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