Commit | Line | Data |
---|---|---|
e005d193 JP |
1 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
2 | ||
5f256bec EB |
3 | #include <linux/workqueue.h> |
4 | #include <linux/rtnetlink.h> | |
5 | #include <linux/cache.h> | |
6 | #include <linux/slab.h> | |
7 | #include <linux/list.h> | |
8 | #include <linux/delay.h> | |
9dd776b6 | 9 | #include <linux/sched.h> |
c93cf61f | 10 | #include <linux/idr.h> |
11a28d37 | 11 | #include <linux/rculist.h> |
30ffee84 | 12 | #include <linux/nsproxy.h> |
0bb80f24 DH |
13 | #include <linux/fs.h> |
14 | #include <linux/proc_ns.h> | |
f0630529 | 15 | #include <linux/file.h> |
bc3b2d7f | 16 | #include <linux/export.h> |
038e7332 | 17 | #include <linux/user_namespace.h> |
0c7aecd4 | 18 | #include <linux/net_namespace.h> |
0c7aecd4 ND |
19 | #include <net/sock.h> |
20 | #include <net/netlink.h> | |
5f256bec | 21 | #include <net/net_namespace.h> |
dec827d1 | 22 | #include <net/netns/generic.h> |
5f256bec EB |
23 | |
24 | /* | |
25 | * Our network namespace constructor/destructor lists | |
26 | */ | |
27 | ||
28 | static LIST_HEAD(pernet_list); | |
29 | static struct list_head *first_device = &pernet_list; | |
200b916f | 30 | DEFINE_MUTEX(net_mutex); |
5f256bec | 31 | |
5f256bec | 32 | LIST_HEAD(net_namespace_list); |
b76a461f | 33 | EXPORT_SYMBOL_GPL(net_namespace_list); |
5f256bec | 34 | |
734b6541 RM |
35 | struct net init_net = { |
36 | .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head), | |
37 | }; | |
ff4b9502 | 38 | EXPORT_SYMBOL(init_net); |
5f256bec | 39 | |
dec827d1 PE |
40 | #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */ |
41 | ||
073862ba ED |
42 | static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS; |
43 | ||
44 | static struct net_generic *net_alloc_generic(void) | |
45 | { | |
46 | struct net_generic *ng; | |
47 | size_t generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]); | |
48 | ||
49 | ng = kzalloc(generic_size, GFP_KERNEL); | |
50 | if (ng) | |
51 | ng->len = max_gen_ptrs; | |
52 | ||
53 | return ng; | |
54 | } | |
55 | ||
05fceb4a JP |
56 | static int net_assign_generic(struct net *net, int id, void *data) |
57 | { | |
58 | struct net_generic *ng, *old_ng; | |
59 | ||
60 | BUG_ON(!mutex_is_locked(&net_mutex)); | |
61 | BUG_ON(id == 0); | |
62 | ||
1c87733d ED |
63 | old_ng = rcu_dereference_protected(net->gen, |
64 | lockdep_is_held(&net_mutex)); | |
65 | ng = old_ng; | |
05fceb4a JP |
66 | if (old_ng->len >= id) |
67 | goto assign; | |
68 | ||
073862ba | 69 | ng = net_alloc_generic(); |
05fceb4a JP |
70 | if (ng == NULL) |
71 | return -ENOMEM; | |
72 | ||
73 | /* | |
74 | * Some synchronisation notes: | |
75 | * | |
76 | * The net_generic explores the net->gen array inside rcu | |
77 | * read section. Besides once set the net->gen->ptr[x] | |
78 | * pointer never changes (see rules in netns/generic.h). | |
79 | * | |
80 | * That said, we simply duplicate this array and schedule | |
81 | * the old copy for kfree after a grace period. | |
82 | */ | |
83 | ||
05fceb4a JP |
84 | memcpy(&ng->ptr, &old_ng->ptr, old_ng->len * sizeof(void*)); |
85 | ||
86 | rcu_assign_pointer(net->gen, ng); | |
04d4dfed | 87 | kfree_rcu(old_ng, rcu); |
05fceb4a JP |
88 | assign: |
89 | ng->ptr[id - 1] = data; | |
90 | return 0; | |
91 | } | |
92 | ||
f875bae0 EB |
93 | static int ops_init(const struct pernet_operations *ops, struct net *net) |
94 | { | |
b922934d JA |
95 | int err = -ENOMEM; |
96 | void *data = NULL; | |
97 | ||
f875bae0 | 98 | if (ops->id && ops->size) { |
b922934d | 99 | data = kzalloc(ops->size, GFP_KERNEL); |
f875bae0 | 100 | if (!data) |
b922934d | 101 | goto out; |
f875bae0 EB |
102 | |
103 | err = net_assign_generic(net, *ops->id, data); | |
b922934d JA |
104 | if (err) |
105 | goto cleanup; | |
f875bae0 | 106 | } |
b922934d | 107 | err = 0; |
f875bae0 | 108 | if (ops->init) |
b922934d JA |
109 | err = ops->init(net); |
110 | if (!err) | |
111 | return 0; | |
112 | ||
113 | cleanup: | |
114 | kfree(data); | |
115 | ||
116 | out: | |
117 | return err; | |
f875bae0 EB |
118 | } |
119 | ||
120 | static void ops_free(const struct pernet_operations *ops, struct net *net) | |
121 | { | |
122 | if (ops->id && ops->size) { | |
123 | int id = *ops->id; | |
124 | kfree(net_generic(net, id)); | |
125 | } | |
126 | } | |
127 | ||
72ad937a EB |
128 | static void ops_exit_list(const struct pernet_operations *ops, |
129 | struct list_head *net_exit_list) | |
130 | { | |
131 | struct net *net; | |
132 | if (ops->exit) { | |
133 | list_for_each_entry(net, net_exit_list, exit_list) | |
134 | ops->exit(net); | |
135 | } | |
72ad937a EB |
136 | if (ops->exit_batch) |
137 | ops->exit_batch(net_exit_list); | |
138 | } | |
139 | ||
140 | static void ops_free_list(const struct pernet_operations *ops, | |
141 | struct list_head *net_exit_list) | |
142 | { | |
143 | struct net *net; | |
144 | if (ops->size && ops->id) { | |
145 | list_for_each_entry(net, net_exit_list, exit_list) | |
146 | ops_free(ops, net); | |
147 | } | |
148 | } | |
149 | ||
9a963454 ND |
150 | static void rtnl_net_notifyid(struct net *net, struct net *peer, int cmd, |
151 | int id); | |
0c7aecd4 ND |
152 | static int alloc_netid(struct net *net, struct net *peer, int reqid) |
153 | { | |
9a963454 | 154 | int min = 0, max = 0, id; |
0c7aecd4 ND |
155 | |
156 | ASSERT_RTNL(); | |
157 | ||
158 | if (reqid >= 0) { | |
159 | min = reqid; | |
160 | max = reqid + 1; | |
161 | } | |
162 | ||
9a963454 ND |
163 | id = idr_alloc(&net->netns_ids, peer, min, max, GFP_KERNEL); |
164 | if (id >= 0) | |
165 | rtnl_net_notifyid(net, peer, RTM_NEWNSID, id); | |
166 | ||
167 | return id; | |
0c7aecd4 ND |
168 | } |
169 | ||
170 | /* This function is used by idr_for_each(). If net is equal to peer, the | |
171 | * function returns the id so that idr_for_each() stops. Because we cannot | |
172 | * returns the id 0 (idr_for_each() will not stop), we return the magic value | |
173 | * NET_ID_ZERO (-1) for it. | |
174 | */ | |
175 | #define NET_ID_ZERO -1 | |
176 | static int net_eq_idr(int id, void *net, void *peer) | |
177 | { | |
178 | if (net_eq(net, peer)) | |
179 | return id ? : NET_ID_ZERO; | |
180 | return 0; | |
181 | } | |
182 | ||
183 | static int __peernet2id(struct net *net, struct net *peer, bool alloc) | |
184 | { | |
185 | int id = idr_for_each(&net->netns_ids, net_eq_idr, peer); | |
186 | ||
187 | ASSERT_RTNL(); | |
188 | ||
189 | /* Magic value for id 0. */ | |
190 | if (id == NET_ID_ZERO) | |
191 | return 0; | |
192 | if (id > 0) | |
193 | return id; | |
194 | ||
195 | if (alloc) | |
196 | return alloc_netid(net, peer, -1); | |
197 | ||
198 | return -ENOENT; | |
199 | } | |
200 | ||
201 | /* This function returns the id of a peer netns. If no id is assigned, one will | |
202 | * be allocated and returned. | |
203 | */ | |
204 | int peernet2id(struct net *net, struct net *peer) | |
205 | { | |
576b7cd2 ND |
206 | bool alloc = atomic_read(&peer->count) == 0 ? false : true; |
207 | int id; | |
0c7aecd4 | 208 | |
576b7cd2 | 209 | id = __peernet2id(net, peer, alloc); |
0c7aecd4 ND |
210 | return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED; |
211 | } | |
193523bf | 212 | EXPORT_SYMBOL(peernet2id); |
0c7aecd4 ND |
213 | |
214 | struct net *get_net_ns_by_id(struct net *net, int id) | |
215 | { | |
216 | struct net *peer; | |
217 | ||
218 | if (id < 0) | |
219 | return NULL; | |
220 | ||
221 | rcu_read_lock(); | |
222 | peer = idr_find(&net->netns_ids, id); | |
223 | if (peer) | |
224 | get_net(peer); | |
225 | rcu_read_unlock(); | |
226 | ||
227 | return peer; | |
228 | } | |
229 | ||
5f256bec EB |
230 | /* |
231 | * setup_net runs the initializers for the network namespace object. | |
232 | */ | |
038e7332 | 233 | static __net_init int setup_net(struct net *net, struct user_namespace *user_ns) |
5f256bec EB |
234 | { |
235 | /* Must be called with net_mutex held */ | |
f875bae0 | 236 | const struct pernet_operations *ops, *saved_ops; |
486a87f1 | 237 | int error = 0; |
72ad937a | 238 | LIST_HEAD(net_exit_list); |
5f256bec | 239 | |
5f256bec | 240 | atomic_set(&net->count, 1); |
a685e089 | 241 | atomic_set(&net->passive, 1); |
4e985ada | 242 | net->dev_base_seq = 1; |
038e7332 | 243 | net->user_ns = user_ns; |
0c7aecd4 | 244 | idr_init(&net->netns_ids); |
486a87f1 | 245 | |
768f3591 | 246 | list_for_each_entry(ops, &pernet_list, list) { |
f875bae0 EB |
247 | error = ops_init(ops, net); |
248 | if (error < 0) | |
249 | goto out_undo; | |
5f256bec EB |
250 | } |
251 | out: | |
252 | return error; | |
768f3591 | 253 | |
5f256bec EB |
254 | out_undo: |
255 | /* Walk through the list backwards calling the exit functions | |
256 | * for the pernet modules whose init functions did not fail. | |
257 | */ | |
72ad937a | 258 | list_add(&net->exit_list, &net_exit_list); |
f875bae0 | 259 | saved_ops = ops; |
72ad937a EB |
260 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) |
261 | ops_exit_list(ops, &net_exit_list); | |
262 | ||
f875bae0 EB |
263 | ops = saved_ops; |
264 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) | |
72ad937a | 265 | ops_free_list(ops, &net_exit_list); |
310928d9 DL |
266 | |
267 | rcu_barrier(); | |
5f256bec EB |
268 | goto out; |
269 | } | |
270 | ||
6a1a3b9f | 271 | |
ebe47d47 CN |
272 | #ifdef CONFIG_NET_NS |
273 | static struct kmem_cache *net_cachep; | |
274 | static struct workqueue_struct *netns_wq; | |
275 | ||
486a87f1 | 276 | static struct net *net_alloc(void) |
45a19b0a | 277 | { |
486a87f1 DL |
278 | struct net *net = NULL; |
279 | struct net_generic *ng; | |
280 | ||
281 | ng = net_alloc_generic(); | |
282 | if (!ng) | |
283 | goto out; | |
284 | ||
285 | net = kmem_cache_zalloc(net_cachep, GFP_KERNEL); | |
45a19b0a | 286 | if (!net) |
486a87f1 | 287 | goto out_free; |
45a19b0a | 288 | |
486a87f1 DL |
289 | rcu_assign_pointer(net->gen, ng); |
290 | out: | |
291 | return net; | |
292 | ||
293 | out_free: | |
294 | kfree(ng); | |
295 | goto out; | |
296 | } | |
297 | ||
298 | static void net_free(struct net *net) | |
299 | { | |
416c51e1 | 300 | kfree(rcu_access_pointer(net->gen)); |
45a19b0a JFS |
301 | kmem_cache_free(net_cachep, net); |
302 | } | |
303 | ||
a685e089 AV |
304 | void net_drop_ns(void *p) |
305 | { | |
306 | struct net *ns = p; | |
307 | if (ns && atomic_dec_and_test(&ns->passive)) | |
308 | net_free(ns); | |
309 | } | |
310 | ||
038e7332 EB |
311 | struct net *copy_net_ns(unsigned long flags, |
312 | struct user_namespace *user_ns, struct net *old_net) | |
9dd776b6 | 313 | { |
088eb2d9 AD |
314 | struct net *net; |
315 | int rv; | |
9dd776b6 | 316 | |
911cb193 RL |
317 | if (!(flags & CLONE_NEWNET)) |
318 | return get_net(old_net); | |
319 | ||
088eb2d9 AD |
320 | net = net_alloc(); |
321 | if (!net) | |
322 | return ERR_PTR(-ENOMEM); | |
038e7332 EB |
323 | |
324 | get_user_ns(user_ns); | |
325 | ||
9dd776b6 | 326 | mutex_lock(&net_mutex); |
038e7332 | 327 | rv = setup_net(net, user_ns); |
088eb2d9 | 328 | if (rv == 0) { |
486a87f1 | 329 | rtnl_lock(); |
11a28d37 | 330 | list_add_tail_rcu(&net->list, &net_namespace_list); |
486a87f1 DL |
331 | rtnl_unlock(); |
332 | } | |
9dd776b6 | 333 | mutex_unlock(&net_mutex); |
088eb2d9 | 334 | if (rv < 0) { |
038e7332 | 335 | put_user_ns(user_ns); |
a685e089 | 336 | net_drop_ns(net); |
088eb2d9 AD |
337 | return ERR_PTR(rv); |
338 | } | |
339 | return net; | |
340 | } | |
486a87f1 | 341 | |
2b035b39 EB |
342 | static DEFINE_SPINLOCK(cleanup_list_lock); |
343 | static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */ | |
344 | ||
6a1a3b9f PE |
345 | static void cleanup_net(struct work_struct *work) |
346 | { | |
f875bae0 | 347 | const struct pernet_operations *ops; |
6d458f5b | 348 | struct net *net, *tmp; |
1818ce4d | 349 | struct list_head net_kill_list; |
72ad937a | 350 | LIST_HEAD(net_exit_list); |
6a1a3b9f | 351 | |
2b035b39 EB |
352 | /* Atomically snapshot the list of namespaces to cleanup */ |
353 | spin_lock_irq(&cleanup_list_lock); | |
354 | list_replace_init(&cleanup_list, &net_kill_list); | |
355 | spin_unlock_irq(&cleanup_list_lock); | |
6a1a3b9f PE |
356 | |
357 | mutex_lock(&net_mutex); | |
358 | ||
359 | /* Don't let anyone else find us. */ | |
360 | rtnl_lock(); | |
72ad937a | 361 | list_for_each_entry(net, &net_kill_list, cleanup_list) { |
2b035b39 | 362 | list_del_rcu(&net->list); |
72ad937a | 363 | list_add_tail(&net->exit_list, &net_exit_list); |
6d458f5b ND |
364 | for_each_net(tmp) { |
365 | int id = __peernet2id(tmp, net, false); | |
366 | ||
9a963454 ND |
367 | if (id >= 0) { |
368 | rtnl_net_notifyid(tmp, net, RTM_DELNSID, id); | |
6d458f5b | 369 | idr_remove(&tmp->netns_ids, id); |
9a963454 | 370 | } |
6d458f5b ND |
371 | } |
372 | idr_destroy(&net->netns_ids); | |
373 | ||
72ad937a | 374 | } |
6a1a3b9f PE |
375 | rtnl_unlock(); |
376 | ||
11a28d37 JB |
377 | /* |
378 | * Another CPU might be rcu-iterating the list, wait for it. | |
379 | * This needs to be before calling the exit() notifiers, so | |
380 | * the rcu_barrier() below isn't sufficient alone. | |
381 | */ | |
382 | synchronize_rcu(); | |
383 | ||
6a1a3b9f | 384 | /* Run all of the network namespace exit methods */ |
72ad937a EB |
385 | list_for_each_entry_reverse(ops, &pernet_list, list) |
386 | ops_exit_list(ops, &net_exit_list); | |
387 | ||
f875bae0 | 388 | /* Free the net generic variables */ |
72ad937a EB |
389 | list_for_each_entry_reverse(ops, &pernet_list, list) |
390 | ops_free_list(ops, &net_exit_list); | |
6a1a3b9f PE |
391 | |
392 | mutex_unlock(&net_mutex); | |
393 | ||
394 | /* Ensure there are no outstanding rcu callbacks using this | |
395 | * network namespace. | |
396 | */ | |
397 | rcu_barrier(); | |
398 | ||
399 | /* Finally it is safe to free my network namespace structure */ | |
72ad937a EB |
400 | list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) { |
401 | list_del_init(&net->exit_list); | |
038e7332 | 402 | put_user_ns(net->user_ns); |
a685e089 | 403 | net_drop_ns(net); |
2b035b39 | 404 | } |
6a1a3b9f | 405 | } |
2b035b39 | 406 | static DECLARE_WORK(net_cleanup_work, cleanup_net); |
6a1a3b9f PE |
407 | |
408 | void __put_net(struct net *net) | |
409 | { | |
410 | /* Cleanup the network namespace in process context */ | |
2b035b39 EB |
411 | unsigned long flags; |
412 | ||
413 | spin_lock_irqsave(&cleanup_list_lock, flags); | |
414 | list_add(&net->cleanup_list, &cleanup_list); | |
415 | spin_unlock_irqrestore(&cleanup_list_lock, flags); | |
416 | ||
417 | queue_work(netns_wq, &net_cleanup_work); | |
6a1a3b9f PE |
418 | } |
419 | EXPORT_SYMBOL_GPL(__put_net); | |
420 | ||
956c9207 SR |
421 | struct net *get_net_ns_by_fd(int fd) |
422 | { | |
956c9207 | 423 | struct file *file; |
33c42940 | 424 | struct ns_common *ns; |
956c9207 SR |
425 | struct net *net; |
426 | ||
956c9207 | 427 | file = proc_ns_fget(fd); |
c316e6a3 AV |
428 | if (IS_ERR(file)) |
429 | return ERR_CAST(file); | |
956c9207 | 430 | |
f77c8014 | 431 | ns = get_proc_ns(file_inode(file)); |
33c42940 AV |
432 | if (ns->ops == &netns_operations) |
433 | net = get_net(container_of(ns, struct net, ns)); | |
c316e6a3 AV |
434 | else |
435 | net = ERR_PTR(-EINVAL); | |
956c9207 | 436 | |
c316e6a3 | 437 | fput(file); |
956c9207 SR |
438 | return net; |
439 | } | |
440 | ||
6a1a3b9f | 441 | #else |
956c9207 SR |
442 | struct net *get_net_ns_by_fd(int fd) |
443 | { | |
444 | return ERR_PTR(-EINVAL); | |
445 | } | |
6a1a3b9f | 446 | #endif |
4b681c82 | 447 | EXPORT_SYMBOL_GPL(get_net_ns_by_fd); |
6a1a3b9f | 448 | |
30ffee84 JB |
449 | struct net *get_net_ns_by_pid(pid_t pid) |
450 | { | |
451 | struct task_struct *tsk; | |
452 | struct net *net; | |
453 | ||
454 | /* Lookup the network namespace */ | |
455 | net = ERR_PTR(-ESRCH); | |
456 | rcu_read_lock(); | |
457 | tsk = find_task_by_vpid(pid); | |
458 | if (tsk) { | |
459 | struct nsproxy *nsproxy; | |
728dba3a EB |
460 | task_lock(tsk); |
461 | nsproxy = tsk->nsproxy; | |
30ffee84 JB |
462 | if (nsproxy) |
463 | net = get_net(nsproxy->net_ns); | |
728dba3a | 464 | task_unlock(tsk); |
30ffee84 JB |
465 | } |
466 | rcu_read_unlock(); | |
467 | return net; | |
468 | } | |
469 | EXPORT_SYMBOL_GPL(get_net_ns_by_pid); | |
470 | ||
98f842e6 EB |
471 | static __net_init int net_ns_net_init(struct net *net) |
472 | { | |
33c42940 AV |
473 | #ifdef CONFIG_NET_NS |
474 | net->ns.ops = &netns_operations; | |
475 | #endif | |
6344c433 | 476 | return ns_alloc_inum(&net->ns); |
98f842e6 EB |
477 | } |
478 | ||
479 | static __net_exit void net_ns_net_exit(struct net *net) | |
480 | { | |
6344c433 | 481 | ns_free_inum(&net->ns); |
98f842e6 EB |
482 | } |
483 | ||
484 | static struct pernet_operations __net_initdata net_ns_ops = { | |
485 | .init = net_ns_net_init, | |
486 | .exit = net_ns_net_exit, | |
487 | }; | |
488 | ||
0c7aecd4 ND |
489 | static struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = { |
490 | [NETNSA_NONE] = { .type = NLA_UNSPEC }, | |
491 | [NETNSA_NSID] = { .type = NLA_S32 }, | |
492 | [NETNSA_PID] = { .type = NLA_U32 }, | |
493 | [NETNSA_FD] = { .type = NLA_U32 }, | |
494 | }; | |
495 | ||
496 | static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
497 | { | |
498 | struct net *net = sock_net(skb->sk); | |
499 | struct nlattr *tb[NETNSA_MAX + 1]; | |
500 | struct net *peer; | |
501 | int nsid, err; | |
502 | ||
503 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
504 | rtnl_net_policy); | |
505 | if (err < 0) | |
506 | return err; | |
507 | if (!tb[NETNSA_NSID]) | |
508 | return -EINVAL; | |
509 | nsid = nla_get_s32(tb[NETNSA_NSID]); | |
510 | ||
511 | if (tb[NETNSA_PID]) | |
512 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); | |
513 | else if (tb[NETNSA_FD]) | |
514 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); | |
515 | else | |
516 | return -EINVAL; | |
517 | if (IS_ERR(peer)) | |
518 | return PTR_ERR(peer); | |
519 | ||
520 | if (__peernet2id(net, peer, false) >= 0) { | |
521 | err = -EEXIST; | |
522 | goto out; | |
523 | } | |
524 | ||
525 | err = alloc_netid(net, peer, nsid); | |
526 | if (err > 0) | |
527 | err = 0; | |
528 | out: | |
529 | put_net(peer); | |
530 | return err; | |
531 | } | |
532 | ||
533 | static int rtnl_net_get_size(void) | |
534 | { | |
535 | return NLMSG_ALIGN(sizeof(struct rtgenmsg)) | |
536 | + nla_total_size(sizeof(s32)) /* NETNSA_NSID */ | |
537 | ; | |
538 | } | |
539 | ||
540 | static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags, | |
9a963454 ND |
541 | int cmd, struct net *net, struct net *peer, |
542 | int nsid) | |
0c7aecd4 ND |
543 | { |
544 | struct nlmsghdr *nlh; | |
545 | struct rtgenmsg *rth; | |
546 | int id; | |
547 | ||
548 | ASSERT_RTNL(); | |
549 | ||
550 | nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags); | |
551 | if (!nlh) | |
552 | return -EMSGSIZE; | |
553 | ||
554 | rth = nlmsg_data(nlh); | |
555 | rth->rtgen_family = AF_UNSPEC; | |
556 | ||
9a963454 ND |
557 | if (nsid >= 0) { |
558 | id = nsid; | |
559 | } else { | |
560 | id = __peernet2id(net, peer, false); | |
561 | if (id < 0) | |
562 | id = NETNSA_NSID_NOT_ASSIGNED; | |
563 | } | |
0c7aecd4 ND |
564 | if (nla_put_s32(skb, NETNSA_NSID, id)) |
565 | goto nla_put_failure; | |
566 | ||
567 | nlmsg_end(skb, nlh); | |
568 | return 0; | |
569 | ||
570 | nla_put_failure: | |
571 | nlmsg_cancel(skb, nlh); | |
572 | return -EMSGSIZE; | |
573 | } | |
574 | ||
575 | static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
576 | { | |
577 | struct net *net = sock_net(skb->sk); | |
578 | struct nlattr *tb[NETNSA_MAX + 1]; | |
579 | struct sk_buff *msg; | |
0c7aecd4 | 580 | struct net *peer; |
b111e4e1 | 581 | int err; |
0c7aecd4 ND |
582 | |
583 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
584 | rtnl_net_policy); | |
585 | if (err < 0) | |
586 | return err; | |
587 | if (tb[NETNSA_PID]) | |
588 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); | |
589 | else if (tb[NETNSA_FD]) | |
590 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); | |
591 | else | |
592 | return -EINVAL; | |
593 | ||
594 | if (IS_ERR(peer)) | |
595 | return PTR_ERR(peer); | |
596 | ||
597 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
598 | if (!msg) { | |
599 | err = -ENOMEM; | |
600 | goto out; | |
601 | } | |
602 | ||
603 | err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, | |
9a963454 | 604 | RTM_GETNSID, net, peer, -1); |
0c7aecd4 ND |
605 | if (err < 0) |
606 | goto err_out; | |
607 | ||
608 | err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid); | |
609 | goto out; | |
610 | ||
611 | err_out: | |
612 | nlmsg_free(msg); | |
613 | out: | |
614 | put_net(peer); | |
615 | return err; | |
616 | } | |
617 | ||
a143c40c ND |
618 | struct rtnl_net_dump_cb { |
619 | struct net *net; | |
620 | struct sk_buff *skb; | |
621 | struct netlink_callback *cb; | |
622 | int idx; | |
623 | int s_idx; | |
624 | }; | |
625 | ||
626 | static int rtnl_net_dumpid_one(int id, void *peer, void *data) | |
627 | { | |
628 | struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data; | |
629 | int ret; | |
630 | ||
631 | if (net_cb->idx < net_cb->s_idx) | |
632 | goto cont; | |
633 | ||
634 | ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid, | |
635 | net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI, | |
636 | RTM_NEWNSID, net_cb->net, peer, id); | |
637 | if (ret < 0) | |
638 | return ret; | |
639 | ||
640 | cont: | |
641 | net_cb->idx++; | |
642 | return 0; | |
643 | } | |
644 | ||
645 | static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb) | |
646 | { | |
647 | struct net *net = sock_net(skb->sk); | |
648 | struct rtnl_net_dump_cb net_cb = { | |
649 | .net = net, | |
650 | .skb = skb, | |
651 | .cb = cb, | |
652 | .idx = 0, | |
653 | .s_idx = cb->args[0], | |
654 | }; | |
655 | ||
656 | ASSERT_RTNL(); | |
657 | ||
658 | idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb); | |
659 | ||
660 | cb->args[0] = net_cb.idx; | |
661 | return skb->len; | |
662 | } | |
663 | ||
9a963454 ND |
664 | static void rtnl_net_notifyid(struct net *net, struct net *peer, int cmd, |
665 | int id) | |
666 | { | |
667 | struct sk_buff *msg; | |
668 | int err = -ENOMEM; | |
669 | ||
670 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
671 | if (!msg) | |
672 | goto out; | |
673 | ||
674 | err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, peer, id); | |
675 | if (err < 0) | |
676 | goto err_out; | |
677 | ||
678 | rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0); | |
679 | return; | |
680 | ||
681 | err_out: | |
682 | nlmsg_free(msg); | |
683 | out: | |
684 | rtnl_set_sk_err(net, RTNLGRP_NSID, err); | |
685 | } | |
686 | ||
5f256bec EB |
687 | static int __init net_ns_init(void) |
688 | { | |
486a87f1 | 689 | struct net_generic *ng; |
5f256bec | 690 | |
d57a9212 | 691 | #ifdef CONFIG_NET_NS |
5f256bec EB |
692 | net_cachep = kmem_cache_create("net_namespace", sizeof(struct net), |
693 | SMP_CACHE_BYTES, | |
694 | SLAB_PANIC, NULL); | |
3ef1355d BT |
695 | |
696 | /* Create workqueue for cleanup */ | |
697 | netns_wq = create_singlethread_workqueue("netns"); | |
698 | if (!netns_wq) | |
699 | panic("Could not create netns workq"); | |
d57a9212 | 700 | #endif |
3ef1355d | 701 | |
486a87f1 DL |
702 | ng = net_alloc_generic(); |
703 | if (!ng) | |
704 | panic("Could not allocate generic netns"); | |
705 | ||
706 | rcu_assign_pointer(init_net.gen, ng); | |
707 | ||
5f256bec | 708 | mutex_lock(&net_mutex); |
038e7332 | 709 | if (setup_net(&init_net, &init_user_ns)) |
ca0f3112 | 710 | panic("Could not setup the initial network namespace"); |
5f256bec | 711 | |
f4618d39 | 712 | rtnl_lock(); |
11a28d37 | 713 | list_add_tail_rcu(&init_net.list, &net_namespace_list); |
f4618d39 | 714 | rtnl_unlock(); |
5f256bec EB |
715 | |
716 | mutex_unlock(&net_mutex); | |
5f256bec | 717 | |
98f842e6 EB |
718 | register_pernet_subsys(&net_ns_ops); |
719 | ||
0c7aecd4 | 720 | rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL); |
a143c40c ND |
721 | rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid, |
722 | NULL); | |
0c7aecd4 | 723 | |
5f256bec EB |
724 | return 0; |
725 | } | |
726 | ||
727 | pure_initcall(net_ns_init); | |
728 | ||
ed160e83 | 729 | #ifdef CONFIG_NET_NS |
f875bae0 EB |
730 | static int __register_pernet_operations(struct list_head *list, |
731 | struct pernet_operations *ops) | |
5f256bec | 732 | { |
72ad937a | 733 | struct net *net; |
5f256bec | 734 | int error; |
72ad937a | 735 | LIST_HEAD(net_exit_list); |
5f256bec | 736 | |
5f256bec | 737 | list_add_tail(&ops->list, list); |
f875bae0 | 738 | if (ops->init || (ops->id && ops->size)) { |
1dba323b | 739 | for_each_net(net) { |
f875bae0 | 740 | error = ops_init(ops, net); |
5f256bec EB |
741 | if (error) |
742 | goto out_undo; | |
72ad937a | 743 | list_add_tail(&net->exit_list, &net_exit_list); |
5f256bec EB |
744 | } |
745 | } | |
1dba323b | 746 | return 0; |
5f256bec EB |
747 | |
748 | out_undo: | |
749 | /* If I have an error cleanup all namespaces I initialized */ | |
750 | list_del(&ops->list); | |
72ad937a EB |
751 | ops_exit_list(ops, &net_exit_list); |
752 | ops_free_list(ops, &net_exit_list); | |
1dba323b | 753 | return error; |
5f256bec EB |
754 | } |
755 | ||
f875bae0 | 756 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
5f256bec EB |
757 | { |
758 | struct net *net; | |
72ad937a | 759 | LIST_HEAD(net_exit_list); |
5f256bec EB |
760 | |
761 | list_del(&ops->list); | |
72ad937a EB |
762 | for_each_net(net) |
763 | list_add_tail(&net->exit_list, &net_exit_list); | |
764 | ops_exit_list(ops, &net_exit_list); | |
765 | ops_free_list(ops, &net_exit_list); | |
5f256bec EB |
766 | } |
767 | ||
ed160e83 DL |
768 | #else |
769 | ||
f875bae0 EB |
770 | static int __register_pernet_operations(struct list_head *list, |
771 | struct pernet_operations *ops) | |
ed160e83 | 772 | { |
b922934d | 773 | return ops_init(ops, &init_net); |
ed160e83 DL |
774 | } |
775 | ||
f875bae0 | 776 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
ed160e83 | 777 | { |
72ad937a EB |
778 | LIST_HEAD(net_exit_list); |
779 | list_add(&init_net.exit_list, &net_exit_list); | |
780 | ops_exit_list(ops, &net_exit_list); | |
781 | ops_free_list(ops, &net_exit_list); | |
ed160e83 | 782 | } |
f875bae0 EB |
783 | |
784 | #endif /* CONFIG_NET_NS */ | |
ed160e83 | 785 | |
c93cf61f PE |
786 | static DEFINE_IDA(net_generic_ids); |
787 | ||
f875bae0 EB |
788 | static int register_pernet_operations(struct list_head *list, |
789 | struct pernet_operations *ops) | |
790 | { | |
791 | int error; | |
792 | ||
793 | if (ops->id) { | |
794 | again: | |
795 | error = ida_get_new_above(&net_generic_ids, 1, ops->id); | |
796 | if (error < 0) { | |
797 | if (error == -EAGAIN) { | |
798 | ida_pre_get(&net_generic_ids, GFP_KERNEL); | |
799 | goto again; | |
800 | } | |
801 | return error; | |
802 | } | |
073862ba | 803 | max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id); |
f875bae0 EB |
804 | } |
805 | error = __register_pernet_operations(list, ops); | |
3a765eda EB |
806 | if (error) { |
807 | rcu_barrier(); | |
808 | if (ops->id) | |
809 | ida_remove(&net_generic_ids, *ops->id); | |
810 | } | |
f875bae0 EB |
811 | |
812 | return error; | |
813 | } | |
814 | ||
815 | static void unregister_pernet_operations(struct pernet_operations *ops) | |
816 | { | |
817 | ||
818 | __unregister_pernet_operations(ops); | |
3a765eda | 819 | rcu_barrier(); |
f875bae0 EB |
820 | if (ops->id) |
821 | ida_remove(&net_generic_ids, *ops->id); | |
822 | } | |
823 | ||
5f256bec EB |
824 | /** |
825 | * register_pernet_subsys - register a network namespace subsystem | |
826 | * @ops: pernet operations structure for the subsystem | |
827 | * | |
828 | * Register a subsystem 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_subsys(struct pernet_operations *ops) | |
844 | { | |
845 | int error; | |
846 | mutex_lock(&net_mutex); | |
847 | error = register_pernet_operations(first_device, ops); | |
848 | mutex_unlock(&net_mutex); | |
849 | return error; | |
850 | } | |
851 | EXPORT_SYMBOL_GPL(register_pernet_subsys); | |
852 | ||
853 | /** | |
854 | * unregister_pernet_subsys - unregister a network namespace subsystem | |
855 | * @ops: pernet operations structure to manipulate | |
856 | * | |
857 | * Remove the pernet operations structure from the list to be | |
53379e57 | 858 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
859 | * addition run the exit method for all existing network |
860 | * namespaces. | |
861 | */ | |
b3c981d2 | 862 | void unregister_pernet_subsys(struct pernet_operations *ops) |
5f256bec EB |
863 | { |
864 | mutex_lock(&net_mutex); | |
b3c981d2 | 865 | unregister_pernet_operations(ops); |
5f256bec EB |
866 | mutex_unlock(&net_mutex); |
867 | } | |
868 | EXPORT_SYMBOL_GPL(unregister_pernet_subsys); | |
869 | ||
870 | /** | |
871 | * register_pernet_device - register a network namespace device | |
872 | * @ops: pernet operations structure for the subsystem | |
873 | * | |
874 | * Register a device which has init and exit functions | |
875 | * that are called when network namespaces are created and | |
876 | * destroyed respectively. | |
877 | * | |
878 | * When registered all network namespace init functions are | |
879 | * called for every existing network namespace. Allowing kernel | |
880 | * modules to have a race free view of the set of network namespaces. | |
881 | * | |
882 | * When a new network namespace is created all of the init | |
883 | * methods are called in the order in which they were registered. | |
884 | * | |
885 | * When a network namespace is destroyed all of the exit methods | |
886 | * are called in the reverse of the order with which they were | |
887 | * registered. | |
888 | */ | |
889 | int register_pernet_device(struct pernet_operations *ops) | |
890 | { | |
891 | int error; | |
892 | mutex_lock(&net_mutex); | |
893 | error = register_pernet_operations(&pernet_list, ops); | |
894 | if (!error && (first_device == &pernet_list)) | |
895 | first_device = &ops->list; | |
896 | mutex_unlock(&net_mutex); | |
897 | return error; | |
898 | } | |
899 | EXPORT_SYMBOL_GPL(register_pernet_device); | |
900 | ||
901 | /** | |
902 | * unregister_pernet_device - unregister a network namespace netdevice | |
903 | * @ops: pernet operations structure to manipulate | |
904 | * | |
905 | * Remove the pernet operations structure from the list to be | |
53379e57 | 906 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
907 | * addition run the exit method for all existing network |
908 | * namespaces. | |
909 | */ | |
910 | void unregister_pernet_device(struct pernet_operations *ops) | |
911 | { | |
912 | mutex_lock(&net_mutex); | |
913 | if (&ops->list == first_device) | |
914 | first_device = first_device->next; | |
915 | unregister_pernet_operations(ops); | |
916 | mutex_unlock(&net_mutex); | |
917 | } | |
918 | EXPORT_SYMBOL_GPL(unregister_pernet_device); | |
13b6f576 EB |
919 | |
920 | #ifdef CONFIG_NET_NS | |
64964528 | 921 | static struct ns_common *netns_get(struct task_struct *task) |
13b6f576 | 922 | { |
f0630529 EB |
923 | struct net *net = NULL; |
924 | struct nsproxy *nsproxy; | |
925 | ||
728dba3a EB |
926 | task_lock(task); |
927 | nsproxy = task->nsproxy; | |
f0630529 EB |
928 | if (nsproxy) |
929 | net = get_net(nsproxy->net_ns); | |
728dba3a | 930 | task_unlock(task); |
f0630529 | 931 | |
ff24870f AV |
932 | return net ? &net->ns : NULL; |
933 | } | |
934 | ||
935 | static inline struct net *to_net_ns(struct ns_common *ns) | |
936 | { | |
937 | return container_of(ns, struct net, ns); | |
13b6f576 EB |
938 | } |
939 | ||
64964528 | 940 | static void netns_put(struct ns_common *ns) |
13b6f576 | 941 | { |
ff24870f | 942 | put_net(to_net_ns(ns)); |
13b6f576 EB |
943 | } |
944 | ||
64964528 | 945 | static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns) |
13b6f576 | 946 | { |
ff24870f | 947 | struct net *net = to_net_ns(ns); |
142e1d1d | 948 | |
5e4a0847 | 949 | if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) || |
c7b96acf | 950 | !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) |
142e1d1d EB |
951 | return -EPERM; |
952 | ||
13b6f576 | 953 | put_net(nsproxy->net_ns); |
142e1d1d | 954 | nsproxy->net_ns = get_net(net); |
13b6f576 EB |
955 | return 0; |
956 | } | |
957 | ||
958 | const struct proc_ns_operations netns_operations = { | |
959 | .name = "net", | |
960 | .type = CLONE_NEWNET, | |
961 | .get = netns_get, | |
962 | .put = netns_put, | |
963 | .install = netns_install, | |
964 | }; | |
965 | #endif |