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 extern struct task_struct *cleanup_net_task; 214 215 #else /* CONFIG_NET_NS */ 216 #include <linux/sched.h> 217 #include <linux/nsproxy.h> 218 static inline struct net *copy_net_ns(unsigned long flags, 219 struct user_namespace *user_ns, struct net *old_net) 220 { 221 if (flags & CLONE_NEWNET) 222 return ERR_PTR(-EINVAL); 223 return old_net; 224 } 225 226 static inline void net_ns_get_ownership(const struct net *net, 227 kuid_t *uid, kgid_t *gid) 228 { 229 *uid = GLOBAL_ROOT_UID; 230 *gid = GLOBAL_ROOT_GID; 231 } 232 233 static inline void net_ns_barrier(void) {} 234 235 static inline struct ns_common *get_net_ns(struct ns_common *ns) 236 { 237 return ERR_PTR(-EINVAL); 238 } 239 240 static inline struct net *get_net_ns_by_fd(int fd) 241 { 242 return ERR_PTR(-EINVAL); 243 } 244 #endif /* CONFIG_NET_NS */ 245 246 247 extern struct list_head net_namespace_list; 248 249 struct net *get_net_ns_by_pid(pid_t pid); 250 251 #ifdef CONFIG_SYSCTL 252 void ipx_register_sysctl(void); 253 void ipx_unregister_sysctl(void); 254 #else 255 #define ipx_register_sysctl() 256 #define ipx_unregister_sysctl() 257 #endif 258 259 #ifdef CONFIG_NET_NS 260 void __put_net(struct net *net); 261 262 /* Try using get_net_track() instead */ 263 static inline struct net *get_net(struct net *net) 264 { 265 refcount_inc(&net->ns.count); 266 return net; 267 } 268 269 static inline struct net *maybe_get_net(struct net *net) 270 { 271 /* Used when we know struct net exists but we 272 * aren't guaranteed a previous reference count 273 * exists. If the reference count is zero this 274 * function fails and returns NULL. 275 */ 276 if (!refcount_inc_not_zero(&net->ns.count)) 277 net = NULL; 278 return net; 279 } 280 281 /* Try using put_net_track() instead */ 282 static inline void put_net(struct net *net) 283 { 284 if (refcount_dec_and_test(&net->ns.count)) 285 __put_net(net); 286 } 287 288 static inline 289 int net_eq(const struct net *net1, const struct net *net2) 290 { 291 return net1 == net2; 292 } 293 294 static inline int check_net(const struct net *net) 295 { 296 return refcount_read(&net->ns.count) != 0; 297 } 298 299 void net_drop_ns(void *); 300 301 #else 302 303 static inline struct net *get_net(struct net *net) 304 { 305 return net; 306 } 307 308 static inline void put_net(struct net *net) 309 { 310 } 311 312 static inline struct net *maybe_get_net(struct net *net) 313 { 314 return net; 315 } 316 317 static inline 318 int net_eq(const struct net *net1, const struct net *net2) 319 { 320 return 1; 321 } 322 323 static inline int check_net(const struct net *net) 324 { 325 return 1; 326 } 327 328 #define net_drop_ns NULL 329 #endif 330 331 /* Returns true if the netns initialization is completed successfully */ 332 static inline bool net_initialized(const struct net *net) 333 { 334 return READ_ONCE(net->list.next); 335 } 336 337 static inline void __netns_tracker_alloc(struct net *net, 338 netns_tracker *tracker, 339 bool refcounted, 340 gfp_t gfp) 341 { 342 #ifdef CONFIG_NET_NS_REFCNT_TRACKER 343 ref_tracker_alloc(refcounted ? &net->refcnt_tracker : 344 &net->notrefcnt_tracker, 345 tracker, gfp); 346 #endif 347 } 348 349 static inline void netns_tracker_alloc(struct net *net, netns_tracker *tracker, 350 gfp_t gfp) 351 { 352 __netns_tracker_alloc(net, tracker, true, gfp); 353 } 354 355 static inline void __netns_tracker_free(struct net *net, 356 netns_tracker *tracker, 357 bool refcounted) 358 { 359 #ifdef CONFIG_NET_NS_REFCNT_TRACKER 360 ref_tracker_free(refcounted ? &net->refcnt_tracker : 361 &net->notrefcnt_tracker, tracker); 362 #endif 363 } 364 365 static inline struct net *get_net_track(struct net *net, 366 netns_tracker *tracker, gfp_t gfp) 367 { 368 get_net(net); 369 netns_tracker_alloc(net, tracker, gfp); 370 return net; 371 } 372 373 static inline void put_net_track(struct net *net, netns_tracker *tracker) 374 { 375 __netns_tracker_free(net, tracker, true); 376 put_net(net); 377 } 378 379 typedef struct { 380 #ifdef CONFIG_NET_NS 381 struct net __rcu *net; 382 #endif 383 } possible_net_t; 384 385 static inline void write_pnet(possible_net_t *pnet, struct net *net) 386 { 387 #ifdef CONFIG_NET_NS 388 rcu_assign_pointer(pnet->net, net); 389 #endif 390 } 391 392 static inline struct net *read_pnet(const possible_net_t *pnet) 393 { 394 #ifdef CONFIG_NET_NS 395 return rcu_dereference_protected(pnet->net, true); 396 #else 397 return &init_net; 398 #endif 399 } 400 401 static inline struct net *read_pnet_rcu(possible_net_t *pnet) 402 { 403 #ifdef CONFIG_NET_NS 404 return rcu_dereference(pnet->net); 405 #else 406 return &init_net; 407 #endif 408 } 409 410 /* Protected by net_rwsem */ 411 #define for_each_net(VAR) \ 412 list_for_each_entry(VAR, &net_namespace_list, list) 413 #define for_each_net_continue_reverse(VAR) \ 414 list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list) 415 #define for_each_net_rcu(VAR) \ 416 list_for_each_entry_rcu(VAR, &net_namespace_list, list) 417 418 #ifdef CONFIG_NET_NS 419 #define __net_init 420 #define __net_exit 421 #define __net_initdata 422 #define __net_initconst 423 #else 424 #define __net_init __init 425 #define __net_exit __ref 426 #define __net_initdata __initdata 427 #define __net_initconst __initconst 428 #endif 429 430 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp); 431 int peernet2id(const struct net *net, struct net *peer); 432 bool peernet_has_id(const struct net *net, struct net *peer); 433 struct net *get_net_ns_by_id(const struct net *net, int id); 434 435 struct pernet_operations { 436 struct list_head list; 437 /* 438 * Below methods are called without any exclusive locks. 439 * More than one net may be constructed and destructed 440 * in parallel on several cpus. Every pernet_operations 441 * have to keep in mind all other pernet_operations and 442 * to introduce a locking, if they share common resources. 443 * 444 * The only time they are called with exclusive lock is 445 * from register_pernet_subsys(), unregister_pernet_subsys() 446 * register_pernet_device() and unregister_pernet_device(). 447 * 448 * Exit methods using blocking RCU primitives, such as 449 * synchronize_rcu(), should be implemented via exit_batch. 450 * Then, destruction of a group of net requires single 451 * synchronize_rcu() related to these pernet_operations, 452 * instead of separate synchronize_rcu() for every net. 453 * Please, avoid synchronize_rcu() at all, where it's possible. 454 * 455 * Note that a combination of pre_exit() and exit() can 456 * be used, since a synchronize_rcu() is guaranteed between 457 * the calls. 458 */ 459 int (*init)(struct net *net); 460 void (*pre_exit)(struct net *net); 461 void (*exit)(struct net *net); 462 void (*exit_batch)(struct list_head *net_exit_list); 463 /* Following method is called with RTNL held. */ 464 void (*exit_batch_rtnl)(struct list_head *net_exit_list, 465 struct list_head *dev_kill_list); 466 unsigned int * const id; 467 const size_t size; 468 }; 469 470 /* 471 * Use these carefully. If you implement a network device and it 472 * needs per network namespace operations use device pernet operations, 473 * otherwise use pernet subsys operations. 474 * 475 * Network interfaces need to be removed from a dying netns _before_ 476 * subsys notifiers can be called, as most of the network code cleanup 477 * (which is done from subsys notifiers) runs with the assumption that 478 * dev_remove_pack has been called so no new packets will arrive during 479 * and after the cleanup functions have been called. dev_remove_pack 480 * is not per namespace so instead the guarantee of no more packets 481 * arriving in a network namespace is provided by ensuring that all 482 * network devices and all sockets have left the network namespace 483 * before the cleanup methods are called. 484 * 485 * For the longest time the ipv4 icmp code was registered as a pernet 486 * device which caused kernel oops, and panics during network 487 * namespace cleanup. So please don't get this wrong. 488 */ 489 int register_pernet_subsys(struct pernet_operations *); 490 void unregister_pernet_subsys(struct pernet_operations *); 491 int register_pernet_device(struct pernet_operations *); 492 void unregister_pernet_device(struct pernet_operations *); 493 494 struct ctl_table; 495 496 #define register_net_sysctl(net, path, table) \ 497 register_net_sysctl_sz(net, path, table, ARRAY_SIZE(table)) 498 #ifdef CONFIG_SYSCTL 499 int net_sysctl_init(void); 500 struct ctl_table_header *register_net_sysctl_sz(struct net *net, const char *path, 501 struct ctl_table *table, size_t table_size); 502 void unregister_net_sysctl_table(struct ctl_table_header *header); 503 #else 504 static inline int net_sysctl_init(void) { return 0; } 505 static inline struct ctl_table_header *register_net_sysctl_sz(struct net *net, 506 const char *path, struct ctl_table *table, size_t table_size) 507 { 508 return NULL; 509 } 510 static inline void unregister_net_sysctl_table(struct ctl_table_header *header) 511 { 512 } 513 #endif 514 515 static inline int rt_genid_ipv4(const struct net *net) 516 { 517 return atomic_read(&net->ipv4.rt_genid); 518 } 519 520 #if IS_ENABLED(CONFIG_IPV6) 521 static inline int rt_genid_ipv6(const struct net *net) 522 { 523 return atomic_read(&net->ipv6.fib6_sernum); 524 } 525 #endif 526 527 static inline void rt_genid_bump_ipv4(struct net *net) 528 { 529 atomic_inc(&net->ipv4.rt_genid); 530 } 531 532 extern void (*__fib6_flush_trees)(struct net *net); 533 static inline void rt_genid_bump_ipv6(struct net *net) 534 { 535 if (__fib6_flush_trees) 536 __fib6_flush_trees(net); 537 } 538 539 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN) 540 static inline struct netns_ieee802154_lowpan * 541 net_ieee802154_lowpan(struct net *net) 542 { 543 return &net->ieee802154_lowpan; 544 } 545 #endif 546 547 /* For callers who don't really care about whether it's IPv4 or IPv6 */ 548 static inline void rt_genid_bump_all(struct net *net) 549 { 550 rt_genid_bump_ipv4(net); 551 rt_genid_bump_ipv6(net); 552 } 553 554 static inline int fnhe_genid(const struct net *net) 555 { 556 return atomic_read(&net->fnhe_genid); 557 } 558 559 static inline void fnhe_genid_bump(struct net *net) 560 { 561 atomic_inc(&net->fnhe_genid); 562 } 563 564 #ifdef CONFIG_NET 565 void net_ns_init(void); 566 #else 567 static inline void net_ns_init(void) {} 568 #endif 569 570 #endif /* __NET_NET_NAMESPACE_H */ 571