xref: /openbmc/linux/include/net/net_namespace.h (revision d2574c33)
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 	atomic64_t		cookie_gen;
63 
64 	struct list_head	list;		/* list of network namespaces */
65 	struct list_head	exit_list;	/* To linked to call pernet exit
66 						 * methods on dead net (
67 						 * pernet_ops_rwsem read locked),
68 						 * or to unregister pernet ops
69 						 * (pernet_ops_rwsem write locked).
70 						 */
71 	struct llist_node	cleanup_list;	/* namespaces on death row */
72 
73 	struct user_namespace   *user_ns;	/* Owning user namespace */
74 	struct ucounts		*ucounts;
75 	spinlock_t		nsid_lock;
76 	struct idr		netns_ids;
77 
78 	struct ns_common	ns;
79 
80 	struct proc_dir_entry 	*proc_net;
81 	struct proc_dir_entry 	*proc_net_stat;
82 
83 #ifdef CONFIG_SYSCTL
84 	struct ctl_table_set	sysctls;
85 #endif
86 
87 	struct sock 		*rtnl;			/* rtnetlink socket */
88 	struct sock		*genl_sock;
89 
90 	struct uevent_sock	*uevent_sock;		/* uevent socket */
91 
92 	struct list_head 	dev_base_head;
93 	struct hlist_head 	*dev_name_head;
94 	struct hlist_head	*dev_index_head;
95 	unsigned int		dev_base_seq;	/* protected by rtnl_mutex */
96 	int			ifindex;
97 	unsigned int		dev_unreg_count;
98 
99 	/* core fib_rules */
100 	struct list_head	rules_ops;
101 
102 	struct list_head	fib_notifier_ops;  /* Populated by
103 						    * register_pernet_subsys()
104 						    */
105 	struct net_device       *loopback_dev;          /* The loopback */
106 	struct netns_core	core;
107 	struct netns_mib	mib;
108 	struct netns_packet	packet;
109 	struct netns_unix	unx;
110 	struct netns_ipv4	ipv4;
111 #if IS_ENABLED(CONFIG_IPV6)
112 	struct netns_ipv6	ipv6;
113 #endif
114 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
115 	struct netns_ieee802154_lowpan	ieee802154_lowpan;
116 #endif
117 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
118 	struct netns_sctp	sctp;
119 #endif
120 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
121 	struct netns_dccp	dccp;
122 #endif
123 #ifdef CONFIG_NETFILTER
124 	struct netns_nf		nf;
125 	struct netns_xt		xt;
126 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
127 	struct netns_ct		ct;
128 #endif
129 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
130 	struct netns_nftables	nft;
131 #endif
132 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
133 	struct netns_nf_frag	nf_frag;
134 	struct ctl_table_header *nf_frag_frags_hdr;
135 #endif
136 	struct sock		*nfnl;
137 	struct sock		*nfnl_stash;
138 #if IS_ENABLED(CONFIG_NETFILTER_NETLINK_ACCT)
139 	struct list_head        nfnl_acct_list;
140 #endif
141 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
142 	struct list_head	nfct_timeout_list;
143 #endif
144 #endif
145 #ifdef CONFIG_WEXT_CORE
146 	struct sk_buff_head	wext_nlevents;
147 #endif
148 	struct net_generic __rcu	*gen;
149 
150 	struct bpf_prog __rcu	*flow_dissector_prog;
151 
152 	/* Note : following structs are cache line aligned */
153 #ifdef CONFIG_XFRM
154 	struct netns_xfrm	xfrm;
155 #endif
156 #if IS_ENABLED(CONFIG_IP_VS)
157 	struct netns_ipvs	*ipvs;
158 #endif
159 #if IS_ENABLED(CONFIG_MPLS)
160 	struct netns_mpls	mpls;
161 #endif
162 #if IS_ENABLED(CONFIG_CAN)
163 	struct netns_can	can;
164 #endif
165 #ifdef CONFIG_XDP_SOCKETS
166 	struct netns_xdp	xdp;
167 #endif
168 	struct sock		*diag_nlsk;
169 	atomic_t		fnhe_genid;
170 } __randomize_layout;
171 
172 #include <linux/seq_file_net.h>
173 
174 /* Init's network namespace */
175 extern struct net init_net;
176 
177 #ifdef CONFIG_NET_NS
178 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
179 			struct net *old_net);
180 
181 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
182 
183 void net_ns_barrier(void);
184 #else /* CONFIG_NET_NS */
185 #include <linux/sched.h>
186 #include <linux/nsproxy.h>
187 static inline struct net *copy_net_ns(unsigned long flags,
188 	struct user_namespace *user_ns, struct net *old_net)
189 {
190 	if (flags & CLONE_NEWNET)
191 		return ERR_PTR(-EINVAL);
192 	return old_net;
193 }
194 
195 static inline void net_ns_get_ownership(const struct net *net,
196 					kuid_t *uid, kgid_t *gid)
197 {
198 	*uid = GLOBAL_ROOT_UID;
199 	*gid = GLOBAL_ROOT_GID;
200 }
201 
202 static inline void net_ns_barrier(void) {}
203 #endif /* CONFIG_NET_NS */
204 
205 
206 extern struct list_head net_namespace_list;
207 
208 struct net *get_net_ns_by_pid(pid_t pid);
209 struct net *get_net_ns_by_fd(int fd);
210 
211 #ifdef CONFIG_SYSCTL
212 void ipx_register_sysctl(void);
213 void ipx_unregister_sysctl(void);
214 #else
215 #define ipx_register_sysctl()
216 #define ipx_unregister_sysctl()
217 #endif
218 
219 #ifdef CONFIG_NET_NS
220 void __put_net(struct net *net);
221 
222 static inline struct net *get_net(struct net *net)
223 {
224 	refcount_inc(&net->count);
225 	return net;
226 }
227 
228 static inline struct net *maybe_get_net(struct net *net)
229 {
230 	/* Used when we know struct net exists but we
231 	 * aren't guaranteed a previous reference count
232 	 * exists.  If the reference count is zero this
233 	 * function fails and returns NULL.
234 	 */
235 	if (!refcount_inc_not_zero(&net->count))
236 		net = NULL;
237 	return net;
238 }
239 
240 static inline void put_net(struct net *net)
241 {
242 	if (refcount_dec_and_test(&net->count))
243 		__put_net(net);
244 }
245 
246 static inline
247 int net_eq(const struct net *net1, const struct net *net2)
248 {
249 	return net1 == net2;
250 }
251 
252 static inline int check_net(const struct net *net)
253 {
254 	return refcount_read(&net->count) != 0;
255 }
256 
257 void net_drop_ns(void *);
258 
259 #else
260 
261 static inline struct net *get_net(struct net *net)
262 {
263 	return net;
264 }
265 
266 static inline void put_net(struct net *net)
267 {
268 }
269 
270 static inline struct net *maybe_get_net(struct net *net)
271 {
272 	return net;
273 }
274 
275 static inline
276 int net_eq(const struct net *net1, const struct net *net2)
277 {
278 	return 1;
279 }
280 
281 static inline int check_net(const struct net *net)
282 {
283 	return 1;
284 }
285 
286 #define net_drop_ns NULL
287 #endif
288 
289 
290 typedef struct {
291 #ifdef CONFIG_NET_NS
292 	struct net *net;
293 #endif
294 } possible_net_t;
295 
296 static inline void write_pnet(possible_net_t *pnet, struct net *net)
297 {
298 #ifdef CONFIG_NET_NS
299 	pnet->net = net;
300 #endif
301 }
302 
303 static inline struct net *read_pnet(const possible_net_t *pnet)
304 {
305 #ifdef CONFIG_NET_NS
306 	return pnet->net;
307 #else
308 	return &init_net;
309 #endif
310 }
311 
312 /* Protected by net_rwsem */
313 #define for_each_net(VAR)				\
314 	list_for_each_entry(VAR, &net_namespace_list, list)
315 
316 #define for_each_net_rcu(VAR)				\
317 	list_for_each_entry_rcu(VAR, &net_namespace_list, list)
318 
319 #ifdef CONFIG_NET_NS
320 #define __net_init
321 #define __net_exit
322 #define __net_initdata
323 #define __net_initconst
324 #else
325 #define __net_init	__init
326 #define __net_exit	__ref
327 #define __net_initdata	__initdata
328 #define __net_initconst	__initconst
329 #endif
330 
331 int peernet2id_alloc(struct net *net, struct net *peer);
332 int peernet2id(struct net *net, struct net *peer);
333 bool peernet_has_id(struct net *net, struct net *peer);
334 struct net *get_net_ns_by_id(struct net *net, int id);
335 
336 struct pernet_operations {
337 	struct list_head list;
338 	/*
339 	 * Below methods are called without any exclusive locks.
340 	 * More than one net may be constructed and destructed
341 	 * in parallel on several cpus. Every pernet_operations
342 	 * have to keep in mind all other pernet_operations and
343 	 * to introduce a locking, if they share common resources.
344 	 *
345 	 * The only time they are called with exclusive lock is
346 	 * from register_pernet_subsys(), unregister_pernet_subsys()
347 	 * register_pernet_device() and unregister_pernet_device().
348 	 *
349 	 * Exit methods using blocking RCU primitives, such as
350 	 * synchronize_rcu(), should be implemented via exit_batch.
351 	 * Then, destruction of a group of net requires single
352 	 * synchronize_rcu() related to these pernet_operations,
353 	 * instead of separate synchronize_rcu() for every net.
354 	 * Please, avoid synchronize_rcu() at all, where it's possible.
355 	 */
356 	int (*init)(struct net *net);
357 	void (*exit)(struct net *net);
358 	void (*exit_batch)(struct list_head *net_exit_list);
359 	unsigned int *id;
360 	size_t size;
361 };
362 
363 /*
364  * Use these carefully.  If you implement a network device and it
365  * needs per network namespace operations use device pernet operations,
366  * otherwise use pernet subsys operations.
367  *
368  * Network interfaces need to be removed from a dying netns _before_
369  * subsys notifiers can be called, as most of the network code cleanup
370  * (which is done from subsys notifiers) runs with the assumption that
371  * dev_remove_pack has been called so no new packets will arrive during
372  * and after the cleanup functions have been called.  dev_remove_pack
373  * is not per namespace so instead the guarantee of no more packets
374  * arriving in a network namespace is provided by ensuring that all
375  * network devices and all sockets have left the network namespace
376  * before the cleanup methods are called.
377  *
378  * For the longest time the ipv4 icmp code was registered as a pernet
379  * device which caused kernel oops, and panics during network
380  * namespace cleanup.   So please don't get this wrong.
381  */
382 int register_pernet_subsys(struct pernet_operations *);
383 void unregister_pernet_subsys(struct pernet_operations *);
384 int register_pernet_device(struct pernet_operations *);
385 void unregister_pernet_device(struct pernet_operations *);
386 
387 struct ctl_table;
388 struct ctl_table_header;
389 
390 #ifdef CONFIG_SYSCTL
391 int net_sysctl_init(void);
392 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
393 					     struct ctl_table *table);
394 void unregister_net_sysctl_table(struct ctl_table_header *header);
395 #else
396 static inline int net_sysctl_init(void) { return 0; }
397 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
398 	const char *path, struct ctl_table *table)
399 {
400 	return NULL;
401 }
402 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
403 {
404 }
405 #endif
406 
407 static inline int rt_genid_ipv4(struct net *net)
408 {
409 	return atomic_read(&net->ipv4.rt_genid);
410 }
411 
412 static inline void rt_genid_bump_ipv4(struct net *net)
413 {
414 	atomic_inc(&net->ipv4.rt_genid);
415 }
416 
417 extern void (*__fib6_flush_trees)(struct net *net);
418 static inline void rt_genid_bump_ipv6(struct net *net)
419 {
420 	if (__fib6_flush_trees)
421 		__fib6_flush_trees(net);
422 }
423 
424 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
425 static inline struct netns_ieee802154_lowpan *
426 net_ieee802154_lowpan(struct net *net)
427 {
428 	return &net->ieee802154_lowpan;
429 }
430 #endif
431 
432 /* For callers who don't really care about whether it's IPv4 or IPv6 */
433 static inline void rt_genid_bump_all(struct net *net)
434 {
435 	rt_genid_bump_ipv4(net);
436 	rt_genid_bump_ipv6(net);
437 }
438 
439 static inline int fnhe_genid(struct net *net)
440 {
441 	return atomic_read(&net->fnhe_genid);
442 }
443 
444 static inline void fnhe_genid_bump(struct net *net)
445 {
446 	atomic_inc(&net->fnhe_genid);
447 }
448 
449 #endif /* __NET_NET_NAMESPACE_H */
450