xref: /openbmc/linux/include/net/net_namespace.h (revision 2fa49589)
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 <linux/ns_common.h>
35 #include <linux/idr.h>
36 #include <linux/skbuff.h>
37 
38 struct user_namespace;
39 struct proc_dir_entry;
40 struct net_device;
41 struct sock;
42 struct ctl_table_header;
43 struct net_generic;
44 struct uevent_sock;
45 struct netns_ipvs;
46 struct bpf_prog;
47 
48 
49 #define NETDEV_HASHBITS    8
50 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
51 
52 struct net {
53 	refcount_t		passive;	/* To decided when the network
54 						 * namespace should be freed.
55 						 */
56 	refcount_t		count;		/* To decided when the network
57 						 *  namespace should be shut down.
58 						 */
59 	spinlock_t		rules_mod_lock;
60 
61 	atomic64_t		cookie_gen;
62 
63 	struct list_head	list;		/* list of network namespaces */
64 	struct list_head	exit_list;	/* To linked to call pernet exit
65 						 * methods on dead net (
66 						 * pernet_ops_rwsem read locked),
67 						 * or to unregister pernet ops
68 						 * (pernet_ops_rwsem write locked).
69 						 */
70 	struct llist_node	cleanup_list;	/* namespaces on death row */
71 
72 	struct user_namespace   *user_ns;	/* Owning user namespace */
73 	struct ucounts		*ucounts;
74 	spinlock_t		nsid_lock;
75 	struct idr		netns_ids;
76 
77 	struct ns_common	ns;
78 
79 	struct proc_dir_entry 	*proc_net;
80 	struct proc_dir_entry 	*proc_net_stat;
81 
82 #ifdef CONFIG_SYSCTL
83 	struct ctl_table_set	sysctls;
84 #endif
85 
86 	struct sock 		*rtnl;			/* rtnetlink socket */
87 	struct sock		*genl_sock;
88 
89 	struct uevent_sock	*uevent_sock;		/* uevent socket */
90 
91 	struct list_head 	dev_base_head;
92 	struct hlist_head 	*dev_name_head;
93 	struct hlist_head	*dev_index_head;
94 	unsigned int		dev_base_seq;	/* protected by rtnl_mutex */
95 	int			ifindex;
96 	unsigned int		dev_unreg_count;
97 
98 	/* core fib_rules */
99 	struct list_head	rules_ops;
100 
101 	struct list_head	fib_notifier_ops;  /* Populated by
102 						    * register_pernet_subsys()
103 						    */
104 	struct net_device       *loopback_dev;          /* The loopback */
105 	struct netns_core	core;
106 	struct netns_mib	mib;
107 	struct netns_packet	packet;
108 	struct netns_unix	unx;
109 	struct netns_ipv4	ipv4;
110 #if IS_ENABLED(CONFIG_IPV6)
111 	struct netns_ipv6	ipv6;
112 #endif
113 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
114 	struct netns_ieee802154_lowpan	ieee802154_lowpan;
115 #endif
116 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
117 	struct netns_sctp	sctp;
118 #endif
119 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
120 	struct netns_dccp	dccp;
121 #endif
122 #ifdef CONFIG_NETFILTER
123 	struct netns_nf		nf;
124 	struct netns_xt		xt;
125 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
126 	struct netns_ct		ct;
127 #endif
128 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
129 	struct netns_nftables	nft;
130 #endif
131 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
132 	struct netns_nf_frag	nf_frag;
133 	struct ctl_table_header *nf_frag_frags_hdr;
134 #endif
135 	struct sock		*nfnl;
136 	struct sock		*nfnl_stash;
137 #if IS_ENABLED(CONFIG_NETFILTER_NETLINK_ACCT)
138 	struct list_head        nfnl_acct_list;
139 #endif
140 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
141 	struct list_head	nfct_timeout_list;
142 #endif
143 #endif
144 #ifdef CONFIG_WEXT_CORE
145 	struct sk_buff_head	wext_nlevents;
146 #endif
147 	struct net_generic __rcu	*gen;
148 
149 	struct bpf_prog __rcu	*flow_dissector_prog;
150 
151 	/* Note : following structs are cache line aligned */
152 #ifdef CONFIG_XFRM
153 	struct netns_xfrm	xfrm;
154 #endif
155 #if IS_ENABLED(CONFIG_IP_VS)
156 	struct netns_ipvs	*ipvs;
157 #endif
158 #if IS_ENABLED(CONFIG_MPLS)
159 	struct netns_mpls	mpls;
160 #endif
161 #if IS_ENABLED(CONFIG_CAN)
162 	struct netns_can	can;
163 #endif
164 	struct sock		*diag_nlsk;
165 	atomic_t		fnhe_genid;
166 } __randomize_layout;
167 
168 #include <linux/seq_file_net.h>
169 
170 /* Init's network namespace */
171 extern struct net init_net;
172 
173 #ifdef CONFIG_NET_NS
174 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
175 			struct net *old_net);
176 
177 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
178 
179 void net_ns_barrier(void);
180 #else /* CONFIG_NET_NS */
181 #include <linux/sched.h>
182 #include <linux/nsproxy.h>
183 static inline struct net *copy_net_ns(unsigned long flags,
184 	struct user_namespace *user_ns, struct net *old_net)
185 {
186 	if (flags & CLONE_NEWNET)
187 		return ERR_PTR(-EINVAL);
188 	return old_net;
189 }
190 
191 static inline void net_ns_get_ownership(const struct net *net,
192 					kuid_t *uid, kgid_t *gid)
193 {
194 	*uid = GLOBAL_ROOT_UID;
195 	*gid = GLOBAL_ROOT_GID;
196 }
197 
198 static inline void net_ns_barrier(void) {}
199 #endif /* CONFIG_NET_NS */
200 
201 
202 extern struct list_head net_namespace_list;
203 
204 struct net *get_net_ns_by_pid(pid_t pid);
205 struct net *get_net_ns_by_fd(int fd);
206 
207 #ifdef CONFIG_SYSCTL
208 void ipx_register_sysctl(void);
209 void ipx_unregister_sysctl(void);
210 #else
211 #define ipx_register_sysctl()
212 #define ipx_unregister_sysctl()
213 #endif
214 
215 #ifdef CONFIG_NET_NS
216 void __put_net(struct net *net);
217 
218 static inline struct net *get_net(struct net *net)
219 {
220 	refcount_inc(&net->count);
221 	return net;
222 }
223 
224 static inline struct net *maybe_get_net(struct net *net)
225 {
226 	/* Used when we know struct net exists but we
227 	 * aren't guaranteed a previous reference count
228 	 * exists.  If the reference count is zero this
229 	 * function fails and returns NULL.
230 	 */
231 	if (!refcount_inc_not_zero(&net->count))
232 		net = NULL;
233 	return net;
234 }
235 
236 static inline void put_net(struct net *net)
237 {
238 	if (refcount_dec_and_test(&net->count))
239 		__put_net(net);
240 }
241 
242 static inline
243 int net_eq(const struct net *net1, const struct net *net2)
244 {
245 	return net1 == net2;
246 }
247 
248 static inline int check_net(const struct net *net)
249 {
250 	return refcount_read(&net->count) != 0;
251 }
252 
253 void net_drop_ns(void *);
254 
255 #else
256 
257 static inline struct net *get_net(struct net *net)
258 {
259 	return net;
260 }
261 
262 static inline void put_net(struct net *net)
263 {
264 }
265 
266 static inline struct net *maybe_get_net(struct net *net)
267 {
268 	return net;
269 }
270 
271 static inline
272 int net_eq(const struct net *net1, const struct net *net2)
273 {
274 	return 1;
275 }
276 
277 static inline int check_net(const struct net *net)
278 {
279 	return 1;
280 }
281 
282 #define net_drop_ns NULL
283 #endif
284 
285 
286 typedef struct {
287 #ifdef CONFIG_NET_NS
288 	struct net *net;
289 #endif
290 } possible_net_t;
291 
292 static inline void write_pnet(possible_net_t *pnet, struct net *net)
293 {
294 #ifdef CONFIG_NET_NS
295 	pnet->net = net;
296 #endif
297 }
298 
299 static inline struct net *read_pnet(const possible_net_t *pnet)
300 {
301 #ifdef CONFIG_NET_NS
302 	return pnet->net;
303 #else
304 	return &init_net;
305 #endif
306 }
307 
308 /* Protected by net_rwsem */
309 #define for_each_net(VAR)				\
310 	list_for_each_entry(VAR, &net_namespace_list, list)
311 
312 #define for_each_net_rcu(VAR)				\
313 	list_for_each_entry_rcu(VAR, &net_namespace_list, list)
314 
315 #ifdef CONFIG_NET_NS
316 #define __net_init
317 #define __net_exit
318 #define __net_initdata
319 #define __net_initconst
320 #else
321 #define __net_init	__init
322 #define __net_exit	__ref
323 #define __net_initdata	__initdata
324 #define __net_initconst	__initconst
325 #endif
326 
327 int peernet2id_alloc(struct net *net, struct net *peer);
328 int peernet2id(struct net *net, struct net *peer);
329 bool peernet_has_id(struct net *net, struct net *peer);
330 struct net *get_net_ns_by_id(struct net *net, int id);
331 
332 struct pernet_operations {
333 	struct list_head list;
334 	/*
335 	 * Below methods are called without any exclusive locks.
336 	 * More than one net may be constructed and destructed
337 	 * in parallel on several cpus. Every pernet_operations
338 	 * have to keep in mind all other pernet_operations and
339 	 * to introduce a locking, if they share common resources.
340 	 *
341 	 * The only time they are called with exclusive lock is
342 	 * from register_pernet_subsys(), unregister_pernet_subsys()
343 	 * register_pernet_device() and unregister_pernet_device().
344 	 *
345 	 * Exit methods using blocking RCU primitives, such as
346 	 * synchronize_rcu(), should be implemented via exit_batch.
347 	 * Then, destruction of a group of net requires single
348 	 * synchronize_rcu() related to these pernet_operations,
349 	 * instead of separate synchronize_rcu() for every net.
350 	 * Please, avoid synchronize_rcu() at all, where it's possible.
351 	 */
352 	int (*init)(struct net *net);
353 	void (*exit)(struct net *net);
354 	void (*exit_batch)(struct list_head *net_exit_list);
355 	unsigned int *id;
356 	size_t size;
357 };
358 
359 /*
360  * Use these carefully.  If you implement a network device and it
361  * needs per network namespace operations use device pernet operations,
362  * otherwise use pernet subsys operations.
363  *
364  * Network interfaces need to be removed from a dying netns _before_
365  * subsys notifiers can be called, as most of the network code cleanup
366  * (which is done from subsys notifiers) runs with the assumption that
367  * dev_remove_pack has been called so no new packets will arrive during
368  * and after the cleanup functions have been called.  dev_remove_pack
369  * is not per namespace so instead the guarantee of no more packets
370  * arriving in a network namespace is provided by ensuring that all
371  * network devices and all sockets have left the network namespace
372  * before the cleanup methods are called.
373  *
374  * For the longest time the ipv4 icmp code was registered as a pernet
375  * device which caused kernel oops, and panics during network
376  * namespace cleanup.   So please don't get this wrong.
377  */
378 int register_pernet_subsys(struct pernet_operations *);
379 void unregister_pernet_subsys(struct pernet_operations *);
380 int register_pernet_device(struct pernet_operations *);
381 void unregister_pernet_device(struct pernet_operations *);
382 
383 struct ctl_table;
384 struct ctl_table_header;
385 
386 #ifdef CONFIG_SYSCTL
387 int net_sysctl_init(void);
388 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
389 					     struct ctl_table *table);
390 void unregister_net_sysctl_table(struct ctl_table_header *header);
391 #else
392 static inline int net_sysctl_init(void) { return 0; }
393 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
394 	const char *path, struct ctl_table *table)
395 {
396 	return NULL;
397 }
398 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
399 {
400 }
401 #endif
402 
403 static inline int rt_genid_ipv4(struct net *net)
404 {
405 	return atomic_read(&net->ipv4.rt_genid);
406 }
407 
408 static inline void rt_genid_bump_ipv4(struct net *net)
409 {
410 	atomic_inc(&net->ipv4.rt_genid);
411 }
412 
413 extern void (*__fib6_flush_trees)(struct net *net);
414 static inline void rt_genid_bump_ipv6(struct net *net)
415 {
416 	if (__fib6_flush_trees)
417 		__fib6_flush_trees(net);
418 }
419 
420 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
421 static inline struct netns_ieee802154_lowpan *
422 net_ieee802154_lowpan(struct net *net)
423 {
424 	return &net->ieee802154_lowpan;
425 }
426 #endif
427 
428 /* For callers who don't really care about whether it's IPv4 or IPv6 */
429 static inline void rt_genid_bump_all(struct net *net)
430 {
431 	rt_genid_bump_ipv4(net);
432 	rt_genid_bump_ipv6(net);
433 }
434 
435 static inline int fnhe_genid(struct net *net)
436 {
437 	return atomic_read(&net->fnhe_genid);
438 }
439 
440 static inline void fnhe_genid_bump(struct net *net)
441 {
442 	atomic_inc(&net->fnhe_genid);
443 }
444 
445 #endif /* __NET_NET_NAMESPACE_H */
446