xref: /linux/include/net/net_namespace.h (revision 50f2944009a25bb39a09f2f7bab64a73ce928bef)
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 #include <net/netns/nftables.h>
30 #include <net/netns/xfrm.h>
31 #include <net/netns/mpls.h>
32 #include <net/netns/can.h>
33 #include <net/netns/xdp.h>
34 #include <net/netns/smc.h>
35 #include <net/netns/bpf.h>
36 #include <net/netns/mctp.h>
37 #include <net/net_trackers.h>
38 #include <linux/ns_common.h>
39 #include <linux/idr.h>
40 #include <linux/skbuff.h>
41 #include <linux/notifier.h>
42 
43 struct user_namespace;
44 struct proc_dir_entry;
45 struct net_device;
46 struct sock;
47 struct ctl_table_header;
48 struct net_generic;
49 struct uevent_sock;
50 struct netns_ipvs;
51 struct bpf_prog;
52 
53 
54 #define NETDEV_HASHBITS    8
55 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
56 
57 struct net {
58 	/* First cache line can be often dirtied.
59 	 * Do not place here read-mostly fields.
60 	 */
61 	refcount_t		passive;	/* To decide when the network
62 						 * namespace should be freed.
63 						 */
64 	spinlock_t		rules_mod_lock;
65 
66 	atomic_t		dev_unreg_count;
67 
68 	unsigned int		dev_base_seq;	/* protected by rtnl_mutex */
69 	int			ifindex;
70 
71 	spinlock_t		nsid_lock;
72 	atomic_t		fnhe_genid;
73 
74 	struct list_head	list;		/* list of network namespaces */
75 	struct list_head	exit_list;	/* To linked to call pernet exit
76 						 * methods on dead net (
77 						 * pernet_ops_rwsem read locked),
78 						 * or to unregister pernet ops
79 						 * (pernet_ops_rwsem write locked).
80 						 */
81 	struct llist_node	cleanup_list;	/* namespaces on death row */
82 
83 #ifdef CONFIG_KEYS
84 	struct key_tag		*key_domain;	/* Key domain of operation tag */
85 #endif
86 	struct user_namespace   *user_ns;	/* Owning user namespace */
87 	struct ucounts		*ucounts;
88 	struct idr		netns_ids;
89 
90 	struct ns_common	ns;
91 	struct ref_tracker_dir  refcnt_tracker;
92 
93 	struct list_head 	dev_base_head;
94 	struct proc_dir_entry 	*proc_net;
95 	struct proc_dir_entry 	*proc_net_stat;
96 
97 #ifdef CONFIG_SYSCTL
98 	struct ctl_table_set	sysctls;
99 #endif
100 
101 	struct sock 		*rtnl;			/* rtnetlink socket */
102 	struct sock		*genl_sock;
103 
104 	struct uevent_sock	*uevent_sock;		/* uevent socket */
105 
106 	struct hlist_head 	*dev_name_head;
107 	struct hlist_head	*dev_index_head;
108 	struct raw_notifier_head	netdev_chain;
109 
110 	/* Note that @hash_mix can be read millions times per second,
111 	 * it is critical that it is on a read_mostly cache line.
112 	 */
113 	u32			hash_mix;
114 
115 	struct net_device       *loopback_dev;          /* The loopback */
116 
117 	/* core fib_rules */
118 	struct list_head	rules_ops;
119 
120 	struct netns_core	core;
121 	struct netns_mib	mib;
122 	struct netns_packet	packet;
123 #if IS_ENABLED(CONFIG_UNIX)
124 	struct netns_unix	unx;
125 #endif
126 	struct netns_nexthop	nexthop;
127 	struct netns_ipv4	ipv4;
128 #if IS_ENABLED(CONFIG_IPV6)
129 	struct netns_ipv6	ipv6;
130 #endif
131 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
132 	struct netns_ieee802154_lowpan	ieee802154_lowpan;
133 #endif
134 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
135 	struct netns_sctp	sctp;
136 #endif
137 #ifdef CONFIG_NETFILTER
138 	struct netns_nf		nf;
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 #endif
146 #ifdef CONFIG_WEXT_CORE
147 	struct sk_buff_head	wext_nlevents;
148 #endif
149 	struct net_generic __rcu	*gen;
150 
151 	/* Used to store attached BPF programs */
152 	struct netns_bpf	bpf;
153 
154 	/* Note : following structs are cache line aligned */
155 #ifdef CONFIG_XFRM
156 	struct netns_xfrm	xfrm;
157 #endif
158 
159 	u64			net_cookie; /* written once */
160 
161 #if IS_ENABLED(CONFIG_IP_VS)
162 	struct netns_ipvs	*ipvs;
163 #endif
164 #if IS_ENABLED(CONFIG_MPLS)
165 	struct netns_mpls	mpls;
166 #endif
167 #if IS_ENABLED(CONFIG_CAN)
168 	struct netns_can	can;
169 #endif
170 #ifdef CONFIG_XDP_SOCKETS
171 	struct netns_xdp	xdp;
172 #endif
173 #if IS_ENABLED(CONFIG_MCTP)
174 	struct netns_mctp	mctp;
175 #endif
176 #if IS_ENABLED(CONFIG_CRYPTO_USER)
177 	struct sock		*crypto_nlsk;
178 #endif
179 	struct sock		*diag_nlsk;
180 #if IS_ENABLED(CONFIG_SMC)
181 	struct netns_smc	smc;
182 #endif
183 } __randomize_layout;
184 
185 #include <linux/seq_file_net.h>
186 
187 /* Init's network namespace */
188 extern struct net init_net;
189 
190 #ifdef CONFIG_NET_NS
191 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
192 			struct net *old_net);
193 
194 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
195 
196 void net_ns_barrier(void);
197 
198 struct ns_common *get_net_ns(struct ns_common *ns);
199 struct net *get_net_ns_by_fd(int fd);
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 
220 static inline struct ns_common *get_net_ns(struct ns_common *ns)
221 {
222 	return ERR_PTR(-EINVAL);
223 }
224 
225 static inline struct net *get_net_ns_by_fd(int fd)
226 {
227 	return ERR_PTR(-EINVAL);
228 }
229 #endif /* CONFIG_NET_NS */
230 
231 
232 extern struct list_head net_namespace_list;
233 
234 struct net *get_net_ns_by_pid(pid_t pid);
235 
236 #ifdef CONFIG_SYSCTL
237 void ipx_register_sysctl(void);
238 void ipx_unregister_sysctl(void);
239 #else
240 #define ipx_register_sysctl()
241 #define ipx_unregister_sysctl()
242 #endif
243 
244 #ifdef CONFIG_NET_NS
245 void __put_net(struct net *net);
246 
247 /* Try using get_net_track() instead */
248 static inline struct net *get_net(struct net *net)
249 {
250 	refcount_inc(&net->ns.count);
251 	return net;
252 }
253 
254 static inline struct net *maybe_get_net(struct net *net)
255 {
256 	/* Used when we know struct net exists but we
257 	 * aren't guaranteed a previous reference count
258 	 * exists.  If the reference count is zero this
259 	 * function fails and returns NULL.
260 	 */
261 	if (!refcount_inc_not_zero(&net->ns.count))
262 		net = NULL;
263 	return net;
264 }
265 
266 /* Try using put_net_track() instead */
267 static inline void put_net(struct net *net)
268 {
269 	if (refcount_dec_and_test(&net->ns.count))
270 		__put_net(net);
271 }
272 
273 static inline
274 int net_eq(const struct net *net1, const struct net *net2)
275 {
276 	return net1 == net2;
277 }
278 
279 static inline int check_net(const struct net *net)
280 {
281 	return refcount_read(&net->ns.count) != 0;
282 }
283 
284 void net_drop_ns(void *);
285 
286 #else
287 
288 static inline struct net *get_net(struct net *net)
289 {
290 	return net;
291 }
292 
293 static inline void put_net(struct net *net)
294 {
295 }
296 
297 static inline struct net *maybe_get_net(struct net *net)
298 {
299 	return net;
300 }
301 
302 static inline
303 int net_eq(const struct net *net1, const struct net *net2)
304 {
305 	return 1;
306 }
307 
308 static inline int check_net(const struct net *net)
309 {
310 	return 1;
311 }
312 
313 #define net_drop_ns NULL
314 #endif
315 
316 
317 static inline void netns_tracker_alloc(struct net *net,
318 				       netns_tracker *tracker, gfp_t gfp)
319 {
320 #ifdef CONFIG_NET_NS_REFCNT_TRACKER
321 	ref_tracker_alloc(&net->refcnt_tracker, tracker, gfp);
322 #endif
323 }
324 
325 static inline void netns_tracker_free(struct net *net,
326 				      netns_tracker *tracker)
327 {
328 #ifdef CONFIG_NET_NS_REFCNT_TRACKER
329        ref_tracker_free(&net->refcnt_tracker, tracker);
330 #endif
331 }
332 
333 static inline struct net *get_net_track(struct net *net,
334 					netns_tracker *tracker, gfp_t gfp)
335 {
336 	get_net(net);
337 	netns_tracker_alloc(net, tracker, gfp);
338 	return net;
339 }
340 
341 static inline void put_net_track(struct net *net, netns_tracker *tracker)
342 {
343 	netns_tracker_free(net, tracker);
344 	put_net(net);
345 }
346 
347 typedef struct {
348 #ifdef CONFIG_NET_NS
349 	struct net *net;
350 #endif
351 } possible_net_t;
352 
353 static inline void write_pnet(possible_net_t *pnet, struct net *net)
354 {
355 #ifdef CONFIG_NET_NS
356 	pnet->net = net;
357 #endif
358 }
359 
360 static inline struct net *read_pnet(const possible_net_t *pnet)
361 {
362 #ifdef CONFIG_NET_NS
363 	return pnet->net;
364 #else
365 	return &init_net;
366 #endif
367 }
368 
369 /* Protected by net_rwsem */
370 #define for_each_net(VAR)				\
371 	list_for_each_entry(VAR, &net_namespace_list, list)
372 #define for_each_net_continue_reverse(VAR)		\
373 	list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list)
374 #define for_each_net_rcu(VAR)				\
375 	list_for_each_entry_rcu(VAR, &net_namespace_list, list)
376 
377 #ifdef CONFIG_NET_NS
378 #define __net_init
379 #define __net_exit
380 #define __net_initdata
381 #define __net_initconst
382 #else
383 #define __net_init	__init
384 #define __net_exit	__ref
385 #define __net_initdata	__initdata
386 #define __net_initconst	__initconst
387 #endif
388 
389 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp);
390 int peernet2id(const struct net *net, struct net *peer);
391 bool peernet_has_id(const struct net *net, struct net *peer);
392 struct net *get_net_ns_by_id(const struct net *net, int id);
393 
394 struct pernet_operations {
395 	struct list_head list;
396 	/*
397 	 * Below methods are called without any exclusive locks.
398 	 * More than one net may be constructed and destructed
399 	 * in parallel on several cpus. Every pernet_operations
400 	 * have to keep in mind all other pernet_operations and
401 	 * to introduce a locking, if they share common resources.
402 	 *
403 	 * The only time they are called with exclusive lock is
404 	 * from register_pernet_subsys(), unregister_pernet_subsys()
405 	 * register_pernet_device() and unregister_pernet_device().
406 	 *
407 	 * Exit methods using blocking RCU primitives, such as
408 	 * synchronize_rcu(), should be implemented via exit_batch.
409 	 * Then, destruction of a group of net requires single
410 	 * synchronize_rcu() related to these pernet_operations,
411 	 * instead of separate synchronize_rcu() for every net.
412 	 * Please, avoid synchronize_rcu() at all, where it's possible.
413 	 *
414 	 * Note that a combination of pre_exit() and exit() can
415 	 * be used, since a synchronize_rcu() is guaranteed between
416 	 * the calls.
417 	 */
418 	int (*init)(struct net *net);
419 	void (*pre_exit)(struct net *net);
420 	void (*exit)(struct net *net);
421 	void (*exit_batch)(struct list_head *net_exit_list);
422 	unsigned int *id;
423 	size_t size;
424 };
425 
426 /*
427  * Use these carefully.  If you implement a network device and it
428  * needs per network namespace operations use device pernet operations,
429  * otherwise use pernet subsys operations.
430  *
431  * Network interfaces need to be removed from a dying netns _before_
432  * subsys notifiers can be called, as most of the network code cleanup
433  * (which is done from subsys notifiers) runs with the assumption that
434  * dev_remove_pack has been called so no new packets will arrive during
435  * and after the cleanup functions have been called.  dev_remove_pack
436  * is not per namespace so instead the guarantee of no more packets
437  * arriving in a network namespace is provided by ensuring that all
438  * network devices and all sockets have left the network namespace
439  * before the cleanup methods are called.
440  *
441  * For the longest time the ipv4 icmp code was registered as a pernet
442  * device which caused kernel oops, and panics during network
443  * namespace cleanup.   So please don't get this wrong.
444  */
445 int register_pernet_subsys(struct pernet_operations *);
446 void unregister_pernet_subsys(struct pernet_operations *);
447 int register_pernet_device(struct pernet_operations *);
448 void unregister_pernet_device(struct pernet_operations *);
449 
450 struct ctl_table;
451 
452 #ifdef CONFIG_SYSCTL
453 int net_sysctl_init(void);
454 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
455 					     struct ctl_table *table);
456 void unregister_net_sysctl_table(struct ctl_table_header *header);
457 #else
458 static inline int net_sysctl_init(void) { return 0; }
459 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
460 	const char *path, struct ctl_table *table)
461 {
462 	return NULL;
463 }
464 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
465 {
466 }
467 #endif
468 
469 static inline int rt_genid_ipv4(const struct net *net)
470 {
471 	return atomic_read(&net->ipv4.rt_genid);
472 }
473 
474 #if IS_ENABLED(CONFIG_IPV6)
475 static inline int rt_genid_ipv6(const struct net *net)
476 {
477 	return atomic_read(&net->ipv6.fib6_sernum);
478 }
479 #endif
480 
481 static inline void rt_genid_bump_ipv4(struct net *net)
482 {
483 	atomic_inc(&net->ipv4.rt_genid);
484 }
485 
486 extern void (*__fib6_flush_trees)(struct net *net);
487 static inline void rt_genid_bump_ipv6(struct net *net)
488 {
489 	if (__fib6_flush_trees)
490 		__fib6_flush_trees(net);
491 }
492 
493 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
494 static inline struct netns_ieee802154_lowpan *
495 net_ieee802154_lowpan(struct net *net)
496 {
497 	return &net->ieee802154_lowpan;
498 }
499 #endif
500 
501 /* For callers who don't really care about whether it's IPv4 or IPv6 */
502 static inline void rt_genid_bump_all(struct net *net)
503 {
504 	rt_genid_bump_ipv4(net);
505 	rt_genid_bump_ipv6(net);
506 }
507 
508 static inline int fnhe_genid(const struct net *net)
509 {
510 	return atomic_read(&net->fnhe_genid);
511 }
512 
513 static inline void fnhe_genid_bump(struct net *net)
514 {
515 	atomic_inc(&net->fnhe_genid);
516 }
517 
518 #ifdef CONFIG_NET
519 void net_ns_init(void);
520 #else
521 static inline void net_ns_init(void) {}
522 #endif
523 
524 #endif /* __NET_NET_NAMESPACE_H */
525