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 | static inline __u8 ip_sock_rt_scope(const struct sock *sk) |
47 | { |
48 | if (sock_flag(sk, SOCK_LOCALROUTE)) |
49 | return RT_SCOPE_LINK; |
50 | |
51 | return RT_SCOPE_UNIVERSE; |
52 | } |
53 | |
54 | static inline __u8 ip_sock_rt_tos(const struct sock *sk) |
55 | { |
56 | return RT_TOS(inet_sk(sk)->tos); |
57 | } |
58 | |
59 | struct ip_tunnel_info; |
60 | struct fib_nh; |
61 | struct fib_info; |
62 | struct uncached_list; |
63 | struct rtable { |
64 | struct dst_entry dst; |
65 | |
66 | int rt_genid; |
67 | unsigned int rt_flags; |
68 | __u16 rt_type; |
69 | __u8 rt_is_input; |
70 | __u8 rt_uses_gateway; |
71 | |
72 | int rt_iif; |
73 | |
74 | u8 rt_gw_family; |
75 | /* Info on neighbour */ |
76 | union { |
77 | __be32 rt_gw4; |
78 | struct in6_addr rt_gw6; |
79 | }; |
80 | |
81 | /* Miscellaneous cached information */ |
82 | u32 rt_mtu_locked:1, |
83 | rt_pmtu:31; |
84 | |
85 | struct list_head rt_uncached; |
86 | struct uncached_list *rt_uncached_list; |
87 | }; |
88 | |
89 | static inline bool rt_is_input_route(const struct rtable *rt) |
90 | { |
91 | return rt->rt_is_input != 0; |
92 | } |
93 | |
94 | static inline bool rt_is_output_route(const struct rtable *rt) |
95 | { |
96 | return rt->rt_is_input == 0; |
97 | } |
98 | |
99 | static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr) |
100 | { |
101 | if (rt->rt_gw_family == AF_INET) |
102 | return rt->rt_gw4; |
103 | return daddr; |
104 | } |
105 | |
106 | struct ip_rt_acct { |
107 | __u32 o_bytes; |
108 | __u32 o_packets; |
109 | __u32 i_bytes; |
110 | __u32 i_packets; |
111 | }; |
112 | |
113 | struct rt_cache_stat { |
114 | unsigned int in_slow_tot; |
115 | unsigned int in_slow_mc; |
116 | unsigned int in_no_route; |
117 | unsigned int in_brd; |
118 | unsigned int in_martian_dst; |
119 | unsigned int in_martian_src; |
120 | unsigned int out_slow_tot; |
121 | unsigned int out_slow_mc; |
122 | }; |
123 | |
124 | extern struct ip_rt_acct __percpu *ip_rt_acct; |
125 | |
126 | struct in_device; |
127 | |
128 | int ip_rt_init(void); |
129 | void rt_cache_flush(struct net *net); |
130 | void rt_flush_dev(struct net_device *dev); |
131 | struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp, |
132 | const struct sk_buff *skb); |
133 | struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp, |
134 | struct fib_result *res, |
135 | const struct sk_buff *skb); |
136 | |
137 | static inline struct rtable *__ip_route_output_key(struct net *net, |
138 | struct flowi4 *flp) |
139 | { |
140 | return ip_route_output_key_hash(net, flp, NULL); |
141 | } |
142 | |
143 | struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp, |
144 | const struct sock *sk); |
145 | struct rtable *ip_route_output_tunnel(struct sk_buff *skb, |
146 | struct net_device *dev, |
147 | struct net *net, __be32 *saddr, |
148 | const struct ip_tunnel_info *info, |
149 | u8 protocol, bool use_cache); |
150 | |
151 | struct dst_entry *ipv4_blackhole_route(struct net *net, |
152 | struct dst_entry *dst_orig); |
153 | |
154 | static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp) |
155 | { |
156 | return ip_route_output_flow(net, flp, NULL); |
157 | } |
158 | |
159 | static inline struct rtable *ip_route_output(struct net *net, __be32 daddr, |
160 | __be32 saddr, u8 tos, int oif) |
161 | { |
162 | struct flowi4 fl4 = { |
163 | .flowi4_oif = oif, |
164 | .flowi4_tos = tos, |
165 | .daddr = daddr, |
166 | .saddr = saddr, |
167 | }; |
168 | return ip_route_output_key(net, &fl4); |
169 | } |
170 | |
171 | static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4, |
172 | struct sock *sk, |
173 | __be32 daddr, __be32 saddr, |
174 | __be16 dport, __be16 sport, |
175 | __u8 proto, __u8 tos, int oif) |
176 | { |
177 | flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos, |
178 | RT_SCOPE_UNIVERSE, proto, |
179 | sk ? inet_sk_flowi_flags(sk) : 0, |
180 | daddr, saddr, dport, sport, sock_net_uid(net, sk)); |
181 | if (sk) |
182 | security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4)); |
183 | return ip_route_output_flow(net, fl4, sk); |
184 | } |
185 | |
186 | static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4, |
187 | __be32 daddr, __be32 saddr, |
188 | __be32 gre_key, __u8 tos, int oif) |
189 | { |
190 | memset(fl4, 0, sizeof(*fl4)); |
191 | fl4->flowi4_oif = oif; |
192 | fl4->daddr = daddr; |
193 | fl4->saddr = saddr; |
194 | fl4->flowi4_tos = tos; |
195 | fl4->flowi4_proto = IPPROTO_GRE; |
196 | fl4->fl4_gre_key = gre_key; |
197 | return ip_route_output_key(net, fl4); |
198 | } |
199 | int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
200 | u8 tos, struct net_device *dev, |
201 | struct in_device *in_dev, u32 *itag); |
202 | int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src, |
203 | u8 tos, struct net_device *devin); |
204 | int ip_route_use_hint(struct sk_buff *skb, __be32 dst, __be32 src, |
205 | u8 tos, struct net_device *devin, |
206 | const struct sk_buff *hint); |
207 | |
208 | static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src, |
209 | u8 tos, struct net_device *devin) |
210 | { |
211 | int err; |
212 | |
213 | rcu_read_lock(); |
214 | err = ip_route_input_noref(skb, dst, src, tos, devin); |
215 | if (!err) { |
216 | skb_dst_force(skb); |
217 | if (!skb_dst(skb)) |
218 | err = -EINVAL; |
219 | } |
220 | rcu_read_unlock(); |
221 | |
222 | return err; |
223 | } |
224 | |
225 | void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif, |
226 | u8 protocol); |
227 | void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu); |
228 | void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol); |
229 | void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk); |
230 | void ip_rt_send_redirect(struct sk_buff *skb); |
231 | |
232 | unsigned int inet_addr_type(struct net *net, __be32 addr); |
233 | unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id); |
234 | unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev, |
235 | __be32 addr); |
236 | unsigned int inet_addr_type_dev_table(struct net *net, |
237 | const struct net_device *dev, |
238 | __be32 addr); |
239 | void ip_rt_multicast_event(struct in_device *); |
240 | int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt); |
241 | void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt); |
242 | struct rtable *rt_dst_alloc(struct net_device *dev, |
243 | unsigned int flags, u16 type, bool noxfrm); |
244 | struct rtable *rt_dst_clone(struct net_device *dev, struct rtable *rt); |
245 | |
246 | struct in_ifaddr; |
247 | void fib_add_ifaddr(struct in_ifaddr *); |
248 | void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *); |
249 | void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric); |
250 | |
251 | void rt_add_uncached_list(struct rtable *rt); |
252 | void rt_del_uncached_list(struct rtable *rt); |
253 | |
254 | int fib_dump_info_fnhe(struct sk_buff *skb, struct netlink_callback *cb, |
255 | u32 table_id, struct fib_info *fi, |
256 | int *fa_index, int fa_start, unsigned int flags); |
257 | |
258 | static inline void ip_rt_put(struct rtable *rt) |
259 | { |
260 | /* dst_release() accepts a NULL parameter. |
261 | * We rely on dst being first structure in struct rtable |
262 | */ |
263 | BUILD_BUG_ON(offsetof(struct rtable, dst) != 0); |
264 | dst_release(&rt->dst); |
265 | } |
266 | |
267 | #define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3) |
268 | |
269 | extern const __u8 ip_tos2prio[16]; |
270 | |
271 | static inline char rt_tos2priority(u8 tos) |
272 | { |
273 | return ip_tos2prio[IPTOS_TOS(tos)>>1]; |
274 | } |
275 | |
276 | /* ip_route_connect() and ip_route_newports() work in tandem whilst |
277 | * binding a socket for a new outgoing connection. |
278 | * |
279 | * In order to use IPSEC properly, we must, in the end, have a |
280 | * route that was looked up using all available keys including source |
281 | * and destination ports. |
282 | * |
283 | * However, if a source port needs to be allocated (the user specified |
284 | * a wildcard source port) we need to obtain addressing information |
285 | * in order to perform that allocation. |
286 | * |
287 | * So ip_route_connect() looks up a route using wildcarded source and |
288 | * destination ports in the key, simply so that we can get a pair of |
289 | * addresses to use for port allocation. |
290 | * |
291 | * Later, once the ports are allocated, ip_route_newports() will make |
292 | * another route lookup if needed to make sure we catch any IPSEC |
293 | * rules keyed on the port information. |
294 | * |
295 | * The callers allocate the flow key on their stack, and must pass in |
296 | * the same flowi4 object to both the ip_route_connect() and the |
297 | * ip_route_newports() calls. |
298 | */ |
299 | |
300 | static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, |
301 | __be32 src, int oif, u8 protocol, |
302 | __be16 sport, __be16 dport, |
303 | const struct sock *sk) |
304 | { |
305 | __u8 flow_flags = 0; |
306 | |
307 | if (inet_sk(sk)->transparent) |
308 | flow_flags |= FLOWI_FLAG_ANYSRC; |
309 | |
310 | flowi4_init_output(fl4, oif, sk->sk_mark, ip_sock_rt_tos(sk), |
311 | ip_sock_rt_scope(sk), protocol, flow_flags, dst, |
312 | src, dport, sport, sk->sk_uid); |
313 | } |
314 | |
315 | static inline struct rtable *ip_route_connect(struct flowi4 *fl4, __be32 dst, |
316 | __be32 src, int oif, u8 protocol, |
317 | __be16 sport, __be16 dport, |
318 | struct sock *sk) |
319 | { |
320 | struct net *net = sock_net(sk); |
321 | struct rtable *rt; |
322 | |
323 | ip_route_connect_init(fl4, dst, src, oif, protocol, sport, dport, sk); |
324 | |
325 | if (!dst || !src) { |
326 | rt = __ip_route_output_key(net, fl4); |
327 | if (IS_ERR(rt)) |
328 | return rt; |
329 | ip_rt_put(rt); |
330 | flowi4_update_output(fl4, oif, fl4->flowi4_tos, fl4->daddr, |
331 | fl4->saddr); |
332 | } |
333 | security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4)); |
334 | return ip_route_output_flow(net, fl4, sk); |
335 | } |
336 | |
337 | static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt, |
338 | __be16 orig_sport, __be16 orig_dport, |
339 | __be16 sport, __be16 dport, |
340 | struct sock *sk) |
341 | { |
342 | if (sport != orig_sport || dport != orig_dport) { |
343 | fl4->fl4_dport = dport; |
344 | fl4->fl4_sport = sport; |
345 | ip_rt_put(rt); |
346 | flowi4_update_output(fl4, sk->sk_bound_dev_if, |
347 | RT_CONN_FLAGS(sk), fl4->daddr, |
348 | fl4->saddr); |
349 | security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4)); |
350 | return ip_route_output_flow(sock_net(sk), fl4, sk); |
351 | } |
352 | return rt; |
353 | } |
354 | |
355 | static inline int inet_iif(const struct sk_buff *skb) |
356 | { |
357 | struct rtable *rt = skb_rtable(skb); |
358 | |
359 | if (rt && rt->rt_iif) |
360 | return rt->rt_iif; |
361 | |
362 | return skb->skb_iif; |
363 | } |
364 | |
365 | static inline int ip4_dst_hoplimit(const struct dst_entry *dst) |
366 | { |
367 | int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT); |
368 | struct net *net = dev_net(dst->dev); |
369 | |
370 | if (hoplimit == 0) |
371 | hoplimit = READ_ONCE(net->ipv4.sysctl_ip_default_ttl); |
372 | return hoplimit; |
373 | } |
374 | |
375 | static inline struct neighbour *ip_neigh_gw4(struct net_device *dev, |
376 | __be32 daddr) |
377 | { |
378 | struct neighbour *neigh; |
379 | |
380 | neigh = __ipv4_neigh_lookup_noref(dev, (__force u32)daddr); |
381 | if (unlikely(!neigh)) |
382 | neigh = __neigh_create(&arp_tbl, &daddr, dev, false); |
383 | |
384 | return neigh; |
385 | } |
386 | |
387 | static inline struct neighbour *ip_neigh_for_gw(struct rtable *rt, |
388 | struct sk_buff *skb, |
389 | bool *is_v6gw) |
390 | { |
391 | struct net_device *dev = rt->dst.dev; |
392 | struct neighbour *neigh; |
393 | |
394 | if (likely(rt->rt_gw_family == AF_INET)) { |
395 | neigh = ip_neigh_gw4(dev, rt->rt_gw4); |
396 | } else if (rt->rt_gw_family == AF_INET6) { |
397 | neigh = ip_neigh_gw6(dev, &rt->rt_gw6); |
398 | *is_v6gw = true; |
399 | } else { |
400 | neigh = ip_neigh_gw4(dev, ip_hdr(skb)->daddr); |
401 | } |
402 | return neigh; |
403 | } |
404 | |
405 | #endif /* _ROUTE_H */ |
406 | |