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