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 | #define RTO_ONLINK 0x01 |
39 | |
40 | #define RT_CONN_FLAGS(sk) (RT_TOS(READ_ONCE(inet_sk(sk)->tos)) | sock_flag(sk, SOCK_LOCALROUTE)) |
41 | #define RT_CONN_FLAGS_TOS(sk,tos) (RT_TOS(tos) | sock_flag(sk, SOCK_LOCALROUTE)) |
42 | |
43 | static inline __u8 ip_sock_rt_scope(const struct sock *sk) |
44 | { |
45 | if (sock_flag(sk, flag: SOCK_LOCALROUTE)) |
46 | return RT_SCOPE_LINK; |
47 | |
48 | return RT_SCOPE_UNIVERSE; |
49 | } |
50 | |
51 | static inline __u8 ip_sock_rt_tos(const struct sock *sk) |
52 | { |
53 | return RT_TOS(READ_ONCE(inet_sk(sk)->tos)); |
54 | } |
55 | |
56 | struct ip_tunnel_info; |
57 | struct fib_nh; |
58 | struct fib_info; |
59 | struct uncached_list; |
60 | struct rtable { |
61 | struct dst_entry dst; |
62 | |
63 | int rt_genid; |
64 | unsigned int rt_flags; |
65 | __u16 rt_type; |
66 | __u8 rt_is_input; |
67 | __u8 rt_uses_gateway; |
68 | |
69 | int rt_iif; |
70 | |
71 | u8 rt_gw_family; |
72 | /* Info on neighbour */ |
73 | union { |
74 | __be32 rt_gw4; |
75 | struct in6_addr rt_gw6; |
76 | }; |
77 | |
78 | /* Miscellaneous cached information */ |
79 | u32 rt_mtu_locked:1, |
80 | rt_pmtu:31; |
81 | }; |
82 | |
83 | static inline bool rt_is_input_route(const struct rtable *rt) |
84 | { |
85 | return rt->rt_is_input != 0; |
86 | } |
87 | |
88 | static inline bool rt_is_output_route(const struct rtable *rt) |
89 | { |
90 | return rt->rt_is_input == 0; |
91 | } |
92 | |
93 | static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr) |
94 | { |
95 | if (rt->rt_gw_family == AF_INET) |
96 | return rt->rt_gw4; |
97 | return daddr; |
98 | } |
99 | |
100 | struct ip_rt_acct { |
101 | __u32 o_bytes; |
102 | __u32 o_packets; |
103 | __u32 i_bytes; |
104 | __u32 i_packets; |
105 | }; |
106 | |
107 | struct rt_cache_stat { |
108 | unsigned int in_slow_tot; |
109 | unsigned int in_slow_mc; |
110 | unsigned int in_no_route; |
111 | unsigned int in_brd; |
112 | unsigned int in_martian_dst; |
113 | unsigned int in_martian_src; |
114 | unsigned int out_slow_tot; |
115 | unsigned int out_slow_mc; |
116 | }; |
117 | |
118 | extern struct ip_rt_acct __percpu *ip_rt_acct; |
119 | |
120 | struct in_device; |
121 | |
122 | int ip_rt_init(void); |
123 | void rt_cache_flush(struct net *net); |
124 | void rt_flush_dev(struct net_device *dev); |
125 | struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp, |
126 | const struct sk_buff *skb); |
127 | struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp, |
128 | struct fib_result *res, |
129 | const struct sk_buff *skb); |
130 | |
131 | static inline struct rtable *__ip_route_output_key(struct net *net, |
132 | struct flowi4 *flp) |
133 | { |
134 | return ip_route_output_key_hash(net, flp, NULL); |
135 | } |
136 | |
137 | struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp, |
138 | const struct sock *sk); |
139 | struct dst_entry *ipv4_blackhole_route(struct net *net, |
140 | struct dst_entry *dst_orig); |
141 | |
142 | static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp) |
143 | { |
144 | return ip_route_output_flow(net, flp, NULL); |
145 | } |
146 | |
147 | static inline struct rtable *ip_route_output(struct net *net, __be32 daddr, |
148 | __be32 saddr, u8 tos, int oif) |
149 | { |
150 | struct flowi4 fl4 = { |
151 | .flowi4_oif = oif, |
152 | .flowi4_tos = tos, |
153 | .daddr = daddr, |
154 | .saddr = saddr, |
155 | }; |
156 | return ip_route_output_key(net, flp: &fl4); |
157 | } |
158 | |
159 | static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4, |
160 | const struct sock *sk, |
161 | __be32 daddr, __be32 saddr, |
162 | __be16 dport, __be16 sport, |
163 | __u8 proto, __u8 tos, int oif) |
164 | { |
165 | flowi4_init_output(fl4, oif, mark: sk ? READ_ONCE(sk->sk_mark) : 0, tos, |
166 | scope: RT_SCOPE_UNIVERSE, proto, |
167 | flags: sk ? inet_sk_flowi_flags(sk) : 0, |
168 | daddr, saddr, dport, sport, uid: sock_net_uid(net, sk)); |
169 | if (sk) |
170 | security_sk_classify_flow(sk, flic: flowi4_to_flowi_common(fl4)); |
171 | return ip_route_output_flow(net, flp: fl4, sk); |
172 | } |
173 | |
174 | static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4, |
175 | __be32 daddr, __be32 saddr, |
176 | __be32 gre_key, __u8 tos, int oif) |
177 | { |
178 | memset(fl4, 0, sizeof(*fl4)); |
179 | fl4->flowi4_oif = oif; |
180 | fl4->daddr = daddr; |
181 | fl4->saddr = saddr; |
182 | fl4->flowi4_tos = tos; |
183 | fl4->flowi4_proto = IPPROTO_GRE; |
184 | fl4->fl4_gre_key = gre_key; |
185 | return ip_route_output_key(net, flp: fl4); |
186 | } |
187 | int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
188 | u8 tos, struct net_device *dev, |
189 | struct in_device *in_dev, u32 *itag); |
190 | int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src, |
191 | u8 tos, struct net_device *devin); |
192 | int ip_route_use_hint(struct sk_buff *skb, __be32 dst, __be32 src, |
193 | u8 tos, struct net_device *devin, |
194 | const struct sk_buff *hint); |
195 | |
196 | static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src, |
197 | u8 tos, struct net_device *devin) |
198 | { |
199 | int err; |
200 | |
201 | rcu_read_lock(); |
202 | err = ip_route_input_noref(skb, dst, src, tos, devin); |
203 | if (!err) { |
204 | skb_dst_force(skb); |
205 | if (!skb_dst(skb)) |
206 | err = -EINVAL; |
207 | } |
208 | rcu_read_unlock(); |
209 | |
210 | return err; |
211 | } |
212 | |
213 | void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif, |
214 | u8 protocol); |
215 | void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu); |
216 | void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol); |
217 | void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk); |
218 | void ip_rt_send_redirect(struct sk_buff *skb); |
219 | |
220 | unsigned int inet_addr_type(struct net *net, __be32 addr); |
221 | unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id); |
222 | unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev, |
223 | __be32 addr); |
224 | unsigned int inet_addr_type_dev_table(struct net *net, |
225 | const struct net_device *dev, |
226 | __be32 addr); |
227 | void ip_rt_multicast_event(struct in_device *); |
228 | int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt); |
229 | void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt); |
230 | struct rtable *rt_dst_alloc(struct net_device *dev, |
231 | unsigned int flags, u16 type, bool noxfrm); |
232 | struct rtable *rt_dst_clone(struct net_device *dev, struct rtable *rt); |
233 | |
234 | struct in_ifaddr; |
235 | void fib_add_ifaddr(struct in_ifaddr *); |
236 | void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *); |
237 | void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric); |
238 | |
239 | void rt_add_uncached_list(struct rtable *rt); |
240 | void rt_del_uncached_list(struct rtable *rt); |
241 | |
242 | int fib_dump_info_fnhe(struct sk_buff *skb, struct netlink_callback *cb, |
243 | u32 table_id, struct fib_info *fi, |
244 | int *fa_index, int fa_start, unsigned int flags); |
245 | |
246 | static inline void ip_rt_put(struct rtable *rt) |
247 | { |
248 | /* dst_release() accepts a NULL parameter. |
249 | * We rely on dst being first structure in struct rtable |
250 | */ |
251 | BUILD_BUG_ON(offsetof(struct rtable, dst) != 0); |
252 | dst_release(dst: &rt->dst); |
253 | } |
254 | |
255 | #define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3) |
256 | |
257 | extern const __u8 ip_tos2prio[16]; |
258 | |
259 | static inline char rt_tos2priority(u8 tos) |
260 | { |
261 | return ip_tos2prio[IPTOS_TOS(tos)>>1]; |
262 | } |
263 | |
264 | /* ip_route_connect() and ip_route_newports() work in tandem whilst |
265 | * binding a socket for a new outgoing connection. |
266 | * |
267 | * In order to use IPSEC properly, we must, in the end, have a |
268 | * route that was looked up using all available keys including source |
269 | * and destination ports. |
270 | * |
271 | * However, if a source port needs to be allocated (the user specified |
272 | * a wildcard source port) we need to obtain addressing information |
273 | * in order to perform that allocation. |
274 | * |
275 | * So ip_route_connect() looks up a route using wildcarded source and |
276 | * destination ports in the key, simply so that we can get a pair of |
277 | * addresses to use for port allocation. |
278 | * |
279 | * Later, once the ports are allocated, ip_route_newports() will make |
280 | * another route lookup if needed to make sure we catch any IPSEC |
281 | * rules keyed on the port information. |
282 | * |
283 | * The callers allocate the flow key on their stack, and must pass in |
284 | * the same flowi4 object to both the ip_route_connect() and the |
285 | * ip_route_newports() calls. |
286 | */ |
287 | |
288 | static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, |
289 | __be32 src, int oif, u8 protocol, |
290 | __be16 sport, __be16 dport, |
291 | const struct sock *sk) |
292 | { |
293 | __u8 flow_flags = 0; |
294 | |
295 | if (inet_test_bit(TRANSPARENT, sk)) |
296 | flow_flags |= FLOWI_FLAG_ANYSRC; |
297 | |
298 | flowi4_init_output(fl4, oif, READ_ONCE(sk->sk_mark), tos: ip_sock_rt_tos(sk), |
299 | scope: ip_sock_rt_scope(sk), proto: protocol, flags: flow_flags, daddr: dst, |
300 | saddr: src, dport, sport, uid: sk->sk_uid); |
301 | } |
302 | |
303 | static inline struct rtable *ip_route_connect(struct flowi4 *fl4, __be32 dst, |
304 | __be32 src, int oif, u8 protocol, |
305 | __be16 sport, __be16 dport, |
306 | const struct sock *sk) |
307 | { |
308 | struct net *net = sock_net(sk); |
309 | struct rtable *rt; |
310 | |
311 | ip_route_connect_init(fl4, dst, src, oif, protocol, sport, dport, sk); |
312 | |
313 | if (!dst || !src) { |
314 | rt = __ip_route_output_key(net, flp: fl4); |
315 | if (IS_ERR(ptr: rt)) |
316 | return rt; |
317 | ip_rt_put(rt); |
318 | flowi4_update_output(fl4, oif, daddr: fl4->daddr, saddr: fl4->saddr); |
319 | } |
320 | security_sk_classify_flow(sk, flic: flowi4_to_flowi_common(fl4)); |
321 | return ip_route_output_flow(net, flp: fl4, sk); |
322 | } |
323 | |
324 | static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt, |
325 | __be16 orig_sport, __be16 orig_dport, |
326 | __be16 sport, __be16 dport, |
327 | const struct sock *sk) |
328 | { |
329 | if (sport != orig_sport || dport != orig_dport) { |
330 | fl4->fl4_dport = dport; |
331 | fl4->fl4_sport = sport; |
332 | ip_rt_put(rt); |
333 | flowi4_update_output(fl4, oif: sk->sk_bound_dev_if, daddr: fl4->daddr, |
334 | saddr: fl4->saddr); |
335 | security_sk_classify_flow(sk, flic: flowi4_to_flowi_common(fl4)); |
336 | return ip_route_output_flow(sock_net(sk), flp: fl4, sk); |
337 | } |
338 | return rt; |
339 | } |
340 | |
341 | static inline int inet_iif(const struct sk_buff *skb) |
342 | { |
343 | struct rtable *rt = skb_rtable(skb); |
344 | |
345 | if (rt && rt->rt_iif) |
346 | return rt->rt_iif; |
347 | |
348 | return skb->skb_iif; |
349 | } |
350 | |
351 | static inline int ip4_dst_hoplimit(const struct dst_entry *dst) |
352 | { |
353 | int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT); |
354 | struct net *net = dev_net(dev: dst->dev); |
355 | |
356 | if (hoplimit == 0) |
357 | hoplimit = READ_ONCE(net->ipv4.sysctl_ip_default_ttl); |
358 | return hoplimit; |
359 | } |
360 | |
361 | static inline struct neighbour *ip_neigh_gw4(struct net_device *dev, |
362 | __be32 daddr) |
363 | { |
364 | struct neighbour *neigh; |
365 | |
366 | neigh = __ipv4_neigh_lookup_noref(dev, key: (__force u32)daddr); |
367 | if (unlikely(!neigh)) |
368 | neigh = __neigh_create(tbl: &arp_tbl, pkey: &daddr, dev, want_ref: false); |
369 | |
370 | return neigh; |
371 | } |
372 | |
373 | static inline struct neighbour *ip_neigh_for_gw(struct rtable *rt, |
374 | struct sk_buff *skb, |
375 | bool *is_v6gw) |
376 | { |
377 | struct net_device *dev = rt->dst.dev; |
378 | struct neighbour *neigh; |
379 | |
380 | if (likely(rt->rt_gw_family == AF_INET)) { |
381 | neigh = ip_neigh_gw4(dev, daddr: rt->rt_gw4); |
382 | } else if (rt->rt_gw_family == AF_INET6) { |
383 | neigh = ip_neigh_gw6(dev, addr: &rt->rt_gw6); |
384 | *is_v6gw = true; |
385 | } else { |
386 | neigh = ip_neigh_gw4(dev, daddr: ip_hdr(skb)->daddr); |
387 | } |
388 | return neigh; |
389 | } |
390 | |
391 | #endif /* _ROUTE_H */ |
392 | |