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 dst_entry *ipv4_blackhole_route(struct net *net, 132 struct dst_entry *dst_orig); 133 134 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp) 135 { 136 return ip_route_output_flow(net, flp, NULL); 137 } 138 139 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr, 140 __be32 saddr, u8 tos, int oif) 141 { 142 struct flowi4 fl4 = { 143 .flowi4_oif = oif, 144 .flowi4_tos = tos, 145 .daddr = daddr, 146 .saddr = saddr, 147 }; 148 return ip_route_output_key(net, &fl4); 149 } 150 151 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4, 152 struct sock *sk, 153 __be32 daddr, __be32 saddr, 154 __be16 dport, __be16 sport, 155 __u8 proto, __u8 tos, int oif) 156 { 157 flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos, 158 RT_SCOPE_UNIVERSE, proto, 159 sk ? inet_sk_flowi_flags(sk) : 0, 160 daddr, saddr, dport, sport, sock_net_uid(net, sk)); 161 if (sk) 162 security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 163 return ip_route_output_flow(net, fl4, sk); 164 } 165 166 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4, 167 __be32 daddr, __be32 saddr, 168 __be32 gre_key, __u8 tos, int oif) 169 { 170 memset(fl4, 0, sizeof(*fl4)); 171 fl4->flowi4_oif = oif; 172 fl4->daddr = daddr; 173 fl4->saddr = saddr; 174 fl4->flowi4_tos = tos; 175 fl4->flowi4_proto = IPPROTO_GRE; 176 fl4->fl4_gre_key = gre_key; 177 return ip_route_output_key(net, fl4); 178 } 179 int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr, 180 u8 tos, struct net_device *dev, 181 struct in_device *in_dev, u32 *itag); 182 int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src, 183 u8 tos, struct net_device *devin); 184 int ip_route_input_rcu(struct sk_buff *skb, __be32 dst, __be32 src, 185 u8 tos, struct net_device *devin, 186 struct fib_result *res); 187 188 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src, 189 u8 tos, struct net_device *devin) 190 { 191 int err; 192 193 rcu_read_lock(); 194 err = ip_route_input_noref(skb, dst, src, tos, devin); 195 if (!err) { 196 skb_dst_force(skb); 197 if (!skb_dst(skb)) 198 err = -EINVAL; 199 } 200 rcu_read_unlock(); 201 202 return err; 203 } 204 205 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif, 206 u8 protocol); 207 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu); 208 void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol); 209 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk); 210 void ip_rt_send_redirect(struct sk_buff *skb); 211 212 unsigned int inet_addr_type(struct net *net, __be32 addr); 213 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id); 214 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev, 215 __be32 addr); 216 unsigned int inet_addr_type_dev_table(struct net *net, 217 const struct net_device *dev, 218 __be32 addr); 219 void ip_rt_multicast_event(struct in_device *); 220 int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt); 221 void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt); 222 struct rtable *rt_dst_alloc(struct net_device *dev, 223 unsigned int flags, u16 type, 224 bool nopolicy, bool noxfrm, bool will_cache); 225 struct rtable *rt_dst_clone(struct net_device *dev, struct rtable *rt); 226 227 struct in_ifaddr; 228 void fib_add_ifaddr(struct in_ifaddr *); 229 void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *); 230 void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric); 231 232 void rt_add_uncached_list(struct rtable *rt); 233 void rt_del_uncached_list(struct rtable *rt); 234 235 int fib_dump_info_fnhe(struct sk_buff *skb, struct netlink_callback *cb, 236 u32 table_id, struct fib_info *fi, 237 int *fa_index, int fa_start, unsigned int flags); 238 239 static inline void ip_rt_put(struct rtable *rt) 240 { 241 /* dst_release() accepts a NULL parameter. 242 * We rely on dst being first structure in struct rtable 243 */ 244 BUILD_BUG_ON(offsetof(struct rtable, dst) != 0); 245 dst_release(&rt->dst); 246 } 247 248 #define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3) 249 250 extern const __u8 ip_tos2prio[16]; 251 252 static inline char rt_tos2priority(u8 tos) 253 { 254 return ip_tos2prio[IPTOS_TOS(tos)>>1]; 255 } 256 257 /* ip_route_connect() and ip_route_newports() work in tandem whilst 258 * binding a socket for a new outgoing connection. 259 * 260 * In order to use IPSEC properly, we must, in the end, have a 261 * route that was looked up using all available keys including source 262 * and destination ports. 263 * 264 * However, if a source port needs to be allocated (the user specified 265 * a wildcard source port) we need to obtain addressing information 266 * in order to perform that allocation. 267 * 268 * So ip_route_connect() looks up a route using wildcarded source and 269 * destination ports in the key, simply so that we can get a pair of 270 * addresses to use for port allocation. 271 * 272 * Later, once the ports are allocated, ip_route_newports() will make 273 * another route lookup if needed to make sure we catch any IPSEC 274 * rules keyed on the port information. 275 * 276 * The callers allocate the flow key on their stack, and must pass in 277 * the same flowi4 object to both the ip_route_connect() and the 278 * ip_route_newports() calls. 279 */ 280 281 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, __be32 src, 282 u32 tos, int oif, u8 protocol, 283 __be16 sport, __be16 dport, 284 struct sock *sk) 285 { 286 __u8 flow_flags = 0; 287 288 if (inet_sk(sk)->transparent) 289 flow_flags |= FLOWI_FLAG_ANYSRC; 290 291 flowi4_init_output(fl4, oif, sk->sk_mark, tos, RT_SCOPE_UNIVERSE, 292 protocol, flow_flags, dst, src, dport, sport, 293 sk->sk_uid); 294 } 295 296 static inline struct rtable *ip_route_connect(struct flowi4 *fl4, 297 __be32 dst, __be32 src, u32 tos, 298 int oif, u8 protocol, 299 __be16 sport, __be16 dport, 300 struct sock *sk) 301 { 302 struct net *net = sock_net(sk); 303 struct rtable *rt; 304 305 ip_route_connect_init(fl4, dst, src, tos, oif, protocol, 306 sport, dport, sk); 307 308 if (!dst || !src) { 309 rt = __ip_route_output_key(net, fl4); 310 if (IS_ERR(rt)) 311 return rt; 312 ip_rt_put(rt); 313 flowi4_update_output(fl4, oif, tos, fl4->daddr, fl4->saddr); 314 } 315 security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 316 return ip_route_output_flow(net, fl4, sk); 317 } 318 319 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt, 320 __be16 orig_sport, __be16 orig_dport, 321 __be16 sport, __be16 dport, 322 struct sock *sk) 323 { 324 if (sport != orig_sport || dport != orig_dport) { 325 fl4->fl4_dport = dport; 326 fl4->fl4_sport = sport; 327 ip_rt_put(rt); 328 flowi4_update_output(fl4, sk->sk_bound_dev_if, 329 RT_CONN_FLAGS(sk), fl4->daddr, 330 fl4->saddr); 331 security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); 332 return ip_route_output_flow(sock_net(sk), fl4, sk); 333 } 334 return rt; 335 } 336 337 static inline int inet_iif(const struct sk_buff *skb) 338 { 339 struct rtable *rt = skb_rtable(skb); 340 341 if (rt && rt->rt_iif) 342 return rt->rt_iif; 343 344 return skb->skb_iif; 345 } 346 347 static inline int ip4_dst_hoplimit(const struct dst_entry *dst) 348 { 349 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT); 350 struct net *net = dev_net(dst->dev); 351 352 if (hoplimit == 0) 353 hoplimit = net->ipv4.sysctl_ip_default_ttl; 354 return hoplimit; 355 } 356 357 static inline struct neighbour *ip_neigh_gw4(struct net_device *dev, 358 __be32 daddr) 359 { 360 struct neighbour *neigh; 361 362 neigh = __ipv4_neigh_lookup_noref(dev, daddr); 363 if (unlikely(!neigh)) 364 neigh = __neigh_create(&arp_tbl, &daddr, dev, false); 365 366 return neigh; 367 } 368 369 static inline struct neighbour *ip_neigh_for_gw(struct rtable *rt, 370 struct sk_buff *skb, 371 bool *is_v6gw) 372 { 373 struct net_device *dev = rt->dst.dev; 374 struct neighbour *neigh; 375 376 if (likely(rt->rt_gw_family == AF_INET)) { 377 neigh = ip_neigh_gw4(dev, rt->rt_gw4); 378 } else if (rt->rt_gw_family == AF_INET6) { 379 neigh = ip_neigh_gw6(dev, &rt->rt_gw6); 380 *is_v6gw = true; 381 } else { 382 neigh = ip_neigh_gw4(dev, ip_hdr(skb)->daddr); 383 } 384 return neigh; 385 } 386 387 #endif /* _ROUTE_H */ 388