xref: /openbmc/linux/include/net/net_namespace.h (revision 55b37d9c)
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 #if IS_ENABLED(CONFIG_NF_FLOW_TABLE)
30 #include <net/netns/flow_table.h>
31 #endif
32 #include <net/netns/nftables.h>
33 #include <net/netns/xfrm.h>
34 #include <net/netns/mpls.h>
35 #include <net/netns/can.h>
36 #include <net/netns/xdp.h>
37 #include <net/netns/smc.h>
38 #include <net/netns/bpf.h>
39 #include <net/netns/mctp.h>
40 #include <net/net_trackers.h>
41 #include <linux/ns_common.h>
42 #include <linux/idr.h>
43 #include <linux/skbuff.h>
44 #include <linux/notifier.h>
45 
46 struct user_namespace;
47 struct proc_dir_entry;
48 struct net_device;
49 struct sock;
50 struct ctl_table_header;
51 struct net_generic;
52 struct uevent_sock;
53 struct netns_ipvs;
54 struct bpf_prog;
55 
56 
57 #define NETDEV_HASHBITS    8
58 #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS)
59 
60 struct net {
61 	/* First cache line can be often dirtied.
62 	 * Do not place here read-mostly fields.
63 	 */
64 	refcount_t		passive;	/* To decide when the network
65 						 * namespace should be freed.
66 						 */
67 	spinlock_t		rules_mod_lock;
68 
69 	atomic_t		dev_unreg_count;
70 
71 	unsigned int		dev_base_seq;	/* protected by rtnl_mutex */
72 	int			ifindex;
73 
74 	spinlock_t		nsid_lock;
75 	atomic_t		fnhe_genid;
76 
77 	struct list_head	list;		/* list of network namespaces */
78 	struct list_head	exit_list;	/* To linked to call pernet exit
79 						 * methods on dead net (
80 						 * pernet_ops_rwsem read locked),
81 						 * or to unregister pernet ops
82 						 * (pernet_ops_rwsem write locked).
83 						 */
84 	struct llist_node	cleanup_list;	/* namespaces on death row */
85 
86 #ifdef CONFIG_KEYS
87 	struct key_tag		*key_domain;	/* Key domain of operation tag */
88 #endif
89 	struct user_namespace   *user_ns;	/* Owning user namespace */
90 	struct ucounts		*ucounts;
91 	struct idr		netns_ids;
92 
93 	struct ns_common	ns;
94 	struct ref_tracker_dir  refcnt_tracker;
95 	struct ref_tracker_dir  notrefcnt_tracker; /* tracker for objects not
96 						    * refcounted against netns
97 						    */
98 	struct list_head 	dev_base_head;
99 	struct proc_dir_entry 	*proc_net;
100 	struct proc_dir_entry 	*proc_net_stat;
101 
102 #ifdef CONFIG_SYSCTL
103 	struct ctl_table_set	sysctls;
104 #endif
105 
106 	struct sock 		*rtnl;			/* rtnetlink socket */
107 	struct sock		*genl_sock;
108 
109 	struct uevent_sock	*uevent_sock;		/* uevent socket */
110 
111 	struct hlist_head 	*dev_name_head;
112 	struct hlist_head	*dev_index_head;
113 	struct raw_notifier_head	netdev_chain;
114 
115 	/* Note that @hash_mix can be read millions times per second,
116 	 * it is critical that it is on a read_mostly cache line.
117 	 */
118 	u32			hash_mix;
119 
120 	struct net_device       *loopback_dev;          /* The loopback */
121 
122 	/* core fib_rules */
123 	struct list_head	rules_ops;
124 
125 	struct netns_core	core;
126 	struct netns_mib	mib;
127 	struct netns_packet	packet;
128 #if IS_ENABLED(CONFIG_UNIX)
129 	struct netns_unix	unx;
130 #endif
131 	struct netns_nexthop	nexthop;
132 	struct netns_ipv4	ipv4;
133 #if IS_ENABLED(CONFIG_IPV6)
134 	struct netns_ipv6	ipv6;
135 #endif
136 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
137 	struct netns_ieee802154_lowpan	ieee802154_lowpan;
138 #endif
139 #if defined(CONFIG_IP_SCTP) || defined(CONFIG_IP_SCTP_MODULE)
140 	struct netns_sctp	sctp;
141 #endif
142 #ifdef CONFIG_NETFILTER
143 	struct netns_nf		nf;
144 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
145 	struct netns_ct		ct;
146 #endif
147 #if defined(CONFIG_NF_TABLES) || defined(CONFIG_NF_TABLES_MODULE)
148 	struct netns_nftables	nft;
149 #endif
150 #if IS_ENABLED(CONFIG_NF_FLOW_TABLE)
151 	struct netns_ft ft;
152 #endif
153 #endif
154 #ifdef CONFIG_WEXT_CORE
155 	struct sk_buff_head	wext_nlevents;
156 #endif
157 	struct net_generic __rcu	*gen;
158 
159 	/* Used to store attached BPF programs */
160 	struct netns_bpf	bpf;
161 
162 	/* Note : following structs are cache line aligned */
163 #ifdef CONFIG_XFRM
164 	struct netns_xfrm	xfrm;
165 #endif
166 
167 	u64			net_cookie; /* written once */
168 
169 #if IS_ENABLED(CONFIG_IP_VS)
170 	struct netns_ipvs	*ipvs;
171 #endif
172 #if IS_ENABLED(CONFIG_MPLS)
173 	struct netns_mpls	mpls;
174 #endif
175 #if IS_ENABLED(CONFIG_CAN)
176 	struct netns_can	can;
177 #endif
178 #ifdef CONFIG_XDP_SOCKETS
179 	struct netns_xdp	xdp;
180 #endif
181 #if IS_ENABLED(CONFIG_MCTP)
182 	struct netns_mctp	mctp;
183 #endif
184 #if IS_ENABLED(CONFIG_CRYPTO_USER)
185 	struct sock		*crypto_nlsk;
186 #endif
187 	struct sock		*diag_nlsk;
188 #if IS_ENABLED(CONFIG_SMC)
189 	struct netns_smc	smc;
190 #endif
191 } __randomize_layout;
192 
193 #include <linux/seq_file_net.h>
194 
195 /* Init's network namespace */
196 extern struct net init_net;
197 
198 #ifdef CONFIG_NET_NS
199 struct net *copy_net_ns(unsigned long flags, struct user_namespace *user_ns,
200 			struct net *old_net);
201 
202 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid);
203 
204 void net_ns_barrier(void);
205 
206 struct ns_common *get_net_ns(struct ns_common *ns);
207 struct net *get_net_ns_by_fd(int fd);
208 #else /* CONFIG_NET_NS */
209 #include <linux/sched.h>
210 #include <linux/nsproxy.h>
211 static inline struct net *copy_net_ns(unsigned long flags,
212 	struct user_namespace *user_ns, struct net *old_net)
213 {
214 	if (flags & CLONE_NEWNET)
215 		return ERR_PTR(-EINVAL);
216 	return old_net;
217 }
218 
219 static inline void net_ns_get_ownership(const struct net *net,
220 					kuid_t *uid, kgid_t *gid)
221 {
222 	*uid = GLOBAL_ROOT_UID;
223 	*gid = GLOBAL_ROOT_GID;
224 }
225 
226 static inline void net_ns_barrier(void) {}
227 
228 static inline struct ns_common *get_net_ns(struct ns_common *ns)
229 {
230 	return ERR_PTR(-EINVAL);
231 }
232 
233 static inline struct net *get_net_ns_by_fd(int fd)
234 {
235 	return ERR_PTR(-EINVAL);
236 }
237 #endif /* CONFIG_NET_NS */
238 
239 
240 extern struct list_head net_namespace_list;
241 
242 struct net *get_net_ns_by_pid(pid_t pid);
243 
244 #ifdef CONFIG_SYSCTL
245 void ipx_register_sysctl(void);
246 void ipx_unregister_sysctl(void);
247 #else
248 #define ipx_register_sysctl()
249 #define ipx_unregister_sysctl()
250 #endif
251 
252 #ifdef CONFIG_NET_NS
253 void __put_net(struct net *net);
254 
255 /* Try using get_net_track() instead */
256 static inline struct net *get_net(struct net *net)
257 {
258 	refcount_inc(&net->ns.count);
259 	return net;
260 }
261 
262 static inline struct net *maybe_get_net(struct net *net)
263 {
264 	/* Used when we know struct net exists but we
265 	 * aren't guaranteed a previous reference count
266 	 * exists.  If the reference count is zero this
267 	 * function fails and returns NULL.
268 	 */
269 	if (!refcount_inc_not_zero(&net->ns.count))
270 		net = NULL;
271 	return net;
272 }
273 
274 /* Try using put_net_track() instead */
275 static inline void put_net(struct net *net)
276 {
277 	if (refcount_dec_and_test(&net->ns.count))
278 		__put_net(net);
279 }
280 
281 static inline
282 int net_eq(const struct net *net1, const struct net *net2)
283 {
284 	return net1 == net2;
285 }
286 
287 static inline int check_net(const struct net *net)
288 {
289 	return refcount_read(&net->ns.count) != 0;
290 }
291 
292 void net_drop_ns(void *);
293 
294 #else
295 
296 static inline struct net *get_net(struct net *net)
297 {
298 	return net;
299 }
300 
301 static inline void put_net(struct net *net)
302 {
303 }
304 
305 static inline struct net *maybe_get_net(struct net *net)
306 {
307 	return net;
308 }
309 
310 static inline
311 int net_eq(const struct net *net1, const struct net *net2)
312 {
313 	return 1;
314 }
315 
316 static inline int check_net(const struct net *net)
317 {
318 	return 1;
319 }
320 
321 #define net_drop_ns NULL
322 #endif
323 
324 
325 static inline void __netns_tracker_alloc(struct net *net,
326 					 netns_tracker *tracker,
327 					 bool refcounted,
328 					 gfp_t gfp)
329 {
330 #ifdef CONFIG_NET_NS_REFCNT_TRACKER
331 	ref_tracker_alloc(refcounted ? &net->refcnt_tracker :
332 				       &net->notrefcnt_tracker,
333 			  tracker, gfp);
334 #endif
335 }
336 
337 static inline void netns_tracker_alloc(struct net *net, netns_tracker *tracker,
338 				       gfp_t gfp)
339 {
340 	__netns_tracker_alloc(net, tracker, true, gfp);
341 }
342 
343 static inline void __netns_tracker_free(struct net *net,
344 					netns_tracker *tracker,
345 					bool refcounted)
346 {
347 #ifdef CONFIG_NET_NS_REFCNT_TRACKER
348        ref_tracker_free(refcounted ? &net->refcnt_tracker :
349 				     &net->notrefcnt_tracker, tracker);
350 #endif
351 }
352 
353 static inline struct net *get_net_track(struct net *net,
354 					netns_tracker *tracker, gfp_t gfp)
355 {
356 	get_net(net);
357 	netns_tracker_alloc(net, tracker, gfp);
358 	return net;
359 }
360 
361 static inline void put_net_track(struct net *net, netns_tracker *tracker)
362 {
363 	__netns_tracker_free(net, tracker, true);
364 	put_net(net);
365 }
366 
367 typedef struct {
368 #ifdef CONFIG_NET_NS
369 	struct net *net;
370 #endif
371 } possible_net_t;
372 
373 static inline void write_pnet(possible_net_t *pnet, struct net *net)
374 {
375 #ifdef CONFIG_NET_NS
376 	pnet->net = net;
377 #endif
378 }
379 
380 static inline struct net *read_pnet(const possible_net_t *pnet)
381 {
382 #ifdef CONFIG_NET_NS
383 	return pnet->net;
384 #else
385 	return &init_net;
386 #endif
387 }
388 
389 /* Protected by net_rwsem */
390 #define for_each_net(VAR)				\
391 	list_for_each_entry(VAR, &net_namespace_list, list)
392 #define for_each_net_continue_reverse(VAR)		\
393 	list_for_each_entry_continue_reverse(VAR, &net_namespace_list, list)
394 #define for_each_net_rcu(VAR)				\
395 	list_for_each_entry_rcu(VAR, &net_namespace_list, list)
396 
397 #ifdef CONFIG_NET_NS
398 #define __net_init
399 #define __net_exit
400 #define __net_initdata
401 #define __net_initconst
402 #else
403 #define __net_init	__init
404 #define __net_exit	__ref
405 #define __net_initdata	__initdata
406 #define __net_initconst	__initconst
407 #endif
408 
409 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp);
410 int peernet2id(const struct net *net, struct net *peer);
411 bool peernet_has_id(const struct net *net, struct net *peer);
412 struct net *get_net_ns_by_id(const struct net *net, int id);
413 
414 struct pernet_operations {
415 	struct list_head list;
416 	/*
417 	 * Below methods are called without any exclusive locks.
418 	 * More than one net may be constructed and destructed
419 	 * in parallel on several cpus. Every pernet_operations
420 	 * have to keep in mind all other pernet_operations and
421 	 * to introduce a locking, if they share common resources.
422 	 *
423 	 * The only time they are called with exclusive lock is
424 	 * from register_pernet_subsys(), unregister_pernet_subsys()
425 	 * register_pernet_device() and unregister_pernet_device().
426 	 *
427 	 * Exit methods using blocking RCU primitives, such as
428 	 * synchronize_rcu(), should be implemented via exit_batch.
429 	 * Then, destruction of a group of net requires single
430 	 * synchronize_rcu() related to these pernet_operations,
431 	 * instead of separate synchronize_rcu() for every net.
432 	 * Please, avoid synchronize_rcu() at all, where it's possible.
433 	 *
434 	 * Note that a combination of pre_exit() and exit() can
435 	 * be used, since a synchronize_rcu() is guaranteed between
436 	 * the calls.
437 	 */
438 	int (*init)(struct net *net);
439 	void (*pre_exit)(struct net *net);
440 	void (*exit)(struct net *net);
441 	void (*exit_batch)(struct list_head *net_exit_list);
442 	unsigned int *id;
443 	size_t size;
444 };
445 
446 /*
447  * Use these carefully.  If you implement a network device and it
448  * needs per network namespace operations use device pernet operations,
449  * otherwise use pernet subsys operations.
450  *
451  * Network interfaces need to be removed from a dying netns _before_
452  * subsys notifiers can be called, as most of the network code cleanup
453  * (which is done from subsys notifiers) runs with the assumption that
454  * dev_remove_pack has been called so no new packets will arrive during
455  * and after the cleanup functions have been called.  dev_remove_pack
456  * is not per namespace so instead the guarantee of no more packets
457  * arriving in a network namespace is provided by ensuring that all
458  * network devices and all sockets have left the network namespace
459  * before the cleanup methods are called.
460  *
461  * For the longest time the ipv4 icmp code was registered as a pernet
462  * device which caused kernel oops, and panics during network
463  * namespace cleanup.   So please don't get this wrong.
464  */
465 int register_pernet_subsys(struct pernet_operations *);
466 void unregister_pernet_subsys(struct pernet_operations *);
467 int register_pernet_device(struct pernet_operations *);
468 void unregister_pernet_device(struct pernet_operations *);
469 
470 struct ctl_table;
471 
472 #ifdef CONFIG_SYSCTL
473 int net_sysctl_init(void);
474 struct ctl_table_header *register_net_sysctl(struct net *net, const char *path,
475 					     struct ctl_table *table);
476 void unregister_net_sysctl_table(struct ctl_table_header *header);
477 #else
478 static inline int net_sysctl_init(void) { return 0; }
479 static inline struct ctl_table_header *register_net_sysctl(struct net *net,
480 	const char *path, struct ctl_table *table)
481 {
482 	return NULL;
483 }
484 static inline void unregister_net_sysctl_table(struct ctl_table_header *header)
485 {
486 }
487 #endif
488 
489 static inline int rt_genid_ipv4(const struct net *net)
490 {
491 	return atomic_read(&net->ipv4.rt_genid);
492 }
493 
494 #if IS_ENABLED(CONFIG_IPV6)
495 static inline int rt_genid_ipv6(const struct net *net)
496 {
497 	return atomic_read(&net->ipv6.fib6_sernum);
498 }
499 #endif
500 
501 static inline void rt_genid_bump_ipv4(struct net *net)
502 {
503 	atomic_inc(&net->ipv4.rt_genid);
504 }
505 
506 extern void (*__fib6_flush_trees)(struct net *net);
507 static inline void rt_genid_bump_ipv6(struct net *net)
508 {
509 	if (__fib6_flush_trees)
510 		__fib6_flush_trees(net);
511 }
512 
513 #if IS_ENABLED(CONFIG_IEEE802154_6LOWPAN)
514 static inline struct netns_ieee802154_lowpan *
515 net_ieee802154_lowpan(struct net *net)
516 {
517 	return &net->ieee802154_lowpan;
518 }
519 #endif
520 
521 /* For callers who don't really care about whether it's IPv4 or IPv6 */
522 static inline void rt_genid_bump_all(struct net *net)
523 {
524 	rt_genid_bump_ipv4(net);
525 	rt_genid_bump_ipv6(net);
526 }
527 
528 static inline int fnhe_genid(const struct net *net)
529 {
530 	return atomic_read(&net->fnhe_genid);
531 }
532 
533 static inline void fnhe_genid_bump(struct net *net)
534 {
535 	atomic_inc(&net->fnhe_genid);
536 }
537 
538 #ifdef CONFIG_NET
539 void net_ns_init(void);
540 #else
541 static inline void net_ns_init(void) {}
542 #endif
543 
544 #endif /* __NET_NET_NAMESPACE_H */
545