blob: f61c0e02a4130318060e35180d3cab391daf64af [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
Paul Moore2a733062016-10-21 21:49:14 -0400222 spin_lock_irqsave(&net->nsid_lock, flags);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200223 alloc = atomic_read(&peer->count) == 0 ? false : true;
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200224 id = __peernet2id_alloc(net, peer, &alloc);
Paul Moore2a733062016-10-21 21:49:14 -0400225 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200226 if (alloc && id >= 0)
227 rtnl_net_notifyid(net, RTM_NEWNSID, id);
228 return id;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100229}
230
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200231/* This function returns, if assigned, the id of a peer netns. */
Nicolas Dichtel59324cf2015-05-07 11:02:53 +0200232int peernet2id(struct net *net, struct net *peer)
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200233{
Paul Moore2a733062016-10-21 21:49:14 -0400234 unsigned long flags;
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200235 int id;
236
Paul Moore2a733062016-10-21 21:49:14 -0400237 spin_lock_irqsave(&net->nsid_lock, flags);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200238 id = __peernet2id(net, peer);
Paul Moore2a733062016-10-21 21:49:14 -0400239 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200240 return id;
241}
WANG Cong38f507f2016-09-01 21:53:44 -0700242EXPORT_SYMBOL(peernet2id);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200243
Nicolas Dichtel59324cf2015-05-07 11:02:53 +0200244/* This function returns true is the peer netns has an id assigned into the
245 * current netns.
246 */
247bool peernet_has_id(struct net *net, struct net *peer)
248{
249 return peernet2id(net, peer) >= 0;
250}
251
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100252struct net *get_net_ns_by_id(struct net *net, int id)
253{
Paul Moore2a733062016-10-21 21:49:14 -0400254 unsigned long flags;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100255 struct net *peer;
256
257 if (id < 0)
258 return NULL;
259
260 rcu_read_lock();
Paul Moore2a733062016-10-21 21:49:14 -0400261 spin_lock_irqsave(&net->nsid_lock, flags);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100262 peer = idr_find(&net->netns_ids, id);
263 if (peer)
264 get_net(peer);
Paul Moore2a733062016-10-21 21:49:14 -0400265 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100266 rcu_read_unlock();
267
268 return peer;
269}
270
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700271/*
272 * setup_net runs the initializers for the network namespace object.
273 */
Eric W. Biederman038e7332012-06-14 02:31:10 -0700274static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700275{
276 /* Must be called with net_mutex held */
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000277 const struct pernet_operations *ops, *saved_ops;
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800278 int error = 0;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000279 LIST_HEAD(net_exit_list);
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700280
281 atomic_set(&net->count, 1);
Al Viroa685e082011-06-08 21:13:01 -0400282 atomic_set(&net->passive, 1);
Thomas Graf4e985ad2011-06-21 03:11:20 +0000283 net->dev_base_seq = 1;
Eric W. Biederman038e7332012-06-14 02:31:10 -0700284 net->user_ns = user_ns;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100285 idr_init(&net->netns_ids);
WANG Congde133462015-05-15 14:47:32 -0700286 spin_lock_init(&net->nsid_lock);
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800287
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700288 list_for_each_entry(ops, &pernet_list, list) {
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000289 error = ops_init(ops, net);
290 if (error < 0)
291 goto out_undo;
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700292 }
293out:
294 return error;
295
296out_undo:
297 /* Walk through the list backwards calling the exit functions
298 * for the pernet modules whose init functions did not fail.
299 */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000300 list_add(&net->exit_list, &net_exit_list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000301 saved_ops = ops;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000302 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
303 ops_exit_list(ops, &net_exit_list);
304
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000305 ops = saved_ops;
306 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000307 ops_free_list(ops, &net_exit_list);
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700308
309 rcu_barrier();
310 goto out;
311}
312
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800313
Clemens Nossebe47d42009-02-23 15:37:35 -0800314#ifdef CONFIG_NET_NS
Arnd Bergmann2ed6afd2016-09-23 18:06:12 +0200315static struct ucounts *inc_net_namespaces(struct user_namespace *ns)
316{
317 return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES);
318}
319
320static void dec_net_namespaces(struct ucounts *ucounts)
321{
322 dec_ucount(ucounts, UCOUNT_NET_NAMESPACES);
323}
324
Clemens Nossebe47d42009-02-23 15:37:35 -0800325static struct kmem_cache *net_cachep;
326static struct workqueue_struct *netns_wq;
327
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200328static struct net *net_alloc(void)
329{
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800330 struct net *net = NULL;
331 struct net_generic *ng;
332
333 ng = net_alloc_generic();
334 if (!ng)
335 goto out;
336
337 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL);
338 if (!net)
339 goto out_free;
340
341 rcu_assign_pointer(net->gen, ng);
342out:
343 return net;
344
345out_free:
346 kfree(ng);
347 goto out;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200348}
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200349
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800350static void net_free(struct net *net)
351{
Eric Dumazet416c51e2014-09-09 08:24:53 -0700352 kfree(rcu_access_pointer(net->gen));
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800353 kmem_cache_free(net_cachep, net);
354}
355
Al Viroa685e082011-06-08 21:13:01 -0400356void net_drop_ns(void *p)
357{
358 struct net *ns = p;
359 if (ns && atomic_dec_and_test(&ns->passive))
360 net_free(ns);
361}
362
Eric W. Biederman038e7332012-06-14 02:31:10 -0700363struct net *copy_net_ns(unsigned long flags,
364 struct user_namespace *user_ns, struct net *old_net)
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700365{
Eric W. Biederman70328662016-08-08 14:33:23 -0500366 struct ucounts *ucounts;
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700367 struct net *net;
368 int rv;
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700369
Rob Landley911cb192011-04-15 02:26:25 +0000370 if (!(flags & CLONE_NEWNET))
371 return get_net(old_net);
372
Eric W. Biederman70328662016-08-08 14:33:23 -0500373 ucounts = inc_net_namespaces(user_ns);
374 if (!ucounts)
Eric W. Biedermandf75e772016-09-22 13:08:36 -0500375 return ERR_PTR(-ENOSPC);
Eric W. Biederman70328662016-08-08 14:33:23 -0500376
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700377 net = net_alloc();
Eric W. Biederman70328662016-08-08 14:33:23 -0500378 if (!net) {
379 dec_net_namespaces(ucounts);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700380 return ERR_PTR(-ENOMEM);
Eric W. Biederman70328662016-08-08 14:33:23 -0500381 }
Eric W. Biederman038e7332012-06-14 02:31:10 -0700382
383 get_user_ns(user_ns);
384
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700385 mutex_lock(&net_mutex);
Eric W. Biederman70328662016-08-08 14:33:23 -0500386 net->ucounts = ucounts;
Eric W. Biederman038e7332012-06-14 02:31:10 -0700387 rv = setup_net(net, user_ns);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700388 if (rv == 0) {
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800389 rtnl_lock();
Johannes Berg11a28d32009-07-10 09:51:33 +0000390 list_add_tail_rcu(&net->list, &net_namespace_list);
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800391 rtnl_unlock();
392 }
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700393 mutex_unlock(&net_mutex);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700394 if (rv < 0) {
Eric W. Biederman70328662016-08-08 14:33:23 -0500395 dec_net_namespaces(ucounts);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700396 put_user_ns(user_ns);
Al Viroa685e082011-06-08 21:13:01 -0400397 net_drop_ns(net);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700398 return ERR_PTR(rv);
399 }
400 return net;
401}
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800402
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000403static DEFINE_SPINLOCK(cleanup_list_lock);
404static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */
405
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200406static void cleanup_net(struct work_struct *work)
407{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000408 const struct pernet_operations *ops;
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200409 struct net *net, *tmp;
xiao jin1818ce42014-04-25 08:50:54 +0800410 struct list_head net_kill_list;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000411 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200412
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000413 /* Atomically snapshot the list of namespaces to cleanup */
414 spin_lock_irq(&cleanup_list_lock);
415 list_replace_init(&cleanup_list, &net_kill_list);
416 spin_unlock_irq(&cleanup_list_lock);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200417
418 mutex_lock(&net_mutex);
419
420 /* Don't let anyone else find us. */
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700421 rtnl_lock();
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000422 list_for_each_entry(net, &net_kill_list, cleanup_list) {
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000423 list_del_rcu(&net->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000424 list_add_tail(&net->exit_list, &net_exit_list);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200425 for_each_net(tmp) {
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200426 int id;
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200427
Paul Moore2a733062016-10-21 21:49:14 -0400428 spin_lock_irq(&tmp->nsid_lock);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200429 id = __peernet2id(tmp, net);
430 if (id >= 0)
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200431 idr_remove(&tmp->netns_ids, id);
Paul Moore2a733062016-10-21 21:49:14 -0400432 spin_unlock_irq(&tmp->nsid_lock);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200433 if (id >= 0)
434 rtnl_net_notifyid(tmp, RTM_DELNSID, id);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200435 }
Paul Moore2a733062016-10-21 21:49:14 -0400436 spin_lock_irq(&net->nsid_lock);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200437 idr_destroy(&net->netns_ids);
Paul Moore2a733062016-10-21 21:49:14 -0400438 spin_unlock_irq(&net->nsid_lock);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200439
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000440 }
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700441 rtnl_unlock();
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200442
Johannes Berg11a28d32009-07-10 09:51:33 +0000443 /*
444 * Another CPU might be rcu-iterating the list, wait for it.
445 * This needs to be before calling the exit() notifiers, so
446 * the rcu_barrier() below isn't sufficient alone.
447 */
448 synchronize_rcu();
449
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200450 /* Run all of the network namespace exit methods */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000451 list_for_each_entry_reverse(ops, &pernet_list, list)
452 ops_exit_list(ops, &net_exit_list);
453
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000454 /* Free the net generic variables */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000455 list_for_each_entry_reverse(ops, &pernet_list, list)
456 ops_free_list(ops, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200457
458 mutex_unlock(&net_mutex);
459
460 /* Ensure there are no outstanding rcu callbacks using this
461 * network namespace.
462 */
463 rcu_barrier();
464
465 /* Finally it is safe to free my network namespace structure */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000466 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
467 list_del_init(&net->exit_list);
Eric W. Biederman70328662016-08-08 14:33:23 -0500468 dec_net_namespaces(net->ucounts);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700469 put_user_ns(net->user_ns);
Al Viroa685e082011-06-08 21:13:01 -0400470 net_drop_ns(net);
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000471 }
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200472}
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000473static DECLARE_WORK(net_cleanup_work, cleanup_net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200474
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200475void __put_net(struct net *net)
476{
477 /* Cleanup the network namespace in process context */
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000478 unsigned long flags;
479
480 spin_lock_irqsave(&cleanup_list_lock, flags);
481 list_add(&net->cleanup_list, &cleanup_list);
482 spin_unlock_irqrestore(&cleanup_list_lock, flags);
483
484 queue_work(netns_wq, &net_cleanup_work);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200485}
486EXPORT_SYMBOL_GPL(__put_net);
487
Stephen Rothwell956c9202011-05-12 13:51:13 +1000488struct net *get_net_ns_by_fd(int fd)
489{
Stephen Rothwell956c9202011-05-12 13:51:13 +1000490 struct file *file;
Al Viro33c42942014-11-01 02:32:53 -0400491 struct ns_common *ns;
Stephen Rothwell956c9202011-05-12 13:51:13 +1000492 struct net *net;
493
Stephen Rothwell956c9202011-05-12 13:51:13 +1000494 file = proc_ns_fget(fd);
Al Viroc316e6a2011-06-05 00:37:35 +0000495 if (IS_ERR(file))
496 return ERR_CAST(file);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000497
Al Virof77c8012014-11-01 03:13:17 -0400498 ns = get_proc_ns(file_inode(file));
Al Viro33c42942014-11-01 02:32:53 -0400499 if (ns->ops == &netns_operations)
500 net = get_net(container_of(ns, struct net, ns));
Al Viroc316e6a2011-06-05 00:37:35 +0000501 else
502 net = ERR_PTR(-EINVAL);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000503
Al Viroc316e6a2011-06-05 00:37:35 +0000504 fput(file);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000505 return net;
506}
507
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700508#else
Stephen Rothwell956c9202011-05-12 13:51:13 +1000509struct net *get_net_ns_by_fd(int fd)
510{
511 return ERR_PTR(-EINVAL);
512}
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700513#endif
Vadim Kochan4b681c82015-01-12 16:34:05 +0200514EXPORT_SYMBOL_GPL(get_net_ns_by_fd);
Eric W. Biederman9dd776b2007-09-26 22:04:26 -0700515
Johannes Berg30ffee82009-07-10 09:51:35 +0000516struct net *get_net_ns_by_pid(pid_t pid)
517{
518 struct task_struct *tsk;
519 struct net *net;
520
521 /* Lookup the network namespace */
522 net = ERR_PTR(-ESRCH);
523 rcu_read_lock();
524 tsk = find_task_by_vpid(pid);
525 if (tsk) {
526 struct nsproxy *nsproxy;
Eric W. Biederman728dba32014-02-03 19:13:49 -0800527 task_lock(tsk);
528 nsproxy = tsk->nsproxy;
Johannes Berg30ffee82009-07-10 09:51:35 +0000529 if (nsproxy)
530 net = get_net(nsproxy->net_ns);
Eric W. Biederman728dba32014-02-03 19:13:49 -0800531 task_unlock(tsk);
Johannes Berg30ffee82009-07-10 09:51:35 +0000532 }
533 rcu_read_unlock();
534 return net;
535}
536EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
537
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700538static __net_init int net_ns_net_init(struct net *net)
539{
Al Viro33c42942014-11-01 02:32:53 -0400540#ifdef CONFIG_NET_NS
541 net->ns.ops = &netns_operations;
542#endif
Al Viro6344c432014-11-01 00:45:45 -0400543 return ns_alloc_inum(&net->ns);
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700544}
545
546static __net_exit void net_ns_net_exit(struct net *net)
547{
Al Viro6344c432014-11-01 00:45:45 -0400548 ns_free_inum(&net->ns);
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700549}
550
551static struct pernet_operations __net_initdata net_ns_ops = {
552 .init = net_ns_net_init,
553 .exit = net_ns_net_exit,
554};
555
stephen hemminger3ee52562016-08-31 15:17:49 -0700556static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = {
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100557 [NETNSA_NONE] = { .type = NLA_UNSPEC },
558 [NETNSA_NSID] = { .type = NLA_S32 },
559 [NETNSA_PID] = { .type = NLA_U32 },
560 [NETNSA_FD] = { .type = NLA_U32 },
561};
562
563static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh)
564{
565 struct net *net = sock_net(skb->sk);
566 struct nlattr *tb[NETNSA_MAX + 1];
Paul Moore2a733062016-10-21 21:49:14 -0400567 unsigned long flags;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100568 struct net *peer;
569 int nsid, err;
570
571 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
572 rtnl_net_policy);
573 if (err < 0)
574 return err;
575 if (!tb[NETNSA_NSID])
576 return -EINVAL;
577 nsid = nla_get_s32(tb[NETNSA_NSID]);
578
579 if (tb[NETNSA_PID])
580 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
581 else if (tb[NETNSA_FD])
582 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
583 else
584 return -EINVAL;
585 if (IS_ERR(peer))
586 return PTR_ERR(peer);
587
Paul Moore2a733062016-10-21 21:49:14 -0400588 spin_lock_irqsave(&net->nsid_lock, flags);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200589 if (__peernet2id(net, peer) >= 0) {
Paul Moore2a733062016-10-21 21:49:14 -0400590 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100591 err = -EEXIST;
592 goto out;
593 }
594
595 err = alloc_netid(net, peer, nsid);
Paul Moore2a733062016-10-21 21:49:14 -0400596 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200597 if (err >= 0) {
598 rtnl_net_notifyid(net, RTM_NEWNSID, err);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100599 err = 0;
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200600 }
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100601out:
602 put_net(peer);
603 return err;
604}
605
606static int rtnl_net_get_size(void)
607{
608 return NLMSG_ALIGN(sizeof(struct rtgenmsg))
609 + nla_total_size(sizeof(s32)) /* NETNSA_NSID */
610 ;
611}
612
613static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags,
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200614 int cmd, struct net *net, int nsid)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100615{
616 struct nlmsghdr *nlh;
617 struct rtgenmsg *rth;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100618
619 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags);
620 if (!nlh)
621 return -EMSGSIZE;
622
623 rth = nlmsg_data(nlh);
624 rth->rtgen_family = AF_UNSPEC;
625
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200626 if (nla_put_s32(skb, NETNSA_NSID, nsid))
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100627 goto nla_put_failure;
628
629 nlmsg_end(skb, nlh);
630 return 0;
631
632nla_put_failure:
633 nlmsg_cancel(skb, nlh);
634 return -EMSGSIZE;
635}
636
637static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh)
638{
639 struct net *net = sock_net(skb->sk);
640 struct nlattr *tb[NETNSA_MAX + 1];
641 struct sk_buff *msg;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100642 struct net *peer;
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200643 int err, id;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100644
645 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
646 rtnl_net_policy);
647 if (err < 0)
648 return err;
649 if (tb[NETNSA_PID])
650 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
651 else if (tb[NETNSA_FD])
652 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
653 else
654 return -EINVAL;
655
656 if (IS_ERR(peer))
657 return PTR_ERR(peer);
658
659 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
660 if (!msg) {
661 err = -ENOMEM;
662 goto out;
663 }
664
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200665 id = peernet2id(net, peer);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100666 err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
David S. Millerb04096f2015-05-13 14:31:43 -0400667 RTM_NEWNSID, net, id);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100668 if (err < 0)
669 goto err_out;
670
671 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid);
672 goto out;
673
674err_out:
675 nlmsg_free(msg);
676out:
677 put_net(peer);
678 return err;
679}
680
Nicolas Dichtela143c402015-04-07 11:51:54 +0200681struct rtnl_net_dump_cb {
682 struct net *net;
683 struct sk_buff *skb;
684 struct netlink_callback *cb;
685 int idx;
686 int s_idx;
687};
688
689static int rtnl_net_dumpid_one(int id, void *peer, void *data)
690{
691 struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data;
692 int ret;
693
694 if (net_cb->idx < net_cb->s_idx)
695 goto cont;
696
697 ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid,
698 net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI,
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200699 RTM_NEWNSID, net_cb->net, id);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200700 if (ret < 0)
701 return ret;
702
703cont:
704 net_cb->idx++;
705 return 0;
706}
707
708static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb)
709{
710 struct net *net = sock_net(skb->sk);
711 struct rtnl_net_dump_cb net_cb = {
712 .net = net,
713 .skb = skb,
714 .cb = cb,
715 .idx = 0,
716 .s_idx = cb->args[0],
717 };
Paul Moore2a733062016-10-21 21:49:14 -0400718 unsigned long flags;
Nicolas Dichtela143c402015-04-07 11:51:54 +0200719
Paul Moore2a733062016-10-21 21:49:14 -0400720 spin_lock_irqsave(&net->nsid_lock, flags);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200721 idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb);
Paul Moore2a733062016-10-21 21:49:14 -0400722 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200723
724 cb->args[0] = net_cb.idx;
725 return skb->len;
726}
727
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200728static void rtnl_net_notifyid(struct net *net, int cmd, int id)
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200729{
730 struct sk_buff *msg;
731 int err = -ENOMEM;
732
733 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
734 if (!msg)
735 goto out;
736
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200737 err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, id);
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200738 if (err < 0)
739 goto err_out;
740
741 rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0);
742 return;
743
744err_out:
745 nlmsg_free(msg);
746out:
747 rtnl_set_sk_err(net, RTNLGRP_NSID, err);
748}
749
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200750static int __init net_ns_init(void)
751{
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800752 struct net_generic *ng;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200753
Pavel Emelyanovd57a9212007-11-01 00:46:50 -0700754#ifdef CONFIG_NET_NS
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200755 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net),
756 SMP_CACHE_BYTES,
757 SLAB_PANIC, NULL);
Benjamin Thery3ef13552007-11-19 23:18:16 -0800758
759 /* Create workqueue for cleanup */
760 netns_wq = create_singlethread_workqueue("netns");
761 if (!netns_wq)
762 panic("Could not create netns workq");
Pavel Emelyanovd57a9212007-11-01 00:46:50 -0700763#endif
Benjamin Thery3ef13552007-11-19 23:18:16 -0800764
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800765 ng = net_alloc_generic();
766 if (!ng)
767 panic("Could not allocate generic netns");
768
769 rcu_assign_pointer(init_net.gen, ng);
770
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200771 mutex_lock(&net_mutex);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700772 if (setup_net(&init_net, &init_user_ns))
Stephen Hemmingerca0f3112009-05-21 15:10:31 -0700773 panic("Could not setup the initial network namespace");
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200774
Dmitry Torokhovf8c46cb2016-08-10 14:36:00 -0700775 init_net_initialized = true;
776
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700777 rtnl_lock();
Johannes Berg11a28d32009-07-10 09:51:33 +0000778 list_add_tail_rcu(&init_net.list, &net_namespace_list);
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700779 rtnl_unlock();
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200780
781 mutex_unlock(&net_mutex);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200782
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700783 register_pernet_subsys(&net_ns_ops);
784
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100785 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200786 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid,
787 NULL);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100788
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200789 return 0;
790}
791
792pure_initcall(net_ns_init);
793
Denis V. Luneved160e82007-11-13 03:23:21 -0800794#ifdef CONFIG_NET_NS
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000795static int __register_pernet_operations(struct list_head *list,
796 struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200797{
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000798 struct net *net;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200799 int error;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000800 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200801
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200802 list_add_tail(&ops->list, list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000803 if (ops->init || (ops->id && ops->size)) {
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700804 for_each_net(net) {
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000805 error = ops_init(ops, net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200806 if (error)
807 goto out_undo;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000808 list_add_tail(&net->exit_list, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200809 }
810 }
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700811 return 0;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200812
813out_undo:
814 /* If I have an error cleanup all namespaces I initialized */
815 list_del(&ops->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000816 ops_exit_list(ops, &net_exit_list);
817 ops_free_list(ops, &net_exit_list);
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700818 return error;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200819}
820
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000821static void __unregister_pernet_operations(struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200822{
823 struct net *net;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000824 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200825
826 list_del(&ops->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000827 for_each_net(net)
828 list_add_tail(&net->exit_list, &net_exit_list);
829 ops_exit_list(ops, &net_exit_list);
830 ops_free_list(ops, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200831}
832
Denis V. Luneved160e82007-11-13 03:23:21 -0800833#else
834
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000835static int __register_pernet_operations(struct list_head *list,
836 struct pernet_operations *ops)
837{
Dmitry Torokhovf8c46cb2016-08-10 14:36:00 -0700838 if (!init_net_initialized) {
839 list_add_tail(&ops->list, list);
840 return 0;
841 }
842
Julian Anastasovb9229342012-04-16 04:43:15 +0000843 return ops_init(ops, &init_net);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000844}
845
846static void __unregister_pernet_operations(struct pernet_operations *ops)
847{
Dmitry Torokhovf8c46cb2016-08-10 14:36:00 -0700848 if (!init_net_initialized) {
849 list_del(&ops->list);
850 } else {
851 LIST_HEAD(net_exit_list);
852 list_add(&init_net.exit_list, &net_exit_list);
853 ops_exit_list(ops, &net_exit_list);
854 ops_free_list(ops, &net_exit_list);
855 }
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000856}
857
858#endif /* CONFIG_NET_NS */
859
860static DEFINE_IDA(net_generic_ids);
861
Denis V. Luneved160e82007-11-13 03:23:21 -0800862static int register_pernet_operations(struct list_head *list,
863 struct pernet_operations *ops)
864{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000865 int error;
866
867 if (ops->id) {
868again:
869 error = ida_get_new_above(&net_generic_ids, 1, ops->id);
870 if (error < 0) {
871 if (error == -EAGAIN) {
872 ida_pre_get(&net_generic_ids, GFP_KERNEL);
873 goto again;
874 }
875 return error;
876 }
Eric Dumazet073862b2012-01-26 00:41:38 +0000877 max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000878 }
879 error = __register_pernet_operations(list, ops);
Eric W. Biederman3a765ed2009-12-03 02:29:06 +0000880 if (error) {
881 rcu_barrier();
882 if (ops->id)
883 ida_remove(&net_generic_ids, *ops->id);
884 }
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000885
886 return error;
Denis V. Luneved160e82007-11-13 03:23:21 -0800887}
888
889static void unregister_pernet_operations(struct pernet_operations *ops)
890{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000891
892 __unregister_pernet_operations(ops);
Eric W. Biederman3a765ed2009-12-03 02:29:06 +0000893 rcu_barrier();
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000894 if (ops->id)
895 ida_remove(&net_generic_ids, *ops->id);
Denis V. Luneved160e82007-11-13 03:23:21 -0800896}
Pavel Emelyanovc93cf612008-04-15 00:35:23 -0700897
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200898/**
899 * register_pernet_subsys - register a network namespace subsystem
900 * @ops: pernet operations structure for the subsystem
901 *
902 * Register a subsystem which has init and exit functions
903 * that are called when network namespaces are created and
904 * destroyed respectively.
905 *
906 * When registered all network namespace init functions are
907 * called for every existing network namespace. Allowing kernel
908 * modules to have a race free view of the set of network namespaces.
909 *
910 * When a new network namespace is created all of the init
911 * methods are called in the order in which they were registered.
912 *
913 * When a network namespace is destroyed all of the exit methods
914 * are called in the reverse of the order with which they were
915 * registered.
916 */
917int register_pernet_subsys(struct pernet_operations *ops)
918{
919 int error;
920 mutex_lock(&net_mutex);
921 error = register_pernet_operations(first_device, ops);
922 mutex_unlock(&net_mutex);
923 return error;
924}
925EXPORT_SYMBOL_GPL(register_pernet_subsys);
926
927/**
928 * unregister_pernet_subsys - unregister a network namespace subsystem
929 * @ops: pernet operations structure to manipulate
930 *
931 * Remove the pernet operations structure from the list to be
Oliver Pinter53379e52008-02-03 17:56:48 +0200932 * used when network namespaces are created or destroyed. In
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200933 * addition run the exit method for all existing network
934 * namespaces.
935 */
Jiri Pirkob3c981d2010-04-25 00:49:56 -0700936void unregister_pernet_subsys(struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200937{
938 mutex_lock(&net_mutex);
Jiri Pirkob3c981d2010-04-25 00:49:56 -0700939 unregister_pernet_operations(ops);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200940 mutex_unlock(&net_mutex);
941}
942EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
943
944/**
945 * register_pernet_device - register a network namespace device
946 * @ops: pernet operations structure for the subsystem
947 *
948 * Register a device which has init and exit functions
949 * that are called when network namespaces are created and
950 * destroyed respectively.
951 *
952 * When registered all network namespace init functions are
953 * called for every existing network namespace. Allowing kernel
954 * modules to have a race free view of the set of network namespaces.
955 *
956 * When a new network namespace is created all of the init
957 * methods are called in the order in which they were registered.
958 *
959 * When a network namespace is destroyed all of the exit methods
960 * are called in the reverse of the order with which they were
961 * registered.
962 */
963int register_pernet_device(struct pernet_operations *ops)
964{
965 int error;
966 mutex_lock(&net_mutex);
967 error = register_pernet_operations(&pernet_list, ops);
968 if (!error && (first_device == &pernet_list))
969 first_device = &ops->list;
970 mutex_unlock(&net_mutex);
971 return error;
972}
973EXPORT_SYMBOL_GPL(register_pernet_device);
974
975/**
976 * unregister_pernet_device - unregister a network namespace netdevice
977 * @ops: pernet operations structure to manipulate
978 *
979 * Remove the pernet operations structure from the list to be
Oliver Pinter53379e52008-02-03 17:56:48 +0200980 * used when network namespaces are created or destroyed. In
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200981 * addition run the exit method for all existing network
982 * namespaces.
983 */
984void unregister_pernet_device(struct pernet_operations *ops)
985{
986 mutex_lock(&net_mutex);
987 if (&ops->list == first_device)
988 first_device = first_device->next;
989 unregister_pernet_operations(ops);
990 mutex_unlock(&net_mutex);
991}
992EXPORT_SYMBOL_GPL(unregister_pernet_device);
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800993
994#ifdef CONFIG_NET_NS
Al Viro64964522014-11-01 00:37:32 -0400995static struct ns_common *netns_get(struct task_struct *task)
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800996{
Eric W. Biedermanf0630522011-05-04 17:51:50 -0700997 struct net *net = NULL;
998 struct nsproxy *nsproxy;
999
Eric W. Biederman728dba32014-02-03 19:13:49 -08001000 task_lock(task);
1001 nsproxy = task->nsproxy;
Eric W. Biedermanf0630522011-05-04 17:51:50 -07001002 if (nsproxy)
1003 net = get_net(nsproxy->net_ns);
Eric W. Biederman728dba32014-02-03 19:13:49 -08001004 task_unlock(task);
Eric W. Biedermanf0630522011-05-04 17:51:50 -07001005
Al Viroff248702014-11-01 00:10:50 -04001006 return net ? &net->ns : NULL;
1007}
1008
1009static inline struct net *to_net_ns(struct ns_common *ns)
1010{
1011 return container_of(ns, struct net, ns);
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001012}
1013
Al Viro64964522014-11-01 00:37:32 -04001014static void netns_put(struct ns_common *ns)
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001015{
Al Viroff248702014-11-01 00:10:50 -04001016 put_net(to_net_ns(ns));
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001017}
1018
Al Viro64964522014-11-01 00:37:32 -04001019static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns)
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001020{
Al Viroff248702014-11-01 00:10:50 -04001021 struct net *net = to_net_ns(ns);
Eric W. Biederman142e1d12012-07-26 01:13:20 -07001022
Eric W. Biederman5e4a0842012-12-14 07:55:36 -08001023 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -07001024 !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
Eric W. Biederman142e1d12012-07-26 01:13:20 -07001025 return -EPERM;
1026
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001027 put_net(nsproxy->net_ns);
Eric W. Biederman142e1d12012-07-26 01:13:20 -07001028 nsproxy->net_ns = get_net(net);
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001029 return 0;
1030}
1031
Andrey Vaginbcac25a2016-09-06 00:47:13 -07001032static struct user_namespace *netns_owner(struct ns_common *ns)
1033{
1034 return to_net_ns(ns)->user_ns;
1035}
1036
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001037const struct proc_ns_operations netns_operations = {
1038 .name = "net",
1039 .type = CLONE_NEWNET,
1040 .get = netns_get,
1041 .put = netns_put,
1042 .install = netns_install,
Andrey Vaginbcac25a2016-09-06 00:47:13 -07001043 .owner = netns_owner,
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001044};
1045#endif