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 #define RT_CONN_FLAGS(sk) (RT_TOS(inet_sk(sk)->tos) | sock_flag(sk, SOCK_LOCALROUTE)) 41 #define RT_CONN_FLAGS_TOS(sk,tos) (RT_TOS(tos) | sock_flag(sk, SOCK_LOCALROUTE)) 42 ip_sock_rt_scope(const struct sock * sk)43 static inline __u8 ip_sock_rt_scope(const struct sock *sk) 44 { 45 if (sock_flag(sk, SOCK_LOCALROUTE)) 46 return RT_SCOPE_LINK; 47 48 return RT_SCOPE_UNIVERSE; 49 } 50 ip_sock_rt_tos(const struct sock * sk)51 static inline __u8 ip_sock_rt_tos(const struct sock *sk) 52 { 53 return RT_TOS(inet_sk(sk)->tos); 54 } 55 56 struct ip_tunnel_info; 57 struct fib_nh; 58 struct fib_info; 59 struct uncached_list; 60 struct rtable { 61 struct dst_entry dst; 62 63 int rt_genid; 64 unsigned int rt_flags; 65 __u16 rt_type; 66 __u8 rt_is_input; 67 __u8 rt_uses_gateway; 68 69 int rt_iif; 70 71 u8 rt_gw_family; 72 /* Info on neighbour */ 73 union { 74 __be32 rt_gw4; 75 struct in6_addr rt_gw6; 76 }; 77 78 /* Miscellaneous cached information */ 79 u32 rt_mtu_locked:1, 80 rt_pmtu:31; 81 }; 82 rt_is_input_route(const struct rtable * rt)83 static inline bool rt_is_input_route(const struct rtable *rt) 84 { 85 return rt->rt_is_input != 0; 86 } 87 rt_is_output_route(const struct rtable * rt)88 static inline bool rt_is_output_route(const struct rtable *rt) 89 { 90 return rt->rt_is_input == 0; 91 } 92 rt_nexthop(const struct rtable * rt,__be32 daddr)93 static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr) 94 { 95 if (rt->rt_gw_family == AF_INET) 96 return rt->rt_gw4; 97 return daddr; 98 } 99 100 struct ip_rt_acct { 101 __u32 o_bytes; 102 __u32 o_packets; 103 __u32 i_bytes; 104 __u32 i_packets; 105 }; 106 107 struct rt_cache_stat { 108 unsigned int in_slow_tot; 109 unsigned int in_slow_mc; 110 unsigned int in_no_route; 111 unsigned int in_brd; 112 unsigned int in_martian_dst; 113 unsigned int in_martian_src; 114 unsigned int out_slow_tot; 115 unsigned int out_slow_mc; 116 }; 117 118 extern struct ip_rt_acct __percpu *ip_rt_acct; 119 120 struct in_device; 121 122 int ip_rt_init(void); 123 void rt_cache_flush(struct net *net); 124 void rt_flush_dev(struct net_device *dev); 125 struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp, 126 const struct sk_buff *skb); 127 struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp, 128 struct fib_result *res, 129 const struct sk_buff *skb); 130 __ip_route_output_key(struct net * net,struct flowi4 * flp)131 static inline struct rtable *__ip_route_output_key(struct net *net, 132 struct flowi4 *flp) 133 { 134 return ip_route_output_key_hash(net, flp, NULL); 135 } 136 137 struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp, 138 const struct sock *sk); 139 struct rtable *ip_route_output_tunnel(struct sk_buff *skb, 140 struct net_device *dev, 141 struct net *net, __be32 *saddr, 142 const struct ip_tunnel_info *info, 143 u8 protocol, bool use_cache); 144 145 struct dst_entry *ipv4_blackhole_route(struct net *net, 146 struct dst_entry *dst_orig); 147 ip_route_output_key(struct net * net,struct flowi4 * flp)148 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp) 149 { 150 return ip_route_output_flow(net, flp, NULL); 151 } 152 ip_route_output(struct net * net,__be32 daddr,__be32 saddr,u8 tos,int oif)153 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr, 154 __be32 saddr, u8 tos, int oif) 155 { 156 struct flowi4 fl4 = { 157 .flowi4_oif = oif, 158 .flowi4_tos = tos, 159 .daddr = daddr, 160 .saddr = saddr, 161 }; 162 return ip_route_output_key(net, &fl4); 163 } 164 ip_route_output_ports(struct net * net,struct flowi4 * fl4,const struct sock * sk,__be32 daddr,__be32 saddr,__be16 dport,__be16 sport,__u8 proto,__u8 tos,int oif)165 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4, 166 const struct sock *sk, 167 __be32 daddr, __be32 saddr, 168 __be16 dport, __be16 sport, 169 __u8 proto, __u8 tos, int oif) 170 { 171 flowi4_init_output(fl4, oif, sk ? READ_ONCE(sk->sk_mark) : 0, tos, 172 RT_SCOPE_UNIVERSE, proto, 173 sk ? inet_sk_flowi_flags(sk) : 0, 174 daddr, saddr, dport, sport, sock_net_uid(net, sk)); 175 if (sk) 176 security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4)); 177 return ip_route_output_flow(net, fl4, sk); 178 } 179 ip_route_output_gre(struct net * net,struct flowi4 * fl4,__be32 daddr,__be32 saddr,__be32 gre_key,__u8 tos,int oif)180 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4, 181 __be32 daddr, __be32 saddr, 182 __be32 gre_key, __u8 tos, int oif) 183 { 184 memset(fl4, 0, sizeof(*fl4)); 185 fl4->flowi4_oif = oif; 186 fl4->daddr = daddr; 187 fl4->saddr = saddr; 188 fl4->flowi4_tos = tos; 189 fl4->flowi4_proto = IPPROTO_GRE; 190 fl4->fl4_gre_key = gre_key; 191 return ip_route_output_key(net, fl4); 192 } 193 int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr, 194 u8 tos, struct net_device *dev, 195 struct in_device *in_dev, u32 *itag); 196 int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src, 197 u8 tos, struct net_device *devin); 198 int ip_route_use_hint(struct sk_buff *skb, __be32 dst, __be32 src, 199 u8 tos, struct net_device *devin, 200 const struct sk_buff *hint); 201 ip_route_input(struct sk_buff * skb,__be32 dst,__be32 src,dscp_t dscp,struct net_device * devin)202 static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src, 203 dscp_t dscp, struct net_device *devin) 204 { 205 int err; 206 207 rcu_read_lock(); 208 err = ip_route_input_noref(skb, dst, src, inet_dscp_to_dsfield(dscp), 209 devin); 210 if (!err) { 211 skb_dst_force(skb); 212 if (!skb_dst(skb)) 213 err = -EINVAL; 214 } 215 rcu_read_unlock(); 216 217 return err; 218 } 219 220 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif, 221 u8 protocol); 222 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu); 223 void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol); 224 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk); 225 void ip_rt_send_redirect(struct sk_buff *skb); 226 227 unsigned int inet_addr_type(struct net *net, __be32 addr); 228 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id); 229 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev, 230 __be32 addr); 231 unsigned int inet_addr_type_dev_table(struct net *net, 232 const struct net_device *dev, 233 __be32 addr); 234 void ip_rt_multicast_event(struct in_device *); 235 int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt); 236 void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt); 237 struct rtable *rt_dst_alloc(struct net_device *dev, 238 unsigned int flags, u16 type, bool noxfrm); 239 struct rtable *rt_dst_clone(struct net_device *dev, struct rtable *rt); 240 241 struct in_ifaddr; 242 void fib_add_ifaddr(struct in_ifaddr *); 243 void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *); 244 void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric); 245 246 void rt_add_uncached_list(struct rtable *rt); 247 void rt_del_uncached_list(struct rtable *rt); 248 249 int fib_dump_info_fnhe(struct sk_buff *skb, struct netlink_callback *cb, 250 u32 table_id, struct fib_info *fi, 251 int *fa_index, int fa_start, unsigned int flags); 252 ip_rt_put(struct rtable * rt)253 static inline void ip_rt_put(struct rtable *rt) 254 { 255 /* dst_release() accepts a NULL parameter. 256 * We rely on dst being first structure in struct rtable 257 */ 258 BUILD_BUG_ON(offsetof(struct rtable, dst) != 0); 259 dst_release(&rt->dst); 260 } 261 262 #define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3) 263 264 extern const __u8 ip_tos2prio[16]; 265 rt_tos2priority(u8 tos)266 static inline char rt_tos2priority(u8 tos) 267 { 268 return ip_tos2prio[IPTOS_TOS(tos)>>1]; 269 } 270 271 /* ip_route_connect() and ip_route_newports() work in tandem whilst 272 * binding a socket for a new outgoing connection. 273 * 274 * In order to use IPSEC properly, we must, in the end, have a 275 * route that was looked up using all available keys including source 276 * and destination ports. 277 * 278 * However, if a source port needs to be allocated (the user specified 279 * a wildcard source port) we need to obtain addressing information 280 * in order to perform that allocation. 281 * 282 * So ip_route_connect() looks up a route using wildcarded source and 283 * destination ports in the key, simply so that we can get a pair of 284 * addresses to use for port allocation. 285 * 286 * Later, once the ports are allocated, ip_route_newports() will make 287 * another route lookup if needed to make sure we catch any IPSEC 288 * rules keyed on the port information. 289 * 290 * The callers allocate the flow key on their stack, and must pass in 291 * the same flowi4 object to both the ip_route_connect() and the 292 * ip_route_newports() calls. 293 */ 294 ip_route_connect_init(struct flowi4 * fl4,__be32 dst,__be32 src,int oif,u8 protocol,__be16 sport,__be16 dport,const struct sock * sk)295 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst, 296 __be32 src, int oif, u8 protocol, 297 __be16 sport, __be16 dport, 298 const struct sock *sk) 299 { 300 __u8 flow_flags = 0; 301 302 if (inet_test_bit(TRANSPARENT, sk)) 303 flow_flags |= FLOWI_FLAG_ANYSRC; 304 305 flowi4_init_output(fl4, oif, READ_ONCE(sk->sk_mark), ip_sock_rt_tos(sk), 306 ip_sock_rt_scope(sk), protocol, flow_flags, dst, 307 src, dport, sport, sk->sk_uid); 308 } 309 ip_route_connect(struct flowi4 * fl4,__be32 dst,__be32 src,int oif,u8 protocol,__be16 sport,__be16 dport,const struct sock * sk)310 static inline struct rtable *ip_route_connect(struct flowi4 *fl4, __be32 dst, 311 __be32 src, int oif, u8 protocol, 312 __be16 sport, __be16 dport, 313 const struct sock *sk) 314 { 315 struct net *net = sock_net(sk); 316 struct rtable *rt; 317 318 ip_route_connect_init(fl4, dst, src, oif, protocol, sport, dport, sk); 319 320 if (!dst || !src) { 321 rt = __ip_route_output_key(net, fl4); 322 if (IS_ERR(rt)) 323 return rt; 324 ip_rt_put(rt); 325 flowi4_update_output(fl4, oif, fl4->daddr, fl4->saddr); 326 } 327 security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4)); 328 return ip_route_output_flow(net, fl4, sk); 329 } 330 ip_route_newports(struct flowi4 * fl4,struct rtable * rt,__be16 orig_sport,__be16 orig_dport,__be16 sport,__be16 dport,const struct sock * sk)331 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt, 332 __be16 orig_sport, __be16 orig_dport, 333 __be16 sport, __be16 dport, 334 const struct sock *sk) 335 { 336 if (sport != orig_sport || dport != orig_dport) { 337 fl4->fl4_dport = dport; 338 fl4->fl4_sport = sport; 339 ip_rt_put(rt); 340 flowi4_update_output(fl4, sk->sk_bound_dev_if, fl4->daddr, 341 fl4->saddr); 342 security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4)); 343 return ip_route_output_flow(sock_net(sk), fl4, sk); 344 } 345 return rt; 346 } 347 inet_iif(const struct sk_buff * skb)348 static inline int inet_iif(const struct sk_buff *skb) 349 { 350 struct rtable *rt = skb_rtable(skb); 351 352 if (rt && rt->rt_iif) 353 return rt->rt_iif; 354 355 return skb->skb_iif; 356 } 357 ip4_dst_hoplimit(const struct dst_entry * dst)358 static inline int ip4_dst_hoplimit(const struct dst_entry *dst) 359 { 360 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT); 361 362 if (hoplimit == 0) { 363 const struct net *net; 364 365 rcu_read_lock(); 366 net = dev_net_rcu(dst->dev); 367 hoplimit = READ_ONCE(net->ipv4.sysctl_ip_default_ttl); 368 rcu_read_unlock(); 369 } 370 return hoplimit; 371 } 372 ip_neigh_gw4(struct net_device * dev,__be32 daddr)373 static inline struct neighbour *ip_neigh_gw4(struct net_device *dev, 374 __be32 daddr) 375 { 376 struct neighbour *neigh; 377 378 neigh = __ipv4_neigh_lookup_noref(dev, (__force u32)daddr); 379 if (unlikely(!neigh)) 380 neigh = __neigh_create(&arp_tbl, &daddr, dev, false); 381 382 return neigh; 383 } 384 ip_neigh_for_gw(struct rtable * rt,struct sk_buff * skb,bool * is_v6gw)385 static inline struct neighbour *ip_neigh_for_gw(struct rtable *rt, 386 struct sk_buff *skb, 387 bool *is_v6gw) 388 { 389 struct net_device *dev = rt->dst.dev; 390 struct neighbour *neigh; 391 392 if (likely(rt->rt_gw_family == AF_INET)) { 393 neigh = ip_neigh_gw4(dev, rt->rt_gw4); 394 } else if (rt->rt_gw_family == AF_INET6) { 395 neigh = ip_neigh_gw6(dev, &rt->rt_gw6); 396 *is_v6gw = true; 397 } else { 398 neigh = ip_neigh_gw4(dev, ip_hdr(skb)->daddr); 399 } 400 return neigh; 401 } 402 403 #endif /* _ROUTE_H */ 404