xref: /linux/include/net/net_namespace.h (revision ed63b9c873601ca113da5c7b1745e3946493e9f3)
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/ieee802154_6lowpan.h>
23 #include <net/netns/sctp.h>
24 #include <net/netns/dccp.h>
25 #include <net/netns/netfilter.h>
26 #include <net/netns/x_tables.h>
27 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
28 #include <net/netns/conntrack.h>
29 #endif
30 #include <net/netns/nftables.h>
31 #include <net/netns/xfrm.h>
32 #include <net/netns/mpls.h>
33 #include <net/netns/can.h>
34 #include <net/netns/xdp.h>
35 #include <linux/ns_common.h>
36 #include <linux/idr.h>
37 #include <linux/skbuff.h>
38 
39 struct user_namespace;
40 struct proc_dir_entry;
41 struct net_device;
42 struct sock;
43 struct ctl_table_header;
44 struct net_generic;
45 struct uevent_sock;
46 struct netns_ipvs;
47 struct bpf_prog;
48 
49 
50 #define NETDEV_HASHBITS    8
51 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
52 
53 struct net {
54 	refcount_t		passive;	/* To decided when the network
55 						 * namespace should be freed.
56 						 */
57 	refcount_t		count;		/* To decided when the network
58 						 *  namespace should be shut down.
59 						 */
60 	spinlock_t		rules_mod_lock;
61 
62 	u32			hash_mix;
63 	atomic64_t		cookie_gen;
64 
65 	struct list_head	list;		/* list of network namespaces */
66 	struct list_head	exit_list;	/* To linked to call pernet exit
67 						 * methods on dead net (
68 						 * pernet_ops_rwsem read locked),
69 						 * or to unregister pernet ops
70 						 * (pernet_ops_rwsem write locked).
71 						 */
72 	struct llist_node	cleanup_list;	/* namespaces on death row */
73 
74 #ifdef CONFIG_KEYS
75 	struct key_tag		*key_domain;	/* Key domain of operation tag */
76 #endif
77 	struct user_namespace   *user_ns;	/* Owning user namespace */
78 	struct ucounts		*ucounts;
79 	spinlock_t		nsid_lock;
80 	struct idr		netns_ids;
81 
82 	struct ns_common	ns;
83 
84 	struct proc_dir_entry 	*proc_net;
85 	struct proc_dir_entry 	*proc_net_stat;
86 
87 #ifdef CONFIG_SYSCTL
88 	struct ctl_table_set	sysctls;
89 #endif
90 
91 	struct sock 		*rtnl;			/* rtnetlink socket */
92 	struct sock		*genl_sock;
93 
94 	struct uevent_sock	*uevent_sock;		/* uevent socket */
95 
96 	struct list_head 	dev_base_head;
97 	struct hlist_head 	*dev_name_head;
98 	struct hlist_head	*dev_index_head;
99 	unsigned int		dev_base_seq;	/* protected by rtnl_mutex */
100 	int			ifindex;
101 	unsigned int		dev_unreg_count;
102 
103 	/* core fib_rules */
104 	struct list_head	rules_ops;
105 
106 	struct list_head	fib_notifier_ops;  /* Populated by
107 						    * register_pernet_subsys()
108 						    */
109 	struct net_device       *loopback_dev;          /* The loopback */
110 	struct netns_core	core;
111 	struct netns_mib	mib;
112 	struct netns_packet	packet;
113 	struct netns_unix	unx;
114 	struct netns_ipv4	ipv4;
115 #if IS_ENABLED(CONFIG_IPV6)
116 	struct netns_ipv6	ipv6;
117 #endif
118 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
119 	struct netns_ieee802154_lowpan	ieee802154_lowpan;
120 #endif
121 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
122 	struct netns_sctp	sctp;
123 #endif
124 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
125 	struct netns_dccp	dccp;
126 #endif
127 #ifdef CONFIG_NETFILTER
128 	struct netns_nf		nf;
129 	struct netns_xt		xt;
130 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
131 	struct netns_ct		ct;
132 #endif
133 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
134 	struct netns_nftables	nft;
135 #endif
136 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
137 	struct netns_nf_frag	nf_frag;
138 	struct ctl_table_header *nf_frag_frags_hdr;
139 #endif
140 	struct sock		*nfnl;
141 	struct sock		*nfnl_stash;
142 #if IS_ENABLED(CONFIG_NETFILTER_NETLINK_ACCT)
143 	struct list_head        nfnl_acct_list;
144 #endif
145 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
146 	struct list_head	nfct_timeout_list;
147 #endif
148 #endif
149 #ifdef CONFIG_WEXT_CORE
150 	struct sk_buff_head	wext_nlevents;
151 #endif
152 	struct net_generic __rcu	*gen;
153 
154 	struct bpf_prog __rcu	*flow_dissector_prog;
155 
156 	/* Note : following structs are cache line aligned */
157 #ifdef CONFIG_XFRM
158 	struct netns_xfrm	xfrm;
159 #endif
160 #if IS_ENABLED(CONFIG_IP_VS)
161 	struct netns_ipvs	*ipvs;
162 #endif
163 #if IS_ENABLED(CONFIG_MPLS)
164 	struct netns_mpls	mpls;
165 #endif
166 #if IS_ENABLED(CONFIG_CAN)
167 	struct netns_can	can;
168 #endif
169 #ifdef CONFIG_XDP_SOCKETS
170 	struct netns_xdp	xdp;
171 #endif
172 	struct sock		*diag_nlsk;
173 	atomic_t		fnhe_genid;
174 } __randomize_layout;
175 
176 #include <linux/seq_file_net.h>
177 
178 /* Init's network namespace */
179 extern struct net init_net;
180 
181 #ifdef CONFIG_NET_NS
182 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
183 			struct net *old_net);
184 
185 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
186 
187 void net_ns_barrier(void);
188 #else /* CONFIG_NET_NS */
189 #include <linux/sched.h>
190 #include <linux/nsproxy.h>
191 static inline struct net *copy_net_ns(unsigned long flags,
192 	struct user_namespace *user_ns, struct net *old_net)
193 {
194 	if (flags & CLONE_NEWNET)
195 		return ERR_PTR(-EINVAL);
196 	return old_net;
197 }
198 
199 static inline void net_ns_get_ownership(const struct net *net,
200 					kuid_t *uid, kgid_t *gid)
201 {
202 	*uid = GLOBAL_ROOT_UID;
203 	*gid = GLOBAL_ROOT_GID;
204 }
205 
206 static inline void net_ns_barrier(void) {}
207 #endif /* CONFIG_NET_NS */
208 
209 
210 extern struct list_head net_namespace_list;
211 
212 struct net *get_net_ns_by_pid(pid_t pid);
213 struct net *get_net_ns_by_fd(int fd);
214 
215 #ifdef CONFIG_SYSCTL
216 void ipx_register_sysctl(void);
217 void ipx_unregister_sysctl(void);
218 #else
219 #define ipx_register_sysctl()
220 #define ipx_unregister_sysctl()
221 #endif
222 
223 #ifdef CONFIG_NET_NS
224 void __put_net(struct net *net);
225 
226 static inline struct net *get_net(struct net *net)
227 {
228 	refcount_inc(&net->count);
229 	return net;
230 }
231 
232 static inline struct net *maybe_get_net(struct net *net)
233 {
234 	/* Used when we know struct net exists but we
235 	 * aren't guaranteed a previous reference count
236 	 * exists.  If the reference count is zero this
237 	 * function fails and returns NULL.
238 	 */
239 	if (!refcount_inc_not_zero(&net->count))
240 		net = NULL;
241 	return net;
242 }
243 
244 static inline void put_net(struct net *net)
245 {
246 	if (refcount_dec_and_test(&net->count))
247 		__put_net(net);
248 }
249 
250 static inline
251 int net_eq(const struct net *net1, const struct net *net2)
252 {
253 	return net1 == net2;
254 }
255 
256 static inline int check_net(const struct net *net)
257 {
258 	return refcount_read(&net->count) != 0;
259 }
260 
261 void net_drop_ns(void *);
262 
263 #else
264 
265 static inline struct net *get_net(struct net *net)
266 {
267 	return net;
268 }
269 
270 static inline void put_net(struct net *net)
271 {
272 }
273 
274 static inline struct net *maybe_get_net(struct net *net)
275 {
276 	return net;
277 }
278 
279 static inline
280 int net_eq(const struct net *net1, const struct net *net2)
281 {
282 	return 1;
283 }
284 
285 static inline int check_net(const struct net *net)
286 {
287 	return 1;
288 }
289 
290 #define net_drop_ns NULL
291 #endif
292 
293 
294 typedef struct {
295 #ifdef CONFIG_NET_NS
296 	struct net *net;
297 #endif
298 } possible_net_t;
299 
300 static inline void write_pnet(possible_net_t *pnet, struct net *net)
301 {
302 #ifdef CONFIG_NET_NS
303 	pnet->net = net;
304 #endif
305 }
306 
307 static inline struct net *read_pnet(const possible_net_t *pnet)
308 {
309 #ifdef CONFIG_NET_NS
310 	return pnet->net;
311 #else
312 	return &init_net;
313 #endif
314 }
315 
316 /* Protected by net_rwsem */
317 #define for_each_net(VAR)				\
318 	list_for_each_entry(VAR, &net_namespace_list, list)
319 
320 #define for_each_net_rcu(VAR)				\
321 	list_for_each_entry_rcu(VAR, &net_namespace_list, list)
322 
323 #ifdef CONFIG_NET_NS
324 #define __net_init
325 #define __net_exit
326 #define __net_initdata
327 #define __net_initconst
328 #else
329 #define __net_init	__init
330 #define __net_exit	__ref
331 #define __net_initdata	__initdata
332 #define __net_initconst	__initconst
333 #endif
334 
335 int peernet2id_alloc(struct net *net, struct net *peer);
336 int peernet2id(struct net *net, struct net *peer);
337 bool peernet_has_id(struct net *net, struct net *peer);
338 struct net *get_net_ns_by_id(struct net *net, int id);
339 
340 struct pernet_operations {
341 	struct list_head list;
342 	/*
343 	 * Below methods are called without any exclusive locks.
344 	 * More than one net may be constructed and destructed
345 	 * in parallel on several cpus. Every pernet_operations
346 	 * have to keep in mind all other pernet_operations and
347 	 * to introduce a locking, if they share common resources.
348 	 *
349 	 * The only time they are called with exclusive lock is
350 	 * from register_pernet_subsys(), unregister_pernet_subsys()
351 	 * register_pernet_device() and unregister_pernet_device().
352 	 *
353 	 * Exit methods using blocking RCU primitives, such as
354 	 * synchronize_rcu(), should be implemented via exit_batch.
355 	 * Then, destruction of a group of net requires single
356 	 * synchronize_rcu() related to these pernet_operations,
357 	 * instead of separate synchronize_rcu() for every net.
358 	 * Please, avoid synchronize_rcu() at all, where it's possible.
359 	 */
360 	int (*init)(struct net *net);
361 	void (*exit)(struct net *net);
362 	void (*exit_batch)(struct list_head *net_exit_list);
363 	unsigned int *id;
364 	size_t size;
365 };
366 
367 /*
368  * Use these carefully.  If you implement a network device and it
369  * needs per network namespace operations use device pernet operations,
370  * otherwise use pernet subsys operations.
371  *
372  * Network interfaces need to be removed from a dying netns _before_
373  * subsys notifiers can be called, as most of the network code cleanup
374  * (which is done from subsys notifiers) runs with the assumption that
375  * dev_remove_pack has been called so no new packets will arrive during
376  * and after the cleanup functions have been called.  dev_remove_pack
377  * is not per namespace so instead the guarantee of no more packets
378  * arriving in a network namespace is provided by ensuring that all
379  * network devices and all sockets have left the network namespace
380  * before the cleanup methods are called.
381  *
382  * For the longest time the ipv4 icmp code was registered as a pernet
383  * device which caused kernel oops, and panics during network
384  * namespace cleanup.   So please don't get this wrong.
385  */
386 int register_pernet_subsys(struct pernet_operations *);
387 void unregister_pernet_subsys(struct pernet_operations *);
388 int register_pernet_device(struct pernet_operations *);
389 void unregister_pernet_device(struct pernet_operations *);
390 
391 struct ctl_table;
392 struct ctl_table_header;
393 
394 #ifdef CONFIG_SYSCTL
395 int net_sysctl_init(void);
396 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
397 					     struct ctl_table *table);
398 void unregister_net_sysctl_table(struct ctl_table_header *header);
399 #else
400 static inline int net_sysctl_init(void) { return 0; }
401 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
402 	const char *path, struct ctl_table *table)
403 {
404 	return NULL;
405 }
406 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
407 {
408 }
409 #endif
410 
411 static inline int rt_genid_ipv4(struct net *net)
412 {
413 	return atomic_read(&net->ipv4.rt_genid);
414 }
415 
416 static inline void rt_genid_bump_ipv4(struct net *net)
417 {
418 	atomic_inc(&net->ipv4.rt_genid);
419 }
420 
421 extern void (*__fib6_flush_trees)(struct net *net);
422 static inline void rt_genid_bump_ipv6(struct net *net)
423 {
424 	if (__fib6_flush_trees)
425 		__fib6_flush_trees(net);
426 }
427 
428 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
429 static inline struct netns_ieee802154_lowpan *
430 net_ieee802154_lowpan(struct net *net)
431 {
432 	return &net->ieee802154_lowpan;
433 }
434 #endif
435 
436 /* For callers who don't really care about whether it's IPv4 or IPv6 */
437 static inline void rt_genid_bump_all(struct net *net)
438 {
439 	rt_genid_bump_ipv4(net);
440 	rt_genid_bump_ipv6(net);
441 }
442 
443 static inline int fnhe_genid(struct net *net)
444 {
445 	return atomic_read(&net->fnhe_genid);
446 }
447 
448 static inline void fnhe_genid_bump(struct net *net)
449 {
450 	atomic_inc(&net->fnhe_genid);
451 }
452 
453 #endif /* __NET_NET_NAMESPACE_H */
454