xref: /openbmc/linux/include/net/route.h (revision ae213c44)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET  is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		Definitions for the IP router.
7  *
8  * Version:	@(#)route.h	1.0.4	05/27/93
9  *
10  * Authors:	Ross Biro
11  *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12  * Fixes:
13  *		Alan Cox	:	Reformatted. Added ip_rt_local()
14  *		Alan Cox	:	Support for TCP parameters.
15  *		Alexey Kuznetsov:	Major changes for new routing code.
16  *		Mike McLagan    :	Routing by source
17  *		Robert Olsson   :	Added rt_cache statistics
18  *
19  *		This program is free software; you can redistribute it and/or
20  *		modify it under the terms of the GNU General Public License
21  *		as published by the Free Software Foundation; either version
22  *		2 of the License, or (at your option) any later version.
23  */
24 #ifndef _ROUTE_H
25 #define _ROUTE_H
26 
27 #include <net/dst.h>
28 #include <net/inetpeer.h>
29 #include <net/flow.h>
30 #include <net/inet_sock.h>
31 #include <net/ip_fib.h>
32 #include <net/arp.h>
33 #include <net/ndisc.h>
34 #include <linux/in_route.h>
35 #include <linux/rtnetlink.h>
36 #include <linux/rcupdate.h>
37 #include <linux/route.h>
38 #include <linux/ip.h>
39 #include <linux/cache.h>
40 #include <linux/security.h>
41 
42 /* IPv4 datagram length is stored into 16bit field (tot_len) */
43 #define IP_MAX_MTU	0xFFFFU
44 
45 #define RTO_ONLINK	0x01
46 
47 #define RT_CONN_FLAGS(sk)   (RT_TOS(inet_sk(sk)->tos) | sock_flag(sk, SOCK_LOCALROUTE))
48 #define RT_CONN_FLAGS_TOS(sk,tos)   (RT_TOS(tos) | sock_flag(sk, SOCK_LOCALROUTE))
49 
50 struct fib_nh;
51 struct fib_info;
52 struct uncached_list;
53 struct rtable {
54 	struct dst_entry	dst;
55 
56 	int			rt_genid;
57 	unsigned int		rt_flags;
58 	__u16			rt_type;
59 	__u8			rt_is_input;
60 	u8			rt_gw_family;
61 
62 	int			rt_iif;
63 
64 	/* Info on neighbour */
65 	union {
66 		__be32		rt_gw4;
67 		struct in6_addr	rt_gw6;
68 	};
69 
70 	/* Miscellaneous cached information */
71 	u32			rt_mtu_locked:1,
72 				rt_pmtu:31;
73 
74 	struct list_head	rt_uncached;
75 	struct uncached_list	*rt_uncached_list;
76 };
77 
78 static inline bool rt_is_input_route(const struct rtable *rt)
79 {
80 	return rt->rt_is_input != 0;
81 }
82 
83 static inline bool rt_is_output_route(const struct rtable *rt)
84 {
85 	return rt->rt_is_input == 0;
86 }
87 
88 static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr)
89 {
90 	if (rt->rt_gw_family == AF_INET)
91 		return rt->rt_gw4;
92 	return daddr;
93 }
94 
95 struct ip_rt_acct {
96 	__u32 	o_bytes;
97 	__u32 	o_packets;
98 	__u32 	i_bytes;
99 	__u32 	i_packets;
100 };
101 
102 struct rt_cache_stat {
103         unsigned int in_slow_tot;
104         unsigned int in_slow_mc;
105         unsigned int in_no_route;
106         unsigned int in_brd;
107         unsigned int in_martian_dst;
108         unsigned int in_martian_src;
109         unsigned int out_slow_tot;
110         unsigned int out_slow_mc;
111 };
112 
113 extern struct ip_rt_acct __percpu *ip_rt_acct;
114 
115 struct in_device;
116 
117 int ip_rt_init(void);
118 void rt_cache_flush(struct net *net);
119 void rt_flush_dev(struct net_device *dev);
120 struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp,
121 					const struct sk_buff *skb);
122 struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp,
123 					    struct fib_result *res,
124 					    const struct sk_buff *skb);
125 
126 static inline struct rtable *__ip_route_output_key(struct net *net,
127 						   struct flowi4 *flp)
128 {
129 	return ip_route_output_key_hash(net, flp, NULL);
130 }
131 
132 struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
133 				    const struct sock *sk);
134 struct dst_entry *ipv4_blackhole_route(struct net *net,
135 				       struct dst_entry *dst_orig);
136 
137 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
138 {
139 	return ip_route_output_flow(net, flp, NULL);
140 }
141 
142 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
143 					     __be32 saddr, u8 tos, int oif)
144 {
145 	struct flowi4 fl4 = {
146 		.flowi4_oif = oif,
147 		.flowi4_tos = tos,
148 		.daddr = daddr,
149 		.saddr = saddr,
150 	};
151 	return ip_route_output_key(net, &fl4);
152 }
153 
154 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
155 						   struct sock *sk,
156 						   __be32 daddr, __be32 saddr,
157 						   __be16 dport, __be16 sport,
158 						   __u8 proto, __u8 tos, int oif)
159 {
160 	flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos,
161 			   RT_SCOPE_UNIVERSE, proto,
162 			   sk ? inet_sk_flowi_flags(sk) : 0,
163 			   daddr, saddr, dport, sport, sock_net_uid(net, sk));
164 	if (sk)
165 		security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
166 	return ip_route_output_flow(net, fl4, sk);
167 }
168 
169 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4,
170 						 __be32 daddr, __be32 saddr,
171 						 __be32 gre_key, __u8 tos, int oif)
172 {
173 	memset(fl4, 0, sizeof(*fl4));
174 	fl4->flowi4_oif = oif;
175 	fl4->daddr = daddr;
176 	fl4->saddr = saddr;
177 	fl4->flowi4_tos = tos;
178 	fl4->flowi4_proto = IPPROTO_GRE;
179 	fl4->fl4_gre_key = gre_key;
180 	return ip_route_output_key(net, fl4);
181 }
182 int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr,
183 			  u8 tos, struct net_device *dev,
184 			  struct in_device *in_dev, u32 *itag);
185 int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src,
186 			 u8 tos, struct net_device *devin);
187 int ip_route_input_rcu(struct sk_buff *skb, __be32 dst, __be32 src,
188 		       u8 tos, struct net_device *devin,
189 		       struct fib_result *res);
190 
191 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
192 				 u8 tos, struct net_device *devin)
193 {
194 	int err;
195 
196 	rcu_read_lock();
197 	err = ip_route_input_noref(skb, dst, src, tos, devin);
198 	if (!err) {
199 		skb_dst_force(skb);
200 		if (!skb_dst(skb))
201 			err = -EINVAL;
202 	}
203 	rcu_read_unlock();
204 
205 	return err;
206 }
207 
208 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif,
209 		      u8 protocol);
210 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
211 void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol);
212 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
213 void ip_rt_send_redirect(struct sk_buff *skb);
214 
215 unsigned int inet_addr_type(struct net *net, __be32 addr);
216 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id);
217 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
218 				__be32 addr);
219 unsigned int inet_addr_type_dev_table(struct net *net,
220 				      const struct net_device *dev,
221 				      __be32 addr);
222 void ip_rt_multicast_event(struct in_device *);
223 int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt);
224 void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
225 struct rtable *rt_dst_alloc(struct net_device *dev,
226 			     unsigned int flags, u16 type,
227 			     bool nopolicy, bool noxfrm, bool will_cache);
228 
229 struct in_ifaddr;
230 void fib_add_ifaddr(struct in_ifaddr *);
231 void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
232 void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric);
233 
234 void rt_add_uncached_list(struct rtable *rt);
235 void rt_del_uncached_list(struct rtable *rt);
236 
237 static inline void ip_rt_put(struct rtable *rt)
238 {
239 	/* dst_release() accepts a NULL parameter.
240 	 * We rely on dst being first structure in struct rtable
241 	 */
242 	BUILD_BUG_ON(offsetof(struct rtable, dst) != 0);
243 	dst_release(&rt->dst);
244 }
245 
246 #define IPTOS_RT_MASK	(IPTOS_TOS_MASK & ~3)
247 
248 extern const __u8 ip_tos2prio[16];
249 
250 static inline char rt_tos2priority(u8 tos)
251 {
252 	return ip_tos2prio[IPTOS_TOS(tos)>>1];
253 }
254 
255 /* ip_route_connect() and ip_route_newports() work in tandem whilst
256  * binding a socket for a new outgoing connection.
257  *
258  * In order to use IPSEC properly, we must, in the end, have a
259  * route that was looked up using all available keys including source
260  * and destination ports.
261  *
262  * However, if a source port needs to be allocated (the user specified
263  * a wildcard source port) we need to obtain addressing information
264  * in order to perform that allocation.
265  *
266  * So ip_route_connect() looks up a route using wildcarded source and
267  * destination ports in the key, simply so that we can get a pair of
268  * addresses to use for port allocation.
269  *
270  * Later, once the ports are allocated, ip_route_newports() will make
271  * another route lookup if needed to make sure we catch any IPSEC
272  * rules keyed on the port information.
273  *
274  * The callers allocate the flow key on their stack, and must pass in
275  * the same flowi4 object to both the ip_route_connect() and the
276  * ip_route_newports() calls.
277  */
278 
279 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src,
280 					 u32 tos, int oif, u8 protocol,
281 					 __be16 sport, __be16 dport,
282 					 struct sock *sk)
283 {
284 	__u8 flow_flags = 0;
285 
286 	if (inet_sk(sk)->transparent)
287 		flow_flags |= FLOWI_FLAG_ANYSRC;
288 
289 	flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE,
290 			   protocol, flow_flags, dst, src, dport, sport,
291 			   sk->sk_uid);
292 }
293 
294 static inline struct rtable *ip_route_connect(struct flowi4 *fl4,
295 					      __be32 dst, __be32 src, u32 tos,
296 					      int oif, u8 protocol,
297 					      __be16 sport, __be16 dport,
298 					      struct sock *sk)
299 {
300 	struct net *net = sock_net(sk);
301 	struct rtable *rt;
302 
303 	ip_route_connect_init(fl4, dst, src, tos, oif, protocol,
304 			      sport, dport, sk);
305 
306 	if (!dst || !src) {
307 		rt = __ip_route_output_key(net, fl4);
308 		if (IS_ERR(rt))
309 			return rt;
310 		ip_rt_put(rt);
311 		flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr);
312 	}
313 	security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
314 	return ip_route_output_flow(net, fl4, sk);
315 }
316 
317 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
318 					       __be16 orig_sport, __be16 orig_dport,
319 					       __be16 sport, __be16 dport,
320 					       struct sock *sk)
321 {
322 	if (sport != orig_sport || dport != orig_dport) {
323 		fl4->fl4_dport = dport;
324 		fl4->fl4_sport = sport;
325 		ip_rt_put(rt);
326 		flowi4_update_output(fl4, sk->sk_bound_dev_if,
327 				     RT_CONN_FLAGS(sk), fl4->daddr,
328 				     fl4->saddr);
329 		security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
330 		return ip_route_output_flow(sock_net(sk), fl4, sk);
331 	}
332 	return rt;
333 }
334 
335 static inline int inet_iif(const struct sk_buff *skb)
336 {
337 	struct rtable *rt = skb_rtable(skb);
338 
339 	if (rt && rt->rt_iif)
340 		return rt->rt_iif;
341 
342 	return skb->skb_iif;
343 }
344 
345 static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
346 {
347 	int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
348 	struct net *net = dev_net(dst->dev);
349 
350 	if (hoplimit == 0)
351 		hoplimit = net->ipv4.sysctl_ip_default_ttl;
352 	return hoplimit;
353 }
354 
355 static inline struct neighbour *ip_neigh_gw4(struct net_device *dev,
356 					     __be32 daddr)
357 {
358 	struct neighbour *neigh;
359 
360 	neigh = __ipv4_neigh_lookup_noref(dev, daddr);
361 	if (unlikely(!neigh))
362 		neigh = __neigh_create(&arp_tbl, &daddr, dev, false);
363 
364 	return neigh;
365 }
366 
367 static inline struct neighbour *ip_neigh_for_gw(struct rtable *rt,
368 						struct sk_buff *skb,
369 						bool *is_v6gw)
370 {
371 	struct net_device *dev = rt->dst.dev;
372 	struct neighbour *neigh;
373 
374 	if (likely(rt->rt_gw_family == AF_INET)) {
375 		neigh = ip_neigh_gw4(dev, rt->rt_gw4);
376 	} else if (rt->rt_gw_family == AF_INET6) {
377 		neigh = ip_neigh_gw6(dev, &rt->rt_gw6);
378 		*is_v6gw = true;
379 	} else {
380 		neigh = ip_neigh_gw4(dev, ip_hdr(skb)->daddr);
381 	}
382 	return neigh;
383 }
384 
385 #endif	/* _ROUTE_H */
386