blob: 06af5d6a883cfeb6c67b7fb5d0e4bcb204a60a99 [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
Pavel Emelyanovdec827d2008-04-15 00:36:08 -070040#define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
41
Eric Dumazet073862b2012-01-26 00:41:38 +000042static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS;
43
44static struct net_generic *net_alloc_generic(void)
45{
46 struct net_generic *ng;
47 size_t generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]);
48
49 ng = kzalloc(generic_size, GFP_KERNEL);
50 if (ng)
51 ng->len = max_gen_ptrs;
52
53 return ng;
54}
55
Jiri Pirko05fceb42010-04-23 01:40:47 +000056static int net_assign_generic(struct net *net, int id, void *data)
57{
58 struct net_generic *ng, *old_ng;
59
60 BUG_ON(!mutex_is_locked(&net_mutex));
61 BUG_ON(id == 0);
62
Eric Dumazet1c877332010-10-25 03:20:11 +000063 old_ng = rcu_dereference_protected(net->gen,
64 lockdep_is_held(&net_mutex));
65 ng = old_ng;
Jiri Pirko05fceb42010-04-23 01:40:47 +000066 if (old_ng->len >= id)
67 goto assign;
68
Eric Dumazet073862b2012-01-26 00:41:38 +000069 ng = net_alloc_generic();
Jiri Pirko05fceb42010-04-23 01:40:47 +000070 if (ng == NULL)
71 return -ENOMEM;
72
73 /*
74 * Some synchronisation notes:
75 *
76 * The net_generic explores the net->gen array inside rcu
77 * read section. Besides once set the net->gen->ptr[x]
78 * pointer never changes (see rules in netns/generic.h).
79 *
80 * That said, we simply duplicate this array and schedule
81 * the old copy for kfree after a grace period.
82 */
83
Jiri Pirko05fceb42010-04-23 01:40:47 +000084 memcpy(&ng->ptr, &old_ng->ptr, old_ng->len * sizeof(void*));
85
86 rcu_assign_pointer(net->gen, ng);
Lai Jiangshan04d4dfe2011-03-18 12:06:32 +080087 kfree_rcu(old_ng, rcu);
Jiri Pirko05fceb42010-04-23 01:40:47 +000088assign:
89 ng->ptr[id - 1] = data;
90 return 0;
91}
92
Eric W. Biedermanf875bae2009-11-29 22:25:28 +000093static int ops_init(const struct pernet_operations *ops, struct net *net)
94{
Julian Anastasovb9229342012-04-16 04:43:15 +000095 int err = -ENOMEM;
96 void *data = NULL;
97
Eric W. Biedermanf875bae2009-11-29 22:25:28 +000098 if (ops->id && ops->size) {
Julian Anastasovb9229342012-04-16 04:43:15 +000099 data = kzalloc(ops->size, GFP_KERNEL);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000100 if (!data)
Julian Anastasovb9229342012-04-16 04:43:15 +0000101 goto out;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000102
103 err = net_assign_generic(net, *ops->id, data);
Julian Anastasovb9229342012-04-16 04:43:15 +0000104 if (err)
105 goto cleanup;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000106 }
Julian Anastasovb9229342012-04-16 04:43:15 +0000107 err = 0;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000108 if (ops->init)
Julian Anastasovb9229342012-04-16 04:43:15 +0000109 err = ops->init(net);
110 if (!err)
111 return 0;
112
113cleanup:
114 kfree(data);
115
116out:
117 return err;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000118}
119
120static void ops_free(const struct pernet_operations *ops, struct net *net)
121{
122 if (ops->id && ops->size) {
123 int id = *ops->id;
124 kfree(net_generic(net, id));
125 }
126}
127
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000128static void ops_exit_list(const struct pernet_operations *ops,
129 struct list_head *net_exit_list)
130{
131 struct net *net;
132 if (ops->exit) {
133 list_for_each_entry(net, net_exit_list, exit_list)
134 ops->exit(net);
135 }
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000136 if (ops->exit_batch)
137 ops->exit_batch(net_exit_list);
138}
139
140static void ops_free_list(const struct pernet_operations *ops,
141 struct list_head *net_exit_list)
142{
143 struct net *net;
144 if (ops->size && ops->id) {
145 list_for_each_entry(net, net_exit_list, exit_list)
146 ops_free(ops, net);
147 }
148}
149
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200150/* should be called with nsid_lock held */
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100151static int alloc_netid(struct net *net, struct net *peer, int reqid)
152{
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200153 int min = 0, max = 0;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100154
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100155 if (reqid >= 0) {
156 min = reqid;
157 max = reqid + 1;
158 }
159
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200160 return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100161}
162
163/* This function is used by idr_for_each(). If net is equal to peer, the
164 * function returns the id so that idr_for_each() stops. Because we cannot
165 * returns the id 0 (idr_for_each() will not stop), we return the magic value
166 * NET_ID_ZERO (-1) for it.
167 */
168#define NET_ID_ZERO -1
169static int net_eq_idr(int id, void *net, void *peer)
170{
171 if (net_eq(net, peer))
172 return id ? : NET_ID_ZERO;
173 return 0;
174}
175
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200176/* Should be called with nsid_lock held. If a new id is assigned, the bool alloc
177 * is set to true, thus the caller knows that the new id must be notified via
178 * rtnl.
179 */
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200180static int __peernet2id_alloc(struct net *net, struct net *peer, bool *alloc)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100181{
182 int id = idr_for_each(&net->netns_ids, net_eq_idr, peer);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200183 bool alloc_it = *alloc;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100184
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200185 *alloc = false;
186
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100187 /* Magic value for id 0. */
188 if (id == NET_ID_ZERO)
189 return 0;
190 if (id > 0)
191 return id;
192
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200193 if (alloc_it) {
Nicolas Dichtel109582af2015-05-07 11:02:47 +0200194 id = alloc_netid(net, peer, -1);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200195 *alloc = true;
Nicolas Dichtel109582af2015-05-07 11:02:47 +0200196 return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED;
197 }
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100198
Nicolas Dichtel109582af2015-05-07 11:02:47 +0200199 return NETNSA_NSID_NOT_ASSIGNED;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100200}
201
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200202/* should be called with nsid_lock held */
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200203static int __peernet2id(struct net *net, struct net *peer)
204{
205 bool no = false;
206
207 return __peernet2id_alloc(net, peer, &no);
208}
209
210static void rtnl_net_notifyid(struct net *net, int cmd, int id);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100211/* This function returns the id of a peer netns. If no id is assigned, one will
212 * be allocated and returned.
213 */
Nicolas Dichtel7a0877d2015-05-07 11:02:49 +0200214int peernet2id_alloc(struct net *net, struct net *peer)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100215{
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200216 unsigned long flags;
217 bool alloc;
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200218 int id;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100219
WANG Congde133462015-05-15 14:47:32 -0700220 spin_lock_irqsave(&net->nsid_lock, flags);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200221 alloc = atomic_read(&peer->count) == 0 ? false : true;
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200222 id = __peernet2id_alloc(net, peer, &alloc);
WANG Congde133462015-05-15 14:47:32 -0700223 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200224 if (alloc && id >= 0)
225 rtnl_net_notifyid(net, RTM_NEWNSID, id);
226 return id;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100227}
Nicolas Dichtel7a0877d2015-05-07 11:02:49 +0200228EXPORT_SYMBOL(peernet2id_alloc);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100229
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200230/* This function returns, if assigned, the id of a peer netns. */
Nicolas Dichtel59324cf2015-05-07 11:02:53 +0200231int peernet2id(struct net *net, struct net *peer)
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200232{
233 unsigned long flags;
234 int id;
235
WANG Congde133462015-05-15 14:47:32 -0700236 spin_lock_irqsave(&net->nsid_lock, flags);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200237 id = __peernet2id(net, peer);
WANG Congde133462015-05-15 14:47:32 -0700238 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200239 return id;
240}
241
Nicolas Dichtel59324cf2015-05-07 11:02:53 +0200242/* This function returns true is the peer netns has an id assigned into the
243 * current netns.
244 */
245bool peernet_has_id(struct net *net, struct net *peer)
246{
247 return peernet2id(net, peer) >= 0;
248}
249
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100250struct net *get_net_ns_by_id(struct net *net, int id)
251{
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200252 unsigned long flags;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100253 struct net *peer;
254
255 if (id < 0)
256 return NULL;
257
258 rcu_read_lock();
WANG Congde133462015-05-15 14:47:32 -0700259 spin_lock_irqsave(&net->nsid_lock, flags);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100260 peer = idr_find(&net->netns_ids, id);
261 if (peer)
262 get_net(peer);
WANG Congde133462015-05-15 14:47:32 -0700263 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100264 rcu_read_unlock();
265
266 return peer;
267}
268
Eric W. Biederman70328662016-08-08 14:33:23 -0500269static struct ucounts *inc_net_namespaces(struct user_namespace *ns)
270{
271 return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES);
272}
273
274static void dec_net_namespaces(struct ucounts *ucounts)
275{
276 dec_ucount(ucounts, UCOUNT_NET_NAMESPACES);
277}
278
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700279/*
280 * setup_net runs the initializers for the network namespace object.
281 */
Eric W. Biederman038e7332012-06-14 02:31:10 -0700282static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700283{
284 /* Must be called with net_mutex held */
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000285 const struct pernet_operations *ops, *saved_ops;
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800286 int error = 0;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000287 LIST_HEAD(net_exit_list);
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700288
289 atomic_set(&net->count, 1);
Al Viroa685e082011-06-08 21:13:01 -0400290 atomic_set(&net->passive, 1);
Thomas Graf4e985ad2011-06-21 03:11:20 +0000291 net->dev_base_seq = 1;
Eric W. Biederman038e7332012-06-14 02:31:10 -0700292 net->user_ns = user_ns;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100293 idr_init(&net->netns_ids);
WANG Congde133462015-05-15 14:47:32 -0700294 spin_lock_init(&net->nsid_lock);
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800295
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700296 list_for_each_entry(ops, &pernet_list, list) {
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000297 error = ops_init(ops, net);
298 if (error < 0)
299 goto out_undo;
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700300 }
301out:
302 return error;
303
304out_undo:
305 /* Walk through the list backwards calling the exit functions
306 * for the pernet modules whose init functions did not fail.
307 */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000308 list_add(&net->exit_list, &net_exit_list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000309 saved_ops = ops;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000310 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
311 ops_exit_list(ops, &net_exit_list);
312
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000313 ops = saved_ops;
314 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000315 ops_free_list(ops, &net_exit_list);
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700316
317 rcu_barrier();
318 goto out;
319}
320
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800321
Clemens Nossebe47d42009-02-23 15:37:35 -0800322#ifdef CONFIG_NET_NS
323static struct kmem_cache *net_cachep;
324static struct workqueue_struct *netns_wq;
325
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200326static struct net *net_alloc(void)
327{
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800328 struct net *net = NULL;
329 struct net_generic *ng;
330
331 ng = net_alloc_generic();
332 if (!ng)
333 goto out;
334
335 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL);
336 if (!net)
337 goto out_free;
338
339 rcu_assign_pointer(net->gen, ng);
340out:
341 return net;
342
343out_free:
344 kfree(ng);
345 goto out;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200346}
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200347
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800348static void net_free(struct net *net)
349{
Eric Dumazet416c51e2014-09-09 08:24:53 -0700350 kfree(rcu_access_pointer(net->gen));
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800351 kmem_cache_free(net_cachep, net);
352}
353
Al Viroa685e082011-06-08 21:13:01 -0400354void net_drop_ns(void *p)
355{
356 struct net *ns = p;
357 if (ns && atomic_dec_and_test(&ns->passive))
358 net_free(ns);
359}
360
Eric W. Biederman038e7332012-06-14 02:31:10 -0700361struct net *copy_net_ns(unsigned long flags,
362 struct user_namespace *user_ns, struct net *old_net)
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700363{
Eric W. Biederman70328662016-08-08 14:33:23 -0500364 struct ucounts *ucounts;
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700365 struct net *net;
366 int rv;
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700367
Rob Landley911cb192011-04-15 02:26:25 +0000368 if (!(flags & CLONE_NEWNET))
369 return get_net(old_net);
370
Eric W. Biederman70328662016-08-08 14:33:23 -0500371 ucounts = inc_net_namespaces(user_ns);
372 if (!ucounts)
Eric W. Biedermandf75e772016-09-22 13:08:36 -0500373 return ERR_PTR(-ENOSPC);
Eric W. Biederman70328662016-08-08 14:33:23 -0500374
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700375 net = net_alloc();
Eric W. Biederman70328662016-08-08 14:33:23 -0500376 if (!net) {
377 dec_net_namespaces(ucounts);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700378 return ERR_PTR(-ENOMEM);
Eric W. Biederman70328662016-08-08 14:33:23 -0500379 }
Eric W. Biederman038e7332012-06-14 02:31:10 -0700380
381 get_user_ns(user_ns);
382
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700383 mutex_lock(&net_mutex);
Eric W. Biederman70328662016-08-08 14:33:23 -0500384 net->ucounts = ucounts;
Eric W. Biederman038e7332012-06-14 02:31:10 -0700385 rv = setup_net(net, user_ns);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700386 if (rv == 0) {
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800387 rtnl_lock();
Johannes Berg11a28d32009-07-10 09:51:33 +0000388 list_add_tail_rcu(&net->list, &net_namespace_list);
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800389 rtnl_unlock();
390 }
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700391 mutex_unlock(&net_mutex);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700392 if (rv < 0) {
Eric W. Biederman70328662016-08-08 14:33:23 -0500393 dec_net_namespaces(ucounts);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700394 put_user_ns(user_ns);
Al Viroa685e082011-06-08 21:13:01 -0400395 net_drop_ns(net);
Alexey Dobriyan088eb2d2009-05-04 11:12:14 -0700396 return ERR_PTR(rv);
397 }
398 return net;
399}
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800400
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000401static DEFINE_SPINLOCK(cleanup_list_lock);
402static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */
403
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200404static void cleanup_net(struct work_struct *work)
405{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000406 const struct pernet_operations *ops;
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200407 struct net *net, *tmp;
xiao jin1818ce42014-04-25 08:50:54 +0800408 struct list_head net_kill_list;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000409 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200410
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000411 /* Atomically snapshot the list of namespaces to cleanup */
412 spin_lock_irq(&cleanup_list_lock);
413 list_replace_init(&cleanup_list, &net_kill_list);
414 spin_unlock_irq(&cleanup_list_lock);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200415
416 mutex_lock(&net_mutex);
417
418 /* Don't let anyone else find us. */
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700419 rtnl_lock();
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000420 list_for_each_entry(net, &net_kill_list, cleanup_list) {
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000421 list_del_rcu(&net->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000422 list_add_tail(&net->exit_list, &net_exit_list);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200423 for_each_net(tmp) {
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200424 int id;
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200425
WANG Congde133462015-05-15 14:47:32 -0700426 spin_lock_irq(&tmp->nsid_lock);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200427 id = __peernet2id(tmp, net);
428 if (id >= 0)
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200429 idr_remove(&tmp->netns_ids, id);
WANG Congde133462015-05-15 14:47:32 -0700430 spin_unlock_irq(&tmp->nsid_lock);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200431 if (id >= 0)
432 rtnl_net_notifyid(tmp, RTM_DELNSID, id);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200433 }
WANG Congde133462015-05-15 14:47:32 -0700434 spin_lock_irq(&net->nsid_lock);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200435 idr_destroy(&net->netns_ids);
WANG Congde133462015-05-15 14:47:32 -0700436 spin_unlock_irq(&net->nsid_lock);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200437
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000438 }
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700439 rtnl_unlock();
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200440
Johannes Berg11a28d32009-07-10 09:51:33 +0000441 /*
442 * Another CPU might be rcu-iterating the list, wait for it.
443 * This needs to be before calling the exit() notifiers, so
444 * the rcu_barrier() below isn't sufficient alone.
445 */
446 synchronize_rcu();
447
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200448 /* Run all of the network namespace exit methods */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000449 list_for_each_entry_reverse(ops, &pernet_list, list)
450 ops_exit_list(ops, &net_exit_list);
451
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000452 /* Free the net generic variables */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000453 list_for_each_entry_reverse(ops, &pernet_list, list)
454 ops_free_list(ops, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200455
456 mutex_unlock(&net_mutex);
457
458 /* Ensure there are no outstanding rcu callbacks using this
459 * network namespace.
460 */
461 rcu_barrier();
462
463 /* Finally it is safe to free my network namespace structure */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000464 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
465 list_del_init(&net->exit_list);
Eric W. Biederman70328662016-08-08 14:33:23 -0500466 dec_net_namespaces(net->ucounts);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700467 put_user_ns(net->user_ns);
Al Viroa685e082011-06-08 21:13:01 -0400468 net_drop_ns(net);
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000469 }
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200470}
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000471static DECLARE_WORK(net_cleanup_work, cleanup_net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200472
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200473void __put_net(struct net *net)
474{
475 /* Cleanup the network namespace in process context */
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000476 unsigned long flags;
477
478 spin_lock_irqsave(&cleanup_list_lock, flags);
479 list_add(&net->cleanup_list, &cleanup_list);
480 spin_unlock_irqrestore(&cleanup_list_lock, flags);
481
482 queue_work(netns_wq, &net_cleanup_work);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200483}
484EXPORT_SYMBOL_GPL(__put_net);
485
Stephen Rothwell956c9202011-05-12 13:51:13 +1000486struct net *get_net_ns_by_fd(int fd)
487{
Stephen Rothwell956c9202011-05-12 13:51:13 +1000488 struct file *file;
Al Viro33c42942014-11-01 02:32:53 -0400489 struct ns_common *ns;
Stephen Rothwell956c9202011-05-12 13:51:13 +1000490 struct net *net;
491
Stephen Rothwell956c9202011-05-12 13:51:13 +1000492 file = proc_ns_fget(fd);
Al Viroc316e6a2011-06-05 00:37:35 +0000493 if (IS_ERR(file))
494 return ERR_CAST(file);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000495
Al Virof77c8012014-11-01 03:13:17 -0400496 ns = get_proc_ns(file_inode(file));
Al Viro33c42942014-11-01 02:32:53 -0400497 if (ns->ops == &netns_operations)
498 net = get_net(container_of(ns, struct net, ns));
Al Viroc316e6a2011-06-05 00:37:35 +0000499 else
500 net = ERR_PTR(-EINVAL);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000501
Al Viroc316e6a2011-06-05 00:37:35 +0000502 fput(file);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000503 return net;
504}
505
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700506#else
Stephen Rothwell956c9202011-05-12 13:51:13 +1000507struct net *get_net_ns_by_fd(int fd)
508{
509 return ERR_PTR(-EINVAL);
510}
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700511#endif
Vadim Kochan4b681c82015-01-12 16:34:05 +0200512EXPORT_SYMBOL_GPL(get_net_ns_by_fd);
Eric W. Biederman9dd776b2007-09-26 22:04:26 -0700513
Johannes Berg30ffee82009-07-10 09:51:35 +0000514struct net *get_net_ns_by_pid(pid_t pid)
515{
516 struct task_struct *tsk;
517 struct net *net;
518
519 /* Lookup the network namespace */
520 net = ERR_PTR(-ESRCH);
521 rcu_read_lock();
522 tsk = find_task_by_vpid(pid);
523 if (tsk) {
524 struct nsproxy *nsproxy;
Eric W. Biederman728dba32014-02-03 19:13:49 -0800525 task_lock(tsk);
526 nsproxy = tsk->nsproxy;
Johannes Berg30ffee82009-07-10 09:51:35 +0000527 if (nsproxy)
528 net = get_net(nsproxy->net_ns);
Eric W. Biederman728dba32014-02-03 19:13:49 -0800529 task_unlock(tsk);
Johannes Berg30ffee82009-07-10 09:51:35 +0000530 }
531 rcu_read_unlock();
532 return net;
533}
534EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
535
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700536static __net_init int net_ns_net_init(struct net *net)
537{
Al Viro33c42942014-11-01 02:32:53 -0400538#ifdef CONFIG_NET_NS
539 net->ns.ops = &netns_operations;
540#endif
Al Viro6344c432014-11-01 00:45:45 -0400541 return ns_alloc_inum(&net->ns);
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700542}
543
544static __net_exit void net_ns_net_exit(struct net *net)
545{
Al Viro6344c432014-11-01 00:45:45 -0400546 ns_free_inum(&net->ns);
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700547}
548
549static struct pernet_operations __net_initdata net_ns_ops = {
550 .init = net_ns_net_init,
551 .exit = net_ns_net_exit,
552};
553
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100554static struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = {
555 [NETNSA_NONE] = { .type = NLA_UNSPEC },
556 [NETNSA_NSID] = { .type = NLA_S32 },
557 [NETNSA_PID] = { .type = NLA_U32 },
558 [NETNSA_FD] = { .type = NLA_U32 },
559};
560
561static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh)
562{
563 struct net *net = sock_net(skb->sk);
564 struct nlattr *tb[NETNSA_MAX + 1];
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200565 unsigned long flags;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100566 struct net *peer;
567 int nsid, err;
568
569 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
570 rtnl_net_policy);
571 if (err < 0)
572 return err;
573 if (!tb[NETNSA_NSID])
574 return -EINVAL;
575 nsid = nla_get_s32(tb[NETNSA_NSID]);
576
577 if (tb[NETNSA_PID])
578 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
579 else if (tb[NETNSA_FD])
580 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
581 else
582 return -EINVAL;
583 if (IS_ERR(peer))
584 return PTR_ERR(peer);
585
WANG Congde133462015-05-15 14:47:32 -0700586 spin_lock_irqsave(&net->nsid_lock, flags);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200587 if (__peernet2id(net, peer) >= 0) {
WANG Congde133462015-05-15 14:47:32 -0700588 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100589 err = -EEXIST;
590 goto out;
591 }
592
593 err = alloc_netid(net, peer, nsid);
WANG Congde133462015-05-15 14:47:32 -0700594 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200595 if (err >= 0) {
596 rtnl_net_notifyid(net, RTM_NEWNSID, err);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100597 err = 0;
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200598 }
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100599out:
600 put_net(peer);
601 return err;
602}
603
604static int rtnl_net_get_size(void)
605{
606 return NLMSG_ALIGN(sizeof(struct rtgenmsg))
607 + nla_total_size(sizeof(s32)) /* NETNSA_NSID */
608 ;
609}
610
611static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags,
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200612 int cmd, struct net *net, int nsid)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100613{
614 struct nlmsghdr *nlh;
615 struct rtgenmsg *rth;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100616
617 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags);
618 if (!nlh)
619 return -EMSGSIZE;
620
621 rth = nlmsg_data(nlh);
622 rth->rtgen_family = AF_UNSPEC;
623
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200624 if (nla_put_s32(skb, NETNSA_NSID, nsid))
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100625 goto nla_put_failure;
626
627 nlmsg_end(skb, nlh);
628 return 0;
629
630nla_put_failure:
631 nlmsg_cancel(skb, nlh);
632 return -EMSGSIZE;
633}
634
635static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh)
636{
637 struct net *net = sock_net(skb->sk);
638 struct nlattr *tb[NETNSA_MAX + 1];
639 struct sk_buff *msg;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100640 struct net *peer;
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200641 int err, id;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100642
643 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
644 rtnl_net_policy);
645 if (err < 0)
646 return err;
647 if (tb[NETNSA_PID])
648 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
649 else if (tb[NETNSA_FD])
650 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
651 else
652 return -EINVAL;
653
654 if (IS_ERR(peer))
655 return PTR_ERR(peer);
656
657 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
658 if (!msg) {
659 err = -ENOMEM;
660 goto out;
661 }
662
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200663 id = peernet2id(net, peer);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100664 err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
David S. Millerb04096f2015-05-13 14:31:43 -0400665 RTM_NEWNSID, net, id);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100666 if (err < 0)
667 goto err_out;
668
669 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid);
670 goto out;
671
672err_out:
673 nlmsg_free(msg);
674out:
675 put_net(peer);
676 return err;
677}
678
Nicolas Dichtela143c402015-04-07 11:51:54 +0200679struct rtnl_net_dump_cb {
680 struct net *net;
681 struct sk_buff *skb;
682 struct netlink_callback *cb;
683 int idx;
684 int s_idx;
685};
686
687static int rtnl_net_dumpid_one(int id, void *peer, void *data)
688{
689 struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data;
690 int ret;
691
692 if (net_cb->idx < net_cb->s_idx)
693 goto cont;
694
695 ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid,
696 net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI,
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200697 RTM_NEWNSID, net_cb->net, id);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200698 if (ret < 0)
699 return ret;
700
701cont:
702 net_cb->idx++;
703 return 0;
704}
705
706static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb)
707{
708 struct net *net = sock_net(skb->sk);
709 struct rtnl_net_dump_cb net_cb = {
710 .net = net,
711 .skb = skb,
712 .cb = cb,
713 .idx = 0,
714 .s_idx = cb->args[0],
715 };
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200716 unsigned long flags;
Nicolas Dichtela143c402015-04-07 11:51:54 +0200717
WANG Congde133462015-05-15 14:47:32 -0700718 spin_lock_irqsave(&net->nsid_lock, flags);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200719 idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb);
WANG Congde133462015-05-15 14:47:32 -0700720 spin_unlock_irqrestore(&net->nsid_lock, flags);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200721
722 cb->args[0] = net_cb.idx;
723 return skb->len;
724}
725
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200726static void rtnl_net_notifyid(struct net *net, int cmd, int id)
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200727{
728 struct sk_buff *msg;
729 int err = -ENOMEM;
730
731 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
732 if (!msg)
733 goto out;
734
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200735 err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, id);
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200736 if (err < 0)
737 goto err_out;
738
739 rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0);
740 return;
741
742err_out:
743 nlmsg_free(msg);
744out:
745 rtnl_set_sk_err(net, RTNLGRP_NSID, err);
746}
747
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200748static int __init net_ns_init(void)
749{
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800750 struct net_generic *ng;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200751
Pavel Emelyanovd57a9212007-11-01 00:46:50 -0700752#ifdef CONFIG_NET_NS
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200753 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net),
754 SMP_CACHE_BYTES,
755 SLAB_PANIC, NULL);
Benjamin Thery3ef13552007-11-19 23:18:16 -0800756
757 /* Create workqueue for cleanup */
758 netns_wq = create_singlethread_workqueue("netns");
759 if (!netns_wq)
760 panic("Could not create netns workq");
Pavel Emelyanovd57a9212007-11-01 00:46:50 -0700761#endif
Benjamin Thery3ef13552007-11-19 23:18:16 -0800762
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800763 ng = net_alloc_generic();
764 if (!ng)
765 panic("Could not allocate generic netns");
766
767 rcu_assign_pointer(init_net.gen, ng);
768
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200769 mutex_lock(&net_mutex);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700770 if (setup_net(&init_net, &init_user_ns))
Stephen Hemmingerca0f3112009-05-21 15:10:31 -0700771 panic("Could not setup the initial network namespace");
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200772
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700773 rtnl_lock();
Johannes Berg11a28d32009-07-10 09:51:33 +0000774 list_add_tail_rcu(&init_net.list, &net_namespace_list);
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700775 rtnl_unlock();
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200776
777 mutex_unlock(&net_mutex);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200778
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700779 register_pernet_subsys(&net_ns_ops);
780
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100781 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200782 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid,
783 NULL);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100784
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200785 return 0;
786}
787
788pure_initcall(net_ns_init);
789
Denis V. Luneved160e82007-11-13 03:23:21 -0800790#ifdef CONFIG_NET_NS
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000791static int __register_pernet_operations(struct list_head *list,
792 struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200793{
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000794 struct net *net;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200795 int error;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000796 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200797
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200798 list_add_tail(&ops->list, list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000799 if (ops->init || (ops->id && ops->size)) {
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700800 for_each_net(net) {
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000801 error = ops_init(ops, net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200802 if (error)
803 goto out_undo;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000804 list_add_tail(&net->exit_list, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200805 }
806 }
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700807 return 0;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200808
809out_undo:
810 /* If I have an error cleanup all namespaces I initialized */
811 list_del(&ops->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000812 ops_exit_list(ops, &net_exit_list);
813 ops_free_list(ops, &net_exit_list);
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700814 return error;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200815}
816
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000817static void __unregister_pernet_operations(struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200818{
819 struct net *net;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000820 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200821
822 list_del(&ops->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000823 for_each_net(net)
824 list_add_tail(&net->exit_list, &net_exit_list);
825 ops_exit_list(ops, &net_exit_list);
826 ops_free_list(ops, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200827}
828
Denis V. Luneved160e82007-11-13 03:23:21 -0800829#else
830
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000831static int __register_pernet_operations(struct list_head *list,
832 struct pernet_operations *ops)
833{
Julian Anastasovb9229342012-04-16 04:43:15 +0000834 return ops_init(ops, &init_net);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000835}
836
837static void __unregister_pernet_operations(struct pernet_operations *ops)
838{
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000839 LIST_HEAD(net_exit_list);
840 list_add(&init_net.exit_list, &net_exit_list);
841 ops_exit_list(ops, &net_exit_list);
842 ops_free_list(ops, &net_exit_list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000843}
844
845#endif /* CONFIG_NET_NS */
846
847static DEFINE_IDA(net_generic_ids);
848
Denis V. Luneved160e82007-11-13 03:23:21 -0800849static int register_pernet_operations(struct list_head *list,
850 struct pernet_operations *ops)
851{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000852 int error;
853
854 if (ops->id) {
855again:
856 error = ida_get_new_above(&net_generic_ids, 1, ops->id);
857 if (error < 0) {
858 if (error == -EAGAIN) {
859 ida_pre_get(&net_generic_ids, GFP_KERNEL);
860 goto again;
861 }
862 return error;
863 }
Eric Dumazet073862b2012-01-26 00:41:38 +0000864 max_gen_ptrs = max_t(unsigned int, max_gen_ptrs, *ops->id);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000865 }
866 error = __register_pernet_operations(list, ops);
Eric W. Biederman3a765ed2009-12-03 02:29:06 +0000867 if (error) {
868 rcu_barrier();
869 if (ops->id)
870 ida_remove(&net_generic_ids, *ops->id);
871 }
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000872
873 return error;
Denis V. Luneved160e82007-11-13 03:23:21 -0800874}
875
876static void unregister_pernet_operations(struct pernet_operations *ops)
877{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000878
879 __unregister_pernet_operations(ops);
Eric W. Biederman3a765ed2009-12-03 02:29:06 +0000880 rcu_barrier();
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000881 if (ops->id)
882 ida_remove(&net_generic_ids, *ops->id);
Denis V. Luneved160e82007-11-13 03:23:21 -0800883}
Pavel Emelyanovc93cf612008-04-15 00:35:23 -0700884
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200885/**
886 * register_pernet_subsys - register a network namespace subsystem
887 * @ops: pernet operations structure for the subsystem
888 *
889 * Register a subsystem which has init and exit functions
890 * that are called when network namespaces are created and
891 * destroyed respectively.
892 *
893 * When registered all network namespace init functions are
894 * called for every existing network namespace. Allowing kernel
895 * modules to have a race free view of the set of network namespaces.
896 *
897 * When a new network namespace is created all of the init
898 * methods are called in the order in which they were registered.
899 *
900 * When a network namespace is destroyed all of the exit methods
901 * are called in the reverse of the order with which they were
902 * registered.
903 */
904int register_pernet_subsys(struct pernet_operations *ops)
905{
906 int error;
907 mutex_lock(&net_mutex);
908 error = register_pernet_operations(first_device, ops);
909 mutex_unlock(&net_mutex);
910 return error;
911}
912EXPORT_SYMBOL_GPL(register_pernet_subsys);
913
914/**
915 * unregister_pernet_subsys - unregister a network namespace subsystem
916 * @ops: pernet operations structure to manipulate
917 *
918 * Remove the pernet operations structure from the list to be
Oliver Pinter53379e52008-02-03 17:56:48 +0200919 * used when network namespaces are created or destroyed. In
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200920 * addition run the exit method for all existing network
921 * namespaces.
922 */
Jiri Pirkob3c981d2010-04-25 00:49:56 -0700923void unregister_pernet_subsys(struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200924{
925 mutex_lock(&net_mutex);
Jiri Pirkob3c981d2010-04-25 00:49:56 -0700926 unregister_pernet_operations(ops);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200927 mutex_unlock(&net_mutex);
928}
929EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
930
931/**
932 * register_pernet_device - register a network namespace device
933 * @ops: pernet operations structure for the subsystem
934 *
935 * Register a device which has init and exit functions
936 * that are called when network namespaces are created and
937 * destroyed respectively.
938 *
939 * When registered all network namespace init functions are
940 * called for every existing network namespace. Allowing kernel
941 * modules to have a race free view of the set of network namespaces.
942 *
943 * When a new network namespace is created all of the init
944 * methods are called in the order in which they were registered.
945 *
946 * When a network namespace is destroyed all of the exit methods
947 * are called in the reverse of the order with which they were
948 * registered.
949 */
950int register_pernet_device(struct pernet_operations *ops)
951{
952 int error;
953 mutex_lock(&net_mutex);
954 error = register_pernet_operations(&pernet_list, ops);
955 if (!error && (first_device == &pernet_list))
956 first_device = &ops->list;
957 mutex_unlock(&net_mutex);
958 return error;
959}
960EXPORT_SYMBOL_GPL(register_pernet_device);
961
962/**
963 * unregister_pernet_device - unregister a network namespace netdevice
964 * @ops: pernet operations structure to manipulate
965 *
966 * Remove the pernet operations structure from the list to be
Oliver Pinter53379e52008-02-03 17:56:48 +0200967 * used when network namespaces are created or destroyed. In
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200968 * addition run the exit method for all existing network
969 * namespaces.
970 */
971void unregister_pernet_device(struct pernet_operations *ops)
972{
973 mutex_lock(&net_mutex);
974 if (&ops->list == first_device)
975 first_device = first_device->next;
976 unregister_pernet_operations(ops);
977 mutex_unlock(&net_mutex);
978}
979EXPORT_SYMBOL_GPL(unregister_pernet_device);
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800980
981#ifdef CONFIG_NET_NS
Al Viro64964522014-11-01 00:37:32 -0400982static struct ns_common *netns_get(struct task_struct *task)
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800983{
Eric W. Biedermanf0630522011-05-04 17:51:50 -0700984 struct net *net = NULL;
985 struct nsproxy *nsproxy;
986
Eric W. Biederman728dba32014-02-03 19:13:49 -0800987 task_lock(task);
988 nsproxy = task->nsproxy;
Eric W. Biedermanf0630522011-05-04 17:51:50 -0700989 if (nsproxy)
990 net = get_net(nsproxy->net_ns);
Eric W. Biederman728dba32014-02-03 19:13:49 -0800991 task_unlock(task);
Eric W. Biedermanf0630522011-05-04 17:51:50 -0700992
Al Viroff248702014-11-01 00:10:50 -0400993 return net ? &net->ns : NULL;
994}
995
996static inline struct net *to_net_ns(struct ns_common *ns)
997{
998 return container_of(ns, struct net, ns);
Eric W. Biederman13b6f572010-03-07 18:14:23 -0800999}
1000
Al Viro64964522014-11-01 00:37:32 -04001001static void netns_put(struct ns_common *ns)
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001002{
Al Viroff248702014-11-01 00:10:50 -04001003 put_net(to_net_ns(ns));
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001004}
1005
Al Viro64964522014-11-01 00:37:32 -04001006static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns)
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001007{
Al Viroff248702014-11-01 00:10:50 -04001008 struct net *net = to_net_ns(ns);
Eric W. Biederman142e1d12012-07-26 01:13:20 -07001009
Eric W. Biederman5e4a0842012-12-14 07:55:36 -08001010 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -07001011 !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
Eric W. Biederman142e1d12012-07-26 01:13:20 -07001012 return -EPERM;
1013
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001014 put_net(nsproxy->net_ns);
Eric W. Biederman142e1d12012-07-26 01:13:20 -07001015 nsproxy->net_ns = get_net(net);
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001016 return 0;
1017}
1018
1019const struct proc_ns_operations netns_operations = {
1020 .name = "net",
1021 .type = CLONE_NEWNET,
1022 .get = netns_get,
1023 .put = netns_put,
1024 .install = netns_install,
1025};
1026#endif