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 <net/neighbour_tables.h> 43 #include <linux/ns_common.h> 44 #include <linux/idr.h> 45 #include <linux/skbuff.h> 46 #include <linux/notifier.h> 47 #include <linux/xarray.h> 48 49 struct user_namespace; 50 struct proc_dir_entry; 51 struct neigh_table; 52 struct net_device; 53 struct sock; 54 struct ctl_table_header; 55 struct net_generic; 56 struct uevent_sock; 57 struct netns_ipvs; 58 struct bpf_prog; 59 60 61 #define NETDEV_HASHBITS 8 62 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS) 63 64 struct net { 65 /* First cache line can be often dirtied. 66 * Do not place here read-mostly fields. 67 */ 68 refcount_t passive; /* To decide when the network 69 * namespace should be freed. 70 */ 71 spinlock_t rules_mod_lock; 72 73 unsigned int dev_base_seq; /* protected by rtnl_mutex */ 74 atomic_t dev_mc_genid; /* dev->mc membership changes */ 75 u32 ifindex; 76 77 spinlock_t nsid_lock; 78 atomic_t fnhe_genid; 79 80 struct list_head list; /* list of network namespaces */ 81 struct list_head exit_list; /* To linked to call pernet exit 82 * methods on dead net ( 83 * pernet_ops_rwsem read locked), 84 * or to unregister pernet ops 85 * (pernet_ops_rwsem write locked). 86 */ 87 struct llist_node defer_free_list; 88 struct llist_node cleanup_list; /* namespaces on death row */ 89 90 struct list_head ptype_all; 91 struct list_head ptype_specific; 92 93 #ifdef CONFIG_KEYS 94 struct key_tag *key_domain; /* Key domain of operation tag */ 95 #endif 96 struct user_namespace *user_ns; /* Owning user namespace */ 97 struct ucounts *ucounts; 98 struct idr netns_ids; 99 100 struct ns_common ns; 101 struct ref_tracker_dir refcnt_tracker; 102 struct ref_tracker_dir notrefcnt_tracker; /* tracker for objects not 103 * refcounted against netns 104 */ 105 struct list_head dev_base_head; 106 struct proc_dir_entry *proc_net; 107 struct proc_dir_entry *proc_net_stat; 108 109 struct neigh_table *neigh_tables[NEIGH_NR_TABLES]; 110 111 #ifdef CONFIG_SYSCTL 112 struct ctl_table_set sysctls; 113 #endif 114 115 struct sock *rtnl; /* rtnetlink socket */ 116 struct sock *genl_sock; 117 118 struct uevent_sock *uevent_sock; /* uevent socket */ 119 120 struct hlist_head *dev_name_head; 121 struct hlist_head *dev_index_head; 122 struct xarray dev_by_index; 123 struct raw_notifier_head netdev_chain; 124 125 /* Note that @hash_mix can be read millions times per second, 126 * it is critical that it is on a read_mostly cache line. 127 */ 128 u32 hash_mix; 129 bool is_dying; 130 131 struct net_device *loopback_dev; /* The loopback */ 132 133 /* core fib_rules */ 134 struct list_head rules_ops; 135 136 struct netns_core core; 137 struct netns_mib mib; 138 struct netns_packet packet; 139 #if IS_ENABLED(CONFIG_UNIX) 140 struct netns_unix unx; 141 #endif 142 struct netns_nexthop nexthop; 143 struct netns_ipv4 ipv4; 144 #if IS_ENABLED(CONFIG_IPV6) 145 struct netns_ipv6 ipv6; 146 #endif 147 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN) 148 struct netns_ieee802154_lowpan ieee802154_lowpan; 149 #endif 150 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE) 151 struct netns_sctp sctp; 152 #endif 153 #ifdef CONFIG_NETFILTER 154 struct netns_nf nf; 155 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 156 struct netns_ct ct; 157 #endif 158 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE) 159 struct netns_nftables nft; 160 #endif 161 #if IS_ENABLED(CONFIG_NF_FLOW_TABLE) 162 struct netns_ft ft; 163 #endif 164 #endif 165 #ifdef CONFIG_WEXT_CORE 166 struct sk_buff_head wext_nlevents; 167 #endif 168 struct net_generic __rcu *gen; 169 170 /* Used to store attached BPF programs */ 171 struct netns_bpf bpf; 172 173 /* Note : following structs are cache line aligned */ 174 #ifdef CONFIG_XFRM 175 struct netns_xfrm xfrm; 176 #endif 177 178 u64 net_cookie; /* written once */ 179 180 #if IS_ENABLED(CONFIG_IP_VS) 181 struct netns_ipvs *ipvs; 182 #endif 183 #if IS_ENABLED(CONFIG_MPLS) 184 struct netns_mpls mpls; 185 #endif 186 #if IS_ENABLED(CONFIG_CAN) 187 struct netns_can can; 188 #endif 189 #ifdef CONFIG_XDP_SOCKETS 190 struct netns_xdp xdp; 191 #endif 192 #if IS_ENABLED(CONFIG_MCTP) 193 struct netns_mctp mctp; 194 #endif 195 #if IS_ENABLED(CONFIG_CRYPTO_USER) 196 struct sock *crypto_nlsk; 197 #endif 198 struct sock *diag_nlsk; 199 #if IS_ENABLED(CONFIG_SMC) 200 struct netns_smc smc; 201 #endif 202 #ifdef CONFIG_DEBUG_NET_SMALL_RTNL 203 /* Move to a better place when the config guard is removed. */ 204 struct mutex rtnl_mutex; 205 struct work_struct rtnl_work; 206 struct list_head dev_unreg_head; 207 spinlock_t dev_unreg_lock; 208 #endif 209 #if IS_ENABLED(CONFIG_VSOCKETS) 210 struct netns_vsock vsock; 211 #endif 212 } __randomize_layout; 213 214 #include <linux/seq_file_net.h> 215 216 /* Init's network namespace */ 217 extern struct net init_net; 218 219 #ifdef CONFIG_NET_NS 220 struct net *copy_net_ns(u64 flags, struct user_namespace *user_ns, 221 struct net *old_net); 222 223 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid); 224 225 void net_ns_barrier(void); 226 227 struct ns_common *get_net_ns(struct ns_common *ns); 228 struct net *get_net_ns_by_fd(int fd); 229 extern struct task_struct *cleanup_net_task; 230 231 #else /* CONFIG_NET_NS */ 232 #include <linux/sched.h> 233 #include <linux/nsproxy.h> 234 static inline struct net *copy_net_ns(u64 flags, 235 struct user_namespace *user_ns, struct net *old_net) 236 { 237 if (flags & CLONE_NEWNET) 238 return ERR_PTR(-EINVAL); 239 return old_net; 240 } 241 242 static inline void net_ns_get_ownership(const struct net *net, 243 kuid_t *uid, kgid_t *gid) 244 { 245 *uid = GLOBAL_ROOT_UID; 246 *gid = GLOBAL_ROOT_GID; 247 } 248 249 static inline void net_ns_barrier(void) {} 250 251 static inline struct ns_common *get_net_ns(struct ns_common *ns) 252 { 253 return ERR_PTR(-EINVAL); 254 } 255 256 static inline struct net *get_net_ns_by_fd(int fd) 257 { 258 return ERR_PTR(-EINVAL); 259 } 260 #endif /* CONFIG_NET_NS */ 261 262 263 extern struct list_head net_namespace_list; 264 265 struct net *get_net_ns_by_pid(pid_t pid); 266 267 #ifdef CONFIG_SYSCTL 268 void ipx_register_sysctl(void); 269 void ipx_unregister_sysctl(void); 270 #else 271 #define ipx_register_sysctl() 272 #define ipx_unregister_sysctl() 273 #endif 274 275 static inline struct net *to_net_ns(struct ns_common *ns) 276 { 277 return container_of(ns, struct net, ns); 278 } 279 280 #ifdef CONFIG_NET_NS 281 void __put_net(struct net *net); 282 283 /* Try using get_net_track() instead */ 284 static inline struct net *get_net(struct net *net) 285 { 286 ns_ref_inc(net); 287 return net; 288 } 289 290 static inline struct net *maybe_get_net(struct net *net) 291 { 292 /* Used when we know struct net exists but we 293 * aren't guaranteed a previous reference count 294 * exists. If the reference count is zero this 295 * function fails and returns NULL. 296 */ 297 if (!ns_ref_get(net)) 298 net = NULL; 299 return net; 300 } 301 302 /* Try using put_net_track() instead */ 303 static inline void put_net(struct net *net) 304 { 305 if (ns_ref_put(net)) 306 __put_net(net); 307 } 308 309 static inline 310 int net_eq(const struct net *net1, const struct net *net2) 311 { 312 return net1 == net2; 313 } 314 315 static inline int check_net(const struct net *net) 316 { 317 return ns_ref_read(net) != 0; 318 } 319 320 void net_drop_ns(struct ns_common *); 321 void net_passive_dec(struct net *net); 322 323 #else 324 325 static inline struct net *get_net(struct net *net) 326 { 327 return net; 328 } 329 330 static inline void put_net(struct net *net) 331 { 332 } 333 334 static inline struct net *maybe_get_net(struct net *net) 335 { 336 return net; 337 } 338 339 static inline 340 int net_eq(const struct net *net1, const struct net *net2) 341 { 342 return 1; 343 } 344 345 static inline int check_net(const struct net *net) 346 { 347 return 1; 348 } 349 350 #define net_drop_ns NULL 351 352 static inline void net_passive_dec(struct net *net) 353 { 354 refcount_dec(&net->passive); 355 } 356 #endif 357 358 static inline void net_passive_inc(struct net *net) 359 { 360 refcount_inc(&net->passive); 361 } 362 363 /* Returns true if the netns initialization is completed successfully */ 364 static inline bool net_initialized(const struct net *net) 365 { 366 return READ_ONCE(net->list.next); 367 } 368 369 static inline void __netns_tracker_alloc(struct net *net, 370 netns_tracker *tracker, 371 bool refcounted, 372 gfp_t gfp) 373 { 374 #ifdef CONFIG_NET_NS_REFCNT_TRACKER 375 ref_tracker_alloc(refcounted ? &net->refcnt_tracker : 376 &net->notrefcnt_tracker, 377 tracker, gfp); 378 #endif 379 } 380 381 static inline void netns_tracker_alloc(struct net *net, netns_tracker *tracker, 382 gfp_t gfp) 383 { 384 __netns_tracker_alloc(net, tracker, true, gfp); 385 } 386 387 static inline void __netns_tracker_free(struct net *net, 388 netns_tracker *tracker, 389 bool refcounted) 390 { 391 #ifdef CONFIG_NET_NS_REFCNT_TRACKER 392 ref_tracker_free(refcounted ? &net->refcnt_tracker : 393 &net->notrefcnt_tracker, tracker); 394 #endif 395 } 396 397 static inline struct net *get_net_track(struct net *net, 398 netns_tracker *tracker, gfp_t gfp) 399 { 400 get_net(net); 401 netns_tracker_alloc(net, tracker, gfp); 402 return net; 403 } 404 405 static inline void put_net_track(struct net *net, netns_tracker *tracker) 406 { 407 __netns_tracker_free(net, tracker, true); 408 put_net(net); 409 } 410 411 typedef struct { 412 #ifdef CONFIG_NET_NS 413 struct net __rcu *net; 414 #endif 415 } possible_net_t; 416 417 static inline void write_pnet(possible_net_t *pnet, struct net *net) 418 { 419 #ifdef CONFIG_NET_NS 420 rcu_assign_pointer(pnet->net, net); 421 #endif 422 } 423 424 static inline struct net *read_pnet(const possible_net_t *pnet) 425 { 426 #ifdef CONFIG_NET_NS 427 return rcu_dereference_protected(pnet->net, true); 428 #else 429 return &init_net; 430 #endif 431 } 432 433 static inline struct net *read_pnet_rcu(const possible_net_t *pnet) 434 { 435 #ifdef CONFIG_NET_NS 436 return rcu_dereference(pnet->net); 437 #else 438 return &init_net; 439 #endif 440 } 441 442 /* Protected by net_rwsem */ 443 #define for_each_net(VAR) \ 444 list_for_each_entry(VAR, &net_namespace_list, list) 445 #define for_each_net_continue_reverse(VAR) \ 446 list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list) 447 #define for_each_net_rcu(VAR) \ 448 list_for_each_entry_rcu(VAR, &net_namespace_list, list) 449 450 #ifdef CONFIG_NET_NS 451 #define __net_init 452 #define __net_exit 453 #define __net_initdata 454 #define __net_initconst 455 #else 456 #define __net_init __init 457 #define __net_exit __ref 458 #define __net_initdata __initdata 459 #define __net_initconst __initconst 460 #endif 461 462 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp); 463 int peernet2id(const struct net *net, struct net *peer); 464 bool peernet_has_id(const struct net *net, struct net *peer); 465 struct net *get_net_ns_by_id(const struct net *net, int id); 466 467 struct pernet_operations { 468 struct list_head list; 469 /* 470 * Below methods are called without any exclusive locks. 471 * More than one net may be constructed and destructed 472 * in parallel on several cpus. Every pernet_operations 473 * have to keep in mind all other pernet_operations and 474 * to introduce a locking, if they share common resources. 475 * 476 * The only time they are called with exclusive lock is 477 * from register_pernet_subsys(), unregister_pernet_subsys() 478 * register_pernet_device() and unregister_pernet_device(). 479 * 480 * Exit methods using blocking RCU primitives, such as 481 * synchronize_rcu(), should be implemented via exit_batch. 482 * Then, destruction of a group of net requires single 483 * synchronize_rcu() related to these pernet_operations, 484 * instead of separate synchronize_rcu() for every net. 485 * Please, avoid synchronize_rcu() at all, where it's possible. 486 * 487 * Note that a combination of pre_exit() and exit() can 488 * be used, since a synchronize_rcu() is guaranteed between 489 * the calls. 490 */ 491 int (*init)(struct net *net); 492 void (*pre_exit)(struct net *net); 493 void (*exit)(struct net *net); 494 void (*exit_batch)(struct list_head *net_exit_list); 495 /* Following method is called with RTNL held. */ 496 void (*exit_rtnl)(struct net *net, 497 struct list_head *dev_kill_list); 498 unsigned int * const id; 499 const size_t size; 500 }; 501 502 /* 503 * Use these carefully. If you implement a network device and it 504 * needs per network namespace operations use device pernet operations, 505 * otherwise use pernet subsys operations. 506 * 507 * Network interfaces need to be removed from a dying netns _before_ 508 * subsys notifiers can be called, as most of the network code cleanup 509 * (which is done from subsys notifiers) runs with the assumption that 510 * dev_remove_pack has been called so no new packets will arrive during 511 * and after the cleanup functions have been called. dev_remove_pack 512 * is not per namespace so instead the guarantee of no more packets 513 * arriving in a network namespace is provided by ensuring that all 514 * network devices and all sockets have left the network namespace 515 * before the cleanup methods are called. 516 * 517 * For the longest time the ipv4 icmp code was registered as a pernet 518 * device which caused kernel oops, and panics during network 519 * namespace cleanup. So please don't get this wrong. 520 */ 521 int register_pernet_subsys(struct pernet_operations *); 522 void unregister_pernet_subsys(struct pernet_operations *); 523 int register_pernet_device(struct pernet_operations *); 524 void unregister_pernet_device(struct pernet_operations *); 525 526 struct ctl_table; 527 528 #define register_net_sysctl(net, path, table) \ 529 register_net_sysctl_sz(net, path, table, ARRAY_SIZE(table)) 530 #ifdef CONFIG_SYSCTL 531 int net_sysctl_init(void); 532 struct ctl_table_header *register_net_sysctl_sz(struct net *net, const char *path, 533 const struct ctl_table *table, 534 size_t table_size); 535 void unregister_net_sysctl_table(struct ctl_table_header *header); 536 #else 537 static inline int net_sysctl_init(void) { return 0; } 538 static inline struct ctl_table_header *register_net_sysctl_sz(struct net *net, 539 const char *path, const struct ctl_table *table, size_t table_size) 540 { 541 return NULL; 542 } 543 static inline void unregister_net_sysctl_table(struct ctl_table_header *header) 544 { 545 } 546 #endif 547 548 static inline int rt_genid_ipv4(const struct net *net) 549 { 550 return atomic_read(&net->ipv4.rt_genid); 551 } 552 553 #if IS_ENABLED(CONFIG_IPV6) 554 static inline int rt_genid_ipv6(const struct net *net) 555 { 556 return atomic_read(&net->ipv6.fib6_sernum); 557 } 558 #endif 559 560 static inline void rt_genid_bump_ipv4(struct net *net) 561 { 562 atomic_inc(&net->ipv4.rt_genid); 563 } 564 565 extern void (*__fib6_flush_trees)(struct net *net); 566 static inline void rt_genid_bump_ipv6(struct net *net) 567 { 568 if (__fib6_flush_trees) 569 __fib6_flush_trees(net); 570 } 571 572 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN) 573 static inline struct netns_ieee802154_lowpan * 574 net_ieee802154_lowpan(struct net *net) 575 { 576 return &net->ieee802154_lowpan; 577 } 578 #endif 579 580 /* For callers who don't really care about whether it's IPv4 or IPv6 */ 581 static inline void rt_genid_bump_all(struct net *net) 582 { 583 rt_genid_bump_ipv4(net); 584 rt_genid_bump_ipv6(net); 585 } 586 587 static inline int fnhe_genid(const struct net *net) 588 { 589 return atomic_read(&net->fnhe_genid); 590 } 591 592 static inline void fnhe_genid_bump(struct net *net) 593 { 594 atomic_inc(&net->fnhe_genid); 595 } 596 597 #ifdef CONFIG_NET 598 void net_ns_init(void); 599 #else 600 static inline void net_ns_init(void) {} 601 #endif 602 603 #endif /* __NET_NET_NAMESPACE_H */ 604