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/netns/vsock.h> 41 #include <net/net_trackers.h> 42 #include <linux/ns_common.h> 43 #include <linux/idr.h> 44 #include <linux/skbuff.h> 45 #include <linux/notifier.h> 46 #include <linux/xarray.h> 47 48 struct user_namespace; 49 struct proc_dir_entry; 50 struct net_device; 51 struct sock; 52 struct ctl_table_header; 53 struct net_generic; 54 struct uevent_sock; 55 struct netns_ipvs; 56 struct bpf_prog; 57 58 59 #define NETDEV_HASHBITS 8 60 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS) 61 62 struct net { 63 /* First cache line can be often dirtied. 64 * Do not place here read-mostly fields. 65 */ 66 refcount_t passive; /* To decide when the network 67 * namespace should be freed. 68 */ 69 spinlock_t rules_mod_lock; 70 71 unsigned int dev_base_seq; /* protected by rtnl_mutex */ 72 u32 ifindex; 73 74 spinlock_t nsid_lock; 75 atomic_t fnhe_genid; 76 77 struct list_head list; /* list of network namespaces */ 78 struct list_head exit_list; /* To linked to call pernet exit 79 * methods on dead net ( 80 * pernet_ops_rwsem read locked), 81 * or to unregister pernet ops 82 * (pernet_ops_rwsem write locked). 83 */ 84 struct llist_node defer_free_list; 85 struct llist_node cleanup_list; /* namespaces on death row */ 86 87 struct list_head ptype_all; 88 struct list_head ptype_specific; 89 90 #ifdef CONFIG_KEYS 91 struct key_tag *key_domain; /* Key domain of operation tag */ 92 #endif 93 struct user_namespace *user_ns; /* Owning user namespace */ 94 struct ucounts *ucounts; 95 struct idr netns_ids; 96 97 struct ns_common ns; 98 struct ref_tracker_dir refcnt_tracker; 99 struct ref_tracker_dir notrefcnt_tracker; /* tracker for objects not 100 * refcounted against netns 101 */ 102 struct list_head dev_base_head; 103 struct proc_dir_entry *proc_net; 104 struct proc_dir_entry *proc_net_stat; 105 106 #ifdef CONFIG_SYSCTL 107 struct ctl_table_set sysctls; 108 #endif 109 110 struct sock *rtnl; /* rtnetlink socket */ 111 struct sock *genl_sock; 112 113 struct uevent_sock *uevent_sock; /* uevent socket */ 114 115 struct hlist_head *dev_name_head; 116 struct hlist_head *dev_index_head; 117 struct xarray dev_by_index; 118 struct raw_notifier_head netdev_chain; 119 120 /* Note that @hash_mix can be read millions times per second, 121 * it is critical that it is on a read_mostly cache line. 122 */ 123 u32 hash_mix; 124 bool is_dying; 125 126 struct net_device *loopback_dev; /* The loopback */ 127 128 /* core fib_rules */ 129 struct list_head rules_ops; 130 131 struct netns_core core; 132 struct netns_mib mib; 133 struct netns_packet packet; 134 #if IS_ENABLED(CONFIG_UNIX) 135 struct netns_unix unx; 136 #endif 137 struct netns_nexthop nexthop; 138 struct netns_ipv4 ipv4; 139 #if IS_ENABLED(CONFIG_IPV6) 140 struct netns_ipv6 ipv6; 141 #endif 142 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN) 143 struct netns_ieee802154_lowpan ieee802154_lowpan; 144 #endif 145 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE) 146 struct netns_sctp sctp; 147 #endif 148 #ifdef CONFIG_NETFILTER 149 struct netns_nf nf; 150 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 151 struct netns_ct ct; 152 #endif 153 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE) 154 struct netns_nftables nft; 155 #endif 156 #if IS_ENABLED(CONFIG_NF_FLOW_TABLE) 157 struct netns_ft ft; 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 /* Used to store attached BPF programs */ 166 struct netns_bpf bpf; 167 168 /* Note : following structs are cache line aligned */ 169 #ifdef CONFIG_XFRM 170 struct netns_xfrm xfrm; 171 #endif 172 173 u64 net_cookie; /* written once */ 174 175 #if IS_ENABLED(CONFIG_IP_VS) 176 struct netns_ipvs *ipvs; 177 #endif 178 #if IS_ENABLED(CONFIG_MPLS) 179 struct netns_mpls mpls; 180 #endif 181 #if IS_ENABLED(CONFIG_CAN) 182 struct netns_can can; 183 #endif 184 #ifdef CONFIG_XDP_SOCKETS 185 struct netns_xdp xdp; 186 #endif 187 #if IS_ENABLED(CONFIG_MCTP) 188 struct netns_mctp mctp; 189 #endif 190 #if IS_ENABLED(CONFIG_CRYPTO_USER) 191 struct sock *crypto_nlsk; 192 #endif 193 struct sock *diag_nlsk; 194 #if IS_ENABLED(CONFIG_SMC) 195 struct netns_smc smc; 196 #endif 197 #ifdef CONFIG_DEBUG_NET_SMALL_RTNL 198 /* Move to a better place when the config guard is removed. */ 199 struct mutex rtnl_mutex; 200 struct work_struct rtnl_work; 201 struct list_head dev_unreg_head; 202 spinlock_t dev_unreg_lock; 203 #endif 204 #if IS_ENABLED(CONFIG_VSOCKETS) 205 struct netns_vsock vsock; 206 #endif 207 } __randomize_layout; 208 209 #include <linux/seq_file_net.h> 210 211 /* Init's network namespace */ 212 extern struct net init_net; 213 214 #ifdef CONFIG_NET_NS 215 struct net *copy_net_ns(u64 flags, struct user_namespace *user_ns, 216 struct net *old_net); 217 218 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid); 219 220 void net_ns_barrier(void); 221 222 struct ns_common *get_net_ns(struct ns_common *ns); 223 struct net *get_net_ns_by_fd(int fd); 224 extern struct task_struct *cleanup_net_task; 225 226 #else /* CONFIG_NET_NS */ 227 #include <linux/sched.h> 228 #include <linux/nsproxy.h> 229 static inline struct net *copy_net_ns(u64 flags, 230 struct user_namespace *user_ns, struct net *old_net) 231 { 232 if (flags & CLONE_NEWNET) 233 return ERR_PTR(-EINVAL); 234 return old_net; 235 } 236 237 static inline void net_ns_get_ownership(const struct net *net, 238 kuid_t *uid, kgid_t *gid) 239 { 240 *uid = GLOBAL_ROOT_UID; 241 *gid = GLOBAL_ROOT_GID; 242 } 243 244 static inline void net_ns_barrier(void) {} 245 246 static inline struct ns_common *get_net_ns(struct ns_common *ns) 247 { 248 return ERR_PTR(-EINVAL); 249 } 250 251 static inline struct net *get_net_ns_by_fd(int fd) 252 { 253 return ERR_PTR(-EINVAL); 254 } 255 #endif /* CONFIG_NET_NS */ 256 257 258 extern struct list_head net_namespace_list; 259 260 struct net *get_net_ns_by_pid(pid_t pid); 261 262 #ifdef CONFIG_SYSCTL 263 void ipx_register_sysctl(void); 264 void ipx_unregister_sysctl(void); 265 #else 266 #define ipx_register_sysctl() 267 #define ipx_unregister_sysctl() 268 #endif 269 270 static inline struct net *to_net_ns(struct ns_common *ns) 271 { 272 return container_of(ns, struct net, ns); 273 } 274 275 #ifdef CONFIG_NET_NS 276 void __put_net(struct net *net); 277 278 /* Try using get_net_track() instead */ 279 static inline struct net *get_net(struct net *net) 280 { 281 ns_ref_inc(net); 282 return net; 283 } 284 285 static inline struct net *maybe_get_net(struct net *net) 286 { 287 /* Used when we know struct net exists but we 288 * aren't guaranteed a previous reference count 289 * exists. If the reference count is zero this 290 * function fails and returns NULL. 291 */ 292 if (!ns_ref_get(net)) 293 net = NULL; 294 return net; 295 } 296 297 /* Try using put_net_track() instead */ 298 static inline void put_net(struct net *net) 299 { 300 if (ns_ref_put(net)) 301 __put_net(net); 302 } 303 304 static inline 305 int net_eq(const struct net *net1, const struct net *net2) 306 { 307 return net1 == net2; 308 } 309 310 static inline int check_net(const struct net *net) 311 { 312 return ns_ref_read(net) != 0; 313 } 314 315 void net_drop_ns(struct ns_common *); 316 void net_passive_dec(struct net *net); 317 318 #else 319 320 static inline struct net *get_net(struct net *net) 321 { 322 return net; 323 } 324 325 static inline void put_net(struct net *net) 326 { 327 } 328 329 static inline struct net *maybe_get_net(struct net *net) 330 { 331 return net; 332 } 333 334 static inline 335 int net_eq(const struct net *net1, const struct net *net2) 336 { 337 return 1; 338 } 339 340 static inline int check_net(const struct net *net) 341 { 342 return 1; 343 } 344 345 #define net_drop_ns NULL 346 347 static inline void net_passive_dec(struct net *net) 348 { 349 refcount_dec(&net->passive); 350 } 351 #endif 352 353 static inline void net_passive_inc(struct net *net) 354 { 355 refcount_inc(&net->passive); 356 } 357 358 /* Returns true if the netns initialization is completed successfully */ 359 static inline bool net_initialized(const struct net *net) 360 { 361 return READ_ONCE(net->list.next); 362 } 363 364 static inline void __netns_tracker_alloc(struct net *net, 365 netns_tracker *tracker, 366 bool refcounted, 367 gfp_t gfp) 368 { 369 #ifdef CONFIG_NET_NS_REFCNT_TRACKER 370 ref_tracker_alloc(refcounted ? &net->refcnt_tracker : 371 &net->notrefcnt_tracker, 372 tracker, gfp); 373 #endif 374 } 375 376 static inline void netns_tracker_alloc(struct net *net, netns_tracker *tracker, 377 gfp_t gfp) 378 { 379 __netns_tracker_alloc(net, tracker, true, gfp); 380 } 381 382 static inline void __netns_tracker_free(struct net *net, 383 netns_tracker *tracker, 384 bool refcounted) 385 { 386 #ifdef CONFIG_NET_NS_REFCNT_TRACKER 387 ref_tracker_free(refcounted ? &net->refcnt_tracker : 388 &net->notrefcnt_tracker, tracker); 389 #endif 390 } 391 392 static inline struct net *get_net_track(struct net *net, 393 netns_tracker *tracker, gfp_t gfp) 394 { 395 get_net(net); 396 netns_tracker_alloc(net, tracker, gfp); 397 return net; 398 } 399 400 static inline void put_net_track(struct net *net, netns_tracker *tracker) 401 { 402 __netns_tracker_free(net, tracker, true); 403 put_net(net); 404 } 405 406 typedef struct { 407 #ifdef CONFIG_NET_NS 408 struct net __rcu *net; 409 #endif 410 } possible_net_t; 411 412 static inline void write_pnet(possible_net_t *pnet, struct net *net) 413 { 414 #ifdef CONFIG_NET_NS 415 rcu_assign_pointer(pnet->net, net); 416 #endif 417 } 418 419 static inline struct net *read_pnet(const possible_net_t *pnet) 420 { 421 #ifdef CONFIG_NET_NS 422 return rcu_dereference_protected(pnet->net, true); 423 #else 424 return &init_net; 425 #endif 426 } 427 428 static inline struct net *read_pnet_rcu(const possible_net_t *pnet) 429 { 430 #ifdef CONFIG_NET_NS 431 return rcu_dereference(pnet->net); 432 #else 433 return &init_net; 434 #endif 435 } 436 437 /* Protected by net_rwsem */ 438 #define for_each_net(VAR) \ 439 list_for_each_entry(VAR, &net_namespace_list, list) 440 #define for_each_net_continue_reverse(VAR) \ 441 list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list) 442 #define for_each_net_rcu(VAR) \ 443 list_for_each_entry_rcu(VAR, &net_namespace_list, list) 444 445 #ifdef CONFIG_NET_NS 446 #define __net_init 447 #define __net_exit 448 #define __net_initdata 449 #define __net_initconst 450 #else 451 #define __net_init __init 452 #define __net_exit __ref 453 #define __net_initdata __initdata 454 #define __net_initconst __initconst 455 #endif 456 457 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp); 458 int peernet2id(const struct net *net, struct net *peer); 459 bool peernet_has_id(const struct net *net, struct net *peer); 460 struct net *get_net_ns_by_id(const struct net *net, int id); 461 462 struct pernet_operations { 463 struct list_head list; 464 /* 465 * Below methods are called without any exclusive locks. 466 * More than one net may be constructed and destructed 467 * in parallel on several cpus. Every pernet_operations 468 * have to keep in mind all other pernet_operations and 469 * to introduce a locking, if they share common resources. 470 * 471 * The only time they are called with exclusive lock is 472 * from register_pernet_subsys(), unregister_pernet_subsys() 473 * register_pernet_device() and unregister_pernet_device(). 474 * 475 * Exit methods using blocking RCU primitives, such as 476 * synchronize_rcu(), should be implemented via exit_batch. 477 * Then, destruction of a group of net requires single 478 * synchronize_rcu() related to these pernet_operations, 479 * instead of separate synchronize_rcu() for every net. 480 * Please, avoid synchronize_rcu() at all, where it's possible. 481 * 482 * Note that a combination of pre_exit() and exit() can 483 * be used, since a synchronize_rcu() is guaranteed between 484 * the calls. 485 */ 486 int (*init)(struct net *net); 487 void (*pre_exit)(struct net *net); 488 void (*exit)(struct net *net); 489 void (*exit_batch)(struct list_head *net_exit_list); 490 /* Following method is called with RTNL held. */ 491 void (*exit_rtnl)(struct net *net, 492 struct list_head *dev_kill_list); 493 unsigned int * const id; 494 const size_t size; 495 }; 496 497 /* 498 * Use these carefully. If you implement a network device and it 499 * needs per network namespace operations use device pernet operations, 500 * otherwise use pernet subsys operations. 501 * 502 * Network interfaces need to be removed from a dying netns _before_ 503 * subsys notifiers can be called, as most of the network code cleanup 504 * (which is done from subsys notifiers) runs with the assumption that 505 * dev_remove_pack has been called so no new packets will arrive during 506 * and after the cleanup functions have been called. dev_remove_pack 507 * is not per namespace so instead the guarantee of no more packets 508 * arriving in a network namespace is provided by ensuring that all 509 * network devices and all sockets have left the network namespace 510 * before the cleanup methods are called. 511 * 512 * For the longest time the ipv4 icmp code was registered as a pernet 513 * device which caused kernel oops, and panics during network 514 * namespace cleanup. So please don't get this wrong. 515 */ 516 int register_pernet_subsys(struct pernet_operations *); 517 void unregister_pernet_subsys(struct pernet_operations *); 518 int register_pernet_device(struct pernet_operations *); 519 void unregister_pernet_device(struct pernet_operations *); 520 521 struct ctl_table; 522 523 #define register_net_sysctl(net, path, table) \ 524 register_net_sysctl_sz(net, path, table, ARRAY_SIZE(table)) 525 #ifdef CONFIG_SYSCTL 526 int net_sysctl_init(void); 527 struct ctl_table_header *register_net_sysctl_sz(struct net *net, const char *path, 528 struct ctl_table *table, size_t table_size); 529 void unregister_net_sysctl_table(struct ctl_table_header *header); 530 #else 531 static inline int net_sysctl_init(void) { return 0; } 532 static inline struct ctl_table_header *register_net_sysctl_sz(struct net *net, 533 const char *path, struct ctl_table *table, size_t table_size) 534 { 535 return NULL; 536 } 537 static inline void unregister_net_sysctl_table(struct ctl_table_header *header) 538 { 539 } 540 #endif 541 542 static inline int rt_genid_ipv4(const struct net *net) 543 { 544 return atomic_read(&net->ipv4.rt_genid); 545 } 546 547 #if IS_ENABLED(CONFIG_IPV6) 548 static inline int rt_genid_ipv6(const struct net *net) 549 { 550 return atomic_read(&net->ipv6.fib6_sernum); 551 } 552 #endif 553 554 static inline void rt_genid_bump_ipv4(struct net *net) 555 { 556 atomic_inc(&net->ipv4.rt_genid); 557 } 558 559 extern void (*__fib6_flush_trees)(struct net *net); 560 static inline void rt_genid_bump_ipv6(struct net *net) 561 { 562 if (__fib6_flush_trees) 563 __fib6_flush_trees(net); 564 } 565 566 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN) 567 static inline struct netns_ieee802154_lowpan * 568 net_ieee802154_lowpan(struct net *net) 569 { 570 return &net->ieee802154_lowpan; 571 } 572 #endif 573 574 /* For callers who don't really care about whether it's IPv4 or IPv6 */ 575 static inline void rt_genid_bump_all(struct net *net) 576 { 577 rt_genid_bump_ipv4(net); 578 rt_genid_bump_ipv6(net); 579 } 580 581 static inline int fnhe_genid(const struct net *net) 582 { 583 return atomic_read(&net->fnhe_genid); 584 } 585 586 static inline void fnhe_genid_bump(struct net *net) 587 { 588 atomic_inc(&net->fnhe_genid); 589 } 590 591 #ifdef CONFIG_NET 592 void net_ns_init(void); 593 #else 594 static inline void net_ns_init(void) {} 595 #endif 596 597 #endif /* __NET_NET_NAMESPACE_H */ 598