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 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 27 #include <net/netns/conntrack.h> 28 #endif 29 #if IS_ENABLED(CONFIG_NF_FLOW_TABLE) 30 #include <net/netns/flow_table.h> 31 #endif 32 #include <net/netns/nftables.h> 33 #include <net/netns/xfrm.h> 34 #include <net/netns/mpls.h> 35 #include <net/netns/can.h> 36 #include <net/netns/xdp.h> 37 #include <net/netns/smc.h> 38 #include <net/netns/bpf.h> 39 #include <net/netns/mctp.h> 40 #include <net/net_trackers.h> 41 #include <linux/ns_common.h> 42 #include <linux/idr.h> 43 #include <linux/skbuff.h> 44 #include <linux/notifier.h> 45 #include <linux/xarray.h> 46 47 struct user_namespace; 48 struct proc_dir_entry; 49 struct net_device; 50 struct sock; 51 struct ctl_table_header; 52 struct net_generic; 53 struct uevent_sock; 54 struct netns_ipvs; 55 struct bpf_prog; 56 57 58 #define NETDEV_HASHBITS 8 59 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS) 60 61 struct net { 62 /* First cache line can be often dirtied. 63 * Do not place here read-mostly fields. 64 */ 65 refcount_t passive; /* To decide when the network 66 * namespace should be freed. 67 */ 68 spinlock_t rules_mod_lock; 69 70 unsigned int dev_base_seq; /* protected by rtnl_mutex */ 71 u32 ifindex; 72 73 spinlock_t nsid_lock; 74 atomic_t fnhe_genid; 75 76 struct list_head list; /* list of network namespaces */ 77 struct list_head exit_list; /* To linked to call pernet exit 78 * methods on dead net ( 79 * pernet_ops_rwsem read locked), 80 * or to unregister pernet ops 81 * (pernet_ops_rwsem write locked). 82 */ 83 struct llist_node defer_free_list; 84 struct llist_node cleanup_list; /* namespaces on death row */ 85 86 #ifdef CONFIG_KEYS 87 struct key_tag *key_domain; /* Key domain of operation tag */ 88 #endif 89 struct user_namespace *user_ns; /* Owning user namespace */ 90 struct ucounts *ucounts; 91 struct idr netns_ids; 92 93 struct ns_common ns; 94 struct ref_tracker_dir refcnt_tracker; 95 struct ref_tracker_dir notrefcnt_tracker; /* tracker for objects not 96 * refcounted against netns 97 */ 98 struct list_head dev_base_head; 99 struct proc_dir_entry *proc_net; 100 struct proc_dir_entry *proc_net_stat; 101 102 #ifdef CONFIG_SYSCTL 103 struct ctl_table_set sysctls; 104 #endif 105 106 struct sock *rtnl; /* rtnetlink socket */ 107 struct sock *genl_sock; 108 109 struct uevent_sock *uevent_sock; /* uevent socket */ 110 111 struct hlist_head *dev_name_head; 112 struct hlist_head *dev_index_head; 113 struct xarray dev_by_index; 114 struct raw_notifier_head netdev_chain; 115 116 /* Note that @hash_mix can be read millions times per second, 117 * it is critical that it is on a read_mostly cache line. 118 */ 119 u32 hash_mix; 120 121 struct net_device *loopback_dev; /* The loopback */ 122 123 /* core fib_rules */ 124 struct list_head rules_ops; 125 126 struct netns_core core; 127 struct netns_mib mib; 128 struct netns_packet packet; 129 #if IS_ENABLED(CONFIG_UNIX) 130 struct netns_unix unx; 131 #endif 132 struct netns_nexthop nexthop; 133 struct netns_ipv4 ipv4; 134 #if IS_ENABLED(CONFIG_IPV6) 135 struct netns_ipv6 ipv6; 136 #endif 137 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN) 138 struct netns_ieee802154_lowpan ieee802154_lowpan; 139 #endif 140 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE) 141 struct netns_sctp sctp; 142 #endif 143 #ifdef CONFIG_NETFILTER 144 struct netns_nf nf; 145 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 146 struct netns_ct ct; 147 #endif 148 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE) 149 struct netns_nftables nft; 150 #endif 151 #if IS_ENABLED(CONFIG_NF_FLOW_TABLE) 152 struct netns_ft ft; 153 #endif 154 #endif 155 #ifdef CONFIG_WEXT_CORE 156 struct sk_buff_head wext_nlevents; 157 #endif 158 struct net_generic __rcu *gen; 159 160 /* Used to store attached BPF programs */ 161 struct netns_bpf bpf; 162 163 /* Note : following structs are cache line aligned */ 164 #ifdef CONFIG_XFRM 165 struct netns_xfrm xfrm; 166 #endif 167 168 u64 net_cookie; /* written once */ 169 170 #if IS_ENABLED(CONFIG_IP_VS) 171 struct netns_ipvs *ipvs; 172 #endif 173 #if IS_ENABLED(CONFIG_MPLS) 174 struct netns_mpls mpls; 175 #endif 176 #if IS_ENABLED(CONFIG_CAN) 177 struct netns_can can; 178 #endif 179 #ifdef CONFIG_XDP_SOCKETS 180 struct netns_xdp xdp; 181 #endif 182 #if IS_ENABLED(CONFIG_MCTP) 183 struct netns_mctp mctp; 184 #endif 185 #if IS_ENABLED(CONFIG_CRYPTO_USER) 186 struct sock *crypto_nlsk; 187 #endif 188 struct sock *diag_nlsk; 189 #if IS_ENABLED(CONFIG_SMC) 190 struct netns_smc smc; 191 #endif 192 #ifdef CONFIG_DEBUG_NET_SMALL_RTNL 193 /* Move to a better place when the config guard is removed. */ 194 struct mutex rtnl_mutex; 195 #endif 196 } __randomize_layout; 197 198 #include <linux/seq_file_net.h> 199 200 /* Init's network namespace */ 201 extern struct net init_net; 202 203 #ifdef CONFIG_NET_NS 204 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns, 205 struct net *old_net); 206 207 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid); 208 209 void net_ns_barrier(void); 210 211 struct ns_common *get_net_ns(struct ns_common *ns); 212 struct net *get_net_ns_by_fd(int fd); 213 #else /* CONFIG_NET_NS */ 214 #include <linux/sched.h> 215 #include <linux/nsproxy.h> 216 static inline struct net *copy_net_ns(unsigned long flags, 217 struct user_namespace *user_ns, struct net *old_net) 218 { 219 if (flags & CLONE_NEWNET) 220 return ERR_PTR(-EINVAL); 221 return old_net; 222 } 223 224 static inline void net_ns_get_ownership(const struct net *net, 225 kuid_t *uid, kgid_t *gid) 226 { 227 *uid = GLOBAL_ROOT_UID; 228 *gid = GLOBAL_ROOT_GID; 229 } 230 231 static inline void net_ns_barrier(void) {} 232 233 static inline struct ns_common *get_net_ns(struct ns_common *ns) 234 { 235 return ERR_PTR(-EINVAL); 236 } 237 238 static inline struct net *get_net_ns_by_fd(int fd) 239 { 240 return ERR_PTR(-EINVAL); 241 } 242 #endif /* CONFIG_NET_NS */ 243 244 245 extern struct list_head net_namespace_list; 246 247 struct net *get_net_ns_by_pid(pid_t pid); 248 249 #ifdef CONFIG_SYSCTL 250 void ipx_register_sysctl(void); 251 void ipx_unregister_sysctl(void); 252 #else 253 #define ipx_register_sysctl() 254 #define ipx_unregister_sysctl() 255 #endif 256 257 #ifdef CONFIG_NET_NS 258 void __put_net(struct net *net); 259 260 /* Try using get_net_track() instead */ 261 static inline struct net *get_net(struct net *net) 262 { 263 refcount_inc(&net->ns.count); 264 return net; 265 } 266 267 static inline struct net *maybe_get_net(struct net *net) 268 { 269 /* Used when we know struct net exists but we 270 * aren't guaranteed a previous reference count 271 * exists. If the reference count is zero this 272 * function fails and returns NULL. 273 */ 274 if (!refcount_inc_not_zero(&net->ns.count)) 275 net = NULL; 276 return net; 277 } 278 279 /* Try using put_net_track() instead */ 280 static inline void put_net(struct net *net) 281 { 282 if (refcount_dec_and_test(&net->ns.count)) 283 __put_net(net); 284 } 285 286 static inline 287 int net_eq(const struct net *net1, const struct net *net2) 288 { 289 return net1 == net2; 290 } 291 292 static inline int check_net(const struct net *net) 293 { 294 return refcount_read(&net->ns.count) != 0; 295 } 296 297 void net_drop_ns(void *); 298 299 #else 300 301 static inline struct net *get_net(struct net *net) 302 { 303 return net; 304 } 305 306 static inline void put_net(struct net *net) 307 { 308 } 309 310 static inline struct net *maybe_get_net(struct net *net) 311 { 312 return net; 313 } 314 315 static inline 316 int net_eq(const struct net *net1, const struct net *net2) 317 { 318 return 1; 319 } 320 321 static inline int check_net(const struct net *net) 322 { 323 return 1; 324 } 325 326 #define net_drop_ns NULL 327 #endif 328 329 /* Returns true if the netns initialization is completed successfully */ 330 static inline bool net_initialized(const struct net *net) 331 { 332 return READ_ONCE(net->list.next); 333 } 334 335 static inline void __netns_tracker_alloc(struct net *net, 336 netns_tracker *tracker, 337 bool refcounted, 338 gfp_t gfp) 339 { 340 #ifdef CONFIG_NET_NS_REFCNT_TRACKER 341 ref_tracker_alloc(refcounted ? &net->refcnt_tracker : 342 &net->notrefcnt_tracker, 343 tracker, gfp); 344 #endif 345 } 346 347 static inline void netns_tracker_alloc(struct net *net, netns_tracker *tracker, 348 gfp_t gfp) 349 { 350 __netns_tracker_alloc(net, tracker, true, gfp); 351 } 352 353 static inline void __netns_tracker_free(struct net *net, 354 netns_tracker *tracker, 355 bool refcounted) 356 { 357 #ifdef CONFIG_NET_NS_REFCNT_TRACKER 358 ref_tracker_free(refcounted ? &net->refcnt_tracker : 359 &net->notrefcnt_tracker, tracker); 360 #endif 361 } 362 363 static inline struct net *get_net_track(struct net *net, 364 netns_tracker *tracker, gfp_t gfp) 365 { 366 get_net(net); 367 netns_tracker_alloc(net, tracker, gfp); 368 return net; 369 } 370 371 static inline void put_net_track(struct net *net, netns_tracker *tracker) 372 { 373 __netns_tracker_free(net, tracker, true); 374 put_net(net); 375 } 376 377 typedef struct { 378 #ifdef CONFIG_NET_NS 379 struct net __rcu *net; 380 #endif 381 } possible_net_t; 382 383 static inline void write_pnet(possible_net_t *pnet, struct net *net) 384 { 385 #ifdef CONFIG_NET_NS 386 rcu_assign_pointer(pnet->net, net); 387 #endif 388 } 389 390 static inline struct net *read_pnet(const possible_net_t *pnet) 391 { 392 #ifdef CONFIG_NET_NS 393 return rcu_dereference_protected(pnet->net, true); 394 #else 395 return &init_net; 396 #endif 397 } 398 399 static inline struct net *read_pnet_rcu(possible_net_t *pnet) 400 { 401 #ifdef CONFIG_NET_NS 402 return rcu_dereference(pnet->net); 403 #else 404 return &init_net; 405 #endif 406 } 407 408 /* Protected by net_rwsem */ 409 #define for_each_net(VAR) \ 410 list_for_each_entry(VAR, &net_namespace_list, list) 411 #define for_each_net_continue_reverse(VAR) \ 412 list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list) 413 #define for_each_net_rcu(VAR) \ 414 list_for_each_entry_rcu(VAR, &net_namespace_list, list) 415 416 #ifdef CONFIG_NET_NS 417 #define __net_init 418 #define __net_exit 419 #define __net_initdata 420 #define __net_initconst 421 #else 422 #define __net_init __init 423 #define __net_exit __ref 424 #define __net_initdata __initdata 425 #define __net_initconst __initconst 426 #endif 427 428 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp); 429 int peernet2id(const struct net *net, struct net *peer); 430 bool peernet_has_id(const struct net *net, struct net *peer); 431 struct net *get_net_ns_by_id(const struct net *net, int id); 432 433 struct pernet_operations { 434 struct list_head list; 435 /* 436 * Below methods are called without any exclusive locks. 437 * More than one net may be constructed and destructed 438 * in parallel on several cpus. Every pernet_operations 439 * have to keep in mind all other pernet_operations and 440 * to introduce a locking, if they share common resources. 441 * 442 * The only time they are called with exclusive lock is 443 * from register_pernet_subsys(), unregister_pernet_subsys() 444 * register_pernet_device() and unregister_pernet_device(). 445 * 446 * Exit methods using blocking RCU primitives, such as 447 * synchronize_rcu(), should be implemented via exit_batch. 448 * Then, destruction of a group of net requires single 449 * synchronize_rcu() related to these pernet_operations, 450 * instead of separate synchronize_rcu() for every net. 451 * Please, avoid synchronize_rcu() at all, where it's possible. 452 * 453 * Note that a combination of pre_exit() and exit() can 454 * be used, since a synchronize_rcu() is guaranteed between 455 * the calls. 456 */ 457 int (*init)(struct net *net); 458 void (*pre_exit)(struct net *net); 459 void (*exit)(struct net *net); 460 void (*exit_batch)(struct list_head *net_exit_list); 461 /* Following method is called with RTNL held. */ 462 void (*exit_batch_rtnl)(struct list_head *net_exit_list, 463 struct list_head *dev_kill_list); 464 unsigned int * const id; 465 const size_t size; 466 }; 467 468 /* 469 * Use these carefully. If you implement a network device and it 470 * needs per network namespace operations use device pernet operations, 471 * otherwise use pernet subsys operations. 472 * 473 * Network interfaces need to be removed from a dying netns _before_ 474 * subsys notifiers can be called, as most of the network code cleanup 475 * (which is done from subsys notifiers) runs with the assumption that 476 * dev_remove_pack has been called so no new packets will arrive during 477 * and after the cleanup functions have been called. dev_remove_pack 478 * is not per namespace so instead the guarantee of no more packets 479 * arriving in a network namespace is provided by ensuring that all 480 * network devices and all sockets have left the network namespace 481 * before the cleanup methods are called. 482 * 483 * For the longest time the ipv4 icmp code was registered as a pernet 484 * device which caused kernel oops, and panics during network 485 * namespace cleanup. So please don't get this wrong. 486 */ 487 int register_pernet_subsys(struct pernet_operations *); 488 void unregister_pernet_subsys(struct pernet_operations *); 489 int register_pernet_device(struct pernet_operations *); 490 void unregister_pernet_device(struct pernet_operations *); 491 492 struct ctl_table; 493 494 #define register_net_sysctl(net, path, table) \ 495 register_net_sysctl_sz(net, path, table, ARRAY_SIZE(table)) 496 #ifdef CONFIG_SYSCTL 497 int net_sysctl_init(void); 498 struct ctl_table_header *register_net_sysctl_sz(struct net *net, const char *path, 499 struct ctl_table *table, size_t table_size); 500 void unregister_net_sysctl_table(struct ctl_table_header *header); 501 #else 502 static inline int net_sysctl_init(void) { return 0; } 503 static inline struct ctl_table_header *register_net_sysctl_sz(struct net *net, 504 const char *path, struct ctl_table *table, size_t table_size) 505 { 506 return NULL; 507 } 508 static inline void unregister_net_sysctl_table(struct ctl_table_header *header) 509 { 510 } 511 #endif 512 513 static inline int rt_genid_ipv4(const struct net *net) 514 { 515 return atomic_read(&net->ipv4.rt_genid); 516 } 517 518 #if IS_ENABLED(CONFIG_IPV6) 519 static inline int rt_genid_ipv6(const struct net *net) 520 { 521 return atomic_read(&net->ipv6.fib6_sernum); 522 } 523 #endif 524 525 static inline void rt_genid_bump_ipv4(struct net *net) 526 { 527 atomic_inc(&net->ipv4.rt_genid); 528 } 529 530 extern void (*__fib6_flush_trees)(struct net *net); 531 static inline void rt_genid_bump_ipv6(struct net *net) 532 { 533 if (__fib6_flush_trees) 534 __fib6_flush_trees(net); 535 } 536 537 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN) 538 static inline struct netns_ieee802154_lowpan * 539 net_ieee802154_lowpan(struct net *net) 540 { 541 return &net->ieee802154_lowpan; 542 } 543 #endif 544 545 /* For callers who don't really care about whether it's IPv4 or IPv6 */ 546 static inline void rt_genid_bump_all(struct net *net) 547 { 548 rt_genid_bump_ipv4(net); 549 rt_genid_bump_ipv6(net); 550 } 551 552 static inline int fnhe_genid(const struct net *net) 553 { 554 return atomic_read(&net->fnhe_genid); 555 } 556 557 static inline void fnhe_genid_bump(struct net *net) 558 { 559 atomic_inc(&net->fnhe_genid); 560 } 561 562 #ifdef CONFIG_NET 563 void net_ns_init(void); 564 #else 565 static inline void net_ns_init(void) {} 566 #endif 567 568 #endif /* __NET_NET_NAMESPACE_H */ 569