]>
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 | ||
cab3c8ec | 150 | static void rtnl_net_notifyid(struct net *net, int cmd, int id); |
0c7aecd4 ND |
151 | static int alloc_netid(struct net *net, struct net *peer, int reqid) |
152 | { | |
9a963454 | 153 | int min = 0, max = 0, id; |
0c7aecd4 ND |
154 | |
155 | ASSERT_RTNL(); | |
156 | ||
157 | if (reqid >= 0) { | |
158 | min = reqid; | |
159 | max = reqid + 1; | |
160 | } | |
161 | ||
9a963454 ND |
162 | id = idr_alloc(&net->netns_ids, peer, min, max, GFP_KERNEL); |
163 | if (id >= 0) | |
cab3c8ec | 164 | rtnl_net_notifyid(net, RTM_NEWNSID, id); |
9a963454 ND |
165 | |
166 | return id; | |
0c7aecd4 ND |
167 | } |
168 | ||
169 | /* This function is used by idr_for_each(). If net is equal to peer, the | |
170 | * function returns the id so that idr_for_each() stops. Because we cannot | |
171 | * returns the id 0 (idr_for_each() will not stop), we return the magic value | |
172 | * NET_ID_ZERO (-1) for it. | |
173 | */ | |
174 | #define NET_ID_ZERO -1 | |
175 | static int net_eq_idr(int id, void *net, void *peer) | |
176 | { | |
177 | if (net_eq(net, peer)) | |
178 | return id ? : NET_ID_ZERO; | |
179 | return 0; | |
180 | } | |
181 | ||
182 | static int __peernet2id(struct net *net, struct net *peer, bool alloc) | |
183 | { | |
184 | int id = idr_for_each(&net->netns_ids, net_eq_idr, peer); | |
185 | ||
186 | ASSERT_RTNL(); | |
187 | ||
188 | /* Magic value for id 0. */ | |
189 | if (id == NET_ID_ZERO) | |
190 | return 0; | |
191 | if (id > 0) | |
192 | return id; | |
193 | ||
109582af ND |
194 | if (alloc) { |
195 | id = alloc_netid(net, peer, -1); | |
196 | return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED; | |
197 | } | |
0c7aecd4 | 198 | |
109582af | 199 | return NETNSA_NSID_NOT_ASSIGNED; |
0c7aecd4 ND |
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 | 207 | bool alloc = atomic_read(&peer->count) == 0 ? false : true; |
0c7aecd4 | 208 | |
109582af | 209 | return __peernet2id(net, peer, alloc); |
0c7aecd4 | 210 | } |
193523bf | 211 | EXPORT_SYMBOL(peernet2id); |
0c7aecd4 ND |
212 | |
213 | struct net *get_net_ns_by_id(struct net *net, int id) | |
214 | { | |
215 | struct net *peer; | |
216 | ||
217 | if (id < 0) | |
218 | return NULL; | |
219 | ||
220 | rcu_read_lock(); | |
221 | peer = idr_find(&net->netns_ids, id); | |
222 | if (peer) | |
223 | get_net(peer); | |
224 | rcu_read_unlock(); | |
225 | ||
226 | return peer; | |
227 | } | |
228 | ||
5f256bec EB |
229 | /* |
230 | * setup_net runs the initializers for the network namespace object. | |
231 | */ | |
038e7332 | 232 | static __net_init int setup_net(struct net *net, struct user_namespace *user_ns) |
5f256bec EB |
233 | { |
234 | /* Must be called with net_mutex held */ | |
f875bae0 | 235 | const struct pernet_operations *ops, *saved_ops; |
486a87f1 | 236 | int error = 0; |
72ad937a | 237 | LIST_HEAD(net_exit_list); |
5f256bec | 238 | |
5f256bec | 239 | atomic_set(&net->count, 1); |
a685e089 | 240 | atomic_set(&net->passive, 1); |
4e985ada | 241 | net->dev_base_seq = 1; |
038e7332 | 242 | net->user_ns = user_ns; |
0c7aecd4 | 243 | idr_init(&net->netns_ids); |
486a87f1 | 244 | |
768f3591 | 245 | list_for_each_entry(ops, &pernet_list, list) { |
f875bae0 EB |
246 | error = ops_init(ops, net); |
247 | if (error < 0) | |
248 | goto out_undo; | |
5f256bec EB |
249 | } |
250 | out: | |
251 | return error; | |
768f3591 | 252 | |
5f256bec EB |
253 | out_undo: |
254 | /* Walk through the list backwards calling the exit functions | |
255 | * for the pernet modules whose init functions did not fail. | |
256 | */ | |
72ad937a | 257 | list_add(&net->exit_list, &net_exit_list); |
f875bae0 | 258 | saved_ops = ops; |
72ad937a EB |
259 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) |
260 | ops_exit_list(ops, &net_exit_list); | |
261 | ||
f875bae0 EB |
262 | ops = saved_ops; |
263 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) | |
72ad937a | 264 | ops_free_list(ops, &net_exit_list); |
310928d9 DL |
265 | |
266 | rcu_barrier(); | |
5f256bec EB |
267 | goto out; |
268 | } | |
269 | ||
6a1a3b9f | 270 | |
ebe47d47 CN |
271 | #ifdef CONFIG_NET_NS |
272 | static struct kmem_cache *net_cachep; | |
273 | static struct workqueue_struct *netns_wq; | |
274 | ||
486a87f1 | 275 | static struct net *net_alloc(void) |
45a19b0a | 276 | { |
486a87f1 DL |
277 | struct net *net = NULL; |
278 | struct net_generic *ng; | |
279 | ||
280 | ng = net_alloc_generic(); | |
281 | if (!ng) | |
282 | goto out; | |
283 | ||
284 | net = kmem_cache_zalloc(net_cachep, GFP_KERNEL); | |
45a19b0a | 285 | if (!net) |
486a87f1 | 286 | goto out_free; |
45a19b0a | 287 | |
486a87f1 DL |
288 | rcu_assign_pointer(net->gen, ng); |
289 | out: | |
290 | return net; | |
291 | ||
292 | out_free: | |
293 | kfree(ng); | |
294 | goto out; | |
295 | } | |
296 | ||
297 | static void net_free(struct net *net) | |
298 | { | |
416c51e1 | 299 | kfree(rcu_access_pointer(net->gen)); |
45a19b0a JFS |
300 | kmem_cache_free(net_cachep, net); |
301 | } | |
302 | ||
a685e089 AV |
303 | void net_drop_ns(void *p) |
304 | { | |
305 | struct net *ns = p; | |
306 | if (ns && atomic_dec_and_test(&ns->passive)) | |
307 | net_free(ns); | |
308 | } | |
309 | ||
038e7332 EB |
310 | struct net *copy_net_ns(unsigned long flags, |
311 | struct user_namespace *user_ns, struct net *old_net) | |
9dd776b6 | 312 | { |
088eb2d9 AD |
313 | struct net *net; |
314 | int rv; | |
9dd776b6 | 315 | |
911cb193 RL |
316 | if (!(flags & CLONE_NEWNET)) |
317 | return get_net(old_net); | |
318 | ||
088eb2d9 AD |
319 | net = net_alloc(); |
320 | if (!net) | |
321 | return ERR_PTR(-ENOMEM); | |
038e7332 EB |
322 | |
323 | get_user_ns(user_ns); | |
324 | ||
9dd776b6 | 325 | mutex_lock(&net_mutex); |
038e7332 | 326 | rv = setup_net(net, user_ns); |
088eb2d9 | 327 | if (rv == 0) { |
486a87f1 | 328 | rtnl_lock(); |
11a28d37 | 329 | list_add_tail_rcu(&net->list, &net_namespace_list); |
486a87f1 DL |
330 | rtnl_unlock(); |
331 | } | |
9dd776b6 | 332 | mutex_unlock(&net_mutex); |
088eb2d9 | 333 | if (rv < 0) { |
038e7332 | 334 | put_user_ns(user_ns); |
a685e089 | 335 | net_drop_ns(net); |
088eb2d9 AD |
336 | return ERR_PTR(rv); |
337 | } | |
338 | return net; | |
339 | } | |
486a87f1 | 340 | |
2b035b39 EB |
341 | static DEFINE_SPINLOCK(cleanup_list_lock); |
342 | static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */ | |
343 | ||
6a1a3b9f PE |
344 | static void cleanup_net(struct work_struct *work) |
345 | { | |
f875bae0 | 346 | const struct pernet_operations *ops; |
6d458f5b | 347 | struct net *net, *tmp; |
1818ce4d | 348 | struct list_head net_kill_list; |
72ad937a | 349 | LIST_HEAD(net_exit_list); |
6a1a3b9f | 350 | |
2b035b39 EB |
351 | /* Atomically snapshot the list of namespaces to cleanup */ |
352 | spin_lock_irq(&cleanup_list_lock); | |
353 | list_replace_init(&cleanup_list, &net_kill_list); | |
354 | spin_unlock_irq(&cleanup_list_lock); | |
6a1a3b9f PE |
355 | |
356 | mutex_lock(&net_mutex); | |
357 | ||
358 | /* Don't let anyone else find us. */ | |
359 | rtnl_lock(); | |
72ad937a | 360 | list_for_each_entry(net, &net_kill_list, cleanup_list) { |
2b035b39 | 361 | list_del_rcu(&net->list); |
72ad937a | 362 | list_add_tail(&net->exit_list, &net_exit_list); |
6d458f5b ND |
363 | for_each_net(tmp) { |
364 | int id = __peernet2id(tmp, net, false); | |
365 | ||
9a963454 | 366 | if (id >= 0) { |
cab3c8ec | 367 | rtnl_net_notifyid(tmp, RTM_DELNSID, id); |
6d458f5b | 368 | idr_remove(&tmp->netns_ids, id); |
9a963454 | 369 | } |
6d458f5b ND |
370 | } |
371 | idr_destroy(&net->netns_ids); | |
372 | ||
72ad937a | 373 | } |
6a1a3b9f PE |
374 | rtnl_unlock(); |
375 | ||
11a28d37 JB |
376 | /* |
377 | * Another CPU might be rcu-iterating the list, wait for it. | |
378 | * This needs to be before calling the exit() notifiers, so | |
379 | * the rcu_barrier() below isn't sufficient alone. | |
380 | */ | |
381 | synchronize_rcu(); | |
382 | ||
6a1a3b9f | 383 | /* Run all of the network namespace exit methods */ |
72ad937a EB |
384 | list_for_each_entry_reverse(ops, &pernet_list, list) |
385 | ops_exit_list(ops, &net_exit_list); | |
386 | ||
f875bae0 | 387 | /* Free the net generic variables */ |
72ad937a EB |
388 | list_for_each_entry_reverse(ops, &pernet_list, list) |
389 | ops_free_list(ops, &net_exit_list); | |
6a1a3b9f PE |
390 | |
391 | mutex_unlock(&net_mutex); | |
392 | ||
393 | /* Ensure there are no outstanding rcu callbacks using this | |
394 | * network namespace. | |
395 | */ | |
396 | rcu_barrier(); | |
397 | ||
398 | /* Finally it is safe to free my network namespace structure */ | |
72ad937a EB |
399 | list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) { |
400 | list_del_init(&net->exit_list); | |
038e7332 | 401 | put_user_ns(net->user_ns); |
a685e089 | 402 | net_drop_ns(net); |
2b035b39 | 403 | } |
6a1a3b9f | 404 | } |
2b035b39 | 405 | static DECLARE_WORK(net_cleanup_work, cleanup_net); |
6a1a3b9f PE |
406 | |
407 | void __put_net(struct net *net) | |
408 | { | |
409 | /* Cleanup the network namespace in process context */ | |
2b035b39 EB |
410 | unsigned long flags; |
411 | ||
412 | spin_lock_irqsave(&cleanup_list_lock, flags); | |
413 | list_add(&net->cleanup_list, &cleanup_list); | |
414 | spin_unlock_irqrestore(&cleanup_list_lock, flags); | |
415 | ||
416 | queue_work(netns_wq, &net_cleanup_work); | |
6a1a3b9f PE |
417 | } |
418 | EXPORT_SYMBOL_GPL(__put_net); | |
419 | ||
956c9207 SR |
420 | struct net *get_net_ns_by_fd(int fd) |
421 | { | |
956c9207 | 422 | struct file *file; |
33c42940 | 423 | struct ns_common *ns; |
956c9207 SR |
424 | struct net *net; |
425 | ||
956c9207 | 426 | file = proc_ns_fget(fd); |
c316e6a3 AV |
427 | if (IS_ERR(file)) |
428 | return ERR_CAST(file); | |
956c9207 | 429 | |
f77c8014 | 430 | ns = get_proc_ns(file_inode(file)); |
33c42940 AV |
431 | if (ns->ops == &netns_operations) |
432 | net = get_net(container_of(ns, struct net, ns)); | |
c316e6a3 AV |
433 | else |
434 | net = ERR_PTR(-EINVAL); | |
956c9207 | 435 | |
c316e6a3 | 436 | fput(file); |
956c9207 SR |
437 | return net; |
438 | } | |
439 | ||
6a1a3b9f | 440 | #else |
956c9207 SR |
441 | struct net *get_net_ns_by_fd(int fd) |
442 | { | |
443 | return ERR_PTR(-EINVAL); | |
444 | } | |
6a1a3b9f | 445 | #endif |
4b681c82 | 446 | EXPORT_SYMBOL_GPL(get_net_ns_by_fd); |
6a1a3b9f | 447 | |
30ffee84 JB |
448 | struct net *get_net_ns_by_pid(pid_t pid) |
449 | { | |
450 | struct task_struct *tsk; | |
451 | struct net *net; | |
452 | ||
453 | /* Lookup the network namespace */ | |
454 | net = ERR_PTR(-ESRCH); | |
455 | rcu_read_lock(); | |
456 | tsk = find_task_by_vpid(pid); | |
457 | if (tsk) { | |
458 | struct nsproxy *nsproxy; | |
728dba3a EB |
459 | task_lock(tsk); |
460 | nsproxy = tsk->nsproxy; | |
30ffee84 JB |
461 | if (nsproxy) |
462 | net = get_net(nsproxy->net_ns); | |
728dba3a | 463 | task_unlock(tsk); |
30ffee84 JB |
464 | } |
465 | rcu_read_unlock(); | |
466 | return net; | |
467 | } | |
468 | EXPORT_SYMBOL_GPL(get_net_ns_by_pid); | |
469 | ||
98f842e6 EB |
470 | static __net_init int net_ns_net_init(struct net *net) |
471 | { | |
33c42940 AV |
472 | #ifdef CONFIG_NET_NS |
473 | net->ns.ops = &netns_operations; | |
474 | #endif | |
6344c433 | 475 | return ns_alloc_inum(&net->ns); |
98f842e6 EB |
476 | } |
477 | ||
478 | static __net_exit void net_ns_net_exit(struct net *net) | |
479 | { | |
6344c433 | 480 | ns_free_inum(&net->ns); |
98f842e6 EB |
481 | } |
482 | ||
483 | static struct pernet_operations __net_initdata net_ns_ops = { | |
484 | .init = net_ns_net_init, | |
485 | .exit = net_ns_net_exit, | |
486 | }; | |
487 | ||
0c7aecd4 ND |
488 | static struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = { |
489 | [NETNSA_NONE] = { .type = NLA_UNSPEC }, | |
490 | [NETNSA_NSID] = { .type = NLA_S32 }, | |
491 | [NETNSA_PID] = { .type = NLA_U32 }, | |
492 | [NETNSA_FD] = { .type = NLA_U32 }, | |
493 | }; | |
494 | ||
495 | static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
496 | { | |
497 | struct net *net = sock_net(skb->sk); | |
498 | struct nlattr *tb[NETNSA_MAX + 1]; | |
499 | struct net *peer; | |
500 | int nsid, err; | |
501 | ||
502 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
503 | rtnl_net_policy); | |
504 | if (err < 0) | |
505 | return err; | |
506 | if (!tb[NETNSA_NSID]) | |
507 | return -EINVAL; | |
508 | nsid = nla_get_s32(tb[NETNSA_NSID]); | |
509 | ||
510 | if (tb[NETNSA_PID]) | |
511 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); | |
512 | else if (tb[NETNSA_FD]) | |
513 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); | |
514 | else | |
515 | return -EINVAL; | |
516 | if (IS_ERR(peer)) | |
517 | return PTR_ERR(peer); | |
518 | ||
519 | if (__peernet2id(net, peer, false) >= 0) { | |
520 | err = -EEXIST; | |
521 | goto out; | |
522 | } | |
523 | ||
524 | err = alloc_netid(net, peer, nsid); | |
525 | if (err > 0) | |
526 | err = 0; | |
527 | out: | |
528 | put_net(peer); | |
529 | return err; | |
530 | } | |
531 | ||
532 | static int rtnl_net_get_size(void) | |
533 | { | |
534 | return NLMSG_ALIGN(sizeof(struct rtgenmsg)) | |
535 | + nla_total_size(sizeof(s32)) /* NETNSA_NSID */ | |
536 | ; | |
537 | } | |
538 | ||
539 | static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags, | |
cab3c8ec | 540 | int cmd, struct net *net, int nsid) |
0c7aecd4 ND |
541 | { |
542 | struct nlmsghdr *nlh; | |
543 | struct rtgenmsg *rth; | |
0c7aecd4 ND |
544 | |
545 | nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags); | |
546 | if (!nlh) | |
547 | return -EMSGSIZE; | |
548 | ||
549 | rth = nlmsg_data(nlh); | |
550 | rth->rtgen_family = AF_UNSPEC; | |
551 | ||
cab3c8ec | 552 | if (nla_put_s32(skb, NETNSA_NSID, nsid)) |
0c7aecd4 ND |
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; |
cab3c8ec | 569 | int err, id; |
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 | ||
cab3c8ec | 591 | id = __peernet2id(net, peer, false); |
0c7aecd4 | 592 | err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, |
cab3c8ec | 593 | RTM_GETNSID, net, id); |
0c7aecd4 ND |
594 | if (err < 0) |
595 | goto err_out; | |
596 | ||
597 | err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid); | |
598 | goto out; | |
599 | ||
600 | err_out: | |
601 | nlmsg_free(msg); | |
602 | out: | |
603 | put_net(peer); | |
604 | return err; | |
605 | } | |
606 | ||
a143c40c ND |
607 | struct rtnl_net_dump_cb { |
608 | struct net *net; | |
609 | struct sk_buff *skb; | |
610 | struct netlink_callback *cb; | |
611 | int idx; | |
612 | int s_idx; | |
613 | }; | |
614 | ||
615 | static int rtnl_net_dumpid_one(int id, void *peer, void *data) | |
616 | { | |
617 | struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data; | |
618 | int ret; | |
619 | ||
620 | if (net_cb->idx < net_cb->s_idx) | |
621 | goto cont; | |
622 | ||
623 | ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid, | |
624 | net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI, | |
cab3c8ec | 625 | RTM_NEWNSID, net_cb->net, id); |
a143c40c ND |
626 | if (ret < 0) |
627 | return ret; | |
628 | ||
629 | cont: | |
630 | net_cb->idx++; | |
631 | return 0; | |
632 | } | |
633 | ||
634 | static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb) | |
635 | { | |
636 | struct net *net = sock_net(skb->sk); | |
637 | struct rtnl_net_dump_cb net_cb = { | |
638 | .net = net, | |
639 | .skb = skb, | |
640 | .cb = cb, | |
641 | .idx = 0, | |
642 | .s_idx = cb->args[0], | |
643 | }; | |
644 | ||
645 | ASSERT_RTNL(); | |
646 | ||
647 | idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb); | |
648 | ||
649 | cb->args[0] = net_cb.idx; | |
650 | return skb->len; | |
651 | } | |
652 | ||
cab3c8ec | 653 | static void rtnl_net_notifyid(struct net *net, int cmd, int id) |
9a963454 ND |
654 | { |
655 | struct sk_buff *msg; | |
656 | int err = -ENOMEM; | |
657 | ||
658 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
659 | if (!msg) | |
660 | goto out; | |
661 | ||
cab3c8ec | 662 | err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, id); |
9a963454 ND |
663 | if (err < 0) |
664 | goto err_out; | |
665 | ||
666 | rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0); | |
667 | return; | |
668 | ||
669 | err_out: | |
670 | nlmsg_free(msg); | |
671 | out: | |
672 | rtnl_set_sk_err(net, RTNLGRP_NSID, err); | |
673 | } | |
674 | ||
5f256bec EB |
675 | static int __init net_ns_init(void) |
676 | { | |
486a87f1 | 677 | struct net_generic *ng; |
5f256bec | 678 | |
d57a9212 | 679 | #ifdef CONFIG_NET_NS |
5f256bec EB |
680 | net_cachep = kmem_cache_create("net_namespace", sizeof(struct net), |
681 | SMP_CACHE_BYTES, | |
682 | SLAB_PANIC, NULL); | |
3ef1355d BT |
683 | |
684 | /* Create workqueue for cleanup */ | |
685 | netns_wq = create_singlethread_workqueue("netns"); | |
686 | if (!netns_wq) | |
687 | panic("Could not create netns workq"); | |
d57a9212 | 688 | #endif |
3ef1355d | 689 | |
486a87f1 DL |
690 | ng = net_alloc_generic(); |
691 | if (!ng) | |
692 | panic("Could not allocate generic netns"); | |
693 | ||
694 | rcu_assign_pointer(init_net.gen, ng); | |
695 | ||
5f256bec | 696 | mutex_lock(&net_mutex); |
038e7332 | 697 | if (setup_net(&init_net, &init_user_ns)) |
ca0f3112 | 698 | panic("Could not setup the initial network namespace"); |
5f256bec | 699 | |
f4618d39 | 700 | rtnl_lock(); |
11a28d37 | 701 | list_add_tail_rcu(&init_net.list, &net_namespace_list); |
f4618d39 | 702 | rtnl_unlock(); |
5f256bec EB |
703 | |
704 | mutex_unlock(&net_mutex); | |
5f256bec | 705 | |
98f842e6 EB |
706 | register_pernet_subsys(&net_ns_ops); |
707 | ||
0c7aecd4 | 708 | rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL); |
a143c40c ND |
709 | rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid, |
710 | NULL); | |
0c7aecd4 | 711 | |
5f256bec EB |
712 | return 0; |
713 | } | |
714 | ||
715 | pure_initcall(net_ns_init); | |
716 | ||
ed160e83 | 717 | #ifdef CONFIG_NET_NS |
f875bae0 EB |
718 | static int __register_pernet_operations(struct list_head *list, |
719 | struct pernet_operations *ops) | |
5f256bec | 720 | { |
72ad937a | 721 | struct net *net; |
5f256bec | 722 | int error; |
72ad937a | 723 | LIST_HEAD(net_exit_list); |
5f256bec | 724 | |
5f256bec | 725 | list_add_tail(&ops->list, list); |
f875bae0 | 726 | if (ops->init || (ops->id && ops->size)) { |
1dba323b | 727 | for_each_net(net) { |
f875bae0 | 728 | error = ops_init(ops, net); |
5f256bec EB |
729 | if (error) |
730 | goto out_undo; | |
72ad937a | 731 | list_add_tail(&net->exit_list, &net_exit_list); |
5f256bec EB |
732 | } |
733 | } | |
1dba323b | 734 | return 0; |
5f256bec EB |
735 | |
736 | out_undo: | |
737 | /* If I have an error cleanup all namespaces I initialized */ | |
738 | list_del(&ops->list); | |
72ad937a EB |
739 | ops_exit_list(ops, &net_exit_list); |
740 | ops_free_list(ops, &net_exit_list); | |
1dba323b | 741 | return error; |
5f256bec EB |
742 | } |
743 | ||
f875bae0 | 744 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
5f256bec EB |
745 | { |
746 | struct net *net; | |
72ad937a | 747 | LIST_HEAD(net_exit_list); |
5f256bec EB |
748 | |
749 | list_del(&ops->list); | |
72ad937a EB |
750 | for_each_net(net) |
751 | list_add_tail(&net->exit_list, &net_exit_list); | |
752 | ops_exit_list(ops, &net_exit_list); | |
753 | ops_free_list(ops, &net_exit_list); | |
5f256bec EB |
754 | } |
755 | ||
ed160e83 DL |
756 | #else |
757 | ||
f875bae0 EB |
758 | static int __register_pernet_operations(struct list_head *list, |
759 | struct pernet_operations *ops) | |
ed160e83 | 760 | { |
b922934d | 761 | return ops_init(ops, &init_net); |
ed160e83 DL |
762 | } |
763 | ||
f875bae0 | 764 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
ed160e83 | 765 | { |
72ad937a EB |
766 | LIST_HEAD(net_exit_list); |
767 | list_add(&init_net.exit_list, &net_exit_list); | |
768 | ops_exit_list(ops, &net_exit_list); | |
769 | ops_free_list(ops, &net_exit_list); | |
ed160e83 | 770 | } |
f875bae0 EB |
771 | |
772 | #endif /* CONFIG_NET_NS */ | |
ed160e83 | 773 | |
c93cf61f PE |
774 | static DEFINE_IDA(net_generic_ids); |
775 | ||
f875bae0 EB |
776 | static int register_pernet_operations(struct list_head *list, |
777 | struct pernet_operations *ops) | |
778 | { | |
779 | int error; | |
780 | ||
781 | if (ops->id) { | |
782 | again: | |
783 | error = ida_get_new_above(&net_generic_ids, 1, ops->id); | |
784 | if (error < 0) { | |
785 | if (error == -EAGAIN) { | |
786 | ida_pre_get(&net_generic_ids, GFP_KERNEL); | |
787 | goto again; | |
788 | } | |
789 | return error; | |
790 | } | |
073862ba | 791 | max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id); |
f875bae0 EB |
792 | } |
793 | error = __register_pernet_operations(list, ops); | |
3a765eda EB |
794 | if (error) { |
795 | rcu_barrier(); | |
796 | if (ops->id) | |
797 | ida_remove(&net_generic_ids, *ops->id); | |
798 | } | |
f875bae0 EB |
799 | |
800 | return error; | |
801 | } | |
802 | ||
803 | static void unregister_pernet_operations(struct pernet_operations *ops) | |
804 | { | |
805 | ||
806 | __unregister_pernet_operations(ops); | |
3a765eda | 807 | rcu_barrier(); |
f875bae0 EB |
808 | if (ops->id) |
809 | ida_remove(&net_generic_ids, *ops->id); | |
810 | } | |
811 | ||
5f256bec EB |
812 | /** |
813 | * register_pernet_subsys - register a network namespace subsystem | |
814 | * @ops: pernet operations structure for the subsystem | |
815 | * | |
816 | * Register a subsystem which has init and exit functions | |
817 | * that are called when network namespaces are created and | |
818 | * destroyed respectively. | |
819 | * | |
820 | * When registered all network namespace init functions are | |
821 | * called for every existing network namespace. Allowing kernel | |
822 | * modules to have a race free view of the set of network namespaces. | |
823 | * | |
824 | * When a new network namespace is created all of the init | |
825 | * methods are called in the order in which they were registered. | |
826 | * | |
827 | * When a network namespace is destroyed all of the exit methods | |
828 | * are called in the reverse of the order with which they were | |
829 | * registered. | |
830 | */ | |
831 | int register_pernet_subsys(struct pernet_operations *ops) | |
832 | { | |
833 | int error; | |
834 | mutex_lock(&net_mutex); | |
835 | error = register_pernet_operations(first_device, ops); | |
836 | mutex_unlock(&net_mutex); | |
837 | return error; | |
838 | } | |
839 | EXPORT_SYMBOL_GPL(register_pernet_subsys); | |
840 | ||
841 | /** | |
842 | * unregister_pernet_subsys - unregister a network namespace subsystem | |
843 | * @ops: pernet operations structure to manipulate | |
844 | * | |
845 | * Remove the pernet operations structure from the list to be | |
53379e57 | 846 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
847 | * addition run the exit method for all existing network |
848 | * namespaces. | |
849 | */ | |
b3c981d2 | 850 | void unregister_pernet_subsys(struct pernet_operations *ops) |
5f256bec EB |
851 | { |
852 | mutex_lock(&net_mutex); | |
b3c981d2 | 853 | unregister_pernet_operations(ops); |
5f256bec EB |
854 | mutex_unlock(&net_mutex); |
855 | } | |
856 | EXPORT_SYMBOL_GPL(unregister_pernet_subsys); | |
857 | ||
858 | /** | |
859 | * register_pernet_device - register a network namespace device | |
860 | * @ops: pernet operations structure for the subsystem | |
861 | * | |
862 | * Register a device which has init and exit functions | |
863 | * that are called when network namespaces are created and | |
864 | * destroyed respectively. | |
865 | * | |
866 | * When registered all network namespace init functions are | |
867 | * called for every existing network namespace. Allowing kernel | |
868 | * modules to have a race free view of the set of network namespaces. | |
869 | * | |
870 | * When a new network namespace is created all of the init | |
871 | * methods are called in the order in which they were registered. | |
872 | * | |
873 | * When a network namespace is destroyed all of the exit methods | |
874 | * are called in the reverse of the order with which they were | |
875 | * registered. | |
876 | */ | |
877 | int register_pernet_device(struct pernet_operations *ops) | |
878 | { | |
879 | int error; | |
880 | mutex_lock(&net_mutex); | |
881 | error = register_pernet_operations(&pernet_list, ops); | |
882 | if (!error && (first_device == &pernet_list)) | |
883 | first_device = &ops->list; | |
884 | mutex_unlock(&net_mutex); | |
885 | return error; | |
886 | } | |
887 | EXPORT_SYMBOL_GPL(register_pernet_device); | |
888 | ||
889 | /** | |
890 | * unregister_pernet_device - unregister a network namespace netdevice | |
891 | * @ops: pernet operations structure to manipulate | |
892 | * | |
893 | * Remove the pernet operations structure from the list to be | |
53379e57 | 894 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
895 | * addition run the exit method for all existing network |
896 | * namespaces. | |
897 | */ | |
898 | void unregister_pernet_device(struct pernet_operations *ops) | |
899 | { | |
900 | mutex_lock(&net_mutex); | |
901 | if (&ops->list == first_device) | |
902 | first_device = first_device->next; | |
903 | unregister_pernet_operations(ops); | |
904 | mutex_unlock(&net_mutex); | |
905 | } | |
906 | EXPORT_SYMBOL_GPL(unregister_pernet_device); | |
13b6f576 EB |
907 | |
908 | #ifdef CONFIG_NET_NS | |
64964528 | 909 | static struct ns_common *netns_get(struct task_struct *task) |
13b6f576 | 910 | { |
f0630529 EB |
911 | struct net *net = NULL; |
912 | struct nsproxy *nsproxy; | |
913 | ||
728dba3a EB |
914 | task_lock(task); |
915 | nsproxy = task->nsproxy; | |
f0630529 EB |
916 | if (nsproxy) |
917 | net = get_net(nsproxy->net_ns); | |
728dba3a | 918 | task_unlock(task); |
f0630529 | 919 | |
ff24870f AV |
920 | return net ? &net->ns : NULL; |
921 | } | |
922 | ||
923 | static inline struct net *to_net_ns(struct ns_common *ns) | |
924 | { | |
925 | return container_of(ns, struct net, ns); | |
13b6f576 EB |
926 | } |
927 | ||
64964528 | 928 | static void netns_put(struct ns_common *ns) |
13b6f576 | 929 | { |
ff24870f | 930 | put_net(to_net_ns(ns)); |
13b6f576 EB |
931 | } |
932 | ||
64964528 | 933 | static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns) |
13b6f576 | 934 | { |
ff24870f | 935 | struct net *net = to_net_ns(ns); |
142e1d1d | 936 | |
5e4a0847 | 937 | if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) || |
c7b96acf | 938 | !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) |
142e1d1d EB |
939 | return -EPERM; |
940 | ||
13b6f576 | 941 | put_net(nsproxy->net_ns); |
142e1d1d | 942 | nsproxy->net_ns = get_net(net); |
13b6f576 EB |
943 | return 0; |
944 | } | |
945 | ||
946 | const struct proc_ns_operations netns_operations = { | |
947 | .name = "net", | |
948 | .type = CLONE_NEWNET, | |
949 | .get = netns_get, | |
950 | .put = netns_put, | |
951 | .install = netns_install, | |
952 | }; | |
953 | #endif |