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 ND |
18 | #include <linux/net_namespace.h> |
19 | #include <linux/rtnetlink.h> | |
20 | #include <net/sock.h> | |
21 | #include <net/netlink.h> | |
5f256bec | 22 | #include <net/net_namespace.h> |
dec827d1 | 23 | #include <net/netns/generic.h> |
5f256bec EB |
24 | |
25 | /* | |
26 | * Our network namespace constructor/destructor lists | |
27 | */ | |
28 | ||
29 | static LIST_HEAD(pernet_list); | |
30 | static struct list_head *first_device = &pernet_list; | |
200b916f | 31 | DEFINE_MUTEX(net_mutex); |
5f256bec | 32 | |
5f256bec | 33 | LIST_HEAD(net_namespace_list); |
b76a461f | 34 | EXPORT_SYMBOL_GPL(net_namespace_list); |
5f256bec | 35 | |
734b6541 RM |
36 | struct net init_net = { |
37 | .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head), | |
38 | }; | |
ff4b9502 | 39 | EXPORT_SYMBOL(init_net); |
5f256bec | 40 | |
dec827d1 PE |
41 | #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */ |
42 | ||
073862ba ED |
43 | static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS; |
44 | ||
45 | static struct net_generic *net_alloc_generic(void) | |
46 | { | |
47 | struct net_generic *ng; | |
48 | size_t generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]); | |
49 | ||
50 | ng = kzalloc(generic_size, GFP_KERNEL); | |
51 | if (ng) | |
52 | ng->len = max_gen_ptrs; | |
53 | ||
54 | return ng; | |
55 | } | |
56 | ||
05fceb4a JP |
57 | static int net_assign_generic(struct net *net, int id, void *data) |
58 | { | |
59 | struct net_generic *ng, *old_ng; | |
60 | ||
61 | BUG_ON(!mutex_is_locked(&net_mutex)); | |
62 | BUG_ON(id == 0); | |
63 | ||
1c87733d ED |
64 | old_ng = rcu_dereference_protected(net->gen, |
65 | lockdep_is_held(&net_mutex)); | |
66 | ng = old_ng; | |
05fceb4a JP |
67 | if (old_ng->len >= id) |
68 | goto assign; | |
69 | ||
073862ba | 70 | ng = net_alloc_generic(); |
05fceb4a JP |
71 | if (ng == NULL) |
72 | return -ENOMEM; | |
73 | ||
74 | /* | |
75 | * Some synchronisation notes: | |
76 | * | |
77 | * The net_generic explores the net->gen array inside rcu | |
78 | * read section. Besides once set the net->gen->ptr[x] | |
79 | * pointer never changes (see rules in netns/generic.h). | |
80 | * | |
81 | * That said, we simply duplicate this array and schedule | |
82 | * the old copy for kfree after a grace period. | |
83 | */ | |
84 | ||
05fceb4a JP |
85 | memcpy(&ng->ptr, &old_ng->ptr, old_ng->len * sizeof(void*)); |
86 | ||
87 | rcu_assign_pointer(net->gen, ng); | |
04d4dfed | 88 | kfree_rcu(old_ng, rcu); |
05fceb4a JP |
89 | assign: |
90 | ng->ptr[id - 1] = data; | |
91 | return 0; | |
92 | } | |
93 | ||
f875bae0 EB |
94 | static int ops_init(const struct pernet_operations *ops, struct net *net) |
95 | { | |
b922934d JA |
96 | int err = -ENOMEM; |
97 | void *data = NULL; | |
98 | ||
f875bae0 | 99 | if (ops->id && ops->size) { |
b922934d | 100 | data = kzalloc(ops->size, GFP_KERNEL); |
f875bae0 | 101 | if (!data) |
b922934d | 102 | goto out; |
f875bae0 EB |
103 | |
104 | err = net_assign_generic(net, *ops->id, data); | |
b922934d JA |
105 | if (err) |
106 | goto cleanup; | |
f875bae0 | 107 | } |
b922934d | 108 | err = 0; |
f875bae0 | 109 | if (ops->init) |
b922934d JA |
110 | err = ops->init(net); |
111 | if (!err) | |
112 | return 0; | |
113 | ||
114 | cleanup: | |
115 | kfree(data); | |
116 | ||
117 | out: | |
118 | return err; | |
f875bae0 EB |
119 | } |
120 | ||
121 | static void ops_free(const struct pernet_operations *ops, struct net *net) | |
122 | { | |
123 | if (ops->id && ops->size) { | |
124 | int id = *ops->id; | |
125 | kfree(net_generic(net, id)); | |
126 | } | |
127 | } | |
128 | ||
72ad937a EB |
129 | static void ops_exit_list(const struct pernet_operations *ops, |
130 | struct list_head *net_exit_list) | |
131 | { | |
132 | struct net *net; | |
133 | if (ops->exit) { | |
134 | list_for_each_entry(net, net_exit_list, exit_list) | |
135 | ops->exit(net); | |
136 | } | |
72ad937a EB |
137 | if (ops->exit_batch) |
138 | ops->exit_batch(net_exit_list); | |
139 | } | |
140 | ||
141 | static void ops_free_list(const struct pernet_operations *ops, | |
142 | struct list_head *net_exit_list) | |
143 | { | |
144 | struct net *net; | |
145 | if (ops->size && ops->id) { | |
146 | list_for_each_entry(net, net_exit_list, exit_list) | |
147 | ops_free(ops, net); | |
148 | } | |
149 | } | |
150 | ||
0c7aecd4 ND |
151 | static int alloc_netid(struct net *net, struct net *peer, int reqid) |
152 | { | |
153 | int min = 0, max = 0; | |
154 | ||
155 | ASSERT_RTNL(); | |
156 | ||
157 | if (reqid >= 0) { | |
158 | min = reqid; | |
159 | max = reqid + 1; | |
160 | } | |
161 | ||
162 | return idr_alloc(&net->netns_ids, peer, min, max, GFP_KERNEL); | |
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 | ||
178 | static int __peernet2id(struct net *net, struct net *peer, bool alloc) | |
179 | { | |
180 | int id = idr_for_each(&net->netns_ids, net_eq_idr, peer); | |
181 | ||
182 | ASSERT_RTNL(); | |
183 | ||
184 | /* Magic value for id 0. */ | |
185 | if (id == NET_ID_ZERO) | |
186 | return 0; | |
187 | if (id > 0) | |
188 | return id; | |
189 | ||
190 | if (alloc) | |
191 | return alloc_netid(net, peer, -1); | |
192 | ||
193 | return -ENOENT; | |
194 | } | |
195 | ||
196 | /* This function returns the id of a peer netns. If no id is assigned, one will | |
197 | * be allocated and returned. | |
198 | */ | |
199 | int peernet2id(struct net *net, struct net *peer) | |
200 | { | |
576b7cd2 ND |
201 | bool alloc = atomic_read(&peer->count) == 0 ? false : true; |
202 | int id; | |
0c7aecd4 | 203 | |
576b7cd2 | 204 | id = __peernet2id(net, peer, alloc); |
0c7aecd4 ND |
205 | return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED; |
206 | } | |
193523bf | 207 | EXPORT_SYMBOL(peernet2id); |
0c7aecd4 ND |
208 | |
209 | struct net *get_net_ns_by_id(struct net *net, int id) | |
210 | { | |
211 | struct net *peer; | |
212 | ||
213 | if (id < 0) | |
214 | return NULL; | |
215 | ||
216 | rcu_read_lock(); | |
217 | peer = idr_find(&net->netns_ids, id); | |
218 | if (peer) | |
219 | get_net(peer); | |
220 | rcu_read_unlock(); | |
221 | ||
222 | return peer; | |
223 | } | |
224 | ||
5f256bec EB |
225 | /* |
226 | * setup_net runs the initializers for the network namespace object. | |
227 | */ | |
038e7332 | 228 | static __net_init int setup_net(struct net *net, struct user_namespace *user_ns) |
5f256bec EB |
229 | { |
230 | /* Must be called with net_mutex held */ | |
f875bae0 | 231 | const struct pernet_operations *ops, *saved_ops; |
486a87f1 | 232 | int error = 0; |
72ad937a | 233 | LIST_HEAD(net_exit_list); |
5f256bec | 234 | |
5f256bec | 235 | atomic_set(&net->count, 1); |
a685e089 | 236 | atomic_set(&net->passive, 1); |
4e985ada | 237 | net->dev_base_seq = 1; |
038e7332 | 238 | net->user_ns = user_ns; |
0c7aecd4 | 239 | idr_init(&net->netns_ids); |
486a87f1 | 240 | |
5d1e4468 | 241 | #ifdef NETNS_REFCNT_DEBUG |
5f256bec | 242 | atomic_set(&net->use_count, 0); |
5d1e4468 | 243 | #endif |
5f256bec | 244 | |
768f3591 | 245 | list_for_each_entry(ops, &pernet_list, list) { |
f875bae0 EB |
246 | error = ops_init(ops, net); |
247 | if (error < 0) | |
248 | goto out_undo; | |
5f256bec EB |
249 | } |
250 | out: | |
251 | return error; | |
768f3591 | 252 | |
5f256bec EB |
253 | out_undo: |
254 | /* Walk through the list backwards calling the exit functions | |
255 | * for the pernet modules whose init functions did not fail. | |
256 | */ | |
72ad937a | 257 | list_add(&net->exit_list, &net_exit_list); |
f875bae0 | 258 | saved_ops = ops; |
72ad937a EB |
259 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) |
260 | ops_exit_list(ops, &net_exit_list); | |
261 | ||
f875bae0 EB |
262 | ops = saved_ops; |
263 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) | |
72ad937a | 264 | ops_free_list(ops, &net_exit_list); |
310928d9 DL |
265 | |
266 | rcu_barrier(); | |
5f256bec EB |
267 | goto out; |
268 | } | |
269 | ||
6a1a3b9f | 270 | |
ebe47d47 CN |
271 | #ifdef CONFIG_NET_NS |
272 | static struct kmem_cache *net_cachep; | |
273 | static struct workqueue_struct *netns_wq; | |
274 | ||
486a87f1 | 275 | static struct net *net_alloc(void) |
45a19b0a | 276 | { |
486a87f1 DL |
277 | struct net *net = NULL; |
278 | struct net_generic *ng; | |
279 | ||
280 | ng = net_alloc_generic(); | |
281 | if (!ng) | |
282 | goto out; | |
283 | ||
284 | net = kmem_cache_zalloc(net_cachep, GFP_KERNEL); | |
45a19b0a | 285 | if (!net) |
486a87f1 | 286 | goto out_free; |
45a19b0a | 287 | |
486a87f1 DL |
288 | rcu_assign_pointer(net->gen, ng); |
289 | out: | |
290 | return net; | |
291 | ||
292 | out_free: | |
293 | kfree(ng); | |
294 | goto out; | |
295 | } | |
296 | ||
297 | static void net_free(struct net *net) | |
298 | { | |
5d1e4468 | 299 | #ifdef NETNS_REFCNT_DEBUG |
45a19b0a | 300 | if (unlikely(atomic_read(&net->use_count) != 0)) { |
e005d193 JP |
301 | pr_emerg("network namespace not free! Usage: %d\n", |
302 | atomic_read(&net->use_count)); | |
45a19b0a JFS |
303 | return; |
304 | } | |
5d1e4468 | 305 | #endif |
416c51e1 | 306 | kfree(rcu_access_pointer(net->gen)); |
45a19b0a JFS |
307 | kmem_cache_free(net_cachep, net); |
308 | } | |
309 | ||
a685e089 AV |
310 | void net_drop_ns(void *p) |
311 | { | |
312 | struct net *ns = p; | |
313 | if (ns && atomic_dec_and_test(&ns->passive)) | |
314 | net_free(ns); | |
315 | } | |
316 | ||
038e7332 EB |
317 | struct net *copy_net_ns(unsigned long flags, |
318 | struct user_namespace *user_ns, struct net *old_net) | |
9dd776b6 | 319 | { |
088eb2d9 AD |
320 | struct net *net; |
321 | int rv; | |
9dd776b6 | 322 | |
911cb193 RL |
323 | if (!(flags & CLONE_NEWNET)) |
324 | return get_net(old_net); | |
325 | ||
088eb2d9 AD |
326 | net = net_alloc(); |
327 | if (!net) | |
328 | return ERR_PTR(-ENOMEM); | |
038e7332 EB |
329 | |
330 | get_user_ns(user_ns); | |
331 | ||
9dd776b6 | 332 | mutex_lock(&net_mutex); |
038e7332 | 333 | rv = setup_net(net, user_ns); |
088eb2d9 | 334 | if (rv == 0) { |
486a87f1 | 335 | rtnl_lock(); |
11a28d37 | 336 | list_add_tail_rcu(&net->list, &net_namespace_list); |
486a87f1 DL |
337 | rtnl_unlock(); |
338 | } | |
9dd776b6 | 339 | mutex_unlock(&net_mutex); |
088eb2d9 | 340 | if (rv < 0) { |
038e7332 | 341 | put_user_ns(user_ns); |
a685e089 | 342 | net_drop_ns(net); |
088eb2d9 AD |
343 | return ERR_PTR(rv); |
344 | } | |
345 | return net; | |
346 | } | |
486a87f1 | 347 | |
2b035b39 EB |
348 | static DEFINE_SPINLOCK(cleanup_list_lock); |
349 | static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */ | |
350 | ||
6a1a3b9f PE |
351 | static void cleanup_net(struct work_struct *work) |
352 | { | |
f875bae0 | 353 | const struct pernet_operations *ops; |
6d458f5b | 354 | struct net *net, *tmp; |
1818ce4d | 355 | struct list_head net_kill_list; |
72ad937a | 356 | LIST_HEAD(net_exit_list); |
6a1a3b9f | 357 | |
2b035b39 EB |
358 | /* Atomically snapshot the list of namespaces to cleanup */ |
359 | spin_lock_irq(&cleanup_list_lock); | |
360 | list_replace_init(&cleanup_list, &net_kill_list); | |
361 | spin_unlock_irq(&cleanup_list_lock); | |
6a1a3b9f PE |
362 | |
363 | mutex_lock(&net_mutex); | |
364 | ||
365 | /* Don't let anyone else find us. */ | |
366 | rtnl_lock(); | |
72ad937a | 367 | list_for_each_entry(net, &net_kill_list, cleanup_list) { |
2b035b39 | 368 | list_del_rcu(&net->list); |
72ad937a | 369 | list_add_tail(&net->exit_list, &net_exit_list); |
6d458f5b ND |
370 | for_each_net(tmp) { |
371 | int id = __peernet2id(tmp, net, false); | |
372 | ||
373 | if (id >= 0) | |
374 | idr_remove(&tmp->netns_ids, id); | |
375 | } | |
376 | idr_destroy(&net->netns_ids); | |
377 | ||
72ad937a | 378 | } |
6a1a3b9f PE |
379 | rtnl_unlock(); |
380 | ||
11a28d37 JB |
381 | /* |
382 | * Another CPU might be rcu-iterating the list, wait for it. | |
383 | * This needs to be before calling the exit() notifiers, so | |
384 | * the rcu_barrier() below isn't sufficient alone. | |
385 | */ | |
386 | synchronize_rcu(); | |
387 | ||
6a1a3b9f | 388 | /* Run all of the network namespace exit methods */ |
72ad937a EB |
389 | list_for_each_entry_reverse(ops, &pernet_list, list) |
390 | ops_exit_list(ops, &net_exit_list); | |
391 | ||
f875bae0 | 392 | /* Free the net generic variables */ |
72ad937a EB |
393 | list_for_each_entry_reverse(ops, &pernet_list, list) |
394 | ops_free_list(ops, &net_exit_list); | |
6a1a3b9f PE |
395 | |
396 | mutex_unlock(&net_mutex); | |
397 | ||
398 | /* Ensure there are no outstanding rcu callbacks using this | |
399 | * network namespace. | |
400 | */ | |
401 | rcu_barrier(); | |
402 | ||
403 | /* Finally it is safe to free my network namespace structure */ | |
72ad937a EB |
404 | list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) { |
405 | list_del_init(&net->exit_list); | |
038e7332 | 406 | put_user_ns(net->user_ns); |
a685e089 | 407 | net_drop_ns(net); |
2b035b39 | 408 | } |
6a1a3b9f | 409 | } |
2b035b39 | 410 | static DECLARE_WORK(net_cleanup_work, cleanup_net); |
6a1a3b9f PE |
411 | |
412 | void __put_net(struct net *net) | |
413 | { | |
414 | /* Cleanup the network namespace in process context */ | |
2b035b39 EB |
415 | unsigned long flags; |
416 | ||
417 | spin_lock_irqsave(&cleanup_list_lock, flags); | |
418 | list_add(&net->cleanup_list, &cleanup_list); | |
419 | spin_unlock_irqrestore(&cleanup_list_lock, flags); | |
420 | ||
421 | queue_work(netns_wq, &net_cleanup_work); | |
6a1a3b9f PE |
422 | } |
423 | EXPORT_SYMBOL_GPL(__put_net); | |
424 | ||
956c9207 SR |
425 | struct net *get_net_ns_by_fd(int fd) |
426 | { | |
956c9207 | 427 | struct file *file; |
33c42940 | 428 | struct ns_common *ns; |
956c9207 SR |
429 | struct net *net; |
430 | ||
956c9207 | 431 | file = proc_ns_fget(fd); |
c316e6a3 AV |
432 | if (IS_ERR(file)) |
433 | return ERR_CAST(file); | |
956c9207 | 434 | |
f77c8014 | 435 | ns = get_proc_ns(file_inode(file)); |
33c42940 AV |
436 | if (ns->ops == &netns_operations) |
437 | net = get_net(container_of(ns, struct net, ns)); | |
c316e6a3 AV |
438 | else |
439 | net = ERR_PTR(-EINVAL); | |
956c9207 | 440 | |
c316e6a3 | 441 | fput(file); |
956c9207 SR |
442 | return net; |
443 | } | |
444 | ||
6a1a3b9f | 445 | #else |
956c9207 SR |
446 | struct net *get_net_ns_by_fd(int fd) |
447 | { | |
448 | return ERR_PTR(-EINVAL); | |
449 | } | |
6a1a3b9f | 450 | #endif |
4b681c82 | 451 | EXPORT_SYMBOL_GPL(get_net_ns_by_fd); |
6a1a3b9f | 452 | |
30ffee84 JB |
453 | struct net *get_net_ns_by_pid(pid_t pid) |
454 | { | |
455 | struct task_struct *tsk; | |
456 | struct net *net; | |
457 | ||
458 | /* Lookup the network namespace */ | |
459 | net = ERR_PTR(-ESRCH); | |
460 | rcu_read_lock(); | |
461 | tsk = find_task_by_vpid(pid); | |
462 | if (tsk) { | |
463 | struct nsproxy *nsproxy; | |
728dba3a EB |
464 | task_lock(tsk); |
465 | nsproxy = tsk->nsproxy; | |
30ffee84 JB |
466 | if (nsproxy) |
467 | net = get_net(nsproxy->net_ns); | |
728dba3a | 468 | task_unlock(tsk); |
30ffee84 JB |
469 | } |
470 | rcu_read_unlock(); | |
471 | return net; | |
472 | } | |
473 | EXPORT_SYMBOL_GPL(get_net_ns_by_pid); | |
474 | ||
98f842e6 EB |
475 | static __net_init int net_ns_net_init(struct net *net) |
476 | { | |
33c42940 AV |
477 | #ifdef CONFIG_NET_NS |
478 | net->ns.ops = &netns_operations; | |
479 | #endif | |
6344c433 | 480 | return ns_alloc_inum(&net->ns); |
98f842e6 EB |
481 | } |
482 | ||
483 | static __net_exit void net_ns_net_exit(struct net *net) | |
484 | { | |
6344c433 | 485 | ns_free_inum(&net->ns); |
98f842e6 EB |
486 | } |
487 | ||
488 | static struct pernet_operations __net_initdata net_ns_ops = { | |
489 | .init = net_ns_net_init, | |
490 | .exit = net_ns_net_exit, | |
491 | }; | |
492 | ||
0c7aecd4 ND |
493 | static struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = { |
494 | [NETNSA_NONE] = { .type = NLA_UNSPEC }, | |
495 | [NETNSA_NSID] = { .type = NLA_S32 }, | |
496 | [NETNSA_PID] = { .type = NLA_U32 }, | |
497 | [NETNSA_FD] = { .type = NLA_U32 }, | |
498 | }; | |
499 | ||
500 | static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
501 | { | |
502 | struct net *net = sock_net(skb->sk); | |
503 | struct nlattr *tb[NETNSA_MAX + 1]; | |
504 | struct net *peer; | |
505 | int nsid, err; | |
506 | ||
507 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
508 | rtnl_net_policy); | |
509 | if (err < 0) | |
510 | return err; | |
511 | if (!tb[NETNSA_NSID]) | |
512 | return -EINVAL; | |
513 | nsid = nla_get_s32(tb[NETNSA_NSID]); | |
514 | ||
515 | if (tb[NETNSA_PID]) | |
516 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); | |
517 | else if (tb[NETNSA_FD]) | |
518 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); | |
519 | else | |
520 | return -EINVAL; | |
521 | if (IS_ERR(peer)) | |
522 | return PTR_ERR(peer); | |
523 | ||
524 | if (__peernet2id(net, peer, false) >= 0) { | |
525 | err = -EEXIST; | |
526 | goto out; | |
527 | } | |
528 | ||
529 | err = alloc_netid(net, peer, nsid); | |
530 | if (err > 0) | |
531 | err = 0; | |
532 | out: | |
533 | put_net(peer); | |
534 | return err; | |
535 | } | |
536 | ||
537 | static int rtnl_net_get_size(void) | |
538 | { | |
539 | return NLMSG_ALIGN(sizeof(struct rtgenmsg)) | |
540 | + nla_total_size(sizeof(s32)) /* NETNSA_NSID */ | |
541 | ; | |
542 | } | |
543 | ||
544 | static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags, | |
545 | int cmd, struct net *net, struct net *peer) | |
546 | { | |
547 | struct nlmsghdr *nlh; | |
548 | struct rtgenmsg *rth; | |
549 | int id; | |
550 | ||
551 | ASSERT_RTNL(); | |
552 | ||
553 | nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags); | |
554 | if (!nlh) | |
555 | return -EMSGSIZE; | |
556 | ||
557 | rth = nlmsg_data(nlh); | |
558 | rth->rtgen_family = AF_UNSPEC; | |
559 | ||
560 | id = __peernet2id(net, peer, false); | |
561 | if (id < 0) | |
562 | id = NETNSA_NSID_NOT_ASSIGNED; | |
563 | if (nla_put_s32(skb, NETNSA_NSID, id)) | |
564 | goto nla_put_failure; | |
565 | ||
566 | nlmsg_end(skb, nlh); | |
567 | return 0; | |
568 | ||
569 | nla_put_failure: | |
570 | nlmsg_cancel(skb, nlh); | |
571 | return -EMSGSIZE; | |
572 | } | |
573 | ||
574 | static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
575 | { | |
576 | struct net *net = sock_net(skb->sk); | |
577 | struct nlattr *tb[NETNSA_MAX + 1]; | |
578 | struct sk_buff *msg; | |
579 | int err = -ENOBUFS; | |
580 | struct net *peer; | |
581 | ||
582 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
583 | rtnl_net_policy); | |
584 | if (err < 0) | |
585 | return err; | |
586 | if (tb[NETNSA_PID]) | |
587 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); | |
588 | else if (tb[NETNSA_FD]) | |
589 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); | |
590 | else | |
591 | return -EINVAL; | |
592 | ||
593 | if (IS_ERR(peer)) | |
594 | return PTR_ERR(peer); | |
595 | ||
596 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
597 | if (!msg) { | |
598 | err = -ENOMEM; | |
599 | goto out; | |
600 | } | |
601 | ||
602 | err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, | |
603 | RTM_GETNSID, net, peer); | |
604 | if (err < 0) | |
605 | goto err_out; | |
606 | ||
607 | err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid); | |
608 | goto out; | |
609 | ||
610 | err_out: | |
611 | nlmsg_free(msg); | |
612 | out: | |
613 | put_net(peer); | |
614 | return err; | |
615 | } | |
616 | ||
5f256bec EB |
617 | static int __init net_ns_init(void) |
618 | { | |
486a87f1 | 619 | struct net_generic *ng; |
5f256bec | 620 | |
d57a9212 | 621 | #ifdef CONFIG_NET_NS |
5f256bec EB |
622 | net_cachep = kmem_cache_create("net_namespace", sizeof(struct net), |
623 | SMP_CACHE_BYTES, | |
624 | SLAB_PANIC, NULL); | |
3ef1355d BT |
625 | |
626 | /* Create workqueue for cleanup */ | |
627 | netns_wq = create_singlethread_workqueue("netns"); | |
628 | if (!netns_wq) | |
629 | panic("Could not create netns workq"); | |
d57a9212 | 630 | #endif |
3ef1355d | 631 | |
486a87f1 DL |
632 | ng = net_alloc_generic(); |
633 | if (!ng) | |
634 | panic("Could not allocate generic netns"); | |
635 | ||
636 | rcu_assign_pointer(init_net.gen, ng); | |
637 | ||
5f256bec | 638 | mutex_lock(&net_mutex); |
038e7332 | 639 | if (setup_net(&init_net, &init_user_ns)) |
ca0f3112 | 640 | panic("Could not setup the initial network namespace"); |
5f256bec | 641 | |
f4618d39 | 642 | rtnl_lock(); |
11a28d37 | 643 | list_add_tail_rcu(&init_net.list, &net_namespace_list); |
f4618d39 | 644 | rtnl_unlock(); |
5f256bec EB |
645 | |
646 | mutex_unlock(&net_mutex); | |
5f256bec | 647 | |
98f842e6 EB |
648 | register_pernet_subsys(&net_ns_ops); |
649 | ||
0c7aecd4 ND |
650 | rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL); |
651 | rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, NULL, NULL); | |
652 | ||
5f256bec EB |
653 | return 0; |
654 | } | |
655 | ||
656 | pure_initcall(net_ns_init); | |
657 | ||
ed160e83 | 658 | #ifdef CONFIG_NET_NS |
f875bae0 EB |
659 | static int __register_pernet_operations(struct list_head *list, |
660 | struct pernet_operations *ops) | |
5f256bec | 661 | { |
72ad937a | 662 | struct net *net; |
5f256bec | 663 | int error; |
72ad937a | 664 | LIST_HEAD(net_exit_list); |
5f256bec | 665 | |
5f256bec | 666 | list_add_tail(&ops->list, list); |
f875bae0 | 667 | if (ops->init || (ops->id && ops->size)) { |
1dba323b | 668 | for_each_net(net) { |
f875bae0 | 669 | error = ops_init(ops, net); |
5f256bec EB |
670 | if (error) |
671 | goto out_undo; | |
72ad937a | 672 | list_add_tail(&net->exit_list, &net_exit_list); |
5f256bec EB |
673 | } |
674 | } | |
1dba323b | 675 | return 0; |
5f256bec EB |
676 | |
677 | out_undo: | |
678 | /* If I have an error cleanup all namespaces I initialized */ | |
679 | list_del(&ops->list); | |
72ad937a EB |
680 | ops_exit_list(ops, &net_exit_list); |
681 | ops_free_list(ops, &net_exit_list); | |
1dba323b | 682 | return error; |
5f256bec EB |
683 | } |
684 | ||
f875bae0 | 685 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
5f256bec EB |
686 | { |
687 | struct net *net; | |
72ad937a | 688 | LIST_HEAD(net_exit_list); |
5f256bec EB |
689 | |
690 | list_del(&ops->list); | |
72ad937a EB |
691 | for_each_net(net) |
692 | list_add_tail(&net->exit_list, &net_exit_list); | |
693 | ops_exit_list(ops, &net_exit_list); | |
694 | ops_free_list(ops, &net_exit_list); | |
5f256bec EB |
695 | } |
696 | ||
ed160e83 DL |
697 | #else |
698 | ||
f875bae0 EB |
699 | static int __register_pernet_operations(struct list_head *list, |
700 | struct pernet_operations *ops) | |
ed160e83 | 701 | { |
b922934d | 702 | return ops_init(ops, &init_net); |
ed160e83 DL |
703 | } |
704 | ||
f875bae0 | 705 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
ed160e83 | 706 | { |
72ad937a EB |
707 | LIST_HEAD(net_exit_list); |
708 | list_add(&init_net.exit_list, &net_exit_list); | |
709 | ops_exit_list(ops, &net_exit_list); | |
710 | ops_free_list(ops, &net_exit_list); | |
ed160e83 | 711 | } |
f875bae0 EB |
712 | |
713 | #endif /* CONFIG_NET_NS */ | |
ed160e83 | 714 | |
c93cf61f PE |
715 | static DEFINE_IDA(net_generic_ids); |
716 | ||
f875bae0 EB |
717 | static int register_pernet_operations(struct list_head *list, |
718 | struct pernet_operations *ops) | |
719 | { | |
720 | int error; | |
721 | ||
722 | if (ops->id) { | |
723 | again: | |
724 | error = ida_get_new_above(&net_generic_ids, 1, ops->id); | |
725 | if (error < 0) { | |
726 | if (error == -EAGAIN) { | |
727 | ida_pre_get(&net_generic_ids, GFP_KERNEL); | |
728 | goto again; | |
729 | } | |
730 | return error; | |
731 | } | |
073862ba | 732 | max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id); |
f875bae0 EB |
733 | } |
734 | error = __register_pernet_operations(list, ops); | |
3a765eda EB |
735 | if (error) { |
736 | rcu_barrier(); | |
737 | if (ops->id) | |
738 | ida_remove(&net_generic_ids, *ops->id); | |
739 | } | |
f875bae0 EB |
740 | |
741 | return error; | |
742 | } | |
743 | ||
744 | static void unregister_pernet_operations(struct pernet_operations *ops) | |
745 | { | |
746 | ||
747 | __unregister_pernet_operations(ops); | |
3a765eda | 748 | rcu_barrier(); |
f875bae0 EB |
749 | if (ops->id) |
750 | ida_remove(&net_generic_ids, *ops->id); | |
751 | } | |
752 | ||
5f256bec EB |
753 | /** |
754 | * register_pernet_subsys - register a network namespace subsystem | |
755 | * @ops: pernet operations structure for the subsystem | |
756 | * | |
757 | * Register a subsystem which has init and exit functions | |
758 | * that are called when network namespaces are created and | |
759 | * destroyed respectively. | |
760 | * | |
761 | * When registered all network namespace init functions are | |
762 | * called for every existing network namespace. Allowing kernel | |
763 | * modules to have a race free view of the set of network namespaces. | |
764 | * | |
765 | * When a new network namespace is created all of the init | |
766 | * methods are called in the order in which they were registered. | |
767 | * | |
768 | * When a network namespace is destroyed all of the exit methods | |
769 | * are called in the reverse of the order with which they were | |
770 | * registered. | |
771 | */ | |
772 | int register_pernet_subsys(struct pernet_operations *ops) | |
773 | { | |
774 | int error; | |
775 | mutex_lock(&net_mutex); | |
776 | error = register_pernet_operations(first_device, ops); | |
777 | mutex_unlock(&net_mutex); | |
778 | return error; | |
779 | } | |
780 | EXPORT_SYMBOL_GPL(register_pernet_subsys); | |
781 | ||
782 | /** | |
783 | * unregister_pernet_subsys - unregister a network namespace subsystem | |
784 | * @ops: pernet operations structure to manipulate | |
785 | * | |
786 | * Remove the pernet operations structure from the list to be | |
53379e57 | 787 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
788 | * addition run the exit method for all existing network |
789 | * namespaces. | |
790 | */ | |
b3c981d2 | 791 | void unregister_pernet_subsys(struct pernet_operations *ops) |
5f256bec EB |
792 | { |
793 | mutex_lock(&net_mutex); | |
b3c981d2 | 794 | unregister_pernet_operations(ops); |
5f256bec EB |
795 | mutex_unlock(&net_mutex); |
796 | } | |
797 | EXPORT_SYMBOL_GPL(unregister_pernet_subsys); | |
798 | ||
799 | /** | |
800 | * register_pernet_device - register a network namespace device | |
801 | * @ops: pernet operations structure for the subsystem | |
802 | * | |
803 | * Register a device which has init and exit functions | |
804 | * that are called when network namespaces are created and | |
805 | * destroyed respectively. | |
806 | * | |
807 | * When registered all network namespace init functions are | |
808 | * called for every existing network namespace. Allowing kernel | |
809 | * modules to have a race free view of the set of network namespaces. | |
810 | * | |
811 | * When a new network namespace is created all of the init | |
812 | * methods are called in the order in which they were registered. | |
813 | * | |
814 | * When a network namespace is destroyed all of the exit methods | |
815 | * are called in the reverse of the order with which they were | |
816 | * registered. | |
817 | */ | |
818 | int register_pernet_device(struct pernet_operations *ops) | |
819 | { | |
820 | int error; | |
821 | mutex_lock(&net_mutex); | |
822 | error = register_pernet_operations(&pernet_list, ops); | |
823 | if (!error && (first_device == &pernet_list)) | |
824 | first_device = &ops->list; | |
825 | mutex_unlock(&net_mutex); | |
826 | return error; | |
827 | } | |
828 | EXPORT_SYMBOL_GPL(register_pernet_device); | |
829 | ||
830 | /** | |
831 | * unregister_pernet_device - unregister a network namespace netdevice | |
832 | * @ops: pernet operations structure to manipulate | |
833 | * | |
834 | * Remove the pernet operations structure from the list to be | |
53379e57 | 835 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
836 | * addition run the exit method for all existing network |
837 | * namespaces. | |
838 | */ | |
839 | void unregister_pernet_device(struct pernet_operations *ops) | |
840 | { | |
841 | mutex_lock(&net_mutex); | |
842 | if (&ops->list == first_device) | |
843 | first_device = first_device->next; | |
844 | unregister_pernet_operations(ops); | |
845 | mutex_unlock(&net_mutex); | |
846 | } | |
847 | EXPORT_SYMBOL_GPL(unregister_pernet_device); | |
13b6f576 EB |
848 | |
849 | #ifdef CONFIG_NET_NS | |
64964528 | 850 | static struct ns_common *netns_get(struct task_struct *task) |
13b6f576 | 851 | { |
f0630529 EB |
852 | struct net *net = NULL; |
853 | struct nsproxy *nsproxy; | |
854 | ||
728dba3a EB |
855 | task_lock(task); |
856 | nsproxy = task->nsproxy; | |
f0630529 EB |
857 | if (nsproxy) |
858 | net = get_net(nsproxy->net_ns); | |
728dba3a | 859 | task_unlock(task); |
f0630529 | 860 | |
ff24870f AV |
861 | return net ? &net->ns : NULL; |
862 | } | |
863 | ||
864 | static inline struct net *to_net_ns(struct ns_common *ns) | |
865 | { | |
866 | return container_of(ns, struct net, ns); | |
13b6f576 EB |
867 | } |
868 | ||
64964528 | 869 | static void netns_put(struct ns_common *ns) |
13b6f576 | 870 | { |
ff24870f | 871 | put_net(to_net_ns(ns)); |
13b6f576 EB |
872 | } |
873 | ||
64964528 | 874 | static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns) |
13b6f576 | 875 | { |
ff24870f | 876 | struct net *net = to_net_ns(ns); |
142e1d1d | 877 | |
5e4a0847 | 878 | if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) || |
c7b96acf | 879 | !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) |
142e1d1d EB |
880 | return -EPERM; |
881 | ||
13b6f576 | 882 | put_net(nsproxy->net_ns); |
142e1d1d | 883 | nsproxy->net_ns = get_net(net); |
13b6f576 EB |
884 | return 0; |
885 | } | |
886 | ||
887 | const struct proc_ns_operations netns_operations = { | |
888 | .name = "net", | |
889 | .type = CLONE_NEWNET, | |
890 | .get = netns_get, | |
891 | .put = netns_put, | |
892 | .install = netns_install, | |
893 | }; | |
894 | #endif |