include/net/route.h

   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 <linux/in_route.h>
  32 #include <linux/rtnetlink.h>
  33 #include <linux/rcupdate.h>
  34 #include <linux/route.h>
  35 #include <linux/ip.h>
  36 #include <linux/cache.h>
  37 #include <linux/security.h>
  38
  39 static inline __u8 ip_sock_rt_scope(const struct sock *sk)
  40 {
  41         if (sock_flag(sk, SOCK_LOCALROUTE))
  42                 return RT_SCOPE_LINK;
  43
  44         return RT_SCOPE_UNIVERSE;
  45 }
  46
  47 static inline __u8 ip_sock_rt_tos(const struct sock *sk)
  48 {
  49         return READ_ONCE(inet_sk(sk)->tos) & INET_DSCP_MASK;
  50 }
  51
  52 struct ip_tunnel_info;
  53 struct fib_nh;
  54 struct fib_info;
  55 struct uncached_list;
  56 struct rtable {
  57         struct dst_entry        dst;
  58
  59         int                     rt_genid;
  60         unsigned int            rt_flags;
  61         __u16                   rt_type;
  62         __u8                    rt_is_input;
  63         __u8                    rt_uses_gateway;
  64
  65         int                     rt_iif;
  66
  67         u8                      rt_gw_family;
  68         /* Info on neighbour */
  69         union {
  70                 __be32          rt_gw4;
  71                 struct in6_addr rt_gw6;
  72         };
  73
  74         /* Miscellaneous cached information */
  75         u32                     rt_mtu_locked:1,
  76                                 rt_pmtu:31;
  77 };
  78
  79 #define dst_rtable(_ptr) container_of_const(_ptr, struct rtable, dst)
  80
  81 /**
  82  * skb_rtable - Returns the skb &rtable
  83  * @skb: buffer
  84  */
  85 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  86 {
  87         return dst_rtable(skb_dst(skb));
  88 }
  89
  90 static inline bool rt_is_input_route(const struct rtable *rt)
  91 {
  92         return rt->rt_is_input != 0;
  93 }
  94
  95 static inline bool rt_is_output_route(const struct rtable *rt)
  96 {
  97         return rt->rt_is_input == 0;
  98 }
  99
 100 static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr)
 101 {
 102         if (rt->rt_gw_family == AF_INET)
 103                 return rt->rt_gw4;
 104         return daddr;
 105 }
 106
 107 struct ip_rt_acct {
 108         __u32   o_bytes;
 109         __u32   o_packets;
 110         __u32   i_bytes;
 111         __u32   i_packets;
 112 };
 113
 114 struct rt_cache_stat {
 115         unsigned int in_slow_tot;
 116         unsigned int in_slow_mc;
 117         unsigned int in_no_route;
 118         unsigned int in_brd;
 119         unsigned int in_martian_dst;
 120         unsigned int in_martian_src;
 121         unsigned int out_slow_tot;
 122         unsigned int out_slow_mc;
 123 };
 124
 125 extern struct ip_rt_acct __percpu *ip_rt_acct;
 126
 127 struct in_device;
 128
 129 int ip_rt_init(void);
 130 void rt_cache_flush(struct net *net);
 131 void rt_flush_dev(struct net_device *dev);
 132 struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp,
 133                                         const struct sk_buff *skb);
 134 struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp,
 135                                             struct fib_result *res,
 136                                             const struct sk_buff *skb);
 137
 138 static inline struct rtable *__ip_route_output_key(struct net *net,
 139                                                    struct flowi4 *flp)
 140 {
 141         return ip_route_output_key_hash(net, flp, NULL);
 142 }
 143
 144 struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
 145                                     const struct sock *sk);
 146 struct dst_entry *ipv4_blackhole_route(struct net *net,
 147                                        struct dst_entry *dst_orig);
 148
 149 static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
 150 {
 151         return ip_route_output_flow(net, flp, NULL);
 152 }
 153
 154 /* Simplistic IPv4 route lookup function.
 155  * This is only suitable for some particular use cases: since the flowi4
 156  * structure is only partially set, it may bypass some fib-rules.
 157  */
 158 static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
 159                                              __be32 saddr, dscp_t dscp,
 160                                              int oif, __u8 scope)
 161 {
 162         struct flowi4 fl4 = {
 163                 .flowi4_oif = oif,
 164                 .flowi4_tos = inet_dscp_to_dsfield(dscp),
 165                 .flowi4_scope = scope,
 166                 .daddr = daddr,
 167                 .saddr = saddr,
 168         };
 169
 170         return ip_route_output_key(net, &fl4);
 171 }
 172
 173 static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
 174                                                    const struct sock *sk,
 175                                                    __be32 daddr, __be32 saddr,
 176                                                    __be16 dport, __be16 sport,
 177                                                    __u8 proto, __u8 tos, int oif)
 178 {
 179         flowi4_init_output(fl4, oif, sk ? READ_ONCE(sk->sk_mark) : 0, tos,
 180                            sk ? ip_sock_rt_scope(sk) : RT_SCOPE_UNIVERSE,
 181                            proto, sk ? inet_sk_flowi_flags(sk) : 0,
 182                            daddr, saddr, dport, sport, sock_net_uid(net, sk));
 183         if (sk)
 184                 security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
 185         return ip_route_output_flow(net, fl4, sk);
 186 }
 187
 188 static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4,
 189                                                  __be32 daddr, __be32 saddr,
 190                                                  __be32 gre_key, __u8 tos, int oif)
 191 {
 192         memset(fl4, 0, sizeof(*fl4));
 193         fl4->flowi4_oif = oif;
 194         fl4->daddr = daddr;
 195         fl4->saddr = saddr;
 196         fl4->flowi4_tos = tos;
 197         fl4->flowi4_proto = IPPROTO_GRE;
 198         fl4->fl4_gre_key = gre_key;
 199         return ip_route_output_key(net, fl4);
 200 }
 201
 202 enum skb_drop_reason
 203 ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr,
 204                       dscp_t dscp, struct net_device *dev,
 205                       struct in_device *in_dev, u32 *itag);
 206 enum skb_drop_reason
 207 ip_route_input_noref(struct sk_buff *skb, __be32 daddr, __be32 saddr,
 208                      dscp_t dscp, struct net_device *dev);
 209 enum skb_drop_reason
 210 ip_route_use_hint(struct sk_buff *skb, __be32 daddr, __be32 saddr,
 211                   dscp_t dscp, struct net_device *dev,
 212                   const struct sk_buff *hint);
 213
 214 static inline enum skb_drop_reason
 215 ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src, dscp_t dscp,
 216                struct net_device *devin)
 217 {
 218         enum skb_drop_reason reason;
 219
 220         rcu_read_lock();
 221         reason = ip_route_input_noref(skb, dst, src, dscp, devin);
 222         if (!reason) {
 223                 skb_dst_force(skb);
 224                 if (!skb_dst(skb))
 225                         reason = SKB_DROP_REASON_NOT_SPECIFIED;
 226         }
 227         rcu_read_unlock();
 228
 229         return reason;
 230 }
 231
 232 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif,
 233                       u8 protocol);
 234 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
 235 void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol);
 236 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
 237 void ip_rt_send_redirect(struct sk_buff *skb);
 238
 239 unsigned int inet_addr_type(struct net *net, __be32 addr);
 240 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id);
 241 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
 242                                 __be32 addr);
 243 unsigned int inet_addr_type_dev_table(struct net *net,
 244                                       const struct net_device *dev,
 245                                       __be32 addr);
 246 void ip_rt_multicast_event(struct in_device *);
 247 int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt);
 248 void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
 249 struct rtable *rt_dst_alloc(struct net_device *dev,
 250                             unsigned int flags, u16 type, bool noxfrm);
 251 struct rtable *rt_dst_clone(struct net_device *dev, struct rtable *rt);
 252
 253 struct in_ifaddr;
 254 void fib_add_ifaddr(struct in_ifaddr *);
 255 void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
 256 void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric);
 257
 258 void rt_add_uncached_list(struct rtable *rt);
 259 void rt_del_uncached_list(struct rtable *rt);
 260
 261 int fib_dump_info_fnhe(struct sk_buff *skb, struct netlink_callback *cb,
 262                        u32 table_id, struct fib_info *fi,
 263                        int *fa_index, int fa_start, unsigned int flags);
 264
 265 static inline void ip_rt_put(struct rtable *rt)
 266 {
 267         /* dst_release() accepts a NULL parameter.
 268          * We rely on dst being first structure in struct rtable
 269          */
 270         BUILD_BUG_ON(offsetof(struct rtable, dst) != 0);
 271         dst_release(&rt->dst);
 272 }
 273
 274 extern const __u8 ip_tos2prio[16];
 275
 276 static inline char rt_tos2priority(u8 tos)
 277 {
 278         return ip_tos2prio[IPTOS_TOS(tos)>>1];
 279 }
 280
 281 /* ip_route_connect() and ip_route_newports() work in tandem whilst
 282  * binding a socket for a new outgoing connection.
 283  *
 284  * In order to use IPSEC properly, we must, in the end, have a
 285  * route that was looked up using all available keys including source
 286  * and destination ports.
 287  *
 288  * However, if a source port needs to be allocated (the user specified
 289  * a wildcard source port) we need to obtain addressing information
 290  * in order to perform that allocation.
 291  *
 292  * So ip_route_connect() looks up a route using wildcarded source and
 293  * destination ports in the key, simply so that we can get a pair of
 294  * addresses to use for port allocation.
 295  *
 296  * Later, once the ports are allocated, ip_route_newports() will make
 297  * another route lookup if needed to make sure we catch any IPSEC
 298  * rules keyed on the port information.
 299  *
 300  * The callers allocate the flow key on their stack, and must pass in
 301  * the same flowi4 object to both the ip_route_connect() and the
 302  * ip_route_newports() calls.
 303  */
 304
 305 static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst,
 306                                          __be32 src, int oif, u8 protocol,
 307                                          __be16 sport, __be16 dport,
 308                                          const struct sock *sk)
 309 {
 310         __u8 flow_flags = 0;
 311
 312         if (inet_test_bit(TRANSPARENT, sk))
 313                 flow_flags |= FLOWI_FLAG_ANYSRC;
 314
 315         flowi4_init_output(fl4, oif, READ_ONCE(sk->sk_mark), ip_sock_rt_tos(sk),
 316                            ip_sock_rt_scope(sk), protocol, flow_flags, dst,
 317                            src, dport, sport, sk->sk_uid);
 318 }
 319
 320 static inline struct rtable *ip_route_connect(struct flowi4 *fl4, __be32 dst,
 321                                               __be32 src, int oif, u8 protocol,
 322                                               __be16 sport, __be16 dport,
 323                                               const struct sock *sk)
 324 {
 325         struct net *net = sock_net(sk);
 326         struct rtable *rt;
 327
 328         ip_route_connect_init(fl4, dst, src, oif, protocol, sport, dport, sk);
 329
 330         if (!dst || !src) {
 331                 rt = __ip_route_output_key(net, fl4);
 332                 if (IS_ERR(rt))
 333                         return rt;
 334                 ip_rt_put(rt);
 335                 flowi4_update_output(fl4, oif, fl4->daddr, fl4->saddr);
 336         }
 337         security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
 338         return ip_route_output_flow(net, fl4, sk);
 339 }
 340
 341 static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
 342                                                __be16 orig_sport, __be16 orig_dport,
 343                                                __be16 sport, __be16 dport,
 344                                                const struct sock *sk)
 345 {
 346         if (sport != orig_sport || dport != orig_dport) {
 347                 fl4->fl4_dport = dport;
 348                 fl4->fl4_sport = sport;
 349                 ip_rt_put(rt);
 350                 flowi4_update_output(fl4, sk->sk_bound_dev_if, fl4->daddr,
 351                                      fl4->saddr);
 352                 security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
 353                 return ip_route_output_flow(sock_net(sk), fl4, sk);
 354         }
 355         return rt;
 356 }
 357
 358 static inline int inet_iif(const struct sk_buff *skb)
 359 {
 360         struct rtable *rt = skb_rtable(skb);
 361
 362         if (rt && rt->rt_iif)
 363                 return rt->rt_iif;
 364
 365         return skb->skb_iif;
 366 }
 367
 368 static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
 369 {
 370         int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
 371         struct net *net = dev_net(dst->dev);
 372
 373         if (hoplimit == 0)
 374                 hoplimit = READ_ONCE(net->ipv4.sysctl_ip_default_ttl);
 375         return hoplimit;
 376 }
 377
 378 static inline struct neighbour *ip_neigh_gw4(struct net_device *dev,
 379                                              __be32 daddr)
 380 {
 381         struct neighbour *neigh;
 382
 383         neigh = __ipv4_neigh_lookup_noref(dev, (__force u32)daddr);
 384         if (unlikely(!neigh))
 385                 neigh = __neigh_create(&arp_tbl, &daddr, dev, false);
 386
 387         return neigh;
 388 }
 389
 390 static inline struct neighbour *ip_neigh_for_gw(struct rtable *rt,
 391                                                 struct sk_buff *skb,
 392                                                 bool *is_v6gw)
 393 {
 394         struct net_device *dev = rt->dst.dev;
 395         struct neighbour *neigh;
 396
 397         if (likely(rt->rt_gw_family == AF_INET)) {
 398                 neigh = ip_neigh_gw4(dev, rt->rt_gw4);
 399         } else if (rt->rt_gw_family == AF_INET6) {
 400                 neigh = ip_neigh_gw6(dev, &rt->rt_gw6);
 401                 *is_v6gw = true;
 402         } else {
 403                 neigh = ip_neigh_gw4(dev, ip_hdr(skb)->daddr);
 404         }
 405         return neigh;
 406 }
 407
 408 #endif  /* _ROUTE_H */