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 /* 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 refcount_t count; /* To decided when the network 63 * namespace should be shut down. 64 */ 65 spinlock_t rules_mod_lock; 66 67 unsigned int dev_unreg_count; 68 69 unsigned int dev_base_seq; /* protected by rtnl_mutex */ 70 int ifindex; 71 72 spinlock_t nsid_lock; 73 atomic_t fnhe_genid; 74 75 struct list_head list; /* list of network namespaces */ 76 struct list_head exit_list; /* To linked to call pernet exit 77 * methods on dead net ( 78 * pernet_ops_rwsem read locked), 79 * or to unregister pernet ops 80 * (pernet_ops_rwsem write locked). 81 */ 82 struct llist_node cleanup_list; /* namespaces on death row */ 83 84 #ifdef CONFIG_KEYS 85 struct key_tag *key_domain; /* Key domain of operation tag */ 86 #endif 87 struct user_namespace *user_ns; /* Owning user namespace */ 88 struct ucounts *ucounts; 89 struct idr netns_ids; 90 91 struct ns_common ns; 92 93 struct list_head dev_base_head; 94 struct proc_dir_entry *proc_net; 95 struct proc_dir_entry *proc_net_stat; 96 97 #ifdef CONFIG_SYSCTL 98 struct ctl_table_set sysctls; 99 #endif 100 101 struct sock *rtnl; /* rtnetlink socket */ 102 struct sock *genl_sock; 103 104 struct uevent_sock *uevent_sock; /* uevent socket */ 105 106 struct hlist_head *dev_name_head; 107 struct hlist_head *dev_index_head; 108 struct raw_notifier_head netdev_chain; 109 110 /* Note that @hash_mix can be read millions times per second, 111 * it is critical that it is on a read_mostly cache line. 112 */ 113 u32 hash_mix; 114 115 struct net_device *loopback_dev; /* The loopback */ 116 117 /* core fib_rules */ 118 struct list_head rules_ops; 119 120 struct netns_core core; 121 struct netns_mib mib; 122 struct netns_packet packet; 123 struct netns_unix unx; 124 struct netns_nexthop nexthop; 125 struct netns_ipv4 ipv4; 126 #if IS_ENABLED(CONFIG_IPV6) 127 struct netns_ipv6 ipv6; 128 #endif 129 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN) 130 struct netns_ieee802154_lowpan ieee802154_lowpan; 131 #endif 132 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE) 133 struct netns_sctp sctp; 134 #endif 135 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE) 136 struct netns_dccp dccp; 137 #endif 138 #ifdef CONFIG_NETFILTER 139 struct netns_nf nf; 140 struct netns_xt xt; 141 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 142 struct netns_ct ct; 143 #endif 144 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE) 145 struct netns_nftables nft; 146 #endif 147 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6) 148 struct netns_nf_frag nf_frag; 149 struct ctl_table_header *nf_frag_frags_hdr; 150 #endif 151 struct sock *nfnl; 152 struct sock *nfnl_stash; 153 #if IS_ENABLED(CONFIG_NETFILTER_NETLINK_ACCT) 154 struct list_head nfnl_acct_list; 155 #endif 156 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT) 157 struct list_head nfct_timeout_list; 158 #endif 159 #endif 160 #ifdef CONFIG_WEXT_CORE 161 struct sk_buff_head wext_nlevents; 162 #endif 163 struct net_generic __rcu *gen; 164 165 struct bpf_prog __rcu *flow_dissector_prog; 166 167 /* Note : following structs are cache line aligned */ 168 #ifdef CONFIG_XFRM 169 struct netns_xfrm xfrm; 170 #endif 171 172 atomic64_t net_cookie; /* written once */ 173 174 #if IS_ENABLED(CONFIG_IP_VS) 175 struct netns_ipvs *ipvs; 176 #endif 177 #if IS_ENABLED(CONFIG_MPLS) 178 struct netns_mpls mpls; 179 #endif 180 #if IS_ENABLED(CONFIG_CAN) 181 struct netns_can can; 182 #endif 183 #ifdef CONFIG_XDP_SOCKETS 184 struct netns_xdp xdp; 185 #endif 186 #if IS_ENABLED(CONFIG_CRYPTO_USER) 187 struct sock *crypto_nlsk; 188 #endif 189 struct sock *diag_nlsk; 190 } __randomize_layout; 191 192 #include <linux/seq_file_net.h> 193 194 /* Init's network namespace */ 195 extern struct net init_net; 196 197 #ifdef CONFIG_NET_NS 198 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns, 199 struct net *old_net); 200 201 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid); 202 203 void net_ns_barrier(void); 204 #else /* CONFIG_NET_NS */ 205 #include <linux/sched.h> 206 #include <linux/nsproxy.h> 207 static inline struct net *copy_net_ns(unsigned long flags, 208 struct user_namespace *user_ns, struct net *old_net) 209 { 210 if (flags & CLONE_NEWNET) 211 return ERR_PTR(-EINVAL); 212 return old_net; 213 } 214 215 static inline void net_ns_get_ownership(const struct net *net, 216 kuid_t *uid, kgid_t *gid) 217 { 218 *uid = GLOBAL_ROOT_UID; 219 *gid = GLOBAL_ROOT_GID; 220 } 221 222 static inline void net_ns_barrier(void) {} 223 #endif /* CONFIG_NET_NS */ 224 225 226 extern struct list_head net_namespace_list; 227 228 struct net *get_net_ns_by_pid(pid_t pid); 229 struct net *get_net_ns_by_fd(int fd); 230 231 u64 net_gen_cookie(struct net *net); 232 233 #ifdef CONFIG_SYSCTL 234 void ipx_register_sysctl(void); 235 void ipx_unregister_sysctl(void); 236 #else 237 #define ipx_register_sysctl() 238 #define ipx_unregister_sysctl() 239 #endif 240 241 #ifdef CONFIG_NET_NS 242 void __put_net(struct net *net); 243 244 static inline struct net *get_net(struct net *net) 245 { 246 refcount_inc(&net->count); 247 return net; 248 } 249 250 static inline struct net *maybe_get_net(struct net *net) 251 { 252 /* Used when we know struct net exists but we 253 * aren't guaranteed a previous reference count 254 * exists. If the reference count is zero this 255 * function fails and returns NULL. 256 */ 257 if (!refcount_inc_not_zero(&net->count)) 258 net = NULL; 259 return net; 260 } 261 262 static inline void put_net(struct net *net) 263 { 264 if (refcount_dec_and_test(&net->count)) 265 __put_net(net); 266 } 267 268 static inline 269 int net_eq(const struct net *net1, const struct net *net2) 270 { 271 return net1 == net2; 272 } 273 274 static inline int check_net(const struct net *net) 275 { 276 return refcount_read(&net->count) != 0; 277 } 278 279 void net_drop_ns(void *); 280 281 #else 282 283 static inline struct net *get_net(struct net *net) 284 { 285 return net; 286 } 287 288 static inline void put_net(struct net *net) 289 { 290 } 291 292 static inline struct net *maybe_get_net(struct net *net) 293 { 294 return net; 295 } 296 297 static inline 298 int net_eq(const struct net *net1, const struct net *net2) 299 { 300 return 1; 301 } 302 303 static inline int check_net(const struct net *net) 304 { 305 return 1; 306 } 307 308 #define net_drop_ns NULL 309 #endif 310 311 312 typedef struct { 313 #ifdef CONFIG_NET_NS 314 struct net *net; 315 #endif 316 } possible_net_t; 317 318 static inline void write_pnet(possible_net_t *pnet, struct net *net) 319 { 320 #ifdef CONFIG_NET_NS 321 pnet->net = net; 322 #endif 323 } 324 325 static inline struct net *read_pnet(const possible_net_t *pnet) 326 { 327 #ifdef CONFIG_NET_NS 328 return pnet->net; 329 #else 330 return &init_net; 331 #endif 332 } 333 334 /* Protected by net_rwsem */ 335 #define for_each_net(VAR) \ 336 list_for_each_entry(VAR, &net_namespace_list, list) 337 #define for_each_net_continue_reverse(VAR) \ 338 list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list) 339 #define for_each_net_rcu(VAR) \ 340 list_for_each_entry_rcu(VAR, &net_namespace_list, list) 341 342 #ifdef CONFIG_NET_NS 343 #define __net_init 344 #define __net_exit 345 #define __net_initdata 346 #define __net_initconst 347 #else 348 #define __net_init __init 349 #define __net_exit __ref 350 #define __net_initdata __initdata 351 #define __net_initconst __initconst 352 #endif 353 354 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp); 355 int peernet2id(const struct net *net, struct net *peer); 356 bool peernet_has_id(const struct net *net, struct net *peer); 357 struct net *get_net_ns_by_id(const struct net *net, int id); 358 359 struct pernet_operations { 360 struct list_head list; 361 /* 362 * Below methods are called without any exclusive locks. 363 * More than one net may be constructed and destructed 364 * in parallel on several cpus. Every pernet_operations 365 * have to keep in mind all other pernet_operations and 366 * to introduce a locking, if they share common resources. 367 * 368 * The only time they are called with exclusive lock is 369 * from register_pernet_subsys(), unregister_pernet_subsys() 370 * register_pernet_device() and unregister_pernet_device(). 371 * 372 * Exit methods using blocking RCU primitives, such as 373 * synchronize_rcu(), should be implemented via exit_batch. 374 * Then, destruction of a group of net requires single 375 * synchronize_rcu() related to these pernet_operations, 376 * instead of separate synchronize_rcu() for every net. 377 * Please, avoid synchronize_rcu() at all, where it's possible. 378 * 379 * Note that a combination of pre_exit() and exit() can 380 * be used, since a synchronize_rcu() is guaranteed between 381 * the calls. 382 */ 383 int (*init)(struct net *net); 384 void (*pre_exit)(struct net *net); 385 void (*exit)(struct net *net); 386 void (*exit_batch)(struct list_head *net_exit_list); 387 unsigned int *id; 388 size_t size; 389 }; 390 391 /* 392 * Use these carefully. If you implement a network device and it 393 * needs per network namespace operations use device pernet operations, 394 * otherwise use pernet subsys operations. 395 * 396 * Network interfaces need to be removed from a dying netns _before_ 397 * subsys notifiers can be called, as most of the network code cleanup 398 * (which is done from subsys notifiers) runs with the assumption that 399 * dev_remove_pack has been called so no new packets will arrive during 400 * and after the cleanup functions have been called. dev_remove_pack 401 * is not per namespace so instead the guarantee of no more packets 402 * arriving in a network namespace is provided by ensuring that all 403 * network devices and all sockets have left the network namespace 404 * before the cleanup methods are called. 405 * 406 * For the longest time the ipv4 icmp code was registered as a pernet 407 * device which caused kernel oops, and panics during network 408 * namespace cleanup. So please don't get this wrong. 409 */ 410 int register_pernet_subsys(struct pernet_operations *); 411 void unregister_pernet_subsys(struct pernet_operations *); 412 int register_pernet_device(struct pernet_operations *); 413 void unregister_pernet_device(struct pernet_operations *); 414 415 struct ctl_table; 416 struct ctl_table_header; 417 418 #ifdef CONFIG_SYSCTL 419 int net_sysctl_init(void); 420 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path, 421 struct ctl_table *table); 422 void unregister_net_sysctl_table(struct ctl_table_header *header); 423 #else 424 static inline int net_sysctl_init(void) { return 0; } 425 static inline struct ctl_table_header *register_net_sysctl(struct net *net, 426 const char *path, struct ctl_table *table) 427 { 428 return NULL; 429 } 430 static inline void unregister_net_sysctl_table(struct ctl_table_header *header) 431 { 432 } 433 #endif 434 435 static inline int rt_genid_ipv4(const struct net *net) 436 { 437 return atomic_read(&net->ipv4.rt_genid); 438 } 439 440 #if IS_ENABLED(CONFIG_IPV6) 441 static inline int rt_genid_ipv6(const struct net *net) 442 { 443 return atomic_read(&net->ipv6.fib6_sernum); 444 } 445 #endif 446 447 static inline void rt_genid_bump_ipv4(struct net *net) 448 { 449 atomic_inc(&net->ipv4.rt_genid); 450 } 451 452 extern void (*__fib6_flush_trees)(struct net *net); 453 static inline void rt_genid_bump_ipv6(struct net *net) 454 { 455 if (__fib6_flush_trees) 456 __fib6_flush_trees(net); 457 } 458 459 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN) 460 static inline struct netns_ieee802154_lowpan * 461 net_ieee802154_lowpan(struct net *net) 462 { 463 return &net->ieee802154_lowpan; 464 } 465 #endif 466 467 /* For callers who don't really care about whether it's IPv4 or IPv6 */ 468 static inline void rt_genid_bump_all(struct net *net) 469 { 470 rt_genid_bump_ipv4(net); 471 rt_genid_bump_ipv6(net); 472 } 473 474 static inline int fnhe_genid(const struct net *net) 475 { 476 return atomic_read(&net->fnhe_genid); 477 } 478 479 static inline void fnhe_genid_bump(struct net *net) 480 { 481 atomic_inc(&net->fnhe_genid); 482 } 483 484 #endif /* __NET_NET_NAMESPACE_H */ 485