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 refcount_t passive; /* To decided when the network 56 * namespace should be freed. 57 */ 58 refcount_t count; /* To decided when the network 59 * namespace should be shut down. 60 */ 61 spinlock_t rules_mod_lock; 62 63 u32 hash_mix; 64 atomic64_t cookie_gen; 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 unsigned int dev_base_seq; /* protected by rtnl_mutex */ 101 int ifindex; 102 unsigned int dev_unreg_count; 103 104 /* core fib_rules */ 105 struct list_head rules_ops; 106 107 struct list_head fib_notifier_ops; /* Populated by 108 * register_pernet_subsys() 109 */ 110 struct net_device *loopback_dev; /* The loopback */ 111 struct netns_core core; 112 struct netns_mib mib; 113 struct netns_packet packet; 114 struct netns_unix unx; 115 struct netns_nexthop nexthop; 116 struct netns_ipv4 ipv4; 117 #if IS_ENABLED(CONFIG_IPV6) 118 struct netns_ipv6 ipv6; 119 #endif 120 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN) 121 struct netns_ieee802154_lowpan ieee802154_lowpan; 122 #endif 123 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE) 124 struct netns_sctp sctp; 125 #endif 126 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE) 127 struct netns_dccp dccp; 128 #endif 129 #ifdef CONFIG_NETFILTER 130 struct netns_nf nf; 131 struct netns_xt xt; 132 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 133 struct netns_ct ct; 134 #endif 135 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE) 136 struct netns_nftables nft; 137 #endif 138 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6) 139 struct netns_nf_frag nf_frag; 140 struct ctl_table_header *nf_frag_frags_hdr; 141 #endif 142 struct sock *nfnl; 143 struct sock *nfnl_stash; 144 #if IS_ENABLED(CONFIG_NETFILTER_NETLINK_ACCT) 145 struct list_head nfnl_acct_list; 146 #endif 147 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT) 148 struct list_head nfct_timeout_list; 149 #endif 150 #endif 151 #ifdef CONFIG_WEXT_CORE 152 struct sk_buff_head wext_nlevents; 153 #endif 154 struct net_generic __rcu *gen; 155 156 struct bpf_prog __rcu *flow_dissector_prog; 157 158 /* Note : following structs are cache line aligned */ 159 #ifdef CONFIG_XFRM 160 struct netns_xfrm xfrm; 161 #endif 162 #if IS_ENABLED(CONFIG_IP_VS) 163 struct netns_ipvs *ipvs; 164 #endif 165 #if IS_ENABLED(CONFIG_MPLS) 166 struct netns_mpls mpls; 167 #endif 168 #if IS_ENABLED(CONFIG_CAN) 169 struct netns_can can; 170 #endif 171 #ifdef CONFIG_XDP_SOCKETS 172 struct netns_xdp xdp; 173 #endif 174 #if IS_ENABLED(CONFIG_CRYPTO_USER) 175 struct sock *crypto_nlsk; 176 #endif 177 struct sock *diag_nlsk; 178 atomic_t fnhe_genid; 179 } __randomize_layout; 180 181 #include <linux/seq_file_net.h> 182 183 /* Init's network namespace */ 184 extern struct net init_net; 185 186 #ifdef CONFIG_NET_NS 187 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns, 188 struct net *old_net); 189 190 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid); 191 192 void net_ns_barrier(void); 193 #else /* CONFIG_NET_NS */ 194 #include <linux/sched.h> 195 #include <linux/nsproxy.h> 196 static inline struct net *copy_net_ns(unsigned long flags, 197 struct user_namespace *user_ns, struct net *old_net) 198 { 199 if (flags & CLONE_NEWNET) 200 return ERR_PTR(-EINVAL); 201 return old_net; 202 } 203 204 static inline void net_ns_get_ownership(const struct net *net, 205 kuid_t *uid, kgid_t *gid) 206 { 207 *uid = GLOBAL_ROOT_UID; 208 *gid = GLOBAL_ROOT_GID; 209 } 210 211 static inline void net_ns_barrier(void) {} 212 #endif /* CONFIG_NET_NS */ 213 214 215 extern struct list_head net_namespace_list; 216 217 struct net *get_net_ns_by_pid(pid_t pid); 218 struct net *get_net_ns_by_fd(int fd); 219 220 #ifdef CONFIG_SYSCTL 221 void ipx_register_sysctl(void); 222 void ipx_unregister_sysctl(void); 223 #else 224 #define ipx_register_sysctl() 225 #define ipx_unregister_sysctl() 226 #endif 227 228 #ifdef CONFIG_NET_NS 229 void __put_net(struct net *net); 230 231 static inline struct net *get_net(struct net *net) 232 { 233 refcount_inc(&net->count); 234 return net; 235 } 236 237 static inline struct net *maybe_get_net(struct net *net) 238 { 239 /* Used when we know struct net exists but we 240 * aren't guaranteed a previous reference count 241 * exists. If the reference count is zero this 242 * function fails and returns NULL. 243 */ 244 if (!refcount_inc_not_zero(&net->count)) 245 net = NULL; 246 return net; 247 } 248 249 static inline void put_net(struct net *net) 250 { 251 if (refcount_dec_and_test(&net->count)) 252 __put_net(net); 253 } 254 255 static inline 256 int net_eq(const struct net *net1, const struct net *net2) 257 { 258 return net1 == net2; 259 } 260 261 static inline int check_net(const struct net *net) 262 { 263 return refcount_read(&net->count) != 0; 264 } 265 266 void net_drop_ns(void *); 267 268 #else 269 270 static inline struct net *get_net(struct net *net) 271 { 272 return net; 273 } 274 275 static inline void put_net(struct net *net) 276 { 277 } 278 279 static inline struct net *maybe_get_net(struct net *net) 280 { 281 return net; 282 } 283 284 static inline 285 int net_eq(const struct net *net1, const struct net *net2) 286 { 287 return 1; 288 } 289 290 static inline int check_net(const struct net *net) 291 { 292 return 1; 293 } 294 295 #define net_drop_ns NULL 296 #endif 297 298 299 typedef struct { 300 #ifdef CONFIG_NET_NS 301 struct net *net; 302 #endif 303 } possible_net_t; 304 305 static inline void write_pnet(possible_net_t *pnet, struct net *net) 306 { 307 #ifdef CONFIG_NET_NS 308 pnet->net = net; 309 #endif 310 } 311 312 static inline struct net *read_pnet(const possible_net_t *pnet) 313 { 314 #ifdef CONFIG_NET_NS 315 return pnet->net; 316 #else 317 return &init_net; 318 #endif 319 } 320 321 /* Protected by net_rwsem */ 322 #define for_each_net(VAR) \ 323 list_for_each_entry(VAR, &net_namespace_list, list) 324 325 #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