route.h

  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 <net/sock.h>
 32#include <linux/in_route.h>
 33#include <linux/rtnetlink.h>
 34#include <linux/rcupdate.h>
 35#include <linux/route.h>
 36#include <linux/ip.h>
 37#include <linux/cache.h>
 38#include <linux/security.h>
 39
 40static inline __u8 ip_sock_rt_scope(const struct sock *sk)
 41{
 42	if (sock_flag(sk, SOCK_LOCALROUTE))
 43		return RT_SCOPE_LINK;
 44
 45	return RT_SCOPE_UNIVERSE;
 46}
 47
 48static inline __u8 ip_sock_rt_tos(const struct sock *sk)
 49{
 50	return READ_ONCE(inet_sk(sk)->tos) & INET_DSCP_MASK;
 51}
 52
 53struct ip_tunnel_info;
 54struct fib_nh;
 55struct fib_info;
 56struct uncached_list;
 57struct rtable {
 58	struct dst_entry	dst;
 59
 60	int			rt_genid;
 61	unsigned int		rt_flags;
 62	__u16			rt_type;
 63	__u8			rt_is_input;
 64	__u8			rt_uses_gateway;
 65
 66	int			rt_iif;
 67
 68	u8			rt_gw_family;
 69	/* Info on neighbour */
 70	union {
 71		__be32		rt_gw4;
 72		struct in6_addr	rt_gw6;
 73	};
 74
 75	/* Miscellaneous cached information */
 76	u32			rt_mtu_locked:1,
 77				rt_pmtu:31;
 78};
 79
 80#define dst_rtable(_ptr) container_of_const(_ptr, struct rtable, dst)
 81
 82/**
 83 * skb_rtable - Returns the skb &rtable
 84 * @skb: buffer
 85 */
 86static inline struct rtable *skb_rtable(const struct sk_buff *skb)
 87{
 88	return dst_rtable(skb_dst(skb));
 89}
 90
 91static inline bool rt_is_input_route(const struct rtable *rt)
 92{
 93	return rt->rt_is_input != 0;
 94}
 95
 96static inline bool rt_is_output_route(const struct rtable *rt)
 97{
 98	return rt->rt_is_input == 0;
 99}
100
101static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr)
102{
103	if (rt->rt_gw_family == AF_INET)
104		return rt->rt_gw4;
105	return daddr;
106}
107
108struct ip_rt_acct {
109	__u32 	o_bytes;
110	__u32 	o_packets;
111	__u32 	i_bytes;
112	__u32 	i_packets;
113};
114
115struct rt_cache_stat {
116        unsigned int in_slow_tot;
117        unsigned int in_slow_mc;
118        unsigned int in_no_route;
119        unsigned int in_brd;
120        unsigned int in_martian_dst;
121        unsigned int in_martian_src;
122        unsigned int out_slow_tot;
123        unsigned int out_slow_mc;
124};
125
126extern struct ip_rt_acct __percpu *ip_rt_acct;
127
128struct in_device;
129
130int ip_rt_init(void);
131void rt_cache_flush(struct net *net);
132void rt_flush_dev(struct net_device *dev);
133
134static inline void inet_sk_init_flowi4(const struct inet_sock *inet,
135				       struct flowi4 *fl4)
136{
137	const struct ip_options_rcu *ip4_opt;
138	const struct sock *sk;
139	__be32 daddr;
140
141	rcu_read_lock();
142	ip4_opt = rcu_dereference(inet->inet_opt);
143
144	/* Source routing option overrides the socket destination address */
145	if (ip4_opt && ip4_opt->opt.srr)
146		daddr = ip4_opt->opt.faddr;
147	else
148		daddr = inet->inet_daddr;
149	rcu_read_unlock();
150
151	sk = &inet->sk;
152	flowi4_init_output(fl4, sk->sk_bound_dev_if, READ_ONCE(sk->sk_mark),
153			   ip_sock_rt_tos(sk), ip_sock_rt_scope(sk),
154			   sk->sk_protocol, inet_sk_flowi_flags(sk), daddr,
155			   inet->inet_saddr, inet->inet_dport,
156			   inet->inet_sport, sk_uid(sk));
157	security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
158}
159
160struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp,
161					const struct sk_buff *skb);
162struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp,
163					    struct fib_result *res,
164					    const struct sk_buff *skb);
165
166static inline struct rtable *__ip_route_output_key(struct net *net,
167						   struct flowi4 *flp)
168{
169	return ip_route_output_key_hash(net, flp, NULL);
170}
171
172struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
173				    const struct sock *sk);
174struct dst_entry *ipv4_blackhole_route(struct net *net,
175				       struct dst_entry *dst_orig);
176
177static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
178{
179	return ip_route_output_flow(net, flp, NULL);
180}
181
182/* Simplistic IPv4 route lookup function.
183 * This is only suitable for some particular use cases: since the flowi4
184 * structure is only partially set, it may bypass some fib-rules.
185 */
186static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
187					     __be32 saddr, dscp_t dscp,
188					     int oif, __u8 scope)
189{
190	struct flowi4 fl4 = {
191		.flowi4_oif = oif,
192		.flowi4_dscp = dscp,
193		.flowi4_scope = scope,
194		.daddr = daddr,
195		.saddr = saddr,
196	};
197
198	return ip_route_output_key(net, &fl4);
199}
200
201static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
202						   const struct sock *sk,
203						   __be32 daddr, __be32 saddr,
204						   __be16 dport, __be16 sport,
205						   __u8 proto, __u8 tos, int oif)
206{
207	flowi4_init_output(fl4, oif, sk ? READ_ONCE(sk->sk_mark) : 0, tos,
208			   sk ? ip_sock_rt_scope(sk) : RT_SCOPE_UNIVERSE,
209			   proto, sk ? inet_sk_flowi_flags(sk) : 0,
210			   daddr, saddr, dport, sport, sock_net_uid(net, sk));
211	if (sk)
212		security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
213	return ip_route_output_flow(net, fl4, sk);
214}
215
216enum skb_drop_reason
217ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr,
218		      dscp_t dscp, struct net_device *dev,
219		      struct in_device *in_dev, u32 *itag);
220enum skb_drop_reason
221ip_route_input_noref(struct sk_buff *skb, __be32 daddr, __be32 saddr,
222		     dscp_t dscp, struct net_device *dev);
223enum skb_drop_reason
224ip_route_use_hint(struct sk_buff *skb, __be32 daddr, __be32 saddr,
225		  dscp_t dscp, struct net_device *dev,
226		  const struct sk_buff *hint);
227
228static inline enum skb_drop_reason
229ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src, dscp_t dscp,
230	       struct net_device *devin)
231{
232	enum skb_drop_reason reason;
233
234	rcu_read_lock();
235	reason = ip_route_input_noref(skb, dst, src, dscp, devin);
236	if (!reason) {
237		skb_dst_force(skb);
238		if (!skb_dst(skb))
239			reason = SKB_DROP_REASON_NOT_SPECIFIED;
240	}
241	rcu_read_unlock();
242
243	return reason;
244}
245
246void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif,
247		      u8 protocol);
248void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
249void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol);
250void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
251void ip_rt_send_redirect(struct sk_buff *skb);
252
253unsigned int inet_addr_type(struct net *net, __be32 addr);
254unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id);
255unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
256				__be32 addr);
257unsigned int inet_addr_type_dev_table(struct net *net,
258				      const struct net_device *dev,
259				      __be32 addr);
260void ip_rt_multicast_event(struct in_device *);
261int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt);
262void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
263struct rtable *rt_dst_alloc(struct net_device *dev,
264			    unsigned int flags, u16 type, bool noxfrm);
265struct rtable *rt_dst_clone(struct net_device *dev, struct rtable *rt);
266
267struct in_ifaddr;
268void fib_add_ifaddr(struct in_ifaddr *);
269void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
270void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric);
271
272void rt_add_uncached_list(struct rtable *rt);
273void rt_del_uncached_list(struct rtable *rt);
274
275int fib_dump_info_fnhe(struct sk_buff *skb, struct netlink_callback *cb,
276		       u32 table_id, struct fib_info *fi,
277		       int *fa_index, int fa_start, unsigned int flags);
278
279static inline void ip_rt_put(struct rtable *rt)
280{
281	/* dst_release() accepts a NULL parameter.
282	 * We rely on dst being first structure in struct rtable
283	 */
284	BUILD_BUG_ON(offsetof(struct rtable, dst) != 0);
285	dst_release(&rt->dst);
286}
287
288extern const __u8 ip_tos2prio[16];
289
290static inline char rt_tos2priority(u8 tos)
291{
292	return ip_tos2prio[IPTOS_TOS(tos)>>1];
293}
294
295/* ip_route_connect() and ip_route_newports() work in tandem whilst
296 * binding a socket for a new outgoing connection.
297 *
298 * In order to use IPSEC properly, we must, in the end, have a
299 * route that was looked up using all available keys including source
300 * and destination ports.
301 *
302 * However, if a source port needs to be allocated (the user specified
303 * a wildcard source port) we need to obtain addressing information
304 * in order to perform that allocation.
305 *
306 * So ip_route_connect() looks up a route using wildcarded source and
307 * destination ports in the key, simply so that we can get a pair of
308 * addresses to use for port allocation.
309 *
310 * Later, once the ports are allocated, ip_route_newports() will make
311 * another route lookup if needed to make sure we catch any IPSEC
312 * rules keyed on the port information.
313 *
314 * The callers allocate the flow key on their stack, and must pass in
315 * the same flowi4 object to both the ip_route_connect() and the
316 * ip_route_newports() calls.
317 */
318
319static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst,
320					 __be32 src, int oif, u8 protocol,
321					 __be16 sport, __be16 dport,
322					 const struct sock *sk)
323{
324	__u8 flow_flags = 0;
325
326	if (inet_test_bit(TRANSPARENT, sk))
327		flow_flags |= FLOWI_FLAG_ANYSRC;
328
329	if (IS_ENABLED(CONFIG_IP_ROUTE_MULTIPATH) && !sport)
330		flow_flags |= FLOWI_FLAG_ANY_SPORT;
331
332	flowi4_init_output(fl4, oif, READ_ONCE(sk->sk_mark), ip_sock_rt_tos(sk),
333			   ip_sock_rt_scope(sk), protocol, flow_flags, dst,
334			   src, dport, sport, sk_uid(sk));
335}
336
337static inline struct rtable *ip_route_connect(struct flowi4 *fl4, __be32 dst,
338					      __be32 src, int oif, u8 protocol,
339					      __be16 sport, __be16 dport,
340					      const struct sock *sk)
341{
342	struct net *net = sock_net(sk);
343	struct rtable *rt;
344
345	ip_route_connect_init(fl4, dst, src, oif, protocol, sport, dport, sk);
346
347	if (!dst || !src) {
348		rt = __ip_route_output_key(net, fl4);
349		if (IS_ERR(rt))
350			return rt;
351		ip_rt_put(rt);
352		flowi4_update_output(fl4, oif, fl4->daddr, fl4->saddr);
353	}
354	security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
355	return ip_route_output_flow(net, fl4, sk);
356}
357
358static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
359					       __be16 orig_sport, __be16 orig_dport,
360					       __be16 sport, __be16 dport,
361					       const struct sock *sk)
362{
363	if (sport != orig_sport || dport != orig_dport) {
364		fl4->fl4_dport = dport;
365		fl4->fl4_sport = sport;
366		ip_rt_put(rt);
367		flowi4_update_output(fl4, sk->sk_bound_dev_if, fl4->daddr,
368				     fl4->saddr);
369		security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
370		return ip_route_output_flow(sock_net(sk), fl4, sk);
371	}
372	return rt;
373}
374
375static inline int inet_iif(const struct sk_buff *skb)
376{
377	struct rtable *rt = skb_rtable(skb);
378
379	if (rt && rt->rt_iif)
380		return rt->rt_iif;
381
382	return skb->skb_iif;
383}
384
385static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
386{
387	int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
388
389	if (hoplimit == 0) {
390		const struct net *net;
391
392		rcu_read_lock();
393		net = dst_dev_net_rcu(dst);
394		hoplimit = READ_ONCE(net->ipv4.sysctl_ip_default_ttl);
395		rcu_read_unlock();
396	}
397	return hoplimit;
398}
399
400static inline struct neighbour *ip_neigh_gw4(struct net_device *dev,
401					     __be32 daddr)
402{
403	struct neighbour *neigh;
404
405	neigh = __ipv4_neigh_lookup_noref(dev, (__force u32)daddr);
406	if (unlikely(!neigh))
407		neigh = __neigh_create(&arp_tbl, &daddr, dev, false);
408
409	return neigh;
410}
411
412static inline struct neighbour *ip_neigh_for_gw(struct rtable *rt,
413						struct sk_buff *skb,
414						bool *is_v6gw)
415{
416	struct net_device *dev = rt->dst.dev;
417	struct neighbour *neigh;
418
419	if (likely(rt->rt_gw_family == AF_INET)) {
420		neigh = ip_neigh_gw4(dev, rt->rt_gw4);
421	} else if (rt->rt_gw_family == AF_INET6) {
422		neigh = ip_neigh_gw6(dev, &rt->rt_gw6);
423		*is_v6gw = true;
424	} else {
425		neigh = ip_neigh_gw4(dev, ip_hdr(skb)->daddr);
426	}
427	return neigh;
428}
429
430#endif	/* _ROUTE_H */