]>
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 | ||
0c7aecd4 ND |
150 | static int alloc_netid(struct net *net, struct net *peer, int reqid) |
151 | { | |
3138dbf8 | 152 | int min = 0, max = 0; |
0c7aecd4 ND |
153 | |
154 | ASSERT_RTNL(); | |
155 | ||
156 | if (reqid >= 0) { | |
157 | min = reqid; | |
158 | max = reqid + 1; | |
159 | } | |
160 | ||
3138dbf8 | 161 | return idr_alloc(&net->netns_ids, peer, min, max, GFP_KERNEL); |
0c7aecd4 ND |
162 | } |
163 | ||
164 | /* This function is used by idr_for_each(). If net is equal to peer, the | |
165 | * function returns the id so that idr_for_each() stops. Because we cannot | |
166 | * returns the id 0 (idr_for_each() will not stop), we return the magic value | |
167 | * NET_ID_ZERO (-1) for it. | |
168 | */ | |
169 | #define NET_ID_ZERO -1 | |
170 | static int net_eq_idr(int id, void *net, void *peer) | |
171 | { | |
172 | if (net_eq(net, peer)) | |
173 | return id ? : NET_ID_ZERO; | |
174 | return 0; | |
175 | } | |
176 | ||
3138dbf8 | 177 | static int __peernet2id_alloc(struct net *net, struct net *peer, bool *alloc) |
0c7aecd4 ND |
178 | { |
179 | int id = idr_for_each(&net->netns_ids, net_eq_idr, peer); | |
3138dbf8 | 180 | bool alloc_it = *alloc; |
0c7aecd4 ND |
181 | |
182 | ASSERT_RTNL(); | |
183 | ||
3138dbf8 ND |
184 | *alloc = false; |
185 | ||
0c7aecd4 ND |
186 | /* Magic value for id 0. */ |
187 | if (id == NET_ID_ZERO) | |
188 | return 0; | |
189 | if (id > 0) | |
190 | return id; | |
191 | ||
3138dbf8 | 192 | if (alloc_it) { |
109582af | 193 | id = alloc_netid(net, peer, -1); |
3138dbf8 | 194 | *alloc = true; |
109582af ND |
195 | return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED; |
196 | } | |
0c7aecd4 | 197 | |
109582af | 198 | return NETNSA_NSID_NOT_ASSIGNED; |
0c7aecd4 ND |
199 | } |
200 | ||
3138dbf8 ND |
201 | static int __peernet2id(struct net *net, struct net *peer) |
202 | { | |
203 | bool no = false; | |
204 | ||
205 | return __peernet2id_alloc(net, peer, &no); | |
206 | } | |
207 | ||
208 | static void rtnl_net_notifyid(struct net *net, int cmd, int id); | |
0c7aecd4 ND |
209 | /* This function returns the id of a peer netns. If no id is assigned, one will |
210 | * be allocated and returned. | |
211 | */ | |
7a0877d4 | 212 | int peernet2id_alloc(struct net *net, struct net *peer) |
0c7aecd4 | 213 | { |
576b7cd2 | 214 | bool alloc = atomic_read(&peer->count) == 0 ? false : true; |
3138dbf8 | 215 | int id; |
0c7aecd4 | 216 | |
3138dbf8 ND |
217 | id = __peernet2id_alloc(net, peer, &alloc); |
218 | if (alloc && id >= 0) | |
219 | rtnl_net_notifyid(net, RTM_NEWNSID, id); | |
220 | return id; | |
0c7aecd4 | 221 | } |
7a0877d4 | 222 | EXPORT_SYMBOL(peernet2id_alloc); |
0c7aecd4 ND |
223 | |
224 | struct net *get_net_ns_by_id(struct net *net, int id) | |
225 | { | |
226 | struct net *peer; | |
227 | ||
228 | if (id < 0) | |
229 | return NULL; | |
230 | ||
231 | rcu_read_lock(); | |
232 | peer = idr_find(&net->netns_ids, id); | |
233 | if (peer) | |
234 | get_net(peer); | |
235 | rcu_read_unlock(); | |
236 | ||
237 | return peer; | |
238 | } | |
239 | ||
5f256bec EB |
240 | /* |
241 | * setup_net runs the initializers for the network namespace object. | |
242 | */ | |
038e7332 | 243 | static __net_init int setup_net(struct net *net, struct user_namespace *user_ns) |
5f256bec EB |
244 | { |
245 | /* Must be called with net_mutex held */ | |
f875bae0 | 246 | const struct pernet_operations *ops, *saved_ops; |
486a87f1 | 247 | int error = 0; |
72ad937a | 248 | LIST_HEAD(net_exit_list); |
5f256bec | 249 | |
5f256bec | 250 | atomic_set(&net->count, 1); |
a685e089 | 251 | atomic_set(&net->passive, 1); |
4e985ada | 252 | net->dev_base_seq = 1; |
038e7332 | 253 | net->user_ns = user_ns; |
0c7aecd4 | 254 | idr_init(&net->netns_ids); |
486a87f1 | 255 | |
768f3591 | 256 | list_for_each_entry(ops, &pernet_list, list) { |
f875bae0 EB |
257 | error = ops_init(ops, net); |
258 | if (error < 0) | |
259 | goto out_undo; | |
5f256bec EB |
260 | } |
261 | out: | |
262 | return error; | |
768f3591 | 263 | |
5f256bec EB |
264 | out_undo: |
265 | /* Walk through the list backwards calling the exit functions | |
266 | * for the pernet modules whose init functions did not fail. | |
267 | */ | |
72ad937a | 268 | list_add(&net->exit_list, &net_exit_list); |
f875bae0 | 269 | saved_ops = ops; |
72ad937a EB |
270 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) |
271 | ops_exit_list(ops, &net_exit_list); | |
272 | ||
f875bae0 EB |
273 | ops = saved_ops; |
274 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) | |
72ad937a | 275 | ops_free_list(ops, &net_exit_list); |
310928d9 DL |
276 | |
277 | rcu_barrier(); | |
5f256bec EB |
278 | goto out; |
279 | } | |
280 | ||
6a1a3b9f | 281 | |
ebe47d47 CN |
282 | #ifdef CONFIG_NET_NS |
283 | static struct kmem_cache *net_cachep; | |
284 | static struct workqueue_struct *netns_wq; | |
285 | ||
486a87f1 | 286 | static struct net *net_alloc(void) |
45a19b0a | 287 | { |
486a87f1 DL |
288 | struct net *net = NULL; |
289 | struct net_generic *ng; | |
290 | ||
291 | ng = net_alloc_generic(); | |
292 | if (!ng) | |
293 | goto out; | |
294 | ||
295 | net = kmem_cache_zalloc(net_cachep, GFP_KERNEL); | |
45a19b0a | 296 | if (!net) |
486a87f1 | 297 | goto out_free; |
45a19b0a | 298 | |
486a87f1 DL |
299 | rcu_assign_pointer(net->gen, ng); |
300 | out: | |
301 | return net; | |
302 | ||
303 | out_free: | |
304 | kfree(ng); | |
305 | goto out; | |
306 | } | |
307 | ||
308 | static void net_free(struct net *net) | |
309 | { | |
416c51e1 | 310 | kfree(rcu_access_pointer(net->gen)); |
45a19b0a JFS |
311 | kmem_cache_free(net_cachep, net); |
312 | } | |
313 | ||
a685e089 AV |
314 | void net_drop_ns(void *p) |
315 | { | |
316 | struct net *ns = p; | |
317 | if (ns && atomic_dec_and_test(&ns->passive)) | |
318 | net_free(ns); | |
319 | } | |
320 | ||
038e7332 EB |
321 | struct net *copy_net_ns(unsigned long flags, |
322 | struct user_namespace *user_ns, struct net *old_net) | |
9dd776b6 | 323 | { |
088eb2d9 AD |
324 | struct net *net; |
325 | int rv; | |
9dd776b6 | 326 | |
911cb193 RL |
327 | if (!(flags & CLONE_NEWNET)) |
328 | return get_net(old_net); | |
329 | ||
088eb2d9 AD |
330 | net = net_alloc(); |
331 | if (!net) | |
332 | return ERR_PTR(-ENOMEM); | |
038e7332 EB |
333 | |
334 | get_user_ns(user_ns); | |
335 | ||
9dd776b6 | 336 | mutex_lock(&net_mutex); |
038e7332 | 337 | rv = setup_net(net, user_ns); |
088eb2d9 | 338 | if (rv == 0) { |
486a87f1 | 339 | rtnl_lock(); |
11a28d37 | 340 | list_add_tail_rcu(&net->list, &net_namespace_list); |
486a87f1 DL |
341 | rtnl_unlock(); |
342 | } | |
9dd776b6 | 343 | mutex_unlock(&net_mutex); |
088eb2d9 | 344 | if (rv < 0) { |
038e7332 | 345 | put_user_ns(user_ns); |
a685e089 | 346 | net_drop_ns(net); |
088eb2d9 AD |
347 | return ERR_PTR(rv); |
348 | } | |
349 | return net; | |
350 | } | |
486a87f1 | 351 | |
2b035b39 EB |
352 | static DEFINE_SPINLOCK(cleanup_list_lock); |
353 | static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */ | |
354 | ||
6a1a3b9f PE |
355 | static void cleanup_net(struct work_struct *work) |
356 | { | |
f875bae0 | 357 | const struct pernet_operations *ops; |
6d458f5b | 358 | struct net *net, *tmp; |
1818ce4d | 359 | struct list_head net_kill_list; |
72ad937a | 360 | LIST_HEAD(net_exit_list); |
6a1a3b9f | 361 | |
2b035b39 EB |
362 | /* Atomically snapshot the list of namespaces to cleanup */ |
363 | spin_lock_irq(&cleanup_list_lock); | |
364 | list_replace_init(&cleanup_list, &net_kill_list); | |
365 | spin_unlock_irq(&cleanup_list_lock); | |
6a1a3b9f PE |
366 | |
367 | mutex_lock(&net_mutex); | |
368 | ||
369 | /* Don't let anyone else find us. */ | |
370 | rtnl_lock(); | |
72ad937a | 371 | list_for_each_entry(net, &net_kill_list, cleanup_list) { |
2b035b39 | 372 | list_del_rcu(&net->list); |
72ad937a | 373 | list_add_tail(&net->exit_list, &net_exit_list); |
6d458f5b | 374 | for_each_net(tmp) { |
3138dbf8 | 375 | int id = __peernet2id(tmp, net); |
6d458f5b | 376 | |
9a963454 | 377 | if (id >= 0) { |
cab3c8ec | 378 | rtnl_net_notifyid(tmp, RTM_DELNSID, id); |
6d458f5b | 379 | idr_remove(&tmp->netns_ids, id); |
9a963454 | 380 | } |
6d458f5b ND |
381 | } |
382 | idr_destroy(&net->netns_ids); | |
383 | ||
72ad937a | 384 | } |
6a1a3b9f PE |
385 | rtnl_unlock(); |
386 | ||
11a28d37 JB |
387 | /* |
388 | * Another CPU might be rcu-iterating the list, wait for it. | |
389 | * This needs to be before calling the exit() notifiers, so | |
390 | * the rcu_barrier() below isn't sufficient alone. | |
391 | */ | |
392 | synchronize_rcu(); | |
393 | ||
6a1a3b9f | 394 | /* Run all of the network namespace exit methods */ |
72ad937a EB |
395 | list_for_each_entry_reverse(ops, &pernet_list, list) |
396 | ops_exit_list(ops, &net_exit_list); | |
397 | ||
f875bae0 | 398 | /* Free the net generic variables */ |
72ad937a EB |
399 | list_for_each_entry_reverse(ops, &pernet_list, list) |
400 | ops_free_list(ops, &net_exit_list); | |
6a1a3b9f PE |
401 | |
402 | mutex_unlock(&net_mutex); | |
403 | ||
404 | /* Ensure there are no outstanding rcu callbacks using this | |
405 | * network namespace. | |
406 | */ | |
407 | rcu_barrier(); | |
408 | ||
409 | /* Finally it is safe to free my network namespace structure */ | |
72ad937a EB |
410 | list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) { |
411 | list_del_init(&net->exit_list); | |
038e7332 | 412 | put_user_ns(net->user_ns); |
a685e089 | 413 | net_drop_ns(net); |
2b035b39 | 414 | } |
6a1a3b9f | 415 | } |
2b035b39 | 416 | static DECLARE_WORK(net_cleanup_work, cleanup_net); |
6a1a3b9f PE |
417 | |
418 | void __put_net(struct net *net) | |
419 | { | |
420 | /* Cleanup the network namespace in process context */ | |
2b035b39 EB |
421 | unsigned long flags; |
422 | ||
423 | spin_lock_irqsave(&cleanup_list_lock, flags); | |
424 | list_add(&net->cleanup_list, &cleanup_list); | |
425 | spin_unlock_irqrestore(&cleanup_list_lock, flags); | |
426 | ||
427 | queue_work(netns_wq, &net_cleanup_work); | |
6a1a3b9f PE |
428 | } |
429 | EXPORT_SYMBOL_GPL(__put_net); | |
430 | ||
956c9207 SR |
431 | struct net *get_net_ns_by_fd(int fd) |
432 | { | |
956c9207 | 433 | struct file *file; |
33c42940 | 434 | struct ns_common *ns; |
956c9207 SR |
435 | struct net *net; |
436 | ||
956c9207 | 437 | file = proc_ns_fget(fd); |
c316e6a3 AV |
438 | if (IS_ERR(file)) |
439 | return ERR_CAST(file); | |
956c9207 | 440 | |
f77c8014 | 441 | ns = get_proc_ns(file_inode(file)); |
33c42940 AV |
442 | if (ns->ops == &netns_operations) |
443 | net = get_net(container_of(ns, struct net, ns)); | |
c316e6a3 AV |
444 | else |
445 | net = ERR_PTR(-EINVAL); | |
956c9207 | 446 | |
c316e6a3 | 447 | fput(file); |
956c9207 SR |
448 | return net; |
449 | } | |
450 | ||
6a1a3b9f | 451 | #else |
956c9207 SR |
452 | struct net *get_net_ns_by_fd(int fd) |
453 | { | |
454 | return ERR_PTR(-EINVAL); | |
455 | } | |
6a1a3b9f | 456 | #endif |
4b681c82 | 457 | EXPORT_SYMBOL_GPL(get_net_ns_by_fd); |
6a1a3b9f | 458 | |
30ffee84 JB |
459 | struct net *get_net_ns_by_pid(pid_t pid) |
460 | { | |
461 | struct task_struct *tsk; | |
462 | struct net *net; | |
463 | ||
464 | /* Lookup the network namespace */ | |
465 | net = ERR_PTR(-ESRCH); | |
466 | rcu_read_lock(); | |
467 | tsk = find_task_by_vpid(pid); | |
468 | if (tsk) { | |
469 | struct nsproxy *nsproxy; | |
728dba3a EB |
470 | task_lock(tsk); |
471 | nsproxy = tsk->nsproxy; | |
30ffee84 JB |
472 | if (nsproxy) |
473 | net = get_net(nsproxy->net_ns); | |
728dba3a | 474 | task_unlock(tsk); |
30ffee84 JB |
475 | } |
476 | rcu_read_unlock(); | |
477 | return net; | |
478 | } | |
479 | EXPORT_SYMBOL_GPL(get_net_ns_by_pid); | |
480 | ||
98f842e6 EB |
481 | static __net_init int net_ns_net_init(struct net *net) |
482 | { | |
33c42940 AV |
483 | #ifdef CONFIG_NET_NS |
484 | net->ns.ops = &netns_operations; | |
485 | #endif | |
6344c433 | 486 | return ns_alloc_inum(&net->ns); |
98f842e6 EB |
487 | } |
488 | ||
489 | static __net_exit void net_ns_net_exit(struct net *net) | |
490 | { | |
6344c433 | 491 | ns_free_inum(&net->ns); |
98f842e6 EB |
492 | } |
493 | ||
494 | static struct pernet_operations __net_initdata net_ns_ops = { | |
495 | .init = net_ns_net_init, | |
496 | .exit = net_ns_net_exit, | |
497 | }; | |
498 | ||
0c7aecd4 ND |
499 | static struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = { |
500 | [NETNSA_NONE] = { .type = NLA_UNSPEC }, | |
501 | [NETNSA_NSID] = { .type = NLA_S32 }, | |
502 | [NETNSA_PID] = { .type = NLA_U32 }, | |
503 | [NETNSA_FD] = { .type = NLA_U32 }, | |
504 | }; | |
505 | ||
506 | static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
507 | { | |
508 | struct net *net = sock_net(skb->sk); | |
509 | struct nlattr *tb[NETNSA_MAX + 1]; | |
510 | struct net *peer; | |
511 | int nsid, err; | |
512 | ||
513 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
514 | rtnl_net_policy); | |
515 | if (err < 0) | |
516 | return err; | |
517 | if (!tb[NETNSA_NSID]) | |
518 | return -EINVAL; | |
519 | nsid = nla_get_s32(tb[NETNSA_NSID]); | |
520 | ||
521 | if (tb[NETNSA_PID]) | |
522 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); | |
523 | else if (tb[NETNSA_FD]) | |
524 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); | |
525 | else | |
526 | return -EINVAL; | |
527 | if (IS_ERR(peer)) | |
528 | return PTR_ERR(peer); | |
529 | ||
3138dbf8 | 530 | if (__peernet2id(net, peer) >= 0) { |
0c7aecd4 ND |
531 | err = -EEXIST; |
532 | goto out; | |
533 | } | |
534 | ||
535 | err = alloc_netid(net, peer, nsid); | |
3138dbf8 ND |
536 | if (err >= 0) { |
537 | rtnl_net_notifyid(net, RTM_NEWNSID, err); | |
0c7aecd4 | 538 | err = 0; |
3138dbf8 | 539 | } |
0c7aecd4 ND |
540 | out: |
541 | put_net(peer); | |
542 | return err; | |
543 | } | |
544 | ||
545 | static int rtnl_net_get_size(void) | |
546 | { | |
547 | return NLMSG_ALIGN(sizeof(struct rtgenmsg)) | |
548 | + nla_total_size(sizeof(s32)) /* NETNSA_NSID */ | |
549 | ; | |
550 | } | |
551 | ||
552 | static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags, | |
cab3c8ec | 553 | int cmd, struct net *net, int nsid) |
0c7aecd4 ND |
554 | { |
555 | struct nlmsghdr *nlh; | |
556 | struct rtgenmsg *rth; | |
0c7aecd4 ND |
557 | |
558 | nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags); | |
559 | if (!nlh) | |
560 | return -EMSGSIZE; | |
561 | ||
562 | rth = nlmsg_data(nlh); | |
563 | rth->rtgen_family = AF_UNSPEC; | |
564 | ||
cab3c8ec | 565 | if (nla_put_s32(skb, NETNSA_NSID, nsid)) |
0c7aecd4 ND |
566 | goto nla_put_failure; |
567 | ||
568 | nlmsg_end(skb, nlh); | |
569 | return 0; | |
570 | ||
571 | nla_put_failure: | |
572 | nlmsg_cancel(skb, nlh); | |
573 | return -EMSGSIZE; | |
574 | } | |
575 | ||
576 | static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
577 | { | |
578 | struct net *net = sock_net(skb->sk); | |
579 | struct nlattr *tb[NETNSA_MAX + 1]; | |
580 | struct sk_buff *msg; | |
0c7aecd4 | 581 | struct net *peer; |
cab3c8ec | 582 | int err, id; |
0c7aecd4 ND |
583 | |
584 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
585 | rtnl_net_policy); | |
586 | if (err < 0) | |
587 | return err; | |
588 | if (tb[NETNSA_PID]) | |
589 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); | |
590 | else if (tb[NETNSA_FD]) | |
591 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); | |
592 | else | |
593 | return -EINVAL; | |
594 | ||
595 | if (IS_ERR(peer)) | |
596 | return PTR_ERR(peer); | |
597 | ||
598 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
599 | if (!msg) { | |
600 | err = -ENOMEM; | |
601 | goto out; | |
602 | } | |
603 | ||
3138dbf8 | 604 | id = __peernet2id(net, peer); |
0c7aecd4 | 605 | err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, |
cab3c8ec | 606 | RTM_GETNSID, net, id); |
0c7aecd4 ND |
607 | if (err < 0) |
608 | goto err_out; | |
609 | ||
610 | err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid); | |
611 | goto out; | |
612 | ||
613 | err_out: | |
614 | nlmsg_free(msg); | |
615 | out: | |
616 | put_net(peer); | |
617 | return err; | |
618 | } | |
619 | ||
a143c40c ND |
620 | struct rtnl_net_dump_cb { |
621 | struct net *net; | |
622 | struct sk_buff *skb; | |
623 | struct netlink_callback *cb; | |
624 | int idx; | |
625 | int s_idx; | |
626 | }; | |
627 | ||
628 | static int rtnl_net_dumpid_one(int id, void *peer, void *data) | |
629 | { | |
630 | struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data; | |
631 | int ret; | |
632 | ||
633 | if (net_cb->idx < net_cb->s_idx) | |
634 | goto cont; | |
635 | ||
636 | ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid, | |
637 | net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI, | |
cab3c8ec | 638 | RTM_NEWNSID, net_cb->net, id); |
a143c40c ND |
639 | if (ret < 0) |
640 | return ret; | |
641 | ||
642 | cont: | |
643 | net_cb->idx++; | |
644 | return 0; | |
645 | } | |
646 | ||
647 | static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb) | |
648 | { | |
649 | struct net *net = sock_net(skb->sk); | |
650 | struct rtnl_net_dump_cb net_cb = { | |
651 | .net = net, | |
652 | .skb = skb, | |
653 | .cb = cb, | |
654 | .idx = 0, | |
655 | .s_idx = cb->args[0], | |
656 | }; | |
657 | ||
658 | ASSERT_RTNL(); | |
659 | ||
660 | idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb); | |
661 | ||
662 | cb->args[0] = net_cb.idx; | |
663 | return skb->len; | |
664 | } | |
665 | ||
cab3c8ec | 666 | static void rtnl_net_notifyid(struct net *net, int cmd, int id) |
9a963454 ND |
667 | { |
668 | struct sk_buff *msg; | |
669 | int err = -ENOMEM; | |
670 | ||
671 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
672 | if (!msg) | |
673 | goto out; | |
674 | ||
cab3c8ec | 675 | err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, id); |
9a963454 ND |
676 | if (err < 0) |
677 | goto err_out; | |
678 | ||
679 | rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0); | |
680 | return; | |
681 | ||
682 | err_out: | |
683 | nlmsg_free(msg); | |
684 | out: | |
685 | rtnl_set_sk_err(net, RTNLGRP_NSID, err); | |
686 | } | |
687 | ||
5f256bec EB |
688 | static int __init net_ns_init(void) |
689 | { | |
486a87f1 | 690 | struct net_generic *ng; |
5f256bec | 691 | |
d57a9212 | 692 | #ifdef CONFIG_NET_NS |
5f256bec EB |
693 | net_cachep = kmem_cache_create("net_namespace", sizeof(struct net), |
694 | SMP_CACHE_BYTES, | |
695 | SLAB_PANIC, NULL); | |
3ef1355d BT |
696 | |
697 | /* Create workqueue for cleanup */ | |
698 | netns_wq = create_singlethread_workqueue("netns"); | |
699 | if (!netns_wq) | |
700 | panic("Could not create netns workq"); | |
d57a9212 | 701 | #endif |
3ef1355d | 702 | |
486a87f1 DL |
703 | ng = net_alloc_generic(); |
704 | if (!ng) | |
705 | panic("Could not allocate generic netns"); | |
706 | ||
707 | rcu_assign_pointer(init_net.gen, ng); | |
708 | ||
5f256bec | 709 | mutex_lock(&net_mutex); |
038e7332 | 710 | if (setup_net(&init_net, &init_user_ns)) |
ca0f3112 | 711 | panic("Could not setup the initial network namespace"); |
5f256bec | 712 | |
f4618d39 | 713 | rtnl_lock(); |
11a28d37 | 714 | list_add_tail_rcu(&init_net.list, &net_namespace_list); |
f4618d39 | 715 | rtnl_unlock(); |
5f256bec EB |
716 | |
717 | mutex_unlock(&net_mutex); | |
5f256bec | 718 | |
98f842e6 EB |
719 | register_pernet_subsys(&net_ns_ops); |
720 | ||
0c7aecd4 | 721 | rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL); |
a143c40c ND |
722 | rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid, |
723 | NULL); | |
0c7aecd4 | 724 | |
5f256bec EB |
725 | return 0; |
726 | } | |
727 | ||
728 | pure_initcall(net_ns_init); | |
729 | ||
ed160e83 | 730 | #ifdef CONFIG_NET_NS |
f875bae0 EB |
731 | static int __register_pernet_operations(struct list_head *list, |
732 | struct pernet_operations *ops) | |
5f256bec | 733 | { |
72ad937a | 734 | struct net *net; |
5f256bec | 735 | int error; |
72ad937a | 736 | LIST_HEAD(net_exit_list); |
5f256bec | 737 | |
5f256bec | 738 | list_add_tail(&ops->list, list); |
f875bae0 | 739 | if (ops->init || (ops->id && ops->size)) { |
1dba323b | 740 | for_each_net(net) { |
f875bae0 | 741 | error = ops_init(ops, net); |
5f256bec EB |
742 | if (error) |
743 | goto out_undo; | |
72ad937a | 744 | list_add_tail(&net->exit_list, &net_exit_list); |
5f256bec EB |
745 | } |
746 | } | |
1dba323b | 747 | return 0; |
5f256bec EB |
748 | |
749 | out_undo: | |
750 | /* If I have an error cleanup all namespaces I initialized */ | |
751 | list_del(&ops->list); | |
72ad937a EB |
752 | ops_exit_list(ops, &net_exit_list); |
753 | ops_free_list(ops, &net_exit_list); | |
1dba323b | 754 | return error; |
5f256bec EB |
755 | } |
756 | ||
f875bae0 | 757 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
5f256bec EB |
758 | { |
759 | struct net *net; | |
72ad937a | 760 | LIST_HEAD(net_exit_list); |
5f256bec EB |
761 | |
762 | list_del(&ops->list); | |
72ad937a EB |
763 | for_each_net(net) |
764 | list_add_tail(&net->exit_list, &net_exit_list); | |
765 | ops_exit_list(ops, &net_exit_list); | |
766 | ops_free_list(ops, &net_exit_list); | |
5f256bec EB |
767 | } |
768 | ||
ed160e83 DL |
769 | #else |
770 | ||
f875bae0 EB |
771 | static int __register_pernet_operations(struct list_head *list, |
772 | struct pernet_operations *ops) | |
ed160e83 | 773 | { |
b922934d | 774 | return ops_init(ops, &init_net); |
ed160e83 DL |
775 | } |
776 | ||
f875bae0 | 777 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
ed160e83 | 778 | { |
72ad937a EB |
779 | LIST_HEAD(net_exit_list); |
780 | list_add(&init_net.exit_list, &net_exit_list); | |
781 | ops_exit_list(ops, &net_exit_list); | |
782 | ops_free_list(ops, &net_exit_list); | |
ed160e83 | 783 | } |
f875bae0 EB |
784 | |
785 | #endif /* CONFIG_NET_NS */ | |
ed160e83 | 786 | |
c93cf61f PE |
787 | static DEFINE_IDA(net_generic_ids); |
788 | ||
f875bae0 EB |
789 | static int register_pernet_operations(struct list_head *list, |
790 | struct pernet_operations *ops) | |
791 | { | |
792 | int error; | |
793 | ||
794 | if (ops->id) { | |
795 | again: | |
796 | error = ida_get_new_above(&net_generic_ids, 1, ops->id); | |
797 | if (error < 0) { | |
798 | if (error == -EAGAIN) { | |
799 | ida_pre_get(&net_generic_ids, GFP_KERNEL); | |
800 | goto again; | |
801 | } | |
802 | return error; | |
803 | } | |
073862ba | 804 | max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id); |
f875bae0 EB |
805 | } |
806 | error = __register_pernet_operations(list, ops); | |
3a765eda EB |
807 | if (error) { |
808 | rcu_barrier(); | |
809 | if (ops->id) | |
810 | ida_remove(&net_generic_ids, *ops->id); | |
811 | } | |
f875bae0 EB |
812 | |
813 | return error; | |
814 | } | |
815 | ||
816 | static void unregister_pernet_operations(struct pernet_operations *ops) | |
817 | { | |
818 | ||
819 | __unregister_pernet_operations(ops); | |
3a765eda | 820 | rcu_barrier(); |
f875bae0 EB |
821 | if (ops->id) |
822 | ida_remove(&net_generic_ids, *ops->id); | |
823 | } | |
824 | ||
5f256bec EB |
825 | /** |
826 | * register_pernet_subsys - register a network namespace subsystem | |
827 | * @ops: pernet operations structure for the subsystem | |
828 | * | |
829 | * Register a subsystem which has init and exit functions | |
830 | * that are called when network namespaces are created and | |
831 | * destroyed respectively. | |
832 | * | |
833 | * When registered all network namespace init functions are | |
834 | * called for every existing network namespace. Allowing kernel | |
835 | * modules to have a race free view of the set of network namespaces. | |
836 | * | |
837 | * When a new network namespace is created all of the init | |
838 | * methods are called in the order in which they were registered. | |
839 | * | |
840 | * When a network namespace is destroyed all of the exit methods | |
841 | * are called in the reverse of the order with which they were | |
842 | * registered. | |
843 | */ | |
844 | int register_pernet_subsys(struct pernet_operations *ops) | |
845 | { | |
846 | int error; | |
847 | mutex_lock(&net_mutex); | |
848 | error = register_pernet_operations(first_device, ops); | |
849 | mutex_unlock(&net_mutex); | |
850 | return error; | |
851 | } | |
852 | EXPORT_SYMBOL_GPL(register_pernet_subsys); | |
853 | ||
854 | /** | |
855 | * unregister_pernet_subsys - unregister a network namespace subsystem | |
856 | * @ops: pernet operations structure to manipulate | |
857 | * | |
858 | * Remove the pernet operations structure from the list to be | |
53379e57 | 859 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
860 | * addition run the exit method for all existing network |
861 | * namespaces. | |
862 | */ | |
b3c981d2 | 863 | void unregister_pernet_subsys(struct pernet_operations *ops) |
5f256bec EB |
864 | { |
865 | mutex_lock(&net_mutex); | |
b3c981d2 | 866 | unregister_pernet_operations(ops); |
5f256bec EB |
867 | mutex_unlock(&net_mutex); |
868 | } | |
869 | EXPORT_SYMBOL_GPL(unregister_pernet_subsys); | |
870 | ||
871 | /** | |
872 | * register_pernet_device - register a network namespace device | |
873 | * @ops: pernet operations structure for the subsystem | |
874 | * | |
875 | * Register a device which has init and exit functions | |
876 | * that are called when network namespaces are created and | |
877 | * destroyed respectively. | |
878 | * | |
879 | * When registered all network namespace init functions are | |
880 | * called for every existing network namespace. Allowing kernel | |
881 | * modules to have a race free view of the set of network namespaces. | |
882 | * | |
883 | * When a new network namespace is created all of the init | |
884 | * methods are called in the order in which they were registered. | |
885 | * | |
886 | * When a network namespace is destroyed all of the exit methods | |
887 | * are called in the reverse of the order with which they were | |
888 | * registered. | |
889 | */ | |
890 | int register_pernet_device(struct pernet_operations *ops) | |
891 | { | |
892 | int error; | |
893 | mutex_lock(&net_mutex); | |
894 | error = register_pernet_operations(&pernet_list, ops); | |
895 | if (!error && (first_device == &pernet_list)) | |
896 | first_device = &ops->list; | |
897 | mutex_unlock(&net_mutex); | |
898 | return error; | |
899 | } | |
900 | EXPORT_SYMBOL_GPL(register_pernet_device); | |
901 | ||
902 | /** | |
903 | * unregister_pernet_device - unregister a network namespace netdevice | |
904 | * @ops: pernet operations structure to manipulate | |
905 | * | |
906 | * Remove the pernet operations structure from the list to be | |
53379e57 | 907 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
908 | * addition run the exit method for all existing network |
909 | * namespaces. | |
910 | */ | |
911 | void unregister_pernet_device(struct pernet_operations *ops) | |
912 | { | |
913 | mutex_lock(&net_mutex); | |
914 | if (&ops->list == first_device) | |
915 | first_device = first_device->next; | |
916 | unregister_pernet_operations(ops); | |
917 | mutex_unlock(&net_mutex); | |
918 | } | |
919 | EXPORT_SYMBOL_GPL(unregister_pernet_device); | |
13b6f576 EB |
920 | |
921 | #ifdef CONFIG_NET_NS | |
64964528 | 922 | static struct ns_common *netns_get(struct task_struct *task) |
13b6f576 | 923 | { |
f0630529 EB |
924 | struct net *net = NULL; |
925 | struct nsproxy *nsproxy; | |
926 | ||
728dba3a EB |
927 | task_lock(task); |
928 | nsproxy = task->nsproxy; | |
f0630529 EB |
929 | if (nsproxy) |
930 | net = get_net(nsproxy->net_ns); | |
728dba3a | 931 | task_unlock(task); |
f0630529 | 932 | |
ff24870f AV |
933 | return net ? &net->ns : NULL; |
934 | } | |
935 | ||
936 | static inline struct net *to_net_ns(struct ns_common *ns) | |
937 | { | |
938 | return container_of(ns, struct net, ns); | |
13b6f576 EB |
939 | } |
940 | ||
64964528 | 941 | static void netns_put(struct ns_common *ns) |
13b6f576 | 942 | { |
ff24870f | 943 | put_net(to_net_ns(ns)); |
13b6f576 EB |
944 | } |
945 | ||
64964528 | 946 | static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns) |
13b6f576 | 947 | { |
ff24870f | 948 | struct net *net = to_net_ns(ns); |
142e1d1d | 949 | |
5e4a0847 | 950 | if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) || |
c7b96acf | 951 | !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) |
142e1d1d EB |
952 | return -EPERM; |
953 | ||
13b6f576 | 954 | put_net(nsproxy->net_ns); |
142e1d1d | 955 | nsproxy->net_ns = get_net(net); |
13b6f576 EB |
956 | return 0; |
957 | } | |
958 | ||
959 | const struct proc_ns_operations netns_operations = { | |
960 | .name = "net", | |
961 | .type = CLONE_NEWNET, | |
962 | .get = netns_get, | |
963 | .put = netns_put, | |
964 | .install = netns_install, | |
965 | }; | |
966 | #endif |