xref: /openbmc/linux/include/net/route.h (revision 5e012745)
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 /* IPv4 datagram length is stored into 16bit field (tot_len) */
39 #define IP_MAX_MTU	0xFFFFU
40 
41 #define RTO_ONLINK	0x01
42 
43 #define RT_CONN_FLAGS(sk)   (RT_TOS(inet_sk(sk)->tos) | sock_flag(sk, SOCK_LOCALROUTE))
44 #define RT_CONN_FLAGS_TOS(sk,tos)   (RT_TOS(tos) | sock_flag(sk, SOCK_LOCALROUTE))
45 
46 struct fib_nh;
47 struct fib_info;
48 struct uncached_list;
49 struct rtable {
50 	struct dst_entry	dst;
51 
52 	int			rt_genid;
53 	unsigned int		rt_flags;
54 	__u16			rt_type;
55 	__u8			rt_is_input;
56 	u8			rt_gw_family;
57 
58 	int			rt_iif;
59 
60 	/* Info on neighbour */
61 	union {
62 		__be32		rt_gw4;
63 		struct in6_addr	rt_gw6;
64 	};
65 
66 	/* Miscellaneous cached information */
67 	u32			rt_mtu_locked:1,
68 				rt_pmtu:31;
69 
70 	struct list_head	rt_uncached;
71 	struct uncached_list	*rt_uncached_list;
72 };
73 
74 static inline bool rt_is_input_route(const struct rtable *rt)
75 {
76 	return rt->rt_is_input != 0;
77 }
78 
79 static inline bool rt_is_output_route(const struct rtable *rt)
80 {
81 	return rt->rt_is_input == 0;
82 }
83 
84 static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr)
85 {
86 	if (rt->rt_gw_family == AF_INET)
87 		return rt->rt_gw4;
88 	return daddr;
89 }
90 
91 struct ip_rt_acct {
92 	__u32 	o_bytes;
93 	__u32 	o_packets;
94 	__u32 	i_bytes;
95 	__u32 	i_packets;
96 };
97 
98 struct rt_cache_stat {
99         unsigned int in_slow_tot;
100         unsigned int in_slow_mc;
101         unsigned int in_no_route;
102         unsigned int in_brd;
103         unsigned int in_martian_dst;
104         unsigned int in_martian_src;
105         unsigned int out_slow_tot;
106         unsigned int out_slow_mc;
107 };
108 
109 extern struct ip_rt_acct __percpu *ip_rt_acct;
110 
111 struct in_device;
112 
113 int ip_rt_init(void);
114 void rt_cache_flush(struct net *net);
115 void rt_flush_dev(struct net_device *dev);
116 struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp,
117 					const struct sk_buff *skb);
118 struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp,
119 					    struct fib_result *res,
120 					    const struct sk_buff *skb);
121 
122 static inline struct rtable *__ip_route_output_key(struct net *net,
123 						   struct flowi4 *flp)
124 {
125 	return ip_route_output_key_hash(net, flp, NULL);
126 }
127 
128 struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
129 				    const struct sock *sk);
130 struct dst_entry *ipv4_blackhole_route(struct net *net,
131 				       struct dst_entry *dst_orig);
132 
133 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
134 {
135 	return ip_route_output_flow(net, flp, NULL);
136 }
137 
138 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
139 					     __be32 saddr, u8 tos, int oif)
140 {
141 	struct flowi4 fl4 = {
142 		.flowi4_oif = oif,
143 		.flowi4_tos = tos,
144 		.daddr = daddr,
145 		.saddr = saddr,
146 	};
147 	return ip_route_output_key(net, &fl4);
148 }
149 
150 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
151 						   struct sock *sk,
152 						   __be32 daddr, __be32 saddr,
153 						   __be16 dport, __be16 sport,
154 						   __u8 proto, __u8 tos, int oif)
155 {
156 	flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos,
157 			   RT_SCOPE_UNIVERSE, proto,
158 			   sk ? inet_sk_flowi_flags(sk) : 0,
159 			   daddr, saddr, dport, sport, sock_net_uid(net, sk));
160 	if (sk)
161 		security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
162 	return ip_route_output_flow(net, fl4, sk);
163 }
164 
165 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4,
166 						 __be32 daddr, __be32 saddr,
167 						 __be32 gre_key, __u8 tos, int oif)
168 {
169 	memset(fl4, 0, sizeof(*fl4));
170 	fl4->flowi4_oif = oif;
171 	fl4->daddr = daddr;
172 	fl4->saddr = saddr;
173 	fl4->flowi4_tos = tos;
174 	fl4->flowi4_proto = IPPROTO_GRE;
175 	fl4->fl4_gre_key = gre_key;
176 	return ip_route_output_key(net, fl4);
177 }
178 int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr,
179 			  u8 tos, struct net_device *dev,
180 			  struct in_device *in_dev, u32 *itag);
181 int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src,
182 			 u8 tos, struct net_device *devin);
183 int ip_route_input_rcu(struct sk_buff *skb, __be32 dst, __be32 src,
184 		       u8 tos, struct net_device *devin,
185 		       struct fib_result *res);
186 
187 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
188 				 u8 tos, struct net_device *devin)
189 {
190 	int err;
191 
192 	rcu_read_lock();
193 	err = ip_route_input_noref(skb, dst, src, tos, devin);
194 	if (!err) {
195 		skb_dst_force(skb);
196 		if (!skb_dst(skb))
197 			err = -EINVAL;
198 	}
199 	rcu_read_unlock();
200 
201 	return err;
202 }
203 
204 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif,
205 		      u8 protocol);
206 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
207 void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol);
208 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
209 void ip_rt_send_redirect(struct sk_buff *skb);
210 
211 unsigned int inet_addr_type(struct net *net, __be32 addr);
212 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id);
213 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
214 				__be32 addr);
215 unsigned int inet_addr_type_dev_table(struct net *net,
216 				      const struct net_device *dev,
217 				      __be32 addr);
218 void ip_rt_multicast_event(struct in_device *);
219 int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt);
220 void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
221 struct rtable *rt_dst_alloc(struct net_device *dev,
222 			     unsigned int flags, u16 type,
223 			     bool nopolicy, bool noxfrm, bool will_cache);
224 struct rtable *rt_dst_clone(struct net_device *dev, struct rtable *rt);
225 
226 struct in_ifaddr;
227 void fib_add_ifaddr(struct in_ifaddr *);
228 void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
229 void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric);
230 
231 void rt_add_uncached_list(struct rtable *rt);
232 void rt_del_uncached_list(struct rtable *rt);
233 
234 int fib_dump_info_fnhe(struct sk_buff *skb, struct netlink_callback *cb,
235 		       u32 table_id, struct fib_info *fi,
236 		       int *fa_index, int fa_start);
237 
238 static inline void ip_rt_put(struct rtable *rt)
239 {
240 	/* dst_release() accepts a NULL parameter.
241 	 * We rely on dst being first structure in struct rtable
242 	 */
243 	BUILD_BUG_ON(offsetof(struct rtable, dst) != 0);
244 	dst_release(&rt->dst);
245 }
246 
247 #define IPTOS_RT_MASK	(IPTOS_TOS_MASK & ~3)
248 
249 extern const __u8 ip_tos2prio[16];
250 
251 static inline char rt_tos2priority(u8 tos)
252 {
253 	return ip_tos2prio[IPTOS_TOS(tos)>>1];
254 }
255 
256 /* ip_route_connect() and ip_route_newports() work in tandem whilst
257  * binding a socket for a new outgoing connection.
258  *
259  * In order to use IPSEC properly, we must, in the end, have a
260  * route that was looked up using all available keys including source
261  * and destination ports.
262  *
263  * However, if a source port needs to be allocated (the user specified
264  * a wildcard source port) we need to obtain addressing information
265  * in order to perform that allocation.
266  *
267  * So ip_route_connect() looks up a route using wildcarded source and
268  * destination ports in the key, simply so that we can get a pair of
269  * addresses to use for port allocation.
270  *
271  * Later, once the ports are allocated, ip_route_newports() will make
272  * another route lookup if needed to make sure we catch any IPSEC
273  * rules keyed on the port information.
274  *
275  * The callers allocate the flow key on their stack, and must pass in
276  * the same flowi4 object to both the ip_route_connect() and the
277  * ip_route_newports() calls.
278  */
279 
280 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src,
281 					 u32 tos, int oif, u8 protocol,
282 					 __be16 sport, __be16 dport,
283 					 struct sock *sk)
284 {
285 	__u8 flow_flags = 0;
286 
287 	if (inet_sk(sk)->transparent)
288 		flow_flags |= FLOWI_FLAG_ANYSRC;
289 
290 	flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE,
291 			   protocol, flow_flags, dst, src, dport, sport,
292 			   sk->sk_uid);
293 }
294 
295 static inline struct rtable *ip_route_connect(struct flowi4 *fl4,
296 					      __be32 dst, __be32 src, u32 tos,
297 					      int oif, u8 protocol,
298 					      __be16 sport, __be16 dport,
299 					      struct sock *sk)
300 {
301 	struct net *net = sock_net(sk);
302 	struct rtable *rt;
303 
304 	ip_route_connect_init(fl4, dst, src, tos, oif, protocol,
305 			      sport, dport, sk);
306 
307 	if (!dst || !src) {
308 		rt = __ip_route_output_key(net, fl4);
309 		if (IS_ERR(rt))
310 			return rt;
311 		ip_rt_put(rt);
312 		flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr);
313 	}
314 	security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
315 	return ip_route_output_flow(net, fl4, sk);
316 }
317 
318 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
319 					       __be16 orig_sport, __be16 orig_dport,
320 					       __be16 sport, __be16 dport,
321 					       struct sock *sk)
322 {
323 	if (sport != orig_sport || dport != orig_dport) {
324 		fl4->fl4_dport = dport;
325 		fl4->fl4_sport = sport;
326 		ip_rt_put(rt);
327 		flowi4_update_output(fl4, sk->sk_bound_dev_if,
328 				     RT_CONN_FLAGS(sk), fl4->daddr,
329 				     fl4->saddr);
330 		security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
331 		return ip_route_output_flow(sock_net(sk), fl4, sk);
332 	}
333 	return rt;
334 }
335 
336 static inline int inet_iif(const struct sk_buff *skb)
337 {
338 	struct rtable *rt = skb_rtable(skb);
339 
340 	if (rt && rt->rt_iif)
341 		return rt->rt_iif;
342 
343 	return skb->skb_iif;
344 }
345 
346 static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
347 {
348 	int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
349 	struct net *net = dev_net(dst->dev);
350 
351 	if (hoplimit == 0)
352 		hoplimit = net->ipv4.sysctl_ip_default_ttl;
353 	return hoplimit;
354 }
355 
356 static inline struct neighbour *ip_neigh_gw4(struct net_device *dev,
357 					     __be32 daddr)
358 {
359 	struct neighbour *neigh;
360 
361 	neigh = __ipv4_neigh_lookup_noref(dev, daddr);
362 	if (unlikely(!neigh))
363 		neigh = __neigh_create(&arp_tbl, &daddr, dev, false);
364 
365 	return neigh;
366 }
367 
368 static inline struct neighbour *ip_neigh_for_gw(struct rtable *rt,
369 						struct sk_buff *skb,
370 						bool *is_v6gw)
371 {
372 	struct net_device *dev = rt->dst.dev;
373 	struct neighbour *neigh;
374 
375 	if (likely(rt->rt_gw_family == AF_INET)) {
376 		neigh = ip_neigh_gw4(dev, rt->rt_gw4);
377 	} else if (rt->rt_gw_family == AF_INET6) {
378 		neigh = ip_neigh_gw6(dev, &rt->rt_gw6);
379 		*is_v6gw = true;
380 	} else {
381 		neigh = ip_neigh_gw4(dev, ip_hdr(skb)->daddr);
382 	}
383 	return neigh;
384 }
385 
386 #endif	/* _ROUTE_H */
387