xref: /openbmc/linux/include/net/net_namespace.h (revision 8bf3cbe32b180836720f735e6de5dee700052317)
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 decide 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 
65 	struct list_head	list;		/* list of network namespaces */
66 	struct list_head	exit_list;	/* To linked to call pernet exit
67 						 * methods on dead net (
68 						 * pernet_ops_rwsem read locked),
69 						 * or to unregister pernet ops
70 						 * (pernet_ops_rwsem write locked).
71 						 */
72 	struct llist_node	cleanup_list;	/* namespaces on death row */
73 
74 #ifdef CONFIG_KEYS
75 	struct key_tag		*key_domain;	/* Key domain of operation tag */
76 #endif
77 	struct user_namespace   *user_ns;	/* Owning user namespace */
78 	struct ucounts		*ucounts;
79 	spinlock_t		nsid_lock;
80 	struct idr		netns_ids;
81 
82 	struct ns_common	ns;
83 
84 	struct proc_dir_entry 	*proc_net;
85 	struct proc_dir_entry 	*proc_net_stat;
86 
87 #ifdef CONFIG_SYSCTL
88 	struct ctl_table_set	sysctls;
89 #endif
90 
91 	struct sock 		*rtnl;			/* rtnetlink socket */
92 	struct sock		*genl_sock;
93 
94 	struct uevent_sock	*uevent_sock;		/* uevent socket */
95 
96 	struct list_head 	dev_base_head;
97 	struct hlist_head 	*dev_name_head;
98 	struct hlist_head	*dev_index_head;
99 	unsigned int		dev_base_seq;	/* protected by rtnl_mutex */
100 	int			ifindex;
101 	unsigned int		dev_unreg_count;
102 
103 	/* core fib_rules */
104 	struct list_head	rules_ops;
105 
106 	struct net_device       *loopback_dev;          /* The loopback */
107 	struct netns_core	core;
108 	struct netns_mib	mib;
109 	struct netns_packet	packet;
110 	struct netns_unix	unx;
111 	struct netns_nexthop	nexthop;
112 	struct netns_ipv4	ipv4;
113 #if IS_ENABLED(CONFIG_IPV6)
114 	struct netns_ipv6	ipv6;
115 #endif
116 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
117 	struct netns_ieee802154_lowpan	ieee802154_lowpan;
118 #endif
119 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
120 	struct netns_sctp	sctp;
121 #endif
122 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
123 	struct netns_dccp	dccp;
124 #endif
125 #ifdef CONFIG_NETFILTER
126 	struct netns_nf		nf;
127 	struct netns_xt		xt;
128 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
129 	struct netns_ct		ct;
130 #endif
131 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
132 	struct netns_nftables	nft;
133 #endif
134 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
135 	struct netns_nf_frag	nf_frag;
136 	struct ctl_table_header *nf_frag_frags_hdr;
137 #endif
138 	struct sock		*nfnl;
139 	struct sock		*nfnl_stash;
140 #if IS_ENABLED(CONFIG_NETFILTER_NETLINK_ACCT)
141 	struct list_head        nfnl_acct_list;
142 #endif
143 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
144 	struct list_head	nfct_timeout_list;
145 #endif
146 #endif
147 #ifdef CONFIG_WEXT_CORE
148 	struct sk_buff_head	wext_nlevents;
149 #endif
150 	struct net_generic __rcu	*gen;
151 
152 	struct bpf_prog __rcu	*flow_dissector_prog;
153 
154 	/* Note : following structs are cache line aligned */
155 #ifdef CONFIG_XFRM
156 	struct netns_xfrm	xfrm;
157 #endif
158 #if IS_ENABLED(CONFIG_IP_VS)
159 	struct netns_ipvs	*ipvs;
160 #endif
161 #if IS_ENABLED(CONFIG_MPLS)
162 	struct netns_mpls	mpls;
163 #endif
164 #if IS_ENABLED(CONFIG_CAN)
165 	struct netns_can	can;
166 #endif
167 #ifdef CONFIG_XDP_SOCKETS
168 	struct netns_xdp	xdp;
169 #endif
170 	struct sock		*diag_nlsk;
171 	atomic_t		fnhe_genid;
172 } __randomize_layout;
173 
174 #include <linux/seq_file_net.h>
175 
176 /* Init's network namespace */
177 extern struct net init_net;
178 
179 #ifdef CONFIG_NET_NS
180 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
181 			struct net *old_net);
182 
183 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
184 
185 void net_ns_barrier(void);
186 #else /* CONFIG_NET_NS */
187 #include <linux/sched.h>
188 #include <linux/nsproxy.h>
189 static inline struct net *copy_net_ns(unsigned long flags,
190 	struct user_namespace *user_ns, struct net *old_net)
191 {
192 	if (flags & CLONE_NEWNET)
193 		return ERR_PTR(-EINVAL);
194 	return old_net;
195 }
196 
197 static inline void net_ns_get_ownership(const struct net *net,
198 					kuid_t *uid, kgid_t *gid)
199 {
200 	*uid = GLOBAL_ROOT_UID;
201 	*gid = GLOBAL_ROOT_GID;
202 }
203 
204 static inline void net_ns_barrier(void) {}
205 #endif /* CONFIG_NET_NS */
206 
207 
208 extern struct list_head net_namespace_list;
209 
210 struct net *get_net_ns_by_pid(pid_t pid);
211 struct net *get_net_ns_by_fd(int fd);
212 
213 #ifdef CONFIG_SYSCTL
214 void ipx_register_sysctl(void);
215 void ipx_unregister_sysctl(void);
216 #else
217 #define ipx_register_sysctl()
218 #define ipx_unregister_sysctl()
219 #endif
220 
221 #ifdef CONFIG_NET_NS
222 void __put_net(struct net *net);
223 
224 static inline struct net *get_net(struct net *net)
225 {
226 	refcount_inc(&net->count);
227 	return net;
228 }
229 
230 static inline struct net *maybe_get_net(struct net *net)
231 {
232 	/* Used when we know struct net exists but we
233 	 * aren't guaranteed a previous reference count
234 	 * exists.  If the reference count is zero this
235 	 * function fails and returns NULL.
236 	 */
237 	if (!refcount_inc_not_zero(&net->count))
238 		net = NULL;
239 	return net;
240 }
241 
242 static inline void put_net(struct net *net)
243 {
244 	if (refcount_dec_and_test(&net->count))
245 		__put_net(net);
246 }
247 
248 static inline
249 int net_eq(const struct net *net1, const struct net *net2)
250 {
251 	return net1 == net2;
252 }
253 
254 static inline int check_net(const struct net *net)
255 {
256 	return refcount_read(&net->count) != 0;
257 }
258 
259 void net_drop_ns(void *);
260 
261 #else
262 
263 static inline struct net *get_net(struct net *net)
264 {
265 	return net;
266 }
267 
268 static inline void put_net(struct net *net)
269 {
270 }
271 
272 static inline struct net *maybe_get_net(struct net *net)
273 {
274 	return net;
275 }
276 
277 static inline
278 int net_eq(const struct net *net1, const struct net *net2)
279 {
280 	return 1;
281 }
282 
283 static inline int check_net(const struct net *net)
284 {
285 	return 1;
286 }
287 
288 #define net_drop_ns NULL
289 #endif
290 
291 
292 typedef struct {
293 #ifdef CONFIG_NET_NS
294 	struct net *net;
295 #endif
296 } possible_net_t;
297 
298 static inline void write_pnet(possible_net_t *pnet, struct net *net)
299 {
300 #ifdef CONFIG_NET_NS
301 	pnet->net = net;
302 #endif
303 }
304 
305 static inline struct net *read_pnet(const possible_net_t *pnet)
306 {
307 #ifdef CONFIG_NET_NS
308 	return pnet->net;
309 #else
310 	return &init_net;
311 #endif
312 }
313 
314 /* Protected by net_rwsem */
315 #define for_each_net(VAR)				\
316 	list_for_each_entry(VAR, &net_namespace_list, list)
317 
318 #define for_each_net_rcu(VAR)				\
319 	list_for_each_entry_rcu(VAR, &net_namespace_list, list)
320 
321 #ifdef CONFIG_NET_NS
322 #define __net_init
323 #define __net_exit
324 #define __net_initdata
325 #define __net_initconst
326 #else
327 #define __net_init	__init
328 #define __net_exit	__ref
329 #define __net_initdata	__initdata
330 #define __net_initconst	__initconst
331 #endif
332 
333 int peernet2id_alloc(struct net *net, struct net *peer);
334 int peernet2id(struct net *net, struct net *peer);
335 bool peernet_has_id(struct net *net, struct net *peer);
336 struct net *get_net_ns_by_id(struct net *net, int id);
337 
338 struct pernet_operations {
339 	struct list_head list;
340 	/*
341 	 * Below methods are called without any exclusive locks.
342 	 * More than one net may be constructed and destructed
343 	 * in parallel on several cpus. Every pernet_operations
344 	 * have to keep in mind all other pernet_operations and
345 	 * to introduce a locking, if they share common resources.
346 	 *
347 	 * The only time they are called with exclusive lock is
348 	 * from register_pernet_subsys(), unregister_pernet_subsys()
349 	 * register_pernet_device() and unregister_pernet_device().
350 	 *
351 	 * Exit methods using blocking RCU primitives, such as
352 	 * synchronize_rcu(), should be implemented via exit_batch.
353 	 * Then, destruction of a group of net requires single
354 	 * synchronize_rcu() related to these pernet_operations,
355 	 * instead of separate synchronize_rcu() for every net.
356 	 * Please, avoid synchronize_rcu() at all, where it's possible.
357 	 *
358 	 * Note that a combination of pre_exit() and exit() can
359 	 * be used, since a synchronize_rcu() is guaranteed between
360 	 * the calls.
361 	 */
362 	int (*init)(struct net *net);
363 	void (*pre_exit)(struct net *net);
364 	void (*exit)(struct net *net);
365 	void (*exit_batch)(struct list_head *net_exit_list);
366 	unsigned int *id;
367 	size_t size;
368 };
369 
370 /*
371  * Use these carefully.  If you implement a network device and it
372  * needs per network namespace operations use device pernet operations,
373  * otherwise use pernet subsys operations.
374  *
375  * Network interfaces need to be removed from a dying netns _before_
376  * subsys notifiers can be called, as most of the network code cleanup
377  * (which is done from subsys notifiers) runs with the assumption that
378  * dev_remove_pack has been called so no new packets will arrive during
379  * and after the cleanup functions have been called.  dev_remove_pack
380  * is not per namespace so instead the guarantee of no more packets
381  * arriving in a network namespace is provided by ensuring that all
382  * network devices and all sockets have left the network namespace
383  * before the cleanup methods are called.
384  *
385  * For the longest time the ipv4 icmp code was registered as a pernet
386  * device which caused kernel oops, and panics during network
387  * namespace cleanup.   So please don't get this wrong.
388  */
389 int register_pernet_subsys(struct pernet_operations *);
390 void unregister_pernet_subsys(struct pernet_operations *);
391 int register_pernet_device(struct pernet_operations *);
392 void unregister_pernet_device(struct pernet_operations *);
393 
394 struct ctl_table;
395 struct ctl_table_header;
396 
397 #ifdef CONFIG_SYSCTL
398 int net_sysctl_init(void);
399 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
400 					     struct ctl_table *table);
401 void unregister_net_sysctl_table(struct ctl_table_header *header);
402 #else
403 static inline int net_sysctl_init(void) { return 0; }
404 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
405 	const char *path, struct ctl_table *table)
406 {
407 	return NULL;
408 }
409 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
410 {
411 }
412 #endif
413 
414 static inline int rt_genid_ipv4(struct net *net)
415 {
416 	return atomic_read(&net->ipv4.rt_genid);
417 }
418 
419 static inline void rt_genid_bump_ipv4(struct net *net)
420 {
421 	atomic_inc(&net->ipv4.rt_genid);
422 }
423 
424 extern void (*__fib6_flush_trees)(struct net *net);
425 static inline void rt_genid_bump_ipv6(struct net *net)
426 {
427 	if (__fib6_flush_trees)
428 		__fib6_flush_trees(net);
429 }
430 
431 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
432 static inline struct netns_ieee802154_lowpan *
433 net_ieee802154_lowpan(struct net *net)
434 {
435 	return &net->ieee802154_lowpan;
436 }
437 #endif
438 
439 /* For callers who don't really care about whether it's IPv4 or IPv6 */
440 static inline void rt_genid_bump_all(struct net *net)
441 {
442 	rt_genid_bump_ipv4(net);
443 	rt_genid_bump_ipv6(net);
444 }
445 
446 static inline int fnhe_genid(struct net *net)
447 {
448 	return atomic_read(&net->fnhe_genid);
449 }
450 
451 static inline void fnhe_genid_bump(struct net *net)
452 {
453 	atomic_inc(&net->fnhe_genid);
454 }
455 
456 #endif /* __NET_NET_NAMESPACE_H */
457