be16 sport,
172 						   __u8 proto, __u8 tos, int oif)
173 {
174 	flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos,
175 			   RT_SCOPE_UNIVERSE, proto,
176 			   sk ? inet_sk_flowi_flags(sk) : 0,
177 			   daddr, saddr, dport, sport, sock_net_uid(net, sk));
178 	if (sk)
179 		security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
180 	return ip_route_output_flow(net, fl4, sk);
181 }
182 
183 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4,
184 						 __be32 daddr, __be32 saddr,
185 						 __be32 gre_key, __u8 tos, int oif)
186 {
187 	memset(fl4, 0, sizeof(*fl4));
188 	fl4->flowi4_oif = oif;
189 	fl4->daddr = daddr;
190 	fl4->saddr = saddr;
191 	fl4->flowi4_tos = tos;
192 	fl4->flowi4_proto = IPPROTO_GRE;
193 	fl4->fl4_gre_key = gre_key;
194 	return ip_route_output_key(net, fl4);
195 }
196 int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr,
197 			  u8 tos, struct net_device *dev,
198 			  struct in_device *in_dev, u32 *itag);
199 int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src,
200 			 u8 tos, struct net_device *devin);
201 int ip_route_use_hint(struct sk_buff *skb, __be32 dst, __be32 src,
202 		      u8 tos, struct net_device *devin,
203 		      const struct sk_buff *hint);
204 
205 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
206 				 u8 tos, struct net_device *devin)
207 {
208 	int err;
209 
210 	rcu_read_lock();
211 	err = ip_route_input_noref(skb, dst, src, tos, devin);
212 	if (!err) {
213 		skb_dst_force(skb);
214 		if (!skb_dst(skb))
215 			err = -EINVAL;
216 	}
217 	rcu_read_unlock();
218 
219 	return err;
220 }
221 
222 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif,
223 		      u8 protocol);
224 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
225 void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol);
226 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
227 void ip_rt_send_redirect(struct sk_buff *skb);
228 
229 unsigned int inet_addr_type(struct net *net, __be32 addr);
230 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id);
231 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
232 				__be32 addr);
233 unsigned int inet_addr_type_dev_table(struct net *net,
234 				      const struct net_device *dev,
235 				      __be32 addr);
236 void ip_rt_multicast_event(struct in_device *);
237 int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt);
238 void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
239 struct rtable *rt_dst_alloc(struct net_device *dev,
240 			    unsigned int flags, u16 type, bool noxfrm);
241 struct rtable *rt_dst_clone(struct net_device *dev, struct rtable *rt);
242 
243 struct in_ifaddr;
244 void fib_add_ifaddr(struct in_ifaddr *);
245 void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
246 void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric);
247 
248 void rt_add_uncached_list(struct rtable *rt);
249 void rt_del_uncached_list(struct rtable *rt);
250 
251 int fib_dump_info_fnhe(struct sk_buff *skb, struct netlink_callback *cb,
252 		       u32 table_id, struct fib_info *fi,
253 		       int *fa_index, int fa_start, unsigned int flags);
254 
255 static inline void ip_rt_put(struct rtable *rt)
256 {
257 	/* dst_release() accepts a NULL parameter.
258 	 * We rely on dst being first structure in struct rtable
259 	 */
260 	BUILD_BUG_ON(offsetof(struct rtable, dst) != 0);
261 	dst_release(&rt->dst);
262 }
263 
264 #define IPTOS_RT_MASK	(IPTOS_TOS_MASK & ~3)
265 
266 extern const __u8 ip_tos2prio[16];
267 
268 static inline char rt_tos2priority(u8 tos)
269 {
270 	return ip_tos2prio[IPTOS_TOS(tos)>>1];
271 }
272 
273 /* ip_route_connect() and ip_route_newports() work in tandem whilst
274  * binding a socket for a new outgoing connection.
275  *
276  * In order to use IPSEC properly, we must, in the end, have a
277  * route that was looked up using all available keys including source
278  * and destination ports.
279  *
280  * However, if a source port needs to be allocated (the user specified
281  * a wildcard source port) we need to obtain addressing information
282  * in order to perform that allocation.
283  *
284  * So ip_route_connect() looks up a route using wildcarded source and
285  * destination ports in the key, simply so that we can get a pair of
286  * addresses to use for port allocation.
287  *
288  * Later, once the ports are allocated, ip_route_newports() will make
289  * another route lookup if needed to make sure we catch any IPSEC
290  * rules keyed on the port information.
291  *
292  * The callers allocate the flow key on their stack, and must pass in
293  * the same flowi4 object to both the ip_route_connect() and the
294  * ip_route_newports() calls.
295  */
296 
297 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst,
298 					 __be32 src, int oif, u8 protocol,
299 					 __be16 sport, __be16 dport,
300 					 const struct sock *sk)
301 {
302 	__u8 flow_flags = 0;
303 
304 	if (inet_sk(sk)->transparent)
305 		flow_flags |= FLOWI_FLAG_ANYSRC;
306 
307 	flowi4_init_output(fl4, oif, sk->sk_mark, ip_sock_rt_tos(sk),
308 			   ip_sock_rt_scope(sk), protocol, flow_flags, dst,
309 			   src, dport, sport, sk->sk_uid);
310 }
311 
312 static inline struct rtable *ip_route_connect(struct flowi4 *fl4, __be32 dst,
313 					      __be32 src, int oif, u8 protocol,
314 					      __be16 sport, __be16 dport,
315 					      struct sock *sk)
316 {
317 	struct net *net = sock_net(sk);
318 	struct rtable *rt;
319 
320 	ip_route_connect_init(fl4, dst, src, oif, protocol, sport, dport, sk);
321 
322 	if (!dst || !src) {
323 		rt = __ip_route_output_key(net, fl4);
324 		if (IS_ERR(rt))
325 			return rt;
326 		ip_rt_put(rt);
327 		flowi4_update_output(fl4, oif, fl4->flowi4_tos, fl4->daddr,
328 				     fl4->saddr);
329 	}
330 	security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
331 	return ip_route_output_flow(net, fl4, sk);
332 }
333 
334 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
335 					       __be16 orig_sport, __be16 orig_dport,
336 					       __be16 sport, __be16 dport,
337 					       struct sock *sk)
338 {
339 	if (sport != orig_sport || dport != orig_dport) {
340 		fl4->fl4_dport = dport;
341 		fl4->fl4_sport = sport;
342 		ip_rt_put(rt);
343 		flowi4_update_output(fl4, sk->sk_bound_dev_if,
344 				     RT_CONN_FLAGS(sk), fl4->daddr,
345 				     fl4->saddr);
346 		security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
347 		return ip_route_output_flow(sock_net(sk), fl4, sk);
348 	}
349 	return rt;
350 }
351 
352 static inline int inet_iif(const struct sk_buff *skb)
353 {
354 	struct rtable *rt = skb_rtable(skb);
355 
356 	if (rt && rt->rt_iif)
357 		return rt->rt_iif;
358 
359 	return skb->skb_iif;
360 }
361 
362 static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
363 {
364 	int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
365 	struct net *net = dev_net(dst->dev);
366 
367 	if (hoplimit == 0)
368 		hoplimit = READ_ONCE(net->ipv4.sysctl_ip_default_ttl);
369 	return hoplimit;
370 }
371 
372 static inline struct neighbour *ip_neigh_gw4(struct net_device *dev,
373 					     __be32 daddr)
374 {
375 	struct neighbour *neigh;
376 
377 	neigh = __ipv4_neigh_lookup_noref(dev, (__force u32)daddr);
378 	if (unlikely(!neigh))
379 		neigh = __neigh_create(&arp_tbl, &daddr, dev, false);
380 
381 	return neigh;
382 }
383 
384 static inline struct neighbour *ip_neigh_for_gw(struct rtable *rt,
385 						struct sk_buff *skb,
386 						bool *is_v6gw)
387 {
388 	struct net_device *dev = rt->dst.dev;
389 	struct neighbour *neigh;
390 
391 	if (likely(rt->rt_gw_family == AF_INET)) {
392 		neigh = ip_neigh_gw4(dev, rt->rt_gw4);
393 	} else if (rt->rt_gw_family == AF_INET6) {
394 		neigh = ip_neigh_gw6(dev, &rt->rt_gw6);
395 		*is_v6gw = true;
396 	} else {
397 		neigh = ip_neigh_gw4(dev, ip_hdr(skb)->daddr);
398 	}
399 	return neigh;
400 }
401 
402 #endif	/* _ROUTE_H */
403