1 /* SPDX-License-Identifier: GPL-2.0 */ 2 /* 3 * Operations on the network namespace 4 */ 5 #ifndef __NET_NET_NAMESPACE_H 6 #define __NET_NET_NAMESPACE_H 7 8 #include <linux/atomic.h> 9 #include <linux/refcount.h> 10 #include <linux/workqueue.h> 11 #include <linux/list.h> 12 #include <linux/sysctl.h> 13 #include <linux/uidgid.h> 14 15 #include <net/flow.h> 16 #include <net/netns/core.h> 17 #include <net/netns/mib.h> 18 #include <net/netns/unix.h> 19 #include <net/netns/packet.h> 20 #include <net/netns/ipv4.h> 21 #include <net/netns/ipv6.h> 22 #include <net/netns/nexthop.h> 23 #include <net/netns/ieee802154_6lowpan.h> 24 #include <net/netns/sctp.h> 25 #include <net/netns/dccp.h> 26 #include <net/netns/netfilter.h> 27 #include <net/netns/x_tables.h> 28 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 29 #include <net/netns/conntrack.h> 30 #endif 31 #include <net/netns/nftables.h> 32 #include <net/netns/xfrm.h> 33 #include <net/netns/mpls.h> 34 #include <net/netns/can.h> 35 #include <net/netns/xdp.h> 36 #include <linux/ns_common.h> 37 #include <linux/idr.h> 38 #include <linux/skbuff.h> 39 40 struct user_namespace; 41 struct proc_dir_entry; 42 struct net_device; 43 struct sock; 44 struct ctl_table_header; 45 struct net_generic; 46 struct uevent_sock; 47 struct netns_ipvs; 48 struct bpf_prog; 49 50 51 #define NETDEV_HASHBITS 8 52 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS) 53 54 struct net { 55 /* First cache line can be often dirtied. 56 * Do not place here read-mostly fields. 57 */ 58 refcount_t passive; /* To decide when the network 59 * namespace should be freed. 60 */ 61 refcount_t count; /* To decided when the network 62 * namespace should be shut down. 63 */ 64 spinlock_t rules_mod_lock; 65 66 unsigned int dev_unreg_count; 67 68 unsigned int dev_base_seq; /* protected by rtnl_mutex */ 69 int ifindex; 70 71 spinlock_t nsid_lock; 72 atomic_t fnhe_genid; 73 74 struct list_head list; /* list of network namespaces */ 75 struct list_head exit_list; /* To linked to call pernet exit 76 * methods on dead net ( 77 * pernet_ops_rwsem read locked), 78 * or to unregister pernet ops 79 * (pernet_ops_rwsem write locked). 80 */ 81 struct llist_node cleanup_list; /* namespaces on death row */ 82 83 #ifdef CONFIG_KEYS 84 struct key_tag *key_domain; /* Key domain of operation tag */ 85 #endif 86 struct user_namespace *user_ns; /* Owning user namespace */ 87 struct ucounts *ucounts; 88 struct idr netns_ids; 89 90 struct ns_common ns; 91 92 struct list_head dev_base_head; 93 struct proc_dir_entry *proc_net; 94 struct proc_dir_entry *proc_net_stat; 95 96 #ifdef CONFIG_SYSCTL 97 struct ctl_table_set sysctls; 98 #endif 99 100 struct sock *rtnl; /* rtnetlink socket */ 101 struct sock *genl_sock; 102 103 struct uevent_sock *uevent_sock; /* uevent socket */ 104 105 struct hlist_head *dev_name_head; 106 struct hlist_head *dev_index_head; 107 /* Note that @hash_mix can be read millions times per second, 108 * it is critical that it is on a read_mostly cache line. 109 */ 110 u32 hash_mix; 111 112 struct net_device *loopback_dev; /* The loopback */ 113 114 /* core fib_rules */ 115 struct list_head rules_ops; 116 117 struct netns_core core; 118 struct netns_mib mib; 119 struct netns_packet packet; 120 struct netns_unix unx; 121 struct netns_nexthop nexthop; 122 struct netns_ipv4 ipv4; 123 #if IS_ENABLED(CONFIG_IPV6) 124 struct netns_ipv6 ipv6; 125 #endif 126 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN) 127 struct netns_ieee802154_lowpan ieee802154_lowpan; 128 #endif 129 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE) 130 struct netns_sctp sctp; 131 #endif 132 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE) 133 struct netns_dccp dccp; 134 #endif 135 #ifdef CONFIG_NETFILTER 136 struct netns_nf nf; 137 struct netns_xt xt; 138 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 139 struct netns_ct ct; 140 #endif 141 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE) 142 struct netns_nftables nft; 143 #endif 144 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6) 145 struct netns_nf_frag nf_frag; 146 struct ctl_table_header *nf_frag_frags_hdr; 147 #endif 148 struct sock *nfnl; 149 struct sock *nfnl_stash; 150 #if IS_ENABLED(CONFIG_NETFILTER_NETLINK_ACCT) 151 struct list_head nfnl_acct_list; 152 #endif 153 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT) 154 struct list_head nfct_timeout_list; 155 #endif 156 #endif 157 #ifdef CONFIG_WEXT_CORE 158 struct sk_buff_head wext_nlevents; 159 #endif 160 struct net_generic __rcu *gen; 161 162 struct bpf_prog __rcu *flow_dissector_prog; 163 164 /* Note : following structs are cache line aligned */ 165 #ifdef CONFIG_XFRM 166 struct netns_xfrm xfrm; 167 #endif 168 #if IS_ENABLED(CONFIG_IP_VS) 169 struct netns_ipvs *ipvs; 170 #endif 171 #if IS_ENABLED(CONFIG_MPLS) 172 struct netns_mpls mpls; 173 #endif 174 #if IS_ENABLED(CONFIG_CAN) 175 struct netns_can can; 176 #endif 177 #ifdef CONFIG_XDP_SOCKETS 178 struct netns_xdp xdp; 179 #endif 180 #if IS_ENABLED(CONFIG_CRYPTO_USER) 181 struct sock *crypto_nlsk; 182 #endif 183 struct sock *diag_nlsk; 184 } __randomize_layout; 185 186 #include <linux/seq_file_net.h> 187 188 /* Init's network namespace */ 189 extern struct net init_net; 190 191 #ifdef CONFIG_NET_NS 192 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns, 193 struct net *old_net); 194 195 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid); 196 197 void net_ns_barrier(void); 198 #else /* CONFIG_NET_NS */ 199 #include <linux/sched.h> 200 #include <linux/nsproxy.h> 201 static inline struct net *copy_net_ns(unsigned long flags, 202 struct user_namespace *user_ns, struct net *old_net) 203 { 204 if (flags & CLONE_NEWNET) 205 return ERR_PTR(-EINVAL); 206 return old_net; 207 } 208 209 static inline void net_ns_get_ownership(const struct net *net, 210 kuid_t *uid, kgid_t *gid) 211 { 212 *uid = GLOBAL_ROOT_UID; 213 *gid = GLOBAL_ROOT_GID; 214 } 215 216 static inline void net_ns_barrier(void) {} 217 #endif /* CONFIG_NET_NS */ 218 219 220 extern struct list_head net_namespace_list; 221 222 struct net *get_net_ns_by_pid(pid_t pid); 223 struct net *get_net_ns_by_fd(int fd); 224 225 #ifdef CONFIG_SYSCTL 226 void ipx_register_sysctl(void); 227 void ipx_unregister_sysctl(void); 228 #else 229 #define ipx_register_sysctl() 230 #define ipx_unregister_sysctl() 231 #endif 232 233 #ifdef CONFIG_NET_NS 234 void __put_net(struct net *net); 235 236 static inline struct net *get_net(struct net *net) 237 { 238 refcount_inc(&net->count); 239 return net; 240 } 241 242 static inline struct net *maybe_get_net(struct net *net) 243 { 244 /* Used when we know struct net exists but we 245 * aren't guaranteed a previous reference count 246 * exists. If the reference count is zero this 247 * function fails and returns NULL. 248 */ 249 if (!refcount_inc_not_zero(&net->count)) 250 net = NULL; 251 return net; 252 } 253 254 static inline void put_net(struct net *net) 255 { 256 if (refcount_dec_and_test(&net->count)) 257 __put_net(net); 258 } 259 260 static inline 261 int net_eq(const struct net *net1, const struct net *net2) 262 { 263 return net1 == net2; 264 } 265 266 static inline int check_net(const struct net *net) 267 { 268 return refcount_read(&net->count) != 0; 269 } 270 271 void net_drop_ns(void *); 272 273 #else 274 275 static inline struct net *get_net(struct net *net) 276 { 277 return net; 278 } 279 280 static inline void put_net(struct net *net) 281 { 282 } 283 284 static inline struct net *maybe_get_net(struct net *net) 285 { 286 return net; 287 } 288 289 static inline 290 int net_eq(const struct net *net1, const struct net *net2) 291 { 292 return 1; 293 } 294 295 static inline int check_net(const struct net *net) 296 { 297 return 1; 298 } 299 300 #define net_drop_ns NULL 301 #endif 302 303 304 typedef struct { 305 #ifdef CONFIG_NET_NS 306 struct net *net; 307 #endif 308 } possible_net_t; 309 310 static inline void write_pnet(possible_net_t *pnet, struct net *net) 311 { 312 #ifdef CONFIG_NET_NS 313 pnet->net = net; 314 #endif 315 } 316 317 static inline struct net *read_pnet(const possible_net_t *pnet) 318 { 319 #ifdef CONFIG_NET_NS 320 return pnet->net; 321 #else 322 return &init_net; 323 #endif 324 } 325 326 /* Protected by net_rwsem */ 327 #define for_each_net(VAR) \ 328 list_for_each_entry(VAR, &net_namespace_list, list) 329 330 #define for_each_net_rcu(VAR) \ 331 list_for_each_entry_rcu(VAR, &net_namespace_list, list) 332 333 #ifdef CONFIG_NET_NS 334 #define __net_init 335 #define __net_exit 336 #define __net_initdata 337 #define __net_initconst 338 #else 339 #define __net_init __init 340 #define __net_exit __ref 341 #define __net_initdata __initdata 342 #define __net_initconst __initconst 343 #endif 344 345 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp); 346 int peernet2id(struct net *net, struct net *peer); 347 bool peernet_has_id(struct net *net, struct net *peer); 348 struct net *get_net_ns_by_id(struct net *net, int id); 349 350 struct pernet_operations { 351 struct list_head list; 352 /* 353 * Below methods are called without any exclusive locks. 354 * More than one net may be constructed and destructed 355 * in parallel on several cpus. Every pernet_operations 356 * have to keep in mind all other pernet_operations and 357 * to introduce a locking, if they share common resources. 358 * 359 * The only time they are called with exclusive lock is 360 * from register_pernet_subsys(), unregister_pernet_subsys() 361 * register_pernet_device() and unregister_pernet_device(). 362 * 363 * Exit methods using blocking RCU primitives, such as 364 * synchronize_rcu(), should be implemented via exit_batch. 365 * Then, destruction of a group of net requires single 366 * synchronize_rcu() related to these pernet_operations, 367 * instead of separate synchronize_rcu() for every net. 368 * Please, avoid synchronize_rcu() at all, where it's possible. 369 * 370 * Note that a combination of pre_exit() and exit() can 371 * be used, since a synchronize_rcu() is guaranteed between 372 * the calls. 373 */ 374 int (*init)(struct net *net); 375 void (*pre_exit)(struct net *net); 376 void (*exit)(struct net *net); 377 void (*exit_batch)(struct list_head *net_exit_list); 378 unsigned int *id; 379 size_t size; 380 }; 381 382 /* 383 * Use these carefully. If you implement a network device and it 384 * needs per network namespace operations use device pernet operations, 385 * otherwise use pernet subsys operations. 386 * 387 * Network interfaces need to be removed from a dying netns _before_ 388 * subsys notifiers can be called, as most of the network code cleanup 389 * (which is done from subsys notifiers) runs with the assumption that 390 * dev_remove_pack has been called so no new packets will arrive during 391 * and after the cleanup functions have been called. dev_remove_pack 392 * is not per namespace so instead the guarantee of no more packets 393 * arriving in a network namespace is provided by ensuring that all 394 * network devices and all sockets have left the network namespace 395 * before the cleanup methods are called. 396 * 397 * For the longest time the ipv4 icmp code was registered as a pernet 398 * device which caused kernel oops, and panics during network 399 * namespace cleanup. So please don't get this wrong. 400 */ 401 int register_pernet_subsys(struct pernet_operations *); 402 void unregister_pernet_subsys(struct pernet_operations *); 403 int register_pernet_device(struct pernet_operations *); 404 void unregister_pernet_device(struct pernet_operations *); 405 406 struct ctl_table; 407 struct ctl_table_header; 408 409 #ifdef CONFIG_SYSCTL 410 int net_sysctl_init(void); 411 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path, 412 struct ctl_table *table); 413 void unregister_net_sysctl_table(struct ctl_table_header *header); 414 #else 415 static inline int net_sysctl_init(void) { return 0; } 416 static inline struct ctl_table_header *register_net_sysctl(struct net *net, 417 const char *path, struct ctl_table *table) 418 { 419 return NULL; 420 } 421 static inline void unregister_net_sysctl_table(struct ctl_table_header *header) 422 { 423 } 424 #endif 425 426 static inline int rt_genid_ipv4(struct net *net) 427 { 428 return atomic_read(&net->ipv4.rt_genid); 429 } 430 431 static inline void rt_genid_bump_ipv4(struct net *net) 432 { 433 atomic_inc(&net->ipv4.rt_genid); 434 } 435 436 extern void (*__fib6_flush_trees)(struct net *net); 437 static inline void rt_genid_bump_ipv6(struct net *net) 438 { 439 if (__fib6_flush_trees) 440 __fib6_flush_trees(net); 441 } 442 443 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN) 444 static inline struct netns_ieee802154_lowpan * 445 net_ieee802154_lowpan(struct net *net) 446 { 447 return &net->ieee802154_lowpan; 448 } 449 #endif 450 451 /* For callers who don't really care about whether it's IPv4 or IPv6 */ 452 static inline void rt_genid_bump_all(struct net *net) 453 { 454 rt_genid_bump_ipv4(net); 455 rt_genid_bump_ipv6(net); 456 } 457 458 static inline int fnhe_genid(struct net *net) 459 { 460 return atomic_read(&net->fnhe_genid); 461 } 462 463 static inline void fnhe_genid_bump(struct net *net) 464 { 465 atomic_inc(&net->fnhe_genid); 466 } 467 468 #endif /* __NET_NET_NAMESPACE_H */ 469