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 | 218 | bool alloc; |
3138dbf8 | 219 | int id; |
0c7aecd4 | 220 | |
bc51dddf | 221 | spin_lock_bh(&net->nsid_lock); |
95f38411 | 222 | alloc = atomic_read(&peer->count) == 0 ? false : true; |
3138dbf8 | 223 | id = __peernet2id_alloc(net, peer, &alloc); |
bc51dddf | 224 | spin_unlock_bh(&net->nsid_lock); |
3138dbf8 ND |
225 | if (alloc && id >= 0) |
226 | rtnl_net_notifyid(net, RTM_NEWNSID, id); | |
227 | return id; | |
0c7aecd4 ND |
228 | } |
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 | 232 | { |
95f38411 ND |
233 | int id; |
234 | ||
bc51dddf | 235 | spin_lock_bh(&net->nsid_lock); |
95f38411 | 236 | id = __peernet2id(net, peer); |
bc51dddf | 237 | spin_unlock_bh(&net->nsid_lock); |
95f38411 ND |
238 | return id; |
239 | } | |
38f507f1 | 240 | EXPORT_SYMBOL(peernet2id); |
95f38411 | 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 | { | |
252 | struct net *peer; | |
253 | ||
254 | if (id < 0) | |
255 | return NULL; | |
256 | ||
257 | rcu_read_lock(); | |
bc51dddf | 258 | spin_lock_bh(&net->nsid_lock); |
0c7aecd4 ND |
259 | peer = idr_find(&net->netns_ids, id); |
260 | if (peer) | |
261 | get_net(peer); | |
bc51dddf | 262 | spin_unlock_bh(&net->nsid_lock); |
0c7aecd4 ND |
263 | rcu_read_unlock(); |
264 | ||
265 | return peer; | |
266 | } | |
267 | ||
5f256bec EB |
268 | /* |
269 | * setup_net runs the initializers for the network namespace object. | |
270 | */ | |
038e7332 | 271 | static __net_init int setup_net(struct net *net, struct user_namespace *user_ns) |
5f256bec EB |
272 | { |
273 | /* Must be called with net_mutex held */ | |
f875bae0 | 274 | const struct pernet_operations *ops, *saved_ops; |
486a87f1 | 275 | int error = 0; |
72ad937a | 276 | LIST_HEAD(net_exit_list); |
5f256bec | 277 | |
5f256bec | 278 | atomic_set(&net->count, 1); |
a685e089 | 279 | atomic_set(&net->passive, 1); |
4e985ada | 280 | net->dev_base_seq = 1; |
038e7332 | 281 | net->user_ns = user_ns; |
0c7aecd4 | 282 | idr_init(&net->netns_ids); |
de133464 | 283 | spin_lock_init(&net->nsid_lock); |
486a87f1 | 284 | |
768f3591 | 285 | list_for_each_entry(ops, &pernet_list, list) { |
f875bae0 EB |
286 | error = ops_init(ops, net); |
287 | if (error < 0) | |
288 | goto out_undo; | |
5f256bec EB |
289 | } |
290 | out: | |
291 | return error; | |
768f3591 | 292 | |
5f256bec EB |
293 | out_undo: |
294 | /* Walk through the list backwards calling the exit functions | |
295 | * for the pernet modules whose init functions did not fail. | |
296 | */ | |
72ad937a | 297 | list_add(&net->exit_list, &net_exit_list); |
f875bae0 | 298 | saved_ops = ops; |
72ad937a EB |
299 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) |
300 | ops_exit_list(ops, &net_exit_list); | |
301 | ||
f875bae0 EB |
302 | ops = saved_ops; |
303 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) | |
72ad937a | 304 | ops_free_list(ops, &net_exit_list); |
310928d9 DL |
305 | |
306 | rcu_barrier(); | |
5f256bec EB |
307 | goto out; |
308 | } | |
309 | ||
6a1a3b9f | 310 | |
ebe47d47 | 311 | #ifdef CONFIG_NET_NS |
2ed6afde AB |
312 | static struct ucounts *inc_net_namespaces(struct user_namespace *ns) |
313 | { | |
314 | return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES); | |
315 | } | |
316 | ||
317 | static void dec_net_namespaces(struct ucounts *ucounts) | |
318 | { | |
319 | dec_ucount(ucounts, UCOUNT_NET_NAMESPACES); | |
320 | } | |
321 | ||
ebe47d47 CN |
322 | static struct kmem_cache *net_cachep; |
323 | static struct workqueue_struct *netns_wq; | |
324 | ||
486a87f1 | 325 | static struct net *net_alloc(void) |
45a19b0a | 326 | { |
486a87f1 DL |
327 | struct net *net = NULL; |
328 | struct net_generic *ng; | |
329 | ||
330 | ng = net_alloc_generic(); | |
331 | if (!ng) | |
332 | goto out; | |
333 | ||
334 | net = kmem_cache_zalloc(net_cachep, GFP_KERNEL); | |
45a19b0a | 335 | if (!net) |
486a87f1 | 336 | goto out_free; |
45a19b0a | 337 | |
486a87f1 DL |
338 | rcu_assign_pointer(net->gen, ng); |
339 | out: | |
340 | return net; | |
341 | ||
342 | out_free: | |
343 | kfree(ng); | |
344 | goto out; | |
345 | } | |
346 | ||
347 | static void net_free(struct net *net) | |
348 | { | |
416c51e1 | 349 | kfree(rcu_access_pointer(net->gen)); |
45a19b0a JFS |
350 | kmem_cache_free(net_cachep, net); |
351 | } | |
352 | ||
a685e089 AV |
353 | void net_drop_ns(void *p) |
354 | { | |
355 | struct net *ns = p; | |
356 | if (ns && atomic_dec_and_test(&ns->passive)) | |
357 | net_free(ns); | |
358 | } | |
359 | ||
038e7332 EB |
360 | struct net *copy_net_ns(unsigned long flags, |
361 | struct user_namespace *user_ns, struct net *old_net) | |
9dd776b6 | 362 | { |
70328660 | 363 | struct ucounts *ucounts; |
088eb2d9 AD |
364 | struct net *net; |
365 | int rv; | |
9dd776b6 | 366 | |
911cb193 RL |
367 | if (!(flags & CLONE_NEWNET)) |
368 | return get_net(old_net); | |
369 | ||
70328660 EB |
370 | ucounts = inc_net_namespaces(user_ns); |
371 | if (!ucounts) | |
df75e774 | 372 | return ERR_PTR(-ENOSPC); |
70328660 | 373 | |
088eb2d9 | 374 | net = net_alloc(); |
70328660 EB |
375 | if (!net) { |
376 | dec_net_namespaces(ucounts); | |
088eb2d9 | 377 | return ERR_PTR(-ENOMEM); |
70328660 | 378 | } |
038e7332 EB |
379 | |
380 | get_user_ns(user_ns); | |
381 | ||
9dd776b6 | 382 | mutex_lock(&net_mutex); |
70328660 | 383 | net->ucounts = ucounts; |
038e7332 | 384 | rv = setup_net(net, user_ns); |
088eb2d9 | 385 | if (rv == 0) { |
486a87f1 | 386 | rtnl_lock(); |
11a28d37 | 387 | list_add_tail_rcu(&net->list, &net_namespace_list); |
486a87f1 DL |
388 | rtnl_unlock(); |
389 | } | |
9dd776b6 | 390 | mutex_unlock(&net_mutex); |
088eb2d9 | 391 | if (rv < 0) { |
70328660 | 392 | dec_net_namespaces(ucounts); |
038e7332 | 393 | put_user_ns(user_ns); |
a685e089 | 394 | net_drop_ns(net); |
088eb2d9 AD |
395 | return ERR_PTR(rv); |
396 | } | |
397 | return net; | |
398 | } | |
486a87f1 | 399 | |
2b035b39 EB |
400 | static DEFINE_SPINLOCK(cleanup_list_lock); |
401 | static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */ | |
402 | ||
6a1a3b9f PE |
403 | static void cleanup_net(struct work_struct *work) |
404 | { | |
f875bae0 | 405 | const struct pernet_operations *ops; |
6d458f5b | 406 | struct net *net, *tmp; |
1818ce4d | 407 | struct list_head net_kill_list; |
72ad937a | 408 | LIST_HEAD(net_exit_list); |
6a1a3b9f | 409 | |
2b035b39 EB |
410 | /* Atomically snapshot the list of namespaces to cleanup */ |
411 | spin_lock_irq(&cleanup_list_lock); | |
412 | list_replace_init(&cleanup_list, &net_kill_list); | |
413 | spin_unlock_irq(&cleanup_list_lock); | |
6a1a3b9f PE |
414 | |
415 | mutex_lock(&net_mutex); | |
416 | ||
417 | /* Don't let anyone else find us. */ | |
418 | rtnl_lock(); | |
72ad937a | 419 | list_for_each_entry(net, &net_kill_list, cleanup_list) { |
2b035b39 | 420 | list_del_rcu(&net->list); |
72ad937a | 421 | list_add_tail(&net->exit_list, &net_exit_list); |
6d458f5b | 422 | for_each_net(tmp) { |
95f38411 | 423 | int id; |
6d458f5b | 424 | |
bc51dddf | 425 | spin_lock_bh(&tmp->nsid_lock); |
95f38411 ND |
426 | id = __peernet2id(tmp, net); |
427 | if (id >= 0) | |
6d458f5b | 428 | idr_remove(&tmp->netns_ids, id); |
bc51dddf | 429 | spin_unlock_bh(&tmp->nsid_lock); |
95f38411 ND |
430 | if (id >= 0) |
431 | rtnl_net_notifyid(tmp, RTM_DELNSID, id); | |
6d458f5b | 432 | } |
bc51dddf | 433 | spin_lock_bh(&net->nsid_lock); |
6d458f5b | 434 | idr_destroy(&net->netns_ids); |
bc51dddf | 435 | spin_unlock_bh(&net->nsid_lock); |
6d458f5b | 436 | |
72ad937a | 437 | } |
6a1a3b9f PE |
438 | rtnl_unlock(); |
439 | ||
11a28d37 JB |
440 | /* |
441 | * Another CPU might be rcu-iterating the list, wait for it. | |
442 | * This needs to be before calling the exit() notifiers, so | |
443 | * the rcu_barrier() below isn't sufficient alone. | |
444 | */ | |
445 | synchronize_rcu(); | |
446 | ||
6a1a3b9f | 447 | /* Run all of the network namespace exit methods */ |
72ad937a EB |
448 | list_for_each_entry_reverse(ops, &pernet_list, list) |
449 | ops_exit_list(ops, &net_exit_list); | |
450 | ||
f875bae0 | 451 | /* Free the net generic variables */ |
72ad937a EB |
452 | list_for_each_entry_reverse(ops, &pernet_list, list) |
453 | ops_free_list(ops, &net_exit_list); | |
6a1a3b9f PE |
454 | |
455 | mutex_unlock(&net_mutex); | |
456 | ||
457 | /* Ensure there are no outstanding rcu callbacks using this | |
458 | * network namespace. | |
459 | */ | |
460 | rcu_barrier(); | |
461 | ||
462 | /* Finally it is safe to free my network namespace structure */ | |
72ad937a EB |
463 | list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) { |
464 | list_del_init(&net->exit_list); | |
70328660 | 465 | dec_net_namespaces(net->ucounts); |
038e7332 | 466 | put_user_ns(net->user_ns); |
a685e089 | 467 | net_drop_ns(net); |
2b035b39 | 468 | } |
6a1a3b9f | 469 | } |
2b035b39 | 470 | static DECLARE_WORK(net_cleanup_work, cleanup_net); |
6a1a3b9f PE |
471 | |
472 | void __put_net(struct net *net) | |
473 | { | |
474 | /* Cleanup the network namespace in process context */ | |
2b035b39 EB |
475 | unsigned long flags; |
476 | ||
477 | spin_lock_irqsave(&cleanup_list_lock, flags); | |
478 | list_add(&net->cleanup_list, &cleanup_list); | |
479 | spin_unlock_irqrestore(&cleanup_list_lock, flags); | |
480 | ||
481 | queue_work(netns_wq, &net_cleanup_work); | |
6a1a3b9f PE |
482 | } |
483 | EXPORT_SYMBOL_GPL(__put_net); | |
484 | ||
956c9207 SR |
485 | struct net *get_net_ns_by_fd(int fd) |
486 | { | |
956c9207 | 487 | struct file *file; |
33c42940 | 488 | struct ns_common *ns; |
956c9207 SR |
489 | struct net *net; |
490 | ||
956c9207 | 491 | file = proc_ns_fget(fd); |
c316e6a3 AV |
492 | if (IS_ERR(file)) |
493 | return ERR_CAST(file); | |
956c9207 | 494 | |
f77c8014 | 495 | ns = get_proc_ns(file_inode(file)); |
33c42940 AV |
496 | if (ns->ops == &netns_operations) |
497 | net = get_net(container_of(ns, struct net, ns)); | |
c316e6a3 AV |
498 | else |
499 | net = ERR_PTR(-EINVAL); | |
956c9207 | 500 | |
c316e6a3 | 501 | fput(file); |
956c9207 SR |
502 | return net; |
503 | } | |
504 | ||
6a1a3b9f | 505 | #else |
956c9207 SR |
506 | struct net *get_net_ns_by_fd(int fd) |
507 | { | |
508 | return ERR_PTR(-EINVAL); | |
509 | } | |
6a1a3b9f | 510 | #endif |
4b681c82 | 511 | EXPORT_SYMBOL_GPL(get_net_ns_by_fd); |
6a1a3b9f | 512 | |
30ffee84 JB |
513 | struct net *get_net_ns_by_pid(pid_t pid) |
514 | { | |
515 | struct task_struct *tsk; | |
516 | struct net *net; | |
517 | ||
518 | /* Lookup the network namespace */ | |
519 | net = ERR_PTR(-ESRCH); | |
520 | rcu_read_lock(); | |
521 | tsk = find_task_by_vpid(pid); | |
522 | if (tsk) { | |
523 | struct nsproxy *nsproxy; | |
728dba3a EB |
524 | task_lock(tsk); |
525 | nsproxy = tsk->nsproxy; | |
30ffee84 JB |
526 | if (nsproxy) |
527 | net = get_net(nsproxy->net_ns); | |
728dba3a | 528 | task_unlock(tsk); |
30ffee84 JB |
529 | } |
530 | rcu_read_unlock(); | |
531 | return net; | |
532 | } | |
533 | EXPORT_SYMBOL_GPL(get_net_ns_by_pid); | |
534 | ||
98f842e6 EB |
535 | static __net_init int net_ns_net_init(struct net *net) |
536 | { | |
33c42940 AV |
537 | #ifdef CONFIG_NET_NS |
538 | net->ns.ops = &netns_operations; | |
539 | #endif | |
6344c433 | 540 | return ns_alloc_inum(&net->ns); |
98f842e6 EB |
541 | } |
542 | ||
543 | static __net_exit void net_ns_net_exit(struct net *net) | |
544 | { | |
6344c433 | 545 | ns_free_inum(&net->ns); |
98f842e6 EB |
546 | } |
547 | ||
548 | static struct pernet_operations __net_initdata net_ns_ops = { | |
549 | .init = net_ns_net_init, | |
550 | .exit = net_ns_net_exit, | |
551 | }; | |
552 | ||
3ee5256d | 553 | static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = { |
0c7aecd4 ND |
554 | [NETNSA_NONE] = { .type = NLA_UNSPEC }, |
555 | [NETNSA_NSID] = { .type = NLA_S32 }, | |
556 | [NETNSA_PID] = { .type = NLA_U32 }, | |
557 | [NETNSA_FD] = { .type = NLA_U32 }, | |
558 | }; | |
559 | ||
560 | static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
561 | { | |
562 | struct net *net = sock_net(skb->sk); | |
563 | struct nlattr *tb[NETNSA_MAX + 1]; | |
564 | struct net *peer; | |
565 | int nsid, err; | |
566 | ||
567 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
568 | rtnl_net_policy); | |
569 | if (err < 0) | |
570 | return err; | |
571 | if (!tb[NETNSA_NSID]) | |
572 | return -EINVAL; | |
573 | nsid = nla_get_s32(tb[NETNSA_NSID]); | |
574 | ||
575 | if (tb[NETNSA_PID]) | |
576 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); | |
577 | else if (tb[NETNSA_FD]) | |
578 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); | |
579 | else | |
580 | return -EINVAL; | |
581 | if (IS_ERR(peer)) | |
582 | return PTR_ERR(peer); | |
583 | ||
bc51dddf | 584 | spin_lock_bh(&net->nsid_lock); |
3138dbf8 | 585 | if (__peernet2id(net, peer) >= 0) { |
bc51dddf | 586 | spin_unlock_bh(&net->nsid_lock); |
0c7aecd4 ND |
587 | err = -EEXIST; |
588 | goto out; | |
589 | } | |
590 | ||
591 | err = alloc_netid(net, peer, nsid); | |
bc51dddf | 592 | spin_unlock_bh(&net->nsid_lock); |
3138dbf8 ND |
593 | if (err >= 0) { |
594 | rtnl_net_notifyid(net, RTM_NEWNSID, err); | |
0c7aecd4 | 595 | err = 0; |
3138dbf8 | 596 | } |
0c7aecd4 ND |
597 | out: |
598 | put_net(peer); | |
599 | return err; | |
600 | } | |
601 | ||
602 | static int rtnl_net_get_size(void) | |
603 | { | |
604 | return NLMSG_ALIGN(sizeof(struct rtgenmsg)) | |
605 | + nla_total_size(sizeof(s32)) /* NETNSA_NSID */ | |
606 | ; | |
607 | } | |
608 | ||
609 | static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags, | |
cab3c8ec | 610 | int cmd, struct net *net, int nsid) |
0c7aecd4 ND |
611 | { |
612 | struct nlmsghdr *nlh; | |
613 | struct rtgenmsg *rth; | |
0c7aecd4 ND |
614 | |
615 | nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags); | |
616 | if (!nlh) | |
617 | return -EMSGSIZE; | |
618 | ||
619 | rth = nlmsg_data(nlh); | |
620 | rth->rtgen_family = AF_UNSPEC; | |
621 | ||
cab3c8ec | 622 | if (nla_put_s32(skb, NETNSA_NSID, nsid)) |
0c7aecd4 ND |
623 | goto nla_put_failure; |
624 | ||
625 | nlmsg_end(skb, nlh); | |
626 | return 0; | |
627 | ||
628 | nla_put_failure: | |
629 | nlmsg_cancel(skb, nlh); | |
630 | return -EMSGSIZE; | |
631 | } | |
632 | ||
633 | static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
634 | { | |
635 | struct net *net = sock_net(skb->sk); | |
636 | struct nlattr *tb[NETNSA_MAX + 1]; | |
637 | struct sk_buff *msg; | |
0c7aecd4 | 638 | struct net *peer; |
cab3c8ec | 639 | int err, id; |
0c7aecd4 ND |
640 | |
641 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
642 | rtnl_net_policy); | |
643 | if (err < 0) | |
644 | return err; | |
645 | if (tb[NETNSA_PID]) | |
646 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); | |
647 | else if (tb[NETNSA_FD]) | |
648 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); | |
649 | else | |
650 | return -EINVAL; | |
651 | ||
652 | if (IS_ERR(peer)) | |
653 | return PTR_ERR(peer); | |
654 | ||
655 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
656 | if (!msg) { | |
657 | err = -ENOMEM; | |
658 | goto out; | |
659 | } | |
660 | ||
95f38411 | 661 | id = peernet2id(net, peer); |
0c7aecd4 | 662 | err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, |
b04096ff | 663 | RTM_NEWNSID, net, id); |
0c7aecd4 ND |
664 | if (err < 0) |
665 | goto err_out; | |
666 | ||
667 | err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid); | |
668 | goto out; | |
669 | ||
670 | err_out: | |
671 | nlmsg_free(msg); | |
672 | out: | |
673 | put_net(peer); | |
674 | return err; | |
675 | } | |
676 | ||
a143c40c ND |
677 | struct rtnl_net_dump_cb { |
678 | struct net *net; | |
679 | struct sk_buff *skb; | |
680 | struct netlink_callback *cb; | |
681 | int idx; | |
682 | int s_idx; | |
683 | }; | |
684 | ||
685 | static int rtnl_net_dumpid_one(int id, void *peer, void *data) | |
686 | { | |
687 | struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data; | |
688 | int ret; | |
689 | ||
690 | if (net_cb->idx < net_cb->s_idx) | |
691 | goto cont; | |
692 | ||
693 | ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid, | |
694 | net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI, | |
cab3c8ec | 695 | RTM_NEWNSID, net_cb->net, id); |
a143c40c ND |
696 | if (ret < 0) |
697 | return ret; | |
698 | ||
699 | cont: | |
700 | net_cb->idx++; | |
701 | return 0; | |
702 | } | |
703 | ||
704 | static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb) | |
705 | { | |
706 | struct net *net = sock_net(skb->sk); | |
707 | struct rtnl_net_dump_cb net_cb = { | |
708 | .net = net, | |
709 | .skb = skb, | |
710 | .cb = cb, | |
711 | .idx = 0, | |
712 | .s_idx = cb->args[0], | |
713 | }; | |
714 | ||
bc51dddf | 715 | spin_lock_bh(&net->nsid_lock); |
a143c40c | 716 | idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb); |
bc51dddf | 717 | spin_unlock_bh(&net->nsid_lock); |
a143c40c ND |
718 | |
719 | cb->args[0] = net_cb.idx; | |
720 | return skb->len; | |
721 | } | |
722 | ||
cab3c8ec | 723 | static void rtnl_net_notifyid(struct net *net, int cmd, int id) |
9a963454 ND |
724 | { |
725 | struct sk_buff *msg; | |
726 | int err = -ENOMEM; | |
727 | ||
728 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
729 | if (!msg) | |
730 | goto out; | |
731 | ||
cab3c8ec | 732 | err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, id); |
9a963454 ND |
733 | if (err < 0) |
734 | goto err_out; | |
735 | ||
736 | rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0); | |
737 | return; | |
738 | ||
739 | err_out: | |
740 | nlmsg_free(msg); | |
741 | out: | |
742 | rtnl_set_sk_err(net, RTNLGRP_NSID, err); | |
743 | } | |
744 | ||
5f256bec EB |
745 | static int __init net_ns_init(void) |
746 | { | |
486a87f1 | 747 | struct net_generic *ng; |
5f256bec | 748 | |
d57a9212 | 749 | #ifdef CONFIG_NET_NS |
5f256bec EB |
750 | net_cachep = kmem_cache_create("net_namespace", sizeof(struct net), |
751 | SMP_CACHE_BYTES, | |
752 | SLAB_PANIC, NULL); | |
3ef1355d BT |
753 | |
754 | /* Create workqueue for cleanup */ | |
755 | netns_wq = create_singlethread_workqueue("netns"); | |
756 | if (!netns_wq) | |
757 | panic("Could not create netns workq"); | |
d57a9212 | 758 | #endif |
3ef1355d | 759 | |
486a87f1 DL |
760 | ng = net_alloc_generic(); |
761 | if (!ng) | |
762 | panic("Could not allocate generic netns"); | |
763 | ||
764 | rcu_assign_pointer(init_net.gen, ng); | |
765 | ||
5f256bec | 766 | mutex_lock(&net_mutex); |
038e7332 | 767 | if (setup_net(&init_net, &init_user_ns)) |
ca0f3112 | 768 | panic("Could not setup the initial network namespace"); |
5f256bec | 769 | |
f8c46cb3 DT |
770 | init_net_initialized = true; |
771 | ||
f4618d39 | 772 | rtnl_lock(); |
11a28d37 | 773 | list_add_tail_rcu(&init_net.list, &net_namespace_list); |
f4618d39 | 774 | rtnl_unlock(); |
5f256bec EB |
775 | |
776 | mutex_unlock(&net_mutex); | |
5f256bec | 777 | |
98f842e6 EB |
778 | register_pernet_subsys(&net_ns_ops); |
779 | ||
0c7aecd4 | 780 | rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL); |
a143c40c ND |
781 | rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid, |
782 | NULL); | |
0c7aecd4 | 783 | |
5f256bec EB |
784 | return 0; |
785 | } | |
786 | ||
787 | pure_initcall(net_ns_init); | |
788 | ||
ed160e83 | 789 | #ifdef CONFIG_NET_NS |
f875bae0 EB |
790 | static int __register_pernet_operations(struct list_head *list, |
791 | struct pernet_operations *ops) | |
5f256bec | 792 | { |
72ad937a | 793 | struct net *net; |
5f256bec | 794 | int error; |
72ad937a | 795 | LIST_HEAD(net_exit_list); |
5f256bec | 796 | |
5f256bec | 797 | list_add_tail(&ops->list, list); |
f875bae0 | 798 | if (ops->init || (ops->id && ops->size)) { |
1dba323b | 799 | for_each_net(net) { |
f875bae0 | 800 | error = ops_init(ops, net); |
5f256bec EB |
801 | if (error) |
802 | goto out_undo; | |
72ad937a | 803 | list_add_tail(&net->exit_list, &net_exit_list); |
5f256bec EB |
804 | } |
805 | } | |
1dba323b | 806 | return 0; |
5f256bec EB |
807 | |
808 | out_undo: | |
809 | /* If I have an error cleanup all namespaces I initialized */ | |
810 | list_del(&ops->list); | |
72ad937a EB |
811 | ops_exit_list(ops, &net_exit_list); |
812 | ops_free_list(ops, &net_exit_list); | |
1dba323b | 813 | return error; |
5f256bec EB |
814 | } |
815 | ||
f875bae0 | 816 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
5f256bec EB |
817 | { |
818 | struct net *net; | |
72ad937a | 819 | LIST_HEAD(net_exit_list); |
5f256bec EB |
820 | |
821 | list_del(&ops->list); | |
72ad937a EB |
822 | for_each_net(net) |
823 | list_add_tail(&net->exit_list, &net_exit_list); | |
824 | ops_exit_list(ops, &net_exit_list); | |
825 | ops_free_list(ops, &net_exit_list); | |
5f256bec EB |
826 | } |
827 | ||
ed160e83 DL |
828 | #else |
829 | ||
f875bae0 EB |
830 | static int __register_pernet_operations(struct list_head *list, |
831 | struct pernet_operations *ops) | |
ed160e83 | 832 | { |
f8c46cb3 DT |
833 | if (!init_net_initialized) { |
834 | list_add_tail(&ops->list, list); | |
835 | return 0; | |
836 | } | |
837 | ||
b922934d | 838 | return ops_init(ops, &init_net); |
ed160e83 DL |
839 | } |
840 | ||
f875bae0 | 841 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
ed160e83 | 842 | { |
f8c46cb3 DT |
843 | if (!init_net_initialized) { |
844 | list_del(&ops->list); | |
845 | } else { | |
846 | LIST_HEAD(net_exit_list); | |
847 | list_add(&init_net.exit_list, &net_exit_list); | |
848 | ops_exit_list(ops, &net_exit_list); | |
849 | ops_free_list(ops, &net_exit_list); | |
850 | } | |
ed160e83 | 851 | } |
f875bae0 EB |
852 | |
853 | #endif /* CONFIG_NET_NS */ | |
ed160e83 | 854 | |
c93cf61f PE |
855 | static DEFINE_IDA(net_generic_ids); |
856 | ||
f875bae0 EB |
857 | static int register_pernet_operations(struct list_head *list, |
858 | struct pernet_operations *ops) | |
859 | { | |
860 | int error; | |
861 | ||
862 | if (ops->id) { | |
863 | again: | |
864 | error = ida_get_new_above(&net_generic_ids, 1, ops->id); | |
865 | if (error < 0) { | |
866 | if (error == -EAGAIN) { | |
867 | ida_pre_get(&net_generic_ids, GFP_KERNEL); | |
868 | goto again; | |
869 | } | |
870 | return error; | |
871 | } | |
073862ba | 872 | max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id); |
f875bae0 EB |
873 | } |
874 | error = __register_pernet_operations(list, ops); | |
3a765eda EB |
875 | if (error) { |
876 | rcu_barrier(); | |
877 | if (ops->id) | |
878 | ida_remove(&net_generic_ids, *ops->id); | |
879 | } | |
f875bae0 EB |
880 | |
881 | return error; | |
882 | } | |
883 | ||
884 | static void unregister_pernet_operations(struct pernet_operations *ops) | |
885 | { | |
886 | ||
887 | __unregister_pernet_operations(ops); | |
3a765eda | 888 | rcu_barrier(); |
f875bae0 EB |
889 | if (ops->id) |
890 | ida_remove(&net_generic_ids, *ops->id); | |
891 | } | |
892 | ||
5f256bec EB |
893 | /** |
894 | * register_pernet_subsys - register a network namespace subsystem | |
895 | * @ops: pernet operations structure for the subsystem | |
896 | * | |
897 | * Register a subsystem which has init and exit functions | |
898 | * that are called when network namespaces are created and | |
899 | * destroyed respectively. | |
900 | * | |
901 | * When registered all network namespace init functions are | |
902 | * called for every existing network namespace. Allowing kernel | |
903 | * modules to have a race free view of the set of network namespaces. | |
904 | * | |
905 | * When a new network namespace is created all of the init | |
906 | * methods are called in the order in which they were registered. | |
907 | * | |
908 | * When a network namespace is destroyed all of the exit methods | |
909 | * are called in the reverse of the order with which they were | |
910 | * registered. | |
911 | */ | |
912 | int register_pernet_subsys(struct pernet_operations *ops) | |
913 | { | |
914 | int error; | |
915 | mutex_lock(&net_mutex); | |
916 | error = register_pernet_operations(first_device, ops); | |
917 | mutex_unlock(&net_mutex); | |
918 | return error; | |
919 | } | |
920 | EXPORT_SYMBOL_GPL(register_pernet_subsys); | |
921 | ||
922 | /** | |
923 | * unregister_pernet_subsys - unregister a network namespace subsystem | |
924 | * @ops: pernet operations structure to manipulate | |
925 | * | |
926 | * Remove the pernet operations structure from the list to be | |
53379e57 | 927 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
928 | * addition run the exit method for all existing network |
929 | * namespaces. | |
930 | */ | |
b3c981d2 | 931 | void unregister_pernet_subsys(struct pernet_operations *ops) |
5f256bec EB |
932 | { |
933 | mutex_lock(&net_mutex); | |
b3c981d2 | 934 | unregister_pernet_operations(ops); |
5f256bec EB |
935 | mutex_unlock(&net_mutex); |
936 | } | |
937 | EXPORT_SYMBOL_GPL(unregister_pernet_subsys); | |
938 | ||
939 | /** | |
940 | * register_pernet_device - register a network namespace device | |
941 | * @ops: pernet operations structure for the subsystem | |
942 | * | |
943 | * Register a device which has init and exit functions | |
944 | * that are called when network namespaces are created and | |
945 | * destroyed respectively. | |
946 | * | |
947 | * When registered all network namespace init functions are | |
948 | * called for every existing network namespace. Allowing kernel | |
949 | * modules to have a race free view of the set of network namespaces. | |
950 | * | |
951 | * When a new network namespace is created all of the init | |
952 | * methods are called in the order in which they were registered. | |
953 | * | |
954 | * When a network namespace is destroyed all of the exit methods | |
955 | * are called in the reverse of the order with which they were | |
956 | * registered. | |
957 | */ | |
958 | int register_pernet_device(struct pernet_operations *ops) | |
959 | { | |
960 | int error; | |
961 | mutex_lock(&net_mutex); | |
962 | error = register_pernet_operations(&pernet_list, ops); | |
963 | if (!error && (first_device == &pernet_list)) | |
964 | first_device = &ops->list; | |
965 | mutex_unlock(&net_mutex); | |
966 | return error; | |
967 | } | |
968 | EXPORT_SYMBOL_GPL(register_pernet_device); | |
969 | ||
970 | /** | |
971 | * unregister_pernet_device - unregister a network namespace netdevice | |
972 | * @ops: pernet operations structure to manipulate | |
973 | * | |
974 | * Remove the pernet operations structure from the list to be | |
53379e57 | 975 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
976 | * addition run the exit method for all existing network |
977 | * namespaces. | |
978 | */ | |
979 | void unregister_pernet_device(struct pernet_operations *ops) | |
980 | { | |
981 | mutex_lock(&net_mutex); | |
982 | if (&ops->list == first_device) | |
983 | first_device = first_device->next; | |
984 | unregister_pernet_operations(ops); | |
985 | mutex_unlock(&net_mutex); | |
986 | } | |
987 | EXPORT_SYMBOL_GPL(unregister_pernet_device); | |
13b6f576 EB |
988 | |
989 | #ifdef CONFIG_NET_NS | |
64964528 | 990 | static struct ns_common *netns_get(struct task_struct *task) |
13b6f576 | 991 | { |
f0630529 EB |
992 | struct net *net = NULL; |
993 | struct nsproxy *nsproxy; | |
994 | ||
728dba3a EB |
995 | task_lock(task); |
996 | nsproxy = task->nsproxy; | |
f0630529 EB |
997 | if (nsproxy) |
998 | net = get_net(nsproxy->net_ns); | |
728dba3a | 999 | task_unlock(task); |
f0630529 | 1000 | |
ff24870f AV |
1001 | return net ? &net->ns : NULL; |
1002 | } | |
1003 | ||
1004 | static inline struct net *to_net_ns(struct ns_common *ns) | |
1005 | { | |
1006 | return container_of(ns, struct net, ns); | |
13b6f576 EB |
1007 | } |
1008 | ||
64964528 | 1009 | static void netns_put(struct ns_common *ns) |
13b6f576 | 1010 | { |
ff24870f | 1011 | put_net(to_net_ns(ns)); |
13b6f576 EB |
1012 | } |
1013 | ||
64964528 | 1014 | static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns) |
13b6f576 | 1015 | { |
ff24870f | 1016 | struct net *net = to_net_ns(ns); |
142e1d1d | 1017 | |
5e4a0847 | 1018 | if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) || |
c7b96acf | 1019 | !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) |
142e1d1d EB |
1020 | return -EPERM; |
1021 | ||
13b6f576 | 1022 | put_net(nsproxy->net_ns); |
142e1d1d | 1023 | nsproxy->net_ns = get_net(net); |
13b6f576 EB |
1024 | return 0; |
1025 | } | |
1026 | ||
bcac25a5 AV |
1027 | static struct user_namespace *netns_owner(struct ns_common *ns) |
1028 | { | |
1029 | return to_net_ns(ns)->user_ns; | |
1030 | } | |
1031 | ||
13b6f576 EB |
1032 | const struct proc_ns_operations netns_operations = { |
1033 | .name = "net", | |
1034 | .type = CLONE_NEWNET, | |
1035 | .get = netns_get, | |
1036 | .put = netns_put, | |
1037 | .install = netns_install, | |
bcac25a5 | 1038 | .owner = netns_owner, |
13b6f576 EB |
1039 | }; |
1040 | #endif |