]>
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> |
0c7aecd4 ND |
19 | #include <net/sock.h> |
20 | #include <net/netlink.h> | |
5f256bec | 21 | #include <net/net_namespace.h> |
dec827d1 | 22 | #include <net/netns/generic.h> |
5f256bec EB |
23 | |
24 | /* | |
25 | * Our network namespace constructor/destructor lists | |
26 | */ | |
27 | ||
28 | static LIST_HEAD(pernet_list); | |
29 | static struct list_head *first_device = &pernet_list; | |
200b916f | 30 | DEFINE_MUTEX(net_mutex); |
5f256bec | 31 | |
5f256bec | 32 | LIST_HEAD(net_namespace_list); |
b76a461f | 33 | EXPORT_SYMBOL_GPL(net_namespace_list); |
5f256bec | 34 | |
734b6541 RM |
35 | struct net init_net = { |
36 | .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head), | |
37 | }; | |
ff4b9502 | 38 | EXPORT_SYMBOL(init_net); |
5f256bec | 39 | |
f8c46cb3 DT |
40 | static bool init_net_initialized; |
41 | ||
dec827d1 PE |
42 | #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */ |
43 | ||
073862ba ED |
44 | static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS; |
45 | ||
46 | static struct net_generic *net_alloc_generic(void) | |
47 | { | |
48 | struct net_generic *ng; | |
49 | size_t generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]); | |
50 | ||
51 | ng = kzalloc(generic_size, GFP_KERNEL); | |
52 | if (ng) | |
53 | ng->len = max_gen_ptrs; | |
54 | ||
55 | return ng; | |
56 | } | |
57 | ||
05fceb4a JP |
58 | static int net_assign_generic(struct net *net, int id, void *data) |
59 | { | |
60 | struct net_generic *ng, *old_ng; | |
61 | ||
62 | BUG_ON(!mutex_is_locked(&net_mutex)); | |
63 | BUG_ON(id == 0); | |
64 | ||
1c87733d ED |
65 | old_ng = rcu_dereference_protected(net->gen, |
66 | lockdep_is_held(&net_mutex)); | |
67 | ng = old_ng; | |
05fceb4a JP |
68 | if (old_ng->len >= id) |
69 | goto assign; | |
70 | ||
073862ba | 71 | ng = net_alloc_generic(); |
05fceb4a JP |
72 | if (ng == NULL) |
73 | return -ENOMEM; | |
74 | ||
75 | /* | |
76 | * Some synchronisation notes: | |
77 | * | |
78 | * The net_generic explores the net->gen array inside rcu | |
79 | * read section. Besides once set the net->gen->ptr[x] | |
80 | * pointer never changes (see rules in netns/generic.h). | |
81 | * | |
82 | * That said, we simply duplicate this array and schedule | |
83 | * the old copy for kfree after a grace period. | |
84 | */ | |
85 | ||
05fceb4a JP |
86 | memcpy(&ng->ptr, &old_ng->ptr, old_ng->len * sizeof(void*)); |
87 | ||
88 | rcu_assign_pointer(net->gen, ng); | |
04d4dfed | 89 | kfree_rcu(old_ng, rcu); |
05fceb4a JP |
90 | assign: |
91 | ng->ptr[id - 1] = data; | |
92 | return 0; | |
93 | } | |
94 | ||
f875bae0 EB |
95 | static int ops_init(const struct pernet_operations *ops, struct net *net) |
96 | { | |
b922934d JA |
97 | int err = -ENOMEM; |
98 | void *data = NULL; | |
99 | ||
f875bae0 | 100 | if (ops->id && ops->size) { |
b922934d | 101 | data = kzalloc(ops->size, GFP_KERNEL); |
f875bae0 | 102 | if (!data) |
b922934d | 103 | goto out; |
f875bae0 EB |
104 | |
105 | err = net_assign_generic(net, *ops->id, data); | |
b922934d JA |
106 | if (err) |
107 | goto cleanup; | |
f875bae0 | 108 | } |
b922934d | 109 | err = 0; |
f875bae0 | 110 | if (ops->init) |
b922934d JA |
111 | err = ops->init(net); |
112 | if (!err) | |
113 | return 0; | |
114 | ||
115 | cleanup: | |
116 | kfree(data); | |
117 | ||
118 | out: | |
119 | return err; | |
f875bae0 EB |
120 | } |
121 | ||
122 | static void ops_free(const struct pernet_operations *ops, struct net *net) | |
123 | { | |
124 | if (ops->id && ops->size) { | |
125 | int id = *ops->id; | |
126 | kfree(net_generic(net, id)); | |
127 | } | |
128 | } | |
129 | ||
72ad937a EB |
130 | static void ops_exit_list(const struct pernet_operations *ops, |
131 | struct list_head *net_exit_list) | |
132 | { | |
133 | struct net *net; | |
134 | if (ops->exit) { | |
135 | list_for_each_entry(net, net_exit_list, exit_list) | |
136 | ops->exit(net); | |
137 | } | |
72ad937a EB |
138 | if (ops->exit_batch) |
139 | ops->exit_batch(net_exit_list); | |
140 | } | |
141 | ||
142 | static void ops_free_list(const struct pernet_operations *ops, | |
143 | struct list_head *net_exit_list) | |
144 | { | |
145 | struct net *net; | |
146 | if (ops->size && ops->id) { | |
147 | list_for_each_entry(net, net_exit_list, exit_list) | |
148 | ops_free(ops, net); | |
149 | } | |
150 | } | |
151 | ||
95f38411 | 152 | /* should be called with nsid_lock held */ |
0c7aecd4 ND |
153 | static int alloc_netid(struct net *net, struct net *peer, int reqid) |
154 | { | |
3138dbf8 | 155 | int min = 0, max = 0; |
0c7aecd4 | 156 | |
0c7aecd4 ND |
157 | if (reqid >= 0) { |
158 | min = reqid; | |
159 | max = reqid + 1; | |
160 | } | |
161 | ||
95f38411 | 162 | return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC); |
0c7aecd4 ND |
163 | } |
164 | ||
165 | /* This function is used by idr_for_each(). If net is equal to peer, the | |
166 | * function returns the id so that idr_for_each() stops. Because we cannot | |
167 | * returns the id 0 (idr_for_each() will not stop), we return the magic value | |
168 | * NET_ID_ZERO (-1) for it. | |
169 | */ | |
170 | #define NET_ID_ZERO -1 | |
171 | static int net_eq_idr(int id, void *net, void *peer) | |
172 | { | |
173 | if (net_eq(net, peer)) | |
174 | return id ? : NET_ID_ZERO; | |
175 | return 0; | |
176 | } | |
177 | ||
95f38411 ND |
178 | /* Should be called with nsid_lock held. If a new id is assigned, the bool alloc |
179 | * is set to true, thus the caller knows that the new id must be notified via | |
180 | * rtnl. | |
181 | */ | |
3138dbf8 | 182 | static int __peernet2id_alloc(struct net *net, struct net *peer, bool *alloc) |
0c7aecd4 ND |
183 | { |
184 | int id = idr_for_each(&net->netns_ids, net_eq_idr, peer); | |
3138dbf8 | 185 | bool alloc_it = *alloc; |
0c7aecd4 | 186 | |
3138dbf8 ND |
187 | *alloc = false; |
188 | ||
0c7aecd4 ND |
189 | /* Magic value for id 0. */ |
190 | if (id == NET_ID_ZERO) | |
191 | return 0; | |
192 | if (id > 0) | |
193 | return id; | |
194 | ||
3138dbf8 | 195 | if (alloc_it) { |
109582af | 196 | id = alloc_netid(net, peer, -1); |
3138dbf8 | 197 | *alloc = true; |
109582af ND |
198 | return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED; |
199 | } | |
0c7aecd4 | 200 | |
109582af | 201 | return NETNSA_NSID_NOT_ASSIGNED; |
0c7aecd4 ND |
202 | } |
203 | ||
95f38411 | 204 | /* should be called with nsid_lock held */ |
3138dbf8 ND |
205 | static int __peernet2id(struct net *net, struct net *peer) |
206 | { | |
207 | bool no = false; | |
208 | ||
209 | return __peernet2id_alloc(net, peer, &no); | |
210 | } | |
211 | ||
212 | static void rtnl_net_notifyid(struct net *net, int cmd, int id); | |
0c7aecd4 ND |
213 | /* This function returns the id of a peer netns. If no id is assigned, one will |
214 | * be allocated and returned. | |
215 | */ | |
7a0877d4 | 216 | int peernet2id_alloc(struct net *net, struct net *peer) |
0c7aecd4 | 217 | { |
95f38411 ND |
218 | unsigned long flags; |
219 | bool alloc; | |
3138dbf8 | 220 | int id; |
0c7aecd4 | 221 | |
de133464 | 222 | spin_lock_irqsave(&net->nsid_lock, flags); |
95f38411 | 223 | alloc = atomic_read(&peer->count) == 0 ? false : true; |
3138dbf8 | 224 | id = __peernet2id_alloc(net, peer, &alloc); |
de133464 | 225 | spin_unlock_irqrestore(&net->nsid_lock, flags); |
3138dbf8 ND |
226 | if (alloc && id >= 0) |
227 | rtnl_net_notifyid(net, RTM_NEWNSID, id); | |
228 | return id; | |
0c7aecd4 | 229 | } |
7a0877d4 | 230 | EXPORT_SYMBOL(peernet2id_alloc); |
0c7aecd4 | 231 | |
95f38411 | 232 | /* This function returns, if assigned, the id of a peer netns. */ |
59324cf3 | 233 | int peernet2id(struct net *net, struct net *peer) |
95f38411 ND |
234 | { |
235 | unsigned long flags; | |
236 | int id; | |
237 | ||
de133464 | 238 | spin_lock_irqsave(&net->nsid_lock, flags); |
95f38411 | 239 | id = __peernet2id(net, peer); |
de133464 | 240 | spin_unlock_irqrestore(&net->nsid_lock, flags); |
95f38411 ND |
241 | return id; |
242 | } | |
243 | ||
59324cf3 ND |
244 | /* This function returns true is the peer netns has an id assigned into the |
245 | * current netns. | |
246 | */ | |
247 | bool peernet_has_id(struct net *net, struct net *peer) | |
248 | { | |
249 | return peernet2id(net, peer) >= 0; | |
250 | } | |
251 | ||
0c7aecd4 ND |
252 | struct net *get_net_ns_by_id(struct net *net, int id) |
253 | { | |
95f38411 | 254 | unsigned long flags; |
0c7aecd4 ND |
255 | struct net *peer; |
256 | ||
257 | if (id < 0) | |
258 | return NULL; | |
259 | ||
260 | rcu_read_lock(); | |
de133464 | 261 | spin_lock_irqsave(&net->nsid_lock, flags); |
0c7aecd4 ND |
262 | peer = idr_find(&net->netns_ids, id); |
263 | if (peer) | |
264 | get_net(peer); | |
de133464 | 265 | spin_unlock_irqrestore(&net->nsid_lock, flags); |
0c7aecd4 ND |
266 | rcu_read_unlock(); |
267 | ||
268 | return peer; | |
269 | } | |
270 | ||
5f256bec EB |
271 | /* |
272 | * setup_net runs the initializers for the network namespace object. | |
273 | */ | |
038e7332 | 274 | static __net_init int setup_net(struct net *net, struct user_namespace *user_ns) |
5f256bec EB |
275 | { |
276 | /* Must be called with net_mutex held */ | |
f875bae0 | 277 | const struct pernet_operations *ops, *saved_ops; |
486a87f1 | 278 | int error = 0; |
72ad937a | 279 | LIST_HEAD(net_exit_list); |
5f256bec | 280 | |
5f256bec | 281 | atomic_set(&net->count, 1); |
a685e089 | 282 | atomic_set(&net->passive, 1); |
4e985ada | 283 | net->dev_base_seq = 1; |
038e7332 | 284 | net->user_ns = user_ns; |
0c7aecd4 | 285 | idr_init(&net->netns_ids); |
de133464 | 286 | spin_lock_init(&net->nsid_lock); |
486a87f1 | 287 | |
768f3591 | 288 | list_for_each_entry(ops, &pernet_list, list) { |
f875bae0 EB |
289 | error = ops_init(ops, net); |
290 | if (error < 0) | |
291 | goto out_undo; | |
5f256bec EB |
292 | } |
293 | out: | |
294 | return error; | |
768f3591 | 295 | |
5f256bec EB |
296 | out_undo: |
297 | /* Walk through the list backwards calling the exit functions | |
298 | * for the pernet modules whose init functions did not fail. | |
299 | */ | |
72ad937a | 300 | list_add(&net->exit_list, &net_exit_list); |
f875bae0 | 301 | saved_ops = ops; |
72ad937a EB |
302 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) |
303 | ops_exit_list(ops, &net_exit_list); | |
304 | ||
f875bae0 EB |
305 | ops = saved_ops; |
306 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) | |
72ad937a | 307 | ops_free_list(ops, &net_exit_list); |
310928d9 DL |
308 | |
309 | rcu_barrier(); | |
5f256bec EB |
310 | goto out; |
311 | } | |
312 | ||
6a1a3b9f | 313 | |
ebe47d47 CN |
314 | #ifdef CONFIG_NET_NS |
315 | static struct kmem_cache *net_cachep; | |
316 | static struct workqueue_struct *netns_wq; | |
317 | ||
486a87f1 | 318 | static struct net *net_alloc(void) |
45a19b0a | 319 | { |
486a87f1 DL |
320 | struct net *net = NULL; |
321 | struct net_generic *ng; | |
322 | ||
323 | ng = net_alloc_generic(); | |
324 | if (!ng) | |
325 | goto out; | |
326 | ||
327 | net = kmem_cache_zalloc(net_cachep, GFP_KERNEL); | |
45a19b0a | 328 | if (!net) |
486a87f1 | 329 | goto out_free; |
45a19b0a | 330 | |
486a87f1 DL |
331 | rcu_assign_pointer(net->gen, ng); |
332 | out: | |
333 | return net; | |
334 | ||
335 | out_free: | |
336 | kfree(ng); | |
337 | goto out; | |
338 | } | |
339 | ||
340 | static void net_free(struct net *net) | |
341 | { | |
416c51e1 | 342 | kfree(rcu_access_pointer(net->gen)); |
45a19b0a JFS |
343 | kmem_cache_free(net_cachep, net); |
344 | } | |
345 | ||
a685e089 AV |
346 | void net_drop_ns(void *p) |
347 | { | |
348 | struct net *ns = p; | |
349 | if (ns && atomic_dec_and_test(&ns->passive)) | |
350 | net_free(ns); | |
351 | } | |
352 | ||
038e7332 EB |
353 | struct net *copy_net_ns(unsigned long flags, |
354 | struct user_namespace *user_ns, struct net *old_net) | |
9dd776b6 | 355 | { |
088eb2d9 AD |
356 | struct net *net; |
357 | int rv; | |
9dd776b6 | 358 | |
911cb193 RL |
359 | if (!(flags & CLONE_NEWNET)) |
360 | return get_net(old_net); | |
361 | ||
088eb2d9 AD |
362 | net = net_alloc(); |
363 | if (!net) | |
364 | return ERR_PTR(-ENOMEM); | |
038e7332 EB |
365 | |
366 | get_user_ns(user_ns); | |
367 | ||
9dd776b6 | 368 | mutex_lock(&net_mutex); |
038e7332 | 369 | rv = setup_net(net, user_ns); |
088eb2d9 | 370 | if (rv == 0) { |
486a87f1 | 371 | rtnl_lock(); |
11a28d37 | 372 | list_add_tail_rcu(&net->list, &net_namespace_list); |
486a87f1 DL |
373 | rtnl_unlock(); |
374 | } | |
9dd776b6 | 375 | mutex_unlock(&net_mutex); |
088eb2d9 | 376 | if (rv < 0) { |
038e7332 | 377 | put_user_ns(user_ns); |
a685e089 | 378 | net_drop_ns(net); |
088eb2d9 AD |
379 | return ERR_PTR(rv); |
380 | } | |
381 | return net; | |
382 | } | |
486a87f1 | 383 | |
2b035b39 EB |
384 | static DEFINE_SPINLOCK(cleanup_list_lock); |
385 | static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */ | |
386 | ||
6a1a3b9f PE |
387 | static void cleanup_net(struct work_struct *work) |
388 | { | |
f875bae0 | 389 | const struct pernet_operations *ops; |
6d458f5b | 390 | struct net *net, *tmp; |
1818ce4d | 391 | struct list_head net_kill_list; |
72ad937a | 392 | LIST_HEAD(net_exit_list); |
6a1a3b9f | 393 | |
2b035b39 EB |
394 | /* Atomically snapshot the list of namespaces to cleanup */ |
395 | spin_lock_irq(&cleanup_list_lock); | |
396 | list_replace_init(&cleanup_list, &net_kill_list); | |
397 | spin_unlock_irq(&cleanup_list_lock); | |
6a1a3b9f PE |
398 | |
399 | mutex_lock(&net_mutex); | |
400 | ||
401 | /* Don't let anyone else find us. */ | |
402 | rtnl_lock(); | |
72ad937a | 403 | list_for_each_entry(net, &net_kill_list, cleanup_list) { |
2b035b39 | 404 | list_del_rcu(&net->list); |
72ad937a | 405 | list_add_tail(&net->exit_list, &net_exit_list); |
6d458f5b | 406 | for_each_net(tmp) { |
95f38411 | 407 | int id; |
6d458f5b | 408 | |
de133464 | 409 | spin_lock_irq(&tmp->nsid_lock); |
95f38411 ND |
410 | id = __peernet2id(tmp, net); |
411 | if (id >= 0) | |
6d458f5b | 412 | idr_remove(&tmp->netns_ids, id); |
de133464 | 413 | spin_unlock_irq(&tmp->nsid_lock); |
95f38411 ND |
414 | if (id >= 0) |
415 | rtnl_net_notifyid(tmp, RTM_DELNSID, id); | |
6d458f5b | 416 | } |
de133464 | 417 | spin_lock_irq(&net->nsid_lock); |
6d458f5b | 418 | idr_destroy(&net->netns_ids); |
de133464 | 419 | spin_unlock_irq(&net->nsid_lock); |
6d458f5b | 420 | |
72ad937a | 421 | } |
6a1a3b9f PE |
422 | rtnl_unlock(); |
423 | ||
11a28d37 JB |
424 | /* |
425 | * Another CPU might be rcu-iterating the list, wait for it. | |
426 | * This needs to be before calling the exit() notifiers, so | |
427 | * the rcu_barrier() below isn't sufficient alone. | |
428 | */ | |
429 | synchronize_rcu(); | |
430 | ||
6a1a3b9f | 431 | /* Run all of the network namespace exit methods */ |
72ad937a EB |
432 | list_for_each_entry_reverse(ops, &pernet_list, list) |
433 | ops_exit_list(ops, &net_exit_list); | |
434 | ||
f875bae0 | 435 | /* Free the net generic variables */ |
72ad937a EB |
436 | list_for_each_entry_reverse(ops, &pernet_list, list) |
437 | ops_free_list(ops, &net_exit_list); | |
6a1a3b9f PE |
438 | |
439 | mutex_unlock(&net_mutex); | |
440 | ||
441 | /* Ensure there are no outstanding rcu callbacks using this | |
442 | * network namespace. | |
443 | */ | |
444 | rcu_barrier(); | |
445 | ||
446 | /* Finally it is safe to free my network namespace structure */ | |
72ad937a EB |
447 | list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) { |
448 | list_del_init(&net->exit_list); | |
038e7332 | 449 | put_user_ns(net->user_ns); |
a685e089 | 450 | net_drop_ns(net); |
2b035b39 | 451 | } |
6a1a3b9f | 452 | } |
2b035b39 | 453 | static DECLARE_WORK(net_cleanup_work, cleanup_net); |
6a1a3b9f PE |
454 | |
455 | void __put_net(struct net *net) | |
456 | { | |
457 | /* Cleanup the network namespace in process context */ | |
2b035b39 EB |
458 | unsigned long flags; |
459 | ||
460 | spin_lock_irqsave(&cleanup_list_lock, flags); | |
461 | list_add(&net->cleanup_list, &cleanup_list); | |
462 | spin_unlock_irqrestore(&cleanup_list_lock, flags); | |
463 | ||
464 | queue_work(netns_wq, &net_cleanup_work); | |
6a1a3b9f PE |
465 | } |
466 | EXPORT_SYMBOL_GPL(__put_net); | |
467 | ||
956c9207 SR |
468 | struct net *get_net_ns_by_fd(int fd) |
469 | { | |
956c9207 | 470 | struct file *file; |
33c42940 | 471 | struct ns_common *ns; |
956c9207 SR |
472 | struct net *net; |
473 | ||
956c9207 | 474 | file = proc_ns_fget(fd); |
c316e6a3 AV |
475 | if (IS_ERR(file)) |
476 | return ERR_CAST(file); | |
956c9207 | 477 | |
f77c8014 | 478 | ns = get_proc_ns(file_inode(file)); |
33c42940 AV |
479 | if (ns->ops == &netns_operations) |
480 | net = get_net(container_of(ns, struct net, ns)); | |
c316e6a3 AV |
481 | else |
482 | net = ERR_PTR(-EINVAL); | |
956c9207 | 483 | |
c316e6a3 | 484 | fput(file); |
956c9207 SR |
485 | return net; |
486 | } | |
487 | ||
6a1a3b9f | 488 | #else |
956c9207 SR |
489 | struct net *get_net_ns_by_fd(int fd) |
490 | { | |
491 | return ERR_PTR(-EINVAL); | |
492 | } | |
6a1a3b9f | 493 | #endif |
4b681c82 | 494 | EXPORT_SYMBOL_GPL(get_net_ns_by_fd); |
6a1a3b9f | 495 | |
30ffee84 JB |
496 | struct net *get_net_ns_by_pid(pid_t pid) |
497 | { | |
498 | struct task_struct *tsk; | |
499 | struct net *net; | |
500 | ||
501 | /* Lookup the network namespace */ | |
502 | net = ERR_PTR(-ESRCH); | |
503 | rcu_read_lock(); | |
504 | tsk = find_task_by_vpid(pid); | |
505 | if (tsk) { | |
506 | struct nsproxy *nsproxy; | |
728dba3a EB |
507 | task_lock(tsk); |
508 | nsproxy = tsk->nsproxy; | |
30ffee84 JB |
509 | if (nsproxy) |
510 | net = get_net(nsproxy->net_ns); | |
728dba3a | 511 | task_unlock(tsk); |
30ffee84 JB |
512 | } |
513 | rcu_read_unlock(); | |
514 | return net; | |
515 | } | |
516 | EXPORT_SYMBOL_GPL(get_net_ns_by_pid); | |
517 | ||
98f842e6 EB |
518 | static __net_init int net_ns_net_init(struct net *net) |
519 | { | |
33c42940 AV |
520 | #ifdef CONFIG_NET_NS |
521 | net->ns.ops = &netns_operations; | |
522 | #endif | |
6344c433 | 523 | return ns_alloc_inum(&net->ns); |
98f842e6 EB |
524 | } |
525 | ||
526 | static __net_exit void net_ns_net_exit(struct net *net) | |
527 | { | |
6344c433 | 528 | ns_free_inum(&net->ns); |
98f842e6 EB |
529 | } |
530 | ||
531 | static struct pernet_operations __net_initdata net_ns_ops = { | |
532 | .init = net_ns_net_init, | |
533 | .exit = net_ns_net_exit, | |
534 | }; | |
535 | ||
0c7aecd4 ND |
536 | static struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = { |
537 | [NETNSA_NONE] = { .type = NLA_UNSPEC }, | |
538 | [NETNSA_NSID] = { .type = NLA_S32 }, | |
539 | [NETNSA_PID] = { .type = NLA_U32 }, | |
540 | [NETNSA_FD] = { .type = NLA_U32 }, | |
541 | }; | |
542 | ||
543 | static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
544 | { | |
545 | struct net *net = sock_net(skb->sk); | |
546 | struct nlattr *tb[NETNSA_MAX + 1]; | |
95f38411 | 547 | unsigned long flags; |
0c7aecd4 ND |
548 | struct net *peer; |
549 | int nsid, err; | |
550 | ||
551 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
552 | rtnl_net_policy); | |
553 | if (err < 0) | |
554 | return err; | |
555 | if (!tb[NETNSA_NSID]) | |
556 | return -EINVAL; | |
557 | nsid = nla_get_s32(tb[NETNSA_NSID]); | |
558 | ||
559 | if (tb[NETNSA_PID]) | |
560 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); | |
561 | else if (tb[NETNSA_FD]) | |
562 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); | |
563 | else | |
564 | return -EINVAL; | |
565 | if (IS_ERR(peer)) | |
566 | return PTR_ERR(peer); | |
567 | ||
de133464 | 568 | spin_lock_irqsave(&net->nsid_lock, flags); |
3138dbf8 | 569 | if (__peernet2id(net, peer) >= 0) { |
de133464 | 570 | spin_unlock_irqrestore(&net->nsid_lock, flags); |
0c7aecd4 ND |
571 | err = -EEXIST; |
572 | goto out; | |
573 | } | |
574 | ||
575 | err = alloc_netid(net, peer, nsid); | |
de133464 | 576 | spin_unlock_irqrestore(&net->nsid_lock, flags); |
3138dbf8 ND |
577 | if (err >= 0) { |
578 | rtnl_net_notifyid(net, RTM_NEWNSID, err); | |
0c7aecd4 | 579 | err = 0; |
3138dbf8 | 580 | } |
0c7aecd4 ND |
581 | out: |
582 | put_net(peer); | |
583 | return err; | |
584 | } | |
585 | ||
586 | static int rtnl_net_get_size(void) | |
587 | { | |
588 | return NLMSG_ALIGN(sizeof(struct rtgenmsg)) | |
589 | + nla_total_size(sizeof(s32)) /* NETNSA_NSID */ | |
590 | ; | |
591 | } | |
592 | ||
593 | static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags, | |
cab3c8ec | 594 | int cmd, struct net *net, int nsid) |
0c7aecd4 ND |
595 | { |
596 | struct nlmsghdr *nlh; | |
597 | struct rtgenmsg *rth; | |
0c7aecd4 ND |
598 | |
599 | nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags); | |
600 | if (!nlh) | |
601 | return -EMSGSIZE; | |
602 | ||
603 | rth = nlmsg_data(nlh); | |
604 | rth->rtgen_family = AF_UNSPEC; | |
605 | ||
cab3c8ec | 606 | if (nla_put_s32(skb, NETNSA_NSID, nsid)) |
0c7aecd4 ND |
607 | goto nla_put_failure; |
608 | ||
609 | nlmsg_end(skb, nlh); | |
610 | return 0; | |
611 | ||
612 | nla_put_failure: | |
613 | nlmsg_cancel(skb, nlh); | |
614 | return -EMSGSIZE; | |
615 | } | |
616 | ||
617 | static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
618 | { | |
619 | struct net *net = sock_net(skb->sk); | |
620 | struct nlattr *tb[NETNSA_MAX + 1]; | |
621 | struct sk_buff *msg; | |
0c7aecd4 | 622 | struct net *peer; |
cab3c8ec | 623 | int err, id; |
0c7aecd4 ND |
624 | |
625 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
626 | rtnl_net_policy); | |
627 | if (err < 0) | |
628 | return err; | |
629 | if (tb[NETNSA_PID]) | |
630 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); | |
631 | else if (tb[NETNSA_FD]) | |
632 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); | |
633 | else | |
634 | return -EINVAL; | |
635 | ||
636 | if (IS_ERR(peer)) | |
637 | return PTR_ERR(peer); | |
638 | ||
639 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
640 | if (!msg) { | |
641 | err = -ENOMEM; | |
642 | goto out; | |
643 | } | |
644 | ||
95f38411 | 645 | id = peernet2id(net, peer); |
0c7aecd4 | 646 | err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, |
b04096ff | 647 | RTM_NEWNSID, net, id); |
0c7aecd4 ND |
648 | if (err < 0) |
649 | goto err_out; | |
650 | ||
651 | err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid); | |
652 | goto out; | |
653 | ||
654 | err_out: | |
655 | nlmsg_free(msg); | |
656 | out: | |
657 | put_net(peer); | |
658 | return err; | |
659 | } | |
660 | ||
a143c40c ND |
661 | struct rtnl_net_dump_cb { |
662 | struct net *net; | |
663 | struct sk_buff *skb; | |
664 | struct netlink_callback *cb; | |
665 | int idx; | |
666 | int s_idx; | |
667 | }; | |
668 | ||
669 | static int rtnl_net_dumpid_one(int id, void *peer, void *data) | |
670 | { | |
671 | struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data; | |
672 | int ret; | |
673 | ||
674 | if (net_cb->idx < net_cb->s_idx) | |
675 | goto cont; | |
676 | ||
677 | ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid, | |
678 | net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI, | |
cab3c8ec | 679 | RTM_NEWNSID, net_cb->net, id); |
a143c40c ND |
680 | if (ret < 0) |
681 | return ret; | |
682 | ||
683 | cont: | |
684 | net_cb->idx++; | |
685 | return 0; | |
686 | } | |
687 | ||
688 | static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb) | |
689 | { | |
690 | struct net *net = sock_net(skb->sk); | |
691 | struct rtnl_net_dump_cb net_cb = { | |
692 | .net = net, | |
693 | .skb = skb, | |
694 | .cb = cb, | |
695 | .idx = 0, | |
696 | .s_idx = cb->args[0], | |
697 | }; | |
95f38411 | 698 | unsigned long flags; |
a143c40c | 699 | |
de133464 | 700 | spin_lock_irqsave(&net->nsid_lock, flags); |
a143c40c | 701 | idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb); |
de133464 | 702 | spin_unlock_irqrestore(&net->nsid_lock, flags); |
a143c40c ND |
703 | |
704 | cb->args[0] = net_cb.idx; | |
705 | return skb->len; | |
706 | } | |
707 | ||
cab3c8ec | 708 | static void rtnl_net_notifyid(struct net *net, int cmd, int id) |
9a963454 ND |
709 | { |
710 | struct sk_buff *msg; | |
711 | int err = -ENOMEM; | |
712 | ||
713 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
714 | if (!msg) | |
715 | goto out; | |
716 | ||
cab3c8ec | 717 | err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, id); |
9a963454 ND |
718 | if (err < 0) |
719 | goto err_out; | |
720 | ||
721 | rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0); | |
722 | return; | |
723 | ||
724 | err_out: | |
725 | nlmsg_free(msg); | |
726 | out: | |
727 | rtnl_set_sk_err(net, RTNLGRP_NSID, err); | |
728 | } | |
729 | ||
5f256bec EB |
730 | static int __init net_ns_init(void) |
731 | { | |
486a87f1 | 732 | struct net_generic *ng; |
5f256bec | 733 | |
d57a9212 | 734 | #ifdef CONFIG_NET_NS |
5f256bec EB |
735 | net_cachep = kmem_cache_create("net_namespace", sizeof(struct net), |
736 | SMP_CACHE_BYTES, | |
737 | SLAB_PANIC, NULL); | |
3ef1355d BT |
738 | |
739 | /* Create workqueue for cleanup */ | |
740 | netns_wq = create_singlethread_workqueue("netns"); | |
741 | if (!netns_wq) | |
742 | panic("Could not create netns workq"); | |
d57a9212 | 743 | #endif |
3ef1355d | 744 | |
486a87f1 DL |
745 | ng = net_alloc_generic(); |
746 | if (!ng) | |
747 | panic("Could not allocate generic netns"); | |
748 | ||
749 | rcu_assign_pointer(init_net.gen, ng); | |
750 | ||
5f256bec | 751 | mutex_lock(&net_mutex); |
038e7332 | 752 | if (setup_net(&init_net, &init_user_ns)) |
ca0f3112 | 753 | panic("Could not setup the initial network namespace"); |
5f256bec | 754 | |
f8c46cb3 DT |
755 | init_net_initialized = true; |
756 | ||
f4618d39 | 757 | rtnl_lock(); |
11a28d37 | 758 | list_add_tail_rcu(&init_net.list, &net_namespace_list); |
f4618d39 | 759 | rtnl_unlock(); |
5f256bec EB |
760 | |
761 | mutex_unlock(&net_mutex); | |
5f256bec | 762 | |
98f842e6 EB |
763 | register_pernet_subsys(&net_ns_ops); |
764 | ||
0c7aecd4 | 765 | rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL); |
a143c40c ND |
766 | rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid, |
767 | NULL); | |
0c7aecd4 | 768 | |
5f256bec EB |
769 | return 0; |
770 | } | |
771 | ||
772 | pure_initcall(net_ns_init); | |
773 | ||
ed160e83 | 774 | #ifdef CONFIG_NET_NS |
f875bae0 EB |
775 | static int __register_pernet_operations(struct list_head *list, |
776 | struct pernet_operations *ops) | |
5f256bec | 777 | { |
72ad937a | 778 | struct net *net; |
5f256bec | 779 | int error; |
72ad937a | 780 | LIST_HEAD(net_exit_list); |
5f256bec | 781 | |
5f256bec | 782 | list_add_tail(&ops->list, list); |
f875bae0 | 783 | if (ops->init || (ops->id && ops->size)) { |
1dba323b | 784 | for_each_net(net) { |
f875bae0 | 785 | error = ops_init(ops, net); |
5f256bec EB |
786 | if (error) |
787 | goto out_undo; | |
72ad937a | 788 | list_add_tail(&net->exit_list, &net_exit_list); |
5f256bec EB |
789 | } |
790 | } | |
1dba323b | 791 | return 0; |
5f256bec EB |
792 | |
793 | out_undo: | |
794 | /* If I have an error cleanup all namespaces I initialized */ | |
795 | list_del(&ops->list); | |
72ad937a EB |
796 | ops_exit_list(ops, &net_exit_list); |
797 | ops_free_list(ops, &net_exit_list); | |
1dba323b | 798 | return error; |
5f256bec EB |
799 | } |
800 | ||
f875bae0 | 801 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
5f256bec EB |
802 | { |
803 | struct net *net; | |
72ad937a | 804 | LIST_HEAD(net_exit_list); |
5f256bec EB |
805 | |
806 | list_del(&ops->list); | |
72ad937a EB |
807 | for_each_net(net) |
808 | list_add_tail(&net->exit_list, &net_exit_list); | |
809 | ops_exit_list(ops, &net_exit_list); | |
810 | ops_free_list(ops, &net_exit_list); | |
5f256bec EB |
811 | } |
812 | ||
ed160e83 DL |
813 | #else |
814 | ||
f875bae0 EB |
815 | static int __register_pernet_operations(struct list_head *list, |
816 | struct pernet_operations *ops) | |
ed160e83 | 817 | { |
f8c46cb3 DT |
818 | if (!init_net_initialized) { |
819 | list_add_tail(&ops->list, list); | |
820 | return 0; | |
821 | } | |
822 | ||
b922934d | 823 | return ops_init(ops, &init_net); |
ed160e83 DL |
824 | } |
825 | ||
f875bae0 | 826 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
ed160e83 | 827 | { |
f8c46cb3 DT |
828 | if (!init_net_initialized) { |
829 | list_del(&ops->list); | |
830 | } else { | |
831 | LIST_HEAD(net_exit_list); | |
832 | list_add(&init_net.exit_list, &net_exit_list); | |
833 | ops_exit_list(ops, &net_exit_list); | |
834 | ops_free_list(ops, &net_exit_list); | |
835 | } | |
ed160e83 | 836 | } |
f875bae0 EB |
837 | |
838 | #endif /* CONFIG_NET_NS */ | |
ed160e83 | 839 | |
c93cf61f PE |
840 | static DEFINE_IDA(net_generic_ids); |
841 | ||
f875bae0 EB |
842 | static int register_pernet_operations(struct list_head *list, |
843 | struct pernet_operations *ops) | |
844 | { | |
845 | int error; | |
846 | ||
847 | if (ops->id) { | |
848 | again: | |
849 | error = ida_get_new_above(&net_generic_ids, 1, ops->id); | |
850 | if (error < 0) { | |
851 | if (error == -EAGAIN) { | |
852 | ida_pre_get(&net_generic_ids, GFP_KERNEL); | |
853 | goto again; | |
854 | } | |
855 | return error; | |
856 | } | |
073862ba | 857 | max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id); |
f875bae0 EB |
858 | } |
859 | error = __register_pernet_operations(list, ops); | |
3a765eda EB |
860 | if (error) { |
861 | rcu_barrier(); | |
862 | if (ops->id) | |
863 | ida_remove(&net_generic_ids, *ops->id); | |
864 | } | |
f875bae0 EB |
865 | |
866 | return error; | |
867 | } | |
868 | ||
869 | static void unregister_pernet_operations(struct pernet_operations *ops) | |
870 | { | |
871 | ||
872 | __unregister_pernet_operations(ops); | |
3a765eda | 873 | rcu_barrier(); |
f875bae0 EB |
874 | if (ops->id) |
875 | ida_remove(&net_generic_ids, *ops->id); | |
876 | } | |
877 | ||
5f256bec EB |
878 | /** |
879 | * register_pernet_subsys - register a network namespace subsystem | |
880 | * @ops: pernet operations structure for the subsystem | |
881 | * | |
882 | * Register a subsystem which has init and exit functions | |
883 | * that are called when network namespaces are created and | |
884 | * destroyed respectively. | |
885 | * | |
886 | * When registered all network namespace init functions are | |
887 | * called for every existing network namespace. Allowing kernel | |
888 | * modules to have a race free view of the set of network namespaces. | |
889 | * | |
890 | * When a new network namespace is created all of the init | |
891 | * methods are called in the order in which they were registered. | |
892 | * | |
893 | * When a network namespace is destroyed all of the exit methods | |
894 | * are called in the reverse of the order with which they were | |
895 | * registered. | |
896 | */ | |
897 | int register_pernet_subsys(struct pernet_operations *ops) | |
898 | { | |
899 | int error; | |
900 | mutex_lock(&net_mutex); | |
901 | error = register_pernet_operations(first_device, ops); | |
902 | mutex_unlock(&net_mutex); | |
903 | return error; | |
904 | } | |
905 | EXPORT_SYMBOL_GPL(register_pernet_subsys); | |
906 | ||
907 | /** | |
908 | * unregister_pernet_subsys - unregister a network namespace subsystem | |
909 | * @ops: pernet operations structure to manipulate | |
910 | * | |
911 | * Remove the pernet operations structure from the list to be | |
53379e57 | 912 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
913 | * addition run the exit method for all existing network |
914 | * namespaces. | |
915 | */ | |
b3c981d2 | 916 | void unregister_pernet_subsys(struct pernet_operations *ops) |
5f256bec EB |
917 | { |
918 | mutex_lock(&net_mutex); | |
b3c981d2 | 919 | unregister_pernet_operations(ops); |
5f256bec EB |
920 | mutex_unlock(&net_mutex); |
921 | } | |
922 | EXPORT_SYMBOL_GPL(unregister_pernet_subsys); | |
923 | ||
924 | /** | |
925 | * register_pernet_device - register a network namespace device | |
926 | * @ops: pernet operations structure for the subsystem | |
927 | * | |
928 | * Register a device which has init and exit functions | |
929 | * that are called when network namespaces are created and | |
930 | * destroyed respectively. | |
931 | * | |
932 | * When registered all network namespace init functions are | |
933 | * called for every existing network namespace. Allowing kernel | |
934 | * modules to have a race free view of the set of network namespaces. | |
935 | * | |
936 | * When a new network namespace is created all of the init | |
937 | * methods are called in the order in which they were registered. | |
938 | * | |
939 | * When a network namespace is destroyed all of the exit methods | |
940 | * are called in the reverse of the order with which they were | |
941 | * registered. | |
942 | */ | |
943 | int register_pernet_device(struct pernet_operations *ops) | |
944 | { | |
945 | int error; | |
946 | mutex_lock(&net_mutex); | |
947 | error = register_pernet_operations(&pernet_list, ops); | |
948 | if (!error && (first_device == &pernet_list)) | |
949 | first_device = &ops->list; | |
950 | mutex_unlock(&net_mutex); | |
951 | return error; | |
952 | } | |
953 | EXPORT_SYMBOL_GPL(register_pernet_device); | |
954 | ||
955 | /** | |
956 | * unregister_pernet_device - unregister a network namespace netdevice | |
957 | * @ops: pernet operations structure to manipulate | |
958 | * | |
959 | * Remove the pernet operations structure from the list to be | |
53379e57 | 960 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
961 | * addition run the exit method for all existing network |
962 | * namespaces. | |
963 | */ | |
964 | void unregister_pernet_device(struct pernet_operations *ops) | |
965 | { | |
966 | mutex_lock(&net_mutex); | |
967 | if (&ops->list == first_device) | |
968 | first_device = first_device->next; | |
969 | unregister_pernet_operations(ops); | |
970 | mutex_unlock(&net_mutex); | |
971 | } | |
972 | EXPORT_SYMBOL_GPL(unregister_pernet_device); | |
13b6f576 EB |
973 | |
974 | #ifdef CONFIG_NET_NS | |
64964528 | 975 | static struct ns_common *netns_get(struct task_struct *task) |
13b6f576 | 976 | { |
f0630529 EB |
977 | struct net *net = NULL; |
978 | struct nsproxy *nsproxy; | |
979 | ||
728dba3a EB |
980 | task_lock(task); |
981 | nsproxy = task->nsproxy; | |
f0630529 EB |
982 | if (nsproxy) |
983 | net = get_net(nsproxy->net_ns); | |
728dba3a | 984 | task_unlock(task); |
f0630529 | 985 | |
ff24870f AV |
986 | return net ? &net->ns : NULL; |
987 | } | |
988 | ||
989 | static inline struct net *to_net_ns(struct ns_common *ns) | |
990 | { | |
991 | return container_of(ns, struct net, ns); | |
13b6f576 EB |
992 | } |
993 | ||
64964528 | 994 | static void netns_put(struct ns_common *ns) |
13b6f576 | 995 | { |
ff24870f | 996 | put_net(to_net_ns(ns)); |
13b6f576 EB |
997 | } |
998 | ||
64964528 | 999 | static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns) |
13b6f576 | 1000 | { |
ff24870f | 1001 | struct net *net = to_net_ns(ns); |
142e1d1d | 1002 | |
5e4a0847 | 1003 | if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) || |
c7b96acf | 1004 | !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) |
142e1d1d EB |
1005 | return -EPERM; |
1006 | ||
13b6f576 | 1007 | put_net(nsproxy->net_ns); |
142e1d1d | 1008 | nsproxy->net_ns = get_net(net); |
13b6f576 EB |
1009 | return 0; |
1010 | } | |
1011 | ||
1012 | const struct proc_ns_operations netns_operations = { | |
1013 | .name = "net", | |
1014 | .type = CLONE_NEWNET, | |
1015 | .get = netns_get, | |
1016 | .put = netns_put, | |
1017 | .install = netns_install, | |
1018 | }; | |
1019 | #endif |