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 <net/net_trackers.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 	atomic_t		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 	struct ref_tracker_dir  refcnt_tracker;
92 
93 	struct list_head 	dev_base_head;
94 	struct proc_dir_entry 	*proc_net;
95 	struct proc_dir_entry 	*proc_net_stat;
96 
97 #ifdef CONFIG_SYSCTL
98 	struct ctl_table_set	sysctls;
99 #endif
100 
101 	struct sock 		*rtnl;			/* rtnetlink socket */
102 	struct sock		*genl_sock;
103 
104 	struct uevent_sock	*uevent_sock;		/* uevent socket */
105 
106 	struct hlist_head 	*dev_name_head;
107 	struct hlist_head	*dev_index_head;
108 	struct raw_notifier_head	netdev_chain;
109 
110 	/* Note that @hash_mix can be read millions times per second,
111 	 * it is critical that it is on a read_mostly cache line.
112 	 */
113 	u32			hash_mix;
114 
115 	struct net_device       *loopback_dev;          /* The loopback */
116 
117 	/* core fib_rules */
118 	struct list_head	rules_ops;
119 
120 	struct netns_core	core;
121 	struct netns_mib	mib;
122 	struct netns_packet	packet;
123 	struct netns_unix	unx;
124 	struct netns_nexthop	nexthop;
125 	struct netns_ipv4	ipv4;
126 #if IS_ENABLED(CONFIG_IPV6)
127 	struct netns_ipv6	ipv6;
128 #endif
129 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
130 	struct netns_ieee802154_lowpan	ieee802154_lowpan;
131 #endif
132 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
133 	struct netns_sctp	sctp;
134 #endif
135 #ifdef CONFIG_NETFILTER
136 	struct netns_nf		nf;
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>
copy_net_ns(unsigned long flags,struct user_namespace * user_ns,struct net * old_net)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 
net_ns_get_ownership(const struct net * net,kuid_t * uid,kgid_t * gid)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 
net_ns_barrier(void)216 static inline void net_ns_barrier(void) {}
217 
get_net_ns(struct ns_common * ns)218 static inline struct ns_common *get_net_ns(struct ns_common *ns)
219 {
220 	return ERR_PTR(-EINVAL);
221 }
222 
get_net_ns_by_fd(int fd)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 /* Try using get_net_track() instead */
get_net(struct net * net)246 static inline struct net *get_net(struct net *net)
247 {
248 	refcount_inc(&net->ns.count);
249 	return net;
250 }
251 
maybe_get_net(struct net * net)252 static inline struct net *maybe_get_net(struct net *net)
253 {
254 	/* Used when we know struct net exists but we
255 	 * aren't guaranteed a previous reference count
256 	 * exists.  If the reference count is zero this
257 	 * function fails and returns NULL.
258 	 */
259 	if (!refcount_inc_not_zero(&net->ns.count))
260 		net = NULL;
261 	return net;
262 }
263 
264 /* Try using put_net_track() instead */
put_net(struct net * net)265 static inline void put_net(struct net *net)
266 {
267 	if (refcount_dec_and_test(&net->ns.count))
268 		__put_net(net);
269 }
270 
271 static inline
net_eq(const struct net * net1,const struct net * net2)272 int net_eq(const struct net *net1, const struct net *net2)
273 {
274 	return net1 == net2;
275 }
276 
check_net(const struct net * net)277 static inline int check_net(const struct net *net)
278 {
279 	return refcount_read(&net->ns.count) != 0;
280 }
281 
282 void net_drop_ns(void *);
283 
284 #else
285 
get_net(struct net * net)286 static inline struct net *get_net(struct net *net)
287 {
288 	return net;
289 }
290 
put_net(struct net * net)291 static inline void put_net(struct net *net)
292 {
293 }
294 
maybe_get_net(struct net * net)295 static inline struct net *maybe_get_net(struct net *net)
296 {
297 	return net;
298 }
299 
300 static inline
net_eq(const struct net * net1,const struct net * net2)301 int net_eq(const struct net *net1, const struct net *net2)
302 {
303 	return 1;
304 }
305 
check_net(const struct net * net)306 static inline int check_net(const struct net *net)
307 {
308 	return 1;
309 }
310 
311 #define net_drop_ns NULL
312 #endif
313 
314 
netns_tracker_alloc(struct net * net,netns_tracker * tracker,gfp_t gfp)315 static inline void netns_tracker_alloc(struct net *net,
316 				       netns_tracker *tracker, gfp_t gfp)
317 {
318 #ifdef CONFIG_NET_NS_REFCNT_TRACKER
319 	ref_tracker_alloc(&net->refcnt_tracker, tracker, gfp);
320 #endif
321 }
322 
netns_tracker_free(struct net * net,netns_tracker * tracker)323 static inline void netns_tracker_free(struct net *net,
324 				      netns_tracker *tracker)
325 {
326 #ifdef CONFIG_NET_NS_REFCNT_TRACKER
327        ref_tracker_free(&net->refcnt_tracker, tracker);
328 #endif
329 }
330 
get_net_track(struct net * net,netns_tracker * tracker,gfp_t gfp)331 static inline struct net *get_net_track(struct net *net,
332 					netns_tracker *tracker, gfp_t gfp)
333 {
334 	get_net(net);
335 	netns_tracker_alloc(net, tracker, gfp);
336 	return net;
337 }
338 
put_net_track(struct net * net,netns_tracker * tracker)339 static inline void put_net_track(struct net *net, netns_tracker *tracker)
340 {
341 	netns_tracker_free(net, tracker);
342 	put_net(net);
343 }
344 
345 typedef struct {
346 #ifdef CONFIG_NET_NS
347 	struct net *net;
348 #endif
349 } possible_net_t;
350 
write_pnet(possible_net_t * pnet,struct net * net)351 static inline void write_pnet(possible_net_t *pnet, struct net *net)
352 {
353 #ifdef CONFIG_NET_NS
354 	pnet->net = net;
355 #endif
356 }
357 
read_pnet(const possible_net_t * pnet)358 static inline struct net *read_pnet(const possible_net_t *pnet)
359 {
360 #ifdef CONFIG_NET_NS
361 	return pnet->net;
362 #else
363 	return &init_net;
364 #endif
365 }
366 
367 /* Protected by net_rwsem */
368 #define for_each_net(VAR)				\
369 	list_for_each_entry(VAR, &net_namespace_list, list)
370 #define for_each_net_continue_reverse(VAR)		\
371 	list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list)
372 #define for_each_net_rcu(VAR)				\
373 	list_for_each_entry_rcu(VAR, &net_namespace_list, list)
374 
375 #ifdef CONFIG_NET_NS
376 #define __net_init
377 #define __net_exit
378 #define __net_initdata
379 #define __net_initconst
380 #else
381 #define __net_init	__init
382 #define __net_exit	__ref
383 #define __net_initdata	__initdata
384 #define __net_initconst	__initconst
385 #endif
386 
387 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp);
388 int peernet2id(const struct net *net, struct net *peer);
389 bool peernet_has_id(const struct net *net, struct net *peer);
390 struct net *get_net_ns_by_id(const struct net *net, int id);
391 
392 struct pernet_operations {
393 	struct list_head list;
394 	/*
395 	 * Below methods are called without any exclusive locks.
396 	 * More than one net may be constructed and destructed
397 	 * in parallel on several cpus. Every pernet_operations
398 	 * have to keep in mind all other pernet_operations and
399 	 * to introduce a locking, if they share common resources.
400 	 *
401 	 * The only time they are called with exclusive lock is
402 	 * from register_pernet_subsys(), unregister_pernet_subsys()
403 	 * register_pernet_device() and unregister_pernet_device().
404 	 *
405 	 * Exit methods using blocking RCU primitives, such as
406 	 * synchronize_rcu(), should be implemented via exit_batch.
407 	 * Then, destruction of a group of net requires single
408 	 * synchronize_rcu() related to these pernet_operations,
409 	 * instead of separate synchronize_rcu() for every net.
410 	 * Please, avoid synchronize_rcu() at all, where it's possible.
411 	 *
412 	 * Note that a combination of pre_exit() and exit() can
413 	 * be used, since a synchronize_rcu() is guaranteed between
414 	 * the calls.
415 	 */
416 	int (*init)(struct net *net);
417 	void (*pre_exit)(struct net *net);
418 	void (*exit)(struct net *net);
419 	void (*exit_batch)(struct list_head *net_exit_list);
420 	unsigned int *id;
421 	size_t size;
422 };
423 
424 /*
425  * Use these carefully.  If you implement a network device and it
426  * needs per network namespace operations use device pernet operations,
427  * otherwise use pernet subsys operations.
428  *
429  * Network interfaces need to be removed from a dying netns _before_
430  * subsys notifiers can be called, as most of the network code cleanup
431  * (which is done from subsys notifiers) runs with the assumption that
432  * dev_remove_pack has been called so no new packets will arrive during
433  * and after the cleanup functions have been called.  dev_remove_pack
434  * is not per namespace so instead the guarantee of no more packets
435  * arriving in a network namespace is provided by ensuring that all
436  * network devices and all sockets have left the network namespace
437  * before the cleanup methods are called.
438  *
439  * For the longest time the ipv4 icmp code was registered as a pernet
440  * device which caused kernel oops, and panics during network
441  * namespace cleanup.   So please don't get this wrong.
442  */
443 int register_pernet_subsys(struct pernet_operations *);
444 void unregister_pernet_subsys(struct pernet_operations *);
445 int register_pernet_device(struct pernet_operations *);
446 void unregister_pernet_device(struct pernet_operations *);
447 
448 struct ctl_table;
449 
450 #ifdef CONFIG_SYSCTL
451 int net_sysctl_init(void);
452 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
453 					     struct ctl_table *table);
454 void unregister_net_sysctl_table(struct ctl_table_header *header);
455 #else
net_sysctl_init(void)456 static inline int net_sysctl_init(void) { return 0; }
register_net_sysctl(struct net * net,const char * path,struct ctl_table * table)457 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
458 	const char *path, struct ctl_table *table)
459 {
460 	return NULL;
461 }
unregister_net_sysctl_table(struct ctl_table_header * header)462 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
463 {
464 }
465 #endif
466 
rt_genid_ipv4(const struct net * net)467 static inline int rt_genid_ipv4(const struct net *net)
468 {
469 	return atomic_read(&net->ipv4.rt_genid);
470 }
471 
472 #if IS_ENABLED(CONFIG_IPV6)
rt_genid_ipv6(const struct net * net)473 static inline int rt_genid_ipv6(const struct net *net)
474 {
475 	return atomic_read(&net->ipv6.fib6_sernum);
476 }
477 #endif
478 
rt_genid_bump_ipv4(struct net * net)479 static inline void rt_genid_bump_ipv4(struct net *net)
480 {
481 	atomic_inc(&net->ipv4.rt_genid);
482 }
483 
484 extern void (*__fib6_flush_trees)(struct net *net);
rt_genid_bump_ipv6(struct net * net)485 static inline void rt_genid_bump_ipv6(struct net *net)
486 {
487 	if (__fib6_flush_trees)
488 		__fib6_flush_trees(net);
489 }
490 
491 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
492 static inline struct netns_ieee802154_lowpan *
net_ieee802154_lowpan(struct net * net)493 net_ieee802154_lowpan(struct net *net)
494 {
495 	return &net->ieee802154_lowpan;
496 }
497 #endif
498 
499 /* For callers who don't really care about whether it's IPv4 or IPv6 */
rt_genid_bump_all(struct net * net)500 static inline void rt_genid_bump_all(struct net *net)
501 {
502 	rt_genid_bump_ipv4(net);
503 	rt_genid_bump_ipv6(net);
504 }
505 
fnhe_genid(const struct net * net)506 static inline int fnhe_genid(const struct net *net)
507 {
508 	return atomic_read(&net->fnhe_genid);
509 }
510 
fnhe_genid_bump(struct net * net)511 static inline void fnhe_genid_bump(struct net *net)
512 {
513 	atomic_inc(&net->fnhe_genid);
514 }
515 
516 #ifdef CONFIG_NET
517 void net_ns_init(void);
518 #else
net_ns_init(void)519 static inline void net_ns_init(void) {}
520 #endif
521 
522 #endif /* __NET_NET_NAMESPACE_H */
523