blob: b3b5f22f0e903e3cbee3f8cf7506dad417ccd183 [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 Dichtel9a963452015-04-07 11:51:53 +0200151static void rtnl_net_notifyid(struct net *net, struct net *peer, int cmd,
152 int id);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100153static int alloc_netid(struct net *net, struct net *peer, int reqid)
154{
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200155 int min = 0, max = 0, id;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100156
157 ASSERT_RTNL();
158
159 if (reqid >= 0) {
160 min = reqid;
161 max = reqid + 1;
162 }
163
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200164 id = idr_alloc(&net->netns_ids, peer, min, max, GFP_KERNEL);
165 if (id >= 0)
166 rtnl_net_notifyid(net, peer, RTM_NEWNSID, id);
167
168 return id;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100169}
170
171/* This function is used by idr_for_each(). If net is equal to peer, the
172 * function returns the id so that idr_for_each() stops. Because we cannot
173 * returns the id 0 (idr_for_each() will not stop), we return the magic value
174 * NET_ID_ZERO (-1) for it.
175 */
176#define NET_ID_ZERO -1
177static int net_eq_idr(int id, void *net, void *peer)
178{
179 if (net_eq(net, peer))
180 return id ? : NET_ID_ZERO;
181 return 0;
182}
183
184static int __peernet2id(struct net *net, struct net *peer, bool alloc)
185{
186 int id = idr_for_each(&net->netns_ids, net_eq_idr, peer);
187
188 ASSERT_RTNL();
189
190 /* Magic value for id 0. */
191 if (id == NET_ID_ZERO)
192 return 0;
193 if (id > 0)
194 return id;
195
196 if (alloc)
197 return alloc_netid(net, peer, -1);
198
199 return -ENOENT;
200}
201
202/* This function returns the id of a peer netns. If no id is assigned, one will
203 * be allocated and returned.
204 */
205int peernet2id(struct net *net, struct net *peer)
206{
Nicolas Dichtel576b7cd2015-04-03 12:02:37 +0200207 bool alloc = atomic_read(&peer->count) == 0 ? false : true;
208 int id;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100209
Nicolas Dichtel576b7cd2015-04-03 12:02:37 +0200210 id = __peernet2id(net, peer, alloc);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100211 return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED;
212}
Nicolas Dichtel193523b2015-01-20 15:15:47 +0100213EXPORT_SYMBOL(peernet2id);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100214
215struct net *get_net_ns_by_id(struct net *net, int id)
216{
217 struct net *peer;
218
219 if (id < 0)
220 return NULL;
221
222 rcu_read_lock();
223 peer = idr_find(&net->netns_ids, id);
224 if (peer)
225 get_net(peer);
226 rcu_read_unlock();
227
228 return peer;
229}
230
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700231/*
232 * setup_net runs the initializers for the network namespace object.
233 */
Eric W. Biederman038e7332012-06-14 02:31:10 -0700234static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700235{
236 /* Must be called with net_mutex held */
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000237 const struct pernet_operations *ops, *saved_ops;
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800238 int error = 0;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000239 LIST_HEAD(net_exit_list);
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700240
241 atomic_set(&net->count, 1);
Al Viroa685e082011-06-08 21:13:01 -0400242 atomic_set(&net->passive, 1);
Thomas Graf4e985ad2011-06-21 03:11:20 +0000243 net->dev_base_seq = 1;
Eric W. Biederman038e7332012-06-14 02:31:10 -0700244 net->user_ns = user_ns;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100245 idr_init(&net->netns_ids);
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800246
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700247 list_for_each_entry(ops, &pernet_list, list) {
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000248 error = ops_init(ops, net);
249 if (error < 0)
250 goto out_undo;
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700251 }
252out:
253 return error;
254
255out_undo:
256 /* Walk through the list backwards calling the exit functions
257 * for the pernet modules whose init functions did not fail.
258 */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000259 list_add(&net->exit_list, &net_exit_list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000260 saved_ops = ops;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000261 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
262 ops_exit_list(ops, &net_exit_list);
263
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000264 ops = saved_ops;
265 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000266 ops_free_list(ops, &net_exit_list);
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700267
268 rcu_barrier();
269 goto out;
270}
271
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800272
Clemens Nossebe47d42009-02-23 15:37:35 -0800273#ifdef CONFIG_NET_NS
274static struct kmem_cache *net_cachep;
275static struct workqueue_struct *netns_wq;
276
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200277static struct net *net_alloc(void)
278{
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800279 struct net *net = NULL;
280 struct net_generic *ng;
281
282 ng = net_alloc_generic();
283 if (!ng)
284 goto out;
285
286 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL);
287 if (!net)
288 goto out_free;
289
290 rcu_assign_pointer(net->gen, ng);
291out:
292 return net;
293
294out_free:
295 kfree(ng);
296 goto out;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200297}
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200298
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800299static void net_free(struct net *net)
300{
Eric Dumazet416c51e2014-09-09 08:24:53 -0700301 kfree(rcu_access_pointer(net->gen));
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800302 kmem_cache_free(net_cachep, net);
303}
304
Al Viroa685e082011-06-08 21:13:01 -0400305void net_drop_ns(void *p)
306{
307 struct net *ns = p;
308 if (ns && atomic_dec_and_test(&ns->passive))
309 net_free(ns);
310}
311
Eric W. Biederman038e7332012-06-14 02:31:10 -0700312struct net *copy_net_ns(unsigned long flags,
313 struct user_namespace *user_ns, struct net *old_net)
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700314{
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700315 struct net *net;
316 int rv;
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700317
Rob Landley911cb192011-04-15 02:26:25 +0000318 if (!(flags & CLONE_NEWNET))
319 return get_net(old_net);
320
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700321 net = net_alloc();
322 if (!net)
323 return ERR_PTR(-ENOMEM);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700324
325 get_user_ns(user_ns);
326
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700327 mutex_lock(&net_mutex);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700328 rv = setup_net(net, user_ns);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700329 if (rv == 0) {
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800330 rtnl_lock();
Johannes Berg11a28d32009-07-10 09:51:33 +0000331 list_add_tail_rcu(&net->list, &net_namespace_list);
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800332 rtnl_unlock();
333 }
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700334 mutex_unlock(&net_mutex);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700335 if (rv < 0) {
Eric W. Biederman038e7332012-06-14 02:31:10 -0700336 put_user_ns(user_ns);
Al Viroa685e082011-06-08 21:13:01 -0400337 net_drop_ns(net);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700338 return ERR_PTR(rv);
339 }
340 return net;
341}
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800342
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000343static DEFINE_SPINLOCK(cleanup_list_lock);
344static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */
345
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200346static void cleanup_net(struct work_struct *work)
347{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000348 const struct pernet_operations *ops;
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200349 struct net *net, *tmp;
xiao jin1818ce42014-04-25 08:50:54 +0800350 struct list_head net_kill_list;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000351 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200352
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000353 /* Atomically snapshot the list of namespaces to cleanup */
354 spin_lock_irq(&cleanup_list_lock);
355 list_replace_init(&cleanup_list, &net_kill_list);
356 spin_unlock_irq(&cleanup_list_lock);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200357
358 mutex_lock(&net_mutex);
359
360 /* Don't let anyone else find us. */
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700361 rtnl_lock();
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000362 list_for_each_entry(net, &net_kill_list, cleanup_list) {
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000363 list_del_rcu(&net->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000364 list_add_tail(&net->exit_list, &net_exit_list);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200365 for_each_net(tmp) {
366 int id = __peernet2id(tmp, net, false);
367
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200368 if (id >= 0) {
369 rtnl_net_notifyid(tmp, net, RTM_DELNSID, id);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200370 idr_remove(&tmp->netns_ids, id);
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200371 }
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200372 }
373 idr_destroy(&net->netns_ids);
374
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000375 }
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700376 rtnl_unlock();
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200377
Johannes Berg11a28d32009-07-10 09:51:33 +0000378 /*
379 * Another CPU might be rcu-iterating the list, wait for it.
380 * This needs to be before calling the exit() notifiers, so
381 * the rcu_barrier() below isn't sufficient alone.
382 */
383 synchronize_rcu();
384
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200385 /* Run all of the network namespace exit methods */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000386 list_for_each_entry_reverse(ops, &pernet_list, list)
387 ops_exit_list(ops, &net_exit_list);
388
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000389 /* Free the net generic variables */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000390 list_for_each_entry_reverse(ops, &pernet_list, list)
391 ops_free_list(ops, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200392
393 mutex_unlock(&net_mutex);
394
395 /* Ensure there are no outstanding rcu callbacks using this
396 * network namespace.
397 */
398 rcu_barrier();
399
400 /* Finally it is safe to free my network namespace structure */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000401 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
402 list_del_init(&net->exit_list);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700403 put_user_ns(net->user_ns);
Al Viroa685e082011-06-08 21:13:01 -0400404 net_drop_ns(net);
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000405 }
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200406}
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000407static DECLARE_WORK(net_cleanup_work, cleanup_net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200408
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200409void __put_net(struct net *net)
410{
411 /* Cleanup the network namespace in process context */
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000412 unsigned long flags;
413
414 spin_lock_irqsave(&cleanup_list_lock, flags);
415 list_add(&net->cleanup_list, &cleanup_list);
416 spin_unlock_irqrestore(&cleanup_list_lock, flags);
417
418 queue_work(netns_wq, &net_cleanup_work);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200419}
420EXPORT_SYMBOL_GPL(__put_net);
421
Stephen Rothwell956c9202011-05-12 13:51:13 +1000422struct net *get_net_ns_by_fd(int fd)
423{
Stephen Rothwell956c9202011-05-12 13:51:13 +1000424 struct file *file;
Al Viro33c42942014-11-01 02:32:53 -0400425 struct ns_common *ns;
Stephen Rothwell956c9202011-05-12 13:51:13 +1000426 struct net *net;
427
Stephen Rothwell956c9202011-05-12 13:51:13 +1000428 file = proc_ns_fget(fd);
Al Viroc316e6a2011-06-05 00:37:35 +0000429 if (IS_ERR(file))
430 return ERR_CAST(file);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000431
Al Virof77c8012014-11-01 03:13:17 -0400432 ns = get_proc_ns(file_inode(file));
Al Viro33c42942014-11-01 02:32:53 -0400433 if (ns->ops == &netns_operations)
434 net = get_net(container_of(ns, struct net, ns));
Al Viroc316e6a2011-06-05 00:37:35 +0000435 else
436 net = ERR_PTR(-EINVAL);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000437
Al Viroc316e6a2011-06-05 00:37:35 +0000438 fput(file);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000439 return net;
440}
441
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700442#else
Stephen Rothwell956c9202011-05-12 13:51:13 +1000443struct net *get_net_ns_by_fd(int fd)
444{
445 return ERR_PTR(-EINVAL);
446}
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700447#endif
Vadim Kochan4b681c82015-01-12 16:34:05 +0200448EXPORT_SYMBOL_GPL(get_net_ns_by_fd);
Eric W. Biederman9dd776b2007-09-26 22:04:26 -0700449
Johannes Berg30ffee82009-07-10 09:51:35 +0000450struct net *get_net_ns_by_pid(pid_t pid)
451{
452 struct task_struct *tsk;
453 struct net *net;
454
455 /* Lookup the network namespace */
456 net = ERR_PTR(-ESRCH);
457 rcu_read_lock();
458 tsk = find_task_by_vpid(pid);
459 if (tsk) {
460 struct nsproxy *nsproxy;
Eric W. Biederman728dba32014-02-03 19:13:49 -0800461 task_lock(tsk);
462 nsproxy = tsk->nsproxy;
Johannes Berg30ffee82009-07-10 09:51:35 +0000463 if (nsproxy)
464 net = get_net(nsproxy->net_ns);
Eric W. Biederman728dba32014-02-03 19:13:49 -0800465 task_unlock(tsk);
Johannes Berg30ffee82009-07-10 09:51:35 +0000466 }
467 rcu_read_unlock();
468 return net;
469}
470EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
471
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700472static __net_init int net_ns_net_init(struct net *net)
473{
Al Viro33c42942014-11-01 02:32:53 -0400474#ifdef CONFIG_NET_NS
475 net->ns.ops = &netns_operations;
476#endif
Al Viro6344c432014-11-01 00:45:45 -0400477 return ns_alloc_inum(&net->ns);
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700478}
479
480static __net_exit void net_ns_net_exit(struct net *net)
481{
Al Viro6344c432014-11-01 00:45:45 -0400482 ns_free_inum(&net->ns);
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700483}
484
485static struct pernet_operations __net_initdata net_ns_ops = {
486 .init = net_ns_net_init,
487 .exit = net_ns_net_exit,
488};
489
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100490static struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = {
491 [NETNSA_NONE] = { .type = NLA_UNSPEC },
492 [NETNSA_NSID] = { .type = NLA_S32 },
493 [NETNSA_PID] = { .type = NLA_U32 },
494 [NETNSA_FD] = { .type = NLA_U32 },
495};
496
497static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh)
498{
499 struct net *net = sock_net(skb->sk);
500 struct nlattr *tb[NETNSA_MAX + 1];
501 struct net *peer;
502 int nsid, err;
503
504 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
505 rtnl_net_policy);
506 if (err < 0)
507 return err;
508 if (!tb[NETNSA_NSID])
509 return -EINVAL;
510 nsid = nla_get_s32(tb[NETNSA_NSID]);
511
512 if (tb[NETNSA_PID])
513 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
514 else if (tb[NETNSA_FD])
515 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
516 else
517 return -EINVAL;
518 if (IS_ERR(peer))
519 return PTR_ERR(peer);
520
521 if (__peernet2id(net, peer, false) >= 0) {
522 err = -EEXIST;
523 goto out;
524 }
525
526 err = alloc_netid(net, peer, nsid);
527 if (err > 0)
528 err = 0;
529out:
530 put_net(peer);
531 return err;
532}
533
534static int rtnl_net_get_size(void)
535{
536 return NLMSG_ALIGN(sizeof(struct rtgenmsg))
537 + nla_total_size(sizeof(s32)) /* NETNSA_NSID */
538 ;
539}
540
541static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags,
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200542 int cmd, struct net *net, struct net *peer,
543 int nsid)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100544{
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
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200558 if (nsid >= 0) {
559 id = nsid;
560 } else {
561 id = __peernet2id(net, peer, false);
562 if (id < 0)
563 id = NETNSA_NSID_NOT_ASSIGNED;
564 }
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100565 if (nla_put_s32(skb, NETNSA_NSID, id))
566 goto nla_put_failure;
567
568 nlmsg_end(skb, nlh);
569 return 0;
570
571nla_put_failure:
572 nlmsg_cancel(skb, nlh);
573 return -EMSGSIZE;
574}
575
576static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh)
577{
578 struct net *net = sock_net(skb->sk);
579 struct nlattr *tb[NETNSA_MAX + 1];
580 struct sk_buff *msg;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100581 struct net *peer;
Nicolas Dichtelb111e4e2015-04-07 11:51:52 +0200582 int err;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100583
584 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
585 rtnl_net_policy);
586 if (err < 0)
587 return err;
588 if (tb[NETNSA_PID])
589 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
590 else if (tb[NETNSA_FD])
591 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
592 else
593 return -EINVAL;
594
595 if (IS_ERR(peer))
596 return PTR_ERR(peer);
597
598 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
599 if (!msg) {
600 err = -ENOMEM;
601 goto out;
602 }
603
604 err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200605 RTM_GETNSID, net, peer, -1);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100606 if (err < 0)
607 goto err_out;
608
609 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid);
610 goto out;
611
612err_out:
613 nlmsg_free(msg);
614out:
615 put_net(peer);
616 return err;
617}
618
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200619static void rtnl_net_notifyid(struct net *net, struct net *peer, int cmd,
620 int id)
621{
622 struct sk_buff *msg;
623 int err = -ENOMEM;
624
625 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
626 if (!msg)
627 goto out;
628
629 err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, peer, id);
630 if (err < 0)
631 goto err_out;
632
633 rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0);
634 return;
635
636err_out:
637 nlmsg_free(msg);
638out:
639 rtnl_set_sk_err(net, RTNLGRP_NSID, err);
640}
641
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200642static int __init net_ns_init(void)
643{
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800644 struct net_generic *ng;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200645
Pavel Emelyanovd57a9212007-11-01 00:46:50 -0700646#ifdef CONFIG_NET_NS
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200647 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net),
648 SMP_CACHE_BYTES,
649 SLAB_PANIC, NULL);
Benjamin Thery3ef13552007-11-19 23:18:16 -0800650
651 /* Create workqueue for cleanup */
652 netns_wq = create_singlethread_workqueue("netns");
653 if (!netns_wq)
654 panic("Could not create netns workq");
Pavel Emelyanovd57a9212007-11-01 00:46:50 -0700655#endif
Benjamin Thery3ef13552007-11-19 23:18:16 -0800656
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800657 ng = net_alloc_generic();
658 if (!ng)
659 panic("Could not allocate generic netns");
660
661 rcu_assign_pointer(init_net.gen, ng);
662
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200663 mutex_lock(&net_mutex);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700664 if (setup_net(&init_net, &init_user_ns))
Stephen Hemmingerca0f3112009-05-21 15:10:31 -0700665 panic("Could not setup the initial network namespace");
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200666
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700667 rtnl_lock();
Johannes Berg11a28d32009-07-10 09:51:33 +0000668 list_add_tail_rcu(&init_net.list, &net_namespace_list);
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700669 rtnl_unlock();
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200670
671 mutex_unlock(&net_mutex);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200672
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700673 register_pernet_subsys(&net_ns_ops);
674
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100675 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL);
676 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, NULL, NULL);
677
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200678 return 0;
679}
680
681pure_initcall(net_ns_init);
682
Denis V. Luneved160e82007-11-13 03:23:21 -0800683#ifdef CONFIG_NET_NS
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000684static int __register_pernet_operations(struct list_head *list,
685 struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200686{
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000687 struct net *net;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200688 int error;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000689 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200690
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200691 list_add_tail(&ops->list, list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000692 if (ops->init || (ops->id && ops->size)) {
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700693 for_each_net(net) {
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000694 error = ops_init(ops, net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200695 if (error)
696 goto out_undo;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000697 list_add_tail(&net->exit_list, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200698 }
699 }
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700700 return 0;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200701
702out_undo:
703 /* If I have an error cleanup all namespaces I initialized */
704 list_del(&ops->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000705 ops_exit_list(ops, &net_exit_list);
706 ops_free_list(ops, &net_exit_list);
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700707 return error;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200708}
709
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000710static void __unregister_pernet_operations(struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200711{
712 struct net *net;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000713 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200714
715 list_del(&ops->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000716 for_each_net(net)
717 list_add_tail(&net->exit_list, &net_exit_list);
718 ops_exit_list(ops, &net_exit_list);
719 ops_free_list(ops, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200720}
721
Denis V. Luneved160e82007-11-13 03:23:21 -0800722#else
723
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000724static int __register_pernet_operations(struct list_head *list,
725 struct pernet_operations *ops)
726{
Julian Anastasovb9229342012-04-16 04:43:15 +0000727 return ops_init(ops, &init_net);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000728}
729
730static void __unregister_pernet_operations(struct pernet_operations *ops)
731{
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000732 LIST_HEAD(net_exit_list);
733 list_add(&init_net.exit_list, &net_exit_list);
734 ops_exit_list(ops, &net_exit_list);
735 ops_free_list(ops, &net_exit_list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000736}
737
738#endif /* CONFIG_NET_NS */
739
740static DEFINE_IDA(net_generic_ids);
741
Denis V. Luneved160e82007-11-13 03:23:21 -0800742static int register_pernet_operations(struct list_head *list,
743 struct pernet_operations *ops)
744{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000745 int error;
746
747 if (ops->id) {
748again:
749 error = ida_get_new_above(&net_generic_ids, 1, ops->id);
750 if (error < 0) {
751 if (error == -EAGAIN) {
752 ida_pre_get(&net_generic_ids, GFP_KERNEL);
753 goto again;
754 }
755 return error;
756 }
Eric Dumazet073862b2012-01-26 00:41:38 +0000757 max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000758 }
759 error = __register_pernet_operations(list, ops);
Eric W. Biederman3a765ed2009-12-03 02:29:06 +0000760 if (error) {
761 rcu_barrier();
762 if (ops->id)
763 ida_remove(&net_generic_ids, *ops->id);
764 }
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000765
766 return error;
Denis V. Luneved160e82007-11-13 03:23:21 -0800767}
768
769static void unregister_pernet_operations(struct pernet_operations *ops)
770{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000771
772 __unregister_pernet_operations(ops);
Eric W. Biederman3a765ed2009-12-03 02:29:06 +0000773 rcu_barrier();
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000774 if (ops->id)
775 ida_remove(&net_generic_ids, *ops->id);
Denis V. Luneved160e82007-11-13 03:23:21 -0800776}
Pavel Emelyanovc93cf612008-04-15 00:35:23 -0700777
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200778/**
779 * register_pernet_subsys - register a network namespace subsystem
780 * @ops: pernet operations structure for the subsystem
781 *
782 * Register a subsystem which has init and exit functions
783 * that are called when network namespaces are created and
784 * destroyed respectively.
785 *
786 * When registered all network namespace init functions are
787 * called for every existing network namespace. Allowing kernel
788 * modules to have a race free view of the set of network namespaces.
789 *
790 * When a new network namespace is created all of the init
791 * methods are called in the order in which they were registered.
792 *
793 * When a network namespace is destroyed all of the exit methods
794 * are called in the reverse of the order with which they were
795 * registered.
796 */
797int register_pernet_subsys(struct pernet_operations *ops)
798{
799 int error;
800 mutex_lock(&net_mutex);
801 error = register_pernet_operations(first_device, ops);
802 mutex_unlock(&net_mutex);
803 return error;
804}
805EXPORT_SYMBOL_GPL(register_pernet_subsys);
806
807/**
808 * unregister_pernet_subsys - unregister a network namespace subsystem
809 * @ops: pernet operations structure to manipulate
810 *
811 * Remove the pernet operations structure from the list to be
Oliver Pinter53379e52008-02-03 17:56:48 +0200812 * used when network namespaces are created or destroyed. In
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200813 * addition run the exit method for all existing network
814 * namespaces.
815 */
Jiri Pirkob3c981d2010-04-25 00:49:56 -0700816void unregister_pernet_subsys(struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200817{
818 mutex_lock(&net_mutex);
Jiri Pirkob3c981d2010-04-25 00:49:56 -0700819 unregister_pernet_operations(ops);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200820 mutex_unlock(&net_mutex);
821}
822EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
823
824/**
825 * register_pernet_device - register a network namespace device
826 * @ops: pernet operations structure for the subsystem
827 *
828 * Register a device which has init and exit functions
829 * that are called when network namespaces are created and
830 * destroyed respectively.
831 *
832 * When registered all network namespace init functions are
833 * called for every existing network namespace. Allowing kernel
834 * modules to have a race free view of the set of network namespaces.
835 *
836 * When a new network namespace is created all of the init
837 * methods are called in the order in which they were registered.
838 *
839 * When a network namespace is destroyed all of the exit methods
840 * are called in the reverse of the order with which they were
841 * registered.
842 */
843int register_pernet_device(struct pernet_operations *ops)
844{
845 int error;
846 mutex_lock(&net_mutex);
847 error = register_pernet_operations(&pernet_list, ops);
848 if (!error && (first_device == &pernet_list))
849 first_device = &ops->list;
850 mutex_unlock(&net_mutex);
851 return error;
852}
853EXPORT_SYMBOL_GPL(register_pernet_device);
854
855/**
856 * unregister_pernet_device - unregister a network namespace netdevice
857 * @ops: pernet operations structure to manipulate
858 *
859 * Remove the pernet operations structure from the list to be
Oliver Pinter53379e52008-02-03 17:56:48 +0200860 * used when network namespaces are created or destroyed. In
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200861 * addition run the exit method for all existing network
862 * namespaces.
863 */
864void unregister_pernet_device(struct pernet_operations *ops)
865{
866 mutex_lock(&net_mutex);
867 if (&ops->list == first_device)
868 first_device = first_device->next;
869 unregister_pernet_operations(ops);
870 mutex_unlock(&net_mutex);
871}
872EXPORT_SYMBOL_GPL(unregister_pernet_device);
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800873
874#ifdef CONFIG_NET_NS
Al Viro64964522014-11-01 00:37:32 -0400875static struct ns_common *netns_get(struct task_struct *task)
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800876{
Eric W. Biedermanf0630522011-05-04 17:51:50 -0700877 struct net *net = NULL;
878 struct nsproxy *nsproxy;
879
Eric W. Biederman728dba32014-02-03 19:13:49 -0800880 task_lock(task);
881 nsproxy = task->nsproxy;
Eric W. Biedermanf0630522011-05-04 17:51:50 -0700882 if (nsproxy)
883 net = get_net(nsproxy->net_ns);
Eric W. Biederman728dba32014-02-03 19:13:49 -0800884 task_unlock(task);
Eric W. Biedermanf0630522011-05-04 17:51:50 -0700885
Al Viroff248702014-11-01 00:10:50 -0400886 return net ? &net->ns : NULL;
887}
888
889static inline struct net *to_net_ns(struct ns_common *ns)
890{
891 return container_of(ns, struct net, ns);
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800892}
893
Al Viro64964522014-11-01 00:37:32 -0400894static void netns_put(struct ns_common *ns)
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800895{
Al Viroff248702014-11-01 00:10:50 -0400896 put_net(to_net_ns(ns));
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800897}
898
Al Viro64964522014-11-01 00:37:32 -0400899static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns)
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800900{
Al Viroff248702014-11-01 00:10:50 -0400901 struct net *net = to_net_ns(ns);
Eric W. Biederman142e1d12012-07-26 01:13:20 -0700902
Eric W. Biederman5e4a0842012-12-14 07:55:36 -0800903 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700904 !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
Eric W. Biederman142e1d12012-07-26 01:13:20 -0700905 return -EPERM;
906
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800907 put_net(nsproxy->net_ns);
Eric W. Biederman142e1d12012-07-26 01:13:20 -0700908 nsproxy->net_ns = get_net(net);
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800909 return 0;
910}
911
912const struct proc_ns_operations netns_operations = {
913 .name = "net",
914 .type = CLONE_NEWNET,
915 .get = netns_get,
916 .put = netns_put,
917 .install = netns_install,
918};
919#endif