]>
Commit | Line | Data |
---|---|---|
1 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt | |
2 | ||
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> | |
9 | #include <linux/sched.h> | |
10 | #include <linux/idr.h> | |
11 | #include <linux/rculist.h> | |
12 | #include <linux/nsproxy.h> | |
13 | #include <linux/fs.h> | |
14 | #include <linux/proc_ns.h> | |
15 | #include <linux/file.h> | |
16 | #include <linux/export.h> | |
17 | #include <linux/user_namespace.h> | |
18 | #include <linux/net_namespace.h> | |
19 | #include <net/sock.h> | |
20 | #include <net/netlink.h> | |
21 | #include <net/net_namespace.h> | |
22 | #include <net/netns/generic.h> | |
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; | |
30 | DEFINE_MUTEX(net_mutex); | |
31 | static DEFINE_SPINLOCK(nsid_lock); | |
32 | ||
33 | LIST_HEAD(net_namespace_list); | |
34 | EXPORT_SYMBOL_GPL(net_namespace_list); | |
35 | ||
36 | struct net init_net = { | |
37 | .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head), | |
38 | }; | |
39 | EXPORT_SYMBOL(init_net); | |
40 | ||
41 | #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */ | |
42 | ||
43 | static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS; | |
44 | ||
45 | static struct net_generic *net_alloc_generic(void) | |
46 | { | |
47 | struct net_generic *ng; | |
48 | size_t generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]); | |
49 | ||
50 | ng = kzalloc(generic_size, GFP_KERNEL); | |
51 | if (ng) | |
52 | ng->len = max_gen_ptrs; | |
53 | ||
54 | return ng; | |
55 | } | |
56 | ||
57 | static int net_assign_generic(struct net *net, int id, void *data) | |
58 | { | |
59 | struct net_generic *ng, *old_ng; | |
60 | ||
61 | BUG_ON(!mutex_is_locked(&net_mutex)); | |
62 | BUG_ON(id == 0); | |
63 | ||
64 | old_ng = rcu_dereference_protected(net->gen, | |
65 | lockdep_is_held(&net_mutex)); | |
66 | ng = old_ng; | |
67 | if (old_ng->len >= id) | |
68 | goto assign; | |
69 | ||
70 | ng = net_alloc_generic(); | |
71 | if (ng == NULL) | |
72 | return -ENOMEM; | |
73 | ||
74 | /* | |
75 | * Some synchronisation notes: | |
76 | * | |
77 | * The net_generic explores the net->gen array inside rcu | |
78 | * read section. Besides once set the net->gen->ptr[x] | |
79 | * pointer never changes (see rules in netns/generic.h). | |
80 | * | |
81 | * That said, we simply duplicate this array and schedule | |
82 | * the old copy for kfree after a grace period. | |
83 | */ | |
84 | ||
85 | memcpy(&ng->ptr, &old_ng->ptr, old_ng->len * sizeof(void*)); | |
86 | ||
87 | rcu_assign_pointer(net->gen, ng); | |
88 | kfree_rcu(old_ng, rcu); | |
89 | assign: | |
90 | ng->ptr[id - 1] = data; | |
91 | return 0; | |
92 | } | |
93 | ||
94 | static int ops_init(const struct pernet_operations *ops, struct net *net) | |
95 | { | |
96 | int err = -ENOMEM; | |
97 | void *data = NULL; | |
98 | ||
99 | if (ops->id && ops->size) { | |
100 | data = kzalloc(ops->size, GFP_KERNEL); | |
101 | if (!data) | |
102 | goto out; | |
103 | ||
104 | err = net_assign_generic(net, *ops->id, data); | |
105 | if (err) | |
106 | goto cleanup; | |
107 | } | |
108 | err = 0; | |
109 | if (ops->init) | |
110 | err = ops->init(net); | |
111 | if (!err) | |
112 | return 0; | |
113 | ||
114 | cleanup: | |
115 | kfree(data); | |
116 | ||
117 | out: | |
118 | return err; | |
119 | } | |
120 | ||
121 | static void ops_free(const struct pernet_operations *ops, struct net *net) | |
122 | { | |
123 | if (ops->id && ops->size) { | |
124 | int id = *ops->id; | |
125 | kfree(net_generic(net, id)); | |
126 | } | |
127 | } | |
128 | ||
129 | static void ops_exit_list(const struct pernet_operations *ops, | |
130 | struct list_head *net_exit_list) | |
131 | { | |
132 | struct net *net; | |
133 | if (ops->exit) { | |
134 | list_for_each_entry(net, net_exit_list, exit_list) | |
135 | ops->exit(net); | |
136 | } | |
137 | if (ops->exit_batch) | |
138 | ops->exit_batch(net_exit_list); | |
139 | } | |
140 | ||
141 | static void ops_free_list(const struct pernet_operations *ops, | |
142 | struct list_head *net_exit_list) | |
143 | { | |
144 | struct net *net; | |
145 | if (ops->size && ops->id) { | |
146 | list_for_each_entry(net, net_exit_list, exit_list) | |
147 | ops_free(ops, net); | |
148 | } | |
149 | } | |
150 | ||
151 | /* should be called with nsid_lock held */ | |
152 | static int alloc_netid(struct net *net, struct net *peer, int reqid) | |
153 | { | |
154 | int min = 0, max = 0; | |
155 | ||
156 | if (reqid >= 0) { | |
157 | min = reqid; | |
158 | max = reqid + 1; | |
159 | } | |
160 | ||
161 | return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC); | |
162 | } | |
163 | ||
164 | /* This function is used by idr_for_each(). If net is equal to peer, the | |
165 | * function returns the id so that idr_for_each() stops. Because we cannot | |
166 | * returns the id 0 (idr_for_each() will not stop), we return the magic value | |
167 | * NET_ID_ZERO (-1) for it. | |
168 | */ | |
169 | #define NET_ID_ZERO -1 | |
170 | static int net_eq_idr(int id, void *net, void *peer) | |
171 | { | |
172 | if (net_eq(net, peer)) | |
173 | return id ? : NET_ID_ZERO; | |
174 | return 0; | |
175 | } | |
176 | ||
177 | /* Should be called with nsid_lock held. If a new id is assigned, the bool alloc | |
178 | * is set to true, thus the caller knows that the new id must be notified via | |
179 | * rtnl. | |
180 | */ | |
181 | static int __peernet2id_alloc(struct net *net, struct net *peer, bool *alloc) | |
182 | { | |
183 | int id = idr_for_each(&net->netns_ids, net_eq_idr, peer); | |
184 | bool alloc_it = *alloc; | |
185 | ||
186 | *alloc = false; | |
187 | ||
188 | /* Magic value for id 0. */ | |
189 | if (id == NET_ID_ZERO) | |
190 | return 0; | |
191 | if (id > 0) | |
192 | return id; | |
193 | ||
194 | if (alloc_it) { | |
195 | id = alloc_netid(net, peer, -1); | |
196 | *alloc = true; | |
197 | return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED; | |
198 | } | |
199 | ||
200 | return NETNSA_NSID_NOT_ASSIGNED; | |
201 | } | |
202 | ||
203 | /* should be called with nsid_lock held */ | |
204 | static int __peernet2id(struct net *net, struct net *peer) | |
205 | { | |
206 | bool no = false; | |
207 | ||
208 | return __peernet2id_alloc(net, peer, &no); | |
209 | } | |
210 | ||
211 | static void rtnl_net_notifyid(struct net *net, int cmd, int id); | |
212 | /* This function returns the id of a peer netns. If no id is assigned, one will | |
213 | * be allocated and returned. | |
214 | */ | |
215 | int peernet2id_alloc(struct net *net, struct net *peer) | |
216 | { | |
217 | unsigned long flags; | |
218 | bool alloc; | |
219 | int id; | |
220 | ||
221 | spin_lock_irqsave(&nsid_lock, flags); | |
222 | alloc = atomic_read(&peer->count) == 0 ? false : true; | |
223 | id = __peernet2id_alloc(net, peer, &alloc); | |
224 | spin_unlock_irqrestore(&nsid_lock, flags); | |
225 | if (alloc && id >= 0) | |
226 | rtnl_net_notifyid(net, RTM_NEWNSID, id); | |
227 | return id; | |
228 | } | |
229 | EXPORT_SYMBOL(peernet2id_alloc); | |
230 | ||
231 | /* This function returns, if assigned, the id of a peer netns. */ | |
232 | int peernet2id(struct net *net, struct net *peer) | |
233 | { | |
234 | unsigned long flags; | |
235 | int id; | |
236 | ||
237 | spin_lock_irqsave(&nsid_lock, flags); | |
238 | id = __peernet2id(net, peer); | |
239 | spin_unlock_irqrestore(&nsid_lock, flags); | |
240 | return id; | |
241 | } | |
242 | ||
243 | /* This function returns true is the peer netns has an id assigned into the | |
244 | * current netns. | |
245 | */ | |
246 | bool peernet_has_id(struct net *net, struct net *peer) | |
247 | { | |
248 | return peernet2id(net, peer) >= 0; | |
249 | } | |
250 | ||
251 | struct net *get_net_ns_by_id(struct net *net, int id) | |
252 | { | |
253 | unsigned long flags; | |
254 | struct net *peer; | |
255 | ||
256 | if (id < 0) | |
257 | return NULL; | |
258 | ||
259 | rcu_read_lock(); | |
260 | spin_lock_irqsave(&nsid_lock, flags); | |
261 | peer = idr_find(&net->netns_ids, id); | |
262 | if (peer) | |
263 | get_net(peer); | |
264 | spin_unlock_irqrestore(&nsid_lock, flags); | |
265 | rcu_read_unlock(); | |
266 | ||
267 | return peer; | |
268 | } | |
269 | ||
270 | /* | |
271 | * setup_net runs the initializers for the network namespace object. | |
272 | */ | |
273 | static __net_init int setup_net(struct net *net, struct user_namespace *user_ns) | |
274 | { | |
275 | /* Must be called with net_mutex held */ | |
276 | const struct pernet_operations *ops, *saved_ops; | |
277 | int error = 0; | |
278 | LIST_HEAD(net_exit_list); | |
279 | ||
280 | atomic_set(&net->count, 1); | |
281 | atomic_set(&net->passive, 1); | |
282 | net->dev_base_seq = 1; | |
283 | net->user_ns = user_ns; | |
284 | idr_init(&net->netns_ids); | |
285 | ||
286 | list_for_each_entry(ops, &pernet_list, list) { | |
287 | error = ops_init(ops, net); | |
288 | if (error < 0) | |
289 | goto out_undo; | |
290 | } | |
291 | out: | |
292 | return error; | |
293 | ||
294 | out_undo: | |
295 | /* Walk through the list backwards calling the exit functions | |
296 | * for the pernet modules whose init functions did not fail. | |
297 | */ | |
298 | list_add(&net->exit_list, &net_exit_list); | |
299 | saved_ops = ops; | |
300 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) | |
301 | ops_exit_list(ops, &net_exit_list); | |
302 | ||
303 | ops = saved_ops; | |
304 | list_for_each_entry_continue_reverse(ops, &pernet_list, list) | |
305 | ops_free_list(ops, &net_exit_list); | |
306 | ||
307 | rcu_barrier(); | |
308 | goto out; | |
309 | } | |
310 | ||
311 | ||
312 | #ifdef CONFIG_NET_NS | |
313 | static struct kmem_cache *net_cachep; | |
314 | static struct workqueue_struct *netns_wq; | |
315 | ||
316 | static struct net *net_alloc(void) | |
317 | { | |
318 | struct net *net = NULL; | |
319 | struct net_generic *ng; | |
320 | ||
321 | ng = net_alloc_generic(); | |
322 | if (!ng) | |
323 | goto out; | |
324 | ||
325 | net = kmem_cache_zalloc(net_cachep, GFP_KERNEL); | |
326 | if (!net) | |
327 | goto out_free; | |
328 | ||
329 | rcu_assign_pointer(net->gen, ng); | |
330 | out: | |
331 | return net; | |
332 | ||
333 | out_free: | |
334 | kfree(ng); | |
335 | goto out; | |
336 | } | |
337 | ||
338 | static void net_free(struct net *net) | |
339 | { | |
340 | kfree(rcu_access_pointer(net->gen)); | |
341 | kmem_cache_free(net_cachep, net); | |
342 | } | |
343 | ||
344 | void net_drop_ns(void *p) | |
345 | { | |
346 | struct net *ns = p; | |
347 | if (ns && atomic_dec_and_test(&ns->passive)) | |
348 | net_free(ns); | |
349 | } | |
350 | ||
351 | struct net *copy_net_ns(unsigned long flags, | |
352 | struct user_namespace *user_ns, struct net *old_net) | |
353 | { | |
354 | struct net *net; | |
355 | int rv; | |
356 | ||
357 | if (!(flags & CLONE_NEWNET)) | |
358 | return get_net(old_net); | |
359 | ||
360 | net = net_alloc(); | |
361 | if (!net) | |
362 | return ERR_PTR(-ENOMEM); | |
363 | ||
364 | get_user_ns(user_ns); | |
365 | ||
366 | mutex_lock(&net_mutex); | |
367 | rv = setup_net(net, user_ns); | |
368 | if (rv == 0) { | |
369 | rtnl_lock(); | |
370 | list_add_tail_rcu(&net->list, &net_namespace_list); | |
371 | rtnl_unlock(); | |
372 | } | |
373 | mutex_unlock(&net_mutex); | |
374 | if (rv < 0) { | |
375 | put_user_ns(user_ns); | |
376 | net_drop_ns(net); | |
377 | return ERR_PTR(rv); | |
378 | } | |
379 | return net; | |
380 | } | |
381 | ||
382 | static DEFINE_SPINLOCK(cleanup_list_lock); | |
383 | static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */ | |
384 | ||
385 | static void cleanup_net(struct work_struct *work) | |
386 | { | |
387 | const struct pernet_operations *ops; | |
388 | struct net *net, *tmp; | |
389 | struct list_head net_kill_list; | |
390 | LIST_HEAD(net_exit_list); | |
391 | ||
392 | /* Atomically snapshot the list of namespaces to cleanup */ | |
393 | spin_lock_irq(&cleanup_list_lock); | |
394 | list_replace_init(&cleanup_list, &net_kill_list); | |
395 | spin_unlock_irq(&cleanup_list_lock); | |
396 | ||
397 | mutex_lock(&net_mutex); | |
398 | ||
399 | /* Don't let anyone else find us. */ | |
400 | rtnl_lock(); | |
401 | list_for_each_entry(net, &net_kill_list, cleanup_list) { | |
402 | list_del_rcu(&net->list); | |
403 | list_add_tail(&net->exit_list, &net_exit_list); | |
404 | for_each_net(tmp) { | |
405 | int id; | |
406 | ||
407 | spin_lock_irq(&nsid_lock); | |
408 | id = __peernet2id(tmp, net); | |
409 | if (id >= 0) | |
410 | idr_remove(&tmp->netns_ids, id); | |
411 | spin_unlock_irq(&nsid_lock); | |
412 | if (id >= 0) | |
413 | rtnl_net_notifyid(tmp, RTM_DELNSID, id); | |
414 | } | |
415 | spin_lock_irq(&nsid_lock); | |
416 | idr_destroy(&net->netns_ids); | |
417 | spin_unlock_irq(&nsid_lock); | |
418 | ||
419 | } | |
420 | rtnl_unlock(); | |
421 | ||
422 | /* | |
423 | * Another CPU might be rcu-iterating the list, wait for it. | |
424 | * This needs to be before calling the exit() notifiers, so | |
425 | * the rcu_barrier() below isn't sufficient alone. | |
426 | */ | |
427 | synchronize_rcu(); | |
428 | ||
429 | /* Run all of the network namespace exit methods */ | |
430 | list_for_each_entry_reverse(ops, &pernet_list, list) | |
431 | ops_exit_list(ops, &net_exit_list); | |
432 | ||
433 | /* Free the net generic variables */ | |
434 | list_for_each_entry_reverse(ops, &pernet_list, list) | |
435 | ops_free_list(ops, &net_exit_list); | |
436 | ||
437 | mutex_unlock(&net_mutex); | |
438 | ||
439 | /* Ensure there are no outstanding rcu callbacks using this | |
440 | * network namespace. | |
441 | */ | |
442 | rcu_barrier(); | |
443 | ||
444 | /* Finally it is safe to free my network namespace structure */ | |
445 | list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) { | |
446 | list_del_init(&net->exit_list); | |
447 | put_user_ns(net->user_ns); | |
448 | net_drop_ns(net); | |
449 | } | |
450 | } | |
451 | static DECLARE_WORK(net_cleanup_work, cleanup_net); | |
452 | ||
453 | void __put_net(struct net *net) | |
454 | { | |
455 | /* Cleanup the network namespace in process context */ | |
456 | unsigned long flags; | |
457 | ||
458 | spin_lock_irqsave(&cleanup_list_lock, flags); | |
459 | list_add(&net->cleanup_list, &cleanup_list); | |
460 | spin_unlock_irqrestore(&cleanup_list_lock, flags); | |
461 | ||
462 | queue_work(netns_wq, &net_cleanup_work); | |
463 | } | |
464 | EXPORT_SYMBOL_GPL(__put_net); | |
465 | ||
466 | struct net *get_net_ns_by_fd(int fd) | |
467 | { | |
468 | struct file *file; | |
469 | struct ns_common *ns; | |
470 | struct net *net; | |
471 | ||
472 | file = proc_ns_fget(fd); | |
473 | if (IS_ERR(file)) | |
474 | return ERR_CAST(file); | |
475 | ||
476 | ns = get_proc_ns(file_inode(file)); | |
477 | if (ns->ops == &netns_operations) | |
478 | net = get_net(container_of(ns, struct net, ns)); | |
479 | else | |
480 | net = ERR_PTR(-EINVAL); | |
481 | ||
482 | fput(file); | |
483 | return net; | |
484 | } | |
485 | ||
486 | #else | |
487 | struct net *get_net_ns_by_fd(int fd) | |
488 | { | |
489 | return ERR_PTR(-EINVAL); | |
490 | } | |
491 | #endif | |
492 | EXPORT_SYMBOL_GPL(get_net_ns_by_fd); | |
493 | ||
494 | struct net *get_net_ns_by_pid(pid_t pid) | |
495 | { | |
496 | struct task_struct *tsk; | |
497 | struct net *net; | |
498 | ||
499 | /* Lookup the network namespace */ | |
500 | net = ERR_PTR(-ESRCH); | |
501 | rcu_read_lock(); | |
502 | tsk = find_task_by_vpid(pid); | |
503 | if (tsk) { | |
504 | struct nsproxy *nsproxy; | |
505 | task_lock(tsk); | |
506 | nsproxy = tsk->nsproxy; | |
507 | if (nsproxy) | |
508 | net = get_net(nsproxy->net_ns); | |
509 | task_unlock(tsk); | |
510 | } | |
511 | rcu_read_unlock(); | |
512 | return net; | |
513 | } | |
514 | EXPORT_SYMBOL_GPL(get_net_ns_by_pid); | |
515 | ||
516 | static __net_init int net_ns_net_init(struct net *net) | |
517 | { | |
518 | #ifdef CONFIG_NET_NS | |
519 | net->ns.ops = &netns_operations; | |
520 | #endif | |
521 | return ns_alloc_inum(&net->ns); | |
522 | } | |
523 | ||
524 | static __net_exit void net_ns_net_exit(struct net *net) | |
525 | { | |
526 | ns_free_inum(&net->ns); | |
527 | } | |
528 | ||
529 | static struct pernet_operations __net_initdata net_ns_ops = { | |
530 | .init = net_ns_net_init, | |
531 | .exit = net_ns_net_exit, | |
532 | }; | |
533 | ||
534 | static struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = { | |
535 | [NETNSA_NONE] = { .type = NLA_UNSPEC }, | |
536 | [NETNSA_NSID] = { .type = NLA_S32 }, | |
537 | [NETNSA_PID] = { .type = NLA_U32 }, | |
538 | [NETNSA_FD] = { .type = NLA_U32 }, | |
539 | }; | |
540 | ||
541 | static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
542 | { | |
543 | struct net *net = sock_net(skb->sk); | |
544 | struct nlattr *tb[NETNSA_MAX + 1]; | |
545 | unsigned long flags; | |
546 | struct net *peer; | |
547 | int nsid, err; | |
548 | ||
549 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
550 | rtnl_net_policy); | |
551 | if (err < 0) | |
552 | return err; | |
553 | if (!tb[NETNSA_NSID]) | |
554 | return -EINVAL; | |
555 | nsid = nla_get_s32(tb[NETNSA_NSID]); | |
556 | ||
557 | if (tb[NETNSA_PID]) | |
558 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); | |
559 | else if (tb[NETNSA_FD]) | |
560 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); | |
561 | else | |
562 | return -EINVAL; | |
563 | if (IS_ERR(peer)) | |
564 | return PTR_ERR(peer); | |
565 | ||
566 | spin_lock_irqsave(&nsid_lock, flags); | |
567 | if (__peernet2id(net, peer) >= 0) { | |
568 | err = -EEXIST; | |
569 | goto out; | |
570 | } | |
571 | ||
572 | err = alloc_netid(net, peer, nsid); | |
573 | spin_unlock_irqrestore(&nsid_lock, flags); | |
574 | if (err >= 0) { | |
575 | rtnl_net_notifyid(net, RTM_NEWNSID, err); | |
576 | err = 0; | |
577 | } | |
578 | out: | |
579 | put_net(peer); | |
580 | return err; | |
581 | } | |
582 | ||
583 | static int rtnl_net_get_size(void) | |
584 | { | |
585 | return NLMSG_ALIGN(sizeof(struct rtgenmsg)) | |
586 | + nla_total_size(sizeof(s32)) /* NETNSA_NSID */ | |
587 | ; | |
588 | } | |
589 | ||
590 | static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags, | |
591 | int cmd, struct net *net, int nsid) | |
592 | { | |
593 | struct nlmsghdr *nlh; | |
594 | struct rtgenmsg *rth; | |
595 | ||
596 | nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags); | |
597 | if (!nlh) | |
598 | return -EMSGSIZE; | |
599 | ||
600 | rth = nlmsg_data(nlh); | |
601 | rth->rtgen_family = AF_UNSPEC; | |
602 | ||
603 | if (nla_put_s32(skb, NETNSA_NSID, nsid)) | |
604 | goto nla_put_failure; | |
605 | ||
606 | nlmsg_end(skb, nlh); | |
607 | return 0; | |
608 | ||
609 | nla_put_failure: | |
610 | nlmsg_cancel(skb, nlh); | |
611 | return -EMSGSIZE; | |
612 | } | |
613 | ||
614 | static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh) | |
615 | { | |
616 | struct net *net = sock_net(skb->sk); | |
617 | struct nlattr *tb[NETNSA_MAX + 1]; | |
618 | struct sk_buff *msg; | |
619 | struct net *peer; | |
620 | int err, id; | |
621 | ||
622 | err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX, | |
623 | rtnl_net_policy); | |
624 | if (err < 0) | |
625 | return err; | |
626 | if (tb[NETNSA_PID]) | |
627 | peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID])); | |
628 | else if (tb[NETNSA_FD]) | |
629 | peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD])); | |
630 | else | |
631 | return -EINVAL; | |
632 | ||
633 | if (IS_ERR(peer)) | |
634 | return PTR_ERR(peer); | |
635 | ||
636 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
637 | if (!msg) { | |
638 | err = -ENOMEM; | |
639 | goto out; | |
640 | } | |
641 | ||
642 | id = peernet2id(net, peer); | |
643 | err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, | |
644 | RTM_GETNSID, net, id); | |
645 | if (err < 0) | |
646 | goto err_out; | |
647 | ||
648 | err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid); | |
649 | goto out; | |
650 | ||
651 | err_out: | |
652 | nlmsg_free(msg); | |
653 | out: | |
654 | put_net(peer); | |
655 | return err; | |
656 | } | |
657 | ||
658 | struct rtnl_net_dump_cb { | |
659 | struct net *net; | |
660 | struct sk_buff *skb; | |
661 | struct netlink_callback *cb; | |
662 | int idx; | |
663 | int s_idx; | |
664 | }; | |
665 | ||
666 | static int rtnl_net_dumpid_one(int id, void *peer, void *data) | |
667 | { | |
668 | struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data; | |
669 | int ret; | |
670 | ||
671 | if (net_cb->idx < net_cb->s_idx) | |
672 | goto cont; | |
673 | ||
674 | ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid, | |
675 | net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI, | |
676 | RTM_NEWNSID, net_cb->net, id); | |
677 | if (ret < 0) | |
678 | return ret; | |
679 | ||
680 | cont: | |
681 | net_cb->idx++; | |
682 | return 0; | |
683 | } | |
684 | ||
685 | static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb) | |
686 | { | |
687 | struct net *net = sock_net(skb->sk); | |
688 | struct rtnl_net_dump_cb net_cb = { | |
689 | .net = net, | |
690 | .skb = skb, | |
691 | .cb = cb, | |
692 | .idx = 0, | |
693 | .s_idx = cb->args[0], | |
694 | }; | |
695 | unsigned long flags; | |
696 | ||
697 | spin_lock_irqsave(&nsid_lock, flags); | |
698 | idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb); | |
699 | spin_unlock_irqrestore(&nsid_lock, flags); | |
700 | ||
701 | cb->args[0] = net_cb.idx; | |
702 | return skb->len; | |
703 | } | |
704 | ||
705 | static void rtnl_net_notifyid(struct net *net, int cmd, int id) | |
706 | { | |
707 | struct sk_buff *msg; | |
708 | int err = -ENOMEM; | |
709 | ||
710 | msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL); | |
711 | if (!msg) | |
712 | goto out; | |
713 | ||
714 | err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, id); | |
715 | if (err < 0) | |
716 | goto err_out; | |
717 | ||
718 | rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0); | |
719 | return; | |
720 | ||
721 | err_out: | |
722 | nlmsg_free(msg); | |
723 | out: | |
724 | rtnl_set_sk_err(net, RTNLGRP_NSID, err); | |
725 | } | |
726 | ||
727 | static int __init net_ns_init(void) | |
728 | { | |
729 | struct net_generic *ng; | |
730 | ||
731 | #ifdef CONFIG_NET_NS | |
732 | net_cachep = kmem_cache_create("net_namespace", sizeof(struct net), | |
733 | SMP_CACHE_BYTES, | |
734 | SLAB_PANIC, NULL); | |
735 | ||
736 | /* Create workqueue for cleanup */ | |
737 | netns_wq = create_singlethread_workqueue("netns"); | |
738 | if (!netns_wq) | |
739 | panic("Could not create netns workq"); | |
740 | #endif | |
741 | ||
742 | ng = net_alloc_generic(); | |
743 | if (!ng) | |
744 | panic("Could not allocate generic netns"); | |
745 | ||
746 | rcu_assign_pointer(init_net.gen, ng); | |
747 | ||
748 | mutex_lock(&net_mutex); | |
749 | if (setup_net(&init_net, &init_user_ns)) | |
750 | panic("Could not setup the initial network namespace"); | |
751 | ||
752 | rtnl_lock(); | |
753 | list_add_tail_rcu(&init_net.list, &net_namespace_list); | |
754 | rtnl_unlock(); | |
755 | ||
756 | mutex_unlock(&net_mutex); | |
757 | ||
758 | register_pernet_subsys(&net_ns_ops); | |
759 | ||
760 | rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL); | |
761 | rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid, | |
762 | NULL); | |
763 | ||
764 | return 0; | |
765 | } | |
766 | ||
767 | pure_initcall(net_ns_init); | |
768 | ||
769 | #ifdef CONFIG_NET_NS | |
770 | static int __register_pernet_operations(struct list_head *list, | |
771 | struct pernet_operations *ops) | |
772 | { | |
773 | struct net *net; | |
774 | int error; | |
775 | LIST_HEAD(net_exit_list); | |
776 | ||
777 | list_add_tail(&ops->list, list); | |
778 | if (ops->init || (ops->id && ops->size)) { | |
779 | for_each_net(net) { | |
780 | error = ops_init(ops, net); | |
781 | if (error) | |
782 | goto out_undo; | |
783 | list_add_tail(&net->exit_list, &net_exit_list); | |
784 | } | |
785 | } | |
786 | return 0; | |
787 | ||
788 | out_undo: | |
789 | /* If I have an error cleanup all namespaces I initialized */ | |
790 | list_del(&ops->list); | |
791 | ops_exit_list(ops, &net_exit_list); | |
792 | ops_free_list(ops, &net_exit_list); | |
793 | return error; | |
794 | } | |
795 | ||
796 | static void __unregister_pernet_operations(struct pernet_operations *ops) | |
797 | { | |
798 | struct net *net; | |
799 | LIST_HEAD(net_exit_list); | |
800 | ||
801 | list_del(&ops->list); | |
802 | for_each_net(net) | |
803 | list_add_tail(&net->exit_list, &net_exit_list); | |
804 | ops_exit_list(ops, &net_exit_list); | |
805 | ops_free_list(ops, &net_exit_list); | |
806 | } | |
807 | ||
808 | #else | |
809 | ||
810 | static int __register_pernet_operations(struct list_head *list, | |
811 | struct pernet_operations *ops) | |
812 | { | |
813 | return ops_init(ops, &init_net); | |
814 | } | |
815 | ||
816 | static void __unregister_pernet_operations(struct pernet_operations *ops) | |
817 | { | |
818 | LIST_HEAD(net_exit_list); | |
819 | list_add(&init_net.exit_list, &net_exit_list); | |
820 | ops_exit_list(ops, &net_exit_list); | |
821 | ops_free_list(ops, &net_exit_list); | |
822 | } | |
823 | ||
824 | #endif /* CONFIG_NET_NS */ | |
825 | ||
826 | static DEFINE_IDA(net_generic_ids); | |
827 | ||
828 | static int register_pernet_operations(struct list_head *list, | |
829 | struct pernet_operations *ops) | |
830 | { | |
831 | int error; | |
832 | ||
833 | if (ops->id) { | |
834 | again: | |
835 | error = ida_get_new_above(&net_generic_ids, 1, ops->id); | |
836 | if (error < 0) { | |
837 | if (error == -EAGAIN) { | |
838 | ida_pre_get(&net_generic_ids, GFP_KERNEL); | |
839 | goto again; | |
840 | } | |
841 | return error; | |
842 | } | |
843 | max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id); | |
844 | } | |
845 | error = __register_pernet_operations(list, ops); | |
846 | if (error) { | |
847 | rcu_barrier(); | |
848 | if (ops->id) | |
849 | ida_remove(&net_generic_ids, *ops->id); | |
850 | } | |
851 | ||
852 | return error; | |
853 | } | |
854 | ||
855 | static void unregister_pernet_operations(struct pernet_operations *ops) | |
856 | { | |
857 | ||
858 | __unregister_pernet_operations(ops); | |
859 | rcu_barrier(); | |
860 | if (ops->id) | |
861 | ida_remove(&net_generic_ids, *ops->id); | |
862 | } | |
863 | ||
864 | /** | |
865 | * register_pernet_subsys - register a network namespace subsystem | |
866 | * @ops: pernet operations structure for the subsystem | |
867 | * | |
868 | * Register a subsystem which has init and exit functions | |
869 | * that are called when network namespaces are created and | |
870 | * destroyed respectively. | |
871 | * | |
872 | * When registered all network namespace init functions are | |
873 | * called for every existing network namespace. Allowing kernel | |
874 | * modules to have a race free view of the set of network namespaces. | |
875 | * | |
876 | * When a new network namespace is created all of the init | |
877 | * methods are called in the order in which they were registered. | |
878 | * | |
879 | * When a network namespace is destroyed all of the exit methods | |
880 | * are called in the reverse of the order with which they were | |
881 | * registered. | |
882 | */ | |
883 | int register_pernet_subsys(struct pernet_operations *ops) | |
884 | { | |
885 | int error; | |
886 | mutex_lock(&net_mutex); | |
887 | error = register_pernet_operations(first_device, ops); | |
888 | mutex_unlock(&net_mutex); | |
889 | return error; | |
890 | } | |
891 | EXPORT_SYMBOL_GPL(register_pernet_subsys); | |
892 | ||
893 | /** | |
894 | * unregister_pernet_subsys - unregister a network namespace subsystem | |
895 | * @ops: pernet operations structure to manipulate | |
896 | * | |
897 | * Remove the pernet operations structure from the list to be | |
898 | * used when network namespaces are created or destroyed. In | |
899 | * addition run the exit method for all existing network | |
900 | * namespaces. | |
901 | */ | |
902 | void unregister_pernet_subsys(struct pernet_operations *ops) | |
903 | { | |
904 | mutex_lock(&net_mutex); | |
905 | unregister_pernet_operations(ops); | |
906 | mutex_unlock(&net_mutex); | |
907 | } | |
908 | EXPORT_SYMBOL_GPL(unregister_pernet_subsys); | |
909 | ||
910 | /** | |
911 | * register_pernet_device - register a network namespace device | |
912 | * @ops: pernet operations structure for the subsystem | |
913 | * | |
914 | * Register a device which has init and exit functions | |
915 | * that are called when network namespaces are created and | |
916 | * destroyed respectively. | |
917 | * | |
918 | * When registered all network namespace init functions are | |
919 | * called for every existing network namespace. Allowing kernel | |
920 | * modules to have a race free view of the set of network namespaces. | |
921 | * | |
922 | * When a new network namespace is created all of the init | |
923 | * methods are called in the order in which they were registered. | |
924 | * | |
925 | * When a network namespace is destroyed all of the exit methods | |
926 | * are called in the reverse of the order with which they were | |
927 | * registered. | |
928 | */ | |
929 | int register_pernet_device(struct pernet_operations *ops) | |
930 | { | |
931 | int error; | |
932 | mutex_lock(&net_mutex); | |
933 | error = register_pernet_operations(&pernet_list, ops); | |
934 | if (!error && (first_device == &pernet_list)) | |
935 | first_device = &ops->list; | |
936 | mutex_unlock(&net_mutex); | |
937 | return error; | |
938 | } | |
939 | EXPORT_SYMBOL_GPL(register_pernet_device); | |
940 | ||
941 | /** | |
942 | * unregister_pernet_device - unregister a network namespace netdevice | |
943 | * @ops: pernet operations structure to manipulate | |
944 | * | |
945 | * Remove the pernet operations structure from the list to be | |
946 | * used when network namespaces are created or destroyed. In | |
947 | * addition run the exit method for all existing network | |
948 | * namespaces. | |
949 | */ | |
950 | void unregister_pernet_device(struct pernet_operations *ops) | |
951 | { | |
952 | mutex_lock(&net_mutex); | |
953 | if (&ops->list == first_device) | |
954 | first_device = first_device->next; | |
955 | unregister_pernet_operations(ops); | |
956 | mutex_unlock(&net_mutex); | |
957 | } | |
958 | EXPORT_SYMBOL_GPL(unregister_pernet_device); | |
959 | ||
960 | #ifdef CONFIG_NET_NS | |
961 | static struct ns_common *netns_get(struct task_struct *task) | |
962 | { | |
963 | struct net *net = NULL; | |
964 | struct nsproxy *nsproxy; | |
965 | ||
966 | task_lock(task); | |
967 | nsproxy = task->nsproxy; | |
968 | if (nsproxy) | |
969 | net = get_net(nsproxy->net_ns); | |
970 | task_unlock(task); | |
971 | ||
972 | return net ? &net->ns : NULL; | |
973 | } | |
974 | ||
975 | static inline struct net *to_net_ns(struct ns_common *ns) | |
976 | { | |
977 | return container_of(ns, struct net, ns); | |
978 | } | |
979 | ||
980 | static void netns_put(struct ns_common *ns) | |
981 | { | |
982 | put_net(to_net_ns(ns)); | |
983 | } | |
984 | ||
985 | static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns) | |
986 | { | |
987 | struct net *net = to_net_ns(ns); | |
988 | ||
989 | if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) || | |
990 | !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) | |
991 | return -EPERM; | |
992 | ||
993 | put_net(nsproxy->net_ns); | |
994 | nsproxy->net_ns = get_net(net); | |
995 | return 0; | |
996 | } | |
997 | ||
998 | const struct proc_ns_operations netns_operations = { | |
999 | .name = "net", | |
1000 | .type = CLONE_NEWNET, | |
1001 | .get = netns_get, | |
1002 | .put = netns_put, | |
1003 | .install = netns_install, | |
1004 | }; | |
1005 | #endif |