xref: /openbmc/linux/include/net/net_namespace.h (revision 8e774e02)
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 <linux/ns_common.h>
37 #include <linux/idr.h>
38 #include <linux/skbuff.h>
39 
40 struct user_namespace;
41 struct proc_dir_entry;
42 struct net_device;
43 struct sock;
44 struct ctl_table_header;
45 struct net_generic;
46 struct uevent_sock;
47 struct netns_ipvs;
48 struct bpf_prog;
49 
50 
51 #define NETDEV_HASHBITS    8
52 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
53 
54 struct net {
55 	refcount_t		passive;	/* To decided when the network
56 						 * namespace should be freed.
57 						 */
58 	refcount_t		count;		/* To decided when the network
59 						 *  namespace should be shut down.
60 						 */
61 	spinlock_t		rules_mod_lock;
62 
63 	u32			hash_mix;
64 	atomic64_t		cookie_gen;
65 
66 	struct list_head	list;		/* list of network namespaces */
67 	struct list_head	exit_list;	/* To linked to call pernet exit
68 						 * methods on dead net (
69 						 * pernet_ops_rwsem read locked),
70 						 * or to unregister pernet ops
71 						 * (pernet_ops_rwsem write locked).
72 						 */
73 	struct llist_node	cleanup_list;	/* namespaces on death row */
74 
75 #ifdef CONFIG_KEYS
76 	struct key_tag		*key_domain;	/* Key domain of operation tag */
77 #endif
78 	struct user_namespace   *user_ns;	/* Owning user namespace */
79 	struct ucounts		*ucounts;
80 	spinlock_t		nsid_lock;
81 	struct idr		netns_ids;
82 
83 	struct ns_common	ns;
84 
85 	struct proc_dir_entry 	*proc_net;
86 	struct proc_dir_entry 	*proc_net_stat;
87 
88 #ifdef CONFIG_SYSCTL
89 	struct ctl_table_set	sysctls;
90 #endif
91 
92 	struct sock 		*rtnl;			/* rtnetlink socket */
93 	struct sock		*genl_sock;
94 
95 	struct uevent_sock	*uevent_sock;		/* uevent socket */
96 
97 	struct list_head 	dev_base_head;
98 	struct hlist_head 	*dev_name_head;
99 	struct hlist_head	*dev_index_head;
100 	unsigned int		dev_base_seq;	/* protected by rtnl_mutex */
101 	int			ifindex;
102 	unsigned int		dev_unreg_count;
103 
104 	/* core fib_rules */
105 	struct list_head	rules_ops;
106 
107 	struct list_head	fib_notifier_ops;  /* Populated by
108 						    * register_pernet_subsys()
109 						    */
110 	struct net_device       *loopback_dev;          /* The loopback */
111 	struct netns_core	core;
112 	struct netns_mib	mib;
113 	struct netns_packet	packet;
114 	struct netns_unix	unx;
115 	struct netns_nexthop	nexthop;
116 	struct netns_ipv4	ipv4;
117 #if IS_ENABLED(CONFIG_IPV6)
118 	struct netns_ipv6	ipv6;
119 #endif
120 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
121 	struct netns_ieee802154_lowpan	ieee802154_lowpan;
122 #endif
123 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
124 	struct netns_sctp	sctp;
125 #endif
126 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
127 	struct netns_dccp	dccp;
128 #endif
129 #ifdef CONFIG_NETFILTER
130 	struct netns_nf		nf;
131 	struct netns_xt		xt;
132 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
133 	struct netns_ct		ct;
134 #endif
135 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
136 	struct netns_nftables	nft;
137 #endif
138 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
139 	struct netns_nf_frag	nf_frag;
140 	struct ctl_table_header *nf_frag_frags_hdr;
141 #endif
142 	struct sock		*nfnl;
143 	struct sock		*nfnl_stash;
144 #if IS_ENABLED(CONFIG_NETFILTER_NETLINK_ACCT)
145 	struct list_head        nfnl_acct_list;
146 #endif
147 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
148 	struct list_head	nfct_timeout_list;
149 #endif
150 #endif
151 #ifdef CONFIG_WEXT_CORE
152 	struct sk_buff_head	wext_nlevents;
153 #endif
154 	struct net_generic __rcu	*gen;
155 
156 	struct bpf_prog __rcu	*flow_dissector_prog;
157 
158 	/* Note : following structs are cache line aligned */
159 #ifdef CONFIG_XFRM
160 	struct netns_xfrm	xfrm;
161 #endif
162 #if IS_ENABLED(CONFIG_IP_VS)
163 	struct netns_ipvs	*ipvs;
164 #endif
165 #if IS_ENABLED(CONFIG_MPLS)
166 	struct netns_mpls	mpls;
167 #endif
168 #if IS_ENABLED(CONFIG_CAN)
169 	struct netns_can	can;
170 #endif
171 #ifdef CONFIG_XDP_SOCKETS
172 	struct netns_xdp	xdp;
173 #endif
174 	struct sock		*diag_nlsk;
175 	atomic_t		fnhe_genid;
176 } __randomize_layout;
177 
178 #include <linux/seq_file_net.h>
179 
180 /* Init's network namespace */
181 extern struct net init_net;
182 
183 #ifdef CONFIG_NET_NS
184 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
185 			struct net *old_net);
186 
187 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
188 
189 void net_ns_barrier(void);
190 #else /* CONFIG_NET_NS */
191 #include <linux/sched.h>
192 #include <linux/nsproxy.h>
193 static inline struct net *copy_net_ns(unsigned long flags,
194 	struct user_namespace *user_ns, struct net *old_net)
195 {
196 	if (flags & CLONE_NEWNET)
197 		return ERR_PTR(-EINVAL);
198 	return old_net;
199 }
200 
201 static inline void net_ns_get_ownership(const struct net *net,
202 					kuid_t *uid, kgid_t *gid)
203 {
204 	*uid = GLOBAL_ROOT_UID;
205 	*gid = GLOBAL_ROOT_GID;
206 }
207 
208 static inline void net_ns_barrier(void) {}
209 #endif /* CONFIG_NET_NS */
210 
211 
212 extern struct list_head net_namespace_list;
213 
214 struct net *get_net_ns_by_pid(pid_t pid);
215 struct net *get_net_ns_by_fd(int fd);
216 
217 #ifdef CONFIG_SYSCTL
218 void ipx_register_sysctl(void);
219 void ipx_unregister_sysctl(void);
220 #else
221 #define ipx_register_sysctl()
222 #define ipx_unregister_sysctl()
223 #endif
224 
225 #ifdef CONFIG_NET_NS
226 void __put_net(struct net *net);
227 
228 static inline struct net *get_net(struct net *net)
229 {
230 	refcount_inc(&net->count);
231 	return net;
232 }
233 
234 static inline struct net *maybe_get_net(struct net *net)
235 {
236 	/* Used when we know struct net exists but we
237 	 * aren't guaranteed a previous reference count
238 	 * exists.  If the reference count is zero this
239 	 * function fails and returns NULL.
240 	 */
241 	if (!refcount_inc_not_zero(&net->count))
242 		net = NULL;
243 	return net;
244 }
245 
246 static inline void put_net(struct net *net)
247 {
248 	if (refcount_dec_and_test(&net->count))
249 		__put_net(net);
250 }
251 
252 static inline
253 int net_eq(const struct net *net1, const struct net *net2)
254 {
255 	return net1 == net2;
256 }
257 
258 static inline int check_net(const struct net *net)
259 {
260 	return refcount_read(&net->count) != 0;
261 }
262 
263 void net_drop_ns(void *);
264 
265 #else
266 
267 static inline struct net *get_net(struct net *net)
268 {
269 	return net;
270 }
271 
272 static inline void put_net(struct net *net)
273 {
274 }
275 
276 static inline struct net *maybe_get_net(struct net *net)
277 {
278 	return net;
279 }
280 
281 static inline
282 int net_eq(const struct net *net1, const struct net *net2)
283 {
284 	return 1;
285 }
286 
287 static inline int check_net(const struct net *net)
288 {
289 	return 1;
290 }
291 
292 #define net_drop_ns NULL
293 #endif
294 
295 
296 typedef struct {
297 #ifdef CONFIG_NET_NS
298 	struct net *net;
299 #endif
300 } possible_net_t;
301 
302 static inline void write_pnet(possible_net_t *pnet, struct net *net)
303 {
304 #ifdef CONFIG_NET_NS
305 	pnet->net = net;
306 #endif
307 }
308 
309 static inline struct net *read_pnet(const possible_net_t *pnet)
310 {
311 #ifdef CONFIG_NET_NS
312 	return pnet->net;
313 #else
314 	return &init_net;
315 #endif
316 }
317 
318 /* Protected by net_rwsem */
319 #define for_each_net(VAR)				\
320 	list_for_each_entry(VAR, &net_namespace_list, list)
321 
322 #define for_each_net_rcu(VAR)				\
323 	list_for_each_entry_rcu(VAR, &net_namespace_list, list)
324 
325 #ifdef CONFIG_NET_NS
326 #define __net_init
327 #define __net_exit
328 #define __net_initdata
329 #define __net_initconst
330 #else
331 #define __net_init	__init
332 #define __net_exit	__ref
333 #define __net_initdata	__initdata
334 #define __net_initconst	__initconst
335 #endif
336 
337 int peernet2id_alloc(struct net *net, struct net *peer);
338 int peernet2id(struct net *net, struct net *peer);
339 bool peernet_has_id(struct net *net, struct net *peer);
340 struct net *get_net_ns_by_id(struct net *net, int id);
341 
342 struct pernet_operations {
343 	struct list_head list;
344 	/*
345 	 * Below methods are called without any exclusive locks.
346 	 * More than one net may be constructed and destructed
347 	 * in parallel on several cpus. Every pernet_operations
348 	 * have to keep in mind all other pernet_operations and
349 	 * to introduce a locking, if they share common resources.
350 	 *
351 	 * The only time they are called with exclusive lock is
352 	 * from register_pernet_subsys(), unregister_pernet_subsys()
353 	 * register_pernet_device() and unregister_pernet_device().
354 	 *
355 	 * Exit methods using blocking RCU primitives, such as
356 	 * synchronize_rcu(), should be implemented via exit_batch.
357 	 * Then, destruction of a group of net requires single
358 	 * synchronize_rcu() related to these pernet_operations,
359 	 * instead of separate synchronize_rcu() for every net.
360 	 * Please, avoid synchronize_rcu() at all, where it's possible.
361 	 *
362 	 * Note that a combination of pre_exit() and exit() can
363 	 * be used, since a synchronize_rcu() is guaranteed between
364 	 * the calls.
365 	 */
366 	int (*init)(struct net *net);
367 	void (*pre_exit)(struct net *net);
368 	void (*exit)(struct net *net);
369 	void (*exit_batch)(struct list_head *net_exit_list);
370 	unsigned int *id;
371 	size_t size;
372 };
373 
374 /*
375  * Use these carefully.  If you implement a network device and it
376  * needs per network namespace operations use device pernet operations,
377  * otherwise use pernet subsys operations.
378  *
379  * Network interfaces need to be removed from a dying netns _before_
380  * subsys notifiers can be called, as most of the network code cleanup
381  * (which is done from subsys notifiers) runs with the assumption that
382  * dev_remove_pack has been called so no new packets will arrive during
383  * and after the cleanup functions have been called.  dev_remove_pack
384  * is not per namespace so instead the guarantee of no more packets
385  * arriving in a network namespace is provided by ensuring that all
386  * network devices and all sockets have left the network namespace
387  * before the cleanup methods are called.
388  *
389  * For the longest time the ipv4 icmp code was registered as a pernet
390  * device which caused kernel oops, and panics during network
391  * namespace cleanup.   So please don't get this wrong.
392  */
393 int register_pernet_subsys(struct pernet_operations *);
394 void unregister_pernet_subsys(struct pernet_operations *);
395 int register_pernet_device(struct pernet_operations *);
396 void unregister_pernet_device(struct pernet_operations *);
397 
398 struct ctl_table;
399 struct ctl_table_header;
400 
401 #ifdef CONFIG_SYSCTL
402 int net_sysctl_init(void);
403 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
404 					     struct ctl_table *table);
405 void unregister_net_sysctl_table(struct ctl_table_header *header);
406 #else
407 static inline int net_sysctl_init(void) { return 0; }
408 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
409 	const char *path, struct ctl_table *table)
410 {
411 	return NULL;
412 }
413 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
414 {
415 }
416 #endif
417 
418 static inline int rt_genid_ipv4(struct net *net)
419 {
420 	return atomic_read(&net->ipv4.rt_genid);
421 }
422 
423 static inline void rt_genid_bump_ipv4(struct net *net)
424 {
425 	atomic_inc(&net->ipv4.rt_genid);
426 }
427 
428 extern void (*__fib6_flush_trees)(struct net *net);
429 static inline void rt_genid_bump_ipv6(struct net *net)
430 {
431 	if (__fib6_flush_trees)
432 		__fib6_flush_trees(net);
433 }
434 
435 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
436 static inline struct netns_ieee802154_lowpan *
437 net_ieee802154_lowpan(struct net *net)
438 {
439 	return &net->ieee802154_lowpan;
440 }
441 #endif
442 
443 /* For callers who don't really care about whether it's IPv4 or IPv6 */
444 static inline void rt_genid_bump_all(struct net *net)
445 {
446 	rt_genid_bump_ipv4(net);
447 	rt_genid_bump_ipv6(net);
448 }
449 
450 static inline int fnhe_genid(struct net *net)
451 {
452 	return atomic_read(&net->fnhe_genid);
453 }
454 
455 static inline void fnhe_genid_bump(struct net *net)
456 {
457 	atomic_inc(&net->fnhe_genid);
458 }
459 
460 #endif /* __NET_NET_NAMESPACE_H */
461