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 struct fib_nh; 47 struct fib_info; 48 struct uncached_list; 49 struct rtable { 50 struct dst_entry dst; 51 52 int rt_genid; 53 unsigned int rt_flags; 54 __u16 rt_type; 55 __u8 rt_is_input; 56 __u8 rt_uses_gateway; 57 58 int rt_iif; 59 60 u8 rt_gw_family; 61 /* Info on neighbour */ 62 union { 63 __be32 rt_gw4; 64 struct in6_addr rt_gw6; 65 }; 66 67 /* Miscellaneous cached information */ 68 u32 rt_mtu_locked:1, 69 rt_pmtu:31; 70 71 struct list_head rt_uncached; 72 struct uncached_list *rt_uncached_list; 73 }; 74 75 static inline bool rt_is_input_route(const struct rtable *rt) 76 { 77 return rt->rt_is_input != 0; 78 } 79 80 static inline bool rt_is_output_route(const struct rtable *rt) 81 { 82 return rt->rt_is_input == 0; 83 } 84 85 static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr) 86 { 87 if (rt->rt_gw_family == AF_INET) 88 return rt->rt_gw4; 89 return daddr; 90 } 91 92 struct ip_rt_acct { 93 __u32 o_bytes; 94 __u32 o_packets; 95 __u32 i_bytes; 96 __u32 i_packets; 97 }; 98 99 struct rt_cache_stat { 100 unsigned int in_slow_tot; 101 unsigned int in_slow_mc; 102 unsigned int in_no_route; 103 unsigned int in_brd; 104 unsigned int in_martian_dst; 105 unsigned int in_martian_src; 106 unsigned int out_slow_tot; 107 unsigned int out_slow_mc; 108 }; 109 110 extern struct ip_rt_acct __percpu *ip_rt_acct; 111 112 struct in_device; 113 114 int ip_rt_init(void); 115 void rt_cache_flush(struct net *net); 116 void rt_flush_dev(struct net_device *dev); 117 struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp, 118 const struct sk_buff *skb); 119 struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp, 120 struct fib_result *res, 121 const struct sk_buff *skb); 122 123 static inline struct rtable *__ip_route_output_key(struct net *net, 124 struct flowi4 *flp) 125 { 126 return ip_route_output_key_hash(net, flp, NULL); 127 } 128 129 struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp, 130 const struct sock *sk); 131 struct rtable *ip_route_output_tunnel(struct sk_buff *skb, 132 struct net_device *dev, 133 struct net *net, __be32 *saddr, 134 const struct ip_tunnel_info *info, 135 u8 protocol, bool use_cache); 136 137 struct dst_entry *ipv4_blackhole_route(struct net *net, 138 struct dst_entry *dst_orig); 139 140 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp) 141 { 142 return ip_route_output_flow(net, flp, NULL); 143 } 144 145 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr, 146 __be32 saddr, u8 tos, int oif) 147 { 148 struct flowi4 fl4 = { 149 .flowi4_oif = oif, 150 .flowi4_tos = tos, 151 .daddr = daddr, 152 .saddr = saddr, 153 }; 154 return ip_route_output_key(net, &fl4); 155 } 156 157 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4, 158 struct sock *sk, 159 __be32 daddr, __be32 saddr, 160 __be16 dport, __be16 sport, 161 __u8 proto, __u8 tos, int oif) 162 { 163 flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos, 164 RT_SCOPE_UNIVERSE, proto, 165 sk ? inet_sk_flowi_flags(sk) : 0, 166 daddr, saddr, dport, sport, sock_net_uid(net, sk)); 167 if (sk) 168 security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 169 return ip_route_output_flow(net, fl4, sk); 170 } 171 172 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4, 173 __be32 daddr, __be32 saddr, 174 __be32 gre_key, __u8 tos, int oif) 175 { 176 memset(fl4, 0, sizeof(*fl4)); 177 fl4->flowi4_oif = oif; 178 fl4->daddr = daddr; 179 fl4->saddr = saddr; 180 fl4->flowi4_tos = tos; 181 fl4->flowi4_proto = IPPROTO_GRE; 182 fl4->fl4_gre_key = gre_key; 183 return ip_route_output_key(net, fl4); 184 } 185 int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr, 186 u8 tos, struct net_device *dev, 187 struct in_device *in_dev, u32 *itag); 188 int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src, 189 u8 tos, struct net_device *devin); 190 int ip_route_input_rcu(struct sk_buff *skb, __be32 dst, __be32 src, 191 u8 tos, struct net_device *devin, 192 struct fib_result *res); 193 194 int ip_route_use_hint(struct sk_buff *skb, __be32 dst, __be32 src, 195 u8 tos, struct net_device *devin, 196 const struct sk_buff *hint); 197 198 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src, 199 u8 tos, struct net_device *devin) 200 { 201 int err; 202 203 rcu_read_lock(); 204 err = ip_route_input_noref(skb, dst, src, tos, devin); 205 if (!err) { 206 skb_dst_force(skb); 207 if (!skb_dst(skb)) 208 err = -EINVAL; 209 } 210 rcu_read_unlock(); 211 212 return err; 213 } 214 215 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif, 216 u8 protocol); 217 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu); 218 void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol); 219 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk); 220 void ip_rt_send_redirect(struct sk_buff *skb); 221 222 unsigned int inet_addr_type(struct net *net, __be32 addr); 223 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id); 224 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev, 225 __be32 addr); 226 unsigned int inet_addr_type_dev_table(struct net *net, 227 const struct net_device *dev, 228 __be32 addr); 229 void ip_rt_multicast_event(struct in_device *); 230 int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt); 231 void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt); 232 struct rtable *rt_dst_alloc(struct net_device *dev, 233 unsigned int flags, u16 type, 234 bool nopolicy, bool noxfrm); 235 struct rtable *rt_dst_clone(struct net_device *dev, struct rtable *rt); 236 237 struct in_ifaddr; 238 void fib_add_ifaddr(struct in_ifaddr *); 239 void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *); 240 void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric); 241 242 void rt_add_uncached_list(struct rtable *rt); 243 void rt_del_uncached_list(struct rtable *rt); 244 245 int fib_dump_info_fnhe(struct sk_buff *skb, struct netlink_callback *cb, 246 u32 table_id, struct fib_info *fi, 247 int *fa_index, int fa_start, unsigned int flags); 248 249 static inline void ip_rt_put(struct rtable *rt) 250 { 251 /* dst_release() accepts a NULL parameter. 252 * We rely on dst being first structure in struct rtable 253 */ 254 BUILD_BUG_ON(offsetof(struct rtable, dst) != 0); 255 dst_release(&rt->dst); 256 } 257 258 #define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3) 259 260 extern const __u8 ip_tos2prio[16]; 261 262 static inline char rt_tos2priority(u8 tos) 263 { 264 return ip_tos2prio[IPTOS_TOS(tos)>>1]; 265 } 266 267 /* ip_route_connect() and ip_route_newports() work in tandem whilst 268 * binding a socket for a new outgoing connection. 269 * 270 * In order to use IPSEC properly, we must, in the end, have a 271 * route that was looked up using all available keys including source 272 * and destination ports. 273 * 274 * However, if a source port needs to be allocated (the user specified 275 * a wildcard source port) we need to obtain addressing information 276 * in order to perform that allocation. 277 * 278 * So ip_route_connect() looks up a route using wildcarded source and 279 * destination ports in the key, simply so that we can get a pair of 280 * addresses to use for port allocation. 281 * 282 * Later, once the ports are allocated, ip_route_newports() will make 283 * another route lookup if needed to make sure we catch any IPSEC 284 * rules keyed on the port information. 285 * 286 * The callers allocate the flow key on their stack, and must pass in 287 * the same flowi4 object to both the ip_route_connect() and the 288 * ip_route_newports() calls. 289 */ 290 291 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src, 292 u32 tos, int oif, u8 protocol, 293 __be16 sport, __be16 dport, 294 struct sock *sk) 295 { 296 __u8 flow_flags = 0; 297 298 if (inet_sk(sk)->transparent) 299 flow_flags |= FLOWI_FLAG_ANYSRC; 300 301 flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE, 302 protocol, flow_flags, dst, src, dport, sport, 303 sk->sk_uid); 304 } 305 306 static inline struct rtable *ip_route_connect(struct flowi4 *fl4, 307 __be32 dst, __be32 src, u32 tos, 308 int oif, u8 protocol, 309 __be16 sport, __be16 dport, 310 struct sock *sk) 311 { 312 struct net *net = sock_net(sk); 313 struct rtable *rt; 314 315 ip_route_connect_init(fl4, dst, src, tos, oif, protocol, 316 sport, dport, sk); 317 318 if (!dst || !src) { 319 rt = __ip_route_output_key(net, fl4); 320 if (IS_ERR(rt)) 321 return rt; 322 ip_rt_put(rt); 323 flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr); 324 } 325 security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 326 return ip_route_output_flow(net, fl4, sk); 327 } 328 329 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt, 330 __be16 orig_sport, __be16 orig_dport, 331 __be16 sport, __be16 dport, 332 struct sock *sk) 333 { 334 if (sport != orig_sport || dport != orig_dport) { 335 fl4->fl4_dport = dport; 336 fl4->fl4_sport = sport; 337 ip_rt_put(rt); 338 flowi4_update_output(fl4, sk->sk_bound_dev_if, 339 RT_CONN_FLAGS(sk), fl4->daddr, 340 fl4->saddr); 341 security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 342 return ip_route_output_flow(sock_net(sk), fl4, sk); 343 } 344 return rt; 345 } 346 347 static inline int inet_iif(const struct sk_buff *skb) 348 { 349 struct rtable *rt = skb_rtable(skb); 350 351 if (rt && rt->rt_iif) 352 return rt->rt_iif; 353 354 return skb->skb_iif; 355 } 356 357 static inline int ip4_dst_hoplimit(const struct dst_entry *dst) 358 { 359 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT); 360 struct net *net = dev_net(dst->dev); 361 362 if (hoplimit == 0) 363 hoplimit = net->ipv4.sysctl_ip_default_ttl; 364 return hoplimit; 365 } 366 367 static inline struct neighbour *ip_neigh_gw4(struct net_device *dev, 368 __be32 daddr) 369 { 370 struct neighbour *neigh; 371 372 neigh = __ipv4_neigh_lookup_noref(dev, daddr); 373 if (unlikely(!neigh)) 374 neigh = __neigh_create(&arp_tbl, &daddr, dev, false); 375 376 return neigh; 377 } 378 379 static inline struct neighbour *ip_neigh_for_gw(struct rtable *rt, 380 struct sk_buff *skb, 381 bool *is_v6gw) 382 { 383 struct net_device *dev = rt->dst.dev; 384 struct neighbour *neigh; 385 386 if (likely(rt->rt_gw_family == AF_INET)) { 387 neigh = ip_neigh_gw4(dev, rt->rt_gw4); 388 } else if (rt->rt_gw_family == AF_INET6) { 389 neigh = ip_neigh_gw6(dev, &rt->rt_gw6); 390 *is_v6gw = true; 391 } else { 392 neigh = ip_neigh_gw4(dev, ip_hdr(skb)->daddr); 393 } 394 return neigh; 395 } 396 397 #endif /* _ROUTE_H */ 398