blob: cb5290b8c428c5c348b25d842ab9cf3797b70eba [file] [log] [blame]
Joe Perchese005d192012-05-16 19:58:40 +00001#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
2
Eric W. Biederman5f256be2007-09-12 11:50:50 +02003#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>
Eric W. Biederman9dd776b2007-09-26 22:04:26 -07009#include <linux/sched.h>
Pavel Emelyanovc93cf612008-04-15 00:35:23 -070010#include <linux/idr.h>
Johannes Berg11a28d32009-07-10 09:51:33 +000011#include <linux/rculist.h>
Johannes Berg30ffee82009-07-10 09:51:35 +000012#include <linux/nsproxy.h>
David Howells0bb80f22013-04-12 01:50:06 +010013#include <linux/fs.h>
14#include <linux/proc_ns.h>
Eric W. Biedermanf0630522011-05-04 17:51:50 -070015#include <linux/file.h>
Paul Gortmakerbc3b2d72011-07-15 11:47:34 -040016#include <linux/export.h>
Eric W. Biederman038e7332012-06-14 02:31:10 -070017#include <linux/user_namespace.h>
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +010018#include <linux/net_namespace.h>
19#include <linux/rtnetlink.h>
20#include <net/sock.h>
21#include <net/netlink.h>
Eric W. Biederman5f256be2007-09-12 11:50:50 +020022#include <net/net_namespace.h>
Pavel Emelyanovdec827d2008-04-15 00:36:08 -070023#include <net/netns/generic.h>
Eric W. Biederman5f256be2007-09-12 11:50:50 +020024
25/*
26 * Our network namespace constructor/destructor lists
27 */
28
29static LIST_HEAD(pernet_list);
30static struct list_head *first_device = &pernet_list;
Cong Wang200b9162014-05-12 15:11:20 -070031DEFINE_MUTEX(net_mutex);
Eric W. Biederman5f256be2007-09-12 11:50:50 +020032
Eric W. Biederman5f256be2007-09-12 11:50:50 +020033LIST_HEAD(net_namespace_list);
Alexey Dobriyanb76a4612008-10-08 11:35:06 +020034EXPORT_SYMBOL_GPL(net_namespace_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +020035
Rustad, Mark D734b6542012-07-18 09:06:07 +000036struct net init_net = {
37 .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head),
38};
Denis V. Lunevff4b9502008-01-22 22:05:33 -080039EXPORT_SYMBOL(init_net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +020040
Pavel Emelyanovdec827d2008-04-15 00:36:08 -070041#define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
42
Eric Dumazet073862b2012-01-26 00:41:38 +000043static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS;
44
45static 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
Jiri Pirko05fceb42010-04-23 01:40:47 +000057static 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
Eric Dumazet1c877332010-10-25 03:20:11 +000064 old_ng = rcu_dereference_protected(net->gen,
65 lockdep_is_held(&net_mutex));
66 ng = old_ng;
Jiri Pirko05fceb42010-04-23 01:40:47 +000067 if (old_ng->len >= id)
68 goto assign;
69
Eric Dumazet073862b2012-01-26 00:41:38 +000070 ng = net_alloc_generic();
Jiri Pirko05fceb42010-04-23 01:40:47 +000071 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
Jiri Pirko05fceb42010-04-23 01:40:47 +000085 memcpy(&ng->ptr, &old_ng->ptr, old_ng->len * sizeof(void*));
86
87 rcu_assign_pointer(net->gen, ng);
Lai Jiangshan04d4dfe2011-03-18 12:06:32 +080088 kfree_rcu(old_ng, rcu);
Jiri Pirko05fceb42010-04-23 01:40:47 +000089assign:
90 ng->ptr[id - 1] = data;
91 return 0;
92}
93
Eric W. Biedermanf875bae2009-11-29 22:25:28 +000094static int ops_init(const struct pernet_operations *ops, struct net *net)
95{
Julian Anastasovb9229342012-04-16 04:43:15 +000096 int err = -ENOMEM;
97 void *data = NULL;
98
Eric W. Biedermanf875bae2009-11-29 22:25:28 +000099 if (ops->id && ops->size) {
Julian Anastasovb9229342012-04-16 04:43:15 +0000100 data = kzalloc(ops->size, GFP_KERNEL);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000101 if (!data)
Julian Anastasovb9229342012-04-16 04:43:15 +0000102 goto out;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000103
104 err = net_assign_generic(net, *ops->id, data);
Julian Anastasovb9229342012-04-16 04:43:15 +0000105 if (err)
106 goto cleanup;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000107 }
Julian Anastasovb9229342012-04-16 04:43:15 +0000108 err = 0;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000109 if (ops->init)
Julian Anastasovb9229342012-04-16 04:43:15 +0000110 err = ops->init(net);
111 if (!err)
112 return 0;
113
114cleanup:
115 kfree(data);
116
117out:
118 return err;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000119}
120
121static 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
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000129static 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 }
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000137 if (ops->exit_batch)
138 ops->exit_batch(net_exit_list);
139}
140
141static 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
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100151static int alloc_netid(struct net *net, struct net *peer, int reqid)
152{
153 int min = 0, max = 0;
154
155 ASSERT_RTNL();
156
157 if (reqid >= 0) {
158 min = reqid;
159 max = reqid + 1;
160 }
161
162 return idr_alloc(&net->netns_ids, peer, min, max, GFP_KERNEL);
163}
164
165/* This function is used by idr_for_each(). If net is equal to peer, the
166 * function returns the id so that idr_for_each() stops. Because we cannot
167 * returns the id 0 (idr_for_each() will not stop), we return the magic value
168 * NET_ID_ZERO (-1) for it.
169 */
170#define NET_ID_ZERO -1
171static int net_eq_idr(int id, void *net, void *peer)
172{
173 if (net_eq(net, peer))
174 return id ? : NET_ID_ZERO;
175 return 0;
176}
177
178static int __peernet2id(struct net *net, struct net *peer, bool alloc)
179{
180 int id = idr_for_each(&net->netns_ids, net_eq_idr, peer);
181
182 ASSERT_RTNL();
183
184 /* Magic value for id 0. */
185 if (id == NET_ID_ZERO)
186 return 0;
187 if (id > 0)
188 return id;
189
190 if (alloc)
191 return alloc_netid(net, peer, -1);
192
193 return -ENOENT;
194}
195
196/* This function returns the id of a peer netns. If no id is assigned, one will
197 * be allocated and returned.
198 */
199int peernet2id(struct net *net, struct net *peer)
200{
201 int id = __peernet2id(net, peer, true);
202
203 return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED;
204}
Nicolas Dichtel193523b2015-01-20 15:15:47 +0100205EXPORT_SYMBOL(peernet2id);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100206
207struct net *get_net_ns_by_id(struct net *net, int id)
208{
209 struct net *peer;
210
211 if (id < 0)
212 return NULL;
213
214 rcu_read_lock();
215 peer = idr_find(&net->netns_ids, id);
216 if (peer)
217 get_net(peer);
218 rcu_read_unlock();
219
220 return peer;
221}
222
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700223/*
224 * setup_net runs the initializers for the network namespace object.
225 */
Eric W. Biederman038e7332012-06-14 02:31:10 -0700226static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700227{
228 /* Must be called with net_mutex held */
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000229 const struct pernet_operations *ops, *saved_ops;
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800230 int error = 0;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000231 LIST_HEAD(net_exit_list);
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700232
233 atomic_set(&net->count, 1);
Al Viroa685e082011-06-08 21:13:01 -0400234 atomic_set(&net->passive, 1);
Thomas Graf4e985ad2011-06-21 03:11:20 +0000235 net->dev_base_seq = 1;
Eric W. Biederman038e7332012-06-14 02:31:10 -0700236 net->user_ns = user_ns;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100237 idr_init(&net->netns_ids);
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800238
Denis V. Lunev5d1e4462008-04-16 01:58:04 -0700239#ifdef NETNS_REFCNT_DEBUG
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700240 atomic_set(&net->use_count, 0);
Denis V. Lunev5d1e4462008-04-16 01:58:04 -0700241#endif
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700242
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700243 list_for_each_entry(ops, &pernet_list, list) {
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000244 error = ops_init(ops, net);
245 if (error < 0)
246 goto out_undo;
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700247 }
248out:
249 return error;
250
251out_undo:
252 /* Walk through the list backwards calling the exit functions
253 * for the pernet modules whose init functions did not fail.
254 */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000255 list_add(&net->exit_list, &net_exit_list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000256 saved_ops = ops;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000257 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
258 ops_exit_list(ops, &net_exit_list);
259
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000260 ops = saved_ops;
261 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000262 ops_free_list(ops, &net_exit_list);
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700263
264 rcu_barrier();
265 goto out;
266}
267
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800268
Clemens Nossebe47d42009-02-23 15:37:35 -0800269#ifdef CONFIG_NET_NS
270static struct kmem_cache *net_cachep;
271static struct workqueue_struct *netns_wq;
272
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200273static struct net *net_alloc(void)
274{
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800275 struct net *net = NULL;
276 struct net_generic *ng;
277
278 ng = net_alloc_generic();
279 if (!ng)
280 goto out;
281
282 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL);
283 if (!net)
284 goto out_free;
285
286 rcu_assign_pointer(net->gen, ng);
287out:
288 return net;
289
290out_free:
291 kfree(ng);
292 goto out;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200293}
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200294
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800295static void net_free(struct net *net)
296{
Denis V. Lunev5d1e4462008-04-16 01:58:04 -0700297#ifdef NETNS_REFCNT_DEBUG
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800298 if (unlikely(atomic_read(&net->use_count) != 0)) {
Joe Perchese005d192012-05-16 19:58:40 +0000299 pr_emerg("network namespace not free! Usage: %d\n",
300 atomic_read(&net->use_count));
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800301 return;
302 }
Denis V. Lunev5d1e4462008-04-16 01:58:04 -0700303#endif
Eric Dumazet416c51e2014-09-09 08:24:53 -0700304 kfree(rcu_access_pointer(net->gen));
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800305 kmem_cache_free(net_cachep, net);
306}
307
Al Viroa685e082011-06-08 21:13:01 -0400308void net_drop_ns(void *p)
309{
310 struct net *ns = p;
311 if (ns && atomic_dec_and_test(&ns->passive))
312 net_free(ns);
313}
314
Eric W. Biederman038e7332012-06-14 02:31:10 -0700315struct net *copy_net_ns(unsigned long flags,
316 struct user_namespace *user_ns, struct net *old_net)
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700317{
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700318 struct net *net;
319 int rv;
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700320
Rob Landley911cb192011-04-15 02:26:25 +0000321 if (!(flags & CLONE_NEWNET))
322 return get_net(old_net);
323
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700324 net = net_alloc();
325 if (!net)
326 return ERR_PTR(-ENOMEM);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700327
328 get_user_ns(user_ns);
329
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700330 mutex_lock(&net_mutex);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700331 rv = setup_net(net, user_ns);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700332 if (rv == 0) {
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800333 rtnl_lock();
Johannes Berg11a28d32009-07-10 09:51:33 +0000334 list_add_tail_rcu(&net->list, &net_namespace_list);
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800335 rtnl_unlock();
336 }
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700337 mutex_unlock(&net_mutex);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700338 if (rv < 0) {
Eric W. Biederman038e7332012-06-14 02:31:10 -0700339 put_user_ns(user_ns);
Al Viroa685e082011-06-08 21:13:01 -0400340 net_drop_ns(net);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700341 return ERR_PTR(rv);
342 }
343 return net;
344}
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800345
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000346static DEFINE_SPINLOCK(cleanup_list_lock);
347static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */
348
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200349static void cleanup_net(struct work_struct *work)
350{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000351 const struct pernet_operations *ops;
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200352 struct net *net, *tmp;
xiao jin1818ce42014-04-25 08:50:54 +0800353 struct list_head net_kill_list;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000354 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200355
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000356 /* Atomically snapshot the list of namespaces to cleanup */
357 spin_lock_irq(&cleanup_list_lock);
358 list_replace_init(&cleanup_list, &net_kill_list);
359 spin_unlock_irq(&cleanup_list_lock);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200360
361 mutex_lock(&net_mutex);
362
363 /* Don't let anyone else find us. */
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700364 rtnl_lock();
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000365 list_for_each_entry(net, &net_kill_list, cleanup_list) {
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000366 list_del_rcu(&net->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000367 list_add_tail(&net->exit_list, &net_exit_list);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200368 for_each_net(tmp) {
369 int id = __peernet2id(tmp, net, false);
370
371 if (id >= 0)
372 idr_remove(&tmp->netns_ids, id);
373 }
374 idr_destroy(&net->netns_ids);
375
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000376 }
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700377 rtnl_unlock();
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200378
Johannes Berg11a28d32009-07-10 09:51:33 +0000379 /*
380 * Another CPU might be rcu-iterating the list, wait for it.
381 * This needs to be before calling the exit() notifiers, so
382 * the rcu_barrier() below isn't sufficient alone.
383 */
384 synchronize_rcu();
385
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200386 /* Run all of the network namespace exit methods */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000387 list_for_each_entry_reverse(ops, &pernet_list, list)
388 ops_exit_list(ops, &net_exit_list);
389
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000390 /* Free the net generic variables */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000391 list_for_each_entry_reverse(ops, &pernet_list, list)
392 ops_free_list(ops, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200393
394 mutex_unlock(&net_mutex);
395
396 /* Ensure there are no outstanding rcu callbacks using this
397 * network namespace.
398 */
399 rcu_barrier();
400
401 /* Finally it is safe to free my network namespace structure */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000402 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
403 list_del_init(&net->exit_list);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700404 put_user_ns(net->user_ns);
Al Viroa685e082011-06-08 21:13:01 -0400405 net_drop_ns(net);
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000406 }
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200407}
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000408static DECLARE_WORK(net_cleanup_work, cleanup_net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200409
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200410void __put_net(struct net *net)
411{
412 /* Cleanup the network namespace in process context */
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000413 unsigned long flags;
414
415 spin_lock_irqsave(&cleanup_list_lock, flags);
416 list_add(&net->cleanup_list, &cleanup_list);
417 spin_unlock_irqrestore(&cleanup_list_lock, flags);
418
419 queue_work(netns_wq, &net_cleanup_work);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200420}
421EXPORT_SYMBOL_GPL(__put_net);
422
Stephen Rothwell956c9202011-05-12 13:51:13 +1000423struct net *get_net_ns_by_fd(int fd)
424{
Stephen Rothwell956c9202011-05-12 13:51:13 +1000425 struct file *file;
Al Viro33c42942014-11-01 02:32:53 -0400426 struct ns_common *ns;
Stephen Rothwell956c9202011-05-12 13:51:13 +1000427 struct net *net;
428
Stephen Rothwell956c9202011-05-12 13:51:13 +1000429 file = proc_ns_fget(fd);
Al Viroc316e6a2011-06-05 00:37:35 +0000430 if (IS_ERR(file))
431 return ERR_CAST(file);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000432
Al Virof77c8012014-11-01 03:13:17 -0400433 ns = get_proc_ns(file_inode(file));
Al Viro33c42942014-11-01 02:32:53 -0400434 if (ns->ops == &netns_operations)
435 net = get_net(container_of(ns, struct net, ns));
Al Viroc316e6a2011-06-05 00:37:35 +0000436 else
437 net = ERR_PTR(-EINVAL);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000438
Al Viroc316e6a2011-06-05 00:37:35 +0000439 fput(file);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000440 return net;
441}
442
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700443#else
Stephen Rothwell956c9202011-05-12 13:51:13 +1000444struct net *get_net_ns_by_fd(int fd)
445{
446 return ERR_PTR(-EINVAL);
447}
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700448#endif
Vadim Kochan4b681c82015-01-12 16:34:05 +0200449EXPORT_SYMBOL_GPL(get_net_ns_by_fd);
Eric W. Biederman9dd776b2007-09-26 22:04:26 -0700450
Johannes Berg30ffee82009-07-10 09:51:35 +0000451struct net *get_net_ns_by_pid(pid_t pid)
452{
453 struct task_struct *tsk;
454 struct net *net;
455
456 /* Lookup the network namespace */
457 net = ERR_PTR(-ESRCH);
458 rcu_read_lock();
459 tsk = find_task_by_vpid(pid);
460 if (tsk) {
461 struct nsproxy *nsproxy;
Eric W. Biederman728dba32014-02-03 19:13:49 -0800462 task_lock(tsk);
463 nsproxy = tsk->nsproxy;
Johannes Berg30ffee82009-07-10 09:51:35 +0000464 if (nsproxy)
465 net = get_net(nsproxy->net_ns);
Eric W. Biederman728dba32014-02-03 19:13:49 -0800466 task_unlock(tsk);
Johannes Berg30ffee82009-07-10 09:51:35 +0000467 }
468 rcu_read_unlock();
469 return net;
470}
471EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
472
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700473static __net_init int net_ns_net_init(struct net *net)
474{
Al Viro33c42942014-11-01 02:32:53 -0400475#ifdef CONFIG_NET_NS
476 net->ns.ops = &netns_operations;
477#endif
Al Viro6344c432014-11-01 00:45:45 -0400478 return ns_alloc_inum(&net->ns);
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700479}
480
481static __net_exit void net_ns_net_exit(struct net *net)
482{
Al Viro6344c432014-11-01 00:45:45 -0400483 ns_free_inum(&net->ns);
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700484}
485
486static struct pernet_operations __net_initdata net_ns_ops = {
487 .init = net_ns_net_init,
488 .exit = net_ns_net_exit,
489};
490
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100491static struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = {
492 [NETNSA_NONE] = { .type = NLA_UNSPEC },
493 [NETNSA_NSID] = { .type = NLA_S32 },
494 [NETNSA_PID] = { .type = NLA_U32 },
495 [NETNSA_FD] = { .type = NLA_U32 },
496};
497
498static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh)
499{
500 struct net *net = sock_net(skb->sk);
501 struct nlattr *tb[NETNSA_MAX + 1];
502 struct net *peer;
503 int nsid, err;
504
505 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
506 rtnl_net_policy);
507 if (err < 0)
508 return err;
509 if (!tb[NETNSA_NSID])
510 return -EINVAL;
511 nsid = nla_get_s32(tb[NETNSA_NSID]);
512
513 if (tb[NETNSA_PID])
514 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
515 else if (tb[NETNSA_FD])
516 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
517 else
518 return -EINVAL;
519 if (IS_ERR(peer))
520 return PTR_ERR(peer);
521
522 if (__peernet2id(net, peer, false) >= 0) {
523 err = -EEXIST;
524 goto out;
525 }
526
527 err = alloc_netid(net, peer, nsid);
528 if (err > 0)
529 err = 0;
530out:
531 put_net(peer);
532 return err;
533}
534
535static int rtnl_net_get_size(void)
536{
537 return NLMSG_ALIGN(sizeof(struct rtgenmsg))
538 + nla_total_size(sizeof(s32)) /* NETNSA_NSID */
539 ;
540}
541
542static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags,
543 int cmd, struct net *net, struct net *peer)
544{
545 struct nlmsghdr *nlh;
546 struct rtgenmsg *rth;
547 int id;
548
549 ASSERT_RTNL();
550
551 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags);
552 if (!nlh)
553 return -EMSGSIZE;
554
555 rth = nlmsg_data(nlh);
556 rth->rtgen_family = AF_UNSPEC;
557
558 id = __peernet2id(net, peer, false);
559 if (id < 0)
560 id = NETNSA_NSID_NOT_ASSIGNED;
561 if (nla_put_s32(skb, NETNSA_NSID, id))
562 goto nla_put_failure;
563
564 nlmsg_end(skb, nlh);
565 return 0;
566
567nla_put_failure:
568 nlmsg_cancel(skb, nlh);
569 return -EMSGSIZE;
570}
571
572static int rtnl_net_getid(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 sk_buff *msg;
577 int err = -ENOBUFS;
578 struct net *peer;
579
580 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
581 rtnl_net_policy);
582 if (err < 0)
583 return err;
584 if (tb[NETNSA_PID])
585 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
586 else if (tb[NETNSA_FD])
587 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
588 else
589 return -EINVAL;
590
591 if (IS_ERR(peer))
592 return PTR_ERR(peer);
593
594 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
595 if (!msg) {
596 err = -ENOMEM;
597 goto out;
598 }
599
600 err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
601 RTM_GETNSID, net, peer);
602 if (err < 0)
603 goto err_out;
604
605 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid);
606 goto out;
607
608err_out:
609 nlmsg_free(msg);
610out:
611 put_net(peer);
612 return err;
613}
614
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200615static int __init net_ns_init(void)
616{
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800617 struct net_generic *ng;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200618
Pavel Emelyanovd57a9212007-11-01 00:46:50 -0700619#ifdef CONFIG_NET_NS
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200620 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net),
621 SMP_CACHE_BYTES,
622 SLAB_PANIC, NULL);
Benjamin Thery3ef13552007-11-19 23:18:16 -0800623
624 /* Create workqueue for cleanup */
625 netns_wq = create_singlethread_workqueue("netns");
626 if (!netns_wq)
627 panic("Could not create netns workq");
Pavel Emelyanovd57a9212007-11-01 00:46:50 -0700628#endif
Benjamin Thery3ef13552007-11-19 23:18:16 -0800629
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800630 ng = net_alloc_generic();
631 if (!ng)
632 panic("Could not allocate generic netns");
633
634 rcu_assign_pointer(init_net.gen, ng);
635
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200636 mutex_lock(&net_mutex);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700637 if (setup_net(&init_net, &init_user_ns))
Stephen Hemmingerca0f3112009-05-21 15:10:31 -0700638 panic("Could not setup the initial network namespace");
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200639
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700640 rtnl_lock();
Johannes Berg11a28d32009-07-10 09:51:33 +0000641 list_add_tail_rcu(&init_net.list, &net_namespace_list);
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700642 rtnl_unlock();
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200643
644 mutex_unlock(&net_mutex);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200645
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700646 register_pernet_subsys(&net_ns_ops);
647
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100648 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL);
649 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, NULL, NULL);
650
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200651 return 0;
652}
653
654pure_initcall(net_ns_init);
655
Denis V. Luneved160e82007-11-13 03:23:21 -0800656#ifdef CONFIG_NET_NS
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000657static int __register_pernet_operations(struct list_head *list,
658 struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200659{
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000660 struct net *net;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200661 int error;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000662 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200663
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200664 list_add_tail(&ops->list, list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000665 if (ops->init || (ops->id && ops->size)) {
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700666 for_each_net(net) {
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000667 error = ops_init(ops, net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200668 if (error)
669 goto out_undo;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000670 list_add_tail(&net->exit_list, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200671 }
672 }
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700673 return 0;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200674
675out_undo:
676 /* If I have an error cleanup all namespaces I initialized */
677 list_del(&ops->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000678 ops_exit_list(ops, &net_exit_list);
679 ops_free_list(ops, &net_exit_list);
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700680 return error;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200681}
682
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000683static void __unregister_pernet_operations(struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200684{
685 struct net *net;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000686 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200687
688 list_del(&ops->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000689 for_each_net(net)
690 list_add_tail(&net->exit_list, &net_exit_list);
691 ops_exit_list(ops, &net_exit_list);
692 ops_free_list(ops, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200693}
694
Denis V. Luneved160e82007-11-13 03:23:21 -0800695#else
696
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000697static int __register_pernet_operations(struct list_head *list,
698 struct pernet_operations *ops)
699{
Julian Anastasovb9229342012-04-16 04:43:15 +0000700 return ops_init(ops, &init_net);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000701}
702
703static void __unregister_pernet_operations(struct pernet_operations *ops)
704{
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000705 LIST_HEAD(net_exit_list);
706 list_add(&init_net.exit_list, &net_exit_list);
707 ops_exit_list(ops, &net_exit_list);
708 ops_free_list(ops, &net_exit_list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000709}
710
711#endif /* CONFIG_NET_NS */
712
713static DEFINE_IDA(net_generic_ids);
714
Denis V. Luneved160e82007-11-13 03:23:21 -0800715static int register_pernet_operations(struct list_head *list,
716 struct pernet_operations *ops)
717{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000718 int error;
719
720 if (ops->id) {
721again:
722 error = ida_get_new_above(&net_generic_ids, 1, ops->id);
723 if (error < 0) {
724 if (error == -EAGAIN) {
725 ida_pre_get(&net_generic_ids, GFP_KERNEL);
726 goto again;
727 }
728 return error;
729 }
Eric Dumazet073862b2012-01-26 00:41:38 +0000730 max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000731 }
732 error = __register_pernet_operations(list, ops);
Eric W. Biederman3a765ed2009-12-03 02:29:06 +0000733 if (error) {
734 rcu_barrier();
735 if (ops->id)
736 ida_remove(&net_generic_ids, *ops->id);
737 }
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000738
739 return error;
Denis V. Luneved160e82007-11-13 03:23:21 -0800740}
741
742static void unregister_pernet_operations(struct pernet_operations *ops)
743{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000744
745 __unregister_pernet_operations(ops);
Eric W. Biederman3a765ed2009-12-03 02:29:06 +0000746 rcu_barrier();
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000747 if (ops->id)
748 ida_remove(&net_generic_ids, *ops->id);
Denis V. Luneved160e82007-11-13 03:23:21 -0800749}
Pavel Emelyanovc93cf612008-04-15 00:35:23 -0700750
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200751/**
752 * register_pernet_subsys - register a network namespace subsystem
753 * @ops: pernet operations structure for the subsystem
754 *
755 * Register a subsystem which has init and exit functions
756 * that are called when network namespaces are created and
757 * destroyed respectively.
758 *
759 * When registered all network namespace init functions are
760 * called for every existing network namespace. Allowing kernel
761 * modules to have a race free view of the set of network namespaces.
762 *
763 * When a new network namespace is created all of the init
764 * methods are called in the order in which they were registered.
765 *
766 * When a network namespace is destroyed all of the exit methods
767 * are called in the reverse of the order with which they were
768 * registered.
769 */
770int register_pernet_subsys(struct pernet_operations *ops)
771{
772 int error;
773 mutex_lock(&net_mutex);
774 error = register_pernet_operations(first_device, ops);
775 mutex_unlock(&net_mutex);
776 return error;
777}
778EXPORT_SYMBOL_GPL(register_pernet_subsys);
779
780/**
781 * unregister_pernet_subsys - unregister a network namespace subsystem
782 * @ops: pernet operations structure to manipulate
783 *
784 * Remove the pernet operations structure from the list to be
Oliver Pinter53379e52008-02-03 17:56:48 +0200785 * used when network namespaces are created or destroyed. In
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200786 * addition run the exit method for all existing network
787 * namespaces.
788 */
Jiri Pirkob3c981d2010-04-25 00:49:56 -0700789void unregister_pernet_subsys(struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200790{
791 mutex_lock(&net_mutex);
Jiri Pirkob3c981d2010-04-25 00:49:56 -0700792 unregister_pernet_operations(ops);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200793 mutex_unlock(&net_mutex);
794}
795EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
796
797/**
798 * register_pernet_device - register a network namespace device
799 * @ops: pernet operations structure for the subsystem
800 *
801 * Register a device which has init and exit functions
802 * that are called when network namespaces are created and
803 * destroyed respectively.
804 *
805 * When registered all network namespace init functions are
806 * called for every existing network namespace. Allowing kernel
807 * modules to have a race free view of the set of network namespaces.
808 *
809 * When a new network namespace is created all of the init
810 * methods are called in the order in which they were registered.
811 *
812 * When a network namespace is destroyed all of the exit methods
813 * are called in the reverse of the order with which they were
814 * registered.
815 */
816int register_pernet_device(struct pernet_operations *ops)
817{
818 int error;
819 mutex_lock(&net_mutex);
820 error = register_pernet_operations(&pernet_list, ops);
821 if (!error && (first_device == &pernet_list))
822 first_device = &ops->list;
823 mutex_unlock(&net_mutex);
824 return error;
825}
826EXPORT_SYMBOL_GPL(register_pernet_device);
827
828/**
829 * unregister_pernet_device - unregister a network namespace netdevice
830 * @ops: pernet operations structure to manipulate
831 *
832 * Remove the pernet operations structure from the list to be
Oliver Pinter53379e52008-02-03 17:56:48 +0200833 * used when network namespaces are created or destroyed. In
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200834 * addition run the exit method for all existing network
835 * namespaces.
836 */
837void unregister_pernet_device(struct pernet_operations *ops)
838{
839 mutex_lock(&net_mutex);
840 if (&ops->list == first_device)
841 first_device = first_device->next;
842 unregister_pernet_operations(ops);
843 mutex_unlock(&net_mutex);
844}
845EXPORT_SYMBOL_GPL(unregister_pernet_device);
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800846
847#ifdef CONFIG_NET_NS
Al Viro64964522014-11-01 00:37:32 -0400848static struct ns_common *netns_get(struct task_struct *task)
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800849{
Eric W. Biedermanf0630522011-05-04 17:51:50 -0700850 struct net *net = NULL;
851 struct nsproxy *nsproxy;
852
Eric W. Biederman728dba32014-02-03 19:13:49 -0800853 task_lock(task);
854 nsproxy = task->nsproxy;
Eric W. Biedermanf0630522011-05-04 17:51:50 -0700855 if (nsproxy)
856 net = get_net(nsproxy->net_ns);
Eric W. Biederman728dba32014-02-03 19:13:49 -0800857 task_unlock(task);
Eric W. Biedermanf0630522011-05-04 17:51:50 -0700858
Al Viroff248702014-11-01 00:10:50 -0400859 return net ? &net->ns : NULL;
860}
861
862static inline struct net *to_net_ns(struct ns_common *ns)
863{
864 return container_of(ns, struct net, ns);
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800865}
866
Al Viro64964522014-11-01 00:37:32 -0400867static void netns_put(struct ns_common *ns)
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800868{
Al Viroff248702014-11-01 00:10:50 -0400869 put_net(to_net_ns(ns));
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800870}
871
Al Viro64964522014-11-01 00:37:32 -0400872static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns)
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800873{
Al Viroff248702014-11-01 00:10:50 -0400874 struct net *net = to_net_ns(ns);
Eric W. Biederman142e1d12012-07-26 01:13:20 -0700875
Eric W. Biederman5e4a0842012-12-14 07:55:36 -0800876 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700877 !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
Eric W. Biederman142e1d12012-07-26 01:13:20 -0700878 return -EPERM;
879
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800880 put_net(nsproxy->net_ns);
Eric W. Biederman142e1d12012-07-26 01:13:20 -0700881 nsproxy->net_ns = get_net(net);
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800882 return 0;
883}
884
885const struct proc_ns_operations netns_operations = {
886 .name = "net",
887 .type = CLONE_NEWNET,
888 .get = netns_get,
889 .put = netns_put,
890 .install = netns_install,
891};
892#endif