blob: 7c6b51a58968613010576f3bd35a3cb6c527db08 [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>
Eric W. Biederman5f256be2007-09-12 11:50:50 +020018#include <net/net_namespace.h>
Pavel Emelyanovdec827d2008-04-15 00:36:08 -070019#include <net/netns/generic.h>
Eric W. Biederman5f256be2007-09-12 11:50:50 +020020
21/*
22 * Our network namespace constructor/destructor lists
23 */
24
25static LIST_HEAD(pernet_list);
26static struct list_head *first_device = &pernet_list;
Cong Wang200b9162014-05-12 15:11:20 -070027DEFINE_MUTEX(net_mutex);
Eric W. Biederman5f256be2007-09-12 11:50:50 +020028
Eric W. Biederman5f256be2007-09-12 11:50:50 +020029LIST_HEAD(net_namespace_list);
Alexey Dobriyanb76a4612008-10-08 11:35:06 +020030EXPORT_SYMBOL_GPL(net_namespace_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +020031
Rustad, Mark D734b6542012-07-18 09:06:07 +000032struct net init_net = {
33 .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head),
34};
Denis V. Lunevff4b9502008-01-22 22:05:33 -080035EXPORT_SYMBOL(init_net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +020036
Pavel Emelyanovdec827d2008-04-15 00:36:08 -070037#define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
38
Eric Dumazet073862b2012-01-26 00:41:38 +000039static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS;
40
41static struct net_generic *net_alloc_generic(void)
42{
43 struct net_generic *ng;
44 size_t generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]);
45
46 ng = kzalloc(generic_size, GFP_KERNEL);
47 if (ng)
48 ng->len = max_gen_ptrs;
49
50 return ng;
51}
52
Jiri Pirko05fceb42010-04-23 01:40:47 +000053static int net_assign_generic(struct net *net, int id, void *data)
54{
55 struct net_generic *ng, *old_ng;
56
57 BUG_ON(!mutex_is_locked(&net_mutex));
58 BUG_ON(id == 0);
59
Eric Dumazet1c877332010-10-25 03:20:11 +000060 old_ng = rcu_dereference_protected(net->gen,
61 lockdep_is_held(&net_mutex));
62 ng = old_ng;
Jiri Pirko05fceb42010-04-23 01:40:47 +000063 if (old_ng->len >= id)
64 goto assign;
65
Eric Dumazet073862b2012-01-26 00:41:38 +000066 ng = net_alloc_generic();
Jiri Pirko05fceb42010-04-23 01:40:47 +000067 if (ng == NULL)
68 return -ENOMEM;
69
70 /*
71 * Some synchronisation notes:
72 *
73 * The net_generic explores the net->gen array inside rcu
74 * read section. Besides once set the net->gen->ptr[x]
75 * pointer never changes (see rules in netns/generic.h).
76 *
77 * That said, we simply duplicate this array and schedule
78 * the old copy for kfree after a grace period.
79 */
80
Jiri Pirko05fceb42010-04-23 01:40:47 +000081 memcpy(&ng->ptr, &old_ng->ptr, old_ng->len * sizeof(void*));
82
83 rcu_assign_pointer(net->gen, ng);
Lai Jiangshan04d4dfe2011-03-18 12:06:32 +080084 kfree_rcu(old_ng, rcu);
Jiri Pirko05fceb42010-04-23 01:40:47 +000085assign:
86 ng->ptr[id - 1] = data;
87 return 0;
88}
89
Eric W. Biedermanf875bae2009-11-29 22:25:28 +000090static int ops_init(const struct pernet_operations *ops, struct net *net)
91{
Julian Anastasovb9229342012-04-16 04:43:15 +000092 int err = -ENOMEM;
93 void *data = NULL;
94
Eric W. Biedermanf875bae2009-11-29 22:25:28 +000095 if (ops->id && ops->size) {
Julian Anastasovb9229342012-04-16 04:43:15 +000096 data = kzalloc(ops->size, GFP_KERNEL);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +000097 if (!data)
Julian Anastasovb9229342012-04-16 04:43:15 +000098 goto out;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +000099
100 err = net_assign_generic(net, *ops->id, data);
Julian Anastasovb9229342012-04-16 04:43:15 +0000101 if (err)
102 goto cleanup;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000103 }
Julian Anastasovb9229342012-04-16 04:43:15 +0000104 err = 0;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000105 if (ops->init)
Julian Anastasovb9229342012-04-16 04:43:15 +0000106 err = ops->init(net);
107 if (!err)
108 return 0;
109
110cleanup:
111 kfree(data);
112
113out:
114 return err;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000115}
116
117static void ops_free(const struct pernet_operations *ops, struct net *net)
118{
119 if (ops->id && ops->size) {
120 int id = *ops->id;
121 kfree(net_generic(net, id));
122 }
123}
124
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000125static void ops_exit_list(const struct pernet_operations *ops,
126 struct list_head *net_exit_list)
127{
128 struct net *net;
129 if (ops->exit) {
130 list_for_each_entry(net, net_exit_list, exit_list)
131 ops->exit(net);
132 }
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000133 if (ops->exit_batch)
134 ops->exit_batch(net_exit_list);
135}
136
137static void ops_free_list(const struct pernet_operations *ops,
138 struct list_head *net_exit_list)
139{
140 struct net *net;
141 if (ops->size && ops->id) {
142 list_for_each_entry(net, net_exit_list, exit_list)
143 ops_free(ops, net);
144 }
145}
146
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700147/*
148 * setup_net runs the initializers for the network namespace object.
149 */
Eric W. Biederman038e7332012-06-14 02:31:10 -0700150static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700151{
152 /* Must be called with net_mutex held */
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000153 const struct pernet_operations *ops, *saved_ops;
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800154 int error = 0;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000155 LIST_HEAD(net_exit_list);
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700156
157 atomic_set(&net->count, 1);
Al Viroa685e082011-06-08 21:13:01 -0400158 atomic_set(&net->passive, 1);
Thomas Graf4e985ad2011-06-21 03:11:20 +0000159 net->dev_base_seq = 1;
Eric W. Biederman038e7332012-06-14 02:31:10 -0700160 net->user_ns = user_ns;
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800161
Denis V. Lunev5d1e4462008-04-16 01:58:04 -0700162#ifdef NETNS_REFCNT_DEBUG
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700163 atomic_set(&net->use_count, 0);
Denis V. Lunev5d1e4462008-04-16 01:58:04 -0700164#endif
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700165
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700166 list_for_each_entry(ops, &pernet_list, list) {
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000167 error = ops_init(ops, net);
168 if (error < 0)
169 goto out_undo;
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700170 }
171out:
172 return error;
173
174out_undo:
175 /* Walk through the list backwards calling the exit functions
176 * for the pernet modules whose init functions did not fail.
177 */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000178 list_add(&net->exit_list, &net_exit_list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000179 saved_ops = ops;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000180 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
181 ops_exit_list(ops, &net_exit_list);
182
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000183 ops = saved_ops;
184 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000185 ops_free_list(ops, &net_exit_list);
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700186
187 rcu_barrier();
188 goto out;
189}
190
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800191
Clemens Nossebe47d42009-02-23 15:37:35 -0800192#ifdef CONFIG_NET_NS
193static struct kmem_cache *net_cachep;
194static struct workqueue_struct *netns_wq;
195
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200196static struct net *net_alloc(void)
197{
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800198 struct net *net = NULL;
199 struct net_generic *ng;
200
201 ng = net_alloc_generic();
202 if (!ng)
203 goto out;
204
205 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL);
206 if (!net)
207 goto out_free;
208
209 rcu_assign_pointer(net->gen, ng);
210out:
211 return net;
212
213out_free:
214 kfree(ng);
215 goto out;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200216}
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200217
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800218static void net_free(struct net *net)
219{
Denis V. Lunev5d1e4462008-04-16 01:58:04 -0700220#ifdef NETNS_REFCNT_DEBUG
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800221 if (unlikely(atomic_read(&net->use_count) != 0)) {
Joe Perchese005d192012-05-16 19:58:40 +0000222 pr_emerg("network namespace not free! Usage: %d\n",
223 atomic_read(&net->use_count));
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800224 return;
225 }
Denis V. Lunev5d1e4462008-04-16 01:58:04 -0700226#endif
Alexey Dobriyan4ef079c2008-10-14 22:54:48 -0700227 kfree(net->gen);
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800228 kmem_cache_free(net_cachep, net);
229}
230
Al Viroa685e082011-06-08 21:13:01 -0400231void net_drop_ns(void *p)
232{
233 struct net *ns = p;
234 if (ns && atomic_dec_and_test(&ns->passive))
235 net_free(ns);
236}
237
Eric W. Biederman038e7332012-06-14 02:31:10 -0700238struct net *copy_net_ns(unsigned long flags,
239 struct user_namespace *user_ns, struct net *old_net)
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700240{
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700241 struct net *net;
242 int rv;
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700243
Rob Landley911cb192011-04-15 02:26:25 +0000244 if (!(flags & CLONE_NEWNET))
245 return get_net(old_net);
246
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700247 net = net_alloc();
248 if (!net)
249 return ERR_PTR(-ENOMEM);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700250
251 get_user_ns(user_ns);
252
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700253 mutex_lock(&net_mutex);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700254 rv = setup_net(net, user_ns);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700255 if (rv == 0) {
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800256 rtnl_lock();
Johannes Berg11a28d32009-07-10 09:51:33 +0000257 list_add_tail_rcu(&net->list, &net_namespace_list);
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800258 rtnl_unlock();
259 }
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700260 mutex_unlock(&net_mutex);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700261 if (rv < 0) {
Eric W. Biederman038e7332012-06-14 02:31:10 -0700262 put_user_ns(user_ns);
Al Viroa685e082011-06-08 21:13:01 -0400263 net_drop_ns(net);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700264 return ERR_PTR(rv);
265 }
266 return net;
267}
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800268
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000269static DEFINE_SPINLOCK(cleanup_list_lock);
270static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */
271
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200272static void cleanup_net(struct work_struct *work)
273{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000274 const struct pernet_operations *ops;
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000275 struct net *net, *tmp;
xiao jin1818ce42014-04-25 08:50:54 +0800276 struct list_head net_kill_list;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000277 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200278
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000279 /* Atomically snapshot the list of namespaces to cleanup */
280 spin_lock_irq(&cleanup_list_lock);
281 list_replace_init(&cleanup_list, &net_kill_list);
282 spin_unlock_irq(&cleanup_list_lock);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200283
284 mutex_lock(&net_mutex);
285
286 /* Don't let anyone else find us. */
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700287 rtnl_lock();
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000288 list_for_each_entry(net, &net_kill_list, cleanup_list) {
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000289 list_del_rcu(&net->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000290 list_add_tail(&net->exit_list, &net_exit_list);
291 }
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700292 rtnl_unlock();
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200293
Johannes Berg11a28d32009-07-10 09:51:33 +0000294 /*
295 * Another CPU might be rcu-iterating the list, wait for it.
296 * This needs to be before calling the exit() notifiers, so
297 * the rcu_barrier() below isn't sufficient alone.
298 */
299 synchronize_rcu();
300
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200301 /* Run all of the network namespace exit methods */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000302 list_for_each_entry_reverse(ops, &pernet_list, list)
303 ops_exit_list(ops, &net_exit_list);
304
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000305 /* Free the net generic variables */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000306 list_for_each_entry_reverse(ops, &pernet_list, list)
307 ops_free_list(ops, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200308
309 mutex_unlock(&net_mutex);
310
311 /* Ensure there are no outstanding rcu callbacks using this
312 * network namespace.
313 */
314 rcu_barrier();
315
316 /* Finally it is safe to free my network namespace structure */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000317 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
318 list_del_init(&net->exit_list);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700319 put_user_ns(net->user_ns);
Al Viroa685e082011-06-08 21:13:01 -0400320 net_drop_ns(net);
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000321 }
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200322}
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000323static DECLARE_WORK(net_cleanup_work, cleanup_net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200324
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200325void __put_net(struct net *net)
326{
327 /* Cleanup the network namespace in process context */
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000328 unsigned long flags;
329
330 spin_lock_irqsave(&cleanup_list_lock, flags);
331 list_add(&net->cleanup_list, &cleanup_list);
332 spin_unlock_irqrestore(&cleanup_list_lock, flags);
333
334 queue_work(netns_wq, &net_cleanup_work);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200335}
336EXPORT_SYMBOL_GPL(__put_net);
337
Stephen Rothwell956c9202011-05-12 13:51:13 +1000338struct net *get_net_ns_by_fd(int fd)
339{
David Howells0bb80f22013-04-12 01:50:06 +0100340 struct proc_ns *ei;
Stephen Rothwell956c9202011-05-12 13:51:13 +1000341 struct file *file;
342 struct net *net;
343
Stephen Rothwell956c9202011-05-12 13:51:13 +1000344 file = proc_ns_fget(fd);
Al Viroc316e6a2011-06-05 00:37:35 +0000345 if (IS_ERR(file))
346 return ERR_CAST(file);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000347
David Howells0bb80f22013-04-12 01:50:06 +0100348 ei = get_proc_ns(file_inode(file));
Al Viroc316e6a2011-06-05 00:37:35 +0000349 if (ei->ns_ops == &netns_operations)
350 net = get_net(ei->ns);
351 else
352 net = ERR_PTR(-EINVAL);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000353
Al Viroc316e6a2011-06-05 00:37:35 +0000354 fput(file);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000355 return net;
356}
357
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700358#else
Stephen Rothwell956c9202011-05-12 13:51:13 +1000359struct net *get_net_ns_by_fd(int fd)
360{
361 return ERR_PTR(-EINVAL);
362}
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700363#endif
Eric W. Biederman9dd776b2007-09-26 22:04:26 -0700364
Johannes Berg30ffee82009-07-10 09:51:35 +0000365struct net *get_net_ns_by_pid(pid_t pid)
366{
367 struct task_struct *tsk;
368 struct net *net;
369
370 /* Lookup the network namespace */
371 net = ERR_PTR(-ESRCH);
372 rcu_read_lock();
373 tsk = find_task_by_vpid(pid);
374 if (tsk) {
375 struct nsproxy *nsproxy;
Eric W. Biederman728dba32014-02-03 19:13:49 -0800376 task_lock(tsk);
377 nsproxy = tsk->nsproxy;
Johannes Berg30ffee82009-07-10 09:51:35 +0000378 if (nsproxy)
379 net = get_net(nsproxy->net_ns);
Eric W. Biederman728dba32014-02-03 19:13:49 -0800380 task_unlock(tsk);
Johannes Berg30ffee82009-07-10 09:51:35 +0000381 }
382 rcu_read_unlock();
383 return net;
384}
385EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
386
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700387static __net_init int net_ns_net_init(struct net *net)
388{
389 return proc_alloc_inum(&net->proc_inum);
390}
391
392static __net_exit void net_ns_net_exit(struct net *net)
393{
394 proc_free_inum(net->proc_inum);
395}
396
397static struct pernet_operations __net_initdata net_ns_ops = {
398 .init = net_ns_net_init,
399 .exit = net_ns_net_exit,
400};
401
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200402static int __init net_ns_init(void)
403{
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800404 struct net_generic *ng;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200405
Pavel Emelyanovd57a9212007-11-01 00:46:50 -0700406#ifdef CONFIG_NET_NS
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200407 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net),
408 SMP_CACHE_BYTES,
409 SLAB_PANIC, NULL);
Benjamin Thery3ef13552007-11-19 23:18:16 -0800410
411 /* Create workqueue for cleanup */
412 netns_wq = create_singlethread_workqueue("netns");
413 if (!netns_wq)
414 panic("Could not create netns workq");
Pavel Emelyanovd57a9212007-11-01 00:46:50 -0700415#endif
Benjamin Thery3ef13552007-11-19 23:18:16 -0800416
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800417 ng = net_alloc_generic();
418 if (!ng)
419 panic("Could not allocate generic netns");
420
421 rcu_assign_pointer(init_net.gen, ng);
422
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200423 mutex_lock(&net_mutex);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700424 if (setup_net(&init_net, &init_user_ns))
Stephen Hemmingerca0f3112009-05-21 15:10:31 -0700425 panic("Could not setup the initial network namespace");
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200426
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700427 rtnl_lock();
Johannes Berg11a28d32009-07-10 09:51:33 +0000428 list_add_tail_rcu(&init_net.list, &net_namespace_list);
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700429 rtnl_unlock();
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200430
431 mutex_unlock(&net_mutex);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200432
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700433 register_pernet_subsys(&net_ns_ops);
434
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200435 return 0;
436}
437
438pure_initcall(net_ns_init);
439
Denis V. Luneved160e82007-11-13 03:23:21 -0800440#ifdef CONFIG_NET_NS
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000441static int __register_pernet_operations(struct list_head *list,
442 struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200443{
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000444 struct net *net;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200445 int error;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000446 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200447
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200448 list_add_tail(&ops->list, list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000449 if (ops->init || (ops->id && ops->size)) {
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700450 for_each_net(net) {
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000451 error = ops_init(ops, net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200452 if (error)
453 goto out_undo;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000454 list_add_tail(&net->exit_list, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200455 }
456 }
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700457 return 0;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200458
459out_undo:
460 /* If I have an error cleanup all namespaces I initialized */
461 list_del(&ops->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000462 ops_exit_list(ops, &net_exit_list);
463 ops_free_list(ops, &net_exit_list);
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700464 return error;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200465}
466
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000467static void __unregister_pernet_operations(struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200468{
469 struct net *net;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000470 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200471
472 list_del(&ops->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000473 for_each_net(net)
474 list_add_tail(&net->exit_list, &net_exit_list);
475 ops_exit_list(ops, &net_exit_list);
476 ops_free_list(ops, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200477}
478
Denis V. Luneved160e82007-11-13 03:23:21 -0800479#else
480
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000481static int __register_pernet_operations(struct list_head *list,
482 struct pernet_operations *ops)
483{
Julian Anastasovb9229342012-04-16 04:43:15 +0000484 return ops_init(ops, &init_net);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000485}
486
487static void __unregister_pernet_operations(struct pernet_operations *ops)
488{
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000489 LIST_HEAD(net_exit_list);
490 list_add(&init_net.exit_list, &net_exit_list);
491 ops_exit_list(ops, &net_exit_list);
492 ops_free_list(ops, &net_exit_list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000493}
494
495#endif /* CONFIG_NET_NS */
496
497static DEFINE_IDA(net_generic_ids);
498
Denis V. Luneved160e82007-11-13 03:23:21 -0800499static int register_pernet_operations(struct list_head *list,
500 struct pernet_operations *ops)
501{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000502 int error;
503
504 if (ops->id) {
505again:
506 error = ida_get_new_above(&net_generic_ids, 1, ops->id);
507 if (error < 0) {
508 if (error == -EAGAIN) {
509 ida_pre_get(&net_generic_ids, GFP_KERNEL);
510 goto again;
511 }
512 return error;
513 }
Eric Dumazet073862b2012-01-26 00:41:38 +0000514 max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000515 }
516 error = __register_pernet_operations(list, ops);
Eric W. Biederman3a765ed2009-12-03 02:29:06 +0000517 if (error) {
518 rcu_barrier();
519 if (ops->id)
520 ida_remove(&net_generic_ids, *ops->id);
521 }
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000522
523 return error;
Denis V. Luneved160e82007-11-13 03:23:21 -0800524}
525
526static void unregister_pernet_operations(struct pernet_operations *ops)
527{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000528
529 __unregister_pernet_operations(ops);
Eric W. Biederman3a765ed2009-12-03 02:29:06 +0000530 rcu_barrier();
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000531 if (ops->id)
532 ida_remove(&net_generic_ids, *ops->id);
Denis V. Luneved160e82007-11-13 03:23:21 -0800533}
Pavel Emelyanovc93cf612008-04-15 00:35:23 -0700534
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200535/**
536 * register_pernet_subsys - register a network namespace subsystem
537 * @ops: pernet operations structure for the subsystem
538 *
539 * Register a subsystem which has init and exit functions
540 * that are called when network namespaces are created and
541 * destroyed respectively.
542 *
543 * When registered all network namespace init functions are
544 * called for every existing network namespace. Allowing kernel
545 * modules to have a race free view of the set of network namespaces.
546 *
547 * When a new network namespace is created all of the init
548 * methods are called in the order in which they were registered.
549 *
550 * When a network namespace is destroyed all of the exit methods
551 * are called in the reverse of the order with which they were
552 * registered.
553 */
554int register_pernet_subsys(struct pernet_operations *ops)
555{
556 int error;
557 mutex_lock(&net_mutex);
558 error = register_pernet_operations(first_device, ops);
559 mutex_unlock(&net_mutex);
560 return error;
561}
562EXPORT_SYMBOL_GPL(register_pernet_subsys);
563
564/**
565 * unregister_pernet_subsys - unregister a network namespace subsystem
566 * @ops: pernet operations structure to manipulate
567 *
568 * Remove the pernet operations structure from the list to be
Oliver Pinter53379e52008-02-03 17:56:48 +0200569 * used when network namespaces are created or destroyed. In
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200570 * addition run the exit method for all existing network
571 * namespaces.
572 */
Jiri Pirkob3c981d2010-04-25 00:49:56 -0700573void unregister_pernet_subsys(struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200574{
575 mutex_lock(&net_mutex);
Jiri Pirkob3c981d2010-04-25 00:49:56 -0700576 unregister_pernet_operations(ops);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200577 mutex_unlock(&net_mutex);
578}
579EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
580
581/**
582 * register_pernet_device - register a network namespace device
583 * @ops: pernet operations structure for the subsystem
584 *
585 * Register a device which has init and exit functions
586 * that are called when network namespaces are created and
587 * destroyed respectively.
588 *
589 * When registered all network namespace init functions are
590 * called for every existing network namespace. Allowing kernel
591 * modules to have a race free view of the set of network namespaces.
592 *
593 * When a new network namespace is created all of the init
594 * methods are called in the order in which they were registered.
595 *
596 * When a network namespace is destroyed all of the exit methods
597 * are called in the reverse of the order with which they were
598 * registered.
599 */
600int register_pernet_device(struct pernet_operations *ops)
601{
602 int error;
603 mutex_lock(&net_mutex);
604 error = register_pernet_operations(&pernet_list, ops);
605 if (!error && (first_device == &pernet_list))
606 first_device = &ops->list;
607 mutex_unlock(&net_mutex);
608 return error;
609}
610EXPORT_SYMBOL_GPL(register_pernet_device);
611
612/**
613 * unregister_pernet_device - unregister a network namespace netdevice
614 * @ops: pernet operations structure to manipulate
615 *
616 * Remove the pernet operations structure from the list to be
Oliver Pinter53379e52008-02-03 17:56:48 +0200617 * used when network namespaces are created or destroyed. In
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200618 * addition run the exit method for all existing network
619 * namespaces.
620 */
621void unregister_pernet_device(struct pernet_operations *ops)
622{
623 mutex_lock(&net_mutex);
624 if (&ops->list == first_device)
625 first_device = first_device->next;
626 unregister_pernet_operations(ops);
627 mutex_unlock(&net_mutex);
628}
629EXPORT_SYMBOL_GPL(unregister_pernet_device);
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800630
631#ifdef CONFIG_NET_NS
632static void *netns_get(struct task_struct *task)
633{
Eric W. Biedermanf0630522011-05-04 17:51:50 -0700634 struct net *net = NULL;
635 struct nsproxy *nsproxy;
636
Eric W. Biederman728dba32014-02-03 19:13:49 -0800637 task_lock(task);
638 nsproxy = task->nsproxy;
Eric W. Biedermanf0630522011-05-04 17:51:50 -0700639 if (nsproxy)
640 net = get_net(nsproxy->net_ns);
Eric W. Biederman728dba32014-02-03 19:13:49 -0800641 task_unlock(task);
Eric W. Biedermanf0630522011-05-04 17:51:50 -0700642
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800643 return net;
644}
645
646static void netns_put(void *ns)
647{
648 put_net(ns);
649}
650
651static int netns_install(struct nsproxy *nsproxy, void *ns)
652{
Eric W. Biederman142e1d12012-07-26 01:13:20 -0700653 struct net *net = ns;
654
Eric W. Biederman5e4a0842012-12-14 07:55:36 -0800655 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700656 !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
Eric W. Biederman142e1d12012-07-26 01:13:20 -0700657 return -EPERM;
658
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800659 put_net(nsproxy->net_ns);
Eric W. Biederman142e1d12012-07-26 01:13:20 -0700660 nsproxy->net_ns = get_net(net);
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800661 return 0;
662}
663
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700664static unsigned int netns_inum(void *ns)
665{
666 struct net *net = ns;
667 return net->proc_inum;
668}
669
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800670const struct proc_ns_operations netns_operations = {
671 .name = "net",
672 .type = CLONE_NEWNET,
673 .get = netns_get,
674 .put = netns_put,
675 .install = netns_install,
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700676 .inum = netns_inum,
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800677};
678#endif