xref: /openbmc/linux/include/net/net_namespace.h (revision 0ca8d3ca4561535f97b31e7b8de569c69bc3b27b)
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 #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 <net/netns/smc.h>
36 #include <net/netns/bpf.h>
37 #include <net/netns/mctp.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 	unsigned int		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 
92 	struct list_head 	dev_base_head;
93 	struct proc_dir_entry 	*proc_net;
94 	struct proc_dir_entry 	*proc_net_stat;
95 
96 #ifdef CONFIG_SYSCTL
97 	struct ctl_table_set	sysctls;
98 #endif
99 
100 	struct sock 		*rtnl;			/* rtnetlink socket */
101 	struct sock		*genl_sock;
102 
103 	struct uevent_sock	*uevent_sock;		/* uevent socket */
104 
105 	struct hlist_head 	*dev_name_head;
106 	struct hlist_head	*dev_index_head;
107 	struct raw_notifier_head	netdev_chain;
108 
109 	/* Note that @hash_mix can be read millions times per second,
110 	 * it is critical that it is on a read_mostly cache line.
111 	 */
112 	u32			hash_mix;
113 
114 	struct net_device       *loopback_dev;          /* The loopback */
115 
116 	/* core fib_rules */
117 	struct list_head	rules_ops;
118 
119 	struct netns_core	core;
120 	struct netns_mib	mib;
121 	struct netns_packet	packet;
122 	struct netns_unix	unx;
123 	struct netns_nexthop	nexthop;
124 	struct netns_ipv4	ipv4;
125 #if IS_ENABLED(CONFIG_IPV6)
126 	struct netns_ipv6	ipv6;
127 #endif
128 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
129 	struct netns_ieee802154_lowpan	ieee802154_lowpan;
130 #endif
131 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
132 	struct netns_sctp	sctp;
133 #endif
134 #ifdef CONFIG_NETFILTER
135 	struct netns_nf		nf;
136 	struct netns_xt		xt;
137 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
138 	struct netns_ct		ct;
139 #endif
140 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
141 	struct netns_nftables	nft;
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 	/* Used to store attached BPF programs */
150 	struct netns_bpf	bpf;
151 
152 	/* Note : following structs are cache line aligned */
153 #ifdef CONFIG_XFRM
154 	struct netns_xfrm	xfrm;
155 #endif
156 
157 	u64			net_cookie; /* written once */
158 
159 #if IS_ENABLED(CONFIG_IP_VS)
160 	struct netns_ipvs	*ipvs;
161 #endif
162 #if IS_ENABLED(CONFIG_MPLS)
163 	struct netns_mpls	mpls;
164 #endif
165 #if IS_ENABLED(CONFIG_CAN)
166 	struct netns_can	can;
167 #endif
168 #ifdef CONFIG_XDP_SOCKETS
169 	struct netns_xdp	xdp;
170 #endif
171 #if IS_ENABLED(CONFIG_MCTP)
172 	struct netns_mctp	mctp;
173 #endif
174 #if IS_ENABLED(CONFIG_CRYPTO_USER)
175 	struct sock		*crypto_nlsk;
176 #endif
177 	struct sock		*diag_nlsk;
178 #if IS_ENABLED(CONFIG_SMC)
179 	struct netns_smc	smc;
180 #endif
181 } __randomize_layout;
182 
183 #include <linux/seq_file_net.h>
184 
185 /* Init's network namespace */
186 extern struct net init_net;
187 
188 #ifdef CONFIG_NET_NS
189 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
190 			struct net *old_net);
191 
192 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
193 
194 void net_ns_barrier(void);
195 
196 struct ns_common *get_net_ns(struct ns_common *ns);
197 struct net *get_net_ns_by_fd(int fd);
198 #else /* CONFIG_NET_NS */
199 #include <linux/sched.h>
200 #include <linux/nsproxy.h>
201 static inline struct net *copy_net_ns(unsigned long flags,
202 	struct user_namespace *user_ns, struct net *old_net)
203 {
204 	if (flags & CLONE_NEWNET)
205 		return ERR_PTR(-EINVAL);
206 	return old_net;
207 }
208 
209 static inline void net_ns_get_ownership(const struct net *net,
210 					kuid_t *uid, kgid_t *gid)
211 {
212 	*uid = GLOBAL_ROOT_UID;
213 	*gid = GLOBAL_ROOT_GID;
214 }
215 
216 static inline void net_ns_barrier(void) {}
217 
218 static inline struct ns_common *get_net_ns(struct ns_common *ns)
219 {
220 	return ERR_PTR(-EINVAL);
221 }
222 
223 static inline struct net *get_net_ns_by_fd(int fd)
224 {
225 	return ERR_PTR(-EINVAL);
226 }
227 #endif /* CONFIG_NET_NS */
228 
229 
230 extern struct list_head net_namespace_list;
231 
232 struct net *get_net_ns_by_pid(pid_t pid);
233 
234 #ifdef CONFIG_SYSCTL
235 void ipx_register_sysctl(void);
236 void ipx_unregister_sysctl(void);
237 #else
238 #define ipx_register_sysctl()
239 #define ipx_unregister_sysctl()
240 #endif
241 
242 #ifdef CONFIG_NET_NS
243 void __put_net(struct net *net);
244 
245 static inline struct net *get_net(struct net *net)
246 {
247 	refcount_inc(&net->ns.count);
248 	return net;
249 }
250 
251 static inline struct net *maybe_get_net(struct net *net)
252 {
253 	/* Used when we know struct net exists but we
254 	 * aren't guaranteed a previous reference count
255 	 * exists.  If the reference count is zero this
256 	 * function fails and returns NULL.
257 	 */
258 	if (!refcount_inc_not_zero(&net->ns.count))
259 		net = NULL;
260 	return net;
261 }
262 
263 static inline void put_net(struct net *net)
264 {
265 	if (refcount_dec_and_test(&net->ns.count))
266 		__put_net(net);
267 }
268 
269 static inline
270 int net_eq(const struct net *net1, const struct net *net2)
271 {
272 	return net1 == net2;
273 }
274 
275 static inline int check_net(const struct net *net)
276 {
277 	return refcount_read(&net->ns.count) != 0;
278 }
279 
280 void net_drop_ns(void *);
281 
282 #else
283 
284 static inline struct net *get_net(struct net *net)
285 {
286 	return net;
287 }
288 
289 static inline void put_net(struct net *net)
290 {
291 }
292 
293 static inline struct net *maybe_get_net(struct net *net)
294 {
295 	return net;
296 }
297 
298 static inline
299 int net_eq(const struct net *net1, const struct net *net2)
300 {
301 	return 1;
302 }
303 
304 static inline int check_net(const struct net *net)
305 {
306 	return 1;
307 }
308 
309 #define net_drop_ns NULL
310 #endif
311 
312 
313 typedef struct {
314 #ifdef CONFIG_NET_NS
315 	struct net *net;
316 #endif
317 } possible_net_t;
318 
319 static inline void write_pnet(possible_net_t *pnet, struct net *net)
320 {
321 #ifdef CONFIG_NET_NS
322 	pnet->net = net;
323 #endif
324 }
325 
326 static inline struct net *read_pnet(const possible_net_t *pnet)
327 {
328 #ifdef CONFIG_NET_NS
329 	return pnet->net;
330 #else
331 	return &init_net;
332 #endif
333 }
334 
335 /* Protected by net_rwsem */
336 #define for_each_net(VAR)				\
337 	list_for_each_entry(VAR, &net_namespace_list, list)
338 #define for_each_net_continue_reverse(VAR)		\
339 	list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list)
340 #define for_each_net_rcu(VAR)				\
341 	list_for_each_entry_rcu(VAR, &net_namespace_list, list)
342 
343 #ifdef CONFIG_NET_NS
344 #define __net_init
345 #define __net_exit
346 #define __net_initdata
347 #define __net_initconst
348 #else
349 #define __net_init	__init
350 #define __net_exit	__ref
351 #define __net_initdata	__initdata
352 #define __net_initconst	__initconst
353 #endif
354 
355 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp);
356 int peernet2id(const struct net *net, struct net *peer);
357 bool peernet_has_id(const struct net *net, struct net *peer);
358 struct net *get_net_ns_by_id(const struct net *net, int id);
359 
360 struct pernet_operations {
361 	struct list_head list;
362 	/*
363 	 * Below methods are called without any exclusive locks.
364 	 * More than one net may be constructed and destructed
365 	 * in parallel on several cpus. Every pernet_operations
366 	 * have to keep in mind all other pernet_operations and
367 	 * to introduce a locking, if they share common resources.
368 	 *
369 	 * The only time they are called with exclusive lock is
370 	 * from register_pernet_subsys(), unregister_pernet_subsys()
371 	 * register_pernet_device() and unregister_pernet_device().
372 	 *
373 	 * Exit methods using blocking RCU primitives, such as
374 	 * synchronize_rcu(), should be implemented via exit_batch.
375 	 * Then, destruction of a group of net requires single
376 	 * synchronize_rcu() related to these pernet_operations,
377 	 * instead of separate synchronize_rcu() for every net.
378 	 * Please, avoid synchronize_rcu() at all, where it's possible.
379 	 *
380 	 * Note that a combination of pre_exit() and exit() can
381 	 * be used, since a synchronize_rcu() is guaranteed between
382 	 * the calls.
383 	 */
384 	int (*init)(struct net *net);
385 	void (*pre_exit)(struct net *net);
386 	void (*exit)(struct net *net);
387 	void (*exit_batch)(struct list_head *net_exit_list);
388 	unsigned int *id;
389 	size_t size;
390 };
391 
392 /*
393  * Use these carefully.  If you implement a network device and it
394  * needs per network namespace operations use device pernet operations,
395  * otherwise use pernet subsys operations.
396  *
397  * Network interfaces need to be removed from a dying netns _before_
398  * subsys notifiers can be called, as most of the network code cleanup
399  * (which is done from subsys notifiers) runs with the assumption that
400  * dev_remove_pack has been called so no new packets will arrive during
401  * and after the cleanup functions have been called.  dev_remove_pack
402  * is not per namespace so instead the guarantee of no more packets
403  * arriving in a network namespace is provided by ensuring that all
404  * network devices and all sockets have left the network namespace
405  * before the cleanup methods are called.
406  *
407  * For the longest time the ipv4 icmp code was registered as a pernet
408  * device which caused kernel oops, and panics during network
409  * namespace cleanup.   So please don't get this wrong.
410  */
411 int register_pernet_subsys(struct pernet_operations *);
412 void unregister_pernet_subsys(struct pernet_operations *);
413 int register_pernet_device(struct pernet_operations *);
414 void unregister_pernet_device(struct pernet_operations *);
415 
416 struct ctl_table;
417 
418 #ifdef CONFIG_SYSCTL
419 int net_sysctl_init(void);
420 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
421 					     struct ctl_table *table);
422 void unregister_net_sysctl_table(struct ctl_table_header *header);
423 #else
424 static inline int net_sysctl_init(void) { return 0; }
425 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
426 	const char *path, struct ctl_table *table)
427 {
428 	return NULL;
429 }
430 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
431 {
432 }
433 #endif
434 
435 static inline int rt_genid_ipv4(const struct net *net)
436 {
437 	return atomic_read(&net->ipv4.rt_genid);
438 }
439 
440 #if IS_ENABLED(CONFIG_IPV6)
441 static inline int rt_genid_ipv6(const struct net *net)
442 {
443 	return atomic_read(&net->ipv6.fib6_sernum);
444 }
445 #endif
446 
447 static inline void rt_genid_bump_ipv4(struct net *net)
448 {
449 	atomic_inc(&net->ipv4.rt_genid);
450 }
451 
452 extern void (*__fib6_flush_trees)(struct net *net);
453 static inline void rt_genid_bump_ipv6(struct net *net)
454 {
455 	if (__fib6_flush_trees)
456 		__fib6_flush_trees(net);
457 }
458 
459 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
460 static inline struct netns_ieee802154_lowpan *
461 net_ieee802154_lowpan(struct net *net)
462 {
463 	return &net->ieee802154_lowpan;
464 }
465 #endif
466 
467 /* For callers who don't really care about whether it's IPv4 or IPv6 */
468 static inline void rt_genid_bump_all(struct net *net)
469 {
470 	rt_genid_bump_ipv4(net);
471 	rt_genid_bump_ipv6(net);
472 }
473 
474 static inline int fnhe_genid(const struct net *net)
475 {
476 	return atomic_read(&net->fnhe_genid);
477 }
478 
479 static inline void fnhe_genid_bump(struct net *net)
480 {
481 	atomic_inc(&net->fnhe_genid);
482 }
483 
484 #endif /* __NET_NET_NAMESPACE_H */
485