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