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/netfilter.h> 26 #include <net/netns/x_tables.h> 27 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 28 #include <net/netns/conntrack.h> 29 #endif 30 #include <net/netns/nftables.h> 31 #include <net/netns/xfrm.h> 32 #include <net/netns/mpls.h> 33 #include <net/netns/can.h> 34 #include <net/netns/xdp.h> 35 #include <net/netns/bpf.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 /* First cache line can be often dirtied. 57 * Do not place here read-mostly fields. 58 */ 59 refcount_t passive; /* To decide when the network 60 * namespace should be freed. 61 */ 62 spinlock_t rules_mod_lock; 63 64 unsigned int dev_unreg_count; 65 66 unsigned int dev_base_seq; /* protected by rtnl_mutex */ 67 int ifindex; 68 69 spinlock_t nsid_lock; 70 atomic_t fnhe_genid; 71 72 struct list_head list; /* list of network namespaces */ 73 struct list_head exit_list; /* To linked to call pernet exit 74 * methods on dead net ( 75 * pernet_ops_rwsem read locked), 76 * or to unregister pernet ops 77 * (pernet_ops_rwsem write locked). 78 */ 79 struct llist_node cleanup_list; /* namespaces on death row */ 80 81 #ifdef CONFIG_KEYS 82 struct key_tag *key_domain; /* Key domain of operation tag */ 83 #endif 84 struct user_namespace *user_ns; /* Owning user namespace */ 85 struct ucounts *ucounts; 86 struct idr netns_ids; 87 88 struct ns_common ns; 89 90 struct list_head dev_base_head; 91 struct proc_dir_entry *proc_net; 92 struct proc_dir_entry *proc_net_stat; 93 94 #ifdef CONFIG_SYSCTL 95 struct ctl_table_set sysctls; 96 #endif 97 98 struct sock *rtnl; /* rtnetlink socket */ 99 struct sock *genl_sock; 100 101 struct uevent_sock *uevent_sock; /* uevent socket */ 102 103 struct hlist_head *dev_name_head; 104 struct hlist_head *dev_index_head; 105 struct raw_notifier_head netdev_chain; 106 107 /* 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 #ifdef CONFIG_NETFILTER 133 struct netns_nf nf; 134 struct netns_xt xt; 135 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 136 struct netns_ct ct; 137 #endif 138 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE) 139 struct netns_nftables nft; 140 #endif 141 #endif 142 #ifdef CONFIG_WEXT_CORE 143 struct sk_buff_head wext_nlevents; 144 #endif 145 struct net_generic __rcu *gen; 146 147 /* Used to store attached BPF programs */ 148 struct netns_bpf bpf; 149 150 /* Note : following structs are cache line aligned */ 151 #ifdef CONFIG_XFRM 152 struct netns_xfrm xfrm; 153 #endif 154 155 u64 net_cookie; /* written once */ 156 157 #if IS_ENABLED(CONFIG_IP_VS) 158 struct netns_ipvs *ipvs; 159 #endif 160 #if IS_ENABLED(CONFIG_MPLS) 161 struct netns_mpls mpls; 162 #endif 163 #if IS_ENABLED(CONFIG_CAN) 164 struct netns_can can; 165 #endif 166 #ifdef CONFIG_XDP_SOCKETS 167 struct netns_xdp xdp; 168 #endif 169 #if IS_ENABLED(CONFIG_CRYPTO_USER) 170 struct sock *crypto_nlsk; 171 #endif 172 struct sock *diag_nlsk; 173 } __randomize_layout; 174 175 #include <linux/seq_file_net.h> 176 177 /* Init's network namespace */ 178 extern struct net init_net; 179 180 #ifdef CONFIG_NET_NS 181 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns, 182 struct net *old_net); 183 184 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid); 185 186 void net_ns_barrier(void); 187 188 struct ns_common *get_net_ns(struct ns_common *ns); 189 struct net *get_net_ns_by_fd(int fd); 190 #else /* CONFIG_NET_NS */ 191 #include <linux/sched.h> 192 #include <linux/nsproxy.h> 193 static inline struct net *copy_net_ns(unsigned long flags, 194 struct user_namespace *user_ns, struct net *old_net) 195 { 196 if (flags & CLONE_NEWNET) 197 return ERR_PTR(-EINVAL); 198 return old_net; 199 } 200 201 static inline void net_ns_get_ownership(const struct net *net, 202 kuid_t *uid, kgid_t *gid) 203 { 204 *uid = GLOBAL_ROOT_UID; 205 *gid = GLOBAL_ROOT_GID; 206 } 207 208 static inline void net_ns_barrier(void) {} 209 210 static inline struct ns_common *get_net_ns(struct ns_common *ns) 211 { 212 return ERR_PTR(-EINVAL); 213 } 214 215 static inline struct net *get_net_ns_by_fd(int fd) 216 { 217 return ERR_PTR(-EINVAL); 218 } 219 #endif /* CONFIG_NET_NS */ 220 221 222 extern struct list_head net_namespace_list; 223 224 struct net *get_net_ns_by_pid(pid_t pid); 225 226 #ifdef CONFIG_SYSCTL 227 void ipx_register_sysctl(void); 228 void ipx_unregister_sysctl(void); 229 #else 230 #define ipx_register_sysctl() 231 #define ipx_unregister_sysctl() 232 #endif 233 234 #ifdef CONFIG_NET_NS 235 void __put_net(struct net *net); 236 237 static inline struct net *get_net(struct net *net) 238 { 239 refcount_inc(&net->ns.count); 240 return net; 241 } 242 243 static inline struct net *maybe_get_net(struct net *net) 244 { 245 /* Used when we know struct net exists but we 246 * aren't guaranteed a previous reference count 247 * exists. If the reference count is zero this 248 * function fails and returns NULL. 249 */ 250 if (!refcount_inc_not_zero(&net->ns.count)) 251 net = NULL; 252 return net; 253 } 254 255 static inline void put_net(struct net *net) 256 { 257 if (refcount_dec_and_test(&net->ns.count)) 258 __put_net(net); 259 } 260 261 static inline 262 int net_eq(const struct net *net1, const struct net *net2) 263 { 264 return net1 == net2; 265 } 266 267 static inline int check_net(const struct net *net) 268 { 269 return refcount_read(&net->ns.count) != 0; 270 } 271 272 void net_drop_ns(void *); 273 274 #else 275 276 static inline struct net *get_net(struct net *net) 277 { 278 return net; 279 } 280 281 static inline void put_net(struct net *net) 282 { 283 } 284 285 static inline struct net *maybe_get_net(struct net *net) 286 { 287 return net; 288 } 289 290 static inline 291 int net_eq(const struct net *net1, const struct net *net2) 292 { 293 return 1; 294 } 295 296 static inline int check_net(const struct net *net) 297 { 298 return 1; 299 } 300 301 #define net_drop_ns NULL 302 #endif 303 304 305 typedef struct { 306 #ifdef CONFIG_NET_NS 307 struct net *net; 308 #endif 309 } possible_net_t; 310 311 static inline void write_pnet(possible_net_t *pnet, struct net *net) 312 { 313 #ifdef CONFIG_NET_NS 314 pnet->net = net; 315 #endif 316 } 317 318 static inline struct net *read_pnet(const possible_net_t *pnet) 319 { 320 #ifdef CONFIG_NET_NS 321 return pnet->net; 322 #else 323 return &init_net; 324 #endif 325 } 326 327 /* Protected by net_rwsem */ 328 #define for_each_net(VAR) \ 329 list_for_each_entry(VAR, &net_namespace_list, list) 330 #define for_each_net_continue_reverse(VAR) \ 331 list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list) 332 #define for_each_net_rcu(VAR) \ 333 list_for_each_entry_rcu(VAR, &net_namespace_list, list) 334 335 #ifdef CONFIG_NET_NS 336 #define __net_init 337 #define __net_exit 338 #define __net_initdata 339 #define __net_initconst 340 #else 341 #define __net_init __init 342 #define __net_exit __ref 343 #define __net_initdata __initdata 344 #define __net_initconst __initconst 345 #endif 346 347 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp); 348 int peernet2id(const struct net *net, struct net *peer); 349 bool peernet_has_id(const struct net *net, struct net *peer); 350 struct net *get_net_ns_by_id(const struct net *net, int id); 351 352 struct pernet_operations { 353 struct list_head list; 354 /* 355 * Below methods are called without any exclusive locks. 356 * More than one net may be constructed and destructed 357 * in parallel on several cpus. Every pernet_operations 358 * have to keep in mind all other pernet_operations and 359 * to introduce a locking, if they share common resources. 360 * 361 * The only time they are called with exclusive lock is 362 * from register_pernet_subsys(), unregister_pernet_subsys() 363 * register_pernet_device() and unregister_pernet_device(). 364 * 365 * Exit methods using blocking RCU primitives, such as 366 * synchronize_rcu(), should be implemented via exit_batch. 367 * Then, destruction of a group of net requires single 368 * synchronize_rcu() related to these pernet_operations, 369 * instead of separate synchronize_rcu() for every net. 370 * Please, avoid synchronize_rcu() at all, where it's possible. 371 * 372 * Note that a combination of pre_exit() and exit() can 373 * be used, since a synchronize_rcu() is guaranteed between 374 * the calls. 375 */ 376 int (*init)(struct net *net); 377 void (*pre_exit)(struct net *net); 378 void (*exit)(struct net *net); 379 void (*exit_batch)(struct list_head *net_exit_list); 380 unsigned int *id; 381 size_t size; 382 }; 383 384 /* 385 * Use these carefully. If you implement a network device and it 386 * needs per network namespace operations use device pernet operations, 387 * otherwise use pernet subsys operations. 388 * 389 * Network interfaces need to be removed from a dying netns _before_ 390 * subsys notifiers can be called, as most of the network code cleanup 391 * (which is done from subsys notifiers) runs with the assumption that 392 * dev_remove_pack has been called so no new packets will arrive during 393 * and after the cleanup functions have been called. dev_remove_pack 394 * is not per namespace so instead the guarantee of no more packets 395 * arriving in a network namespace is provided by ensuring that all 396 * network devices and all sockets have left the network namespace 397 * before the cleanup methods are called. 398 * 399 * For the longest time the ipv4 icmp code was registered as a pernet 400 * device which caused kernel oops, and panics during network 401 * namespace cleanup. So please don't get this wrong. 402 */ 403 int register_pernet_subsys(struct pernet_operations *); 404 void unregister_pernet_subsys(struct pernet_operations *); 405 int register_pernet_device(struct pernet_operations *); 406 void unregister_pernet_device(struct pernet_operations *); 407 408 struct ctl_table; 409 410 #ifdef CONFIG_SYSCTL 411 int net_sysctl_init(void); 412 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path, 413 struct ctl_table *table); 414 void unregister_net_sysctl_table(struct ctl_table_header *header); 415 #else 416 static inline int net_sysctl_init(void) { return 0; } 417 static inline struct ctl_table_header *register_net_sysctl(struct net *net, 418 const char *path, struct ctl_table *table) 419 { 420 return NULL; 421 } 422 static inline void unregister_net_sysctl_table(struct ctl_table_header *header) 423 { 424 } 425 #endif 426 427 static inline int rt_genid_ipv4(const struct net *net) 428 { 429 return atomic_read(&net->ipv4.rt_genid); 430 } 431 432 #if IS_ENABLED(CONFIG_IPV6) 433 static inline int rt_genid_ipv6(const struct net *net) 434 { 435 return atomic_read(&net->ipv6.fib6_sernum); 436 } 437 #endif 438 439 static inline void rt_genid_bump_ipv4(struct net *net) 440 { 441 atomic_inc(&net->ipv4.rt_genid); 442 } 443 444 extern void (*__fib6_flush_trees)(struct net *net); 445 static inline void rt_genid_bump_ipv6(struct net *net) 446 { 447 if (__fib6_flush_trees) 448 __fib6_flush_trees(net); 449 } 450 451 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN) 452 static inline struct netns_ieee802154_lowpan * 453 net_ieee802154_lowpan(struct net *net) 454 { 455 return &net->ieee802154_lowpan; 456 } 457 #endif 458 459 /* For callers who don't really care about whether it's IPv4 or IPv6 */ 460 static inline void rt_genid_bump_all(struct net *net) 461 { 462 rt_genid_bump_ipv4(net); 463 rt_genid_bump_ipv6(net); 464 } 465 466 static inline int fnhe_genid(const struct net *net) 467 { 468 return atomic_read(&net->fnhe_genid); 469 } 470 471 static inline void fnhe_genid_bump(struct net *net) 472 { 473 atomic_inc(&net->fnhe_genid); 474 } 475 476 #endif /* __NET_NET_NAMESPACE_H */ 477