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 | |
768f3591 | 241 | list_for_each_entry(ops, &pernet_list, list) { |
f875bae0 EB |
242 | error = ops_init(ops, net); |
243 | if (error < 0) | |
244 | goto out_undo; | |
5f256bec EB |
245 | } |
246 | out: | |
247 | return error; | |
768f3591 | 248 | |
5f256bec EB |
249 | out_undo: |
250 | /* Walk through the list backwards calling the exit functions | |
251 | * for the pernet modules whose init functions did not fail. | |
252 | */ | |
72ad937a | 253 | list_add(&net->exit_list, &net_exit_list); |
f875bae0 | 254 | saved_ops = ops; |
72ad937a EB |
255 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) |
256 | ops_exit_list(ops, &net_exit_list); | |
257 | ||
f875bae0 EB |
258 | ops = saved_ops; |
259 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) | |
72ad937a | 260 | ops_free_list(ops, &net_exit_list); |
310928d9 DL |
261 | |
262 | rcu_barrier(); | |
5f256bec EB |
263 | goto out; |
264 | } | |
265 | ||
6a1a3b9f | 266 | |
ebe47d47 CN |
267 | #ifdef CONFIG_NET_NS |
268 | static struct kmem_cache *net_cachep; | |
269 | static struct workqueue_struct *netns_wq; | |
270 | ||
486a87f1 | 271 | static struct net *net_alloc(void) |
45a19b0a | 272 | { |
486a87f1 DL |
273 | struct net *net = NULL; |
274 | struct net_generic *ng; | |
275 | ||
276 | ng = net_alloc_generic(); | |
277 | if (!ng) | |
278 | goto out; | |
279 | ||
280 | net = kmem_cache_zalloc(net_cachep, GFP_KERNEL); | |
45a19b0a | 281 | if (!net) |
486a87f1 | 282 | goto out_free; |
45a19b0a | 283 | |
486a87f1 DL |
284 | rcu_assign_pointer(net->gen, ng); |
285 | out: | |
286 | return net; | |
287 | ||
288 | out_free: | |
289 | kfree(ng); | |
290 | goto out; | |
291 | } | |
292 | ||
293 | static void net_free(struct net *net) | |
294 | { | |
416c51e1 | 295 | kfree(rcu_access_pointer(net->gen)); |
45a19b0a JFS |
296 | kmem_cache_free(net_cachep, net); |
297 | } | |
298 | ||
a685e089 AV |
299 | void net_drop_ns(void *p) |
300 | { | |
301 | struct net *ns = p; | |
302 | if (ns && atomic_dec_and_test(&ns->passive)) | |
303 | net_free(ns); | |
304 | } | |
305 | ||
038e7332 EB |
306 | struct net *copy_net_ns(unsigned long flags, |
307 | struct user_namespace *user_ns, struct net *old_net) | |
9dd776b6 | 308 | { |
088eb2d9 AD |
309 | struct net *net; |
310 | int rv; | |
9dd776b6 | 311 | |
911cb193 RL |
312 | if (!(flags & CLONE_NEWNET)) |
313 | return get_net(old_net); | |
314 | ||
088eb2d9 AD |
315 | net = net_alloc(); |
316 | if (!net) | |
317 | return ERR_PTR(-ENOMEM); | |
038e7332 EB |
318 | |
319 | get_user_ns(user_ns); | |
320 | ||
9dd776b6 | 321 | mutex_lock(&net_mutex); |
038e7332 | 322 | rv = setup_net(net, user_ns); |
088eb2d9 | 323 | if (rv == 0) { |
486a87f1 | 324 | rtnl_lock(); |
11a28d37 | 325 | list_add_tail_rcu(&net->list, &net_namespace_list); |
486a87f1 DL |
326 | rtnl_unlock(); |
327 | } | |
9dd776b6 | 328 | mutex_unlock(&net_mutex); |
088eb2d9 | 329 | if (rv < 0) { |
038e7332 | 330 | put_user_ns(user_ns); |
a685e089 | 331 | net_drop_ns(net); |
088eb2d9 AD |
332 | return ERR_PTR(rv); |
333 | } | |
334 | return net; | |
335 | } | |
486a87f1 | 336 | |
2b035b39 EB |
337 | static DEFINE_SPINLOCK(cleanup_list_lock); |
338 | static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */ | |
339 | ||
6a1a3b9f PE |
340 | static void cleanup_net(struct work_struct *work) |
341 | { | |
f875bae0 | 342 | const struct pernet_operations *ops; |
6d458f5b | 343 | struct net *net, *tmp; |
1818ce4d | 344 | struct list_head net_kill_list; |
72ad937a | 345 | LIST_HEAD(net_exit_list); |
6a1a3b9f | 346 | |
2b035b39 EB |
347 | /* Atomically snapshot the list of namespaces to cleanup */ |
348 | spin_lock_irq(&cleanup_list_lock); | |
349 | list_replace_init(&cleanup_list, &net_kill_list); | |
350 | spin_unlock_irq(&cleanup_list_lock); | |
6a1a3b9f PE |
351 | |
352 | mutex_lock(&net_mutex); | |
353 | ||
354 | /* Don't let anyone else find us. */ | |
355 | rtnl_lock(); | |
72ad937a | 356 | list_for_each_entry(net, &net_kill_list, cleanup_list) { |
2b035b39 | 357 | list_del_rcu(&net->list); |
72ad937a | 358 | list_add_tail(&net->exit_list, &net_exit_list); |
6d458f5b ND |
359 | for_each_net(tmp) { |
360 | int id = __peernet2id(tmp, net, false); | |
361 | ||
362 | if (id >= 0) | |
363 | idr_remove(&tmp->netns_ids, id); | |
364 | } | |
365 | idr_destroy(&net->netns_ids); | |
366 | ||
72ad937a | 367 | } |
6a1a3b9f PE |
368 | rtnl_unlock(); |
369 | ||
11a28d37 JB |
370 | /* |
371 | * Another CPU might be rcu-iterating the list, wait for it. | |
372 | * This needs to be before calling the exit() notifiers, so | |
373 | * the rcu_barrier() below isn't sufficient alone. | |
374 | */ | |
375 | synchronize_rcu(); | |
376 | ||
6a1a3b9f | 377 | /* Run all of the network namespace exit methods */ |
72ad937a EB |
378 | list_for_each_entry_reverse(ops, &pernet_list, list) |
379 | ops_exit_list(ops, &net_exit_list); | |
380 | ||
f875bae0 | 381 | /* Free the net generic variables */ |
72ad937a EB |
382 | list_for_each_entry_reverse(ops, &pernet_list, list) |
383 | ops_free_list(ops, &net_exit_list); | |
6a1a3b9f PE |
384 | |
385 | mutex_unlock(&net_mutex); | |
386 | ||
387 | /* Ensure there are no outstanding rcu callbacks using this | |
388 | * network namespace. | |
389 | */ | |
390 | rcu_barrier(); | |
391 | ||
392 | /* Finally it is safe to free my network namespace structure */ | |
72ad937a EB |
393 | list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) { |
394 | list_del_init(&net->exit_list); | |
038e7332 | 395 | put_user_ns(net->user_ns); |
a685e089 | 396 | net_drop_ns(net); |
2b035b39 | 397 | } |
6a1a3b9f | 398 | } |
2b035b39 | 399 | static DECLARE_WORK(net_cleanup_work, cleanup_net); |
6a1a3b9f PE |
400 | |
401 | void __put_net(struct net *net) | |
402 | { | |
403 | /* Cleanup the network namespace in process context */ | |
2b035b39 EB |
404 | unsigned long flags; |
405 | ||
406 | spin_lock_irqsave(&cleanup_list_lock, flags); | |
407 | list_add(&net->cleanup_list, &cleanup_list); | |
408 | spin_unlock_irqrestore(&cleanup_list_lock, flags); | |
409 | ||
410 | queue_work(netns_wq, &net_cleanup_work); | |
6a1a3b9f PE |
411 | } |
412 | EXPORT_SYMBOL_GPL(__put_net); | |
413 | ||
956c9207 SR |
414 | struct net *get_net_ns_by_fd(int fd) |
415 | { | |
956c9207 | 416 | struct file *file; |
33c42940 | 417 | struct ns_common *ns; |
956c9207 SR |
418 | struct net *net; |
419 | ||
956c9207 | 420 | file = proc_ns_fget(fd); |
c316e6a3 AV |
421 | if (IS_ERR(file)) |
422 | return ERR_CAST(file); | |
956c9207 | 423 | |
f77c8014 | 424 | ns = get_proc_ns(file_inode(file)); |
33c42940 AV |
425 | if (ns->ops == &netns_operations) |
426 | net = get_net(container_of(ns, struct net, ns)); | |
c316e6a3 AV |
427 | else |
428 | net = ERR_PTR(-EINVAL); | |
956c9207 | 429 | |
c316e6a3 | 430 | fput(file); |
956c9207 SR |
431 | return net; |
432 | } | |
433 | ||
6a1a3b9f | 434 | #else |
956c9207 SR |
435 | struct net *get_net_ns_by_fd(int fd) |
436 | { | |
437 | return ERR_PTR(-EINVAL); | |
438 | } | |
6a1a3b9f | 439 | #endif |
4b681c82 | 440 | EXPORT_SYMBOL_GPL(get_net_ns_by_fd); |
6a1a3b9f | 441 | |
30ffee84 JB |
442 | struct net *get_net_ns_by_pid(pid_t pid) |
443 | { | |
444 | struct task_struct *tsk; | |
445 | struct net *net; | |
446 | ||
447 | /* Lookup the network namespace */ | |
448 | net = ERR_PTR(-ESRCH); | |
449 | rcu_read_lock(); | |
450 | tsk = find_task_by_vpid(pid); | |
451 | if (tsk) { | |
452 | struct nsproxy *nsproxy; | |
728dba3a EB |
453 | task_lock(tsk); |
454 | nsproxy = tsk->nsproxy; | |
30ffee84 JB |
455 | if (nsproxy) |
456 | net = get_net(nsproxy->net_ns); | |
728dba3a | 457 | task_unlock(tsk); |
30ffee84 JB |
458 | } |
459 | rcu_read_unlock(); | |
460 | return net; | |
461 | } | |
462 | EXPORT_SYMBOL_GPL(get_net_ns_by_pid); | |
463 | ||
98f842e6 EB |
464 | static __net_init int net_ns_net_init(struct net *net) |
465 | { | |
33c42940 AV |
466 | #ifdef CONFIG_NET_NS |
467 | net->ns.ops = &netns_operations; | |
468 | #endif | |
6344c433 | 469 | return ns_alloc_inum(&net->ns); |
98f842e6 EB |
470 | } |
471 | ||
472 | static __net_exit void net_ns_net_exit(struct net *net) | |
473 | { | |
6344c433 | 474 | ns_free_inum(&net->ns); |
98f842e6 EB |
475 | } |
476 | ||
477 | static struct pernet_operations __net_initdata net_ns_ops = { | |
478 | .init = net_ns_net_init, | |
479 | .exit = net_ns_net_exit, | |
480 | }; | |
481 | ||
0c7aecd4 ND |
482 | static struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = { |
483 | [NETNSA_NONE] = { .type = NLA_UNSPEC }, | |
484 | [NETNSA_NSID] = { .type = NLA_S32 }, | |
485 | [NETNSA_PID] = { .type = NLA_U32 }, | |
486 | [NETNSA_FD] = { .type = NLA_U32 }, | |
487 | }; | |
488 | ||
489 | static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
490 | { | |
491 | struct net *net = sock_net(skb->sk); | |
492 | struct nlattr *tb[NETNSA_MAX + 1]; | |
493 | struct net *peer; | |
494 | int nsid, err; | |
495 | ||
496 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
497 | rtnl_net_policy); | |
498 | if (err < 0) | |
499 | return err; | |
500 | if (!tb[NETNSA_NSID]) | |
501 | return -EINVAL; | |
502 | nsid = nla_get_s32(tb[NETNSA_NSID]); | |
503 | ||
504 | if (tb[NETNSA_PID]) | |
505 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); | |
506 | else if (tb[NETNSA_FD]) | |
507 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); | |
508 | else | |
509 | return -EINVAL; | |
510 | if (IS_ERR(peer)) | |
511 | return PTR_ERR(peer); | |
512 | ||
513 | if (__peernet2id(net, peer, false) >= 0) { | |
514 | err = -EEXIST; | |
515 | goto out; | |
516 | } | |
517 | ||
518 | err = alloc_netid(net, peer, nsid); | |
519 | if (err > 0) | |
520 | err = 0; | |
521 | out: | |
522 | put_net(peer); | |
523 | return err; | |
524 | } | |
525 | ||
526 | static int rtnl_net_get_size(void) | |
527 | { | |
528 | return NLMSG_ALIGN(sizeof(struct rtgenmsg)) | |
529 | + nla_total_size(sizeof(s32)) /* NETNSA_NSID */ | |
530 | ; | |
531 | } | |
532 | ||
533 | static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags, | |
534 | int cmd, struct net *net, struct net *peer) | |
535 | { | |
536 | struct nlmsghdr *nlh; | |
537 | struct rtgenmsg *rth; | |
538 | int id; | |
539 | ||
540 | ASSERT_RTNL(); | |
541 | ||
542 | nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags); | |
543 | if (!nlh) | |
544 | return -EMSGSIZE; | |
545 | ||
546 | rth = nlmsg_data(nlh); | |
547 | rth->rtgen_family = AF_UNSPEC; | |
548 | ||
549 | id = __peernet2id(net, peer, false); | |
550 | if (id < 0) | |
551 | id = NETNSA_NSID_NOT_ASSIGNED; | |
552 | if (nla_put_s32(skb, NETNSA_NSID, id)) | |
553 | goto nla_put_failure; | |
554 | ||
555 | nlmsg_end(skb, nlh); | |
556 | return 0; | |
557 | ||
558 | nla_put_failure: | |
559 | nlmsg_cancel(skb, nlh); | |
560 | return -EMSGSIZE; | |
561 | } | |
562 | ||
563 | static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
564 | { | |
565 | struct net *net = sock_net(skb->sk); | |
566 | struct nlattr *tb[NETNSA_MAX + 1]; | |
567 | struct sk_buff *msg; | |
0c7aecd4 | 568 | struct net *peer; |
b111e4e1 | 569 | int err; |
0c7aecd4 ND |
570 | |
571 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
572 | rtnl_net_policy); | |
573 | if (err < 0) | |
574 | return err; | |
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 | ||
582 | if (IS_ERR(peer)) | |
583 | return PTR_ERR(peer); | |
584 | ||
585 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
586 | if (!msg) { | |
587 | err = -ENOMEM; | |
588 | goto out; | |
589 | } | |
590 | ||
591 | err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, | |
592 | RTM_GETNSID, net, peer); | |
593 | if (err < 0) | |
594 | goto err_out; | |
595 | ||
596 | err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid); | |
597 | goto out; | |
598 | ||
599 | err_out: | |
600 | nlmsg_free(msg); | |
601 | out: | |
602 | put_net(peer); | |
603 | return err; | |
604 | } | |
605 | ||
5f256bec EB |
606 | static int __init net_ns_init(void) |
607 | { | |
486a87f1 | 608 | struct net_generic *ng; |
5f256bec | 609 | |
d57a9212 | 610 | #ifdef CONFIG_NET_NS |
5f256bec EB |
611 | net_cachep = kmem_cache_create("net_namespace", sizeof(struct net), |
612 | SMP_CACHE_BYTES, | |
613 | SLAB_PANIC, NULL); | |
3ef1355d BT |
614 | |
615 | /* Create workqueue for cleanup */ | |
616 | netns_wq = create_singlethread_workqueue("netns"); | |
617 | if (!netns_wq) | |
618 | panic("Could not create netns workq"); | |
d57a9212 | 619 | #endif |
3ef1355d | 620 | |
486a87f1 DL |
621 | ng = net_alloc_generic(); |
622 | if (!ng) | |
623 | panic("Could not allocate generic netns"); | |
624 | ||
625 | rcu_assign_pointer(init_net.gen, ng); | |
626 | ||
5f256bec | 627 | mutex_lock(&net_mutex); |
038e7332 | 628 | if (setup_net(&init_net, &init_user_ns)) |
ca0f3112 | 629 | panic("Could not setup the initial network namespace"); |
5f256bec | 630 | |
f4618d39 | 631 | rtnl_lock(); |
11a28d37 | 632 | list_add_tail_rcu(&init_net.list, &net_namespace_list); |
f4618d39 | 633 | rtnl_unlock(); |
5f256bec EB |
634 | |
635 | mutex_unlock(&net_mutex); | |
5f256bec | 636 | |
98f842e6 EB |
637 | register_pernet_subsys(&net_ns_ops); |
638 | ||
0c7aecd4 ND |
639 | rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL); |
640 | rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, NULL, NULL); | |
641 | ||
5f256bec EB |
642 | return 0; |
643 | } | |
644 | ||
645 | pure_initcall(net_ns_init); | |
646 | ||
ed160e83 | 647 | #ifdef CONFIG_NET_NS |
f875bae0 EB |
648 | static int __register_pernet_operations(struct list_head *list, |
649 | struct pernet_operations *ops) | |
5f256bec | 650 | { |
72ad937a | 651 | struct net *net; |
5f256bec | 652 | int error; |
72ad937a | 653 | LIST_HEAD(net_exit_list); |
5f256bec | 654 | |
5f256bec | 655 | list_add_tail(&ops->list, list); |
f875bae0 | 656 | if (ops->init || (ops->id && ops->size)) { |
1dba323b | 657 | for_each_net(net) { |
f875bae0 | 658 | error = ops_init(ops, net); |
5f256bec EB |
659 | if (error) |
660 | goto out_undo; | |
72ad937a | 661 | list_add_tail(&net->exit_list, &net_exit_list); |
5f256bec EB |
662 | } |
663 | } | |
1dba323b | 664 | return 0; |
5f256bec EB |
665 | |
666 | out_undo: | |
667 | /* If I have an error cleanup all namespaces I initialized */ | |
668 | list_del(&ops->list); | |
72ad937a EB |
669 | ops_exit_list(ops, &net_exit_list); |
670 | ops_free_list(ops, &net_exit_list); | |
1dba323b | 671 | return error; |
5f256bec EB |
672 | } |
673 | ||
f875bae0 | 674 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
5f256bec EB |
675 | { |
676 | struct net *net; | |
72ad937a | 677 | LIST_HEAD(net_exit_list); |
5f256bec EB |
678 | |
679 | list_del(&ops->list); | |
72ad937a EB |
680 | for_each_net(net) |
681 | list_add_tail(&net->exit_list, &net_exit_list); | |
682 | ops_exit_list(ops, &net_exit_list); | |
683 | ops_free_list(ops, &net_exit_list); | |
5f256bec EB |
684 | } |
685 | ||
ed160e83 DL |
686 | #else |
687 | ||
f875bae0 EB |
688 | static int __register_pernet_operations(struct list_head *list, |
689 | struct pernet_operations *ops) | |
ed160e83 | 690 | { |
b922934d | 691 | return ops_init(ops, &init_net); |
ed160e83 DL |
692 | } |
693 | ||
f875bae0 | 694 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
ed160e83 | 695 | { |
72ad937a EB |
696 | LIST_HEAD(net_exit_list); |
697 | list_add(&init_net.exit_list, &net_exit_list); | |
698 | ops_exit_list(ops, &net_exit_list); | |
699 | ops_free_list(ops, &net_exit_list); | |
ed160e83 | 700 | } |
f875bae0 EB |
701 | |
702 | #endif /* CONFIG_NET_NS */ | |
ed160e83 | 703 | |
c93cf61f PE |
704 | static DEFINE_IDA(net_generic_ids); |
705 | ||
f875bae0 EB |
706 | static int register_pernet_operations(struct list_head *list, |
707 | struct pernet_operations *ops) | |
708 | { | |
709 | int error; | |
710 | ||
711 | if (ops->id) { | |
712 | again: | |
713 | error = ida_get_new_above(&net_generic_ids, 1, ops->id); | |
714 | if (error < 0) { | |
715 | if (error == -EAGAIN) { | |
716 | ida_pre_get(&net_generic_ids, GFP_KERNEL); | |
717 | goto again; | |
718 | } | |
719 | return error; | |
720 | } | |
073862ba | 721 | max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id); |
f875bae0 EB |
722 | } |
723 | error = __register_pernet_operations(list, ops); | |
3a765eda EB |
724 | if (error) { |
725 | rcu_barrier(); | |
726 | if (ops->id) | |
727 | ida_remove(&net_generic_ids, *ops->id); | |
728 | } | |
f875bae0 EB |
729 | |
730 | return error; | |
731 | } | |
732 | ||
733 | static void unregister_pernet_operations(struct pernet_operations *ops) | |
734 | { | |
735 | ||
736 | __unregister_pernet_operations(ops); | |
3a765eda | 737 | rcu_barrier(); |
f875bae0 EB |
738 | if (ops->id) |
739 | ida_remove(&net_generic_ids, *ops->id); | |
740 | } | |
741 | ||
5f256bec EB |
742 | /** |
743 | * register_pernet_subsys - register a network namespace subsystem | |
744 | * @ops: pernet operations structure for the subsystem | |
745 | * | |
746 | * Register a subsystem which has init and exit functions | |
747 | * that are called when network namespaces are created and | |
748 | * destroyed respectively. | |
749 | * | |
750 | * When registered all network namespace init functions are | |
751 | * called for every existing network namespace. Allowing kernel | |
752 | * modules to have a race free view of the set of network namespaces. | |
753 | * | |
754 | * When a new network namespace is created all of the init | |
755 | * methods are called in the order in which they were registered. | |
756 | * | |
757 | * When a network namespace is destroyed all of the exit methods | |
758 | * are called in the reverse of the order with which they were | |
759 | * registered. | |
760 | */ | |
761 | int register_pernet_subsys(struct pernet_operations *ops) | |
762 | { | |
763 | int error; | |
764 | mutex_lock(&net_mutex); | |
765 | error = register_pernet_operations(first_device, ops); | |
766 | mutex_unlock(&net_mutex); | |
767 | return error; | |
768 | } | |
769 | EXPORT_SYMBOL_GPL(register_pernet_subsys); | |
770 | ||
771 | /** | |
772 | * unregister_pernet_subsys - unregister a network namespace subsystem | |
773 | * @ops: pernet operations structure to manipulate | |
774 | * | |
775 | * Remove the pernet operations structure from the list to be | |
53379e57 | 776 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
777 | * addition run the exit method for all existing network |
778 | * namespaces. | |
779 | */ | |
b3c981d2 | 780 | void unregister_pernet_subsys(struct pernet_operations *ops) |
5f256bec EB |
781 | { |
782 | mutex_lock(&net_mutex); | |
b3c981d2 | 783 | unregister_pernet_operations(ops); |
5f256bec EB |
784 | mutex_unlock(&net_mutex); |
785 | } | |
786 | EXPORT_SYMBOL_GPL(unregister_pernet_subsys); | |
787 | ||
788 | /** | |
789 | * register_pernet_device - register a network namespace device | |
790 | * @ops: pernet operations structure for the subsystem | |
791 | * | |
792 | * Register a device which has init and exit functions | |
793 | * that are called when network namespaces are created and | |
794 | * destroyed respectively. | |
795 | * | |
796 | * When registered all network namespace init functions are | |
797 | * called for every existing network namespace. Allowing kernel | |
798 | * modules to have a race free view of the set of network namespaces. | |
799 | * | |
800 | * When a new network namespace is created all of the init | |
801 | * methods are called in the order in which they were registered. | |
802 | * | |
803 | * When a network namespace is destroyed all of the exit methods | |
804 | * are called in the reverse of the order with which they were | |
805 | * registered. | |
806 | */ | |
807 | int register_pernet_device(struct pernet_operations *ops) | |
808 | { | |
809 | int error; | |
810 | mutex_lock(&net_mutex); | |
811 | error = register_pernet_operations(&pernet_list, ops); | |
812 | if (!error && (first_device == &pernet_list)) | |
813 | first_device = &ops->list; | |
814 | mutex_unlock(&net_mutex); | |
815 | return error; | |
816 | } | |
817 | EXPORT_SYMBOL_GPL(register_pernet_device); | |
818 | ||
819 | /** | |
820 | * unregister_pernet_device - unregister a network namespace netdevice | |
821 | * @ops: pernet operations structure to manipulate | |
822 | * | |
823 | * Remove the pernet operations structure from the list to be | |
53379e57 | 824 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
825 | * addition run the exit method for all existing network |
826 | * namespaces. | |
827 | */ | |
828 | void unregister_pernet_device(struct pernet_operations *ops) | |
829 | { | |
830 | mutex_lock(&net_mutex); | |
831 | if (&ops->list == first_device) | |
832 | first_device = first_device->next; | |
833 | unregister_pernet_operations(ops); | |
834 | mutex_unlock(&net_mutex); | |
835 | } | |
836 | EXPORT_SYMBOL_GPL(unregister_pernet_device); | |
13b6f576 EB |
837 | |
838 | #ifdef CONFIG_NET_NS | |
64964528 | 839 | static struct ns_common *netns_get(struct task_struct *task) |
13b6f576 | 840 | { |
f0630529 EB |
841 | struct net *net = NULL; |
842 | struct nsproxy *nsproxy; | |
843 | ||
728dba3a EB |
844 | task_lock(task); |
845 | nsproxy = task->nsproxy; | |
f0630529 EB |
846 | if (nsproxy) |
847 | net = get_net(nsproxy->net_ns); | |
728dba3a | 848 | task_unlock(task); |
f0630529 | 849 | |
ff24870f AV |
850 | return net ? &net->ns : NULL; |
851 | } | |
852 | ||
853 | static inline struct net *to_net_ns(struct ns_common *ns) | |
854 | { | |
855 | return container_of(ns, struct net, ns); | |
13b6f576 EB |
856 | } |
857 | ||
64964528 | 858 | static void netns_put(struct ns_common *ns) |
13b6f576 | 859 | { |
ff24870f | 860 | put_net(to_net_ns(ns)); |
13b6f576 EB |
861 | } |
862 | ||
64964528 | 863 | static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns) |
13b6f576 | 864 | { |
ff24870f | 865 | struct net *net = to_net_ns(ns); |
142e1d1d | 866 | |
5e4a0847 | 867 | if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) || |
c7b96acf | 868 | !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) |
142e1d1d EB |
869 | return -EPERM; |
870 | ||
13b6f576 | 871 | put_net(nsproxy->net_ns); |
142e1d1d | 872 | nsproxy->net_ns = get_net(net); |
13b6f576 EB |
873 | return 0; |
874 | } | |
875 | ||
876 | const struct proc_ns_operations netns_operations = { | |
877 | .name = "net", | |
878 | .type = CLONE_NEWNET, | |
879 | .get = netns_get, | |
880 | .put = netns_put, | |
881 | .install = netns_install, | |
882 | }; | |
883 | #endif |