xref: /openbmc/linux/include/net/net_namespace.h (revision aa0dc6a7)
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 <linux/ns_common.h>
38 #include <linux/idr.h>
39 #include <linux/skbuff.h>
40 #include <linux/notifier.h>
41 
42 struct user_namespace;
43 struct proc_dir_entry;
44 struct net_device;
45 struct sock;
46 struct ctl_table_header;
47 struct net_generic;
48 struct uevent_sock;
49 struct netns_ipvs;
50 struct bpf_prog;
51 
52 
53 #define NETDEV_HASHBITS    8
54 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
55 
56 struct net {
57 	/* First cache line can be often dirtied.
58 	 * Do not place here read-mostly fields.
59 	 */
60 	refcount_t		passive;	/* To decide when the network
61 						 * namespace should be freed.
62 						 */
63 	spinlock_t		rules_mod_lock;
64 
65 	unsigned int		dev_unreg_count;
66 
67 	unsigned int		dev_base_seq;	/* protected by rtnl_mutex */
68 	int			ifindex;
69 
70 	spinlock_t		nsid_lock;
71 	atomic_t		fnhe_genid;
72 
73 	struct list_head	list;		/* list of network namespaces */
74 	struct list_head	exit_list;	/* To linked to call pernet exit
75 						 * methods on dead net (
76 						 * pernet_ops_rwsem read locked),
77 						 * or to unregister pernet ops
78 						 * (pernet_ops_rwsem write locked).
79 						 */
80 	struct llist_node	cleanup_list;	/* namespaces on death row */
81 
82 #ifdef CONFIG_KEYS
83 	struct key_tag		*key_domain;	/* Key domain of operation tag */
84 #endif
85 	struct user_namespace   *user_ns;	/* Owning user namespace */
86 	struct ucounts		*ucounts;
87 	struct idr		netns_ids;
88 
89 	struct ns_common	ns;
90 
91 	struct list_head 	dev_base_head;
92 	struct proc_dir_entry 	*proc_net;
93 	struct proc_dir_entry 	*proc_net_stat;
94 
95 #ifdef CONFIG_SYSCTL
96 	struct ctl_table_set	sysctls;
97 #endif
98 
99 	struct sock 		*rtnl;			/* rtnetlink socket */
100 	struct sock		*genl_sock;
101 
102 	struct uevent_sock	*uevent_sock;		/* uevent socket */
103 
104 	struct hlist_head 	*dev_name_head;
105 	struct hlist_head	*dev_index_head;
106 	struct raw_notifier_head	netdev_chain;
107 
108 	/* Note that @hash_mix can be read millions times per second,
109 	 * it is critical that it is on a read_mostly cache line.
110 	 */
111 	u32			hash_mix;
112 
113 	struct net_device       *loopback_dev;          /* The loopback */
114 
115 	/* core fib_rules */
116 	struct list_head	rules_ops;
117 
118 	struct netns_core	core;
119 	struct netns_mib	mib;
120 	struct netns_packet	packet;
121 	struct netns_unix	unx;
122 	struct netns_nexthop	nexthop;
123 	struct netns_ipv4	ipv4;
124 #if IS_ENABLED(CONFIG_IPV6)
125 	struct netns_ipv6	ipv6;
126 #endif
127 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
128 	struct netns_ieee802154_lowpan	ieee802154_lowpan;
129 #endif
130 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
131 	struct netns_sctp	sctp;
132 #endif
133 #ifdef CONFIG_NETFILTER
134 	struct netns_nf		nf;
135 	struct netns_xt		xt;
136 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
137 	struct netns_ct		ct;
138 #endif
139 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
140 	struct netns_nftables	nft;
141 #endif
142 #endif
143 #ifdef CONFIG_WEXT_CORE
144 	struct sk_buff_head	wext_nlevents;
145 #endif
146 	struct net_generic __rcu	*gen;
147 
148 	/* Used to store attached BPF programs */
149 	struct netns_bpf	bpf;
150 
151 	/* Note : following structs are cache line aligned */
152 #ifdef CONFIG_XFRM
153 	struct netns_xfrm	xfrm;
154 #endif
155 
156 	u64			net_cookie; /* written once */
157 
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 #if IS_ENABLED(CONFIG_CRYPTO_USER)
171 	struct sock		*crypto_nlsk;
172 #endif
173 	struct sock		*diag_nlsk;
174 #if IS_ENABLED(CONFIG_SMC)
175 	struct netns_smc	smc;
176 #endif
177 } __randomize_layout;
178 
179 #include <linux/seq_file_net.h>
180 
181 /* Init's network namespace */
182 extern struct net init_net;
183 
184 #ifdef CONFIG_NET_NS
185 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
186 			struct net *old_net);
187 
188 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
189 
190 void net_ns_barrier(void);
191 
192 struct ns_common *get_net_ns(struct ns_common *ns);
193 struct net *get_net_ns_by_fd(int fd);
194 #else /* CONFIG_NET_NS */
195 #include <linux/sched.h>
196 #include <linux/nsproxy.h>
197 static inline struct net *copy_net_ns(unsigned long flags,
198 	struct user_namespace *user_ns, struct net *old_net)
199 {
200 	if (flags & CLONE_NEWNET)
201 		return ERR_PTR(-EINVAL);
202 	return old_net;
203 }
204 
205 static inline void net_ns_get_ownership(const struct net *net,
206 					kuid_t *uid, kgid_t *gid)
207 {
208 	*uid = GLOBAL_ROOT_UID;
209 	*gid = GLOBAL_ROOT_GID;
210 }
211 
212 static inline void net_ns_barrier(void) {}
213 
214 static inline struct ns_common *get_net_ns(struct ns_common *ns)
215 {
216 	return ERR_PTR(-EINVAL);
217 }
218 
219 static inline struct net *get_net_ns_by_fd(int fd)
220 {
221 	return ERR_PTR(-EINVAL);
222 }
223 #endif /* CONFIG_NET_NS */
224 
225 
226 extern struct list_head net_namespace_list;
227 
228 struct net *get_net_ns_by_pid(pid_t pid);
229 
230 #ifdef CONFIG_SYSCTL
231 void ipx_register_sysctl(void);
232 void ipx_unregister_sysctl(void);
233 #else
234 #define ipx_register_sysctl()
235 #define ipx_unregister_sysctl()
236 #endif
237 
238 #ifdef CONFIG_NET_NS
239 void __put_net(struct net *net);
240 
241 static inline struct net *get_net(struct net *net)
242 {
243 	refcount_inc(&net->ns.count);
244 	return net;
245 }
246 
247 static inline struct net *maybe_get_net(struct net *net)
248 {
249 	/* Used when we know struct net exists but we
250 	 * aren't guaranteed a previous reference count
251 	 * exists.  If the reference count is zero this
252 	 * function fails and returns NULL.
253 	 */
254 	if (!refcount_inc_not_zero(&net->ns.count))
255 		net = NULL;
256 	return net;
257 }
258 
259 static inline void put_net(struct net *net)
260 {
261 	if (refcount_dec_and_test(&net->ns.count))
262 		__put_net(net);
263 }
264 
265 static inline
266 int net_eq(const struct net *net1, const struct net *net2)
267 {
268 	return net1 == net2;
269 }
270 
271 static inline int check_net(const struct net *net)
272 {
273 	return refcount_read(&net->ns.count) != 0;
274 }
275 
276 void net_drop_ns(void *);
277 
278 #else
279 
280 static inline struct net *get_net(struct net *net)
281 {
282 	return net;
283 }
284 
285 static inline void put_net(struct net *net)
286 {
287 }
288 
289 static inline struct net *maybe_get_net(struct net *net)
290 {
291 	return net;
292 }
293 
294 static inline
295 int net_eq(const struct net *net1, const struct net *net2)
296 {
297 	return 1;
298 }
299 
300 static inline int check_net(const struct net *net)
301 {
302 	return 1;
303 }
304 
305 #define net_drop_ns NULL
306 #endif
307 
308 
309 typedef struct {
310 #ifdef CONFIG_NET_NS
311 	struct net *net;
312 #endif
313 } possible_net_t;
314 
315 static inline void write_pnet(possible_net_t *pnet, struct net *net)
316 {
317 #ifdef CONFIG_NET_NS
318 	pnet->net = net;
319 #endif
320 }
321 
322 static inline struct net *read_pnet(const possible_net_t *pnet)
323 {
324 #ifdef CONFIG_NET_NS
325 	return pnet->net;
326 #else
327 	return &init_net;
328 #endif
329 }
330 
331 /* Protected by net_rwsem */
332 #define for_each_net(VAR)				\
333 	list_for_each_entry(VAR, &net_namespace_list, list)
334 #define for_each_net_continue_reverse(VAR)		\
335 	list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list)
336 #define for_each_net_rcu(VAR)				\
337 	list_for_each_entry_rcu(VAR, &net_namespace_list, list)
338 
339 #ifdef CONFIG_NET_NS
340 #define __net_init
341 #define __net_exit
342 #define __net_initdata
343 #define __net_initconst
344 #else
345 #define __net_init	__init
346 #define __net_exit	__ref
347 #define __net_initdata	__initdata
348 #define __net_initconst	__initconst
349 #endif
350 
351 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp);
352 int peernet2id(const struct net *net, struct net *peer);
353 bool peernet_has_id(const struct net *net, struct net *peer);
354 struct net *get_net_ns_by_id(const struct net *net, int id);
355 
356 struct pernet_operations {
357 	struct list_head list;
358 	/*
359 	 * Below methods are called without any exclusive locks.
360 	 * More than one net may be constructed and destructed
361 	 * in parallel on several cpus. Every pernet_operations
362 	 * have to keep in mind all other pernet_operations and
363 	 * to introduce a locking, if they share common resources.
364 	 *
365 	 * The only time they are called with exclusive lock is
366 	 * from register_pernet_subsys(), unregister_pernet_subsys()
367 	 * register_pernet_device() and unregister_pernet_device().
368 	 *
369 	 * Exit methods using blocking RCU primitives, such as
370 	 * synchronize_rcu(), should be implemented via exit_batch.
371 	 * Then, destruction of a group of net requires single
372 	 * synchronize_rcu() related to these pernet_operations,
373 	 * instead of separate synchronize_rcu() for every net.
374 	 * Please, avoid synchronize_rcu() at all, where it's possible.
375 	 *
376 	 * Note that a combination of pre_exit() and exit() can
377 	 * be used, since a synchronize_rcu() is guaranteed between
378 	 * the calls.
379 	 */
380 	int (*init)(struct net *net);
381 	void (*pre_exit)(struct net *net);
382 	void (*exit)(struct net *net);
383 	void (*exit_batch)(struct list_head *net_exit_list);
384 	unsigned int *id;
385 	size_t size;
386 };
387 
388 /*
389  * Use these carefully.  If you implement a network device and it
390  * needs per network namespace operations use device pernet operations,
391  * otherwise use pernet subsys operations.
392  *
393  * Network interfaces need to be removed from a dying netns _before_
394  * subsys notifiers can be called, as most of the network code cleanup
395  * (which is done from subsys notifiers) runs with the assumption that
396  * dev_remove_pack has been called so no new packets will arrive during
397  * and after the cleanup functions have been called.  dev_remove_pack
398  * is not per namespace so instead the guarantee of no more packets
399  * arriving in a network namespace is provided by ensuring that all
400  * network devices and all sockets have left the network namespace
401  * before the cleanup methods are called.
402  *
403  * For the longest time the ipv4 icmp code was registered as a pernet
404  * device which caused kernel oops, and panics during network
405  * namespace cleanup.   So please don't get this wrong.
406  */
407 int register_pernet_subsys(struct pernet_operations *);
408 void unregister_pernet_subsys(struct pernet_operations *);
409 int register_pernet_device(struct pernet_operations *);
410 void unregister_pernet_device(struct pernet_operations *);
411 
412 struct ctl_table;
413 
414 #ifdef CONFIG_SYSCTL
415 int net_sysctl_init(void);
416 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
417 					     struct ctl_table *table);
418 void unregister_net_sysctl_table(struct ctl_table_header *header);
419 #else
420 static inline int net_sysctl_init(void) { return 0; }
421 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
422 	const char *path, struct ctl_table *table)
423 {
424 	return NULL;
425 }
426 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
427 {
428 }
429 #endif
430 
431 static inline int rt_genid_ipv4(const struct net *net)
432 {
433 	return atomic_read(&net->ipv4.rt_genid);
434 }
435 
436 #if IS_ENABLED(CONFIG_IPV6)
437 static inline int rt_genid_ipv6(const struct net *net)
438 {
439 	return atomic_read(&net->ipv6.fib6_sernum);
440 }
441 #endif
442 
443 static inline void rt_genid_bump_ipv4(struct net *net)
444 {
445 	atomic_inc(&net->ipv4.rt_genid);
446 }
447 
448 extern void (*__fib6_flush_trees)(struct net *net);
449 static inline void rt_genid_bump_ipv6(struct net *net)
450 {
451 	if (__fib6_flush_trees)
452 		__fib6_flush_trees(net);
453 }
454 
455 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
456 static inline struct netns_ieee802154_lowpan *
457 net_ieee802154_lowpan(struct net *net)
458 {
459 	return &net->ieee802154_lowpan;
460 }
461 #endif
462 
463 /* For callers who don't really care about whether it's IPv4 or IPv6 */
464 static inline void rt_genid_bump_all(struct net *net)
465 {
466 	rt_genid_bump_ipv4(net);
467 	rt_genid_bump_ipv6(net);
468 }
469 
470 static inline int fnhe_genid(const struct net *net)
471 {
472 	return atomic_read(&net->fnhe_genid);
473 }
474 
475 static inline void fnhe_genid_bump(struct net *net)
476 {
477 	atomic_inc(&net->fnhe_genid);
478 }
479 
480 #endif /* __NET_NET_NAMESPACE_H */
481