]>
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 | |
f8c46cb3 DT |
40 | static bool init_net_initialized; |
41 | ||
6af2d5ff AD |
42 | #define MIN_PERNET_OPS_ID \ |
43 | ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *)) | |
44 | ||
dec827d1 PE |
45 | #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */ |
46 | ||
073862ba ED |
47 | static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS; |
48 | ||
49 | static struct net_generic *net_alloc_generic(void) | |
50 | { | |
51 | struct net_generic *ng; | |
6af2d5ff | 52 | unsigned int generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]); |
073862ba ED |
53 | |
54 | ng = kzalloc(generic_size, GFP_KERNEL); | |
55 | if (ng) | |
9bfc7b99 | 56 | ng->s.len = max_gen_ptrs; |
073862ba ED |
57 | |
58 | return ng; | |
59 | } | |
60 | ||
c7d03a00 | 61 | static int net_assign_generic(struct net *net, unsigned int id, void *data) |
05fceb4a JP |
62 | { |
63 | struct net_generic *ng, *old_ng; | |
64 | ||
65 | BUG_ON(!mutex_is_locked(&net_mutex)); | |
6af2d5ff | 66 | BUG_ON(id < MIN_PERNET_OPS_ID); |
05fceb4a | 67 | |
1c87733d ED |
68 | old_ng = rcu_dereference_protected(net->gen, |
69 | lockdep_is_held(&net_mutex)); | |
6af2d5ff AD |
70 | if (old_ng->s.len > id) { |
71 | old_ng->ptr[id] = data; | |
1a9a0592 AD |
72 | return 0; |
73 | } | |
05fceb4a | 74 | |
073862ba | 75 | ng = net_alloc_generic(); |
05fceb4a JP |
76 | if (ng == NULL) |
77 | return -ENOMEM; | |
78 | ||
79 | /* | |
80 | * Some synchronisation notes: | |
81 | * | |
82 | * The net_generic explores the net->gen array inside rcu | |
83 | * read section. Besides once set the net->gen->ptr[x] | |
84 | * pointer never changes (see rules in netns/generic.h). | |
85 | * | |
86 | * That said, we simply duplicate this array and schedule | |
87 | * the old copy for kfree after a grace period. | |
88 | */ | |
89 | ||
6af2d5ff AD |
90 | memcpy(&ng->ptr[MIN_PERNET_OPS_ID], &old_ng->ptr[MIN_PERNET_OPS_ID], |
91 | (old_ng->s.len - MIN_PERNET_OPS_ID) * sizeof(void *)); | |
92 | ng->ptr[id] = data; | |
05fceb4a JP |
93 | |
94 | rcu_assign_pointer(net->gen, ng); | |
9bfc7b99 | 95 | kfree_rcu(old_ng, s.rcu); |
05fceb4a JP |
96 | return 0; |
97 | } | |
98 | ||
f875bae0 EB |
99 | static int ops_init(const struct pernet_operations *ops, struct net *net) |
100 | { | |
b922934d JA |
101 | int err = -ENOMEM; |
102 | void *data = NULL; | |
103 | ||
f875bae0 | 104 | if (ops->id && ops->size) { |
b922934d | 105 | data = kzalloc(ops->size, GFP_KERNEL); |
f875bae0 | 106 | if (!data) |
b922934d | 107 | goto out; |
f875bae0 EB |
108 | |
109 | err = net_assign_generic(net, *ops->id, data); | |
b922934d JA |
110 | if (err) |
111 | goto cleanup; | |
f875bae0 | 112 | } |
b922934d | 113 | err = 0; |
f875bae0 | 114 | if (ops->init) |
b922934d JA |
115 | err = ops->init(net); |
116 | if (!err) | |
117 | return 0; | |
118 | ||
119 | cleanup: | |
120 | kfree(data); | |
121 | ||
122 | out: | |
123 | return err; | |
f875bae0 EB |
124 | } |
125 | ||
126 | static void ops_free(const struct pernet_operations *ops, struct net *net) | |
127 | { | |
128 | if (ops->id && ops->size) { | |
c7d03a00 | 129 | kfree(net_generic(net, *ops->id)); |
f875bae0 EB |
130 | } |
131 | } | |
132 | ||
72ad937a EB |
133 | static void ops_exit_list(const struct pernet_operations *ops, |
134 | struct list_head *net_exit_list) | |
135 | { | |
136 | struct net *net; | |
137 | if (ops->exit) { | |
138 | list_for_each_entry(net, net_exit_list, exit_list) | |
139 | ops->exit(net); | |
140 | } | |
72ad937a EB |
141 | if (ops->exit_batch) |
142 | ops->exit_batch(net_exit_list); | |
143 | } | |
144 | ||
145 | static void ops_free_list(const struct pernet_operations *ops, | |
146 | struct list_head *net_exit_list) | |
147 | { | |
148 | struct net *net; | |
149 | if (ops->size && ops->id) { | |
150 | list_for_each_entry(net, net_exit_list, exit_list) | |
151 | ops_free(ops, net); | |
152 | } | |
153 | } | |
154 | ||
95f38411 | 155 | /* should be called with nsid_lock held */ |
0c7aecd4 ND |
156 | static int alloc_netid(struct net *net, struct net *peer, int reqid) |
157 | { | |
3138dbf8 | 158 | int min = 0, max = 0; |
0c7aecd4 | 159 | |
0c7aecd4 ND |
160 | if (reqid >= 0) { |
161 | min = reqid; | |
162 | max = reqid + 1; | |
163 | } | |
164 | ||
95f38411 | 165 | return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC); |
0c7aecd4 ND |
166 | } |
167 | ||
168 | /* This function is used by idr_for_each(). If net is equal to peer, the | |
169 | * function returns the id so that idr_for_each() stops. Because we cannot | |
170 | * returns the id 0 (idr_for_each() will not stop), we return the magic value | |
171 | * NET_ID_ZERO (-1) for it. | |
172 | */ | |
173 | #define NET_ID_ZERO -1 | |
174 | static int net_eq_idr(int id, void *net, void *peer) | |
175 | { | |
176 | if (net_eq(net, peer)) | |
177 | return id ? : NET_ID_ZERO; | |
178 | return 0; | |
179 | } | |
180 | ||
95f38411 ND |
181 | /* Should be called with nsid_lock held. If a new id is assigned, the bool alloc |
182 | * is set to true, thus the caller knows that the new id must be notified via | |
183 | * rtnl. | |
184 | */ | |
3138dbf8 | 185 | static int __peernet2id_alloc(struct net *net, struct net *peer, bool *alloc) |
0c7aecd4 ND |
186 | { |
187 | int id = idr_for_each(&net->netns_ids, net_eq_idr, peer); | |
3138dbf8 | 188 | bool alloc_it = *alloc; |
0c7aecd4 | 189 | |
3138dbf8 ND |
190 | *alloc = false; |
191 | ||
0c7aecd4 ND |
192 | /* Magic value for id 0. */ |
193 | if (id == NET_ID_ZERO) | |
194 | return 0; | |
195 | if (id > 0) | |
196 | return id; | |
197 | ||
3138dbf8 | 198 | if (alloc_it) { |
109582af | 199 | id = alloc_netid(net, peer, -1); |
3138dbf8 | 200 | *alloc = true; |
109582af ND |
201 | return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED; |
202 | } | |
0c7aecd4 | 203 | |
109582af | 204 | return NETNSA_NSID_NOT_ASSIGNED; |
0c7aecd4 ND |
205 | } |
206 | ||
95f38411 | 207 | /* should be called with nsid_lock held */ |
3138dbf8 ND |
208 | static int __peernet2id(struct net *net, struct net *peer) |
209 | { | |
210 | bool no = false; | |
211 | ||
212 | return __peernet2id_alloc(net, peer, &no); | |
213 | } | |
214 | ||
215 | static void rtnl_net_notifyid(struct net *net, int cmd, int id); | |
0c7aecd4 ND |
216 | /* This function returns the id of a peer netns. If no id is assigned, one will |
217 | * be allocated and returned. | |
218 | */ | |
7a0877d4 | 219 | int peernet2id_alloc(struct net *net, struct net *peer) |
0c7aecd4 | 220 | { |
95f38411 | 221 | bool alloc; |
3138dbf8 | 222 | int id; |
0c7aecd4 | 223 | |
cfc44a4d WC |
224 | if (atomic_read(&net->count) == 0) |
225 | return NETNSA_NSID_NOT_ASSIGNED; | |
fba143c6 | 226 | spin_lock_bh(&net->nsid_lock); |
95f38411 | 227 | alloc = atomic_read(&peer->count) == 0 ? false : true; |
3138dbf8 | 228 | id = __peernet2id_alloc(net, peer, &alloc); |
fba143c6 | 229 | spin_unlock_bh(&net->nsid_lock); |
3138dbf8 ND |
230 | if (alloc && id >= 0) |
231 | rtnl_net_notifyid(net, RTM_NEWNSID, id); | |
232 | return id; | |
0c7aecd4 ND |
233 | } |
234 | ||
95f38411 | 235 | /* This function returns, if assigned, the id of a peer netns. */ |
59324cf3 | 236 | int peernet2id(struct net *net, struct net *peer) |
95f38411 | 237 | { |
95f38411 ND |
238 | int id; |
239 | ||
fba143c6 | 240 | spin_lock_bh(&net->nsid_lock); |
95f38411 | 241 | id = __peernet2id(net, peer); |
fba143c6 | 242 | spin_unlock_bh(&net->nsid_lock); |
95f38411 ND |
243 | return id; |
244 | } | |
38f507f1 | 245 | EXPORT_SYMBOL(peernet2id); |
95f38411 | 246 | |
59324cf3 ND |
247 | /* This function returns true is the peer netns has an id assigned into the |
248 | * current netns. | |
249 | */ | |
250 | bool peernet_has_id(struct net *net, struct net *peer) | |
251 | { | |
252 | return peernet2id(net, peer) >= 0; | |
253 | } | |
254 | ||
0c7aecd4 ND |
255 | struct net *get_net_ns_by_id(struct net *net, int id) |
256 | { | |
257 | struct net *peer; | |
258 | ||
259 | if (id < 0) | |
260 | return NULL; | |
261 | ||
262 | rcu_read_lock(); | |
fba143c6 | 263 | spin_lock_bh(&net->nsid_lock); |
0c7aecd4 ND |
264 | peer = idr_find(&net->netns_ids, id); |
265 | if (peer) | |
266 | get_net(peer); | |
fba143c6 | 267 | spin_unlock_bh(&net->nsid_lock); |
0c7aecd4 ND |
268 | rcu_read_unlock(); |
269 | ||
270 | return peer; | |
271 | } | |
272 | ||
5f256bec EB |
273 | /* |
274 | * setup_net runs the initializers for the network namespace object. | |
275 | */ | |
038e7332 | 276 | static __net_init int setup_net(struct net *net, struct user_namespace *user_ns) |
5f256bec EB |
277 | { |
278 | /* Must be called with net_mutex held */ | |
f875bae0 | 279 | const struct pernet_operations *ops, *saved_ops; |
486a87f1 | 280 | int error = 0; |
72ad937a | 281 | LIST_HEAD(net_exit_list); |
5f256bec | 282 | |
5f256bec | 283 | atomic_set(&net->count, 1); |
a685e089 | 284 | atomic_set(&net->passive, 1); |
4e985ada | 285 | net->dev_base_seq = 1; |
038e7332 | 286 | net->user_ns = user_ns; |
0c7aecd4 | 287 | idr_init(&net->netns_ids); |
de133464 | 288 | spin_lock_init(&net->nsid_lock); |
486a87f1 | 289 | |
768f3591 | 290 | list_for_each_entry(ops, &pernet_list, list) { |
f875bae0 EB |
291 | error = ops_init(ops, net); |
292 | if (error < 0) | |
293 | goto out_undo; | |
5f256bec EB |
294 | } |
295 | out: | |
296 | return error; | |
768f3591 | 297 | |
5f256bec EB |
298 | out_undo: |
299 | /* Walk through the list backwards calling the exit functions | |
300 | * for the pernet modules whose init functions did not fail. | |
301 | */ | |
72ad937a | 302 | list_add(&net->exit_list, &net_exit_list); |
f875bae0 | 303 | saved_ops = ops; |
72ad937a EB |
304 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) |
305 | ops_exit_list(ops, &net_exit_list); | |
306 | ||
f875bae0 EB |
307 | ops = saved_ops; |
308 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) | |
72ad937a | 309 | ops_free_list(ops, &net_exit_list); |
310928d9 DL |
310 | |
311 | rcu_barrier(); | |
5f256bec EB |
312 | goto out; |
313 | } | |
314 | ||
6a1a3b9f | 315 | |
ebe47d47 | 316 | #ifdef CONFIG_NET_NS |
2ed6afde AB |
317 | static struct ucounts *inc_net_namespaces(struct user_namespace *ns) |
318 | { | |
319 | return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES); | |
320 | } | |
321 | ||
322 | static void dec_net_namespaces(struct ucounts *ucounts) | |
323 | { | |
324 | dec_ucount(ucounts, UCOUNT_NET_NAMESPACES); | |
325 | } | |
326 | ||
ebe47d47 CN |
327 | static struct kmem_cache *net_cachep; |
328 | static struct workqueue_struct *netns_wq; | |
329 | ||
486a87f1 | 330 | static struct net *net_alloc(void) |
45a19b0a | 331 | { |
486a87f1 DL |
332 | struct net *net = NULL; |
333 | struct net_generic *ng; | |
334 | ||
335 | ng = net_alloc_generic(); | |
336 | if (!ng) | |
337 | goto out; | |
338 | ||
339 | net = kmem_cache_zalloc(net_cachep, GFP_KERNEL); | |
45a19b0a | 340 | if (!net) |
486a87f1 | 341 | goto out_free; |
45a19b0a | 342 | |
486a87f1 DL |
343 | rcu_assign_pointer(net->gen, ng); |
344 | out: | |
345 | return net; | |
346 | ||
347 | out_free: | |
348 | kfree(ng); | |
349 | goto out; | |
350 | } | |
351 | ||
352 | static void net_free(struct net *net) | |
353 | { | |
416c51e1 | 354 | kfree(rcu_access_pointer(net->gen)); |
45a19b0a JFS |
355 | kmem_cache_free(net_cachep, net); |
356 | } | |
357 | ||
a685e089 AV |
358 | void net_drop_ns(void *p) |
359 | { | |
360 | struct net *ns = p; | |
361 | if (ns && atomic_dec_and_test(&ns->passive)) | |
362 | net_free(ns); | |
363 | } | |
364 | ||
038e7332 EB |
365 | struct net *copy_net_ns(unsigned long flags, |
366 | struct user_namespace *user_ns, struct net *old_net) | |
9dd776b6 | 367 | { |
70328660 | 368 | struct ucounts *ucounts; |
088eb2d9 AD |
369 | struct net *net; |
370 | int rv; | |
9dd776b6 | 371 | |
911cb193 RL |
372 | if (!(flags & CLONE_NEWNET)) |
373 | return get_net(old_net); | |
374 | ||
70328660 EB |
375 | ucounts = inc_net_namespaces(user_ns); |
376 | if (!ucounts) | |
df75e774 | 377 | return ERR_PTR(-ENOSPC); |
70328660 | 378 | |
088eb2d9 | 379 | net = net_alloc(); |
70328660 EB |
380 | if (!net) { |
381 | dec_net_namespaces(ucounts); | |
088eb2d9 | 382 | return ERR_PTR(-ENOMEM); |
70328660 | 383 | } |
038e7332 EB |
384 | |
385 | get_user_ns(user_ns); | |
386 | ||
7281a665 AV |
387 | rv = mutex_lock_killable(&net_mutex); |
388 | if (rv < 0) { | |
389 | net_free(net); | |
390 | dec_net_namespaces(ucounts); | |
391 | put_user_ns(user_ns); | |
392 | return ERR_PTR(rv); | |
393 | } | |
394 | ||
70328660 | 395 | net->ucounts = ucounts; |
038e7332 | 396 | rv = setup_net(net, user_ns); |
088eb2d9 | 397 | if (rv == 0) { |
486a87f1 | 398 | rtnl_lock(); |
11a28d37 | 399 | list_add_tail_rcu(&net->list, &net_namespace_list); |
486a87f1 DL |
400 | rtnl_unlock(); |
401 | } | |
9dd776b6 | 402 | mutex_unlock(&net_mutex); |
088eb2d9 | 403 | if (rv < 0) { |
70328660 | 404 | dec_net_namespaces(ucounts); |
038e7332 | 405 | put_user_ns(user_ns); |
a685e089 | 406 | net_drop_ns(net); |
088eb2d9 AD |
407 | return ERR_PTR(rv); |
408 | } | |
409 | return net; | |
410 | } | |
486a87f1 | 411 | |
2b035b39 EB |
412 | static DEFINE_SPINLOCK(cleanup_list_lock); |
413 | static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */ | |
414 | ||
6a1a3b9f PE |
415 | static void cleanup_net(struct work_struct *work) |
416 | { | |
f875bae0 | 417 | const struct pernet_operations *ops; |
6d458f5b | 418 | struct net *net, *tmp; |
1818ce4d | 419 | struct list_head net_kill_list; |
72ad937a | 420 | LIST_HEAD(net_exit_list); |
6a1a3b9f | 421 | |
2b035b39 EB |
422 | /* Atomically snapshot the list of namespaces to cleanup */ |
423 | spin_lock_irq(&cleanup_list_lock); | |
424 | list_replace_init(&cleanup_list, &net_kill_list); | |
425 | spin_unlock_irq(&cleanup_list_lock); | |
6a1a3b9f PE |
426 | |
427 | mutex_lock(&net_mutex); | |
428 | ||
429 | /* Don't let anyone else find us. */ | |
430 | rtnl_lock(); | |
72ad937a | 431 | list_for_each_entry(net, &net_kill_list, cleanup_list) { |
2b035b39 | 432 | list_del_rcu(&net->list); |
72ad937a | 433 | list_add_tail(&net->exit_list, &net_exit_list); |
6d458f5b | 434 | for_each_net(tmp) { |
95f38411 | 435 | int id; |
6d458f5b | 436 | |
fba143c6 | 437 | spin_lock_bh(&tmp->nsid_lock); |
95f38411 ND |
438 | id = __peernet2id(tmp, net); |
439 | if (id >= 0) | |
6d458f5b | 440 | idr_remove(&tmp->netns_ids, id); |
fba143c6 | 441 | spin_unlock_bh(&tmp->nsid_lock); |
95f38411 ND |
442 | if (id >= 0) |
443 | rtnl_net_notifyid(tmp, RTM_DELNSID, id); | |
6d458f5b | 444 | } |
fba143c6 | 445 | spin_lock_bh(&net->nsid_lock); |
6d458f5b | 446 | idr_destroy(&net->netns_ids); |
fba143c6 | 447 | spin_unlock_bh(&net->nsid_lock); |
6d458f5b | 448 | |
72ad937a | 449 | } |
6a1a3b9f PE |
450 | rtnl_unlock(); |
451 | ||
11a28d37 JB |
452 | /* |
453 | * Another CPU might be rcu-iterating the list, wait for it. | |
454 | * This needs to be before calling the exit() notifiers, so | |
455 | * the rcu_barrier() below isn't sufficient alone. | |
456 | */ | |
457 | synchronize_rcu(); | |
458 | ||
6a1a3b9f | 459 | /* Run all of the network namespace exit methods */ |
72ad937a EB |
460 | list_for_each_entry_reverse(ops, &pernet_list, list) |
461 | ops_exit_list(ops, &net_exit_list); | |
462 | ||
f875bae0 | 463 | /* Free the net generic variables */ |
72ad937a EB |
464 | list_for_each_entry_reverse(ops, &pernet_list, list) |
465 | ops_free_list(ops, &net_exit_list); | |
6a1a3b9f PE |
466 | |
467 | mutex_unlock(&net_mutex); | |
468 | ||
469 | /* Ensure there are no outstanding rcu callbacks using this | |
470 | * network namespace. | |
471 | */ | |
472 | rcu_barrier(); | |
473 | ||
474 | /* Finally it is safe to free my network namespace structure */ | |
72ad937a EB |
475 | list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) { |
476 | list_del_init(&net->exit_list); | |
70328660 | 477 | dec_net_namespaces(net->ucounts); |
038e7332 | 478 | put_user_ns(net->user_ns); |
a685e089 | 479 | net_drop_ns(net); |
2b035b39 | 480 | } |
6a1a3b9f | 481 | } |
2b035b39 | 482 | static DECLARE_WORK(net_cleanup_work, cleanup_net); |
6a1a3b9f PE |
483 | |
484 | void __put_net(struct net *net) | |
485 | { | |
486 | /* Cleanup the network namespace in process context */ | |
2b035b39 EB |
487 | unsigned long flags; |
488 | ||
489 | spin_lock_irqsave(&cleanup_list_lock, flags); | |
490 | list_add(&net->cleanup_list, &cleanup_list); | |
491 | spin_unlock_irqrestore(&cleanup_list_lock, flags); | |
492 | ||
493 | queue_work(netns_wq, &net_cleanup_work); | |
6a1a3b9f PE |
494 | } |
495 | EXPORT_SYMBOL_GPL(__put_net); | |
496 | ||
956c9207 SR |
497 | struct net *get_net_ns_by_fd(int fd) |
498 | { | |
956c9207 | 499 | struct file *file; |
33c42940 | 500 | struct ns_common *ns; |
956c9207 SR |
501 | struct net *net; |
502 | ||
956c9207 | 503 | file = proc_ns_fget(fd); |
c316e6a3 AV |
504 | if (IS_ERR(file)) |
505 | return ERR_CAST(file); | |
956c9207 | 506 | |
f77c8014 | 507 | ns = get_proc_ns(file_inode(file)); |
33c42940 AV |
508 | if (ns->ops == &netns_operations) |
509 | net = get_net(container_of(ns, struct net, ns)); | |
c316e6a3 AV |
510 | else |
511 | net = ERR_PTR(-EINVAL); | |
956c9207 | 512 | |
c316e6a3 | 513 | fput(file); |
956c9207 SR |
514 | return net; |
515 | } | |
516 | ||
6a1a3b9f | 517 | #else |
956c9207 SR |
518 | struct net *get_net_ns_by_fd(int fd) |
519 | { | |
520 | return ERR_PTR(-EINVAL); | |
521 | } | |
6a1a3b9f | 522 | #endif |
4b681c82 | 523 | EXPORT_SYMBOL_GPL(get_net_ns_by_fd); |
6a1a3b9f | 524 | |
30ffee84 JB |
525 | struct net *get_net_ns_by_pid(pid_t pid) |
526 | { | |
527 | struct task_struct *tsk; | |
528 | struct net *net; | |
529 | ||
530 | /* Lookup the network namespace */ | |
531 | net = ERR_PTR(-ESRCH); | |
532 | rcu_read_lock(); | |
533 | tsk = find_task_by_vpid(pid); | |
534 | if (tsk) { | |
535 | struct nsproxy *nsproxy; | |
728dba3a EB |
536 | task_lock(tsk); |
537 | nsproxy = tsk->nsproxy; | |
30ffee84 JB |
538 | if (nsproxy) |
539 | net = get_net(nsproxy->net_ns); | |
728dba3a | 540 | task_unlock(tsk); |
30ffee84 JB |
541 | } |
542 | rcu_read_unlock(); | |
543 | return net; | |
544 | } | |
545 | EXPORT_SYMBOL_GPL(get_net_ns_by_pid); | |
546 | ||
98f842e6 EB |
547 | static __net_init int net_ns_net_init(struct net *net) |
548 | { | |
33c42940 AV |
549 | #ifdef CONFIG_NET_NS |
550 | net->ns.ops = &netns_operations; | |
551 | #endif | |
6344c433 | 552 | return ns_alloc_inum(&net->ns); |
98f842e6 EB |
553 | } |
554 | ||
555 | static __net_exit void net_ns_net_exit(struct net *net) | |
556 | { | |
6344c433 | 557 | ns_free_inum(&net->ns); |
98f842e6 EB |
558 | } |
559 | ||
560 | static struct pernet_operations __net_initdata net_ns_ops = { | |
561 | .init = net_ns_net_init, | |
562 | .exit = net_ns_net_exit, | |
563 | }; | |
564 | ||
3ee5256d | 565 | static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = { |
0c7aecd4 ND |
566 | [NETNSA_NONE] = { .type = NLA_UNSPEC }, |
567 | [NETNSA_NSID] = { .type = NLA_S32 }, | |
568 | [NETNSA_PID] = { .type = NLA_U32 }, | |
569 | [NETNSA_FD] = { .type = NLA_U32 }, | |
570 | }; | |
571 | ||
572 | static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
573 | { | |
574 | struct net *net = sock_net(skb->sk); | |
575 | struct nlattr *tb[NETNSA_MAX + 1]; | |
576 | struct net *peer; | |
577 | int nsid, err; | |
578 | ||
579 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
580 | rtnl_net_policy); | |
581 | if (err < 0) | |
582 | return err; | |
583 | if (!tb[NETNSA_NSID]) | |
584 | return -EINVAL; | |
585 | nsid = nla_get_s32(tb[NETNSA_NSID]); | |
586 | ||
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 | if (IS_ERR(peer)) | |
594 | return PTR_ERR(peer); | |
595 | ||
fba143c6 | 596 | spin_lock_bh(&net->nsid_lock); |
3138dbf8 | 597 | if (__peernet2id(net, peer) >= 0) { |
fba143c6 | 598 | spin_unlock_bh(&net->nsid_lock); |
0c7aecd4 ND |
599 | err = -EEXIST; |
600 | goto out; | |
601 | } | |
602 | ||
603 | err = alloc_netid(net, peer, nsid); | |
fba143c6 | 604 | spin_unlock_bh(&net->nsid_lock); |
3138dbf8 ND |
605 | if (err >= 0) { |
606 | rtnl_net_notifyid(net, RTM_NEWNSID, err); | |
0c7aecd4 | 607 | err = 0; |
3138dbf8 | 608 | } |
0c7aecd4 ND |
609 | out: |
610 | put_net(peer); | |
611 | return err; | |
612 | } | |
613 | ||
614 | static int rtnl_net_get_size(void) | |
615 | { | |
616 | return NLMSG_ALIGN(sizeof(struct rtgenmsg)) | |
617 | + nla_total_size(sizeof(s32)) /* NETNSA_NSID */ | |
618 | ; | |
619 | } | |
620 | ||
621 | static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags, | |
cab3c8ec | 622 | int cmd, struct net *net, int nsid) |
0c7aecd4 ND |
623 | { |
624 | struct nlmsghdr *nlh; | |
625 | struct rtgenmsg *rth; | |
0c7aecd4 ND |
626 | |
627 | nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags); | |
628 | if (!nlh) | |
629 | return -EMSGSIZE; | |
630 | ||
631 | rth = nlmsg_data(nlh); | |
632 | rth->rtgen_family = AF_UNSPEC; | |
633 | ||
cab3c8ec | 634 | if (nla_put_s32(skb, NETNSA_NSID, nsid)) |
0c7aecd4 ND |
635 | goto nla_put_failure; |
636 | ||
637 | nlmsg_end(skb, nlh); | |
638 | return 0; | |
639 | ||
640 | nla_put_failure: | |
641 | nlmsg_cancel(skb, nlh); | |
642 | return -EMSGSIZE; | |
643 | } | |
644 | ||
645 | static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
646 | { | |
647 | struct net *net = sock_net(skb->sk); | |
648 | struct nlattr *tb[NETNSA_MAX + 1]; | |
649 | struct sk_buff *msg; | |
0c7aecd4 | 650 | struct net *peer; |
cab3c8ec | 651 | int err, id; |
0c7aecd4 ND |
652 | |
653 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
654 | rtnl_net_policy); | |
655 | if (err < 0) | |
656 | return err; | |
657 | if (tb[NETNSA_PID]) | |
658 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); | |
659 | else if (tb[NETNSA_FD]) | |
660 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); | |
661 | else | |
662 | return -EINVAL; | |
663 | ||
664 | if (IS_ERR(peer)) | |
665 | return PTR_ERR(peer); | |
666 | ||
667 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
668 | if (!msg) { | |
669 | err = -ENOMEM; | |
670 | goto out; | |
671 | } | |
672 | ||
95f38411 | 673 | id = peernet2id(net, peer); |
0c7aecd4 | 674 | err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, |
b04096ff | 675 | RTM_NEWNSID, net, id); |
0c7aecd4 ND |
676 | if (err < 0) |
677 | goto err_out; | |
678 | ||
679 | err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid); | |
680 | goto out; | |
681 | ||
682 | err_out: | |
683 | nlmsg_free(msg); | |
684 | out: | |
685 | put_net(peer); | |
686 | return err; | |
687 | } | |
688 | ||
a143c40c ND |
689 | struct rtnl_net_dump_cb { |
690 | struct net *net; | |
691 | struct sk_buff *skb; | |
692 | struct netlink_callback *cb; | |
693 | int idx; | |
694 | int s_idx; | |
695 | }; | |
696 | ||
697 | static int rtnl_net_dumpid_one(int id, void *peer, void *data) | |
698 | { | |
699 | struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data; | |
700 | int ret; | |
701 | ||
702 | if (net_cb->idx < net_cb->s_idx) | |
703 | goto cont; | |
704 | ||
705 | ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid, | |
706 | net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI, | |
cab3c8ec | 707 | RTM_NEWNSID, net_cb->net, id); |
a143c40c ND |
708 | if (ret < 0) |
709 | return ret; | |
710 | ||
711 | cont: | |
712 | net_cb->idx++; | |
713 | return 0; | |
714 | } | |
715 | ||
716 | static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb) | |
717 | { | |
718 | struct net *net = sock_net(skb->sk); | |
719 | struct rtnl_net_dump_cb net_cb = { | |
720 | .net = net, | |
721 | .skb = skb, | |
722 | .cb = cb, | |
723 | .idx = 0, | |
724 | .s_idx = cb->args[0], | |
725 | }; | |
726 | ||
fba143c6 | 727 | spin_lock_bh(&net->nsid_lock); |
a143c40c | 728 | idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb); |
fba143c6 | 729 | spin_unlock_bh(&net->nsid_lock); |
a143c40c ND |
730 | |
731 | cb->args[0] = net_cb.idx; | |
732 | return skb->len; | |
733 | } | |
734 | ||
cab3c8ec | 735 | static void rtnl_net_notifyid(struct net *net, int cmd, int id) |
9a963454 ND |
736 | { |
737 | struct sk_buff *msg; | |
738 | int err = -ENOMEM; | |
739 | ||
740 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
741 | if (!msg) | |
742 | goto out; | |
743 | ||
cab3c8ec | 744 | err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, id); |
9a963454 ND |
745 | if (err < 0) |
746 | goto err_out; | |
747 | ||
748 | rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0); | |
749 | return; | |
750 | ||
751 | err_out: | |
752 | nlmsg_free(msg); | |
753 | out: | |
754 | rtnl_set_sk_err(net, RTNLGRP_NSID, err); | |
755 | } | |
756 | ||
5f256bec EB |
757 | static int __init net_ns_init(void) |
758 | { | |
486a87f1 | 759 | struct net_generic *ng; |
5f256bec | 760 | |
d57a9212 | 761 | #ifdef CONFIG_NET_NS |
5f256bec EB |
762 | net_cachep = kmem_cache_create("net_namespace", sizeof(struct net), |
763 | SMP_CACHE_BYTES, | |
764 | SLAB_PANIC, NULL); | |
3ef1355d BT |
765 | |
766 | /* Create workqueue for cleanup */ | |
767 | netns_wq = create_singlethread_workqueue("netns"); | |
768 | if (!netns_wq) | |
769 | panic("Could not create netns workq"); | |
d57a9212 | 770 | #endif |
3ef1355d | 771 | |
486a87f1 DL |
772 | ng = net_alloc_generic(); |
773 | if (!ng) | |
774 | panic("Could not allocate generic netns"); | |
775 | ||
776 | rcu_assign_pointer(init_net.gen, ng); | |
777 | ||
5f256bec | 778 | mutex_lock(&net_mutex); |
038e7332 | 779 | if (setup_net(&init_net, &init_user_ns)) |
ca0f3112 | 780 | panic("Could not setup the initial network namespace"); |
5f256bec | 781 | |
f8c46cb3 DT |
782 | init_net_initialized = true; |
783 | ||
f4618d39 | 784 | rtnl_lock(); |
11a28d37 | 785 | list_add_tail_rcu(&init_net.list, &net_namespace_list); |
f4618d39 | 786 | rtnl_unlock(); |
5f256bec EB |
787 | |
788 | mutex_unlock(&net_mutex); | |
5f256bec | 789 | |
98f842e6 EB |
790 | register_pernet_subsys(&net_ns_ops); |
791 | ||
0c7aecd4 | 792 | rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL); |
a143c40c ND |
793 | rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid, |
794 | NULL); | |
0c7aecd4 | 795 | |
5f256bec EB |
796 | return 0; |
797 | } | |
798 | ||
799 | pure_initcall(net_ns_init); | |
800 | ||
ed160e83 | 801 | #ifdef CONFIG_NET_NS |
f875bae0 EB |
802 | static int __register_pernet_operations(struct list_head *list, |
803 | struct pernet_operations *ops) | |
5f256bec | 804 | { |
72ad937a | 805 | struct net *net; |
5f256bec | 806 | int error; |
72ad937a | 807 | LIST_HEAD(net_exit_list); |
5f256bec | 808 | |
5f256bec | 809 | list_add_tail(&ops->list, list); |
f875bae0 | 810 | if (ops->init || (ops->id && ops->size)) { |
1dba323b | 811 | for_each_net(net) { |
f875bae0 | 812 | error = ops_init(ops, net); |
5f256bec EB |
813 | if (error) |
814 | goto out_undo; | |
72ad937a | 815 | list_add_tail(&net->exit_list, &net_exit_list); |
5f256bec EB |
816 | } |
817 | } | |
1dba323b | 818 | return 0; |
5f256bec EB |
819 | |
820 | out_undo: | |
821 | /* If I have an error cleanup all namespaces I initialized */ | |
822 | list_del(&ops->list); | |
72ad937a EB |
823 | ops_exit_list(ops, &net_exit_list); |
824 | ops_free_list(ops, &net_exit_list); | |
1dba323b | 825 | return error; |
5f256bec EB |
826 | } |
827 | ||
f875bae0 | 828 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
5f256bec EB |
829 | { |
830 | struct net *net; | |
72ad937a | 831 | LIST_HEAD(net_exit_list); |
5f256bec EB |
832 | |
833 | list_del(&ops->list); | |
72ad937a EB |
834 | for_each_net(net) |
835 | list_add_tail(&net->exit_list, &net_exit_list); | |
836 | ops_exit_list(ops, &net_exit_list); | |
837 | ops_free_list(ops, &net_exit_list); | |
5f256bec EB |
838 | } |
839 | ||
ed160e83 DL |
840 | #else |
841 | ||
f875bae0 EB |
842 | static int __register_pernet_operations(struct list_head *list, |
843 | struct pernet_operations *ops) | |
ed160e83 | 844 | { |
f8c46cb3 DT |
845 | if (!init_net_initialized) { |
846 | list_add_tail(&ops->list, list); | |
847 | return 0; | |
848 | } | |
849 | ||
b922934d | 850 | return ops_init(ops, &init_net); |
ed160e83 DL |
851 | } |
852 | ||
f875bae0 | 853 | static void __unregister_pernet_operations(struct pernet_operations *ops) |
ed160e83 | 854 | { |
f8c46cb3 DT |
855 | if (!init_net_initialized) { |
856 | list_del(&ops->list); | |
857 | } else { | |
858 | LIST_HEAD(net_exit_list); | |
859 | list_add(&init_net.exit_list, &net_exit_list); | |
860 | ops_exit_list(ops, &net_exit_list); | |
861 | ops_free_list(ops, &net_exit_list); | |
862 | } | |
ed160e83 | 863 | } |
f875bae0 EB |
864 | |
865 | #endif /* CONFIG_NET_NS */ | |
ed160e83 | 866 | |
c93cf61f PE |
867 | static DEFINE_IDA(net_generic_ids); |
868 | ||
f875bae0 EB |
869 | static int register_pernet_operations(struct list_head *list, |
870 | struct pernet_operations *ops) | |
871 | { | |
872 | int error; | |
873 | ||
874 | if (ops->id) { | |
875 | again: | |
6af2d5ff | 876 | error = ida_get_new_above(&net_generic_ids, MIN_PERNET_OPS_ID, ops->id); |
f875bae0 EB |
877 | if (error < 0) { |
878 | if (error == -EAGAIN) { | |
879 | ida_pre_get(&net_generic_ids, GFP_KERNEL); | |
880 | goto again; | |
881 | } | |
882 | return error; | |
883 | } | |
6af2d5ff | 884 | max_gen_ptrs = max(max_gen_ptrs, *ops->id + 1); |
f875bae0 EB |
885 | } |
886 | error = __register_pernet_operations(list, ops); | |
3a765eda EB |
887 | if (error) { |
888 | rcu_barrier(); | |
889 | if (ops->id) | |
890 | ida_remove(&net_generic_ids, *ops->id); | |
891 | } | |
f875bae0 EB |
892 | |
893 | return error; | |
894 | } | |
895 | ||
896 | static void unregister_pernet_operations(struct pernet_operations *ops) | |
897 | { | |
898 | ||
899 | __unregister_pernet_operations(ops); | |
3a765eda | 900 | rcu_barrier(); |
f875bae0 EB |
901 | if (ops->id) |
902 | ida_remove(&net_generic_ids, *ops->id); | |
903 | } | |
904 | ||
5f256bec EB |
905 | /** |
906 | * register_pernet_subsys - register a network namespace subsystem | |
907 | * @ops: pernet operations structure for the subsystem | |
908 | * | |
909 | * Register a subsystem which has init and exit functions | |
910 | * that are called when network namespaces are created and | |
911 | * destroyed respectively. | |
912 | * | |
913 | * When registered all network namespace init functions are | |
914 | * called for every existing network namespace. Allowing kernel | |
915 | * modules to have a race free view of the set of network namespaces. | |
916 | * | |
917 | * When a new network namespace is created all of the init | |
918 | * methods are called in the order in which they were registered. | |
919 | * | |
920 | * When a network namespace is destroyed all of the exit methods | |
921 | * are called in the reverse of the order with which they were | |
922 | * registered. | |
923 | */ | |
924 | int register_pernet_subsys(struct pernet_operations *ops) | |
925 | { | |
926 | int error; | |
927 | mutex_lock(&net_mutex); | |
928 | error = register_pernet_operations(first_device, ops); | |
929 | mutex_unlock(&net_mutex); | |
930 | return error; | |
931 | } | |
932 | EXPORT_SYMBOL_GPL(register_pernet_subsys); | |
933 | ||
934 | /** | |
935 | * unregister_pernet_subsys - unregister a network namespace subsystem | |
936 | * @ops: pernet operations structure to manipulate | |
937 | * | |
938 | * Remove the pernet operations structure from the list to be | |
53379e57 | 939 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
940 | * addition run the exit method for all existing network |
941 | * namespaces. | |
942 | */ | |
b3c981d2 | 943 | void unregister_pernet_subsys(struct pernet_operations *ops) |
5f256bec EB |
944 | { |
945 | mutex_lock(&net_mutex); | |
b3c981d2 | 946 | unregister_pernet_operations(ops); |
5f256bec EB |
947 | mutex_unlock(&net_mutex); |
948 | } | |
949 | EXPORT_SYMBOL_GPL(unregister_pernet_subsys); | |
950 | ||
951 | /** | |
952 | * register_pernet_device - register a network namespace device | |
953 | * @ops: pernet operations structure for the subsystem | |
954 | * | |
955 | * Register a device which has init and exit functions | |
956 | * that are called when network namespaces are created and | |
957 | * destroyed respectively. | |
958 | * | |
959 | * When registered all network namespace init functions are | |
960 | * called for every existing network namespace. Allowing kernel | |
961 | * modules to have a race free view of the set of network namespaces. | |
962 | * | |
963 | * When a new network namespace is created all of the init | |
964 | * methods are called in the order in which they were registered. | |
965 | * | |
966 | * When a network namespace is destroyed all of the exit methods | |
967 | * are called in the reverse of the order with which they were | |
968 | * registered. | |
969 | */ | |
970 | int register_pernet_device(struct pernet_operations *ops) | |
971 | { | |
972 | int error; | |
973 | mutex_lock(&net_mutex); | |
974 | error = register_pernet_operations(&pernet_list, ops); | |
975 | if (!error && (first_device == &pernet_list)) | |
976 | first_device = &ops->list; | |
977 | mutex_unlock(&net_mutex); | |
978 | return error; | |
979 | } | |
980 | EXPORT_SYMBOL_GPL(register_pernet_device); | |
981 | ||
982 | /** | |
983 | * unregister_pernet_device - unregister a network namespace netdevice | |
984 | * @ops: pernet operations structure to manipulate | |
985 | * | |
986 | * Remove the pernet operations structure from the list to be | |
53379e57 | 987 | * used when network namespaces are created or destroyed. In |
5f256bec EB |
988 | * addition run the exit method for all existing network |
989 | * namespaces. | |
990 | */ | |
991 | void unregister_pernet_device(struct pernet_operations *ops) | |
992 | { | |
993 | mutex_lock(&net_mutex); | |
994 | if (&ops->list == first_device) | |
995 | first_device = first_device->next; | |
996 | unregister_pernet_operations(ops); | |
997 | mutex_unlock(&net_mutex); | |
998 | } | |
999 | EXPORT_SYMBOL_GPL(unregister_pernet_device); | |
13b6f576 EB |
1000 | |
1001 | #ifdef CONFIG_NET_NS | |
64964528 | 1002 | static struct ns_common *netns_get(struct task_struct *task) |
13b6f576 | 1003 | { |
f0630529 EB |
1004 | struct net *net = NULL; |
1005 | struct nsproxy *nsproxy; | |
1006 | ||
728dba3a EB |
1007 | task_lock(task); |
1008 | nsproxy = task->nsproxy; | |
f0630529 EB |
1009 | if (nsproxy) |
1010 | net = get_net(nsproxy->net_ns); | |
728dba3a | 1011 | task_unlock(task); |
f0630529 | 1012 | |
ff24870f AV |
1013 | return net ? &net->ns : NULL; |
1014 | } | |
1015 | ||
1016 | static inline struct net *to_net_ns(struct ns_common *ns) | |
1017 | { | |
1018 | return container_of(ns, struct net, ns); | |
13b6f576 EB |
1019 | } |
1020 | ||
64964528 | 1021 | static void netns_put(struct ns_common *ns) |
13b6f576 | 1022 | { |
ff24870f | 1023 | put_net(to_net_ns(ns)); |
13b6f576 EB |
1024 | } |
1025 | ||
64964528 | 1026 | static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns) |
13b6f576 | 1027 | { |
ff24870f | 1028 | struct net *net = to_net_ns(ns); |
142e1d1d | 1029 | |
5e4a0847 | 1030 | if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) || |
c7b96acf | 1031 | !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) |
142e1d1d EB |
1032 | return -EPERM; |
1033 | ||
13b6f576 | 1034 | put_net(nsproxy->net_ns); |
142e1d1d | 1035 | nsproxy->net_ns = get_net(net); |
13b6f576 EB |
1036 | return 0; |
1037 | } | |
1038 | ||
bcac25a5 AV |
1039 | static struct user_namespace *netns_owner(struct ns_common *ns) |
1040 | { | |
1041 | return to_net_ns(ns)->user_ns; | |
1042 | } | |
1043 | ||
13b6f576 EB |
1044 | const struct proc_ns_operations netns_operations = { |
1045 | .name = "net", | |
1046 | .type = CLONE_NEWNET, | |
1047 | .get = netns_get, | |
1048 | .put = netns_put, | |
1049 | .install = netns_install, | |
bcac25a5 | 1050 | .owner = netns_owner, |
13b6f576 EB |
1051 | }; |
1052 | #endif |