]>
Commit | Line | Data |
---|---|---|
e005d193 JP |
1 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
2 | ||
5f256bec EB |
3 | #include <linux/workqueue.h> |
4 | #include <linux/rtnetlink.h> | |
5 | #include <linux/cache.h> | |
6 | #include <linux/slab.h> | |
7 | #include <linux/list.h> | |
8 | #include <linux/delay.h> | |
9dd776b6 | 9 | #include <linux/sched.h> |
c93cf61f | 10 | #include <linux/idr.h> |
11a28d37 | 11 | #include <linux/rculist.h> |
30ffee84 | 12 | #include <linux/nsproxy.h> |
0bb80f24 DH |
13 | #include <linux/fs.h> |
14 | #include <linux/proc_ns.h> | |
f0630529 | 15 | #include <linux/file.h> |
bc3b2d7f | 16 | #include <linux/export.h> |
038e7332 | 17 | #include <linux/user_namespace.h> |
5f256bec | 18 | #include <net/net_namespace.h> |
dec827d1 | 19 | #include <net/netns/generic.h> |
5f256bec EB |
20 | |
21 | /* | |
22 | * Our network namespace constructor/destructor lists | |
23 | */ | |
24 | ||
25 | static LIST_HEAD(pernet_list); | |
26 | static struct list_head *first_device = &pernet_list; | |
200b916f | 27 | DEFINE_MUTEX(net_mutex); |
5f256bec | 28 | |
5f256bec | 29 | LIST_HEAD(net_namespace_list); |
b76a461f | 30 | EXPORT_SYMBOL_GPL(net_namespace_list); |
5f256bec | 31 | |
734b6541 RM |
32 | struct net init_net = { |
33 | .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head), | |
34 | }; | |
ff4b9502 | 35 | EXPORT_SYMBOL(init_net); |
5f256bec | 36 | |
dec827d1 PE |
37 | #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */ |
38 | ||
073862ba ED |
39 | static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS; |
40 | ||
41 | static struct net_generic *net_alloc_generic(void) | |
42 | { | |
43 | struct net_generic *ng; | |
44 | size_t generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]); | |
45 | ||
46 | ng = kzalloc(generic_size, GFP_KERNEL); | |
47 | if (ng) | |
48 | ng->len = max_gen_ptrs; | |
49 | ||
50 | return ng; | |
51 | } | |
52 | ||
05fceb4a JP |
53 | static int net_assign_generic(struct net *net, int id, void *data) |
54 | { | |
55 | struct net_generic *ng, *old_ng; | |
56 | ||
57 | BUG_ON(!mutex_is_locked(&net_mutex)); | |
58 | BUG_ON(id == 0); | |
59 | ||
1c87733d ED |
60 | old_ng = rcu_dereference_protected(net->gen, |
61 | lockdep_is_held(&net_mutex)); | |
62 | ng = old_ng; | |
05fceb4a JP |
63 | if (old_ng->len >= id) |
64 | goto assign; | |
65 | ||
073862ba | 66 | ng = net_alloc_generic(); |
05fceb4a JP |
67 | if (ng == NULL) |
68 | return -ENOMEM; | |
69 | ||
70 | /* | |
71 | * Some synchronisation notes: | |
72 | * | |
73 | * The net_generic explores the net->gen array inside rcu | |
74 | * read section. Besides once set the net->gen->ptr[x] | |
75 | * pointer never changes (see rules in netns/generic.h). | |
76 | * | |
77 | * That said, we simply duplicate this array and schedule | |
78 | * the old copy for kfree after a grace period. | |
79 | */ | |
80 | ||
05fceb4a JP |
81 | memcpy(&ng->ptr, &old_ng->ptr, old_ng->len * sizeof(void*)); |
82 | ||
83 | rcu_assign_pointer(net->gen, ng); | |
04d4dfed | 84 | kfree_rcu(old_ng, rcu); |
05fceb4a JP |
85 | assign: |
86 | ng->ptr[id - 1] = data; | |
87 | return 0; | |
88 | } | |
89 | ||
f875bae0 EB |
90 | static int ops_init(const struct pernet_operations *ops, struct net *net) |
91 | { | |
b922934d JA |
92 | int err = -ENOMEM; |
93 | void *data = NULL; | |
94 | ||
f875bae0 | 95 | if (ops->id && ops->size) { |
b922934d | 96 | data = kzalloc(ops->size, GFP_KERNEL); |
f875bae0 | 97 | if (!data) |
b922934d | 98 | goto out; |
f875bae0 EB |
99 | |
100 | err = net_assign_generic(net, *ops->id, data); | |
b922934d JA |
101 | if (err) |
102 | goto cleanup; | |
f875bae0 | 103 | } |
b922934d | 104 | err = 0; |
f875bae0 | 105 | if (ops->init) |
b922934d JA |
106 | err = ops->init(net); |
107 | if (!err) | |
108 | return 0; | |
109 | ||
110 | cleanup: | |
111 | kfree(data); | |
112 | ||
113 | out: | |
114 | return err; | |
f875bae0 EB |
115 | } |
116 | ||
117 | static void ops_free(const struct pernet_operations *ops, struct net *net) | |
118 | { | |
119 | if (ops->id && ops->size) { | |
120 | int id = *ops->id; | |
121 | kfree(net_generic(net, id)); | |
122 | } | |
123 | } | |
124 | ||
72ad937a EB |
125 | static void ops_exit_list(const struct pernet_operations *ops, |
126 | struct list_head *net_exit_list) | |
127 | { | |
128 | struct net *net; | |
129 | if (ops->exit) { | |
130 | list_for_each_entry(net, net_exit_list, exit_list) | |
131 | ops->exit(net); | |
132 | } | |
72ad937a EB |
133 | if (ops->exit_batch) |
134 | ops->exit_batch(net_exit_list); | |
135 | } | |
136 | ||
137 | static void ops_free_list(const struct pernet_operations *ops, | |
138 | struct list_head *net_exit_list) | |
139 | { | |
140 | struct net *net; | |
141 | if (ops->size && ops->id) { | |
142 | list_for_each_entry(net, net_exit_list, exit_list) | |
143 | ops_free(ops, net); | |
144 | } | |
145 | } | |
146 | ||
5f256bec EB |
147 | /* |
148 | * setup_net runs the initializers for the network namespace object. | |
149 | */ | |
038e7332 | 150 | static __net_init int setup_net(struct net *net, struct user_namespace *user_ns) |
5f256bec EB |
151 | { |
152 | /* Must be called with net_mutex held */ | |
f875bae0 | 153 | const struct pernet_operations *ops, *saved_ops; |
486a87f1 | 154 | int error = 0; |
72ad937a | 155 | LIST_HEAD(net_exit_list); |
5f256bec | 156 | |
5f256bec | 157 | atomic_set(&net->count, 1); |
a685e089 | 158 | atomic_set(&net->passive, 1); |
4e985ada | 159 | net->dev_base_seq = 1; |
038e7332 | 160 | net->user_ns = user_ns; |
486a87f1 | 161 | |
5d1e4468 | 162 | #ifdef NETNS_REFCNT_DEBUG |
5f256bec | 163 | atomic_set(&net->use_count, 0); |
5d1e4468 | 164 | #endif |
5f256bec | 165 | |
768f3591 | 166 | list_for_each_entry(ops, &pernet_list, list) { |
f875bae0 EB |
167 | error = ops_init(ops, net); |
168 | if (error < 0) | |
169 | goto out_undo; | |
5f256bec EB |
170 | } |
171 | out: | |
172 | return error; | |
768f3591 | 173 | |
5f256bec EB |
174 | out_undo: |
175 | /* Walk through the list backwards calling the exit functions | |
176 | * for the pernet modules whose init functions did not fail. | |
177 | */ | |
72ad937a | 178 | list_add(&net->exit_list, &net_exit_list); |
f875bae0 | 179 | saved_ops = ops; |
72ad937a EB |
180 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) |
181 | ops_exit_list(ops, &net_exit_list); | |
182 | ||
f875bae0 EB |
183 | ops = saved_ops; |
184 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) | |
72ad937a | 185 | ops_free_list(ops, &net_exit_list); |
310928d9 DL |
186 | |
187 | rcu_barrier(); | |
5f256bec EB |
188 | goto out; |
189 | } | |
190 | ||
6a1a3b9f | 191 | |
ebe47d47 CN |
192 | #ifdef CONFIG_NET_NS |
193 | static struct kmem_cache *net_cachep; | |
194 | static struct workqueue_struct *netns_wq; | |
195 | ||
486a87f1 | 196 | static struct net *net_alloc(void) |
45a19b0a | 197 | { |
486a87f1 DL |
198 | struct net *net = NULL; |
199 | struct net_generic *ng; | |
200 | ||
201 | ng = net_alloc_generic(); | |
202 | if (!ng) | |
203 | goto out; | |
204 | ||
205 | net = kmem_cache_zalloc(net_cachep, GFP_KERNEL); | |
45a19b0a | 206 | if (!net) |
486a87f1 | 207 | goto out_free; |
45a19b0a | 208 | |
486a87f1 DL |
209 | rcu_assign_pointer(net->gen, ng); |
210 | out: | |
211 | return net; | |
212 | ||
213 | out_free: | |
214 | kfree(ng); | |
215 | goto out; | |
216 | } | |
217 | ||
218 | static void net_free(struct net *net) | |
219 | { | |
5d1e4468 | 220 | #ifdef NETNS_REFCNT_DEBUG |
45a19b0a | 221 | if (unlikely(atomic_read(&net->use_count) != 0)) { |
e005d193 JP |
222 | pr_emerg("network namespace not free! Usage: %d\n", |
223 | atomic_read(&net->use_count)); | |
45a19b0a JFS |
224 | return; |
225 | } | |
5d1e4468 | 226 | #endif |
416c51e1 | 227 | kfree(rcu_access_pointer(net->gen)); |
45a19b0a JFS |
228 | kmem_cache_free(net_cachep, net); |
229 | } | |
230 | ||
a685e089 AV |
231 | void net_drop_ns(void *p) |
232 | { | |
233 | struct net *ns = p; | |
234 | if (ns && atomic_dec_and_test(&ns->passive)) | |
235 | net_free(ns); | |
236 | } | |
237 | ||
038e7332 EB |
238 | struct net *copy_net_ns(unsigned long flags, |
239 | struct user_namespace *user_ns, struct net *old_net) | |
9dd776b6 | 240 | { |
088eb2d9 AD |
241 | struct net *net; |
242 | int rv; | |
9dd776b6 | 243 | |
911cb193 RL |
244 | if (!(flags & CLONE_NEWNET)) |
245 | return get_net(old_net); | |
246 | ||
088eb2d9 AD |
247 | net = net_alloc(); |
248 | if (!net) | |
249 | return ERR_PTR(-ENOMEM); | |
038e7332 EB |
250 | |
251 | get_user_ns(user_ns); | |
252 | ||
9dd776b6 | 253 | mutex_lock(&net_mutex); |
038e7332 | 254 | rv = setup_net(net, user_ns); |
088eb2d9 | 255 | if (rv == 0) { |
486a87f1 | 256 | rtnl_lock(); |
11a28d37 | 257 | list_add_tail_rcu(&net->list, &net_namespace_list); |
486a87f1 DL |
258 | rtnl_unlock(); |
259 | } | |
9dd776b6 | 260 | mutex_unlock(&net_mutex); |
088eb2d9 | 261 | if (rv < 0) { |
038e7332 | 262 | put_user_ns(user_ns); |
a685e089 | 263 | net_drop_ns(net); |
088eb2d9 AD |
264 | return ERR_PTR(rv); |
265 | } | |
266 | return net; | |
267 | } | |
486a87f1 | 268 | |
2b035b39 EB |
269 | static DEFINE_SPINLOCK(cleanup_list_lock); |
270 | static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */ | |
271 | ||
6a1a3b9f PE |
272 | static void cleanup_net(struct work_struct *work) |
273 | { | |
f875bae0 | 274 | const struct pernet_operations *ops; |
2b035b39 | 275 | struct net *net, *tmp; |
1818ce4d | 276 | struct list_head net_kill_list; |
72ad937a | 277 | LIST_HEAD(net_exit_list); |
6a1a3b9f | 278 | |
2b035b39 EB |
279 | /* Atomically snapshot the list of namespaces to cleanup */ |
280 | spin_lock_irq(&cleanup_list_lock); | |
281 | list_replace_init(&cleanup_list, &net_kill_list); | |
282 | spin_unlock_irq(&cleanup_list_lock); | |
6a1a3b9f PE |
283 | |
284 | mutex_lock(&net_mutex); | |
285 | ||
286 | /* Don't let anyone else find us. */ | |
287 | rtnl_lock(); | |
72ad937a | 288 | list_for_each_entry(net, &net_kill_list, cleanup_list) { |
2b035b39 | 289 | list_del_rcu(&net->list); |
72ad937a EB |
290 | list_add_tail(&net->exit_list, &net_exit_list); |
291 | } | |
6a1a3b9f PE |
292 | rtnl_unlock(); |
293 | ||
11a28d37 JB |
294 | /* |
295 | * Another CPU might be rcu-iterating the list, wait for it. | |
296 | * This needs to be before calling the exit() notifiers, so | |
297 | * the rcu_barrier() below isn't sufficient alone. | |
298 | */ | |
299 | synchronize_rcu(); | |
300 | ||
6a1a3b9f | 301 | /* Run all of the network namespace exit methods */ |
72ad937a EB |
302 | list_for_each_entry_reverse(ops, &pernet_list, list) |
303 | ops_exit_list(ops, &net_exit_list); | |
304 | ||
f875bae0 | 305 | /* Free the net generic variables */ |
72ad937a EB |
306 | list_for_each_entry_reverse(ops, &pernet_list, list) |
307 | ops_free_list(ops, &net_exit_list); | |
6a1a3b9f PE |
308 | |
309 | mutex_unlock(&net_mutex); | |
310 | ||
311 | /* Ensure there are no outstanding rcu callbacks using this | |
312 | * network namespace. | |
313 | */ | |
314 | rcu_barrier(); | |
315 | ||
316 | /* Finally it is safe to free my network namespace structure */ | |
72ad937a EB |
317 | list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) { |
318 | list_del_init(&net->exit_list); | |
038e7332 | 319 | put_user_ns(net->user_ns); |
a685e089 | 320 | net_drop_ns(net); |
2b035b39 | 321 | } |
6a1a3b9f | 322 | } |
2b035b39 | 323 | static DECLARE_WORK(net_cleanup_work, cleanup_net); |
6a1a3b9f PE |
324 | |
325 | void __put_net(struct net *net) | |
326 | { | |
327 | /* Cleanup the network namespace in process context */ | |
2b035b39 EB |
328 | unsigned long flags; |
329 | ||
330 | spin_lock_irqsave(&cleanup_list_lock, flags); | |
331 | list_add(&net->cleanup_list, &cleanup_list); | |
332 | spin_unlock_irqrestore(&cleanup_list_lock, flags); | |
333 | ||
334 | queue_work(netns_wq, &net_cleanup_work); | |
6a1a3b9f PE |
335 | } |
336 | EXPORT_SYMBOL_GPL(__put_net); | |
337 | ||
956c9207 SR |
338 | struct net *get_net_ns_by_fd(int fd) |
339 | { | |
0bb80f24 | 340 | struct proc_ns *ei; |
956c9207 | 341 | struct file *file; |
33c42940 | 342 | struct ns_common *ns; |
956c9207 SR |
343 | struct net *net; |
344 | ||
956c9207 | 345 | file = proc_ns_fget(fd); |
c316e6a3 AV |
346 | if (IS_ERR(file)) |
347 | return ERR_CAST(file); | |
956c9207 | 348 | |
0bb80f24 | 349 | ei = get_proc_ns(file_inode(file)); |
33c42940 AV |
350 | ns = ei->ns; |
351 | if (ns->ops == &netns_operations) | |
352 | net = get_net(container_of(ns, struct net, ns)); | |
c316e6a3 AV |
353 | else |
354 | net = ERR_PTR(-EINVAL); | |
956c9207 | 355 | |
c316e6a3 | 356 | fput(file); |
956c9207 SR |
357 | return net; |
358 | } | |
359 | ||
6a1a3b9f | 360 | #else |
956c9207 SR |
361 | struct net *get_net_ns_by_fd(int fd) |
362 | { | |
363 | return ERR_PTR(-EINVAL); | |
364 | } | |
6a1a3b9f PE |
365 | #endif |
366 | ||
30ffee84 JB |
367 | struct net *get_net_ns_by_pid(pid_t pid) |
368 | { | |
369 | struct task_struct *tsk; | |
370 | struct net *net; | |
371 | ||
372 | /* Lookup the network namespace */ | |
373 | net = ERR_PTR(-ESRCH); | |
374 | rcu_read_lock(); | |
375 | tsk = find_task_by_vpid(pid); | |
376 | if (tsk) { | |
377 | struct nsproxy *nsproxy; | |
728dba3a EB |
378 | task_lock(tsk); |
379 | nsproxy = tsk->nsproxy; | |
30ffee84 JB |
380 | if (nsproxy) |
381 | net = get_net(nsproxy->net_ns); | |
728dba3a | 382 | task_unlock(tsk); |
30ffee84 JB |
383 | } |
384 | rcu_read_unlock(); | |
385 | return net; | |
386 | } | |
387 | EXPORT_SYMBOL_GPL(get_net_ns_by_pid); | |
388 | ||
98f842e6 EB |
389 | static __net_init int net_ns_net_init(struct net *net) |
390 | { | |
33c42940 AV |
391 | #ifdef CONFIG_NET_NS |
392 | net->ns.ops = &netns_operations; | |
393 | #endif | |
6344c433 | 394 | return ns_alloc_inum(&net->ns); |
98f842e6 EB |
395 | } |
396 | ||
397 | static __net_exit void net_ns_net_exit(struct net *net) | |
398 | { | |
6344c433 | 399 | ns_free_inum(&net->ns); |
98f842e6 EB |
400 | } |
401 | ||
402 | static struct pernet_operations __net_initdata net_ns_ops = { | |
403 | .init = net_ns_net_init, | |
404 | .exit = net_ns_net_exit, | |
405 | }; | |
406 | ||
5f256bec EB |
407 | static int __init net_ns_init(void) |
408 | { | |
486a87f1 | 409 | struct net_generic *ng; |
5f256bec | 410 | |
d57a9212 | 411 | #ifdef CONFIG_NET_NS |
5f256bec EB |
412 | net_cachep = kmem_cache_create("net_namespace", sizeof(struct net), |
413 | SMP_CACHE_BYTES, | |
414 | SLAB_PANIC, NULL); | |
3ef1355d BT |
415 | |
416 | /* Create workqueue for cleanup */ | |
417 | netns_wq = create_singlethread_workqueue("netns"); | |
418 | if (!netns_wq) | |
419 | panic("Could not create netns workq"); | |
d57a9212 | 420 | #endif |
3ef1355d | 421 | |
486a87f1 DL |
422 | ng = net_alloc_generic(); |
423 | if (!ng) | |
424 | panic("Could not allocate generic netns"); | |
425 | ||
426 | rcu_assign_pointer(init_net.gen, ng); | |
427 | ||
5f256bec | 428 | mutex_lock(&net_mutex); |
038e7332 | 429 | if (setup_net(&init_net, &init_user_ns)) |
ca0f3112 | 430 | panic("Could not setup the initial network namespace"); |
5f256bec | 431 | |
f4618d39 | 432 | rtnl_lock(); |
11a28d37 | 433 | list_add_tail_rcu(&init_net.list, &net_namespace_list); |
f4618d39 | 434 | rtnl_unlock(); |
5f256bec EB |
435 | |
436 | mutex_unlock(&net_mutex); | |
5f256bec | 437 | |
98f842e6 EB |
438 | register_pernet_subsys(&net_ns_ops); |
439 | ||
5f256bec EB |
440 | return 0; |
441 | } | |
442 | ||
443 | pure_initcall(net_ns_init); | |
444 | ||
ed160e83 | 445 | #ifdef CONFIG_NET_NS |
f875bae0 EB |
446 | static int __register_pernet_operations(struct list_head *list, |
447 | struct pernet_operations *ops) | |
5f256bec | 448 | { |
72ad937a | 449 | struct net *net; |
5f256bec | 450 | int error; |
72ad937a | 451 | LIST_HEAD(net_exit_list); |
5f256bec | 452 | |
5f256bec | 453 | list_add_tail(&ops->list, list); |
f875bae0 | 454 | if (ops->init || (ops->id && ops->size)) { |
1dba323b | 455 | for_each_net(net) { |
f875bae0 | 456 | error = ops_init(ops, net); |
5f256bec EB |
457 | if (error) |
458 | goto out_undo; | |
72ad937a | 459 | list_add_tail(&net->exit_list, &net_exit_list); |
5f256bec EB |
460 | } |
461 | } | |
1dba323b | 462 | return 0; |
5f256bec EB |
463 | |
464 | out_undo: | |
465 | /* If I have an error cleanup all namespaces I initialized */ | |
466 | list_del(&ops->list); | |
72ad937a EB |
467 | ops_exit_list(ops, &net_exit_list); |
468 | ops_free_list(ops, &net_exit_list); | |
1dba323b | 469 | return error; |
5f256bec EB |
470 | } |
471 | ||
f875bae0 | 472 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
5f256bec EB |
473 | { |
474 | struct net *net; | |
72ad937a | 475 | LIST_HEAD(net_exit_list); |
5f256bec EB |
476 | |
477 | list_del(&ops->list); | |
72ad937a EB |
478 | for_each_net(net) |
479 | list_add_tail(&net->exit_list, &net_exit_list); | |
480 | ops_exit_list(ops, &net_exit_list); | |
481 | ops_free_list(ops, &net_exit_list); | |
5f256bec EB |
482 | } |
483 | ||
ed160e83 DL |
484 | #else |
485 | ||
f875bae0 EB |
486 | static int __register_pernet_operations(struct list_head *list, |
487 | struct pernet_operations *ops) | |
ed160e83 | 488 | { |
b922934d | 489 | return ops_init(ops, &init_net); |
ed160e83 DL |
490 | } |
491 | ||
f875bae0 | 492 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
ed160e83 | 493 | { |
72ad937a EB |
494 | LIST_HEAD(net_exit_list); |
495 | list_add(&init_net.exit_list, &net_exit_list); | |
496 | ops_exit_list(ops, &net_exit_list); | |
497 | ops_free_list(ops, &net_exit_list); | |
ed160e83 | 498 | } |
f875bae0 EB |
499 | |
500 | #endif /* CONFIG_NET_NS */ | |
ed160e83 | 501 | |
c93cf61f PE |
502 | static DEFINE_IDA(net_generic_ids); |
503 | ||
f875bae0 EB |
504 | static int register_pernet_operations(struct list_head *list, |
505 | struct pernet_operations *ops) | |
506 | { | |
507 | int error; | |
508 | ||
509 | if (ops->id) { | |
510 | again: | |
511 | error = ida_get_new_above(&net_generic_ids, 1, ops->id); | |
512 | if (error < 0) { | |
513 | if (error == -EAGAIN) { | |
514 | ida_pre_get(&net_generic_ids, GFP_KERNEL); | |
515 | goto again; | |
516 | } | |
517 | return error; | |
518 | } | |
073862ba | 519 | max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id); |
f875bae0 EB |
520 | } |
521 | error = __register_pernet_operations(list, ops); | |
3a765eda EB |
522 | if (error) { |
523 | rcu_barrier(); | |
524 | if (ops->id) | |
525 | ida_remove(&net_generic_ids, *ops->id); | |
526 | } | |
f875bae0 EB |
527 | |
528 | return error; | |
529 | } | |
530 | ||
531 | static void unregister_pernet_operations(struct pernet_operations *ops) | |
532 | { | |
533 | ||
534 | __unregister_pernet_operations(ops); | |
3a765eda | 535 | rcu_barrier(); |
f875bae0 EB |
536 | if (ops->id) |
537 | ida_remove(&net_generic_ids, *ops->id); | |
538 | } | |
539 | ||
5f256bec EB |
540 | /** |
541 | * register_pernet_subsys - register a network namespace subsystem | |
542 | * @ops: pernet operations structure for the subsystem | |
543 | * | |
544 | * Register a subsystem which has init and exit functions | |
545 | * that are called when network namespaces are created and | |
546 | * destroyed respectively. | |
547 | * | |
548 | * When registered all network namespace init functions are | |
549 | * called for every existing network namespace. Allowing kernel | |
550 | * modules to have a race free view of the set of network namespaces. | |
551 | * | |
552 | * When a new network namespace is created all of the init | |
553 | * methods are called in the order in which they were registered. | |
554 | * | |
555 | * When a network namespace is destroyed all of the exit methods | |
556 | * are called in the reverse of the order with which they were | |
557 | * registered. | |
558 | */ | |
559 | int register_pernet_subsys(struct pernet_operations *ops) | |
560 | { | |
561 | int error; | |
562 | mutex_lock(&net_mutex); | |
563 | error = register_pernet_operations(first_device, ops); | |
564 | mutex_unlock(&net_mutex); | |
565 | return error; | |
566 | } | |
567 | EXPORT_SYMBOL_GPL(register_pernet_subsys); | |
568 | ||
569 | /** | |
570 | * unregister_pernet_subsys - unregister a network namespace subsystem | |
571 | * @ops: pernet operations structure to manipulate | |
572 | * | |
573 | * Remove the pernet operations structure from the list to be | |
53379e57 | 574 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
575 | * addition run the exit method for all existing network |
576 | * namespaces. | |
577 | */ | |
b3c981d2 | 578 | void unregister_pernet_subsys(struct pernet_operations *ops) |
5f256bec EB |
579 | { |
580 | mutex_lock(&net_mutex); | |
b3c981d2 | 581 | unregister_pernet_operations(ops); |
5f256bec EB |
582 | mutex_unlock(&net_mutex); |
583 | } | |
584 | EXPORT_SYMBOL_GPL(unregister_pernet_subsys); | |
585 | ||
586 | /** | |
587 | * register_pernet_device - register a network namespace device | |
588 | * @ops: pernet operations structure for the subsystem | |
589 | * | |
590 | * Register a device which has init and exit functions | |
591 | * that are called when network namespaces are created and | |
592 | * destroyed respectively. | |
593 | * | |
594 | * When registered all network namespace init functions are | |
595 | * called for every existing network namespace. Allowing kernel | |
596 | * modules to have a race free view of the set of network namespaces. | |
597 | * | |
598 | * When a new network namespace is created all of the init | |
599 | * methods are called in the order in which they were registered. | |
600 | * | |
601 | * When a network namespace is destroyed all of the exit methods | |
602 | * are called in the reverse of the order with which they were | |
603 | * registered. | |
604 | */ | |
605 | int register_pernet_device(struct pernet_operations *ops) | |
606 | { | |
607 | int error; | |
608 | mutex_lock(&net_mutex); | |
609 | error = register_pernet_operations(&pernet_list, ops); | |
610 | if (!error && (first_device == &pernet_list)) | |
611 | first_device = &ops->list; | |
612 | mutex_unlock(&net_mutex); | |
613 | return error; | |
614 | } | |
615 | EXPORT_SYMBOL_GPL(register_pernet_device); | |
616 | ||
617 | /** | |
618 | * unregister_pernet_device - unregister a network namespace netdevice | |
619 | * @ops: pernet operations structure to manipulate | |
620 | * | |
621 | * Remove the pernet operations structure from the list to be | |
53379e57 | 622 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
623 | * addition run the exit method for all existing network |
624 | * namespaces. | |
625 | */ | |
626 | void unregister_pernet_device(struct pernet_operations *ops) | |
627 | { | |
628 | mutex_lock(&net_mutex); | |
629 | if (&ops->list == first_device) | |
630 | first_device = first_device->next; | |
631 | unregister_pernet_operations(ops); | |
632 | mutex_unlock(&net_mutex); | |
633 | } | |
634 | EXPORT_SYMBOL_GPL(unregister_pernet_device); | |
13b6f576 EB |
635 | |
636 | #ifdef CONFIG_NET_NS | |
64964528 | 637 | static struct ns_common *netns_get(struct task_struct *task) |
13b6f576 | 638 | { |
f0630529 EB |
639 | struct net *net = NULL; |
640 | struct nsproxy *nsproxy; | |
641 | ||
728dba3a EB |
642 | task_lock(task); |
643 | nsproxy = task->nsproxy; | |
f0630529 EB |
644 | if (nsproxy) |
645 | net = get_net(nsproxy->net_ns); | |
728dba3a | 646 | task_unlock(task); |
f0630529 | 647 | |
ff24870f AV |
648 | return net ? &net->ns : NULL; |
649 | } | |
650 | ||
651 | static inline struct net *to_net_ns(struct ns_common *ns) | |
652 | { | |
653 | return container_of(ns, struct net, ns); | |
13b6f576 EB |
654 | } |
655 | ||
64964528 | 656 | static void netns_put(struct ns_common *ns) |
13b6f576 | 657 | { |
ff24870f | 658 | put_net(to_net_ns(ns)); |
13b6f576 EB |
659 | } |
660 | ||
64964528 | 661 | static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns) |
13b6f576 | 662 | { |
ff24870f | 663 | struct net *net = to_net_ns(ns); |
142e1d1d | 664 | |
5e4a0847 | 665 | if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) || |
c7b96acf | 666 | !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) |
142e1d1d EB |
667 | return -EPERM; |
668 | ||
13b6f576 | 669 | put_net(nsproxy->net_ns); |
142e1d1d | 670 | nsproxy->net_ns = get_net(net); |
13b6f576 EB |
671 | return 0; |
672 | } | |
673 | ||
674 | const struct proc_ns_operations netns_operations = { | |
675 | .name = "net", | |
676 | .type = CLONE_NEWNET, | |
677 | .get = netns_get, | |
678 | .put = netns_put, | |
679 | .install = netns_install, | |
680 | }; | |
681 | #endif |