]>
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> |
0c7aecd4 | 18 | #include <linux/net_namespace.h> |
f719ff9b IM |
19 | #include <linux/sched/task.h> |
20 | ||
0c7aecd4 ND |
21 | #include <net/sock.h> |
22 | #include <net/netlink.h> | |
5f256bec | 23 | #include <net/net_namespace.h> |
dec827d1 | 24 | #include <net/netns/generic.h> |
5f256bec EB |
25 | |
26 | /* | |
27 | * Our network namespace constructor/destructor lists | |
28 | */ | |
29 | ||
30 | static LIST_HEAD(pernet_list); | |
31 | static struct list_head *first_device = &pernet_list; | |
200b916f | 32 | DEFINE_MUTEX(net_mutex); |
5f256bec | 33 | |
5f256bec | 34 | LIST_HEAD(net_namespace_list); |
b76a461f | 35 | EXPORT_SYMBOL_GPL(net_namespace_list); |
5f256bec | 36 | |
734b6541 | 37 | struct net init_net = { |
b5082df8 DH |
38 | .count = ATOMIC_INIT(1), |
39 | .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head), | |
734b6541 | 40 | }; |
ff4b9502 | 41 | EXPORT_SYMBOL(init_net); |
5f256bec | 42 | |
f8c46cb3 DT |
43 | static bool init_net_initialized; |
44 | ||
6af2d5ff AD |
45 | #define MIN_PERNET_OPS_ID \ |
46 | ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *)) | |
47 | ||
dec827d1 PE |
48 | #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */ |
49 | ||
073862ba ED |
50 | static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS; |
51 | ||
52 | static struct net_generic *net_alloc_generic(void) | |
53 | { | |
54 | struct net_generic *ng; | |
6af2d5ff | 55 | unsigned int generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]); |
073862ba ED |
56 | |
57 | ng = kzalloc(generic_size, GFP_KERNEL); | |
58 | if (ng) | |
9bfc7b99 | 59 | ng->s.len = max_gen_ptrs; |
073862ba ED |
60 | |
61 | return ng; | |
62 | } | |
63 | ||
c7d03a00 | 64 | static int net_assign_generic(struct net *net, unsigned int id, void *data) |
05fceb4a JP |
65 | { |
66 | struct net_generic *ng, *old_ng; | |
67 | ||
68 | BUG_ON(!mutex_is_locked(&net_mutex)); | |
6af2d5ff | 69 | BUG_ON(id < MIN_PERNET_OPS_ID); |
05fceb4a | 70 | |
1c87733d ED |
71 | old_ng = rcu_dereference_protected(net->gen, |
72 | lockdep_is_held(&net_mutex)); | |
6af2d5ff AD |
73 | if (old_ng->s.len > id) { |
74 | old_ng->ptr[id] = data; | |
1a9a0592 AD |
75 | return 0; |
76 | } | |
05fceb4a | 77 | |
073862ba | 78 | ng = net_alloc_generic(); |
05fceb4a JP |
79 | if (ng == NULL) |
80 | return -ENOMEM; | |
81 | ||
82 | /* | |
83 | * Some synchronisation notes: | |
84 | * | |
85 | * The net_generic explores the net->gen array inside rcu | |
86 | * read section. Besides once set the net->gen->ptr[x] | |
87 | * pointer never changes (see rules in netns/generic.h). | |
88 | * | |
89 | * That said, we simply duplicate this array and schedule | |
90 | * the old copy for kfree after a grace period. | |
91 | */ | |
92 | ||
6af2d5ff AD |
93 | memcpy(&ng->ptr[MIN_PERNET_OPS_ID], &old_ng->ptr[MIN_PERNET_OPS_ID], |
94 | (old_ng->s.len - MIN_PERNET_OPS_ID) * sizeof(void *)); | |
95 | ng->ptr[id] = data; | |
05fceb4a JP |
96 | |
97 | rcu_assign_pointer(net->gen, ng); | |
9bfc7b99 | 98 | kfree_rcu(old_ng, s.rcu); |
05fceb4a JP |
99 | return 0; |
100 | } | |
101 | ||
f875bae0 EB |
102 | static int ops_init(const struct pernet_operations *ops, struct net *net) |
103 | { | |
b922934d JA |
104 | int err = -ENOMEM; |
105 | void *data = NULL; | |
106 | ||
f875bae0 | 107 | if (ops->id && ops->size) { |
b922934d | 108 | data = kzalloc(ops->size, GFP_KERNEL); |
f875bae0 | 109 | if (!data) |
b922934d | 110 | goto out; |
f875bae0 EB |
111 | |
112 | err = net_assign_generic(net, *ops->id, data); | |
b922934d JA |
113 | if (err) |
114 | goto cleanup; | |
f875bae0 | 115 | } |
b922934d | 116 | err = 0; |
f875bae0 | 117 | if (ops->init) |
b922934d JA |
118 | err = ops->init(net); |
119 | if (!err) | |
120 | return 0; | |
121 | ||
122 | cleanup: | |
123 | kfree(data); | |
124 | ||
125 | out: | |
126 | return err; | |
f875bae0 EB |
127 | } |
128 | ||
129 | static void ops_free(const struct pernet_operations *ops, struct net *net) | |
130 | { | |
131 | if (ops->id && ops->size) { | |
c7d03a00 | 132 | kfree(net_generic(net, *ops->id)); |
f875bae0 EB |
133 | } |
134 | } | |
135 | ||
72ad937a EB |
136 | static void ops_exit_list(const struct pernet_operations *ops, |
137 | struct list_head *net_exit_list) | |
138 | { | |
139 | struct net *net; | |
140 | if (ops->exit) { | |
141 | list_for_each_entry(net, net_exit_list, exit_list) | |
142 | ops->exit(net); | |
143 | } | |
72ad937a EB |
144 | if (ops->exit_batch) |
145 | ops->exit_batch(net_exit_list); | |
146 | } | |
147 | ||
148 | static void ops_free_list(const struct pernet_operations *ops, | |
149 | struct list_head *net_exit_list) | |
150 | { | |
151 | struct net *net; | |
152 | if (ops->size && ops->id) { | |
153 | list_for_each_entry(net, net_exit_list, exit_list) | |
154 | ops_free(ops, net); | |
155 | } | |
156 | } | |
157 | ||
95f38411 | 158 | /* should be called with nsid_lock held */ |
0c7aecd4 ND |
159 | static int alloc_netid(struct net *net, struct net *peer, int reqid) |
160 | { | |
3138dbf8 | 161 | int min = 0, max = 0; |
0c7aecd4 | 162 | |
0c7aecd4 ND |
163 | if (reqid >= 0) { |
164 | min = reqid; | |
165 | max = reqid + 1; | |
166 | } | |
167 | ||
95f38411 | 168 | return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC); |
0c7aecd4 ND |
169 | } |
170 | ||
171 | /* This function is used by idr_for_each(). If net is equal to peer, the | |
172 | * function returns the id so that idr_for_each() stops. Because we cannot | |
173 | * returns the id 0 (idr_for_each() will not stop), we return the magic value | |
174 | * NET_ID_ZERO (-1) for it. | |
175 | */ | |
176 | #define NET_ID_ZERO -1 | |
177 | static int net_eq_idr(int id, void *net, void *peer) | |
178 | { | |
179 | if (net_eq(net, peer)) | |
180 | return id ? : NET_ID_ZERO; | |
181 | return 0; | |
182 | } | |
183 | ||
95f38411 ND |
184 | /* Should be called with nsid_lock held. If a new id is assigned, the bool alloc |
185 | * is set to true, thus the caller knows that the new id must be notified via | |
186 | * rtnl. | |
187 | */ | |
3138dbf8 | 188 | static int __peernet2id_alloc(struct net *net, struct net *peer, bool *alloc) |
0c7aecd4 ND |
189 | { |
190 | int id = idr_for_each(&net->netns_ids, net_eq_idr, peer); | |
3138dbf8 | 191 | bool alloc_it = *alloc; |
0c7aecd4 | 192 | |
3138dbf8 ND |
193 | *alloc = false; |
194 | ||
0c7aecd4 ND |
195 | /* Magic value for id 0. */ |
196 | if (id == NET_ID_ZERO) | |
197 | return 0; | |
198 | if (id > 0) | |
199 | return id; | |
200 | ||
3138dbf8 | 201 | if (alloc_it) { |
109582af | 202 | id = alloc_netid(net, peer, -1); |
3138dbf8 | 203 | *alloc = true; |
109582af ND |
204 | return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED; |
205 | } | |
0c7aecd4 | 206 | |
109582af | 207 | return NETNSA_NSID_NOT_ASSIGNED; |
0c7aecd4 ND |
208 | } |
209 | ||
95f38411 | 210 | /* should be called with nsid_lock held */ |
3138dbf8 ND |
211 | static int __peernet2id(struct net *net, struct net *peer) |
212 | { | |
213 | bool no = false; | |
214 | ||
215 | return __peernet2id_alloc(net, peer, &no); | |
216 | } | |
217 | ||
218 | static void rtnl_net_notifyid(struct net *net, int cmd, int id); | |
0c7aecd4 ND |
219 | /* This function returns the id of a peer netns. If no id is assigned, one will |
220 | * be allocated and returned. | |
221 | */ | |
7a0877d4 | 222 | int peernet2id_alloc(struct net *net, struct net *peer) |
0c7aecd4 | 223 | { |
95f38411 | 224 | bool alloc; |
3138dbf8 | 225 | int id; |
0c7aecd4 | 226 | |
cfc44a4d WC |
227 | if (atomic_read(&net->count) == 0) |
228 | return NETNSA_NSID_NOT_ASSIGNED; | |
fba143c6 | 229 | spin_lock_bh(&net->nsid_lock); |
95f38411 | 230 | alloc = atomic_read(&peer->count) == 0 ? false : true; |
3138dbf8 | 231 | id = __peernet2id_alloc(net, peer, &alloc); |
fba143c6 | 232 | spin_unlock_bh(&net->nsid_lock); |
3138dbf8 ND |
233 | if (alloc && id >= 0) |
234 | rtnl_net_notifyid(net, RTM_NEWNSID, id); | |
235 | return id; | |
0c7aecd4 | 236 | } |
7cbebc8a | 237 | EXPORT_SYMBOL_GPL(peernet2id_alloc); |
0c7aecd4 | 238 | |
95f38411 | 239 | /* This function returns, if assigned, the id of a peer netns. */ |
59324cf3 | 240 | int peernet2id(struct net *net, struct net *peer) |
95f38411 | 241 | { |
95f38411 ND |
242 | int id; |
243 | ||
fba143c6 | 244 | spin_lock_bh(&net->nsid_lock); |
95f38411 | 245 | id = __peernet2id(net, peer); |
fba143c6 | 246 | spin_unlock_bh(&net->nsid_lock); |
95f38411 ND |
247 | return id; |
248 | } | |
38f507f1 | 249 | EXPORT_SYMBOL(peernet2id); |
95f38411 | 250 | |
59324cf3 ND |
251 | /* This function returns true is the peer netns has an id assigned into the |
252 | * current netns. | |
253 | */ | |
254 | bool peernet_has_id(struct net *net, struct net *peer) | |
255 | { | |
256 | return peernet2id(net, peer) >= 0; | |
257 | } | |
258 | ||
0c7aecd4 ND |
259 | struct net *get_net_ns_by_id(struct net *net, int id) |
260 | { | |
261 | struct net *peer; | |
262 | ||
263 | if (id < 0) | |
264 | return NULL; | |
265 | ||
266 | rcu_read_lock(); | |
fba143c6 | 267 | spin_lock_bh(&net->nsid_lock); |
0c7aecd4 ND |
268 | peer = idr_find(&net->netns_ids, id); |
269 | if (peer) | |
21b59443 | 270 | peer = maybe_get_net(peer); |
fba143c6 | 271 | spin_unlock_bh(&net->nsid_lock); |
0c7aecd4 ND |
272 | rcu_read_unlock(); |
273 | ||
274 | return peer; | |
275 | } | |
276 | ||
5f256bec EB |
277 | /* |
278 | * setup_net runs the initializers for the network namespace object. | |
279 | */ | |
038e7332 | 280 | static __net_init int setup_net(struct net *net, struct user_namespace *user_ns) |
5f256bec EB |
281 | { |
282 | /* Must be called with net_mutex held */ | |
f875bae0 | 283 | const struct pernet_operations *ops, *saved_ops; |
486a87f1 | 284 | int error = 0; |
72ad937a | 285 | LIST_HEAD(net_exit_list); |
5f256bec | 286 | |
5f256bec | 287 | atomic_set(&net->count, 1); |
c122e14d | 288 | refcount_set(&net->passive, 1); |
4e985ada | 289 | net->dev_base_seq = 1; |
038e7332 | 290 | net->user_ns = user_ns; |
0c7aecd4 | 291 | idr_init(&net->netns_ids); |
de133464 | 292 | spin_lock_init(&net->nsid_lock); |
486a87f1 | 293 | |
768f3591 | 294 | list_for_each_entry(ops, &pernet_list, list) { |
f875bae0 EB |
295 | error = ops_init(ops, net); |
296 | if (error < 0) | |
297 | goto out_undo; | |
5f256bec EB |
298 | } |
299 | out: | |
300 | return error; | |
768f3591 | 301 | |
5f256bec EB |
302 | out_undo: |
303 | /* Walk through the list backwards calling the exit functions | |
304 | * for the pernet modules whose init functions did not fail. | |
305 | */ | |
72ad937a | 306 | list_add(&net->exit_list, &net_exit_list); |
f875bae0 | 307 | saved_ops = ops; |
72ad937a EB |
308 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) |
309 | ops_exit_list(ops, &net_exit_list); | |
310 | ||
f875bae0 EB |
311 | ops = saved_ops; |
312 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) | |
72ad937a | 313 | ops_free_list(ops, &net_exit_list); |
310928d9 DL |
314 | |
315 | rcu_barrier(); | |
5f256bec EB |
316 | goto out; |
317 | } | |
318 | ||
7c3f1875 RK |
319 | static int __net_init net_defaults_init_net(struct net *net) |
320 | { | |
321 | net->core.sysctl_somaxconn = SOMAXCONN; | |
322 | return 0; | |
323 | } | |
324 | ||
325 | static struct pernet_operations net_defaults_ops = { | |
326 | .init = net_defaults_init_net, | |
327 | }; | |
328 | ||
329 | static __init int net_defaults_init(void) | |
330 | { | |
331 | if (register_pernet_subsys(&net_defaults_ops)) | |
332 | panic("Cannot initialize net default settings"); | |
333 | ||
334 | return 0; | |
335 | } | |
336 | ||
337 | core_initcall(net_defaults_init); | |
6a1a3b9f | 338 | |
ebe47d47 | 339 | #ifdef CONFIG_NET_NS |
2ed6afde AB |
340 | static struct ucounts *inc_net_namespaces(struct user_namespace *ns) |
341 | { | |
342 | return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES); | |
343 | } | |
344 | ||
345 | static void dec_net_namespaces(struct ucounts *ucounts) | |
346 | { | |
347 | dec_ucount(ucounts, UCOUNT_NET_NAMESPACES); | |
348 | } | |
349 | ||
ebe47d47 CN |
350 | static struct kmem_cache *net_cachep; |
351 | static struct workqueue_struct *netns_wq; | |
352 | ||
486a87f1 | 353 | static struct net *net_alloc(void) |
45a19b0a | 354 | { |
486a87f1 DL |
355 | struct net *net = NULL; |
356 | struct net_generic *ng; | |
357 | ||
358 | ng = net_alloc_generic(); | |
359 | if (!ng) | |
360 | goto out; | |
361 | ||
362 | net = kmem_cache_zalloc(net_cachep, GFP_KERNEL); | |
45a19b0a | 363 | if (!net) |
486a87f1 | 364 | goto out_free; |
45a19b0a | 365 | |
486a87f1 DL |
366 | rcu_assign_pointer(net->gen, ng); |
367 | out: | |
368 | return net; | |
369 | ||
370 | out_free: | |
371 | kfree(ng); | |
372 | goto out; | |
373 | } | |
374 | ||
375 | static void net_free(struct net *net) | |
376 | { | |
416c51e1 | 377 | kfree(rcu_access_pointer(net->gen)); |
45a19b0a JFS |
378 | kmem_cache_free(net_cachep, net); |
379 | } | |
380 | ||
a685e089 AV |
381 | void net_drop_ns(void *p) |
382 | { | |
383 | struct net *ns = p; | |
c122e14d | 384 | if (ns && refcount_dec_and_test(&ns->passive)) |
a685e089 AV |
385 | net_free(ns); |
386 | } | |
387 | ||
038e7332 EB |
388 | struct net *copy_net_ns(unsigned long flags, |
389 | struct user_namespace *user_ns, struct net *old_net) | |
9dd776b6 | 390 | { |
70328660 | 391 | struct ucounts *ucounts; |
088eb2d9 AD |
392 | struct net *net; |
393 | int rv; | |
9dd776b6 | 394 | |
911cb193 RL |
395 | if (!(flags & CLONE_NEWNET)) |
396 | return get_net(old_net); | |
397 | ||
70328660 EB |
398 | ucounts = inc_net_namespaces(user_ns); |
399 | if (!ucounts) | |
df75e774 | 400 | return ERR_PTR(-ENOSPC); |
70328660 | 401 | |
088eb2d9 | 402 | net = net_alloc(); |
70328660 EB |
403 | if (!net) { |
404 | dec_net_namespaces(ucounts); | |
088eb2d9 | 405 | return ERR_PTR(-ENOMEM); |
70328660 | 406 | } |
038e7332 EB |
407 | |
408 | get_user_ns(user_ns); | |
409 | ||
7281a665 AV |
410 | rv = mutex_lock_killable(&net_mutex); |
411 | if (rv < 0) { | |
412 | net_free(net); | |
413 | dec_net_namespaces(ucounts); | |
414 | put_user_ns(user_ns); | |
415 | return ERR_PTR(rv); | |
416 | } | |
417 | ||
70328660 | 418 | net->ucounts = ucounts; |
038e7332 | 419 | rv = setup_net(net, user_ns); |
088eb2d9 | 420 | if (rv == 0) { |
486a87f1 | 421 | rtnl_lock(); |
11a28d37 | 422 | list_add_tail_rcu(&net->list, &net_namespace_list); |
486a87f1 DL |
423 | rtnl_unlock(); |
424 | } | |
9dd776b6 | 425 | mutex_unlock(&net_mutex); |
088eb2d9 | 426 | if (rv < 0) { |
70328660 | 427 | dec_net_namespaces(ucounts); |
038e7332 | 428 | put_user_ns(user_ns); |
a685e089 | 429 | net_drop_ns(net); |
088eb2d9 AD |
430 | return ERR_PTR(rv); |
431 | } | |
432 | return net; | |
433 | } | |
486a87f1 | 434 | |
2b035b39 EB |
435 | static DEFINE_SPINLOCK(cleanup_list_lock); |
436 | static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */ | |
437 | ||
6a1a3b9f PE |
438 | static void cleanup_net(struct work_struct *work) |
439 | { | |
f875bae0 | 440 | const struct pernet_operations *ops; |
6d458f5b | 441 | struct net *net, *tmp; |
1818ce4d | 442 | struct list_head net_kill_list; |
72ad937a | 443 | LIST_HEAD(net_exit_list); |
6a1a3b9f | 444 | |
2b035b39 EB |
445 | /* Atomically snapshot the list of namespaces to cleanup */ |
446 | spin_lock_irq(&cleanup_list_lock); | |
447 | list_replace_init(&cleanup_list, &net_kill_list); | |
448 | spin_unlock_irq(&cleanup_list_lock); | |
6a1a3b9f PE |
449 | |
450 | mutex_lock(&net_mutex); | |
451 | ||
452 | /* Don't let anyone else find us. */ | |
453 | rtnl_lock(); | |
72ad937a | 454 | list_for_each_entry(net, &net_kill_list, cleanup_list) { |
2b035b39 | 455 | list_del_rcu(&net->list); |
72ad937a | 456 | list_add_tail(&net->exit_list, &net_exit_list); |
6d458f5b | 457 | for_each_net(tmp) { |
95f38411 | 458 | int id; |
6d458f5b | 459 | |
fba143c6 | 460 | spin_lock_bh(&tmp->nsid_lock); |
95f38411 ND |
461 | id = __peernet2id(tmp, net); |
462 | if (id >= 0) | |
6d458f5b | 463 | idr_remove(&tmp->netns_ids, id); |
fba143c6 | 464 | spin_unlock_bh(&tmp->nsid_lock); |
95f38411 ND |
465 | if (id >= 0) |
466 | rtnl_net_notifyid(tmp, RTM_DELNSID, id); | |
6d458f5b | 467 | } |
fba143c6 | 468 | spin_lock_bh(&net->nsid_lock); |
6d458f5b | 469 | idr_destroy(&net->netns_ids); |
fba143c6 | 470 | spin_unlock_bh(&net->nsid_lock); |
6d458f5b | 471 | |
72ad937a | 472 | } |
6a1a3b9f PE |
473 | rtnl_unlock(); |
474 | ||
11a28d37 JB |
475 | /* |
476 | * Another CPU might be rcu-iterating the list, wait for it. | |
477 | * This needs to be before calling the exit() notifiers, so | |
478 | * the rcu_barrier() below isn't sufficient alone. | |
479 | */ | |
480 | synchronize_rcu(); | |
481 | ||
6a1a3b9f | 482 | /* Run all of the network namespace exit methods */ |
72ad937a EB |
483 | list_for_each_entry_reverse(ops, &pernet_list, list) |
484 | ops_exit_list(ops, &net_exit_list); | |
485 | ||
f875bae0 | 486 | /* Free the net generic variables */ |
72ad937a EB |
487 | list_for_each_entry_reverse(ops, &pernet_list, list) |
488 | ops_free_list(ops, &net_exit_list); | |
6a1a3b9f PE |
489 | |
490 | mutex_unlock(&net_mutex); | |
491 | ||
492 | /* Ensure there are no outstanding rcu callbacks using this | |
493 | * network namespace. | |
494 | */ | |
495 | rcu_barrier(); | |
496 | ||
497 | /* Finally it is safe to free my network namespace structure */ | |
72ad937a EB |
498 | list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) { |
499 | list_del_init(&net->exit_list); | |
70328660 | 500 | dec_net_namespaces(net->ucounts); |
038e7332 | 501 | put_user_ns(net->user_ns); |
a685e089 | 502 | net_drop_ns(net); |
2b035b39 | 503 | } |
6a1a3b9f | 504 | } |
7866cc57 FW |
505 | |
506 | /** | |
507 | * net_ns_barrier - wait until concurrent net_cleanup_work is done | |
508 | * | |
509 | * cleanup_net runs from work queue and will first remove namespaces | |
510 | * from the global list, then run net exit functions. | |
511 | * | |
512 | * Call this in module exit path to make sure that all netns | |
513 | * ->exit ops have been invoked before the function is removed. | |
514 | */ | |
515 | void net_ns_barrier(void) | |
516 | { | |
517 | mutex_lock(&net_mutex); | |
518 | mutex_unlock(&net_mutex); | |
519 | } | |
520 | EXPORT_SYMBOL(net_ns_barrier); | |
521 | ||
2b035b39 | 522 | static DECLARE_WORK(net_cleanup_work, cleanup_net); |
6a1a3b9f PE |
523 | |
524 | void __put_net(struct net *net) | |
525 | { | |
526 | /* Cleanup the network namespace in process context */ | |
2b035b39 EB |
527 | unsigned long flags; |
528 | ||
529 | spin_lock_irqsave(&cleanup_list_lock, flags); | |
530 | list_add(&net->cleanup_list, &cleanup_list); | |
531 | spin_unlock_irqrestore(&cleanup_list_lock, flags); | |
532 | ||
533 | queue_work(netns_wq, &net_cleanup_work); | |
6a1a3b9f PE |
534 | } |
535 | EXPORT_SYMBOL_GPL(__put_net); | |
536 | ||
956c9207 SR |
537 | struct net *get_net_ns_by_fd(int fd) |
538 | { | |
956c9207 | 539 | struct file *file; |
33c42940 | 540 | struct ns_common *ns; |
956c9207 SR |
541 | struct net *net; |
542 | ||
956c9207 | 543 | file = proc_ns_fget(fd); |
c316e6a3 AV |
544 | if (IS_ERR(file)) |
545 | return ERR_CAST(file); | |
956c9207 | 546 | |
f77c8014 | 547 | ns = get_proc_ns(file_inode(file)); |
33c42940 AV |
548 | if (ns->ops == &netns_operations) |
549 | net = get_net(container_of(ns, struct net, ns)); | |
c316e6a3 AV |
550 | else |
551 | net = ERR_PTR(-EINVAL); | |
956c9207 | 552 | |
c316e6a3 | 553 | fput(file); |
956c9207 SR |
554 | return net; |
555 | } | |
556 | ||
6a1a3b9f | 557 | #else |
956c9207 SR |
558 | struct net *get_net_ns_by_fd(int fd) |
559 | { | |
560 | return ERR_PTR(-EINVAL); | |
561 | } | |
6a1a3b9f | 562 | #endif |
4b681c82 | 563 | EXPORT_SYMBOL_GPL(get_net_ns_by_fd); |
6a1a3b9f | 564 | |
30ffee84 JB |
565 | struct net *get_net_ns_by_pid(pid_t pid) |
566 | { | |
567 | struct task_struct *tsk; | |
568 | struct net *net; | |
569 | ||
570 | /* Lookup the network namespace */ | |
571 | net = ERR_PTR(-ESRCH); | |
572 | rcu_read_lock(); | |
573 | tsk = find_task_by_vpid(pid); | |
574 | if (tsk) { | |
575 | struct nsproxy *nsproxy; | |
728dba3a EB |
576 | task_lock(tsk); |
577 | nsproxy = tsk->nsproxy; | |
30ffee84 JB |
578 | if (nsproxy) |
579 | net = get_net(nsproxy->net_ns); | |
728dba3a | 580 | task_unlock(tsk); |
30ffee84 JB |
581 | } |
582 | rcu_read_unlock(); | |
583 | return net; | |
584 | } | |
585 | EXPORT_SYMBOL_GPL(get_net_ns_by_pid); | |
586 | ||
98f842e6 EB |
587 | static __net_init int net_ns_net_init(struct net *net) |
588 | { | |
33c42940 AV |
589 | #ifdef CONFIG_NET_NS |
590 | net->ns.ops = &netns_operations; | |
591 | #endif | |
6344c433 | 592 | return ns_alloc_inum(&net->ns); |
98f842e6 EB |
593 | } |
594 | ||
595 | static __net_exit void net_ns_net_exit(struct net *net) | |
596 | { | |
6344c433 | 597 | ns_free_inum(&net->ns); |
98f842e6 EB |
598 | } |
599 | ||
600 | static struct pernet_operations __net_initdata net_ns_ops = { | |
601 | .init = net_ns_net_init, | |
602 | .exit = net_ns_net_exit, | |
603 | }; | |
604 | ||
3ee5256d | 605 | static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = { |
0c7aecd4 ND |
606 | [NETNSA_NONE] = { .type = NLA_UNSPEC }, |
607 | [NETNSA_NSID] = { .type = NLA_S32 }, | |
608 | [NETNSA_PID] = { .type = NLA_U32 }, | |
609 | [NETNSA_FD] = { .type = NLA_U32 }, | |
610 | }; | |
611 | ||
c21ef3e3 DA |
612 | static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh, |
613 | struct netlink_ext_ack *extack) | |
0c7aecd4 ND |
614 | { |
615 | struct net *net = sock_net(skb->sk); | |
616 | struct nlattr *tb[NETNSA_MAX + 1]; | |
4a7f7bc6 | 617 | struct nlattr *nla; |
0c7aecd4 ND |
618 | struct net *peer; |
619 | int nsid, err; | |
620 | ||
621 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
c21ef3e3 | 622 | rtnl_net_policy, extack); |
0c7aecd4 ND |
623 | if (err < 0) |
624 | return err; | |
4a7f7bc6 ND |
625 | if (!tb[NETNSA_NSID]) { |
626 | NL_SET_ERR_MSG(extack, "nsid is missing"); | |
0c7aecd4 | 627 | return -EINVAL; |
4a7f7bc6 | 628 | } |
0c7aecd4 ND |
629 | nsid = nla_get_s32(tb[NETNSA_NSID]); |
630 | ||
4a7f7bc6 | 631 | if (tb[NETNSA_PID]) { |
0c7aecd4 | 632 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); |
4a7f7bc6 ND |
633 | nla = tb[NETNSA_PID]; |
634 | } else if (tb[NETNSA_FD]) { | |
0c7aecd4 | 635 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); |
4a7f7bc6 ND |
636 | nla = tb[NETNSA_FD]; |
637 | } else { | |
638 | NL_SET_ERR_MSG(extack, "Peer netns reference is missing"); | |
0c7aecd4 | 639 | return -EINVAL; |
4a7f7bc6 ND |
640 | } |
641 | if (IS_ERR(peer)) { | |
642 | NL_SET_BAD_ATTR(extack, nla); | |
643 | NL_SET_ERR_MSG(extack, "Peer netns reference is invalid"); | |
0c7aecd4 | 644 | return PTR_ERR(peer); |
4a7f7bc6 | 645 | } |
0c7aecd4 | 646 | |
fba143c6 | 647 | spin_lock_bh(&net->nsid_lock); |
3138dbf8 | 648 | if (__peernet2id(net, peer) >= 0) { |
fba143c6 | 649 | spin_unlock_bh(&net->nsid_lock); |
0c7aecd4 | 650 | err = -EEXIST; |
4a7f7bc6 ND |
651 | NL_SET_BAD_ATTR(extack, nla); |
652 | NL_SET_ERR_MSG(extack, | |
653 | "Peer netns already has a nsid assigned"); | |
0c7aecd4 ND |
654 | goto out; |
655 | } | |
656 | ||
657 | err = alloc_netid(net, peer, nsid); | |
fba143c6 | 658 | spin_unlock_bh(&net->nsid_lock); |
3138dbf8 ND |
659 | if (err >= 0) { |
660 | rtnl_net_notifyid(net, RTM_NEWNSID, err); | |
0c7aecd4 | 661 | err = 0; |
4a7f7bc6 | 662 | } else if (err == -ENOSPC && nsid >= 0) { |
10d486a3 | 663 | err = -EEXIST; |
4a7f7bc6 ND |
664 | NL_SET_BAD_ATTR(extack, tb[NETNSA_NSID]); |
665 | NL_SET_ERR_MSG(extack, "The specified nsid is already used"); | |
3138dbf8 | 666 | } |
0c7aecd4 ND |
667 | out: |
668 | put_net(peer); | |
669 | return err; | |
670 | } | |
671 | ||
672 | static int rtnl_net_get_size(void) | |
673 | { | |
674 | return NLMSG_ALIGN(sizeof(struct rtgenmsg)) | |
675 | + nla_total_size(sizeof(s32)) /* NETNSA_NSID */ | |
676 | ; | |
677 | } | |
678 | ||
679 | static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags, | |
cab3c8ec | 680 | int cmd, struct net *net, int nsid) |
0c7aecd4 ND |
681 | { |
682 | struct nlmsghdr *nlh; | |
683 | struct rtgenmsg *rth; | |
0c7aecd4 ND |
684 | |
685 | nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags); | |
686 | if (!nlh) | |
687 | return -EMSGSIZE; | |
688 | ||
689 | rth = nlmsg_data(nlh); | |
690 | rth->rtgen_family = AF_UNSPEC; | |
691 | ||
cab3c8ec | 692 | if (nla_put_s32(skb, NETNSA_NSID, nsid)) |
0c7aecd4 ND |
693 | goto nla_put_failure; |
694 | ||
695 | nlmsg_end(skb, nlh); | |
696 | return 0; | |
697 | ||
698 | nla_put_failure: | |
699 | nlmsg_cancel(skb, nlh); | |
700 | return -EMSGSIZE; | |
701 | } | |
702 | ||
c21ef3e3 DA |
703 | static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh, |
704 | struct netlink_ext_ack *extack) | |
0c7aecd4 ND |
705 | { |
706 | struct net *net = sock_net(skb->sk); | |
707 | struct nlattr *tb[NETNSA_MAX + 1]; | |
4a7f7bc6 | 708 | struct nlattr *nla; |
0c7aecd4 | 709 | struct sk_buff *msg; |
0c7aecd4 | 710 | struct net *peer; |
cab3c8ec | 711 | int err, id; |
0c7aecd4 ND |
712 | |
713 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
c21ef3e3 | 714 | rtnl_net_policy, extack); |
0c7aecd4 ND |
715 | if (err < 0) |
716 | return err; | |
4a7f7bc6 | 717 | if (tb[NETNSA_PID]) { |
0c7aecd4 | 718 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); |
4a7f7bc6 ND |
719 | nla = tb[NETNSA_PID]; |
720 | } else if (tb[NETNSA_FD]) { | |
0c7aecd4 | 721 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); |
4a7f7bc6 ND |
722 | nla = tb[NETNSA_FD]; |
723 | } else { | |
724 | NL_SET_ERR_MSG(extack, "Peer netns reference is missing"); | |
0c7aecd4 | 725 | return -EINVAL; |
4a7f7bc6 | 726 | } |
0c7aecd4 | 727 | |
4a7f7bc6 ND |
728 | if (IS_ERR(peer)) { |
729 | NL_SET_BAD_ATTR(extack, nla); | |
730 | NL_SET_ERR_MSG(extack, "Peer netns reference is invalid"); | |
0c7aecd4 | 731 | return PTR_ERR(peer); |
4a7f7bc6 | 732 | } |
0c7aecd4 ND |
733 | |
734 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
735 | if (!msg) { | |
736 | err = -ENOMEM; | |
737 | goto out; | |
738 | } | |
739 | ||
95f38411 | 740 | id = peernet2id(net, peer); |
0c7aecd4 | 741 | err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, |
b04096ff | 742 | RTM_NEWNSID, net, id); |
0c7aecd4 ND |
743 | if (err < 0) |
744 | goto err_out; | |
745 | ||
746 | err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid); | |
747 | goto out; | |
748 | ||
749 | err_out: | |
750 | nlmsg_free(msg); | |
751 | out: | |
752 | put_net(peer); | |
753 | return err; | |
754 | } | |
755 | ||
a143c40c ND |
756 | struct rtnl_net_dump_cb { |
757 | struct net *net; | |
758 | struct sk_buff *skb; | |
759 | struct netlink_callback *cb; | |
760 | int idx; | |
761 | int s_idx; | |
762 | }; | |
763 | ||
764 | static int rtnl_net_dumpid_one(int id, void *peer, void *data) | |
765 | { | |
766 | struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data; | |
767 | int ret; | |
768 | ||
769 | if (net_cb->idx < net_cb->s_idx) | |
770 | goto cont; | |
771 | ||
772 | ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid, | |
773 | net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI, | |
cab3c8ec | 774 | RTM_NEWNSID, net_cb->net, id); |
a143c40c ND |
775 | if (ret < 0) |
776 | return ret; | |
777 | ||
778 | cont: | |
779 | net_cb->idx++; | |
780 | return 0; | |
781 | } | |
782 | ||
783 | static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb) | |
784 | { | |
785 | struct net *net = sock_net(skb->sk); | |
786 | struct rtnl_net_dump_cb net_cb = { | |
787 | .net = net, | |
788 | .skb = skb, | |
789 | .cb = cb, | |
790 | .idx = 0, | |
791 | .s_idx = cb->args[0], | |
792 | }; | |
793 | ||
fba143c6 | 794 | spin_lock_bh(&net->nsid_lock); |
a143c40c | 795 | idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb); |
fba143c6 | 796 | spin_unlock_bh(&net->nsid_lock); |
a143c40c ND |
797 | |
798 | cb->args[0] = net_cb.idx; | |
799 | return skb->len; | |
800 | } | |
801 | ||
cab3c8ec | 802 | static void rtnl_net_notifyid(struct net *net, int cmd, int id) |
9a963454 ND |
803 | { |
804 | struct sk_buff *msg; | |
805 | int err = -ENOMEM; | |
806 | ||
807 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
808 | if (!msg) | |
809 | goto out; | |
810 | ||
cab3c8ec | 811 | err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, id); |
9a963454 ND |
812 | if (err < 0) |
813 | goto err_out; | |
814 | ||
815 | rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0); | |
816 | return; | |
817 | ||
818 | err_out: | |
819 | nlmsg_free(msg); | |
820 | out: | |
821 | rtnl_set_sk_err(net, RTNLGRP_NSID, err); | |
822 | } | |
823 | ||
5f256bec EB |
824 | static int __init net_ns_init(void) |
825 | { | |
486a87f1 | 826 | struct net_generic *ng; |
5f256bec | 827 | |
d57a9212 | 828 | #ifdef CONFIG_NET_NS |
5f256bec EB |
829 | net_cachep = kmem_cache_create("net_namespace", sizeof(struct net), |
830 | SMP_CACHE_BYTES, | |
831 | SLAB_PANIC, NULL); | |
3ef1355d BT |
832 | |
833 | /* Create workqueue for cleanup */ | |
834 | netns_wq = create_singlethread_workqueue("netns"); | |
835 | if (!netns_wq) | |
836 | panic("Could not create netns workq"); | |
d57a9212 | 837 | #endif |
3ef1355d | 838 | |
486a87f1 DL |
839 | ng = net_alloc_generic(); |
840 | if (!ng) | |
841 | panic("Could not allocate generic netns"); | |
842 | ||
843 | rcu_assign_pointer(init_net.gen, ng); | |
844 | ||
5f256bec | 845 | mutex_lock(&net_mutex); |
038e7332 | 846 | if (setup_net(&init_net, &init_user_ns)) |
ca0f3112 | 847 | panic("Could not setup the initial network namespace"); |
5f256bec | 848 | |
f8c46cb3 DT |
849 | init_net_initialized = true; |
850 | ||
f4618d39 | 851 | rtnl_lock(); |
11a28d37 | 852 | list_add_tail_rcu(&init_net.list, &net_namespace_list); |
f4618d39 | 853 | rtnl_unlock(); |
5f256bec EB |
854 | |
855 | mutex_unlock(&net_mutex); | |
5f256bec | 856 | |
98f842e6 EB |
857 | register_pernet_subsys(&net_ns_ops); |
858 | ||
165b9117 FW |
859 | rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, |
860 | RTNL_FLAG_DOIT_UNLOCKED); | |
a143c40c | 861 | rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid, |
165b9117 | 862 | RTNL_FLAG_DOIT_UNLOCKED); |
0c7aecd4 | 863 | |
5f256bec EB |
864 | return 0; |
865 | } | |
866 | ||
867 | pure_initcall(net_ns_init); | |
868 | ||
ed160e83 | 869 | #ifdef CONFIG_NET_NS |
f875bae0 EB |
870 | static int __register_pernet_operations(struct list_head *list, |
871 | struct pernet_operations *ops) | |
5f256bec | 872 | { |
72ad937a | 873 | struct net *net; |
5f256bec | 874 | int error; |
72ad937a | 875 | LIST_HEAD(net_exit_list); |
5f256bec | 876 | |
5f256bec | 877 | list_add_tail(&ops->list, list); |
f875bae0 | 878 | if (ops->init || (ops->id && ops->size)) { |
1dba323b | 879 | for_each_net(net) { |
f875bae0 | 880 | error = ops_init(ops, net); |
5f256bec EB |
881 | if (error) |
882 | goto out_undo; | |
72ad937a | 883 | list_add_tail(&net->exit_list, &net_exit_list); |
5f256bec EB |
884 | } |
885 | } | |
1dba323b | 886 | return 0; |
5f256bec EB |
887 | |
888 | out_undo: | |
889 | /* If I have an error cleanup all namespaces I initialized */ | |
890 | list_del(&ops->list); | |
72ad937a EB |
891 | ops_exit_list(ops, &net_exit_list); |
892 | ops_free_list(ops, &net_exit_list); | |
1dba323b | 893 | return error; |
5f256bec EB |
894 | } |
895 | ||
f875bae0 | 896 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
5f256bec EB |
897 | { |
898 | struct net *net; | |
72ad937a | 899 | LIST_HEAD(net_exit_list); |
5f256bec EB |
900 | |
901 | list_del(&ops->list); | |
72ad937a EB |
902 | for_each_net(net) |
903 | list_add_tail(&net->exit_list, &net_exit_list); | |
904 | ops_exit_list(ops, &net_exit_list); | |
905 | ops_free_list(ops, &net_exit_list); | |
5f256bec EB |
906 | } |
907 | ||
ed160e83 DL |
908 | #else |
909 | ||
f875bae0 EB |
910 | static int __register_pernet_operations(struct list_head *list, |
911 | struct pernet_operations *ops) | |
ed160e83 | 912 | { |
f8c46cb3 DT |
913 | if (!init_net_initialized) { |
914 | list_add_tail(&ops->list, list); | |
915 | return 0; | |
916 | } | |
917 | ||
b922934d | 918 | return ops_init(ops, &init_net); |
ed160e83 DL |
919 | } |
920 | ||
f875bae0 | 921 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
ed160e83 | 922 | { |
f8c46cb3 DT |
923 | if (!init_net_initialized) { |
924 | list_del(&ops->list); | |
925 | } else { | |
926 | LIST_HEAD(net_exit_list); | |
927 | list_add(&init_net.exit_list, &net_exit_list); | |
928 | ops_exit_list(ops, &net_exit_list); | |
929 | ops_free_list(ops, &net_exit_list); | |
930 | } | |
ed160e83 | 931 | } |
f875bae0 EB |
932 | |
933 | #endif /* CONFIG_NET_NS */ | |
ed160e83 | 934 | |
c93cf61f PE |
935 | static DEFINE_IDA(net_generic_ids); |
936 | ||
f875bae0 EB |
937 | static int register_pernet_operations(struct list_head *list, |
938 | struct pernet_operations *ops) | |
939 | { | |
940 | int error; | |
941 | ||
942 | if (ops->id) { | |
943 | again: | |
6af2d5ff | 944 | error = ida_get_new_above(&net_generic_ids, MIN_PERNET_OPS_ID, ops->id); |
f875bae0 EB |
945 | if (error < 0) { |
946 | if (error == -EAGAIN) { | |
947 | ida_pre_get(&net_generic_ids, GFP_KERNEL); | |
948 | goto again; | |
949 | } | |
950 | return error; | |
951 | } | |
6af2d5ff | 952 | max_gen_ptrs = max(max_gen_ptrs, *ops->id + 1); |
f875bae0 EB |
953 | } |
954 | error = __register_pernet_operations(list, ops); | |
3a765eda EB |
955 | if (error) { |
956 | rcu_barrier(); | |
957 | if (ops->id) | |
958 | ida_remove(&net_generic_ids, *ops->id); | |
959 | } | |
f875bae0 EB |
960 | |
961 | return error; | |
962 | } | |
963 | ||
964 | static void unregister_pernet_operations(struct pernet_operations *ops) | |
965 | { | |
966 | ||
967 | __unregister_pernet_operations(ops); | |
3a765eda | 968 | rcu_barrier(); |
f875bae0 EB |
969 | if (ops->id) |
970 | ida_remove(&net_generic_ids, *ops->id); | |
971 | } | |
972 | ||
5f256bec EB |
973 | /** |
974 | * register_pernet_subsys - register a network namespace subsystem | |
975 | * @ops: pernet operations structure for the subsystem | |
976 | * | |
977 | * Register a subsystem which has init and exit functions | |
978 | * that are called when network namespaces are created and | |
979 | * destroyed respectively. | |
980 | * | |
981 | * When registered all network namespace init functions are | |
982 | * called for every existing network namespace. Allowing kernel | |
983 | * modules to have a race free view of the set of network namespaces. | |
984 | * | |
985 | * When a new network namespace is created all of the init | |
986 | * methods are called in the order in which they were registered. | |
987 | * | |
988 | * When a network namespace is destroyed all of the exit methods | |
989 | * are called in the reverse of the order with which they were | |
990 | * registered. | |
991 | */ | |
992 | int register_pernet_subsys(struct pernet_operations *ops) | |
993 | { | |
994 | int error; | |
995 | mutex_lock(&net_mutex); | |
996 | error = register_pernet_operations(first_device, ops); | |
997 | mutex_unlock(&net_mutex); | |
998 | return error; | |
999 | } | |
1000 | EXPORT_SYMBOL_GPL(register_pernet_subsys); | |
1001 | ||
1002 | /** | |
1003 | * unregister_pernet_subsys - unregister a network namespace subsystem | |
1004 | * @ops: pernet operations structure to manipulate | |
1005 | * | |
1006 | * Remove the pernet operations structure from the list to be | |
53379e57 | 1007 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
1008 | * addition run the exit method for all existing network |
1009 | * namespaces. | |
1010 | */ | |
b3c981d2 | 1011 | void unregister_pernet_subsys(struct pernet_operations *ops) |
5f256bec EB |
1012 | { |
1013 | mutex_lock(&net_mutex); | |
b3c981d2 | 1014 | unregister_pernet_operations(ops); |
5f256bec EB |
1015 | mutex_unlock(&net_mutex); |
1016 | } | |
1017 | EXPORT_SYMBOL_GPL(unregister_pernet_subsys); | |
1018 | ||
1019 | /** | |
1020 | * register_pernet_device - register a network namespace device | |
1021 | * @ops: pernet operations structure for the subsystem | |
1022 | * | |
1023 | * Register a device which has init and exit functions | |
1024 | * that are called when network namespaces are created and | |
1025 | * destroyed respectively. | |
1026 | * | |
1027 | * When registered all network namespace init functions are | |
1028 | * called for every existing network namespace. Allowing kernel | |
1029 | * modules to have a race free view of the set of network namespaces. | |
1030 | * | |
1031 | * When a new network namespace is created all of the init | |
1032 | * methods are called in the order in which they were registered. | |
1033 | * | |
1034 | * When a network namespace is destroyed all of the exit methods | |
1035 | * are called in the reverse of the order with which they were | |
1036 | * registered. | |
1037 | */ | |
1038 | int register_pernet_device(struct pernet_operations *ops) | |
1039 | { | |
1040 | int error; | |
1041 | mutex_lock(&net_mutex); | |
1042 | error = register_pernet_operations(&pernet_list, ops); | |
1043 | if (!error && (first_device == &pernet_list)) | |
1044 | first_device = &ops->list; | |
1045 | mutex_unlock(&net_mutex); | |
1046 | return error; | |
1047 | } | |
1048 | EXPORT_SYMBOL_GPL(register_pernet_device); | |
1049 | ||
1050 | /** | |
1051 | * unregister_pernet_device - unregister a network namespace netdevice | |
1052 | * @ops: pernet operations structure to manipulate | |
1053 | * | |
1054 | * Remove the pernet operations structure from the list to be | |
53379e57 | 1055 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
1056 | * addition run the exit method for all existing network |
1057 | * namespaces. | |
1058 | */ | |
1059 | void unregister_pernet_device(struct pernet_operations *ops) | |
1060 | { | |
1061 | mutex_lock(&net_mutex); | |
1062 | if (&ops->list == first_device) | |
1063 | first_device = first_device->next; | |
1064 | unregister_pernet_operations(ops); | |
1065 | mutex_unlock(&net_mutex); | |
1066 | } | |
1067 | EXPORT_SYMBOL_GPL(unregister_pernet_device); | |
13b6f576 EB |
1068 | |
1069 | #ifdef CONFIG_NET_NS | |
64964528 | 1070 | static struct ns_common *netns_get(struct task_struct *task) |
13b6f576 | 1071 | { |
f0630529 EB |
1072 | struct net *net = NULL; |
1073 | struct nsproxy *nsproxy; | |
1074 | ||
728dba3a EB |
1075 | task_lock(task); |
1076 | nsproxy = task->nsproxy; | |
f0630529 EB |
1077 | if (nsproxy) |
1078 | net = get_net(nsproxy->net_ns); | |
728dba3a | 1079 | task_unlock(task); |
f0630529 | 1080 | |
ff24870f AV |
1081 | return net ? &net->ns : NULL; |
1082 | } | |
1083 | ||
1084 | static inline struct net *to_net_ns(struct ns_common *ns) | |
1085 | { | |
1086 | return container_of(ns, struct net, ns); | |
13b6f576 EB |
1087 | } |
1088 | ||
64964528 | 1089 | static void netns_put(struct ns_common *ns) |
13b6f576 | 1090 | { |
ff24870f | 1091 | put_net(to_net_ns(ns)); |
13b6f576 EB |
1092 | } |
1093 | ||
64964528 | 1094 | static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns) |
13b6f576 | 1095 | { |
ff24870f | 1096 | struct net *net = to_net_ns(ns); |
142e1d1d | 1097 | |
5e4a0847 | 1098 | if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) || |
c7b96acf | 1099 | !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) |
142e1d1d EB |
1100 | return -EPERM; |
1101 | ||
13b6f576 | 1102 | put_net(nsproxy->net_ns); |
142e1d1d | 1103 | nsproxy->net_ns = get_net(net); |
13b6f576 EB |
1104 | return 0; |
1105 | } | |
1106 | ||
bcac25a5 AV |
1107 | static struct user_namespace *netns_owner(struct ns_common *ns) |
1108 | { | |
1109 | return to_net_ns(ns)->user_ns; | |
1110 | } | |
1111 | ||
13b6f576 EB |
1112 | const struct proc_ns_operations netns_operations = { |
1113 | .name = "net", | |
1114 | .type = CLONE_NEWNET, | |
1115 | .get = netns_get, | |
1116 | .put = netns_put, | |
1117 | .install = netns_install, | |
bcac25a5 | 1118 | .owner = netns_owner, |
13b6f576 EB |
1119 | }; |
1120 | #endif |