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 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
27 #include <net/netns/conntrack.h>
28 #endif
29 #include <net/netns/nftables.h>
30 #include <net/netns/xfrm.h>
31 #include <net/netns/mpls.h>
32 #include <net/netns/can.h>
33 #include <net/netns/xdp.h>
34 #include <net/netns/smc.h>
35 #include <net/netns/bpf.h>
36 #include <net/netns/mctp.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 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
136 	struct netns_ct		ct;
137 #endif
138 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
139 	struct netns_nftables	nft;
140 #endif
141 #endif
142 #ifdef CONFIG_WEXT_CORE
143 	struct sk_buff_head	wext_nlevents;
144 #endif
145 	struct net_generic __rcu	*gen;
146 
147 	/* Used to store attached BPF programs */
148 	struct netns_bpf	bpf;
149 
150 	/* Note : following structs are cache line aligned */
151 #ifdef CONFIG_XFRM
152 	struct netns_xfrm	xfrm;
153 #endif
154 
155 	u64			net_cookie; /* written once */
156 
157 #if IS_ENABLED(CONFIG_IP_VS)
158 	struct netns_ipvs	*ipvs;
159 #endif
160 #if IS_ENABLED(CONFIG_MPLS)
161 	struct netns_mpls	mpls;
162 #endif
163 #if IS_ENABLED(CONFIG_CAN)
164 	struct netns_can	can;
165 #endif
166 #ifdef CONFIG_XDP_SOCKETS
167 	struct netns_xdp	xdp;
168 #endif
169 #if IS_ENABLED(CONFIG_MCTP)
170 	struct netns_mctp	mctp;
171 #endif
172 #if IS_ENABLED(CONFIG_CRYPTO_USER)
173 	struct sock		*crypto_nlsk;
174 #endif
175 	struct sock		*diag_nlsk;
176 #if IS_ENABLED(CONFIG_SMC)
177 	struct netns_smc	smc;
178 #endif
179 } __randomize_layout;
180 
181 #include <linux/seq_file_net.h>
182 
183 /* Init's network namespace */
184 extern struct net init_net;
185 
186 #ifdef CONFIG_NET_NS
187 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
188 			struct net *old_net);
189 
190 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
191 
192 void net_ns_barrier(void);
193 
194 struct ns_common *get_net_ns(struct ns_common *ns);
195 struct net *get_net_ns_by_fd(int fd);
196 #else /* CONFIG_NET_NS */
197 #include <linux/sched.h>
198 #include <linux/nsproxy.h>
copy_net_ns(unsigned long flags,struct user_namespace * user_ns,struct net * old_net)199 static inline struct net *copy_net_ns(unsigned long flags,
200 	struct user_namespace *user_ns, struct net *old_net)
201 {
202 	if (flags & CLONE_NEWNET)
203 		return ERR_PTR(-EINVAL);
204 	return old_net;
205 }
206 
net_ns_get_ownership(const struct net * net,kuid_t * uid,kgid_t * gid)207 static inline void net_ns_get_ownership(const struct net *net,
208 					kuid_t *uid, kgid_t *gid)
209 {
210 	*uid = GLOBAL_ROOT_UID;
211 	*gid = GLOBAL_ROOT_GID;
212 }
213 
net_ns_barrier(void)214 static inline void net_ns_barrier(void) {}
215 
get_net_ns(struct ns_common * ns)216 static inline struct ns_common *get_net_ns(struct ns_common *ns)
217 {
218 	return ERR_PTR(-EINVAL);
219 }
220 
get_net_ns_by_fd(int fd)221 static inline struct net *get_net_ns_by_fd(int fd)
222 {
223 	return ERR_PTR(-EINVAL);
224 }
225 #endif /* CONFIG_NET_NS */
226 
227 
228 extern struct list_head net_namespace_list;
229 
230 struct net *get_net_ns_by_pid(pid_t pid);
231 
232 #ifdef CONFIG_SYSCTL
233 void ipx_register_sysctl(void);
234 void ipx_unregister_sysctl(void);
235 #else
236 #define ipx_register_sysctl()
237 #define ipx_unregister_sysctl()
238 #endif
239 
240 #ifdef CONFIG_NET_NS
241 void __put_net(struct net *net);
242 
get_net(struct net * net)243 static inline struct net *get_net(struct net *net)
244 {
245 	refcount_inc(&net->ns.count);
246 	return net;
247 }
248 
maybe_get_net(struct net * net)249 static inline struct net *maybe_get_net(struct net *net)
250 {
251 	/* Used when we know struct net exists but we
252 	 * aren't guaranteed a previous reference count
253 	 * exists.  If the reference count is zero this
254 	 * function fails and returns NULL.
255 	 */
256 	if (!refcount_inc_not_zero(&net->ns.count))
257 		net = NULL;
258 	return net;
259 }
260 
put_net(struct net * net)261 static inline void put_net(struct net *net)
262 {
263 	if (refcount_dec_and_test(&net->ns.count))
264 		__put_net(net);
265 }
266 
267 static inline
net_eq(const struct net * net1,const struct net * net2)268 int net_eq(const struct net *net1, const struct net *net2)
269 {
270 	return net1 == net2;
271 }
272 
check_net(const struct net * net)273 static inline int check_net(const struct net *net)
274 {
275 	return refcount_read(&net->ns.count) != 0;
276 }
277 
278 void net_drop_ns(void *);
279 
280 #else
281 
get_net(struct net * net)282 static inline struct net *get_net(struct net *net)
283 {
284 	return net;
285 }
286 
put_net(struct net * net)287 static inline void put_net(struct net *net)
288 {
289 }
290 
maybe_get_net(struct net * net)291 static inline struct net *maybe_get_net(struct net *net)
292 {
293 	return net;
294 }
295 
296 static inline
net_eq(const struct net * net1,const struct net * net2)297 int net_eq(const struct net *net1, const struct net *net2)
298 {
299 	return 1;
300 }
301 
check_net(const struct net * net)302 static inline int check_net(const struct net *net)
303 {
304 	return 1;
305 }
306 
307 #define net_drop_ns NULL
308 #endif
309 
310 
311 typedef struct {
312 #ifdef CONFIG_NET_NS
313 	struct net *net;
314 #endif
315 } possible_net_t;
316 
write_pnet(possible_net_t * pnet,struct net * net)317 static inline void write_pnet(possible_net_t *pnet, struct net *net)
318 {
319 #ifdef CONFIG_NET_NS
320 	pnet->net = net;
321 #endif
322 }
323 
read_pnet(const possible_net_t * pnet)324 static inline struct net *read_pnet(const possible_net_t *pnet)
325 {
326 #ifdef CONFIG_NET_NS
327 	return pnet->net;
328 #else
329 	return &init_net;
330 #endif
331 }
332 
333 /* Protected by net_rwsem */
334 #define for_each_net(VAR)				\
335 	list_for_each_entry(VAR, &net_namespace_list, list)
336 #define for_each_net_continue_reverse(VAR)		\
337 	list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list)
338 #define for_each_net_rcu(VAR)				\
339 	list_for_each_entry_rcu(VAR, &net_namespace_list, list)
340 
341 #ifdef CONFIG_NET_NS
342 #define __net_init
343 #define __net_exit
344 #define __net_initdata
345 #define __net_initconst
346 #else
347 #define __net_init	__init
348 #define __net_exit	__ref
349 #define __net_initdata	__initdata
350 #define __net_initconst	__initconst
351 #endif
352 
353 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp);
354 int peernet2id(const struct net *net, struct net *peer);
355 bool peernet_has_id(const struct net *net, struct net *peer);
356 struct net *get_net_ns_by_id(const struct net *net, int id);
357 
358 struct pernet_operations {
359 	struct list_head list;
360 	/*
361 	 * Below methods are called without any exclusive locks.
362 	 * More than one net may be constructed and destructed
363 	 * in parallel on several cpus. Every pernet_operations
364 	 * have to keep in mind all other pernet_operations and
365 	 * to introduce a locking, if they share common resources.
366 	 *
367 	 * The only time they are called with exclusive lock is
368 	 * from register_pernet_subsys(), unregister_pernet_subsys()
369 	 * register_pernet_device() and unregister_pernet_device().
370 	 *
371 	 * Exit methods using blocking RCU primitives, such as
372 	 * synchronize_rcu(), should be implemented via exit_batch.
373 	 * Then, destruction of a group of net requires single
374 	 * synchronize_rcu() related to these pernet_operations,
375 	 * instead of separate synchronize_rcu() for every net.
376 	 * Please, avoid synchronize_rcu() at all, where it's possible.
377 	 *
378 	 * Note that a combination of pre_exit() and exit() can
379 	 * be used, since a synchronize_rcu() is guaranteed between
380 	 * the calls.
381 	 */
382 	int (*init)(struct net *net);
383 	void (*pre_exit)(struct net *net);
384 	void (*exit)(struct net *net);
385 	void (*exit_batch)(struct list_head *net_exit_list);
386 	unsigned int *id;
387 	size_t size;
388 };
389 
390 /*
391  * Use these carefully.  If you implement a network device and it
392  * needs per network namespace operations use device pernet operations,
393  * otherwise use pernet subsys operations.
394  *
395  * Network interfaces need to be removed from a dying netns _before_
396  * subsys notifiers can be called, as most of the network code cleanup
397  * (which is done from subsys notifiers) runs with the assumption that
398  * dev_remove_pack has been called so no new packets will arrive during
399  * and after the cleanup functions have been called.  dev_remove_pack
400  * is not per namespace so instead the guarantee of no more packets
401  * arriving in a network namespace is provided by ensuring that all
402  * network devices and all sockets have left the network namespace
403  * before the cleanup methods are called.
404  *
405  * For the longest time the ipv4 icmp code was registered as a pernet
406  * device which caused kernel oops, and panics during network
407  * namespace cleanup.   So please don't get this wrong.
408  */
409 int register_pernet_subsys(struct pernet_operations *);
410 void unregister_pernet_subsys(struct pernet_operations *);
411 int register_pernet_device(struct pernet_operations *);
412 void unregister_pernet_device(struct pernet_operations *);
413 
414 struct ctl_table;
415 
416 #ifdef CONFIG_SYSCTL
417 int net_sysctl_init(void);
418 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
419 					     struct ctl_table *table);
420 void unregister_net_sysctl_table(struct ctl_table_header *header);
421 #else
net_sysctl_init(void)422 static inline int net_sysctl_init(void) { return 0; }
register_net_sysctl(struct net * net,const char * path,struct ctl_table * table)423 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
424 	const char *path, struct ctl_table *table)
425 {
426 	return NULL;
427 }
unregister_net_sysctl_table(struct ctl_table_header * header)428 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
429 {
430 }
431 #endif
432 
rt_genid_ipv4(const struct net * net)433 static inline int rt_genid_ipv4(const struct net *net)
434 {
435 	return atomic_read(&net->ipv4.rt_genid);
436 }
437 
438 #if IS_ENABLED(CONFIG_IPV6)
rt_genid_ipv6(const struct net * net)439 static inline int rt_genid_ipv6(const struct net *net)
440 {
441 	return atomic_read(&net->ipv6.fib6_sernum);
442 }
443 #endif
444 
rt_genid_bump_ipv4(struct net * net)445 static inline void rt_genid_bump_ipv4(struct net *net)
446 {
447 	atomic_inc(&net->ipv4.rt_genid);
448 }
449 
450 extern void (*__fib6_flush_trees)(struct net *net);
rt_genid_bump_ipv6(struct net * net)451 static inline void rt_genid_bump_ipv6(struct net *net)
452 {
453 	if (__fib6_flush_trees)
454 		__fib6_flush_trees(net);
455 }
456 
457 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
458 static inline struct netns_ieee802154_lowpan *
net_ieee802154_lowpan(struct net * net)459 net_ieee802154_lowpan(struct net *net)
460 {
461 	return &net->ieee802154_lowpan;
462 }
463 #endif
464 
465 /* For callers who don't really care about whether it's IPv4 or IPv6 */
rt_genid_bump_all(struct net * net)466 static inline void rt_genid_bump_all(struct net *net)
467 {
468 	rt_genid_bump_ipv4(net);
469 	rt_genid_bump_ipv6(net);
470 }
471 
fnhe_genid(const struct net * net)472 static inline int fnhe_genid(const struct net *net)
473 {
474 	return atomic_read(&net->fnhe_genid);
475 }
476 
fnhe_genid_bump(struct net * net)477 static inline void fnhe_genid_bump(struct net *net)
478 {
479 	atomic_inc(&net->fnhe_genid);
480 }
481 
482 #endif /* __NET_NET_NAMESPACE_H */
483