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, 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 /* Simplistic IPv4 route lookup function. 145 * This is only suitable for some particular use cases: since the flowi4 146 * structure is only partially set, it may bypass some fib-rules. 147 */ 148 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr, 149 __be32 saddr, u8 tos, int oif, 150 __u8 scope) 151 { 152 struct flowi4 fl4 = { 153 .flowi4_oif = oif, 154 .flowi4_tos = tos, 155 .flowi4_scope = scope, 156 .daddr = daddr, 157 .saddr = saddr, 158 }; 159 160 return ip_route_output_key(net, &fl4); 161 } 162 163 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4, 164 const struct sock *sk, 165 __be32 daddr, __be32 saddr, 166 __be16 dport, __be16 sport, 167 __u8 proto, __u8 tos, int oif) 168 { 169 flowi4_init_output(fl4, oif, sk ? READ_ONCE(sk->sk_mark) : 0, tos, 170 sk ? ip_sock_rt_scope(sk) : RT_SCOPE_UNIVERSE, 171 proto, sk ? inet_sk_flowi_flags(sk) : 0, 172 daddr, saddr, dport, sport, sock_net_uid(net, sk)); 173 if (sk) 174 security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4)); 175 return ip_route_output_flow(net, fl4, sk); 176 } 177 178 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4, 179 __be32 daddr, __be32 saddr, 180 __be32 gre_key, __u8 tos, int oif) 181 { 182 memset(fl4, 0, sizeof(*fl4)); 183 fl4->flowi4_oif = oif; 184 fl4->daddr = daddr; 185 fl4->saddr = saddr; 186 fl4->flowi4_tos = tos; 187 fl4->flowi4_proto = IPPROTO_GRE; 188 fl4->fl4_gre_key = gre_key; 189 return ip_route_output_key(net, fl4); 190 } 191 int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr, 192 u8 tos, struct net_device *dev, 193 struct in_device *in_dev, u32 *itag); 194 int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src, 195 u8 tos, struct net_device *devin); 196 int ip_route_use_hint(struct sk_buff *skb, __be32 dst, __be32 src, 197 u8 tos, struct net_device *devin, 198 const struct sk_buff *hint); 199 200 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src, 201 u8 tos, struct net_device *devin) 202 { 203 int err; 204 205 rcu_read_lock(); 206 err = ip_route_input_noref(skb, dst, src, tos, devin); 207 if (!err) { 208 skb_dst_force(skb); 209 if (!skb_dst(skb)) 210 err = -EINVAL; 211 } 212 rcu_read_unlock(); 213 214 return err; 215 } 216 217 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif, 218 u8 protocol); 219 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu); 220 void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol); 221 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk); 222 void ip_rt_send_redirect(struct sk_buff *skb); 223 224 unsigned int inet_addr_type(struct net *net, __be32 addr); 225 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id); 226 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev, 227 __be32 addr); 228 unsigned int inet_addr_type_dev_table(struct net *net, 229 const struct net_device *dev, 230 __be32 addr); 231 void ip_rt_multicast_event(struct in_device *); 232 int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt); 233 void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt); 234 struct rtable *rt_dst_alloc(struct net_device *dev, 235 unsigned int flags, u16 type, bool noxfrm); 236 struct rtable *rt_dst_clone(struct net_device *dev, struct rtable *rt); 237 238 struct in_ifaddr; 239 void fib_add_ifaddr(struct in_ifaddr *); 240 void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *); 241 void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric); 242 243 void rt_add_uncached_list(struct rtable *rt); 244 void rt_del_uncached_list(struct rtable *rt); 245 246 int fib_dump_info_fnhe(struct sk_buff *skb, struct netlink_callback *cb, 247 u32 table_id, struct fib_info *fi, 248 int *fa_index, int fa_start, unsigned int flags); 249 250 static inline void ip_rt_put(struct rtable *rt) 251 { 252 /* dst_release() accepts a NULL parameter. 253 * We rely on dst being first structure in struct rtable 254 */ 255 BUILD_BUG_ON(offsetof(struct rtable, dst) != 0); 256 dst_release(&rt->dst); 257 } 258 259 #define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3) 260 261 extern const __u8 ip_tos2prio[16]; 262 263 static inline char rt_tos2priority(u8 tos) 264 { 265 return ip_tos2prio[IPTOS_TOS(tos)>>1]; 266 } 267 268 /* ip_route_connect() and ip_route_newports() work in tandem whilst 269 * binding a socket for a new outgoing connection. 270 * 271 * In order to use IPSEC properly, we must, in the end, have a 272 * route that was looked up using all available keys including source 273 * and destination ports. 274 * 275 * However, if a source port needs to be allocated (the user specified 276 * a wildcard source port) we need to obtain addressing information 277 * in order to perform that allocation. 278 * 279 * So ip_route_connect() looks up a route using wildcarded source and 280 * destination ports in the key, simply so that we can get a pair of 281 * addresses to use for port allocation. 282 * 283 * Later, once the ports are allocated, ip_route_newports() will make 284 * another route lookup if needed to make sure we catch any IPSEC 285 * rules keyed on the port information. 286 * 287 * The callers allocate the flow key on their stack, and must pass in 288 * the same flowi4 object to both the ip_route_connect() and the 289 * ip_route_newports() calls. 290 */ 291 292 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, 293 __be32 src, int oif, u8 protocol, 294 __be16 sport, __be16 dport, 295 const struct sock *sk) 296 { 297 __u8 flow_flags = 0; 298 299 if (inet_test_bit(TRANSPARENT, sk)) 300 flow_flags |= FLOWI_FLAG_ANYSRC; 301 302 flowi4_init_output(fl4, oif, READ_ONCE(sk->sk_mark), ip_sock_rt_tos(sk), 303 ip_sock_rt_scope(sk), protocol, flow_flags, dst, 304 src, dport, sport, sk->sk_uid); 305 } 306 307 static inline struct rtable *ip_route_connect(struct flowi4 *fl4, __be32 dst, 308 __be32 src, int oif, u8 protocol, 309 __be16 sport, __be16 dport, 310 const struct sock *sk) 311 { 312 struct net *net = sock_net(sk); 313 struct rtable *rt; 314 315 ip_route_connect_init(fl4, dst, src, oif, protocol, sport, dport, sk); 316 317 if (!dst || !src) { 318 rt = __ip_route_output_key(net, fl4); 319 if (IS_ERR(rt)) 320 return rt; 321 ip_rt_put(rt); 322 flowi4_update_output(fl4, oif, fl4->daddr, fl4->saddr); 323 } 324 security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4)); 325 return ip_route_output_flow(net, fl4, sk); 326 } 327 328 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt, 329 __be16 orig_sport, __be16 orig_dport, 330 __be16 sport, __be16 dport, 331 const struct sock *sk) 332 { 333 if (sport != orig_sport || dport != orig_dport) { 334 fl4->fl4_dport = dport; 335 fl4->fl4_sport = sport; 336 ip_rt_put(rt); 337 flowi4_update_output(fl4, sk->sk_bound_dev_if, fl4->daddr, 338 fl4->saddr); 339 security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4)); 340 return ip_route_output_flow(sock_net(sk), fl4, sk); 341 } 342 return rt; 343 } 344 345 static inline int inet_iif(const struct sk_buff *skb) 346 { 347 struct rtable *rt = skb_rtable(skb); 348 349 if (rt && rt->rt_iif) 350 return rt->rt_iif; 351 352 return skb->skb_iif; 353 } 354 355 static inline int ip4_dst_hoplimit(const struct dst_entry *dst) 356 { 357 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT); 358 struct net *net = dev_net(dst->dev); 359 360 if (hoplimit == 0) 361 hoplimit = READ_ONCE(net->ipv4.sysctl_ip_default_ttl); 362 return hoplimit; 363 } 364 365 static inline struct neighbour *ip_neigh_gw4(struct net_device *dev, 366 __be32 daddr) 367 { 368 struct neighbour *neigh; 369 370 neigh = __ipv4_neigh_lookup_noref(dev, (__force u32)daddr); 371 if (unlikely(!neigh)) 372 neigh = __neigh_create(&arp_tbl, &daddr, dev, false); 373 374 return neigh; 375 } 376 377 static inline struct neighbour *ip_neigh_for_gw(struct rtable *rt, 378 struct sk_buff *skb, 379 bool *is_v6gw) 380 { 381 struct net_device *dev = rt->dst.dev; 382 struct neighbour *neigh; 383 384 if (likely(rt->rt_gw_family == AF_INET)) { 385 neigh = ip_neigh_gw4(dev, rt->rt_gw4); 386 } else if (rt->rt_gw_family == AF_INET6) { 387 neigh = ip_neigh_gw6(dev, &rt->rt_gw6); 388 *is_v6gw = true; 389 } else { 390 neigh = ip_neigh_gw4(dev, ip_hdr(skb)->daddr); 391 } 392 return neigh; 393 } 394 395 #endif /* _ROUTE_H */ 396