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 40 static 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 48 static inline __u8 ip_sock_rt_tos(const struct sock *sk) 49 { 50 return READ_ONCE(inet_sk(sk)->tos) & INET_DSCP_MASK; 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 #define dst_rtable(_ptr) container_of_const(_ptr, struct rtable, dst) 81 82 /** 83 * skb_rtable - Returns the skb &rtable 84 * @skb: buffer 85 */ 86 static inline struct rtable *skb_rtable(const struct sk_buff *skb) 87 { 88 return dst_rtable(skb_dst(skb)); 89 } 90 91 static inline bool rt_is_input_route(const struct rtable *rt) 92 { 93 return rt->rt_is_input != 0; 94 } 95 96 static inline bool rt_is_output_route(const struct rtable *rt) 97 { 98 return rt->rt_is_input == 0; 99 } 100 101 static 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 } 107 108 struct ip_rt_acct { 109 __u32 o_bytes; 110 __u32 o_packets; 111 __u32 i_bytes; 112 __u32 i_packets; 113 }; 114 115 struct 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 }; 125 126 extern struct ip_rt_acct __percpu *ip_rt_acct; 127 128 struct in_device; 129 130 int ip_rt_init(void); 131 void rt_cache_flush(struct net *net); 132 #ifdef CONFIG_INET 133 void rt_flush_dev(struct net_device *dev); 134 #else 135 static inline void rt_flush_dev(struct net_device *dev) 136 { 137 } 138 #endif 139 140 static inline void inet_sk_init_flowi4(const struct inet_sock *inet, 141 struct flowi4 *fl4) 142 { 143 const struct ip_options_rcu *ip4_opt; 144 const struct sock *sk; 145 __be32 daddr; 146 147 rcu_read_lock(); 148 ip4_opt = rcu_dereference(inet->inet_opt); 149 150 /* Source routing option overrides the socket destination address */ 151 if (ip4_opt && ip4_opt->opt.srr) 152 daddr = ip4_opt->opt.faddr; 153 else 154 daddr = inet->inet_daddr; 155 rcu_read_unlock(); 156 157 sk = &inet->sk; 158 flowi4_init_output(fl4, sk->sk_bound_dev_if, READ_ONCE(sk->sk_mark), 159 ip_sock_rt_tos(sk), ip_sock_rt_scope(sk), 160 sk->sk_protocol, inet_sk_flowi_flags(sk), daddr, 161 inet->inet_saddr, inet->inet_dport, 162 inet->inet_sport, sk_uid(sk)); 163 security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4)); 164 } 165 166 struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp, 167 const struct sk_buff *skb); 168 struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp, 169 struct fib_result *res, 170 const struct sk_buff *skb); 171 172 static inline struct rtable *__ip_route_output_key(struct net *net, 173 struct flowi4 *flp) 174 { 175 return ip_route_output_key_hash(net, flp, NULL); 176 } 177 178 struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp, 179 const struct sock *sk); 180 struct dst_entry *ipv4_blackhole_route(struct net *net, 181 struct dst_entry *dst_orig); 182 183 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp) 184 { 185 return ip_route_output_flow(net, flp, NULL); 186 } 187 188 /* Simplistic IPv4 route lookup function. 189 * This is only suitable for some particular use cases: since the flowi4 190 * structure is only partially set, it may bypass some fib-rules. 191 */ 192 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr, 193 __be32 saddr, dscp_t dscp, 194 int oif, __u8 scope) 195 { 196 struct flowi4 fl4 = { 197 .flowi4_oif = oif, 198 .flowi4_dscp = dscp, 199 .flowi4_scope = scope, 200 .daddr = daddr, 201 .saddr = saddr, 202 }; 203 204 return ip_route_output_key(net, &fl4); 205 } 206 207 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4, 208 const struct sock *sk, 209 __be32 daddr, __be32 saddr, 210 __be16 dport, __be16 sport, 211 __u8 proto, __u8 tos, int oif) 212 { 213 flowi4_init_output(fl4, oif, sk ? READ_ONCE(sk->sk_mark) : 0, tos, 214 sk ? ip_sock_rt_scope(sk) : RT_SCOPE_UNIVERSE, 215 proto, sk ? inet_sk_flowi_flags(sk) : 0, 216 daddr, saddr, dport, sport, sock_net_uid(net, sk)); 217 if (sk) 218 security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4)); 219 return ip_route_output_flow(net, fl4, sk); 220 } 221 222 enum skb_drop_reason 223 ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr, 224 dscp_t dscp, struct net_device *dev, 225 struct in_device *in_dev, u32 *itag); 226 enum skb_drop_reason 227 ip_route_input_noref(struct sk_buff *skb, __be32 daddr, __be32 saddr, 228 dscp_t dscp, struct net_device *dev); 229 enum skb_drop_reason 230 ip_route_use_hint(struct sk_buff *skb, __be32 daddr, __be32 saddr, 231 dscp_t dscp, struct net_device *dev, 232 const struct sk_buff *hint); 233 234 static inline enum skb_drop_reason 235 ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src, dscp_t dscp, 236 struct net_device *devin) 237 { 238 enum skb_drop_reason reason; 239 240 rcu_read_lock(); 241 reason = ip_route_input_noref(skb, dst, src, dscp, devin); 242 if (!reason) { 243 skb_dst_force(skb); 244 if (!skb_dst(skb)) 245 reason = SKB_DROP_REASON_NOT_SPECIFIED; 246 } 247 rcu_read_unlock(); 248 249 return reason; 250 } 251 252 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif, 253 u8 protocol); 254 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu); 255 void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol); 256 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk); 257 void ip_rt_send_redirect(struct sk_buff *skb); 258 259 unsigned int inet_addr_type(struct net *net, __be32 addr); 260 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id); 261 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev, 262 __be32 addr); 263 unsigned int inet_addr_type_dev_table(struct net *net, 264 const struct net_device *dev, 265 __be32 addr); 266 void ip_rt_multicast_event(struct in_device *); 267 int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt); 268 void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt); 269 struct rtable *rt_dst_alloc(struct net_device *dev, 270 unsigned int flags, u16 type, bool noxfrm); 271 struct rtable *rt_dst_clone(struct net_device *dev, struct rtable *rt); 272 273 struct in_ifaddr; 274 void fib_add_ifaddr(struct in_ifaddr *); 275 void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *); 276 void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric); 277 278 void rt_add_uncached_list(struct rtable *rt); 279 void rt_del_uncached_list(struct rtable *rt); 280 281 int fib_dump_info_fnhe(struct sk_buff *skb, struct netlink_callback *cb, 282 u32 table_id, struct fib_info *fi, 283 int *fa_index, int fa_start, unsigned int flags); 284 285 void fnhe_update_pmtu(struct fib_nh_exception *fnhe, u32 new, u32 orig); 286 287 static inline void ip_rt_put(struct rtable *rt) 288 { 289 /* dst_release() accepts a NULL parameter. 290 * We rely on dst being first structure in struct rtable 291 */ 292 BUILD_BUG_ON(offsetof(struct rtable, dst) != 0); 293 dst_release(&rt->dst); 294 } 295 296 extern const __u8 ip_tos2prio[16]; 297 298 static inline char rt_tos2priority(u8 tos) 299 { 300 return ip_tos2prio[IPTOS_TOS(tos)>>1]; 301 } 302 303 /* ip_route_connect() and ip_route_newports() work in tandem whilst 304 * binding a socket for a new outgoing connection. 305 * 306 * In order to use IPSEC properly, we must, in the end, have a 307 * route that was looked up using all available keys including source 308 * and destination ports. 309 * 310 * However, if a source port needs to be allocated (the user specified 311 * a wildcard source port) we need to obtain addressing information 312 * in order to perform that allocation. 313 * 314 * So ip_route_connect() looks up a route using wildcarded source and 315 * destination ports in the key, simply so that we can get a pair of 316 * addresses to use for port allocation. 317 * 318 * Later, once the ports are allocated, ip_route_newports() will make 319 * another route lookup if needed to make sure we catch any IPSEC 320 * rules keyed on the port information. 321 * 322 * The callers allocate the flow key on their stack, and must pass in 323 * the same flowi4 object to both the ip_route_connect() and the 324 * ip_route_newports() calls. 325 */ 326 327 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, 328 __be32 src, int oif, u8 protocol, 329 __be16 sport, __be16 dport, 330 const struct sock *sk) 331 { 332 __u8 flow_flags = 0; 333 334 if (inet_test_bit(TRANSPARENT, sk)) 335 flow_flags |= FLOWI_FLAG_ANYSRC; 336 337 if (IS_ENABLED(CONFIG_IP_ROUTE_MULTIPATH) && !sport) 338 flow_flags |= FLOWI_FLAG_ANY_SPORT; 339 340 flowi4_init_output(fl4, oif, READ_ONCE(sk->sk_mark), ip_sock_rt_tos(sk), 341 ip_sock_rt_scope(sk), protocol, flow_flags, dst, 342 src, dport, sport, sk_uid(sk)); 343 } 344 345 static inline struct rtable *ip_route_connect(struct flowi4 *fl4, __be32 dst, 346 __be32 src, int oif, u8 protocol, 347 __be16 sport, __be16 dport, 348 const struct sock *sk) 349 { 350 struct net *net = sock_net(sk); 351 struct rtable *rt; 352 353 ip_route_connect_init(fl4, dst, src, oif, protocol, sport, dport, sk); 354 355 if (!dst || !src) { 356 rt = __ip_route_output_key(net, fl4); 357 if (IS_ERR(rt)) 358 return rt; 359 ip_rt_put(rt); 360 flowi4_update_output(fl4, oif, fl4->daddr, fl4->saddr); 361 } 362 security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4)); 363 return ip_route_output_flow(net, fl4, sk); 364 } 365 366 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt, 367 __be16 orig_sport, __be16 orig_dport, 368 __be16 sport, __be16 dport, 369 const struct sock *sk) 370 { 371 if (sport != orig_sport || dport != orig_dport) { 372 fl4->fl4_dport = dport; 373 fl4->fl4_sport = sport; 374 ip_rt_put(rt); 375 flowi4_update_output(fl4, sk->sk_bound_dev_if, fl4->daddr, 376 fl4->saddr); 377 security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4)); 378 return ip_route_output_flow(sock_net(sk), fl4, sk); 379 } 380 return rt; 381 } 382 383 static inline int inet_iif(const struct sk_buff *skb) 384 { 385 struct rtable *rt = skb_rtable(skb); 386 387 if (rt && rt->rt_iif) 388 return rt->rt_iif; 389 390 return skb->skb_iif; 391 } 392 393 static inline int ip4_dst_hoplimit(const struct dst_entry *dst) 394 { 395 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT); 396 397 if (hoplimit == 0) { 398 const struct net *net; 399 400 rcu_read_lock(); 401 net = dst_dev_net_rcu(dst); 402 hoplimit = READ_ONCE(net->ipv4.sysctl_ip_default_ttl); 403 rcu_read_unlock(); 404 } 405 return hoplimit; 406 } 407 408 static inline struct neighbour *ip_neigh_gw4(struct net_device *dev, 409 __be32 daddr) 410 { 411 struct neighbour *neigh; 412 413 neigh = __ipv4_neigh_lookup_noref(dev, (__force u32)daddr); 414 if (unlikely(!neigh)) { 415 struct neigh_table *tbl = arp_table(dev_net(dev)); 416 417 neigh = __neigh_create(tbl, &daddr, dev, false); 418 } 419 420 return neigh; 421 } 422 423 static inline struct neighbour *ip_neigh_for_gw(struct rtable *rt, 424 struct sk_buff *skb, 425 bool *is_v6gw) 426 { 427 struct net_device *dev = rt->dst.dev; 428 struct neighbour *neigh; 429 430 if (likely(rt->rt_gw_family == AF_INET)) { 431 neigh = ip_neigh_gw4(dev, rt->rt_gw4); 432 } else if (rt->rt_gw_family == AF_INET6) { 433 neigh = ip_neigh_gw6(dev, &rt->rt_gw6); 434 *is_v6gw = true; 435 } else { 436 neigh = ip_neigh_gw4(dev, ip_hdr(skb)->daddr); 437 } 438 return neigh; 439 } 440 441 #endif /* _ROUTE_H */ 442