blob: 1f15abb1d733c702d8dfaf62898db3712bdd90e5 [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>
Ingo Molnarf719ff9b2017-02-06 10:57:33 +010019#include <linux/sched/task.h>
20
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +010021#include <net/sock.h>
22#include <net/netlink.h>
Eric W. Biederman5f256be2007-09-12 11:50:50 +020023#include <net/net_namespace.h>
Pavel Emelyanovdec827d2008-04-15 00:36:08 -070024#include <net/netns/generic.h>
Eric W. Biederman5f256be2007-09-12 11:50:50 +020025
26/*
27 * Our network namespace constructor/destructor lists
28 */
29
30static LIST_HEAD(pernet_list);
31static struct list_head *first_device = &pernet_list;
Cong Wang200b9162014-05-12 15:11:20 -070032DEFINE_MUTEX(net_mutex);
Eric W. Biederman5f256be2007-09-12 11:50:50 +020033
Eric W. Biederman5f256be2007-09-12 11:50:50 +020034LIST_HEAD(net_namespace_list);
Alexey Dobriyanb76a4612008-10-08 11:35:06 +020035EXPORT_SYMBOL_GPL(net_namespace_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +020036
Rustad, Mark D734b6542012-07-18 09:06:07 +000037struct net init_net = {
David Howellsb5082df2017-04-27 22:40:23 +010038 .count = ATOMIC_INIT(1),
39 .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head),
Rustad, Mark D734b6542012-07-18 09:06:07 +000040};
Denis V. Lunevff4b9502008-01-22 22:05:33 -080041EXPORT_SYMBOL(init_net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +020042
Dmitry Torokhovf8c46cb2016-08-10 14:36:00 -070043static bool init_net_initialized;
44
Alexey Dobriyan6af2d5f2016-12-02 04:21:32 +030045#define MIN_PERNET_OPS_ID \
46 ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *))
47
Pavel Emelyanovdec827d2008-04-15 00:36:08 -070048#define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
49
Eric Dumazet073862b2012-01-26 00:41:38 +000050static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS;
51
52static struct net_generic *net_alloc_generic(void)
53{
54 struct net_generic *ng;
Alexey Dobriyan6af2d5f2016-12-02 04:21:32 +030055 unsigned int generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]);
Eric Dumazet073862b2012-01-26 00:41:38 +000056
57 ng = kzalloc(generic_size, GFP_KERNEL);
58 if (ng)
Alexey Dobriyan9bfc7b92016-12-02 04:12:58 +030059 ng->s.len = max_gen_ptrs;
Eric Dumazet073862b2012-01-26 00:41:38 +000060
61 return ng;
62}
63
Alexey Dobriyanc7d03a02016-11-17 04:58:21 +030064static int net_assign_generic(struct net *net, unsigned int id, void *data)
Jiri Pirko05fceb42010-04-23 01:40:47 +000065{
66 struct net_generic *ng, *old_ng;
67
68 BUG_ON(!mutex_is_locked(&net_mutex));
Alexey Dobriyan6af2d5f2016-12-02 04:21:32 +030069 BUG_ON(id < MIN_PERNET_OPS_ID);
Jiri Pirko05fceb42010-04-23 01:40:47 +000070
Eric Dumazet1c877332010-10-25 03:20:11 +000071 old_ng = rcu_dereference_protected(net->gen,
72 lockdep_is_held(&net_mutex));
Alexey Dobriyan6af2d5f2016-12-02 04:21:32 +030073 if (old_ng->s.len > id) {
74 old_ng->ptr[id] = data;
Alexey Dobriyan1a9a05922016-12-02 04:11:34 +030075 return 0;
76 }
Jiri Pirko05fceb42010-04-23 01:40:47 +000077
Eric Dumazet073862b2012-01-26 00:41:38 +000078 ng = net_alloc_generic();
Jiri Pirko05fceb42010-04-23 01:40:47 +000079 if (ng == NULL)
80 return -ENOMEM;
81
82 /*
83 * Some synchronisation notes:
84 *
85 * The net_generic explores the net->gen array inside rcu
86 * read section. Besides once set the net->gen->ptr[x]
87 * pointer never changes (see rules in netns/generic.h).
88 *
89 * That said, we simply duplicate this array and schedule
90 * the old copy for kfree after a grace period.
91 */
92
Alexey Dobriyan6af2d5f2016-12-02 04:21:32 +030093 memcpy(&ng->ptr[MIN_PERNET_OPS_ID], &old_ng->ptr[MIN_PERNET_OPS_ID],
94 (old_ng->s.len - MIN_PERNET_OPS_ID) * sizeof(void *));
95 ng->ptr[id] = data;
Jiri Pirko05fceb42010-04-23 01:40:47 +000096
97 rcu_assign_pointer(net->gen, ng);
Alexey Dobriyan9bfc7b92016-12-02 04:12:58 +030098 kfree_rcu(old_ng, s.rcu);
Jiri Pirko05fceb42010-04-23 01:40:47 +000099 return 0;
100}
101
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000102static int ops_init(const struct pernet_operations *ops, struct net *net)
103{
Julian Anastasovb9229342012-04-16 04:43:15 +0000104 int err = -ENOMEM;
105 void *data = NULL;
106
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000107 if (ops->id && ops->size) {
Julian Anastasovb9229342012-04-16 04:43:15 +0000108 data = kzalloc(ops->size, GFP_KERNEL);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000109 if (!data)
Julian Anastasovb9229342012-04-16 04:43:15 +0000110 goto out;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000111
112 err = net_assign_generic(net, *ops->id, data);
Julian Anastasovb9229342012-04-16 04:43:15 +0000113 if (err)
114 goto cleanup;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000115 }
Julian Anastasovb9229342012-04-16 04:43:15 +0000116 err = 0;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000117 if (ops->init)
Julian Anastasovb9229342012-04-16 04:43:15 +0000118 err = ops->init(net);
119 if (!err)
120 return 0;
121
122cleanup:
123 kfree(data);
124
125out:
126 return err;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000127}
128
129static void ops_free(const struct pernet_operations *ops, struct net *net)
130{
131 if (ops->id && ops->size) {
Alexey Dobriyanc7d03a02016-11-17 04:58:21 +0300132 kfree(net_generic(net, *ops->id));
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000133 }
134}
135
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000136static void ops_exit_list(const struct pernet_operations *ops,
137 struct list_head *net_exit_list)
138{
139 struct net *net;
140 if (ops->exit) {
141 list_for_each_entry(net, net_exit_list, exit_list)
142 ops->exit(net);
143 }
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000144 if (ops->exit_batch)
145 ops->exit_batch(net_exit_list);
146}
147
148static void ops_free_list(const struct pernet_operations *ops,
149 struct list_head *net_exit_list)
150{
151 struct net *net;
152 if (ops->size && ops->id) {
153 list_for_each_entry(net, net_exit_list, exit_list)
154 ops_free(ops, net);
155 }
156}
157
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200158/* should be called with nsid_lock held */
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100159static int alloc_netid(struct net *net, struct net *peer, int reqid)
160{
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200161 int min = 0, max = 0;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100162
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100163 if (reqid >= 0) {
164 min = reqid;
165 max = reqid + 1;
166 }
167
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200168 return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100169}
170
171/* This function is used by idr_for_each(). If net is equal to peer, the
172 * function returns the id so that idr_for_each() stops. Because we cannot
173 * returns the id 0 (idr_for_each() will not stop), we return the magic value
174 * NET_ID_ZERO (-1) for it.
175 */
176#define NET_ID_ZERO -1
177static int net_eq_idr(int id, void *net, void *peer)
178{
179 if (net_eq(net, peer))
180 return id ? : NET_ID_ZERO;
181 return 0;
182}
183
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200184/* Should be called with nsid_lock held. If a new id is assigned, the bool alloc
185 * is set to true, thus the caller knows that the new id must be notified via
186 * rtnl.
187 */
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200188static int __peernet2id_alloc(struct net *net, struct net *peer, bool *alloc)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100189{
190 int id = idr_for_each(&net->netns_ids, net_eq_idr, peer);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200191 bool alloc_it = *alloc;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100192
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200193 *alloc = false;
194
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100195 /* Magic value for id 0. */
196 if (id == NET_ID_ZERO)
197 return 0;
198 if (id > 0)
199 return id;
200
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200201 if (alloc_it) {
Nicolas Dichtel109582af2015-05-07 11:02:47 +0200202 id = alloc_netid(net, peer, -1);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200203 *alloc = true;
Nicolas Dichtel109582af2015-05-07 11:02:47 +0200204 return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED;
205 }
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100206
Nicolas Dichtel109582af2015-05-07 11:02:47 +0200207 return NETNSA_NSID_NOT_ASSIGNED;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100208}
209
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200210/* should be called with nsid_lock held */
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200211static int __peernet2id(struct net *net, struct net *peer)
212{
213 bool no = false;
214
215 return __peernet2id_alloc(net, peer, &no);
216}
217
218static void rtnl_net_notifyid(struct net *net, int cmd, int id);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100219/* This function returns the id of a peer netns. If no id is assigned, one will
220 * be allocated and returned.
221 */
Nicolas Dichtel7a0877d2015-05-07 11:02:49 +0200222int peernet2id_alloc(struct net *net, struct net *peer)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100223{
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200224 bool alloc;
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200225 int id;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100226
WANG Congcfc44a42016-11-16 10:27:02 -0800227 if (atomic_read(&net->count) == 0)
228 return NETNSA_NSID_NOT_ASSIGNED;
Paul Moorefba143c2016-11-29 16:57:48 -0500229 spin_lock_bh(&net->nsid_lock);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200230 alloc = atomic_read(&peer->count) == 0 ? false : true;
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200231 id = __peernet2id_alloc(net, peer, &alloc);
Paul Moorefba143c2016-11-29 16:57:48 -0500232 spin_unlock_bh(&net->nsid_lock);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200233 if (alloc && id >= 0)
234 rtnl_net_notifyid(net, RTM_NEWNSID, id);
235 return id;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100236}
237
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200238/* This function returns, if assigned, the id of a peer netns. */
Nicolas Dichtel59324cf2015-05-07 11:02:53 +0200239int peernet2id(struct net *net, struct net *peer)
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200240{
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200241 int id;
242
Paul Moorefba143c2016-11-29 16:57:48 -0500243 spin_lock_bh(&net->nsid_lock);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200244 id = __peernet2id(net, peer);
Paul Moorefba143c2016-11-29 16:57:48 -0500245 spin_unlock_bh(&net->nsid_lock);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200246 return id;
247}
WANG Cong38f507f2016-09-01 21:53:44 -0700248EXPORT_SYMBOL(peernet2id);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200249
Nicolas Dichtel59324cf2015-05-07 11:02:53 +0200250/* This function returns true is the peer netns has an id assigned into the
251 * current netns.
252 */
253bool peernet_has_id(struct net *net, struct net *peer)
254{
255 return peernet2id(net, peer) >= 0;
256}
257
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100258struct net *get_net_ns_by_id(struct net *net, int id)
259{
260 struct net *peer;
261
262 if (id < 0)
263 return NULL;
264
265 rcu_read_lock();
Paul Moorefba143c2016-11-29 16:57:48 -0500266 spin_lock_bh(&net->nsid_lock);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100267 peer = idr_find(&net->netns_ids, id);
268 if (peer)
269 get_net(peer);
Paul Moorefba143c2016-11-29 16:57:48 -0500270 spin_unlock_bh(&net->nsid_lock);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100271 rcu_read_unlock();
272
273 return peer;
274}
275
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700276/*
277 * setup_net runs the initializers for the network namespace object.
278 */
Eric W. Biederman038e7332012-06-14 02:31:10 -0700279static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700280{
281 /* Must be called with net_mutex held */
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000282 const struct pernet_operations *ops, *saved_ops;
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800283 int error = 0;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000284 LIST_HEAD(net_exit_list);
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700285
286 atomic_set(&net->count, 1);
Al Viroa685e082011-06-08 21:13:01 -0400287 atomic_set(&net->passive, 1);
Thomas Graf4e985ad2011-06-21 03:11:20 +0000288 net->dev_base_seq = 1;
Eric W. Biederman038e7332012-06-14 02:31:10 -0700289 net->user_ns = user_ns;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100290 idr_init(&net->netns_ids);
WANG Congde133462015-05-15 14:47:32 -0700291 spin_lock_init(&net->nsid_lock);
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800292
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700293 list_for_each_entry(ops, &pernet_list, list) {
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000294 error = ops_init(ops, net);
295 if (error < 0)
296 goto out_undo;
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700297 }
298out:
299 return error;
300
301out_undo:
302 /* Walk through the list backwards calling the exit functions
303 * for the pernet modules whose init functions did not fail.
304 */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000305 list_add(&net->exit_list, &net_exit_list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000306 saved_ops = ops;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000307 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
308 ops_exit_list(ops, &net_exit_list);
309
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000310 ops = saved_ops;
311 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000312 ops_free_list(ops, &net_exit_list);
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700313
314 rcu_barrier();
315 goto out;
316}
317
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800318
Clemens Nossebe47d472009-02-23 15:37:35 -0800319#ifdef CONFIG_NET_NS
Arnd Bergmann2ed6afd2016-09-23 18:06:12 +0200320static struct ucounts *inc_net_namespaces(struct user_namespace *ns)
321{
322 return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES);
323}
324
325static void dec_net_namespaces(struct ucounts *ucounts)
326{
327 dec_ucount(ucounts, UCOUNT_NET_NAMESPACES);
328}
329
Clemens Nossebe47d472009-02-23 15:37:35 -0800330static struct kmem_cache *net_cachep;
331static struct workqueue_struct *netns_wq;
332
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200333static struct net *net_alloc(void)
334{
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800335 struct net *net = NULL;
336 struct net_generic *ng;
337
338 ng = net_alloc_generic();
339 if (!ng)
340 goto out;
341
342 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL);
343 if (!net)
344 goto out_free;
345
346 rcu_assign_pointer(net->gen, ng);
347out:
348 return net;
349
350out_free:
351 kfree(ng);
352 goto out;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200353}
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200354
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800355static void net_free(struct net *net)
356{
Eric Dumazet416c51e2014-09-09 08:24:53 -0700357 kfree(rcu_access_pointer(net->gen));
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800358 kmem_cache_free(net_cachep, net);
359}
360
Al Viroa685e082011-06-08 21:13:01 -0400361void net_drop_ns(void *p)
362{
363 struct net *ns = p;
364 if (ns && atomic_dec_and_test(&ns->passive))
365 net_free(ns);
366}
367
Eric W. Biederman038e7332012-06-14 02:31:10 -0700368struct net *copy_net_ns(unsigned long flags,
369 struct user_namespace *user_ns, struct net *old_net)
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700370{
Eric W. Biederman70328662016-08-08 14:33:23 -0500371 struct ucounts *ucounts;
Alexey Dobriyan088eb2d92009-05-04 11:12:14 -0700372 struct net *net;
373 int rv;
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700374
Rob Landley911cb192011-04-15 02:26:25 +0000375 if (!(flags & CLONE_NEWNET))
376 return get_net(old_net);
377
Eric W. Biederman70328662016-08-08 14:33:23 -0500378 ucounts = inc_net_namespaces(user_ns);
379 if (!ucounts)
Eric W. Biedermandf75e772016-09-22 13:08:36 -0500380 return ERR_PTR(-ENOSPC);
Eric W. Biederman70328662016-08-08 14:33:23 -0500381
Alexey Dobriyan088eb2d92009-05-04 11:12:14 -0700382 net = net_alloc();
Eric W. Biederman70328662016-08-08 14:33:23 -0500383 if (!net) {
384 dec_net_namespaces(ucounts);
Alexey Dobriyan088eb2d92009-05-04 11:12:14 -0700385 return ERR_PTR(-ENOMEM);
Eric W. Biederman70328662016-08-08 14:33:23 -0500386 }
Eric W. Biederman038e7332012-06-14 02:31:10 -0700387
388 get_user_ns(user_ns);
389
Andrey Vagin7281a662016-10-20 19:45:43 -0700390 rv = mutex_lock_killable(&net_mutex);
391 if (rv < 0) {
392 net_free(net);
393 dec_net_namespaces(ucounts);
394 put_user_ns(user_ns);
395 return ERR_PTR(rv);
396 }
397
Eric W. Biederman70328662016-08-08 14:33:23 -0500398 net->ucounts = ucounts;
Eric W. Biederman038e7332012-06-14 02:31:10 -0700399 rv = setup_net(net, user_ns);
Alexey Dobriyan088eb2d92009-05-04 11:12:14 -0700400 if (rv == 0) {
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800401 rtnl_lock();
Johannes Berg11a28d32009-07-10 09:51:33 +0000402 list_add_tail_rcu(&net->list, &net_namespace_list);
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800403 rtnl_unlock();
404 }
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700405 mutex_unlock(&net_mutex);
Alexey Dobriyan088eb2d92009-05-04 11:12:14 -0700406 if (rv < 0) {
Eric W. Biederman70328662016-08-08 14:33:23 -0500407 dec_net_namespaces(ucounts);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700408 put_user_ns(user_ns);
Al Viroa685e082011-06-08 21:13:01 -0400409 net_drop_ns(net);
Alexey Dobriyan088eb2d92009-05-04 11:12:14 -0700410 return ERR_PTR(rv);
411 }
412 return net;
413}
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800414
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000415static DEFINE_SPINLOCK(cleanup_list_lock);
416static LIST_HEAD(cleanup_list); /* Must hold cleanup_list_lock to touch */
417
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200418static void cleanup_net(struct work_struct *work)
419{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000420 const struct pernet_operations *ops;
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200421 struct net *net, *tmp;
xiao jin1818ce42014-04-25 08:50:54 +0800422 struct list_head net_kill_list;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000423 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200424
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000425 /* Atomically snapshot the list of namespaces to cleanup */
426 spin_lock_irq(&cleanup_list_lock);
427 list_replace_init(&cleanup_list, &net_kill_list);
428 spin_unlock_irq(&cleanup_list_lock);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200429
430 mutex_lock(&net_mutex);
431
432 /* Don't let anyone else find us. */
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700433 rtnl_lock();
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000434 list_for_each_entry(net, &net_kill_list, cleanup_list) {
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000435 list_del_rcu(&net->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000436 list_add_tail(&net->exit_list, &net_exit_list);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200437 for_each_net(tmp) {
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200438 int id;
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200439
Paul Moorefba143c2016-11-29 16:57:48 -0500440 spin_lock_bh(&tmp->nsid_lock);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200441 id = __peernet2id(tmp, net);
442 if (id >= 0)
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200443 idr_remove(&tmp->netns_ids, id);
Paul Moorefba143c2016-11-29 16:57:48 -0500444 spin_unlock_bh(&tmp->nsid_lock);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200445 if (id >= 0)
446 rtnl_net_notifyid(tmp, RTM_DELNSID, id);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200447 }
Paul Moorefba143c2016-11-29 16:57:48 -0500448 spin_lock_bh(&net->nsid_lock);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200449 idr_destroy(&net->netns_ids);
Paul Moorefba143c2016-11-29 16:57:48 -0500450 spin_unlock_bh(&net->nsid_lock);
Nicolas Dichtel6d458f52015-04-03 12:02:36 +0200451
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000452 }
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700453 rtnl_unlock();
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200454
Johannes Berg11a28d32009-07-10 09:51:33 +0000455 /*
456 * Another CPU might be rcu-iterating the list, wait for it.
457 * This needs to be before calling the exit() notifiers, so
458 * the rcu_barrier() below isn't sufficient alone.
459 */
460 synchronize_rcu();
461
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200462 /* Run all of the network namespace exit methods */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000463 list_for_each_entry_reverse(ops, &pernet_list, list)
464 ops_exit_list(ops, &net_exit_list);
465
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000466 /* Free the net generic variables */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000467 list_for_each_entry_reverse(ops, &pernet_list, list)
468 ops_free_list(ops, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200469
470 mutex_unlock(&net_mutex);
471
472 /* Ensure there are no outstanding rcu callbacks using this
473 * network namespace.
474 */
475 rcu_barrier();
476
477 /* Finally it is safe to free my network namespace structure */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000478 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
479 list_del_init(&net->exit_list);
Eric W. Biederman70328662016-08-08 14:33:23 -0500480 dec_net_namespaces(net->ucounts);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700481 put_user_ns(net->user_ns);
Al Viroa685e082011-06-08 21:13:01 -0400482 net_drop_ns(net);
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000483 }
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200484}
Florian Westphal7866cc52017-05-30 11:38:12 +0200485
486/**
487 * net_ns_barrier - wait until concurrent net_cleanup_work is done
488 *
489 * cleanup_net runs from work queue and will first remove namespaces
490 * from the global list, then run net exit functions.
491 *
492 * Call this in module exit path to make sure that all netns
493 * ->exit ops have been invoked before the function is removed.
494 */
495void net_ns_barrier(void)
496{
497 mutex_lock(&net_mutex);
498 mutex_unlock(&net_mutex);
499}
500EXPORT_SYMBOL(net_ns_barrier);
501
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000502static DECLARE_WORK(net_cleanup_work, cleanup_net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200503
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200504void __put_net(struct net *net)
505{
506 /* Cleanup the network namespace in process context */
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000507 unsigned long flags;
508
509 spin_lock_irqsave(&cleanup_list_lock, flags);
510 list_add(&net->cleanup_list, &cleanup_list);
511 spin_unlock_irqrestore(&cleanup_list_lock, flags);
512
513 queue_work(netns_wq, &net_cleanup_work);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200514}
515EXPORT_SYMBOL_GPL(__put_net);
516
Stephen Rothwell956c9202011-05-12 13:51:13 +1000517struct net *get_net_ns_by_fd(int fd)
518{
Stephen Rothwell956c9202011-05-12 13:51:13 +1000519 struct file *file;
Al Viro33c42942014-11-01 02:32:53 -0400520 struct ns_common *ns;
Stephen Rothwell956c9202011-05-12 13:51:13 +1000521 struct net *net;
522
Stephen Rothwell956c9202011-05-12 13:51:13 +1000523 file = proc_ns_fget(fd);
Al Viroc316e6a2011-06-05 00:37:35 +0000524 if (IS_ERR(file))
525 return ERR_CAST(file);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000526
Al Virof77c8012014-11-01 03:13:17 -0400527 ns = get_proc_ns(file_inode(file));
Al Viro33c42942014-11-01 02:32:53 -0400528 if (ns->ops == &netns_operations)
529 net = get_net(container_of(ns, struct net, ns));
Al Viroc316e6a2011-06-05 00:37:35 +0000530 else
531 net = ERR_PTR(-EINVAL);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000532
Al Viroc316e6a2011-06-05 00:37:35 +0000533 fput(file);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000534 return net;
535}
536
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700537#else
Stephen Rothwell956c9202011-05-12 13:51:13 +1000538struct net *get_net_ns_by_fd(int fd)
539{
540 return ERR_PTR(-EINVAL);
541}
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700542#endif
Vadim Kochan4b681c82015-01-12 16:34:05 +0200543EXPORT_SYMBOL_GPL(get_net_ns_by_fd);
Eric W. Biederman9dd776b2007-09-26 22:04:26 -0700544
Johannes Berg30ffee82009-07-10 09:51:35 +0000545struct net *get_net_ns_by_pid(pid_t pid)
546{
547 struct task_struct *tsk;
548 struct net *net;
549
550 /* Lookup the network namespace */
551 net = ERR_PTR(-ESRCH);
552 rcu_read_lock();
553 tsk = find_task_by_vpid(pid);
554 if (tsk) {
555 struct nsproxy *nsproxy;
Eric W. Biederman728dba32014-02-03 19:13:49 -0800556 task_lock(tsk);
557 nsproxy = tsk->nsproxy;
Johannes Berg30ffee82009-07-10 09:51:35 +0000558 if (nsproxy)
559 net = get_net(nsproxy->net_ns);
Eric W. Biederman728dba32014-02-03 19:13:49 -0800560 task_unlock(tsk);
Johannes Berg30ffee82009-07-10 09:51:35 +0000561 }
562 rcu_read_unlock();
563 return net;
564}
565EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
566
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700567static __net_init int net_ns_net_init(struct net *net)
568{
Al Viro33c42942014-11-01 02:32:53 -0400569#ifdef CONFIG_NET_NS
570 net->ns.ops = &netns_operations;
571#endif
Al Viro6344c432014-11-01 00:45:45 -0400572 return ns_alloc_inum(&net->ns);
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700573}
574
575static __net_exit void net_ns_net_exit(struct net *net)
576{
Al Viro6344c432014-11-01 00:45:45 -0400577 ns_free_inum(&net->ns);
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700578}
579
580static struct pernet_operations __net_initdata net_ns_ops = {
581 .init = net_ns_net_init,
582 .exit = net_ns_net_exit,
583};
584
stephen hemminger3ee52562016-08-31 15:17:49 -0700585static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = {
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100586 [NETNSA_NONE] = { .type = NLA_UNSPEC },
587 [NETNSA_NSID] = { .type = NLA_S32 },
588 [NETNSA_PID] = { .type = NLA_U32 },
589 [NETNSA_FD] = { .type = NLA_U32 },
590};
591
David Ahernc21ef3e2017-04-16 09:48:24 -0700592static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh,
593 struct netlink_ext_ack *extack)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100594{
595 struct net *net = sock_net(skb->sk);
596 struct nlattr *tb[NETNSA_MAX + 1];
597 struct net *peer;
598 int nsid, err;
599
600 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
David Ahernc21ef3e2017-04-16 09:48:24 -0700601 rtnl_net_policy, extack);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100602 if (err < 0)
603 return err;
604 if (!tb[NETNSA_NSID])
605 return -EINVAL;
606 nsid = nla_get_s32(tb[NETNSA_NSID]);
607
608 if (tb[NETNSA_PID])
609 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
610 else if (tb[NETNSA_FD])
611 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
612 else
613 return -EINVAL;
614 if (IS_ERR(peer))
615 return PTR_ERR(peer);
616
Paul Moorefba143c2016-11-29 16:57:48 -0500617 spin_lock_bh(&net->nsid_lock);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200618 if (__peernet2id(net, peer) >= 0) {
Paul Moorefba143c2016-11-29 16:57:48 -0500619 spin_unlock_bh(&net->nsid_lock);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100620 err = -EEXIST;
621 goto out;
622 }
623
624 err = alloc_netid(net, peer, nsid);
Paul Moorefba143c2016-11-29 16:57:48 -0500625 spin_unlock_bh(&net->nsid_lock);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200626 if (err >= 0) {
627 rtnl_net_notifyid(net, RTM_NEWNSID, err);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100628 err = 0;
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200629 }
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100630out:
631 put_net(peer);
632 return err;
633}
634
635static int rtnl_net_get_size(void)
636{
637 return NLMSG_ALIGN(sizeof(struct rtgenmsg))
638 + nla_total_size(sizeof(s32)) /* NETNSA_NSID */
639 ;
640}
641
642static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags,
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200643 int cmd, struct net *net, int nsid)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100644{
645 struct nlmsghdr *nlh;
646 struct rtgenmsg *rth;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100647
648 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags);
649 if (!nlh)
650 return -EMSGSIZE;
651
652 rth = nlmsg_data(nlh);
653 rth->rtgen_family = AF_UNSPEC;
654
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200655 if (nla_put_s32(skb, NETNSA_NSID, nsid))
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100656 goto nla_put_failure;
657
658 nlmsg_end(skb, nlh);
659 return 0;
660
661nla_put_failure:
662 nlmsg_cancel(skb, nlh);
663 return -EMSGSIZE;
664}
665
David Ahernc21ef3e2017-04-16 09:48:24 -0700666static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh,
667 struct netlink_ext_ack *extack)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100668{
669 struct net *net = sock_net(skb->sk);
670 struct nlattr *tb[NETNSA_MAX + 1];
671 struct sk_buff *msg;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100672 struct net *peer;
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200673 int err, id;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100674
675 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
David Ahernc21ef3e2017-04-16 09:48:24 -0700676 rtnl_net_policy, extack);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100677 if (err < 0)
678 return err;
679 if (tb[NETNSA_PID])
680 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
681 else if (tb[NETNSA_FD])
682 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
683 else
684 return -EINVAL;
685
686 if (IS_ERR(peer))
687 return PTR_ERR(peer);
688
689 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
690 if (!msg) {
691 err = -ENOMEM;
692 goto out;
693 }
694
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200695 id = peernet2id(net, peer);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100696 err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
David S. Millerb04096f2015-05-13 14:31:43 -0400697 RTM_NEWNSID, net, id);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100698 if (err < 0)
699 goto err_out;
700
701 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid);
702 goto out;
703
704err_out:
705 nlmsg_free(msg);
706out:
707 put_net(peer);
708 return err;
709}
710
Nicolas Dichtela143c402015-04-07 11:51:54 +0200711struct rtnl_net_dump_cb {
712 struct net *net;
713 struct sk_buff *skb;
714 struct netlink_callback *cb;
715 int idx;
716 int s_idx;
717};
718
719static int rtnl_net_dumpid_one(int id, void *peer, void *data)
720{
721 struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data;
722 int ret;
723
724 if (net_cb->idx < net_cb->s_idx)
725 goto cont;
726
727 ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid,
728 net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI,
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200729 RTM_NEWNSID, net_cb->net, id);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200730 if (ret < 0)
731 return ret;
732
733cont:
734 net_cb->idx++;
735 return 0;
736}
737
738static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb)
739{
740 struct net *net = sock_net(skb->sk);
741 struct rtnl_net_dump_cb net_cb = {
742 .net = net,
743 .skb = skb,
744 .cb = cb,
745 .idx = 0,
746 .s_idx = cb->args[0],
747 };
748
Paul Moorefba143c2016-11-29 16:57:48 -0500749 spin_lock_bh(&net->nsid_lock);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200750 idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb);
Paul Moorefba143c2016-11-29 16:57:48 -0500751 spin_unlock_bh(&net->nsid_lock);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200752
753 cb->args[0] = net_cb.idx;
754 return skb->len;
755}
756
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200757static void rtnl_net_notifyid(struct net *net, int cmd, int id)
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200758{
759 struct sk_buff *msg;
760 int err = -ENOMEM;
761
762 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
763 if (!msg)
764 goto out;
765
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200766 err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, id);
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200767 if (err < 0)
768 goto err_out;
769
770 rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0);
771 return;
772
773err_out:
774 nlmsg_free(msg);
775out:
776 rtnl_set_sk_err(net, RTNLGRP_NSID, err);
777}
778
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200779static int __init net_ns_init(void)
780{
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800781 struct net_generic *ng;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200782
Pavel Emelyanovd57a9212007-11-01 00:46:50 -0700783#ifdef CONFIG_NET_NS
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200784 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net),
785 SMP_CACHE_BYTES,
786 SLAB_PANIC, NULL);
Benjamin Thery3ef13552007-11-19 23:18:16 -0800787
788 /* Create workqueue for cleanup */
789 netns_wq = create_singlethread_workqueue("netns");
790 if (!netns_wq)
791 panic("Could not create netns workq");
Pavel Emelyanovd57a9212007-11-01 00:46:50 -0700792#endif
Benjamin Thery3ef13552007-11-19 23:18:16 -0800793
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800794 ng = net_alloc_generic();
795 if (!ng)
796 panic("Could not allocate generic netns");
797
798 rcu_assign_pointer(init_net.gen, ng);
799
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200800 mutex_lock(&net_mutex);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700801 if (setup_net(&init_net, &init_user_ns))
Stephen Hemmingerca0f3112009-05-21 15:10:31 -0700802 panic("Could not setup the initial network namespace");
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200803
Dmitry Torokhovf8c46cb2016-08-10 14:36:00 -0700804 init_net_initialized = true;
805
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700806 rtnl_lock();
Johannes Berg11a28d32009-07-10 09:51:33 +0000807 list_add_tail_rcu(&init_net.list, &net_namespace_list);
Eric W. Biedermanf4618d32007-09-26 22:40:08 -0700808 rtnl_unlock();
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200809
810 mutex_unlock(&net_mutex);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200811
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700812 register_pernet_subsys(&net_ns_ops);
813
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100814 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL, NULL);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200815 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid,
816 NULL);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100817
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200818 return 0;
819}
820
821pure_initcall(net_ns_init);
822
Denis V. Luneved160e82007-11-13 03:23:21 -0800823#ifdef CONFIG_NET_NS
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000824static int __register_pernet_operations(struct list_head *list,
825 struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200826{
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000827 struct net *net;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200828 int error;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000829 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200830
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200831 list_add_tail(&ops->list, list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000832 if (ops->init || (ops->id && ops->size)) {
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700833 for_each_net(net) {
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000834 error = ops_init(ops, net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200835 if (error)
836 goto out_undo;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000837 list_add_tail(&net->exit_list, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200838 }
839 }
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700840 return 0;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200841
842out_undo:
843 /* If I have an error cleanup all namespaces I initialized */
844 list_del(&ops->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000845 ops_exit_list(ops, &net_exit_list);
846 ops_free_list(ops, &net_exit_list);
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700847 return error;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200848}
849
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000850static void __unregister_pernet_operations(struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200851{
852 struct net *net;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000853 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200854
855 list_del(&ops->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000856 for_each_net(net)
857 list_add_tail(&net->exit_list, &net_exit_list);
858 ops_exit_list(ops, &net_exit_list);
859 ops_free_list(ops, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200860}
861
Denis V. Luneved160e82007-11-13 03:23:21 -0800862#else
863
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000864static int __register_pernet_operations(struct list_head *list,
865 struct pernet_operations *ops)
866{
Dmitry Torokhovf8c46cb2016-08-10 14:36:00 -0700867 if (!init_net_initialized) {
868 list_add_tail(&ops->list, list);
869 return 0;
870 }
871
Julian Anastasovb9229342012-04-16 04:43:15 +0000872 return ops_init(ops, &init_net);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000873}
874
875static void __unregister_pernet_operations(struct pernet_operations *ops)
876{
Dmitry Torokhovf8c46cb2016-08-10 14:36:00 -0700877 if (!init_net_initialized) {
878 list_del(&ops->list);
879 } else {
880 LIST_HEAD(net_exit_list);
881 list_add(&init_net.exit_list, &net_exit_list);
882 ops_exit_list(ops, &net_exit_list);
883 ops_free_list(ops, &net_exit_list);
884 }
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000885}
886
887#endif /* CONFIG_NET_NS */
888
889static DEFINE_IDA(net_generic_ids);
890
Denis V. Luneved160e82007-11-13 03:23:21 -0800891static int register_pernet_operations(struct list_head *list,
892 struct pernet_operations *ops)
893{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000894 int error;
895
896 if (ops->id) {
897again:
Alexey Dobriyan6af2d5f2016-12-02 04:21:32 +0300898 error = ida_get_new_above(&net_generic_ids, MIN_PERNET_OPS_ID, ops->id);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000899 if (error < 0) {
900 if (error == -EAGAIN) {
901 ida_pre_get(&net_generic_ids, GFP_KERNEL);
902 goto again;
903 }
904 return error;
905 }
Alexey Dobriyan6af2d5f2016-12-02 04:21:32 +0300906 max_gen_ptrs = max(max_gen_ptrs, *ops->id + 1);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000907 }
908 error = __register_pernet_operations(list, ops);
Eric W. Biederman3a765ed2009-12-03 02:29:06 +0000909 if (error) {
910 rcu_barrier();
911 if (ops->id)
912 ida_remove(&net_generic_ids, *ops->id);
913 }
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000914
915 return error;
Denis V. Luneved160e82007-11-13 03:23:21 -0800916}
917
918static void unregister_pernet_operations(struct pernet_operations *ops)
919{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000920
921 __unregister_pernet_operations(ops);
Eric W. Biederman3a765ed2009-12-03 02:29:06 +0000922 rcu_barrier();
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000923 if (ops->id)
924 ida_remove(&net_generic_ids, *ops->id);
Denis V. Luneved160e82007-11-13 03:23:21 -0800925}
Pavel Emelyanovc93cf612008-04-15 00:35:23 -0700926
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200927/**
928 * register_pernet_subsys - register a network namespace subsystem
929 * @ops: pernet operations structure for the subsystem
930 *
931 * Register a subsystem which has init and exit functions
932 * that are called when network namespaces are created and
933 * destroyed respectively.
934 *
935 * When registered all network namespace init functions are
936 * called for every existing network namespace. Allowing kernel
937 * modules to have a race free view of the set of network namespaces.
938 *
939 * When a new network namespace is created all of the init
940 * methods are called in the order in which they were registered.
941 *
942 * When a network namespace is destroyed all of the exit methods
943 * are called in the reverse of the order with which they were
944 * registered.
945 */
946int register_pernet_subsys(struct pernet_operations *ops)
947{
948 int error;
949 mutex_lock(&net_mutex);
950 error = register_pernet_operations(first_device, ops);
951 mutex_unlock(&net_mutex);
952 return error;
953}
954EXPORT_SYMBOL_GPL(register_pernet_subsys);
955
956/**
957 * unregister_pernet_subsys - unregister a network namespace subsystem
958 * @ops: pernet operations structure to manipulate
959 *
960 * Remove the pernet operations structure from the list to be
Oliver Pinter53379e52008-02-03 17:56:48 +0200961 * used when network namespaces are created or destroyed. In
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200962 * addition run the exit method for all existing network
963 * namespaces.
964 */
Jiri Pirkob3c981d2010-04-25 00:49:56 -0700965void unregister_pernet_subsys(struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200966{
967 mutex_lock(&net_mutex);
Jiri Pirkob3c981d2010-04-25 00:49:56 -0700968 unregister_pernet_operations(ops);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200969 mutex_unlock(&net_mutex);
970}
971EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
972
973/**
974 * register_pernet_device - register a network namespace device
975 * @ops: pernet operations structure for the subsystem
976 *
977 * Register a device which has init and exit functions
978 * that are called when network namespaces are created and
979 * destroyed respectively.
980 *
981 * When registered all network namespace init functions are
982 * called for every existing network namespace. Allowing kernel
983 * modules to have a race free view of the set of network namespaces.
984 *
985 * When a new network namespace is created all of the init
986 * methods are called in the order in which they were registered.
987 *
988 * When a network namespace is destroyed all of the exit methods
989 * are called in the reverse of the order with which they were
990 * registered.
991 */
992int register_pernet_device(struct pernet_operations *ops)
993{
994 int error;
995 mutex_lock(&net_mutex);
996 error = register_pernet_operations(&pernet_list, ops);
997 if (!error && (first_device == &pernet_list))
998 first_device = &ops->list;
999 mutex_unlock(&net_mutex);
1000 return error;
1001}
1002EXPORT_SYMBOL_GPL(register_pernet_device);
1003
1004/**
1005 * unregister_pernet_device - unregister a network namespace netdevice
1006 * @ops: pernet operations structure to manipulate
1007 *
1008 * Remove the pernet operations structure from the list to be
Oliver Pinter53379e52008-02-03 17:56:48 +02001009 * used when network namespaces are created or destroyed. In
Eric W. Biederman5f256be2007-09-12 11:50:50 +02001010 * addition run the exit method for all existing network
1011 * namespaces.
1012 */
1013void unregister_pernet_device(struct pernet_operations *ops)
1014{
1015 mutex_lock(&net_mutex);
1016 if (&ops->list == first_device)
1017 first_device = first_device->next;
1018 unregister_pernet_operations(ops);
1019 mutex_unlock(&net_mutex);
1020}
1021EXPORT_SYMBOL_GPL(unregister_pernet_device);
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001022
1023#ifdef CONFIG_NET_NS
Al Viro64964522014-11-01 00:37:32 -04001024static struct ns_common *netns_get(struct task_struct *task)
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001025{
Eric W. Biedermanf0630522011-05-04 17:51:50 -07001026 struct net *net = NULL;
1027 struct nsproxy *nsproxy;
1028
Eric W. Biederman728dba32014-02-03 19:13:49 -08001029 task_lock(task);
1030 nsproxy = task->nsproxy;
Eric W. Biedermanf0630522011-05-04 17:51:50 -07001031 if (nsproxy)
1032 net = get_net(nsproxy->net_ns);
Eric W. Biederman728dba32014-02-03 19:13:49 -08001033 task_unlock(task);
Eric W. Biedermanf0630522011-05-04 17:51:50 -07001034
Al Viroff248702014-11-01 00:10:50 -04001035 return net ? &net->ns : NULL;
1036}
1037
1038static inline struct net *to_net_ns(struct ns_common *ns)
1039{
1040 return container_of(ns, struct net, ns);
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001041}
1042
Al Viro64964522014-11-01 00:37:32 -04001043static void netns_put(struct ns_common *ns)
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001044{
Al Viroff248702014-11-01 00:10:50 -04001045 put_net(to_net_ns(ns));
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001046}
1047
Al Viro64964522014-11-01 00:37:32 -04001048static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns)
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001049{
Al Viroff248702014-11-01 00:10:50 -04001050 struct net *net = to_net_ns(ns);
Eric W. Biederman142e1d12012-07-26 01:13:20 -07001051
Eric W. Biederman5e4a0842012-12-14 07:55:36 -08001052 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -07001053 !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
Eric W. Biederman142e1d12012-07-26 01:13:20 -07001054 return -EPERM;
1055
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001056 put_net(nsproxy->net_ns);
Eric W. Biederman142e1d12012-07-26 01:13:20 -07001057 nsproxy->net_ns = get_net(net);
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001058 return 0;
1059}
1060
Andrey Vaginbcac25a2016-09-06 00:47:13 -07001061static struct user_namespace *netns_owner(struct ns_common *ns)
1062{
1063 return to_net_ns(ns)->user_ns;
1064}
1065
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001066const struct proc_ns_operations netns_operations = {
1067 .name = "net",
1068 .type = CLONE_NEWNET,
1069 .get = netns_get,
1070 .put = netns_put,
1071 .install = netns_install,
Andrey Vaginbcac25a2016-09-06 00:47:13 -07001072 .owner = netns_owner,
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001073};
1074#endif