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 | ||
9a963454 ND |
151 | static void rtnl_net_notifyid(struct net *net, struct net *peer, int cmd, |
152 | int id); | |
0c7aecd4 ND |
153 | static int alloc_netid(struct net *net, struct net *peer, int reqid) |
154 | { | |
9a963454 | 155 | int min = 0, max = 0, id; |
0c7aecd4 ND |
156 | |
157 | ASSERT_RTNL(); | |
158 | ||
159 | if (reqid >= 0) { | |
160 | min = reqid; | |
161 | max = reqid + 1; | |
162 | } | |
163 | ||
9a963454 ND |
164 | id = idr_alloc(&net->netns_ids, peer, min, max, GFP_KERNEL); |
165 | if (id >= 0) | |
166 | rtnl_net_notifyid(net, peer, RTM_NEWNSID, id); | |
167 | ||
168 | return id; | |
0c7aecd4 ND |
169 | } |
170 | ||
171 | /* This function is used by idr_for_each(). If net is equal to peer, the | |
172 | * function returns the id so that idr_for_each() stops. Because we cannot | |
173 | * returns the id 0 (idr_for_each() will not stop), we return the magic value | |
174 | * NET_ID_ZERO (-1) for it. | |
175 | */ | |
176 | #define NET_ID_ZERO -1 | |
177 | static int net_eq_idr(int id, void *net, void *peer) | |
178 | { | |
179 | if (net_eq(net, peer)) | |
180 | return id ? : NET_ID_ZERO; | |
181 | return 0; | |
182 | } | |
183 | ||
184 | static int __peernet2id(struct net *net, struct net *peer, bool alloc) | |
185 | { | |
186 | int id = idr_for_each(&net->netns_ids, net_eq_idr, peer); | |
187 | ||
188 | ASSERT_RTNL(); | |
189 | ||
190 | /* Magic value for id 0. */ | |
191 | if (id == NET_ID_ZERO) | |
192 | return 0; | |
193 | if (id > 0) | |
194 | return id; | |
195 | ||
196 | if (alloc) | |
197 | return alloc_netid(net, peer, -1); | |
198 | ||
199 | return -ENOENT; | |
200 | } | |
201 | ||
202 | /* This function returns the id of a peer netns. If no id is assigned, one will | |
203 | * be allocated and returned. | |
204 | */ | |
205 | int peernet2id(struct net *net, struct net *peer) | |
206 | { | |
576b7cd2 ND |
207 | bool alloc = atomic_read(&peer->count) == 0 ? false : true; |
208 | int id; | |
0c7aecd4 | 209 | |
576b7cd2 | 210 | id = __peernet2id(net, peer, alloc); |
0c7aecd4 ND |
211 | return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED; |
212 | } | |
193523bf | 213 | EXPORT_SYMBOL(peernet2id); |
0c7aecd4 ND |
214 | |
215 | struct net *get_net_ns_by_id(struct net *net, int id) | |
216 | { | |
217 | struct net *peer; | |
218 | ||
219 | if (id < 0) | |
220 | return NULL; | |
221 | ||
222 | rcu_read_lock(); | |
223 | peer = idr_find(&net->netns_ids, id); | |
224 | if (peer) | |
225 | get_net(peer); | |
226 | rcu_read_unlock(); | |
227 | ||
228 | return peer; | |
229 | } | |
230 | ||
5f256bec EB |
231 | /* |
232 | * setup_net runs the initializers for the network namespace object. | |
233 | */ | |
038e7332 | 234 | static __net_init int setup_net(struct net *net, struct user_namespace *user_ns) |
5f256bec EB |
235 | { |
236 | /* Must be called with net_mutex held */ | |
f875bae0 | 237 | const struct pernet_operations *ops, *saved_ops; |
486a87f1 | 238 | int error = 0; |
72ad937a | 239 | LIST_HEAD(net_exit_list); |
5f256bec | 240 | |
5f256bec | 241 | atomic_set(&net->count, 1); |
a685e089 | 242 | atomic_set(&net->passive, 1); |
4e985ada | 243 | net->dev_base_seq = 1; |
038e7332 | 244 | net->user_ns = user_ns; |
0c7aecd4 | 245 | idr_init(&net->netns_ids); |
486a87f1 | 246 | |
768f3591 | 247 | list_for_each_entry(ops, &pernet_list, list) { |
f875bae0 EB |
248 | error = ops_init(ops, net); |
249 | if (error < 0) | |
250 | goto out_undo; | |
5f256bec EB |
251 | } |
252 | out: | |
253 | return error; | |
768f3591 | 254 | |
5f256bec EB |
255 | out_undo: |
256 | /* Walk through the list backwards calling the exit functions | |
257 | * for the pernet modules whose init functions did not fail. | |
258 | */ | |
72ad937a | 259 | list_add(&net->exit_list, &net_exit_list); |
f875bae0 | 260 | saved_ops = ops; |
72ad937a EB |
261 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) |
262 | ops_exit_list(ops, &net_exit_list); | |
263 | ||
f875bae0 EB |
264 | ops = saved_ops; |
265 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) | |
72ad937a | 266 | ops_free_list(ops, &net_exit_list); |
310928d9 DL |
267 | |
268 | rcu_barrier(); | |
5f256bec EB |
269 | goto out; |
270 | } | |
271 | ||
6a1a3b9f | 272 | |
ebe47d47 CN |
273 | #ifdef CONFIG_NET_NS |
274 | static struct kmem_cache *net_cachep; | |
275 | static struct workqueue_struct *netns_wq; | |
276 | ||
486a87f1 | 277 | static struct net *net_alloc(void) |
45a19b0a | 278 | { |
486a87f1 DL |
279 | struct net *net = NULL; |
280 | struct net_generic *ng; | |
281 | ||
282 | ng = net_alloc_generic(); | |
283 | if (!ng) | |
284 | goto out; | |
285 | ||
286 | net = kmem_cache_zalloc(net_cachep, GFP_KERNEL); | |
45a19b0a | 287 | if (!net) |
486a87f1 | 288 | goto out_free; |
45a19b0a | 289 | |
486a87f1 DL |
290 | rcu_assign_pointer(net->gen, ng); |
291 | out: | |
292 | return net; | |
293 | ||
294 | out_free: | |
295 | kfree(ng); | |
296 | goto out; | |
297 | } | |
298 | ||
299 | static void net_free(struct net *net) | |
300 | { | |
416c51e1 | 301 | kfree(rcu_access_pointer(net->gen)); |
45a19b0a JFS |
302 | kmem_cache_free(net_cachep, net); |
303 | } | |
304 | ||
a685e089 AV |
305 | void net_drop_ns(void *p) |
306 | { | |
307 | struct net *ns = p; | |
308 | if (ns && atomic_dec_and_test(&ns->passive)) | |
309 | net_free(ns); | |
310 | } | |
311 | ||
038e7332 EB |
312 | struct net *copy_net_ns(unsigned long flags, |
313 | struct user_namespace *user_ns, struct net *old_net) | |
9dd776b6 | 314 | { |
088eb2d9 AD |
315 | struct net *net; |
316 | int rv; | |
9dd776b6 | 317 | |
911cb193 RL |
318 | if (!(flags & CLONE_NEWNET)) |
319 | return get_net(old_net); | |
320 | ||
088eb2d9 AD |
321 | net = net_alloc(); |
322 | if (!net) | |
323 | return ERR_PTR(-ENOMEM); | |
038e7332 EB |
324 | |
325 | get_user_ns(user_ns); | |
326 | ||
9dd776b6 | 327 | mutex_lock(&net_mutex); |
038e7332 | 328 | rv = setup_net(net, user_ns); |
088eb2d9 | 329 | if (rv == 0) { |
486a87f1 | 330 | rtnl_lock(); |
11a28d37 | 331 | list_add_tail_rcu(&net->list, &net_namespace_list); |
486a87f1 DL |
332 | rtnl_unlock(); |
333 | } | |
9dd776b6 | 334 | mutex_unlock(&net_mutex); |
088eb2d9 | 335 | if (rv < 0) { |
038e7332 | 336 | put_user_ns(user_ns); |
a685e089 | 337 | net_drop_ns(net); |
088eb2d9 AD |
338 | return ERR_PTR(rv); |
339 | } | |
340 | return net; | |
341 | } | |
486a87f1 | 342 | |
2b035b39 EB |
343 | static DEFINE_SPINLOCK(cleanup_list_lock); |
344 | static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */ | |
345 | ||
6a1a3b9f PE |
346 | static void cleanup_net(struct work_struct *work) |
347 | { | |
f875bae0 | 348 | const struct pernet_operations *ops; |
6d458f5b | 349 | struct net *net, *tmp; |
1818ce4d | 350 | struct list_head net_kill_list; |
72ad937a | 351 | LIST_HEAD(net_exit_list); |
6a1a3b9f | 352 | |
2b035b39 EB |
353 | /* Atomically snapshot the list of namespaces to cleanup */ |
354 | spin_lock_irq(&cleanup_list_lock); | |
355 | list_replace_init(&cleanup_list, &net_kill_list); | |
356 | spin_unlock_irq(&cleanup_list_lock); | |
6a1a3b9f PE |
357 | |
358 | mutex_lock(&net_mutex); | |
359 | ||
360 | /* Don't let anyone else find us. */ | |
361 | rtnl_lock(); | |
72ad937a | 362 | list_for_each_entry(net, &net_kill_list, cleanup_list) { |
2b035b39 | 363 | list_del_rcu(&net->list); |
72ad937a | 364 | list_add_tail(&net->exit_list, &net_exit_list); |
6d458f5b ND |
365 | for_each_net(tmp) { |
366 | int id = __peernet2id(tmp, net, false); | |
367 | ||
9a963454 ND |
368 | if (id >= 0) { |
369 | rtnl_net_notifyid(tmp, net, RTM_DELNSID, id); | |
6d458f5b | 370 | idr_remove(&tmp->netns_ids, id); |
9a963454 | 371 | } |
6d458f5b ND |
372 | } |
373 | idr_destroy(&net->netns_ids); | |
374 | ||
72ad937a | 375 | } |
6a1a3b9f PE |
376 | rtnl_unlock(); |
377 | ||
11a28d37 JB |
378 | /* |
379 | * Another CPU might be rcu-iterating the list, wait for it. | |
380 | * This needs to be before calling the exit() notifiers, so | |
381 | * the rcu_barrier() below isn't sufficient alone. | |
382 | */ | |
383 | synchronize_rcu(); | |
384 | ||
6a1a3b9f | 385 | /* Run all of the network namespace exit methods */ |
72ad937a EB |
386 | list_for_each_entry_reverse(ops, &pernet_list, list) |
387 | ops_exit_list(ops, &net_exit_list); | |
388 | ||
f875bae0 | 389 | /* Free the net generic variables */ |
72ad937a EB |
390 | list_for_each_entry_reverse(ops, &pernet_list, list) |
391 | ops_free_list(ops, &net_exit_list); | |
6a1a3b9f PE |
392 | |
393 | mutex_unlock(&net_mutex); | |
394 | ||
395 | /* Ensure there are no outstanding rcu callbacks using this | |
396 | * network namespace. | |
397 | */ | |
398 | rcu_barrier(); | |
399 | ||
400 | /* Finally it is safe to free my network namespace structure */ | |
72ad937a EB |
401 | list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) { |
402 | list_del_init(&net->exit_list); | |
038e7332 | 403 | put_user_ns(net->user_ns); |
a685e089 | 404 | net_drop_ns(net); |
2b035b39 | 405 | } |
6a1a3b9f | 406 | } |
2b035b39 | 407 | static DECLARE_WORK(net_cleanup_work, cleanup_net); |
6a1a3b9f PE |
408 | |
409 | void __put_net(struct net *net) | |
410 | { | |
411 | /* Cleanup the network namespace in process context */ | |
2b035b39 EB |
412 | unsigned long flags; |
413 | ||
414 | spin_lock_irqsave(&cleanup_list_lock, flags); | |
415 | list_add(&net->cleanup_list, &cleanup_list); | |
416 | spin_unlock_irqrestore(&cleanup_list_lock, flags); | |
417 | ||
418 | queue_work(netns_wq, &net_cleanup_work); | |
6a1a3b9f PE |
419 | } |
420 | EXPORT_SYMBOL_GPL(__put_net); | |
421 | ||
956c9207 SR |
422 | struct net *get_net_ns_by_fd(int fd) |
423 | { | |
956c9207 | 424 | struct file *file; |
33c42940 | 425 | struct ns_common *ns; |
956c9207 SR |
426 | struct net *net; |
427 | ||
956c9207 | 428 | file = proc_ns_fget(fd); |
c316e6a3 AV |
429 | if (IS_ERR(file)) |
430 | return ERR_CAST(file); | |
956c9207 | 431 | |
f77c8014 | 432 | ns = get_proc_ns(file_inode(file)); |
33c42940 AV |
433 | if (ns->ops == &netns_operations) |
434 | net = get_net(container_of(ns, struct net, ns)); | |
c316e6a3 AV |
435 | else |
436 | net = ERR_PTR(-EINVAL); | |
956c9207 | 437 | |
c316e6a3 | 438 | fput(file); |
956c9207 SR |
439 | return net; |
440 | } | |
441 | ||
6a1a3b9f | 442 | #else |
956c9207 SR |
443 | struct net *get_net_ns_by_fd(int fd) |
444 | { | |
445 | return ERR_PTR(-EINVAL); | |
446 | } | |
6a1a3b9f | 447 | #endif |
4b681c82 | 448 | EXPORT_SYMBOL_GPL(get_net_ns_by_fd); |
6a1a3b9f | 449 | |
30ffee84 JB |
450 | struct net *get_net_ns_by_pid(pid_t pid) |
451 | { | |
452 | struct task_struct *tsk; | |
453 | struct net *net; | |
454 | ||
455 | /* Lookup the network namespace */ | |
456 | net = ERR_PTR(-ESRCH); | |
457 | rcu_read_lock(); | |
458 | tsk = find_task_by_vpid(pid); | |
459 | if (tsk) { | |
460 | struct nsproxy *nsproxy; | |
728dba3a EB |
461 | task_lock(tsk); |
462 | nsproxy = tsk->nsproxy; | |
30ffee84 JB |
463 | if (nsproxy) |
464 | net = get_net(nsproxy->net_ns); | |
728dba3a | 465 | task_unlock(tsk); |
30ffee84 JB |
466 | } |
467 | rcu_read_unlock(); | |
468 | return net; | |
469 | } | |
470 | EXPORT_SYMBOL_GPL(get_net_ns_by_pid); | |
471 | ||
98f842e6 EB |
472 | static __net_init int net_ns_net_init(struct net *net) |
473 | { | |
33c42940 AV |
474 | #ifdef CONFIG_NET_NS |
475 | net->ns.ops = &netns_operations; | |
476 | #endif | |
6344c433 | 477 | return ns_alloc_inum(&net->ns); |
98f842e6 EB |
478 | } |
479 | ||
480 | static __net_exit void net_ns_net_exit(struct net *net) | |
481 | { | |
6344c433 | 482 | ns_free_inum(&net->ns); |
98f842e6 EB |
483 | } |
484 | ||
485 | static struct pernet_operations __net_initdata net_ns_ops = { | |
486 | .init = net_ns_net_init, | |
487 | .exit = net_ns_net_exit, | |
488 | }; | |
489 | ||
0c7aecd4 ND |
490 | static struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = { |
491 | [NETNSA_NONE] = { .type = NLA_UNSPEC }, | |
492 | [NETNSA_NSID] = { .type = NLA_S32 }, | |
493 | [NETNSA_PID] = { .type = NLA_U32 }, | |
494 | [NETNSA_FD] = { .type = NLA_U32 }, | |
495 | }; | |
496 | ||
497 | static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
498 | { | |
499 | struct net *net = sock_net(skb->sk); | |
500 | struct nlattr *tb[NETNSA_MAX + 1]; | |
501 | struct net *peer; | |
502 | int nsid, err; | |
503 | ||
504 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
505 | rtnl_net_policy); | |
506 | if (err < 0) | |
507 | return err; | |
508 | if (!tb[NETNSA_NSID]) | |
509 | return -EINVAL; | |
510 | nsid = nla_get_s32(tb[NETNSA_NSID]); | |
511 | ||
512 | if (tb[NETNSA_PID]) | |
513 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); | |
514 | else if (tb[NETNSA_FD]) | |
515 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); | |
516 | else | |
517 | return -EINVAL; | |
518 | if (IS_ERR(peer)) | |
519 | return PTR_ERR(peer); | |
520 | ||
521 | if (__peernet2id(net, peer, false) >= 0) { | |
522 | err = -EEXIST; | |
523 | goto out; | |
524 | } | |
525 | ||
526 | err = alloc_netid(net, peer, nsid); | |
527 | if (err > 0) | |
528 | err = 0; | |
529 | out: | |
530 | put_net(peer); | |
531 | return err; | |
532 | } | |
533 | ||
534 | static int rtnl_net_get_size(void) | |
535 | { | |
536 | return NLMSG_ALIGN(sizeof(struct rtgenmsg)) | |
537 | + nla_total_size(sizeof(s32)) /* NETNSA_NSID */ | |
538 | ; | |
539 | } | |
540 | ||
541 | static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags, | |
9a963454 ND |
542 | int cmd, struct net *net, struct net *peer, |
543 | int nsid) | |
0c7aecd4 ND |
544 | { |
545 | struct nlmsghdr *nlh; | |
546 | struct rtgenmsg *rth; | |
547 | int id; | |
548 | ||
549 | ASSERT_RTNL(); | |
550 | ||
551 | nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags); | |
552 | if (!nlh) | |
553 | return -EMSGSIZE; | |
554 | ||
555 | rth = nlmsg_data(nlh); | |
556 | rth->rtgen_family = AF_UNSPEC; | |
557 | ||
9a963454 ND |
558 | if (nsid >= 0) { |
559 | id = nsid; | |
560 | } else { | |
561 | id = __peernet2id(net, peer, false); | |
562 | if (id < 0) | |
563 | id = NETNSA_NSID_NOT_ASSIGNED; | |
564 | } | |
0c7aecd4 ND |
565 | if (nla_put_s32(skb, NETNSA_NSID, id)) |
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; |
b111e4e1 | 582 | int err; |
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 | ||
604 | err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, | |
9a963454 | 605 | RTM_GETNSID, net, peer, -1); |
0c7aecd4 ND |
606 | if (err < 0) |
607 | goto err_out; | |
608 | ||
609 | err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid); | |
610 | goto out; | |
611 | ||
612 | err_out: | |
613 | nlmsg_free(msg); | |
614 | out: | |
615 | put_net(peer); | |
616 | return err; | |
617 | } | |
618 | ||
9a963454 ND |
619 | static void rtnl_net_notifyid(struct net *net, struct net *peer, int cmd, |
620 | int id) | |
621 | { | |
622 | struct sk_buff *msg; | |
623 | int err = -ENOMEM; | |
624 | ||
625 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
626 | if (!msg) | |
627 | goto out; | |
628 | ||
629 | err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, peer, id); | |
630 | if (err < 0) | |
631 | goto err_out; | |
632 | ||
633 | rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0); | |
634 | return; | |
635 | ||
636 | err_out: | |
637 | nlmsg_free(msg); | |
638 | out: | |
639 | rtnl_set_sk_err(net, RTNLGRP_NSID, err); | |
640 | } | |
641 | ||
5f256bec EB |
642 | static int __init net_ns_init(void) |
643 | { | |
486a87f1 | 644 | struct net_generic *ng; |
5f256bec | 645 | |
d57a9212 | 646 | #ifdef CONFIG_NET_NS |
5f256bec EB |
647 | net_cachep = kmem_cache_create("net_namespace", sizeof(struct net), |
648 | SMP_CACHE_BYTES, | |
649 | SLAB_PANIC, NULL); | |
3ef1355d BT |
650 | |
651 | /* Create workqueue for cleanup */ | |
652 | netns_wq = create_singlethread_workqueue("netns"); | |
653 | if (!netns_wq) | |
654 | panic("Could not create netns workq"); | |
d57a9212 | 655 | #endif |
3ef1355d | 656 | |
486a87f1 DL |
657 | ng = net_alloc_generic(); |
658 | if (!ng) | |
659 | panic("Could not allocate generic netns"); | |
660 | ||
661 | rcu_assign_pointer(init_net.gen, ng); | |
662 | ||
5f256bec | 663 | mutex_lock(&net_mutex); |
038e7332 | 664 | if (setup_net(&init_net, &init_user_ns)) |
ca0f3112 | 665 | panic("Could not setup the initial network namespace"); |
5f256bec | 666 | |
f4618d39 | 667 | rtnl_lock(); |
11a28d37 | 668 | list_add_tail_rcu(&init_net.list, &net_namespace_list); |
f4618d39 | 669 | rtnl_unlock(); |
5f256bec EB |
670 | |
671 | mutex_unlock(&net_mutex); | |
5f256bec | 672 | |
98f842e6 EB |
673 | register_pernet_subsys(&net_ns_ops); |
674 | ||
0c7aecd4 ND |
675 | rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL); |
676 | rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, NULL, NULL); | |
677 | ||
5f256bec EB |
678 | return 0; |
679 | } | |
680 | ||
681 | pure_initcall(net_ns_init); | |
682 | ||
ed160e83 | 683 | #ifdef CONFIG_NET_NS |
f875bae0 EB |
684 | static int __register_pernet_operations(struct list_head *list, |
685 | struct pernet_operations *ops) | |
5f256bec | 686 | { |
72ad937a | 687 | struct net *net; |
5f256bec | 688 | int error; |
72ad937a | 689 | LIST_HEAD(net_exit_list); |
5f256bec | 690 | |
5f256bec | 691 | list_add_tail(&ops->list, list); |
f875bae0 | 692 | if (ops->init || (ops->id && ops->size)) { |
1dba323b | 693 | for_each_net(net) { |
f875bae0 | 694 | error = ops_init(ops, net); |
5f256bec EB |
695 | if (error) |
696 | goto out_undo; | |
72ad937a | 697 | list_add_tail(&net->exit_list, &net_exit_list); |
5f256bec EB |
698 | } |
699 | } | |
1dba323b | 700 | return 0; |
5f256bec EB |
701 | |
702 | out_undo: | |
703 | /* If I have an error cleanup all namespaces I initialized */ | |
704 | list_del(&ops->list); | |
72ad937a EB |
705 | ops_exit_list(ops, &net_exit_list); |
706 | ops_free_list(ops, &net_exit_list); | |
1dba323b | 707 | return error; |
5f256bec EB |
708 | } |
709 | ||
f875bae0 | 710 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
5f256bec EB |
711 | { |
712 | struct net *net; | |
72ad937a | 713 | LIST_HEAD(net_exit_list); |
5f256bec EB |
714 | |
715 | list_del(&ops->list); | |
72ad937a EB |
716 | for_each_net(net) |
717 | list_add_tail(&net->exit_list, &net_exit_list); | |
718 | ops_exit_list(ops, &net_exit_list); | |
719 | ops_free_list(ops, &net_exit_list); | |
5f256bec EB |
720 | } |
721 | ||
ed160e83 DL |
722 | #else |
723 | ||
f875bae0 EB |
724 | static int __register_pernet_operations(struct list_head *list, |
725 | struct pernet_operations *ops) | |
ed160e83 | 726 | { |
b922934d | 727 | return ops_init(ops, &init_net); |
ed160e83 DL |
728 | } |
729 | ||
f875bae0 | 730 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
ed160e83 | 731 | { |
72ad937a EB |
732 | LIST_HEAD(net_exit_list); |
733 | list_add(&init_net.exit_list, &net_exit_list); | |
734 | ops_exit_list(ops, &net_exit_list); | |
735 | ops_free_list(ops, &net_exit_list); | |
ed160e83 | 736 | } |
f875bae0 EB |
737 | |
738 | #endif /* CONFIG_NET_NS */ | |
ed160e83 | 739 | |
c93cf61f PE |
740 | static DEFINE_IDA(net_generic_ids); |
741 | ||
f875bae0 EB |
742 | static int register_pernet_operations(struct list_head *list, |
743 | struct pernet_operations *ops) | |
744 | { | |
745 | int error; | |
746 | ||
747 | if (ops->id) { | |
748 | again: | |
749 | error = ida_get_new_above(&net_generic_ids, 1, ops->id); | |
750 | if (error < 0) { | |
751 | if (error == -EAGAIN) { | |
752 | ida_pre_get(&net_generic_ids, GFP_KERNEL); | |
753 | goto again; | |
754 | } | |
755 | return error; | |
756 | } | |
073862ba | 757 | max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id); |
f875bae0 EB |
758 | } |
759 | error = __register_pernet_operations(list, ops); | |
3a765eda EB |
760 | if (error) { |
761 | rcu_barrier(); | |
762 | if (ops->id) | |
763 | ida_remove(&net_generic_ids, *ops->id); | |
764 | } | |
f875bae0 EB |
765 | |
766 | return error; | |
767 | } | |
768 | ||
769 | static void unregister_pernet_operations(struct pernet_operations *ops) | |
770 | { | |
771 | ||
772 | __unregister_pernet_operations(ops); | |
3a765eda | 773 | rcu_barrier(); |
f875bae0 EB |
774 | if (ops->id) |
775 | ida_remove(&net_generic_ids, *ops->id); | |
776 | } | |
777 | ||
5f256bec EB |
778 | /** |
779 | * register_pernet_subsys - register a network namespace subsystem | |
780 | * @ops: pernet operations structure for the subsystem | |
781 | * | |
782 | * Register a subsystem which has init and exit functions | |
783 | * that are called when network namespaces are created and | |
784 | * destroyed respectively. | |
785 | * | |
786 | * When registered all network namespace init functions are | |
787 | * called for every existing network namespace. Allowing kernel | |
788 | * modules to have a race free view of the set of network namespaces. | |
789 | * | |
790 | * When a new network namespace is created all of the init | |
791 | * methods are called in the order in which they were registered. | |
792 | * | |
793 | * When a network namespace is destroyed all of the exit methods | |
794 | * are called in the reverse of the order with which they were | |
795 | * registered. | |
796 | */ | |
797 | int register_pernet_subsys(struct pernet_operations *ops) | |
798 | { | |
799 | int error; | |
800 | mutex_lock(&net_mutex); | |
801 | error = register_pernet_operations(first_device, ops); | |
802 | mutex_unlock(&net_mutex); | |
803 | return error; | |
804 | } | |
805 | EXPORT_SYMBOL_GPL(register_pernet_subsys); | |
806 | ||
807 | /** | |
808 | * unregister_pernet_subsys - unregister a network namespace subsystem | |
809 | * @ops: pernet operations structure to manipulate | |
810 | * | |
811 | * Remove the pernet operations structure from the list to be | |
53379e57 | 812 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
813 | * addition run the exit method for all existing network |
814 | * namespaces. | |
815 | */ | |
b3c981d2 | 816 | void unregister_pernet_subsys(struct pernet_operations *ops) |
5f256bec EB |
817 | { |
818 | mutex_lock(&net_mutex); | |
b3c981d2 | 819 | unregister_pernet_operations(ops); |
5f256bec EB |
820 | mutex_unlock(&net_mutex); |
821 | } | |
822 | EXPORT_SYMBOL_GPL(unregister_pernet_subsys); | |
823 | ||
824 | /** | |
825 | * register_pernet_device - register a network namespace device | |
826 | * @ops: pernet operations structure for the subsystem | |
827 | * | |
828 | * Register a device which has init and exit functions | |
829 | * that are called when network namespaces are created and | |
830 | * destroyed respectively. | |
831 | * | |
832 | * When registered all network namespace init functions are | |
833 | * called for every existing network namespace. Allowing kernel | |
834 | * modules to have a race free view of the set of network namespaces. | |
835 | * | |
836 | * When a new network namespace is created all of the init | |
837 | * methods are called in the order in which they were registered. | |
838 | * | |
839 | * When a network namespace is destroyed all of the exit methods | |
840 | * are called in the reverse of the order with which they were | |
841 | * registered. | |
842 | */ | |
843 | int register_pernet_device(struct pernet_operations *ops) | |
844 | { | |
845 | int error; | |
846 | mutex_lock(&net_mutex); | |
847 | error = register_pernet_operations(&pernet_list, ops); | |
848 | if (!error && (first_device == &pernet_list)) | |
849 | first_device = &ops->list; | |
850 | mutex_unlock(&net_mutex); | |
851 | return error; | |
852 | } | |
853 | EXPORT_SYMBOL_GPL(register_pernet_device); | |
854 | ||
855 | /** | |
856 | * unregister_pernet_device - unregister a network namespace netdevice | |
857 | * @ops: pernet operations structure to manipulate | |
858 | * | |
859 | * Remove the pernet operations structure from the list to be | |
53379e57 | 860 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
861 | * addition run the exit method for all existing network |
862 | * namespaces. | |
863 | */ | |
864 | void unregister_pernet_device(struct pernet_operations *ops) | |
865 | { | |
866 | mutex_lock(&net_mutex); | |
867 | if (&ops->list == first_device) | |
868 | first_device = first_device->next; | |
869 | unregister_pernet_operations(ops); | |
870 | mutex_unlock(&net_mutex); | |
871 | } | |
872 | EXPORT_SYMBOL_GPL(unregister_pernet_device); | |
13b6f576 EB |
873 | |
874 | #ifdef CONFIG_NET_NS | |
64964528 | 875 | static struct ns_common *netns_get(struct task_struct *task) |
13b6f576 | 876 | { |
f0630529 EB |
877 | struct net *net = NULL; |
878 | struct nsproxy *nsproxy; | |
879 | ||
728dba3a EB |
880 | task_lock(task); |
881 | nsproxy = task->nsproxy; | |
f0630529 EB |
882 | if (nsproxy) |
883 | net = get_net(nsproxy->net_ns); | |
728dba3a | 884 | task_unlock(task); |
f0630529 | 885 | |
ff24870f AV |
886 | return net ? &net->ns : NULL; |
887 | } | |
888 | ||
889 | static inline struct net *to_net_ns(struct ns_common *ns) | |
890 | { | |
891 | return container_of(ns, struct net, ns); | |
13b6f576 EB |
892 | } |
893 | ||
64964528 | 894 | static void netns_put(struct ns_common *ns) |
13b6f576 | 895 | { |
ff24870f | 896 | put_net(to_net_ns(ns)); |
13b6f576 EB |
897 | } |
898 | ||
64964528 | 899 | static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns) |
13b6f576 | 900 | { |
ff24870f | 901 | struct net *net = to_net_ns(ns); |
142e1d1d | 902 | |
5e4a0847 | 903 | if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) || |
c7b96acf | 904 | !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) |
142e1d1d EB |
905 | return -EPERM; |
906 | ||
13b6f576 | 907 | put_net(nsproxy->net_ns); |
142e1d1d | 908 | nsproxy->net_ns = get_net(net); |
13b6f576 EB |
909 | return 0; |
910 | } | |
911 | ||
912 | const struct proc_ns_operations netns_operations = { | |
913 | .name = "net", | |
914 | .type = CLONE_NEWNET, | |
915 | .get = netns_get, | |
916 | .put = netns_put, | |
917 | .install = netns_install, | |
918 | }; | |
919 | #endif |