1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Operations on the network namespace
5 #ifndef __NET_NET_NAMESPACE_H
6 #define __NET_NET_NAMESPACE_H
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>
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>
29 #if IS_ENABLED(CONFIG_NF_FLOW_TABLE)
30 #include <net/netns/flow_table.h>
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>
47 struct user_namespace;
48 struct proc_dir_entry;
51 struct ctl_table_header;
58 #define NETDEV_HASHBITS 8
59 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
62 /* First cache line can be often dirtied.
63 * Do not place here read-mostly fields.
65 refcount_t passive; /* To decide when the network
66 * namespace should be freed.
68 spinlock_t rules_mod_lock;
70 atomic_t dev_unreg_count;
72 unsigned int dev_base_seq; /* protected by rtnl_mutex */
78 struct list_head list; /* list of network namespaces */
79 struct list_head exit_list; /* To linked to call pernet exit
80 * methods on dead net (
81 * pernet_ops_rwsem read locked),
82 * or to unregister pernet ops
83 * (pernet_ops_rwsem write locked).
85 struct llist_node cleanup_list; /* namespaces on death row */
88 struct key_tag *key_domain; /* Key domain of operation tag */
90 struct user_namespace *user_ns; /* Owning user namespace */
91 struct ucounts *ucounts;
95 struct ref_tracker_dir refcnt_tracker;
96 struct ref_tracker_dir notrefcnt_tracker; /* tracker for objects not
97 * refcounted against netns
99 struct list_head dev_base_head;
100 struct proc_dir_entry *proc_net;
101 struct proc_dir_entry *proc_net_stat;
104 struct ctl_table_set sysctls;
107 struct sock *rtnl; /* rtnetlink socket */
108 struct sock *genl_sock;
110 struct uevent_sock *uevent_sock; /* uevent socket */
112 struct hlist_head *dev_name_head;
113 struct hlist_head *dev_index_head;
114 struct xarray dev_by_index;
115 struct raw_notifier_head netdev_chain;
117 /* Note that @hash_mix can be read millions times per second,
118 * it is critical that it is on a read_mostly cache line.
122 struct net_device *loopback_dev; /* The loopback */
125 struct list_head rules_ops;
127 struct netns_core core;
128 struct netns_mib mib;
129 struct netns_packet packet;
130 #if IS_ENABLED(CONFIG_UNIX)
131 struct netns_unix unx;
133 struct netns_nexthop nexthop;
134 struct netns_ipv4 ipv4;
135 #if IS_ENABLED(CONFIG_IPV6)
136 struct netns_ipv6 ipv6;
138 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
139 struct netns_ieee802154_lowpan ieee802154_lowpan;
141 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
142 struct netns_sctp sctp;
144 #ifdef CONFIG_NETFILTER
146 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
149 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
150 struct netns_nftables nft;
152 #if IS_ENABLED(CONFIG_NF_FLOW_TABLE)
156 #ifdef CONFIG_WEXT_CORE
157 struct sk_buff_head wext_nlevents;
159 struct net_generic __rcu *gen;
161 /* Used to store attached BPF programs */
162 struct netns_bpf bpf;
164 /* Note : following structs are cache line aligned */
166 struct netns_xfrm xfrm;
169 u64 net_cookie; /* written once */
171 #if IS_ENABLED(CONFIG_IP_VS)
172 struct netns_ipvs *ipvs;
174 #if IS_ENABLED(CONFIG_MPLS)
175 struct netns_mpls mpls;
177 #if IS_ENABLED(CONFIG_CAN)
178 struct netns_can can;
180 #ifdef CONFIG_XDP_SOCKETS
181 struct netns_xdp xdp;
183 #if IS_ENABLED(CONFIG_MCTP)
184 struct netns_mctp mctp;
186 #if IS_ENABLED(CONFIG_CRYPTO_USER)
187 struct sock *crypto_nlsk;
189 struct sock *diag_nlsk;
190 #if IS_ENABLED(CONFIG_SMC)
191 struct netns_smc smc;
193 } __randomize_layout;
195 #include <linux/seq_file_net.h>
197 /* Init's network namespace */
198 extern struct net init_net;
201 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
202 struct net *old_net);
204 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
206 void net_ns_barrier(void);
208 struct ns_common *get_net_ns(struct ns_common *ns);
209 struct net *get_net_ns_by_fd(int fd);
210 #else /* CONFIG_NET_NS */
211 #include <linux/sched.h>
212 #include <linux/nsproxy.h>
213 static inline struct net *copy_net_ns(unsigned long flags,
214 struct user_namespace *user_ns, struct net *old_net)
216 if (flags & CLONE_NEWNET)
217 return ERR_PTR(-EINVAL);
221 static inline void net_ns_get_ownership(const struct net *net,
222 kuid_t *uid, kgid_t *gid)
224 *uid = GLOBAL_ROOT_UID;
225 *gid = GLOBAL_ROOT_GID;
228 static inline void net_ns_barrier(void) {}
230 static inline struct ns_common *get_net_ns(struct ns_common *ns)
232 return ERR_PTR(-EINVAL);
235 static inline struct net *get_net_ns_by_fd(int fd)
237 return ERR_PTR(-EINVAL);
239 #endif /* CONFIG_NET_NS */
242 extern struct list_head net_namespace_list;
244 struct net *get_net_ns_by_pid(pid_t pid);
247 void ipx_register_sysctl(void);
248 void ipx_unregister_sysctl(void);
250 #define ipx_register_sysctl()
251 #define ipx_unregister_sysctl()
255 void __put_net(struct net *net);
257 /* Try using get_net_track() instead */
258 static inline struct net *get_net(struct net *net)
260 refcount_inc(&net->ns.count);
264 static inline struct net *maybe_get_net(struct net *net)
266 /* Used when we know struct net exists but we
267 * aren't guaranteed a previous reference count
268 * exists. If the reference count is zero this
269 * function fails and returns NULL.
271 if (!refcount_inc_not_zero(&net->ns.count))
276 /* Try using put_net_track() instead */
277 static inline void put_net(struct net *net)
279 if (refcount_dec_and_test(&net->ns.count))
284 int net_eq(const struct net *net1, const struct net *net2)
289 static inline int check_net(const struct net *net)
291 return refcount_read(&net->ns.count) != 0;
294 void net_drop_ns(void *);
298 static inline struct net *get_net(struct net *net)
303 static inline void put_net(struct net *net)
307 static inline struct net *maybe_get_net(struct net *net)
313 int net_eq(const struct net *net1, const struct net *net2)
318 static inline int check_net(const struct net *net)
323 #define net_drop_ns NULL
327 static inline void __netns_tracker_alloc(struct net *net,
328 netns_tracker *tracker,
332 #ifdef CONFIG_NET_NS_REFCNT_TRACKER
333 ref_tracker_alloc(refcounted ? &net->refcnt_tracker :
334 &net->notrefcnt_tracker,
339 static inline void netns_tracker_alloc(struct net *net, netns_tracker *tracker,
342 __netns_tracker_alloc(net, tracker, true, gfp);
345 static inline void __netns_tracker_free(struct net *net,
346 netns_tracker *tracker,
349 #ifdef CONFIG_NET_NS_REFCNT_TRACKER
350 ref_tracker_free(refcounted ? &net->refcnt_tracker :
351 &net->notrefcnt_tracker, tracker);
355 static inline struct net *get_net_track(struct net *net,
356 netns_tracker *tracker, gfp_t gfp)
359 netns_tracker_alloc(net, tracker, gfp);
363 static inline void put_net_track(struct net *net, netns_tracker *tracker)
365 __netns_tracker_free(net, tracker, true);
375 static inline void write_pnet(possible_net_t *pnet, struct net *net)
382 static inline struct net *read_pnet(const possible_net_t *pnet)
391 /* Protected by net_rwsem */
392 #define for_each_net(VAR) \
393 list_for_each_entry(VAR, &net_namespace_list, list)
394 #define for_each_net_continue_reverse(VAR) \
395 list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list)
396 #define for_each_net_rcu(VAR) \
397 list_for_each_entry_rcu(VAR, &net_namespace_list, list)
402 #define __net_initdata
403 #define __net_initconst
405 #define __net_init __init
406 #define __net_exit __ref
407 #define __net_initdata __initdata
408 #define __net_initconst __initconst
411 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp);
412 int peernet2id(const struct net *net, struct net *peer);
413 bool peernet_has_id(const struct net *net, struct net *peer);
414 struct net *get_net_ns_by_id(const struct net *net, int id);
416 struct pernet_operations {
417 struct list_head list;
419 * Below methods are called without any exclusive locks.
420 * More than one net may be constructed and destructed
421 * in parallel on several cpus. Every pernet_operations
422 * have to keep in mind all other pernet_operations and
423 * to introduce a locking, if they share common resources.
425 * The only time they are called with exclusive lock is
426 * from register_pernet_subsys(), unregister_pernet_subsys()
427 * register_pernet_device() and unregister_pernet_device().
429 * Exit methods using blocking RCU primitives, such as
430 * synchronize_rcu(), should be implemented via exit_batch.
431 * Then, destruction of a group of net requires single
432 * synchronize_rcu() related to these pernet_operations,
433 * instead of separate synchronize_rcu() for every net.
434 * Please, avoid synchronize_rcu() at all, where it's possible.
436 * Note that a combination of pre_exit() and exit() can
437 * be used, since a synchronize_rcu() is guaranteed between
440 int (*init)(struct net *net);
441 void (*pre_exit)(struct net *net);
442 void (*exit)(struct net *net);
443 void (*exit_batch)(struct list_head *net_exit_list);
449 * Use these carefully. If you implement a network device and it
450 * needs per network namespace operations use device pernet operations,
451 * otherwise use pernet subsys operations.
453 * Network interfaces need to be removed from a dying netns _before_
454 * subsys notifiers can be called, as most of the network code cleanup
455 * (which is done from subsys notifiers) runs with the assumption that
456 * dev_remove_pack has been called so no new packets will arrive during
457 * and after the cleanup functions have been called. dev_remove_pack
458 * is not per namespace so instead the guarantee of no more packets
459 * arriving in a network namespace is provided by ensuring that all
460 * network devices and all sockets have left the network namespace
461 * before the cleanup methods are called.
463 * For the longest time the ipv4 icmp code was registered as a pernet
464 * device which caused kernel oops, and panics during network
465 * namespace cleanup. So please don't get this wrong.
467 int register_pernet_subsys(struct pernet_operations *);
468 void unregister_pernet_subsys(struct pernet_operations *);
469 int register_pernet_device(struct pernet_operations *);
470 void unregister_pernet_device(struct pernet_operations *);
474 #define register_net_sysctl(net, path, table) \
475 register_net_sysctl_sz(net, path, table, ARRAY_SIZE(table))
477 int net_sysctl_init(void);
478 struct ctl_table_header *register_net_sysctl_sz(struct net *net, const char *path,
479 struct ctl_table *table, size_t table_size);
480 void unregister_net_sysctl_table(struct ctl_table_header *header);
482 static inline int net_sysctl_init(void) { return 0; }
483 static inline struct ctl_table_header *register_net_sysctl_sz(struct net *net,
484 const char *path, struct ctl_table *table, size_t table_size)
488 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
493 static inline int rt_genid_ipv4(const struct net *net)
495 return atomic_read(&net->ipv4.rt_genid);
498 #if IS_ENABLED(CONFIG_IPV6)
499 static inline int rt_genid_ipv6(const struct net *net)
501 return atomic_read(&net->ipv6.fib6_sernum);
505 static inline void rt_genid_bump_ipv4(struct net *net)
507 atomic_inc(&net->ipv4.rt_genid);
510 extern void (*__fib6_flush_trees)(struct net *net);
511 static inline void rt_genid_bump_ipv6(struct net *net)
513 if (__fib6_flush_trees)
514 __fib6_flush_trees(net);
517 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
518 static inline struct netns_ieee802154_lowpan *
519 net_ieee802154_lowpan(struct net *net)
521 return &net->ieee802154_lowpan;
525 /* For callers who don't really care about whether it's IPv4 or IPv6 */
526 static inline void rt_genid_bump_all(struct net *net)
528 rt_genid_bump_ipv4(net);
529 rt_genid_bump_ipv6(net);
532 static inline int fnhe_genid(const struct net *net)
534 return atomic_read(&net->fnhe_genid);
537 static inline void fnhe_genid_bump(struct net *net)
539 atomic_inc(&net->fnhe_genid);
543 void net_ns_init(void);
545 static inline void net_ns_init(void) {}
548 #endif /* __NET_NET_NAMESPACE_H */