xref: /linux/include/net/net_namespace.h (revision c79c3c34f75d72a066e292b10aa50fc758c97c89)
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/dccp.h>
26 #include <net/netns/netfilter.h>
27 #include <net/netns/x_tables.h>
28 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
29 #include <net/netns/conntrack.h>
30 #endif
31 #include <net/netns/nftables.h>
32 #include <net/netns/xfrm.h>
33 #include <net/netns/mpls.h>
34 #include <net/netns/can.h>
35 #include <net/netns/xdp.h>
36 #include <net/netns/bpf.h>
37 #include <linux/ns_common.h>
38 #include <linux/idr.h>
39 #include <linux/skbuff.h>
40 #include <linux/notifier.h>
41 
42 struct user_namespace;
43 struct proc_dir_entry;
44 struct net_device;
45 struct sock;
46 struct ctl_table_header;
47 struct net_generic;
48 struct uevent_sock;
49 struct netns_ipvs;
50 struct bpf_prog;
51 
52 
53 #define NETDEV_HASHBITS    8
54 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
55 
56 struct net {
57 	/* First cache line can be often dirtied.
58 	 * Do not place here read-mostly fields.
59 	 */
60 	refcount_t		passive;	/* To decide when the network
61 						 * namespace should be freed.
62 						 */
63 	spinlock_t		rules_mod_lock;
64 
65 	unsigned int		dev_unreg_count;
66 
67 	unsigned int		dev_base_seq;	/* protected by rtnl_mutex */
68 	int			ifindex;
69 
70 	spinlock_t		nsid_lock;
71 	atomic_t		fnhe_genid;
72 
73 	struct list_head	list;		/* list of network namespaces */
74 	struct list_head	exit_list;	/* To linked to call pernet exit
75 						 * methods on dead net (
76 						 * pernet_ops_rwsem read locked),
77 						 * or to unregister pernet ops
78 						 * (pernet_ops_rwsem write locked).
79 						 */
80 	struct llist_node	cleanup_list;	/* namespaces on death row */
81 
82 #ifdef CONFIG_KEYS
83 	struct key_tag		*key_domain;	/* Key domain of operation tag */
84 #endif
85 	struct user_namespace   *user_ns;	/* Owning user namespace */
86 	struct ucounts		*ucounts;
87 	struct idr		netns_ids;
88 
89 	struct ns_common	ns;
90 
91 	struct list_head 	dev_base_head;
92 	struct proc_dir_entry 	*proc_net;
93 	struct proc_dir_entry 	*proc_net_stat;
94 
95 #ifdef CONFIG_SYSCTL
96 	struct ctl_table_set	sysctls;
97 #endif
98 
99 	struct sock 		*rtnl;			/* rtnetlink socket */
100 	struct sock		*genl_sock;
101 
102 	struct uevent_sock	*uevent_sock;		/* uevent socket */
103 
104 	struct hlist_head 	*dev_name_head;
105 	struct hlist_head	*dev_index_head;
106 	struct raw_notifier_head	netdev_chain;
107 
108 	/* Note that @hash_mix can be read millions times per second,
109 	 * it is critical that it is on a read_mostly cache line.
110 	 */
111 	u32			hash_mix;
112 
113 	struct net_device       *loopback_dev;          /* The loopback */
114 
115 	/* core fib_rules */
116 	struct list_head	rules_ops;
117 
118 	struct netns_core	core;
119 	struct netns_mib	mib;
120 	struct netns_packet	packet;
121 	struct netns_unix	unx;
122 	struct netns_nexthop	nexthop;
123 	struct netns_ipv4	ipv4;
124 #if IS_ENABLED(CONFIG_IPV6)
125 	struct netns_ipv6	ipv6;
126 #endif
127 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
128 	struct netns_ieee802154_lowpan	ieee802154_lowpan;
129 #endif
130 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
131 	struct netns_sctp	sctp;
132 #endif
133 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
134 	struct netns_dccp	dccp;
135 #endif
136 #ifdef CONFIG_NETFILTER
137 	struct netns_nf		nf;
138 	struct netns_xt		xt;
139 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
140 	struct netns_ct		ct;
141 #endif
142 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
143 	struct netns_nftables	nft;
144 #endif
145 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
146 	struct netns_nf_frag	nf_frag;
147 	struct ctl_table_header *nf_frag_frags_hdr;
148 #endif
149 	struct sock		*nfnl;
150 	struct sock		*nfnl_stash;
151 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
152 	struct list_head	nfct_timeout_list;
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_CRYPTO_USER)
183 	struct sock		*crypto_nlsk;
184 #endif
185 	struct sock		*diag_nlsk;
186 } __randomize_layout;
187 
188 #include <linux/seq_file_net.h>
189 
190 /* Init's network namespace */
191 extern struct net init_net;
192 
193 #ifdef CONFIG_NET_NS
194 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
195 			struct net *old_net);
196 
197 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
198 
199 void net_ns_barrier(void);
200 #else /* CONFIG_NET_NS */
201 #include <linux/sched.h>
202 #include <linux/nsproxy.h>
203 static inline struct net *copy_net_ns(unsigned long flags,
204 	struct user_namespace *user_ns, struct net *old_net)
205 {
206 	if (flags & CLONE_NEWNET)
207 		return ERR_PTR(-EINVAL);
208 	return old_net;
209 }
210 
211 static inline void net_ns_get_ownership(const struct net *net,
212 					kuid_t *uid, kgid_t *gid)
213 {
214 	*uid = GLOBAL_ROOT_UID;
215 	*gid = GLOBAL_ROOT_GID;
216 }
217 
218 static inline void net_ns_barrier(void) {}
219 #endif /* CONFIG_NET_NS */
220 
221 
222 extern struct list_head net_namespace_list;
223 
224 struct net *get_net_ns_by_pid(pid_t pid);
225 struct net *get_net_ns_by_fd(int fd);
226 
227 #ifdef CONFIG_SYSCTL
228 void ipx_register_sysctl(void);
229 void ipx_unregister_sysctl(void);
230 #else
231 #define ipx_register_sysctl()
232 #define ipx_unregister_sysctl()
233 #endif
234 
235 #ifdef CONFIG_NET_NS
236 void __put_net(struct net *net);
237 
238 static inline struct net *get_net(struct net *net)
239 {
240 	refcount_inc(&net->ns.count);
241 	return net;
242 }
243 
244 static inline struct net *maybe_get_net(struct net *net)
245 {
246 	/* Used when we know struct net exists but we
247 	 * aren't guaranteed a previous reference count
248 	 * exists.  If the reference count is zero this
249 	 * function fails and returns NULL.
250 	 */
251 	if (!refcount_inc_not_zero(&net->ns.count))
252 		net = NULL;
253 	return net;
254 }
255 
256 static inline void put_net(struct net *net)
257 {
258 	if (refcount_dec_and_test(&net->ns.count))
259 		__put_net(net);
260 }
261 
262 static inline
263 int net_eq(const struct net *net1, const struct net *net2)
264 {
265 	return net1 == net2;
266 }
267 
268 static inline int check_net(const struct net *net)
269 {
270 	return refcount_read(&net->ns.count) != 0;
271 }
272 
273 void net_drop_ns(void *);
274 
275 #else
276 
277 static inline struct net *get_net(struct net *net)
278 {
279 	return net;
280 }
281 
282 static inline void put_net(struct net *net)
283 {
284 }
285 
286 static inline struct net *maybe_get_net(struct net *net)
287 {
288 	return net;
289 }
290 
291 static inline
292 int net_eq(const struct net *net1, const struct net *net2)
293 {
294 	return 1;
295 }
296 
297 static inline int check_net(const struct net *net)
298 {
299 	return 1;
300 }
301 
302 #define net_drop_ns NULL
303 #endif
304 
305 
306 typedef struct {
307 #ifdef CONFIG_NET_NS
308 	struct net *net;
309 #endif
310 } possible_net_t;
311 
312 static inline void write_pnet(possible_net_t *pnet, struct net *net)
313 {
314 #ifdef CONFIG_NET_NS
315 	pnet->net = net;
316 #endif
317 }
318 
319 static inline struct net *read_pnet(const possible_net_t *pnet)
320 {
321 #ifdef CONFIG_NET_NS
322 	return pnet->net;
323 #else
324 	return &init_net;
325 #endif
326 }
327 
328 /* Protected by net_rwsem */
329 #define for_each_net(VAR)				\
330 	list_for_each_entry(VAR, &net_namespace_list, list)
331 #define for_each_net_continue_reverse(VAR)		\
332 	list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list)
333 #define for_each_net_rcu(VAR)				\
334 	list_for_each_entry_rcu(VAR, &net_namespace_list, list)
335 
336 #ifdef CONFIG_NET_NS
337 #define __net_init
338 #define __net_exit
339 #define __net_initdata
340 #define __net_initconst
341 #else
342 #define __net_init	__init
343 #define __net_exit	__ref
344 #define __net_initdata	__initdata
345 #define __net_initconst	__initconst
346 #endif
347 
348 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp);
349 int peernet2id(const struct net *net, struct net *peer);
350 bool peernet_has_id(const struct net *net, struct net *peer);
351 struct net *get_net_ns_by_id(const struct net *net, int id);
352 
353 struct pernet_operations {
354 	struct list_head list;
355 	/*
356 	 * Below methods are called without any exclusive locks.
357 	 * More than one net may be constructed and destructed
358 	 * in parallel on several cpus. Every pernet_operations
359 	 * have to keep in mind all other pernet_operations and
360 	 * to introduce a locking, if they share common resources.
361 	 *
362 	 * The only time they are called with exclusive lock is
363 	 * from register_pernet_subsys(), unregister_pernet_subsys()
364 	 * register_pernet_device() and unregister_pernet_device().
365 	 *
366 	 * Exit methods using blocking RCU primitives, such as
367 	 * synchronize_rcu(), should be implemented via exit_batch.
368 	 * Then, destruction of a group of net requires single
369 	 * synchronize_rcu() related to these pernet_operations,
370 	 * instead of separate synchronize_rcu() for every net.
371 	 * Please, avoid synchronize_rcu() at all, where it's possible.
372 	 *
373 	 * Note that a combination of pre_exit() and exit() can
374 	 * be used, since a synchronize_rcu() is guaranteed between
375 	 * the calls.
376 	 */
377 	int (*init)(struct net *net);
378 	void (*pre_exit)(struct net *net);
379 	void (*exit)(struct net *net);
380 	void (*exit_batch)(struct list_head *net_exit_list);
381 	unsigned int *id;
382 	size_t size;
383 };
384 
385 /*
386  * Use these carefully.  If you implement a network device and it
387  * needs per network namespace operations use device pernet operations,
388  * otherwise use pernet subsys operations.
389  *
390  * Network interfaces need to be removed from a dying netns _before_
391  * subsys notifiers can be called, as most of the network code cleanup
392  * (which is done from subsys notifiers) runs with the assumption that
393  * dev_remove_pack has been called so no new packets will arrive during
394  * and after the cleanup functions have been called.  dev_remove_pack
395  * is not per namespace so instead the guarantee of no more packets
396  * arriving in a network namespace is provided by ensuring that all
397  * network devices and all sockets have left the network namespace
398  * before the cleanup methods are called.
399  *
400  * For the longest time the ipv4 icmp code was registered as a pernet
401  * device which caused kernel oops, and panics during network
402  * namespace cleanup.   So please don't get this wrong.
403  */
404 int register_pernet_subsys(struct pernet_operations *);
405 void unregister_pernet_subsys(struct pernet_operations *);
406 int register_pernet_device(struct pernet_operations *);
407 void unregister_pernet_device(struct pernet_operations *);
408 
409 struct ctl_table;
410 struct ctl_table_header;
411 
412 #ifdef CONFIG_SYSCTL
413 int net_sysctl_init(void);
414 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
415 					     struct ctl_table *table);
416 void unregister_net_sysctl_table(struct ctl_table_header *header);
417 #else
418 static inline int net_sysctl_init(void) { return 0; }
419 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
420 	const char *path, struct ctl_table *table)
421 {
422 	return NULL;
423 }
424 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
425 {
426 }
427 #endif
428 
429 static inline int rt_genid_ipv4(const struct net *net)
430 {
431 	return atomic_read(&net->ipv4.rt_genid);
432 }
433 
434 #if IS_ENABLED(CONFIG_IPV6)
435 static inline int rt_genid_ipv6(const struct net *net)
436 {
437 	return atomic_read(&net->ipv6.fib6_sernum);
438 }
439 #endif
440 
441 static inline void rt_genid_bump_ipv4(struct net *net)
442 {
443 	atomic_inc(&net->ipv4.rt_genid);
444 }
445 
446 extern void (*__fib6_flush_trees)(struct net *net);
447 static inline void rt_genid_bump_ipv6(struct net *net)
448 {
449 	if (__fib6_flush_trees)
450 		__fib6_flush_trees(net);
451 }
452 
453 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
454 static inline struct netns_ieee802154_lowpan *
455 net_ieee802154_lowpan(struct net *net)
456 {
457 	return &net->ieee802154_lowpan;
458 }
459 #endif
460 
461 /* For callers who don't really care about whether it's IPv4 or IPv6 */
462 static inline void rt_genid_bump_all(struct net *net)
463 {
464 	rt_genid_bump_ipv4(net);
465 	rt_genid_bump_ipv6(net);
466 }
467 
468 static inline int fnhe_genid(const struct net *net)
469 {
470 	return atomic_read(&net->fnhe_genid);
471 }
472 
473 static inline void fnhe_genid_bump(struct net *net)
474 {
475 	atomic_inc(&net->fnhe_genid);
476 }
477 
478 #endif /* __NET_NET_NAMESPACE_H */
479