blob: 7001da910c6b26cb2300c38147305a459895cc49 [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>
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +010019#include <net/sock.h>
20#include <net/netlink.h>
Eric W. Biederman5f256be2007-09-12 11:50:50 +020021#include <net/net_namespace.h>
Pavel Emelyanovdec827d2008-04-15 00:36:08 -070022#include <net/netns/generic.h>
Eric W. Biederman5f256be2007-09-12 11:50:50 +020023
24/*
25 * Our network namespace constructor/destructor lists
26 */
27
28static LIST_HEAD(pernet_list);
29static struct list_head *first_device = &pernet_list;
Cong Wang200b9162014-05-12 15:11:20 -070030DEFINE_MUTEX(net_mutex);
Eric W. Biederman5f256be2007-09-12 11:50:50 +020031
Eric W. Biederman5f256be2007-09-12 11:50:50 +020032LIST_HEAD(net_namespace_list);
Alexey Dobriyanb76a4612008-10-08 11:35:06 +020033EXPORT_SYMBOL_GPL(net_namespace_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +020034
Rustad, Mark D734b6542012-07-18 09:06:07 +000035struct net init_net = {
36 .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head),
37};
Denis V. Lunevff4b9502008-01-22 22:05:33 -080038EXPORT_SYMBOL(init_net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +020039
Dmitry Torokhovf8c46cb2016-08-10 14:36:00 -070040static bool init_net_initialized;
41
Pavel Emelyanovdec827d2008-04-15 00:36:08 -070042#define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
43
Eric Dumazet073862b2012-01-26 00:41:38 +000044static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS;
45
46static struct net_generic *net_alloc_generic(void)
47{
48 struct net_generic *ng;
49 size_t generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]);
50
51 ng = kzalloc(generic_size, GFP_KERNEL);
52 if (ng)
53 ng->len = max_gen_ptrs;
54
55 return ng;
56}
57
Jiri Pirko05fceb42010-04-23 01:40:47 +000058static int net_assign_generic(struct net *net, int id, void *data)
59{
60 struct net_generic *ng, *old_ng;
61
62 BUG_ON(!mutex_is_locked(&net_mutex));
63 BUG_ON(id == 0);
64
Eric Dumazet1c877332010-10-25 03:20:11 +000065 old_ng = rcu_dereference_protected(net->gen,
66 lockdep_is_held(&net_mutex));
67 ng = old_ng;
Jiri Pirko05fceb42010-04-23 01:40:47 +000068 if (old_ng->len >= id)
69 goto assign;
70
Eric Dumazet073862b2012-01-26 00:41:38 +000071 ng = net_alloc_generic();
Jiri Pirko05fceb42010-04-23 01:40:47 +000072 if (ng == NULL)
73 return -ENOMEM;
74
75 /*
76 * Some synchronisation notes:
77 *
78 * The net_generic explores the net->gen array inside rcu
79 * read section. Besides once set the net->gen->ptr[x]
80 * pointer never changes (see rules in netns/generic.h).
81 *
82 * That said, we simply duplicate this array and schedule
83 * the old copy for kfree after a grace period.
84 */
85
Jiri Pirko05fceb42010-04-23 01:40:47 +000086 memcpy(&ng->ptr, &old_ng->ptr, old_ng->len * sizeof(void*));
87
88 rcu_assign_pointer(net->gen, ng);
Lai Jiangshan04d4dfe2011-03-18 12:06:32 +080089 kfree_rcu(old_ng, rcu);
Jiri Pirko05fceb42010-04-23 01:40:47 +000090assign:
91 ng->ptr[id - 1] = data;
92 return 0;
93}
94
Eric W. Biedermanf875bae2009-11-29 22:25:28 +000095static int ops_init(const struct pernet_operations *ops, struct net *net)
96{
Julian Anastasovb9229342012-04-16 04:43:15 +000097 int err = -ENOMEM;
98 void *data = NULL;
99
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000100 if (ops->id && ops->size) {
Julian Anastasovb9229342012-04-16 04:43:15 +0000101 data = kzalloc(ops->size, GFP_KERNEL);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000102 if (!data)
Julian Anastasovb9229342012-04-16 04:43:15 +0000103 goto out;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000104
105 err = net_assign_generic(net, *ops->id, data);
Julian Anastasovb9229342012-04-16 04:43:15 +0000106 if (err)
107 goto cleanup;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000108 }
Julian Anastasovb9229342012-04-16 04:43:15 +0000109 err = 0;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000110 if (ops->init)
Julian Anastasovb9229342012-04-16 04:43:15 +0000111 err = ops->init(net);
112 if (!err)
113 return 0;
114
115cleanup:
116 kfree(data);
117
118out:
119 return err;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000120}
121
122static void ops_free(const struct pernet_operations *ops, struct net *net)
123{
124 if (ops->id && ops->size) {
125 int id = *ops->id;
126 kfree(net_generic(net, id));
127 }
128}
129
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000130static void ops_exit_list(const struct pernet_operations *ops,
131 struct list_head *net_exit_list)
132{
133 struct net *net;
134 if (ops->exit) {
135 list_for_each_entry(net, net_exit_list, exit_list)
136 ops->exit(net);
137 }
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000138 if (ops->exit_batch)
139 ops->exit_batch(net_exit_list);
140}
141
142static void ops_free_list(const struct pernet_operations *ops,
143 struct list_head *net_exit_list)
144{
145 struct net *net;
146 if (ops->size && ops->id) {
147 list_for_each_entry(net, net_exit_list, exit_list)
148 ops_free(ops, net);
149 }
150}
151
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200152/* should be called with nsid_lock held */
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100153static int alloc_netid(struct net *net, struct net *peer, int reqid)
154{
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200155 int min = 0, max = 0;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100156
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100157 if (reqid >= 0) {
158 min = reqid;
159 max = reqid + 1;
160 }
161
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200162 return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100163}
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
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200178/* Should be called with nsid_lock held. If a new id is assigned, the bool alloc
179 * is set to true, thus the caller knows that the new id must be notified via
180 * rtnl.
181 */
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200182static int __peernet2id_alloc(struct net *net, struct net *peer, bool *alloc)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100183{
184 int id = idr_for_each(&net->netns_ids, net_eq_idr, peer);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200185 bool alloc_it = *alloc;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100186
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200187 *alloc = false;
188
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100189 /* Magic value for id 0. */
190 if (id == NET_ID_ZERO)
191 return 0;
192 if (id > 0)
193 return id;
194
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200195 if (alloc_it) {
Nicolas Dichtel109582af2015-05-07 11:02:47 +0200196 id = alloc_netid(net, peer, -1);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200197 *alloc = true;
Nicolas Dichtel109582af2015-05-07 11:02:47 +0200198 return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED;
199 }
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100200
Nicolas Dichtel109582af2015-05-07 11:02:47 +0200201 return NETNSA_NSID_NOT_ASSIGNED;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100202}
203
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200204/* should be called with nsid_lock held */
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200205static int __peernet2id(struct net *net, struct net *peer)
206{
207 bool no = false;
208
209 return __peernet2id_alloc(net, peer, &no);
210}
211
212static void rtnl_net_notifyid(struct net *net, int cmd, int id);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100213/* This function returns the id of a peer netns. If no id is assigned, one will
214 * be allocated and returned.
215 */
Nicolas Dichtel7a0877d2015-05-07 11:02:49 +0200216int peernet2id_alloc(struct net *net, struct net *peer)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100217{
Paul Moore2a733062016-10-21 21:49:14 -0400218 unsigned long flags;
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200219 bool alloc;
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200220 int id;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100221
WANG Congcfc44a42016-11-16 10:27:02 -0800222 if (atomic_read(&net->count) == 0)
223 return NETNSA_NSID_NOT_ASSIGNED;
Paul Moore2a733062016-10-21 21:49:14 -0400224 spin_lock_irqsave(&net->nsid_lock, flags);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200225 alloc = atomic_read(&peer->count) == 0 ? false : true;
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200226 id = __peernet2id_alloc(net, peer, &alloc);
Paul Moore2a733062016-10-21 21:49:14 -0400227 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200228 if (alloc && id >= 0)
229 rtnl_net_notifyid(net, RTM_NEWNSID, id);
230 return id;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100231}
232
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200233/* This function returns, if assigned, the id of a peer netns. */
Nicolas Dichtel59324cf2015-05-07 11:02:53 +0200234int peernet2id(struct net *net, struct net *peer)
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200235{
Paul Moore2a733062016-10-21 21:49:14 -0400236 unsigned long flags;
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200237 int id;
238
Paul Moore2a733062016-10-21 21:49:14 -0400239 spin_lock_irqsave(&net->nsid_lock, flags);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200240 id = __peernet2id(net, peer);
Paul Moore2a733062016-10-21 21:49:14 -0400241 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200242 return id;
243}
WANG Cong38f507f2016-09-01 21:53:44 -0700244EXPORT_SYMBOL(peernet2id);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200245
Nicolas Dichtel59324cf2015-05-07 11:02:53 +0200246/* This function returns true is the peer netns has an id assigned into the
247 * current netns.
248 */
249bool peernet_has_id(struct net *net, struct net *peer)
250{
251 return peernet2id(net, peer) >= 0;
252}
253
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100254struct net *get_net_ns_by_id(struct net *net, int id)
255{
Paul Moore2a733062016-10-21 21:49:14 -0400256 unsigned long flags;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100257 struct net *peer;
258
259 if (id < 0)
260 return NULL;
261
262 rcu_read_lock();
Paul Moore2a733062016-10-21 21:49:14 -0400263 spin_lock_irqsave(&net->nsid_lock, flags);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100264 peer = idr_find(&net->netns_ids, id);
265 if (peer)
266 get_net(peer);
Paul Moore2a733062016-10-21 21:49:14 -0400267 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100268 rcu_read_unlock();
269
270 return peer;
271}
272
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700273/*
274 * setup_net runs the initializers for the network namespace object.
275 */
Eric W. Biederman038e7332012-06-14 02:31:10 -0700276static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700277{
278 /* Must be called with net_mutex held */
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000279 const struct pernet_operations *ops, *saved_ops;
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800280 int error = 0;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000281 LIST_HEAD(net_exit_list);
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700282
283 atomic_set(&net->count, 1);
Al Viroa685e082011-06-08 21:13:01 -0400284 atomic_set(&net->passive, 1);
Thomas Graf4e985ad2011-06-21 03:11:20 +0000285 net->dev_base_seq = 1;
Eric W. Biederman038e7332012-06-14 02:31:10 -0700286 net->user_ns = user_ns;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100287 idr_init(&net->netns_ids);
WANG Congde133462015-05-15 14:47:32 -0700288 spin_lock_init(&net->nsid_lock);
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800289
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700290 list_for_each_entry(ops, &pernet_list, list) {
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000291 error = ops_init(ops, net);
292 if (error < 0)
293 goto out_undo;
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700294 }
295out:
296 return error;
297
298out_undo:
299 /* Walk through the list backwards calling the exit functions
300 * for the pernet modules whose init functions did not fail.
301 */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000302 list_add(&net->exit_list, &net_exit_list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000303 saved_ops = ops;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000304 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
305 ops_exit_list(ops, &net_exit_list);
306
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000307 ops = saved_ops;
308 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000309 ops_free_list(ops, &net_exit_list);
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700310
311 rcu_barrier();
312 goto out;
313}
314
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800315
Clemens Nossebe47d42009-02-23 15:37:35 -0800316#ifdef CONFIG_NET_NS
Arnd Bergmann2ed6afd2016-09-23 18:06:12 +0200317static struct ucounts *inc_net_namespaces(struct user_namespace *ns)
318{
319 return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES);
320}
321
322static void dec_net_namespaces(struct ucounts *ucounts)
323{
324 dec_ucount(ucounts, UCOUNT_NET_NAMESPACES);
325}
326
Clemens Nossebe47d42009-02-23 15:37:35 -0800327static struct kmem_cache *net_cachep;
328static struct workqueue_struct *netns_wq;
329
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200330static struct net *net_alloc(void)
331{
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800332 struct net *net = NULL;
333 struct net_generic *ng;
334
335 ng = net_alloc_generic();
336 if (!ng)
337 goto out;
338
339 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL);
340 if (!net)
341 goto out_free;
342
343 rcu_assign_pointer(net->gen, ng);
344out:
345 return net;
346
347out_free:
348 kfree(ng);
349 goto out;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200350}
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200351
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800352static void net_free(struct net *net)
353{
Eric Dumazet416c51e2014-09-09 08:24:53 -0700354 kfree(rcu_access_pointer(net->gen));
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800355 kmem_cache_free(net_cachep, net);
356}
357
Al Viroa685e082011-06-08 21:13:01 -0400358void net_drop_ns(void *p)
359{
360 struct net *ns = p;
361 if (ns && atomic_dec_and_test(&ns->passive))
362 net_free(ns);
363}
364
Eric W. Biederman038e7332012-06-14 02:31:10 -0700365struct net *copy_net_ns(unsigned long flags,
366 struct user_namespace *user_ns, struct net *old_net)
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700367{
Eric W. Biederman70328662016-08-08 14:33:23 -0500368 struct ucounts *ucounts;
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700369 struct net *net;
370 int rv;
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700371
Rob Landley911cb192011-04-15 02:26:25 +0000372 if (!(flags & CLONE_NEWNET))
373 return get_net(old_net);
374
Eric W. Biederman70328662016-08-08 14:33:23 -0500375 ucounts = inc_net_namespaces(user_ns);
376 if (!ucounts)
Eric W. Biedermandf75e772016-09-22 13:08:36 -0500377 return ERR_PTR(-ENOSPC);
Eric W. Biederman70328662016-08-08 14:33:23 -0500378
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700379 net = net_alloc();
Eric W. Biederman70328662016-08-08 14:33:23 -0500380 if (!net) {
381 dec_net_namespaces(ucounts);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700382 return ERR_PTR(-ENOMEM);
Eric W. Biederman70328662016-08-08 14:33:23 -0500383 }
Eric W. Biederman038e7332012-06-14 02:31:10 -0700384
385 get_user_ns(user_ns);
386
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700387 mutex_lock(&net_mutex);
Eric W. Biederman70328662016-08-08 14:33:23 -0500388 net->ucounts = ucounts;
Eric W. Biederman038e7332012-06-14 02:31:10 -0700389 rv = setup_net(net, user_ns);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700390 if (rv == 0) {
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800391 rtnl_lock();
Johannes Berg11a28d32009-07-10 09:51:33 +0000392 list_add_tail_rcu(&net->list, &net_namespace_list);
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800393 rtnl_unlock();
394 }
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700395 mutex_unlock(&net_mutex);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700396 if (rv < 0) {
Eric W. Biederman70328662016-08-08 14:33:23 -0500397 dec_net_namespaces(ucounts);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700398 put_user_ns(user_ns);
Al Viroa685e082011-06-08 21:13:01 -0400399 net_drop_ns(net);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700400 return ERR_PTR(rv);
401 }
402 return net;
403}
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800404
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000405static DEFINE_SPINLOCK(cleanup_list_lock);
406static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */
407
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200408static void cleanup_net(struct work_struct *work)
409{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000410 const struct pernet_operations *ops;
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200411 struct net *net, *tmp;
xiao jin1818ce42014-04-25 08:50:54 +0800412 struct list_head net_kill_list;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000413 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200414
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000415 /* Atomically snapshot the list of namespaces to cleanup */
416 spin_lock_irq(&cleanup_list_lock);
417 list_replace_init(&cleanup_list, &net_kill_list);
418 spin_unlock_irq(&cleanup_list_lock);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200419
420 mutex_lock(&net_mutex);
421
422 /* Don't let anyone else find us. */
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700423 rtnl_lock();
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000424 list_for_each_entry(net, &net_kill_list, cleanup_list) {
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000425 list_del_rcu(&net->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000426 list_add_tail(&net->exit_list, &net_exit_list);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200427 for_each_net(tmp) {
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200428 int id;
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200429
Paul Moore2a733062016-10-21 21:49:14 -0400430 spin_lock_irq(&tmp->nsid_lock);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200431 id = __peernet2id(tmp, net);
432 if (id >= 0)
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200433 idr_remove(&tmp->netns_ids, id);
Paul Moore2a733062016-10-21 21:49:14 -0400434 spin_unlock_irq(&tmp->nsid_lock);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200435 if (id >= 0)
436 rtnl_net_notifyid(tmp, RTM_DELNSID, id);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200437 }
Paul Moore2a733062016-10-21 21:49:14 -0400438 spin_lock_irq(&net->nsid_lock);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200439 idr_destroy(&net->netns_ids);
Paul Moore2a733062016-10-21 21:49:14 -0400440 spin_unlock_irq(&net->nsid_lock);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200441
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000442 }
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700443 rtnl_unlock();
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200444
Johannes Berg11a28d32009-07-10 09:51:33 +0000445 /*
446 * Another CPU might be rcu-iterating the list, wait for it.
447 * This needs to be before calling the exit() notifiers, so
448 * the rcu_barrier() below isn't sufficient alone.
449 */
450 synchronize_rcu();
451
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200452 /* Run all of the network namespace exit methods */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000453 list_for_each_entry_reverse(ops, &pernet_list, list)
454 ops_exit_list(ops, &net_exit_list);
455
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000456 /* Free the net generic variables */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000457 list_for_each_entry_reverse(ops, &pernet_list, list)
458 ops_free_list(ops, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200459
460 mutex_unlock(&net_mutex);
461
462 /* Ensure there are no outstanding rcu callbacks using this
463 * network namespace.
464 */
465 rcu_barrier();
466
467 /* Finally it is safe to free my network namespace structure */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000468 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
469 list_del_init(&net->exit_list);
Eric W. Biederman70328662016-08-08 14:33:23 -0500470 dec_net_namespaces(net->ucounts);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700471 put_user_ns(net->user_ns);
Al Viroa685e082011-06-08 21:13:01 -0400472 net_drop_ns(net);
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000473 }
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200474}
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000475static DECLARE_WORK(net_cleanup_work, cleanup_net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200476
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200477void __put_net(struct net *net)
478{
479 /* Cleanup the network namespace in process context */
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000480 unsigned long flags;
481
482 spin_lock_irqsave(&cleanup_list_lock, flags);
483 list_add(&net->cleanup_list, &cleanup_list);
484 spin_unlock_irqrestore(&cleanup_list_lock, flags);
485
486 queue_work(netns_wq, &net_cleanup_work);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200487}
488EXPORT_SYMBOL_GPL(__put_net);
489
Stephen Rothwell956c9202011-05-12 13:51:13 +1000490struct net *get_net_ns_by_fd(int fd)
491{
Stephen Rothwell956c9202011-05-12 13:51:13 +1000492 struct file *file;
Al Viro33c42942014-11-01 02:32:53 -0400493 struct ns_common *ns;
Stephen Rothwell956c9202011-05-12 13:51:13 +1000494 struct net *net;
495
Stephen Rothwell956c9202011-05-12 13:51:13 +1000496 file = proc_ns_fget(fd);
Al Viroc316e6a2011-06-05 00:37:35 +0000497 if (IS_ERR(file))
498 return ERR_CAST(file);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000499
Al Virof77c8012014-11-01 03:13:17 -0400500 ns = get_proc_ns(file_inode(file));
Al Viro33c42942014-11-01 02:32:53 -0400501 if (ns->ops == &netns_operations)
502 net = get_net(container_of(ns, struct net, ns));
Al Viroc316e6a2011-06-05 00:37:35 +0000503 else
504 net = ERR_PTR(-EINVAL);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000505
Al Viroc316e6a2011-06-05 00:37:35 +0000506 fput(file);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000507 return net;
508}
509
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700510#else
Stephen Rothwell956c9202011-05-12 13:51:13 +1000511struct net *get_net_ns_by_fd(int fd)
512{
513 return ERR_PTR(-EINVAL);
514}
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700515#endif
Vadim Kochan4b681c82015-01-12 16:34:05 +0200516EXPORT_SYMBOL_GPL(get_net_ns_by_fd);
Eric W. Biederman9dd776b2007-09-26 22:04:26 -0700517
Johannes Berg30ffee82009-07-10 09:51:35 +0000518struct net *get_net_ns_by_pid(pid_t pid)
519{
520 struct task_struct *tsk;
521 struct net *net;
522
523 /* Lookup the network namespace */
524 net = ERR_PTR(-ESRCH);
525 rcu_read_lock();
526 tsk = find_task_by_vpid(pid);
527 if (tsk) {
528 struct nsproxy *nsproxy;
Eric W. Biederman728dba32014-02-03 19:13:49 -0800529 task_lock(tsk);
530 nsproxy = tsk->nsproxy;
Johannes Berg30ffee82009-07-10 09:51:35 +0000531 if (nsproxy)
532 net = get_net(nsproxy->net_ns);
Eric W. Biederman728dba32014-02-03 19:13:49 -0800533 task_unlock(tsk);
Johannes Berg30ffee82009-07-10 09:51:35 +0000534 }
535 rcu_read_unlock();
536 return net;
537}
538EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
539
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700540static __net_init int net_ns_net_init(struct net *net)
541{
Al Viro33c42942014-11-01 02:32:53 -0400542#ifdef CONFIG_NET_NS
543 net->ns.ops = &netns_operations;
544#endif
Al Viro6344c432014-11-01 00:45:45 -0400545 return ns_alloc_inum(&net->ns);
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700546}
547
548static __net_exit void net_ns_net_exit(struct net *net)
549{
Al Viro6344c432014-11-01 00:45:45 -0400550 ns_free_inum(&net->ns);
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700551}
552
553static struct pernet_operations __net_initdata net_ns_ops = {
554 .init = net_ns_net_init,
555 .exit = net_ns_net_exit,
556};
557
stephen hemminger3ee52562016-08-31 15:17:49 -0700558static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = {
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100559 [NETNSA_NONE] = { .type = NLA_UNSPEC },
560 [NETNSA_NSID] = { .type = NLA_S32 },
561 [NETNSA_PID] = { .type = NLA_U32 },
562 [NETNSA_FD] = { .type = NLA_U32 },
563};
564
565static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh)
566{
567 struct net *net = sock_net(skb->sk);
568 struct nlattr *tb[NETNSA_MAX + 1];
Paul Moore2a733062016-10-21 21:49:14 -0400569 unsigned long flags;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100570 struct net *peer;
571 int nsid, err;
572
573 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
574 rtnl_net_policy);
575 if (err < 0)
576 return err;
577 if (!tb[NETNSA_NSID])
578 return -EINVAL;
579 nsid = nla_get_s32(tb[NETNSA_NSID]);
580
581 if (tb[NETNSA_PID])
582 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
583 else if (tb[NETNSA_FD])
584 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
585 else
586 return -EINVAL;
587 if (IS_ERR(peer))
588 return PTR_ERR(peer);
589
Paul Moore2a733062016-10-21 21:49:14 -0400590 spin_lock_irqsave(&net->nsid_lock, flags);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200591 if (__peernet2id(net, peer) >= 0) {
Paul Moore2a733062016-10-21 21:49:14 -0400592 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100593 err = -EEXIST;
594 goto out;
595 }
596
597 err = alloc_netid(net, peer, nsid);
Paul Moore2a733062016-10-21 21:49:14 -0400598 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200599 if (err >= 0) {
600 rtnl_net_notifyid(net, RTM_NEWNSID, err);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100601 err = 0;
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200602 }
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100603out:
604 put_net(peer);
605 return err;
606}
607
608static int rtnl_net_get_size(void)
609{
610 return NLMSG_ALIGN(sizeof(struct rtgenmsg))
611 + nla_total_size(sizeof(s32)) /* NETNSA_NSID */
612 ;
613}
614
615static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags,
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200616 int cmd, struct net *net, int nsid)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100617{
618 struct nlmsghdr *nlh;
619 struct rtgenmsg *rth;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100620
621 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags);
622 if (!nlh)
623 return -EMSGSIZE;
624
625 rth = nlmsg_data(nlh);
626 rth->rtgen_family = AF_UNSPEC;
627
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200628 if (nla_put_s32(skb, NETNSA_NSID, nsid))
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100629 goto nla_put_failure;
630
631 nlmsg_end(skb, nlh);
632 return 0;
633
634nla_put_failure:
635 nlmsg_cancel(skb, nlh);
636 return -EMSGSIZE;
637}
638
639static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh)
640{
641 struct net *net = sock_net(skb->sk);
642 struct nlattr *tb[NETNSA_MAX + 1];
643 struct sk_buff *msg;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100644 struct net *peer;
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200645 int err, id;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100646
647 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
648 rtnl_net_policy);
649 if (err < 0)
650 return err;
651 if (tb[NETNSA_PID])
652 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
653 else if (tb[NETNSA_FD])
654 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
655 else
656 return -EINVAL;
657
658 if (IS_ERR(peer))
659 return PTR_ERR(peer);
660
661 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
662 if (!msg) {
663 err = -ENOMEM;
664 goto out;
665 }
666
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200667 id = peernet2id(net, peer);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100668 err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
David S. Millerb04096f2015-05-13 14:31:43 -0400669 RTM_NEWNSID, net, id);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100670 if (err < 0)
671 goto err_out;
672
673 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid);
674 goto out;
675
676err_out:
677 nlmsg_free(msg);
678out:
679 put_net(peer);
680 return err;
681}
682
Nicolas Dichtela143c402015-04-07 11:51:54 +0200683struct rtnl_net_dump_cb {
684 struct net *net;
685 struct sk_buff *skb;
686 struct netlink_callback *cb;
687 int idx;
688 int s_idx;
689};
690
691static int rtnl_net_dumpid_one(int id, void *peer, void *data)
692{
693 struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data;
694 int ret;
695
696 if (net_cb->idx < net_cb->s_idx)
697 goto cont;
698
699 ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid,
700 net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI,
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200701 RTM_NEWNSID, net_cb->net, id);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200702 if (ret < 0)
703 return ret;
704
705cont:
706 net_cb->idx++;
707 return 0;
708}
709
710static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb)
711{
712 struct net *net = sock_net(skb->sk);
713 struct rtnl_net_dump_cb net_cb = {
714 .net = net,
715 .skb = skb,
716 .cb = cb,
717 .idx = 0,
718 .s_idx = cb->args[0],
719 };
Paul Moore2a733062016-10-21 21:49:14 -0400720 unsigned long flags;
Nicolas Dichtela143c402015-04-07 11:51:54 +0200721
Paul Moore2a733062016-10-21 21:49:14 -0400722 spin_lock_irqsave(&net->nsid_lock, flags);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200723 idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb);
Paul Moore2a733062016-10-21 21:49:14 -0400724 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200725
726 cb->args[0] = net_cb.idx;
727 return skb->len;
728}
729
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200730static void rtnl_net_notifyid(struct net *net, int cmd, int id)
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200731{
732 struct sk_buff *msg;
733 int err = -ENOMEM;
734
735 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
736 if (!msg)
737 goto out;
738
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200739 err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, id);
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200740 if (err < 0)
741 goto err_out;
742
743 rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0);
744 return;
745
746err_out:
747 nlmsg_free(msg);
748out:
749 rtnl_set_sk_err(net, RTNLGRP_NSID, err);
750}
751
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200752static int __init net_ns_init(void)
753{
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800754 struct net_generic *ng;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200755
Pavel Emelyanovd57a9212007-11-01 00:46:50 -0700756#ifdef CONFIG_NET_NS
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200757 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net),
758 SMP_CACHE_BYTES,
759 SLAB_PANIC, NULL);
Benjamin Thery3ef13552007-11-19 23:18:16 -0800760
761 /* Create workqueue for cleanup */
762 netns_wq = create_singlethread_workqueue("netns");
763 if (!netns_wq)
764 panic("Could not create netns workq");
Pavel Emelyanovd57a9212007-11-01 00:46:50 -0700765#endif
Benjamin Thery3ef13552007-11-19 23:18:16 -0800766
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800767 ng = net_alloc_generic();
768 if (!ng)
769 panic("Could not allocate generic netns");
770
771 rcu_assign_pointer(init_net.gen, ng);
772
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200773 mutex_lock(&net_mutex);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700774 if (setup_net(&init_net, &init_user_ns))
Stephen Hemmingerca0f3112009-05-21 15:10:31 -0700775 panic("Could not setup the initial network namespace");
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200776
Dmitry Torokhovf8c46cb2016-08-10 14:36:00 -0700777 init_net_initialized = true;
778
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700779 rtnl_lock();
Johannes Berg11a28d32009-07-10 09:51:33 +0000780 list_add_tail_rcu(&init_net.list, &net_namespace_list);
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700781 rtnl_unlock();
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200782
783 mutex_unlock(&net_mutex);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200784
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700785 register_pernet_subsys(&net_ns_ops);
786
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100787 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200788 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid,
789 NULL);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100790
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200791 return 0;
792}
793
794pure_initcall(net_ns_init);
795
Denis V. Luneved160e82007-11-13 03:23:21 -0800796#ifdef CONFIG_NET_NS
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000797static int __register_pernet_operations(struct list_head *list,
798 struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200799{
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000800 struct net *net;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200801 int error;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000802 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200803
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200804 list_add_tail(&ops->list, list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000805 if (ops->init || (ops->id && ops->size)) {
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700806 for_each_net(net) {
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000807 error = ops_init(ops, net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200808 if (error)
809 goto out_undo;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000810 list_add_tail(&net->exit_list, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200811 }
812 }
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700813 return 0;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200814
815out_undo:
816 /* If I have an error cleanup all namespaces I initialized */
817 list_del(&ops->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000818 ops_exit_list(ops, &net_exit_list);
819 ops_free_list(ops, &net_exit_list);
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700820 return error;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200821}
822
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000823static void __unregister_pernet_operations(struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200824{
825 struct net *net;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000826 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200827
828 list_del(&ops->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000829 for_each_net(net)
830 list_add_tail(&net->exit_list, &net_exit_list);
831 ops_exit_list(ops, &net_exit_list);
832 ops_free_list(ops, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200833}
834
Denis V. Luneved160e82007-11-13 03:23:21 -0800835#else
836
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000837static int __register_pernet_operations(struct list_head *list,
838 struct pernet_operations *ops)
839{
Dmitry Torokhovf8c46cb2016-08-10 14:36:00 -0700840 if (!init_net_initialized) {
841 list_add_tail(&ops->list, list);
842 return 0;
843 }
844
Julian Anastasovb9229342012-04-16 04:43:15 +0000845 return ops_init(ops, &init_net);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000846}
847
848static void __unregister_pernet_operations(struct pernet_operations *ops)
849{
Dmitry Torokhovf8c46cb2016-08-10 14:36:00 -0700850 if (!init_net_initialized) {
851 list_del(&ops->list);
852 } else {
853 LIST_HEAD(net_exit_list);
854 list_add(&init_net.exit_list, &net_exit_list);
855 ops_exit_list(ops, &net_exit_list);
856 ops_free_list(ops, &net_exit_list);
857 }
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000858}
859
860#endif /* CONFIG_NET_NS */
861
862static DEFINE_IDA(net_generic_ids);
863
Denis V. Luneved160e82007-11-13 03:23:21 -0800864static int register_pernet_operations(struct list_head *list,
865 struct pernet_operations *ops)
866{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000867 int error;
868
869 if (ops->id) {
870again:
871 error = ida_get_new_above(&net_generic_ids, 1, ops->id);
872 if (error < 0) {
873 if (error == -EAGAIN) {
874 ida_pre_get(&net_generic_ids, GFP_KERNEL);
875 goto again;
876 }
877 return error;
878 }
Eric Dumazet073862b2012-01-26 00:41:38 +0000879 max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000880 }
881 error = __register_pernet_operations(list, ops);
Eric W. Biederman3a765ed2009-12-03 02:29:06 +0000882 if (error) {
883 rcu_barrier();
884 if (ops->id)
885 ida_remove(&net_generic_ids, *ops->id);
886 }
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000887
888 return error;
Denis V. Luneved160e82007-11-13 03:23:21 -0800889}
890
891static void unregister_pernet_operations(struct pernet_operations *ops)
892{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000893
894 __unregister_pernet_operations(ops);
Eric W. Biederman3a765ed2009-12-03 02:29:06 +0000895 rcu_barrier();
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000896 if (ops->id)
897 ida_remove(&net_generic_ids, *ops->id);
Denis V. Luneved160e82007-11-13 03:23:21 -0800898}
Pavel Emelyanovc93cf612008-04-15 00:35:23 -0700899
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200900/**
901 * register_pernet_subsys - register a network namespace subsystem
902 * @ops: pernet operations structure for the subsystem
903 *
904 * Register a subsystem which has init and exit functions
905 * that are called when network namespaces are created and
906 * destroyed respectively.
907 *
908 * When registered all network namespace init functions are
909 * called for every existing network namespace. Allowing kernel
910 * modules to have a race free view of the set of network namespaces.
911 *
912 * When a new network namespace is created all of the init
913 * methods are called in the order in which they were registered.
914 *
915 * When a network namespace is destroyed all of the exit methods
916 * are called in the reverse of the order with which they were
917 * registered.
918 */
919int register_pernet_subsys(struct pernet_operations *ops)
920{
921 int error;
922 mutex_lock(&net_mutex);
923 error = register_pernet_operations(first_device, ops);
924 mutex_unlock(&net_mutex);
925 return error;
926}
927EXPORT_SYMBOL_GPL(register_pernet_subsys);
928
929/**
930 * unregister_pernet_subsys - unregister a network namespace subsystem
931 * @ops: pernet operations structure to manipulate
932 *
933 * Remove the pernet operations structure from the list to be
Oliver Pinter53379e52008-02-03 17:56:48 +0200934 * used when network namespaces are created or destroyed. In
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200935 * addition run the exit method for all existing network
936 * namespaces.
937 */
Jiri Pirkob3c981d2010-04-25 00:49:56 -0700938void unregister_pernet_subsys(struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200939{
940 mutex_lock(&net_mutex);
Jiri Pirkob3c981d2010-04-25 00:49:56 -0700941 unregister_pernet_operations(ops);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200942 mutex_unlock(&net_mutex);
943}
944EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
945
946/**
947 * register_pernet_device - register a network namespace device
948 * @ops: pernet operations structure for the subsystem
949 *
950 * Register a device which has init and exit functions
951 * that are called when network namespaces are created and
952 * destroyed respectively.
953 *
954 * When registered all network namespace init functions are
955 * called for every existing network namespace. Allowing kernel
956 * modules to have a race free view of the set of network namespaces.
957 *
958 * When a new network namespace is created all of the init
959 * methods are called in the order in which they were registered.
960 *
961 * When a network namespace is destroyed all of the exit methods
962 * are called in the reverse of the order with which they were
963 * registered.
964 */
965int register_pernet_device(struct pernet_operations *ops)
966{
967 int error;
968 mutex_lock(&net_mutex);
969 error = register_pernet_operations(&pernet_list, ops);
970 if (!error && (first_device == &pernet_list))
971 first_device = &ops->list;
972 mutex_unlock(&net_mutex);
973 return error;
974}
975EXPORT_SYMBOL_GPL(register_pernet_device);
976
977/**
978 * unregister_pernet_device - unregister a network namespace netdevice
979 * @ops: pernet operations structure to manipulate
980 *
981 * Remove the pernet operations structure from the list to be
Oliver Pinter53379e52008-02-03 17:56:48 +0200982 * used when network namespaces are created or destroyed. In
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200983 * addition run the exit method for all existing network
984 * namespaces.
985 */
986void unregister_pernet_device(struct pernet_operations *ops)
987{
988 mutex_lock(&net_mutex);
989 if (&ops->list == first_device)
990 first_device = first_device->next;
991 unregister_pernet_operations(ops);
992 mutex_unlock(&net_mutex);
993}
994EXPORT_SYMBOL_GPL(unregister_pernet_device);
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800995
996#ifdef CONFIG_NET_NS
Al Viro64964522014-11-01 00:37:32 -0400997static struct ns_common *netns_get(struct task_struct *task)
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800998{
Eric W. Biedermanf0630522011-05-04 17:51:50 -0700999 struct net *net = NULL;
1000 struct nsproxy *nsproxy;
1001
Eric W. Biederman728dba32014-02-03 19:13:49 -08001002 task_lock(task);
1003 nsproxy = task->nsproxy;
Eric W. Biedermanf0630522011-05-04 17:51:50 -07001004 if (nsproxy)
1005 net = get_net(nsproxy->net_ns);
Eric W. Biederman728dba32014-02-03 19:13:49 -08001006 task_unlock(task);
Eric W. Biedermanf0630522011-05-04 17:51:50 -07001007
Al Viroff248702014-11-01 00:10:50 -04001008 return net ? &net->ns : NULL;
1009}
1010
1011static inline struct net *to_net_ns(struct ns_common *ns)
1012{
1013 return container_of(ns, struct net, ns);
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001014}
1015
Al Viro64964522014-11-01 00:37:32 -04001016static void netns_put(struct ns_common *ns)
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001017{
Al Viroff248702014-11-01 00:10:50 -04001018 put_net(to_net_ns(ns));
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001019}
1020
Al Viro64964522014-11-01 00:37:32 -04001021static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns)
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001022{
Al Viroff248702014-11-01 00:10:50 -04001023 struct net *net = to_net_ns(ns);
Eric W. Biederman142e1d12012-07-26 01:13:20 -07001024
Eric W. Biederman5e4a0842012-12-14 07:55:36 -08001025 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -07001026 !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
Eric W. Biederman142e1d12012-07-26 01:13:20 -07001027 return -EPERM;
1028
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001029 put_net(nsproxy->net_ns);
Eric W. Biederman142e1d12012-07-26 01:13:20 -07001030 nsproxy->net_ns = get_net(net);
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001031 return 0;
1032}
1033
Andrey Vaginbcac25a2016-09-06 00:47:13 -07001034static struct user_namespace *netns_owner(struct ns_common *ns)
1035{
1036 return to_net_ns(ns)->user_ns;
1037}
1038
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001039const struct proc_ns_operations netns_operations = {
1040 .name = "net",
1041 .type = CLONE_NEWNET,
1042 .get = netns_get,
1043 .put = netns_put,
1044 .install = netns_install,
Andrey Vaginbcac25a2016-09-06 00:47:13 -07001045 .owner = netns_owner,
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001046};
1047#endif