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