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