blob: a11e03f920d3a7d9d50a37f699ffd66a26a8a057 [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;
Eric W. Biederman5f256be2007-09-12 11:50:50 +020032
Eric W. Biederman5f256be2007-09-12 11:50:50 +020033LIST_HEAD(net_namespace_list);
Alexey Dobriyanb76a4612008-10-08 11:35:06 +020034EXPORT_SYMBOL_GPL(net_namespace_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +020035
Kirill Tkhaif0b07bb12018-03-29 19:20:32 +030036/* Protects net_namespace_list. Nests iside rtnl_lock() */
37DECLARE_RWSEM(net_rwsem);
38EXPORT_SYMBOL_GPL(net_rwsem);
39
Rustad, Mark D734b6542012-07-18 09:06:07 +000040struct net init_net = {
Kirill Tkhai273c28b2018-01-12 18:28:31 +030041 .count = REFCOUNT_INIT(1),
David Howellsb5082df2017-04-27 22:40:23 +010042 .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head),
Rustad, Mark D734b6542012-07-18 09:06:07 +000043};
Denis V. Lunevff4b9502008-01-22 22:05:33 -080044EXPORT_SYMBOL(init_net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +020045
Dmitry Torokhovf8c46cb2016-08-10 14:36:00 -070046static bool init_net_initialized;
Kirill Tkhai1a57feb2018-02-13 12:26:23 +030047/*
Kirill Tkhai4420bf22018-03-27 18:02:23 +030048 * pernet_ops_rwsem: protects: pernet_list, net_generic_ids,
Kirill Tkhai1a57feb2018-02-13 12:26:23 +030049 * init_net_initialized and first_device pointer.
Kirill Tkhai8518e9b2018-03-27 18:02:32 +030050 * This is internal net namespace object. Please, don't use it
51 * outside.
Kirill Tkhai1a57feb2018-02-13 12:26:23 +030052 */
Kirill Tkhai4420bf22018-03-27 18:02:23 +030053DECLARE_RWSEM(pernet_ops_rwsem);
Kirill Tkhai554873e2018-03-30 19:38:37 +030054EXPORT_SYMBOL_GPL(pernet_ops_rwsem);
Dmitry Torokhovf8c46cb2016-08-10 14:36:00 -070055
Alexey Dobriyan6af2d5f2016-12-02 04:21:32 +030056#define MIN_PERNET_OPS_ID \
57 ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *))
58
Pavel Emelyanovdec827d2008-04-15 00:36:08 -070059#define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
60
Eric Dumazet073862b2012-01-26 00:41:38 +000061static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS;
62
63static struct net_generic *net_alloc_generic(void)
64{
65 struct net_generic *ng;
Alexey Dobriyan6af2d5f2016-12-02 04:21:32 +030066 unsigned int generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]);
Eric Dumazet073862b2012-01-26 00:41:38 +000067
68 ng = kzalloc(generic_size, GFP_KERNEL);
69 if (ng)
Alexey Dobriyan9bfc7b92016-12-02 04:12:58 +030070 ng->s.len = max_gen_ptrs;
Eric Dumazet073862b2012-01-26 00:41:38 +000071
72 return ng;
73}
74
Alexey Dobriyanc7d03a02016-11-17 04:58:21 +030075static int net_assign_generic(struct net *net, unsigned int id, void *data)
Jiri Pirko05fceb42010-04-23 01:40:47 +000076{
77 struct net_generic *ng, *old_ng;
78
Alexey Dobriyan6af2d5f2016-12-02 04:21:32 +030079 BUG_ON(id < MIN_PERNET_OPS_ID);
Jiri Pirko05fceb42010-04-23 01:40:47 +000080
Eric Dumazet1c877332010-10-25 03:20:11 +000081 old_ng = rcu_dereference_protected(net->gen,
Kirill Tkhai4420bf22018-03-27 18:02:23 +030082 lockdep_is_held(&pernet_ops_rwsem));
Alexey Dobriyan6af2d5f2016-12-02 04:21:32 +030083 if (old_ng->s.len > id) {
84 old_ng->ptr[id] = data;
Alexey Dobriyan1a9a05922016-12-02 04:11:34 +030085 return 0;
86 }
Jiri Pirko05fceb42010-04-23 01:40:47 +000087
Eric Dumazet073862b2012-01-26 00:41:38 +000088 ng = net_alloc_generic();
Jiri Pirko05fceb42010-04-23 01:40:47 +000089 if (ng == NULL)
90 return -ENOMEM;
91
92 /*
93 * Some synchronisation notes:
94 *
95 * The net_generic explores the net->gen array inside rcu
96 * read section. Besides once set the net->gen->ptr[x]
97 * pointer never changes (see rules in netns/generic.h).
98 *
99 * That said, we simply duplicate this array and schedule
100 * the old copy for kfree after a grace period.
101 */
102
Alexey Dobriyan6af2d5f2016-12-02 04:21:32 +0300103 memcpy(&ng->ptr[MIN_PERNET_OPS_ID], &old_ng->ptr[MIN_PERNET_OPS_ID],
104 (old_ng->s.len - MIN_PERNET_OPS_ID) * sizeof(void *));
105 ng->ptr[id] = data;
Jiri Pirko05fceb42010-04-23 01:40:47 +0000106
107 rcu_assign_pointer(net->gen, ng);
Alexey Dobriyan9bfc7b92016-12-02 04:12:58 +0300108 kfree_rcu(old_ng, s.rcu);
Jiri Pirko05fceb42010-04-23 01:40:47 +0000109 return 0;
110}
111
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000112static int ops_init(const struct pernet_operations *ops, struct net *net)
113{
Julian Anastasovb9229342012-04-16 04:43:15 +0000114 int err = -ENOMEM;
115 void *data = NULL;
116
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000117 if (ops->id && ops->size) {
Julian Anastasovb9229342012-04-16 04:43:15 +0000118 data = kzalloc(ops->size, GFP_KERNEL);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000119 if (!data)
Julian Anastasovb9229342012-04-16 04:43:15 +0000120 goto out;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000121
122 err = net_assign_generic(net, *ops->id, data);
Julian Anastasovb9229342012-04-16 04:43:15 +0000123 if (err)
124 goto cleanup;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000125 }
Julian Anastasovb9229342012-04-16 04:43:15 +0000126 err = 0;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000127 if (ops->init)
Julian Anastasovb9229342012-04-16 04:43:15 +0000128 err = ops->init(net);
129 if (!err)
130 return 0;
131
132cleanup:
133 kfree(data);
134
135out:
136 return err;
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000137}
138
139static void ops_free(const struct pernet_operations *ops, struct net *net)
140{
141 if (ops->id && ops->size) {
Alexey Dobriyanc7d03a02016-11-17 04:58:21 +0300142 kfree(net_generic(net, *ops->id));
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000143 }
144}
145
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000146static void ops_exit_list(const struct pernet_operations *ops,
147 struct list_head *net_exit_list)
148{
149 struct net *net;
150 if (ops->exit) {
151 list_for_each_entry(net, net_exit_list, exit_list)
152 ops->exit(net);
153 }
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000154 if (ops->exit_batch)
155 ops->exit_batch(net_exit_list);
156}
157
158static void ops_free_list(const struct pernet_operations *ops,
159 struct list_head *net_exit_list)
160{
161 struct net *net;
162 if (ops->size && ops->id) {
163 list_for_each_entry(net, net_exit_list, exit_list)
164 ops_free(ops, net);
165 }
166}
167
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200168/* should be called with nsid_lock held */
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100169static int alloc_netid(struct net *net, struct net *peer, int reqid)
170{
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200171 int min = 0, max = 0;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100172
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100173 if (reqid >= 0) {
174 min = reqid;
175 max = reqid + 1;
176 }
177
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200178 return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100179}
180
181/* This function is used by idr_for_each(). If net is equal to peer, the
182 * function returns the id so that idr_for_each() stops. Because we cannot
183 * returns the id 0 (idr_for_each() will not stop), we return the magic value
184 * NET_ID_ZERO (-1) for it.
185 */
186#define NET_ID_ZERO -1
187static int net_eq_idr(int id, void *net, void *peer)
188{
189 if (net_eq(net, peer))
190 return id ? : NET_ID_ZERO;
191 return 0;
192}
193
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200194/* Should be called with nsid_lock held. If a new id is assigned, the bool alloc
195 * is set to true, thus the caller knows that the new id must be notified via
196 * rtnl.
197 */
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200198static int __peernet2id_alloc(struct net *net, struct net *peer, bool *alloc)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100199{
200 int id = idr_for_each(&net->netns_ids, net_eq_idr, peer);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200201 bool alloc_it = *alloc;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100202
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200203 *alloc = false;
204
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100205 /* Magic value for id 0. */
206 if (id == NET_ID_ZERO)
207 return 0;
208 if (id > 0)
209 return id;
210
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200211 if (alloc_it) {
Nicolas Dichtel109582af2015-05-07 11:02:47 +0200212 id = alloc_netid(net, peer, -1);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200213 *alloc = true;
Nicolas Dichtel109582af2015-05-07 11:02:47 +0200214 return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED;
215 }
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100216
Nicolas Dichtel109582af2015-05-07 11:02:47 +0200217 return NETNSA_NSID_NOT_ASSIGNED;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100218}
219
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200220/* should be called with nsid_lock held */
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200221static int __peernet2id(struct net *net, struct net *peer)
222{
223 bool no = false;
224
225 return __peernet2id_alloc(net, peer, &no);
226}
227
228static void rtnl_net_notifyid(struct net *net, int cmd, int id);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100229/* This function returns the id of a peer netns. If no id is assigned, one will
230 * be allocated and returned.
231 */
Nicolas Dichtel7a0877d2015-05-07 11:02:49 +0200232int peernet2id_alloc(struct net *net, struct net *peer)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100233{
Kirill Tkhai0c06bea92018-01-16 12:31:41 +0300234 bool alloc = false, alive = false;
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200235 int id;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100236
Kirill Tkhai273c28b2018-01-12 18:28:31 +0300237 if (refcount_read(&net->count) == 0)
WANG Congcfc44a42016-11-16 10:27:02 -0800238 return NETNSA_NSID_NOT_ASSIGNED;
Paul Moorefba143c2016-11-29 16:57:48 -0500239 spin_lock_bh(&net->nsid_lock);
Kirill Tkhai0c06bea92018-01-16 12:31:41 +0300240 /*
241 * When peer is obtained from RCU lists, we may race with
242 * its cleanup. Check whether it's alive, and this guarantees
243 * we never hash a peer back to net->netns_ids, after it has
244 * just been idr_remove()'d from there in cleanup_net().
245 */
246 if (maybe_get_net(peer))
247 alive = alloc = true;
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200248 id = __peernet2id_alloc(net, peer, &alloc);
Paul Moorefba143c2016-11-29 16:57:48 -0500249 spin_unlock_bh(&net->nsid_lock);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200250 if (alloc && id >= 0)
251 rtnl_net_notifyid(net, RTM_NEWNSID, id);
Kirill Tkhai0c06bea92018-01-16 12:31:41 +0300252 if (alive)
253 put_net(peer);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200254 return id;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100255}
Jiri Benc7cbebc8a2017-11-02 17:04:36 -0200256EXPORT_SYMBOL_GPL(peernet2id_alloc);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100257
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200258/* This function returns, if assigned, the id of a peer netns. */
Nicolas Dichtel59324cf2015-05-07 11:02:53 +0200259int peernet2id(struct net *net, struct net *peer)
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200260{
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200261 int id;
262
Paul Moorefba143c2016-11-29 16:57:48 -0500263 spin_lock_bh(&net->nsid_lock);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200264 id = __peernet2id(net, peer);
Paul Moorefba143c2016-11-29 16:57:48 -0500265 spin_unlock_bh(&net->nsid_lock);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200266 return id;
267}
WANG Cong38f507f2016-09-01 21:53:44 -0700268EXPORT_SYMBOL(peernet2id);
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200269
Nicolas Dichtel59324cf2015-05-07 11:02:53 +0200270/* This function returns true is the peer netns has an id assigned into the
271 * current netns.
272 */
273bool peernet_has_id(struct net *net, struct net *peer)
274{
275 return peernet2id(net, peer) >= 0;
276}
277
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100278struct net *get_net_ns_by_id(struct net *net, int id)
279{
280 struct net *peer;
281
282 if (id < 0)
283 return NULL;
284
285 rcu_read_lock();
286 peer = idr_find(&net->netns_ids, id);
287 if (peer)
Eric W. Biederman21b59442017-12-19 11:27:56 -0600288 peer = maybe_get_net(peer);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100289 rcu_read_unlock();
290
291 return peer;
292}
293
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700294/*
295 * setup_net runs the initializers for the network namespace object.
296 */
Eric W. Biederman038e7332012-06-14 02:31:10 -0700297static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700298{
Kirill Tkhai4420bf22018-03-27 18:02:23 +0300299 /* Must be called with pernet_ops_rwsem held */
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000300 const struct pernet_operations *ops, *saved_ops;
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800301 int error = 0;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000302 LIST_HEAD(net_exit_list);
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700303
Kirill Tkhai273c28b2018-01-12 18:28:31 +0300304 refcount_set(&net->count, 1);
Reshetova, Elenac122e142017-06-30 13:08:08 +0300305 refcount_set(&net->passive, 1);
Thomas Graf4e985ad2011-06-21 03:11:20 +0000306 net->dev_base_seq = 1;
Eric W. Biederman038e7332012-06-14 02:31:10 -0700307 net->user_ns = user_ns;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100308 idr_init(&net->netns_ids);
WANG Congde133462015-05-15 14:47:32 -0700309 spin_lock_init(&net->nsid_lock);
Kirill Tkhaid9ff3042018-03-22 12:45:40 +0300310 mutex_init(&net->ipv4.ra_mutex);
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800311
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700312 list_for_each_entry(ops, &pernet_list, list) {
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000313 error = ops_init(ops, net);
314 if (error < 0)
315 goto out_undo;
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700316 }
Kirill Tkhaif0b07bb12018-03-29 19:20:32 +0300317 down_write(&net_rwsem);
Kirill Tkhai98f6c532018-02-13 12:26:02 +0300318 list_add_tail_rcu(&net->list, &net_namespace_list);
Kirill Tkhaif0b07bb12018-03-29 19:20:32 +0300319 up_write(&net_rwsem);
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700320out:
321 return error;
322
323out_undo:
324 /* Walk through the list backwards calling the exit functions
325 * for the pernet modules whose init functions did not fail.
326 */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000327 list_add(&net->exit_list, &net_exit_list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000328 saved_ops = ops;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000329 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
330 ops_exit_list(ops, &net_exit_list);
331
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000332 ops = saved_ops;
333 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000334 ops_free_list(ops, &net_exit_list);
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700335
336 rcu_barrier();
337 goto out;
338}
339
Roman Kapl7c3f18752017-05-24 10:22:22 +0200340static int __net_init net_defaults_init_net(struct net *net)
341{
342 net->core.sysctl_somaxconn = SOMAXCONN;
343 return 0;
344}
345
346static struct pernet_operations net_defaults_ops = {
347 .init = net_defaults_init_net,
348};
349
350static __init int net_defaults_init(void)
351{
352 if (register_pernet_subsys(&net_defaults_ops))
353 panic("Cannot initialize net default settings");
354
355 return 0;
356}
357
358core_initcall(net_defaults_init);
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800359
Clemens Nossebe47d472009-02-23 15:37:35 -0800360#ifdef CONFIG_NET_NS
Arnd Bergmann2ed6afd2016-09-23 18:06:12 +0200361static struct ucounts *inc_net_namespaces(struct user_namespace *ns)
362{
363 return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES);
364}
365
366static void dec_net_namespaces(struct ucounts *ucounts)
367{
368 dec_ucount(ucounts, UCOUNT_NET_NAMESPACES);
369}
370
Alexey Dobriyan08009a72018-02-24 21:20:33 +0300371static struct kmem_cache *net_cachep __ro_after_init;
Clemens Nossebe47d472009-02-23 15:37:35 -0800372static struct workqueue_struct *netns_wq;
373
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200374static struct net *net_alloc(void)
375{
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800376 struct net *net = NULL;
377 struct net_generic *ng;
378
379 ng = net_alloc_generic();
380 if (!ng)
381 goto out;
382
383 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL);
384 if (!net)
385 goto out_free;
386
387 rcu_assign_pointer(net->gen, ng);
388out:
389 return net;
390
391out_free:
392 kfree(ng);
393 goto out;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200394}
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200395
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800396static void net_free(struct net *net)
397{
Eric Dumazet416c51e2014-09-09 08:24:53 -0700398 kfree(rcu_access_pointer(net->gen));
Johann Felix Soden45a19b02007-11-07 01:30:30 -0800399 kmem_cache_free(net_cachep, net);
400}
401
Al Viroa685e082011-06-08 21:13:01 -0400402void net_drop_ns(void *p)
403{
404 struct net *ns = p;
Reshetova, Elenac122e142017-06-30 13:08:08 +0300405 if (ns && refcount_dec_and_test(&ns->passive))
Al Viroa685e082011-06-08 21:13:01 -0400406 net_free(ns);
407}
408
Eric W. Biederman038e7332012-06-14 02:31:10 -0700409struct net *copy_net_ns(unsigned long flags,
410 struct user_namespace *user_ns, struct net *old_net)
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700411{
Eric W. Biederman70328662016-08-08 14:33:23 -0500412 struct ucounts *ucounts;
Alexey Dobriyan088eb2d92009-05-04 11:12:14 -0700413 struct net *net;
414 int rv;
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700415
Rob Landley911cb192011-04-15 02:26:25 +0000416 if (!(flags & CLONE_NEWNET))
417 return get_net(old_net);
418
Eric W. Biederman70328662016-08-08 14:33:23 -0500419 ucounts = inc_net_namespaces(user_ns);
420 if (!ucounts)
Eric W. Biedermandf75e772016-09-22 13:08:36 -0500421 return ERR_PTR(-ENOSPC);
Eric W. Biederman70328662016-08-08 14:33:23 -0500422
Alexey Dobriyan088eb2d92009-05-04 11:12:14 -0700423 net = net_alloc();
Eric W. Biederman70328662016-08-08 14:33:23 -0500424 if (!net) {
Kirill Tkhai5ba049a2018-02-13 12:26:13 +0300425 rv = -ENOMEM;
426 goto dec_ucounts;
Eric W. Biederman70328662016-08-08 14:33:23 -0500427 }
Kirill Tkhai5ba049a2018-02-13 12:26:13 +0300428 refcount_set(&net->passive, 1);
429 net->ucounts = ucounts;
Eric W. Biederman038e7332012-06-14 02:31:10 -0700430 get_user_ns(user_ns);
Kirill Tkhai094374e2018-03-27 18:02:01 +0300431
Kirill Tkhai4420bf22018-03-27 18:02:23 +0300432 rv = down_read_killable(&pernet_ops_rwsem);
Kirill Tkhai5ba049a2018-02-13 12:26:13 +0300433 if (rv < 0)
434 goto put_userns;
Kirill Tkhai19efbd92018-02-19 12:58:38 +0300435
Eric W. Biederman038e7332012-06-14 02:31:10 -0700436 rv = setup_net(net, user_ns);
Kirill Tkhai19efbd92018-02-19 12:58:38 +0300437
Kirill Tkhai4420bf22018-03-27 18:02:23 +0300438 up_read(&pernet_ops_rwsem);
Kirill Tkhai19efbd92018-02-19 12:58:38 +0300439
Alexey Dobriyan088eb2d92009-05-04 11:12:14 -0700440 if (rv < 0) {
Kirill Tkhai5ba049a2018-02-13 12:26:13 +0300441put_userns:
Eric W. Biederman038e7332012-06-14 02:31:10 -0700442 put_user_ns(user_ns);
Al Viroa685e082011-06-08 21:13:01 -0400443 net_drop_ns(net);
Kirill Tkhai5ba049a2018-02-13 12:26:13 +0300444dec_ucounts:
445 dec_net_namespaces(ucounts);
Alexey Dobriyan088eb2d92009-05-04 11:12:14 -0700446 return ERR_PTR(rv);
447 }
448 return net;
449}
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800450
Kirill Tkhaifb07a822018-01-19 19:14:53 +0300451static void unhash_nsid(struct net *net, struct net *last)
452{
453 struct net *tmp;
454 /* This function is only called from cleanup_net() work,
455 * and this work is the only process, that may delete
456 * a net from net_namespace_list. So, when the below
457 * is executing, the list may only grow. Thus, we do not
Kirill Tkhaif0b07bb12018-03-29 19:20:32 +0300458 * use for_each_net_rcu() or net_rwsem.
Kirill Tkhaifb07a822018-01-19 19:14:53 +0300459 */
460 for_each_net(tmp) {
461 int id;
462
463 spin_lock_bh(&tmp->nsid_lock);
464 id = __peernet2id(tmp, net);
465 if (id >= 0)
466 idr_remove(&tmp->netns_ids, id);
467 spin_unlock_bh(&tmp->nsid_lock);
468 if (id >= 0)
469 rtnl_net_notifyid(tmp, RTM_DELNSID, id);
470 if (tmp == last)
471 break;
472 }
473 spin_lock_bh(&net->nsid_lock);
474 idr_destroy(&net->netns_ids);
475 spin_unlock_bh(&net->nsid_lock);
476}
477
Kirill Tkhai65b7b5b2018-02-19 12:58:45 +0300478static LLIST_HEAD(cleanup_list);
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000479
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200480static void cleanup_net(struct work_struct *work)
481{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000482 const struct pernet_operations *ops;
Kirill Tkhaifb07a822018-01-19 19:14:53 +0300483 struct net *net, *tmp, *last;
Kirill Tkhai65b7b5b2018-02-19 12:58:45 +0300484 struct llist_node *net_kill_list;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000485 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200486
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000487 /* Atomically snapshot the list of namespaces to cleanup */
Kirill Tkhai65b7b5b2018-02-19 12:58:45 +0300488 net_kill_list = llist_del_all(&cleanup_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200489
Kirill Tkhai4420bf22018-03-27 18:02:23 +0300490 down_read(&pernet_ops_rwsem);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200491
492 /* Don't let anyone else find us. */
Kirill Tkhaif0b07bb12018-03-29 19:20:32 +0300493 down_write(&net_rwsem);
Kirill Tkhai65b7b5b2018-02-19 12:58:45 +0300494 llist_for_each_entry(net, net_kill_list, cleanup_list)
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000495 list_del_rcu(&net->list);
Kirill Tkhaifb07a822018-01-19 19:14:53 +0300496 /* Cache last net. After we unlock rtnl, no one new net
497 * added to net_namespace_list can assign nsid pointer
498 * to a net from net_kill_list (see peernet2id_alloc()).
499 * So, we skip them in unhash_nsid().
500 *
501 * Note, that unhash_nsid() does not delete nsid links
502 * between net_kill_list's nets, as they've already
503 * deleted from net_namespace_list. But, this would be
504 * useless anyway, as netns_ids are destroyed there.
505 */
506 last = list_last_entry(&net_namespace_list, struct net, list);
Kirill Tkhaif0b07bb12018-03-29 19:20:32 +0300507 up_write(&net_rwsem);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200508
Kirill Tkhai65b7b5b2018-02-19 12:58:45 +0300509 llist_for_each_entry(net, net_kill_list, cleanup_list) {
Kirill Tkhaifb07a822018-01-19 19:14:53 +0300510 unhash_nsid(net, last);
511 list_add_tail(&net->exit_list, &net_exit_list);
512 }
513
Johannes Berg11a28d32009-07-10 09:51:33 +0000514 /*
515 * Another CPU might be rcu-iterating the list, wait for it.
516 * This needs to be before calling the exit() notifiers, so
517 * the rcu_barrier() below isn't sufficient alone.
518 */
519 synchronize_rcu();
520
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200521 /* Run all of the network namespace exit methods */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000522 list_for_each_entry_reverse(ops, &pernet_list, list)
523 ops_exit_list(ops, &net_exit_list);
524
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000525 /* Free the net generic variables */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000526 list_for_each_entry_reverse(ops, &pernet_list, list)
527 ops_free_list(ops, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200528
Kirill Tkhai4420bf22018-03-27 18:02:23 +0300529 up_read(&pernet_ops_rwsem);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200530
531 /* Ensure there are no outstanding rcu callbacks using this
532 * network namespace.
533 */
534 rcu_barrier();
535
536 /* Finally it is safe to free my network namespace structure */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000537 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
538 list_del_init(&net->exit_list);
Eric W. Biederman70328662016-08-08 14:33:23 -0500539 dec_net_namespaces(net->ucounts);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700540 put_user_ns(net->user_ns);
Al Viroa685e082011-06-08 21:13:01 -0400541 net_drop_ns(net);
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000542 }
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200543}
Florian Westphal7866cc52017-05-30 11:38:12 +0200544
545/**
546 * net_ns_barrier - wait until concurrent net_cleanup_work is done
547 *
548 * cleanup_net runs from work queue and will first remove namespaces
549 * from the global list, then run net exit functions.
550 *
551 * Call this in module exit path to make sure that all netns
552 * ->exit ops have been invoked before the function is removed.
553 */
554void net_ns_barrier(void)
555{
Kirill Tkhai4420bf22018-03-27 18:02:23 +0300556 down_write(&pernet_ops_rwsem);
557 up_write(&pernet_ops_rwsem);
Florian Westphal7866cc52017-05-30 11:38:12 +0200558}
559EXPORT_SYMBOL(net_ns_barrier);
560
Eric W. Biederman2b035b32009-11-29 22:25:27 +0000561static DECLARE_WORK(net_cleanup_work, cleanup_net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200562
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200563void __put_net(struct net *net)
564{
565 /* Cleanup the network namespace in process context */
Kirill Tkhai8349efd2018-02-19 12:58:54 +0300566 if (llist_add(&net->cleanup_list, &cleanup_list))
567 queue_work(netns_wq, &net_cleanup_work);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200568}
569EXPORT_SYMBOL_GPL(__put_net);
570
Stephen Rothwell956c9202011-05-12 13:51:13 +1000571struct net *get_net_ns_by_fd(int fd)
572{
Stephen Rothwell956c9202011-05-12 13:51:13 +1000573 struct file *file;
Al Viro33c42942014-11-01 02:32:53 -0400574 struct ns_common *ns;
Stephen Rothwell956c9202011-05-12 13:51:13 +1000575 struct net *net;
576
Stephen Rothwell956c9202011-05-12 13:51:13 +1000577 file = proc_ns_fget(fd);
Al Viroc316e6a2011-06-05 00:37:35 +0000578 if (IS_ERR(file))
579 return ERR_CAST(file);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000580
Al Virof77c8012014-11-01 03:13:17 -0400581 ns = get_proc_ns(file_inode(file));
Al Viro33c42942014-11-01 02:32:53 -0400582 if (ns->ops == &netns_operations)
583 net = get_net(container_of(ns, struct net, ns));
Al Viroc316e6a2011-06-05 00:37:35 +0000584 else
585 net = ERR_PTR(-EINVAL);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000586
Al Viroc316e6a2011-06-05 00:37:35 +0000587 fput(file);
Stephen Rothwell956c9202011-05-12 13:51:13 +1000588 return net;
589}
590
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700591#else
Stephen Rothwell956c9202011-05-12 13:51:13 +1000592struct net *get_net_ns_by_fd(int fd)
593{
594 return ERR_PTR(-EINVAL);
595}
Pavel Emelyanov6a1a3b92007-11-01 00:44:50 -0700596#endif
Vadim Kochan4b681c82015-01-12 16:34:05 +0200597EXPORT_SYMBOL_GPL(get_net_ns_by_fd);
Eric W. Biederman9dd776b2007-09-26 22:04:26 -0700598
Johannes Berg30ffee82009-07-10 09:51:35 +0000599struct net *get_net_ns_by_pid(pid_t pid)
600{
601 struct task_struct *tsk;
602 struct net *net;
603
604 /* Lookup the network namespace */
605 net = ERR_PTR(-ESRCH);
606 rcu_read_lock();
607 tsk = find_task_by_vpid(pid);
608 if (tsk) {
609 struct nsproxy *nsproxy;
Eric W. Biederman728dba32014-02-03 19:13:49 -0800610 task_lock(tsk);
611 nsproxy = tsk->nsproxy;
Johannes Berg30ffee82009-07-10 09:51:35 +0000612 if (nsproxy)
613 net = get_net(nsproxy->net_ns);
Eric W. Biederman728dba32014-02-03 19:13:49 -0800614 task_unlock(tsk);
Johannes Berg30ffee82009-07-10 09:51:35 +0000615 }
616 rcu_read_unlock();
617 return net;
618}
619EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
620
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700621static __net_init int net_ns_net_init(struct net *net)
622{
Al Viro33c42942014-11-01 02:32:53 -0400623#ifdef CONFIG_NET_NS
624 net->ns.ops = &netns_operations;
625#endif
Al Viro6344c432014-11-01 00:45:45 -0400626 return ns_alloc_inum(&net->ns);
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700627}
628
629static __net_exit void net_ns_net_exit(struct net *net)
630{
Al Viro6344c432014-11-01 00:45:45 -0400631 ns_free_inum(&net->ns);
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700632}
633
634static struct pernet_operations __net_initdata net_ns_ops = {
635 .init = net_ns_net_init,
636 .exit = net_ns_net_exit,
637};
638
stephen hemminger3ee52562016-08-31 15:17:49 -0700639static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = {
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100640 [NETNSA_NONE] = { .type = NLA_UNSPEC },
641 [NETNSA_NSID] = { .type = NLA_S32 },
642 [NETNSA_PID] = { .type = NLA_U32 },
643 [NETNSA_FD] = { .type = NLA_U32 },
644};
645
David Ahernc21ef3e2017-04-16 09:48:24 -0700646static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh,
647 struct netlink_ext_ack *extack)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100648{
649 struct net *net = sock_net(skb->sk);
650 struct nlattr *tb[NETNSA_MAX + 1];
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200651 struct nlattr *nla;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100652 struct net *peer;
653 int nsid, err;
654
655 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
David Ahernc21ef3e2017-04-16 09:48:24 -0700656 rtnl_net_policy, extack);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100657 if (err < 0)
658 return err;
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200659 if (!tb[NETNSA_NSID]) {
660 NL_SET_ERR_MSG(extack, "nsid is missing");
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100661 return -EINVAL;
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200662 }
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100663 nsid = nla_get_s32(tb[NETNSA_NSID]);
664
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200665 if (tb[NETNSA_PID]) {
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100666 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200667 nla = tb[NETNSA_PID];
668 } else if (tb[NETNSA_FD]) {
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100669 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200670 nla = tb[NETNSA_FD];
671 } else {
672 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100673 return -EINVAL;
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200674 }
675 if (IS_ERR(peer)) {
676 NL_SET_BAD_ATTR(extack, nla);
677 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100678 return PTR_ERR(peer);
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200679 }
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100680
Paul Moorefba143c2016-11-29 16:57:48 -0500681 spin_lock_bh(&net->nsid_lock);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200682 if (__peernet2id(net, peer) >= 0) {
Paul Moorefba143c2016-11-29 16:57:48 -0500683 spin_unlock_bh(&net->nsid_lock);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100684 err = -EEXIST;
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200685 NL_SET_BAD_ATTR(extack, nla);
686 NL_SET_ERR_MSG(extack,
687 "Peer netns already has a nsid assigned");
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100688 goto out;
689 }
690
691 err = alloc_netid(net, peer, nsid);
Paul Moorefba143c2016-11-29 16:57:48 -0500692 spin_unlock_bh(&net->nsid_lock);
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200693 if (err >= 0) {
694 rtnl_net_notifyid(net, RTM_NEWNSID, err);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100695 err = 0;
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200696 } else if (err == -ENOSPC && nsid >= 0) {
Nicolas Dichtel10d486a2017-06-09 14:41:57 +0200697 err = -EEXIST;
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200698 NL_SET_BAD_ATTR(extack, tb[NETNSA_NSID]);
699 NL_SET_ERR_MSG(extack, "The specified nsid is already used");
Nicolas Dichtel3138dbf2015-05-07 11:02:50 +0200700 }
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100701out:
702 put_net(peer);
703 return err;
704}
705
706static int rtnl_net_get_size(void)
707{
708 return NLMSG_ALIGN(sizeof(struct rtgenmsg))
709 + nla_total_size(sizeof(s32)) /* NETNSA_NSID */
710 ;
711}
712
713static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags,
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200714 int cmd, struct net *net, int nsid)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100715{
716 struct nlmsghdr *nlh;
717 struct rtgenmsg *rth;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100718
719 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags);
720 if (!nlh)
721 return -EMSGSIZE;
722
723 rth = nlmsg_data(nlh);
724 rth->rtgen_family = AF_UNSPEC;
725
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200726 if (nla_put_s32(skb, NETNSA_NSID, nsid))
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100727 goto nla_put_failure;
728
729 nlmsg_end(skb, nlh);
730 return 0;
731
732nla_put_failure:
733 nlmsg_cancel(skb, nlh);
734 return -EMSGSIZE;
735}
736
David Ahernc21ef3e2017-04-16 09:48:24 -0700737static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh,
738 struct netlink_ext_ack *extack)
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100739{
740 struct net *net = sock_net(skb->sk);
741 struct nlattr *tb[NETNSA_MAX + 1];
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200742 struct nlattr *nla;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100743 struct sk_buff *msg;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100744 struct net *peer;
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200745 int err, id;
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100746
747 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
David Ahernc21ef3e2017-04-16 09:48:24 -0700748 rtnl_net_policy, extack);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100749 if (err < 0)
750 return err;
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200751 if (tb[NETNSA_PID]) {
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100752 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200753 nla = tb[NETNSA_PID];
754 } else if (tb[NETNSA_FD]) {
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100755 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200756 nla = tb[NETNSA_FD];
757 } else {
758 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100759 return -EINVAL;
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200760 }
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100761
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200762 if (IS_ERR(peer)) {
763 NL_SET_BAD_ATTR(extack, nla);
764 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100765 return PTR_ERR(peer);
Nicolas Dichtel4a7f7bc2017-06-09 14:41:56 +0200766 }
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100767
768 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
769 if (!msg) {
770 err = -ENOMEM;
771 goto out;
772 }
773
Nicolas Dichtel95f38412015-05-07 11:02:51 +0200774 id = peernet2id(net, peer);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100775 err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
David S. Millerb04096f2015-05-13 14:31:43 -0400776 RTM_NEWNSID, net, id);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100777 if (err < 0)
778 goto err_out;
779
780 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid);
781 goto out;
782
783err_out:
784 nlmsg_free(msg);
785out:
786 put_net(peer);
787 return err;
788}
789
Nicolas Dichtela143c402015-04-07 11:51:54 +0200790struct rtnl_net_dump_cb {
791 struct net *net;
792 struct sk_buff *skb;
793 struct netlink_callback *cb;
794 int idx;
795 int s_idx;
796};
797
798static int rtnl_net_dumpid_one(int id, void *peer, void *data)
799{
800 struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data;
801 int ret;
802
803 if (net_cb->idx < net_cb->s_idx)
804 goto cont;
805
806 ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid,
807 net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI,
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200808 RTM_NEWNSID, net_cb->net, id);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200809 if (ret < 0)
810 return ret;
811
812cont:
813 net_cb->idx++;
814 return 0;
815}
816
817static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb)
818{
819 struct net *net = sock_net(skb->sk);
820 struct rtnl_net_dump_cb net_cb = {
821 .net = net,
822 .skb = skb,
823 .cb = cb,
824 .idx = 0,
825 .s_idx = cb->args[0],
826 };
827
Paul Moorefba143c2016-11-29 16:57:48 -0500828 spin_lock_bh(&net->nsid_lock);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200829 idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb);
Paul Moorefba143c2016-11-29 16:57:48 -0500830 spin_unlock_bh(&net->nsid_lock);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200831
832 cb->args[0] = net_cb.idx;
833 return skb->len;
834}
835
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200836static void rtnl_net_notifyid(struct net *net, int cmd, int id)
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200837{
838 struct sk_buff *msg;
839 int err = -ENOMEM;
840
841 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
842 if (!msg)
843 goto out;
844
Nicolas Dichtelcab3c8e2015-05-07 11:02:48 +0200845 err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, id);
Nicolas Dichtel9a963452015-04-07 11:51:53 +0200846 if (err < 0)
847 goto err_out;
848
849 rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0);
850 return;
851
852err_out:
853 nlmsg_free(msg);
854out:
855 rtnl_set_sk_err(net, RTNLGRP_NSID, err);
856}
857
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200858static int __init net_ns_init(void)
859{
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800860 struct net_generic *ng;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200861
Pavel Emelyanovd57a9212007-11-01 00:46:50 -0700862#ifdef CONFIG_NET_NS
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200863 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net),
864 SMP_CACHE_BYTES,
Kirill Tkhai30855ff2018-03-01 15:23:28 +0300865 SLAB_PANIC|SLAB_ACCOUNT, NULL);
Benjamin Thery3ef13552007-11-19 23:18:16 -0800866
867 /* Create workqueue for cleanup */
868 netns_wq = create_singlethread_workqueue("netns");
869 if (!netns_wq)
870 panic("Could not create netns workq");
Pavel Emelyanovd57a9212007-11-01 00:46:50 -0700871#endif
Benjamin Thery3ef13552007-11-19 23:18:16 -0800872
Daniel Lezcano486a87f2009-02-22 00:07:53 -0800873 ng = net_alloc_generic();
874 if (!ng)
875 panic("Could not allocate generic netns");
876
877 rcu_assign_pointer(init_net.gen, ng);
878
Kirill Tkhai4420bf22018-03-27 18:02:23 +0300879 down_write(&pernet_ops_rwsem);
Eric W. Biederman038e7332012-06-14 02:31:10 -0700880 if (setup_net(&init_net, &init_user_ns))
Stephen Hemmingerca0f3112009-05-21 15:10:31 -0700881 panic("Could not setup the initial network namespace");
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200882
Dmitry Torokhovf8c46cb2016-08-10 14:36:00 -0700883 init_net_initialized = true;
Kirill Tkhai4420bf22018-03-27 18:02:23 +0300884 up_write(&pernet_ops_rwsem);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200885
Eric W. Biederman98f842e2011-06-15 10:21:48 -0700886 register_pernet_subsys(&net_ns_ops);
887
Florian Westphal165b9112017-08-09 20:41:53 +0200888 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL,
889 RTNL_FLAG_DOIT_UNLOCKED);
Nicolas Dichtela143c402015-04-07 11:51:54 +0200890 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid,
Florian Westphal165b9112017-08-09 20:41:53 +0200891 RTNL_FLAG_DOIT_UNLOCKED);
Nicolas Dichtel0c7aecd2015-01-15 15:11:15 +0100892
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200893 return 0;
894}
895
896pure_initcall(net_ns_init);
897
Denis V. Luneved160e82007-11-13 03:23:21 -0800898#ifdef CONFIG_NET_NS
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000899static int __register_pernet_operations(struct list_head *list,
900 struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200901{
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000902 struct net *net;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200903 int error;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000904 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200905
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200906 list_add_tail(&ops->list, list);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000907 if (ops->init || (ops->id && ops->size)) {
Kirill Tkhaif0b07bb12018-03-29 19:20:32 +0300908 /* We held write locked pernet_ops_rwsem, and parallel
909 * setup_net() and cleanup_net() are not possible.
910 */
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700911 for_each_net(net) {
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000912 error = ops_init(ops, net);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200913 if (error)
914 goto out_undo;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000915 list_add_tail(&net->exit_list, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200916 }
917 }
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700918 return 0;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200919
920out_undo:
921 /* If I have an error cleanup all namespaces I initialized */
922 list_del(&ops->list);
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000923 ops_exit_list(ops, &net_exit_list);
924 ops_free_list(ops, &net_exit_list);
Pavel Emelyanov1dba3232007-11-01 00:42:43 -0700925 return error;
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200926}
927
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000928static void __unregister_pernet_operations(struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200929{
930 struct net *net;
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000931 LIST_HEAD(net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200932
933 list_del(&ops->list);
Kirill Tkhaif0b07bb12018-03-29 19:20:32 +0300934 /* See comment in __register_pernet_operations() */
Eric W. Biederman72ad9372009-12-03 02:29:03 +0000935 for_each_net(net)
936 list_add_tail(&net->exit_list, &net_exit_list);
937 ops_exit_list(ops, &net_exit_list);
938 ops_free_list(ops, &net_exit_list);
Eric W. Biederman5f256be2007-09-12 11:50:50 +0200939}
940
Denis V. Luneved160e82007-11-13 03:23:21 -0800941#else
942
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000943static int __register_pernet_operations(struct list_head *list,
944 struct pernet_operations *ops)
945{
Dmitry Torokhovf8c46cb2016-08-10 14:36:00 -0700946 if (!init_net_initialized) {
947 list_add_tail(&ops->list, list);
948 return 0;
949 }
950
Julian Anastasovb9229342012-04-16 04:43:15 +0000951 return ops_init(ops, &init_net);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000952}
953
954static void __unregister_pernet_operations(struct pernet_operations *ops)
955{
Dmitry Torokhovf8c46cb2016-08-10 14:36:00 -0700956 if (!init_net_initialized) {
957 list_del(&ops->list);
958 } else {
959 LIST_HEAD(net_exit_list);
960 list_add(&init_net.exit_list, &net_exit_list);
961 ops_exit_list(ops, &net_exit_list);
962 ops_free_list(ops, &net_exit_list);
963 }
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000964}
965
966#endif /* CONFIG_NET_NS */
967
968static DEFINE_IDA(net_generic_ids);
969
Denis V. Luneved160e82007-11-13 03:23:21 -0800970static int register_pernet_operations(struct list_head *list,
971 struct pernet_operations *ops)
972{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000973 int error;
974
975 if (ops->id) {
976again:
Alexey Dobriyan6af2d5f2016-12-02 04:21:32 +0300977 error = ida_get_new_above(&net_generic_ids, MIN_PERNET_OPS_ID, ops->id);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000978 if (error < 0) {
979 if (error == -EAGAIN) {
980 ida_pre_get(&net_generic_ids, GFP_KERNEL);
981 goto again;
982 }
983 return error;
984 }
Alexey Dobriyan6af2d5f2016-12-02 04:21:32 +0300985 max_gen_ptrs = max(max_gen_ptrs, *ops->id + 1);
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000986 }
987 error = __register_pernet_operations(list, ops);
Eric W. Biederman3a765ed2009-12-03 02:29:06 +0000988 if (error) {
989 rcu_barrier();
990 if (ops->id)
991 ida_remove(&net_generic_ids, *ops->id);
992 }
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000993
994 return error;
Denis V. Luneved160e82007-11-13 03:23:21 -0800995}
996
997static void unregister_pernet_operations(struct pernet_operations *ops)
998{
Eric W. Biedermanf875bae2009-11-29 22:25:28 +0000999 __unregister_pernet_operations(ops);
Eric W. Biederman3a765ed2009-12-03 02:29:06 +00001000 rcu_barrier();
Eric W. Biedermanf875bae2009-11-29 22:25:28 +00001001 if (ops->id)
1002 ida_remove(&net_generic_ids, *ops->id);
Denis V. Luneved160e82007-11-13 03:23:21 -08001003}
Pavel Emelyanovc93cf612008-04-15 00:35:23 -07001004
Eric W. Biederman5f256be2007-09-12 11:50:50 +02001005/**
1006 * register_pernet_subsys - register a network namespace subsystem
1007 * @ops: pernet operations structure for the subsystem
1008 *
1009 * Register a subsystem which has init and exit functions
1010 * that are called when network namespaces are created and
1011 * destroyed respectively.
1012 *
1013 * When registered all network namespace init functions are
1014 * called for every existing network namespace. Allowing kernel
1015 * modules to have a race free view of the set of network namespaces.
1016 *
1017 * When a new network namespace is created all of the init
1018 * methods are called in the order in which they were registered.
1019 *
1020 * When a network namespace is destroyed all of the exit methods
1021 * are called in the reverse of the order with which they were
1022 * registered.
1023 */
1024int register_pernet_subsys(struct pernet_operations *ops)
1025{
1026 int error;
Kirill Tkhai4420bf22018-03-27 18:02:23 +03001027 down_write(&pernet_ops_rwsem);
Eric W. Biederman5f256be2007-09-12 11:50:50 +02001028 error = register_pernet_operations(first_device, ops);
Kirill Tkhai4420bf22018-03-27 18:02:23 +03001029 up_write(&pernet_ops_rwsem);
Eric W. Biederman5f256be2007-09-12 11:50:50 +02001030 return error;
1031}
1032EXPORT_SYMBOL_GPL(register_pernet_subsys);
1033
1034/**
1035 * unregister_pernet_subsys - unregister a network namespace subsystem
1036 * @ops: pernet operations structure to manipulate
1037 *
1038 * Remove the pernet operations structure from the list to be
Oliver Pinter53379e52008-02-03 17:56:48 +02001039 * used when network namespaces are created or destroyed. In
Eric W. Biederman5f256be2007-09-12 11:50:50 +02001040 * addition run the exit method for all existing network
1041 * namespaces.
1042 */
Jiri Pirkob3c981d2010-04-25 00:49:56 -07001043void unregister_pernet_subsys(struct pernet_operations *ops)
Eric W. Biederman5f256be2007-09-12 11:50:50 +02001044{
Kirill Tkhai4420bf22018-03-27 18:02:23 +03001045 down_write(&pernet_ops_rwsem);
Jiri Pirkob3c981d2010-04-25 00:49:56 -07001046 unregister_pernet_operations(ops);
Kirill Tkhai4420bf22018-03-27 18:02:23 +03001047 up_write(&pernet_ops_rwsem);
Eric W. Biederman5f256be2007-09-12 11:50:50 +02001048}
1049EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
1050
1051/**
1052 * register_pernet_device - register a network namespace device
1053 * @ops: pernet operations structure for the subsystem
1054 *
1055 * Register a device which has init and exit functions
1056 * that are called when network namespaces are created and
1057 * destroyed respectively.
1058 *
1059 * When registered all network namespace init functions are
1060 * called for every existing network namespace. Allowing kernel
1061 * modules to have a race free view of the set of network namespaces.
1062 *
1063 * When a new network namespace is created all of the init
1064 * methods are called in the order in which they were registered.
1065 *
1066 * When a network namespace is destroyed all of the exit methods
1067 * are called in the reverse of the order with which they were
1068 * registered.
1069 */
1070int register_pernet_device(struct pernet_operations *ops)
1071{
1072 int error;
Kirill Tkhai4420bf22018-03-27 18:02:23 +03001073 down_write(&pernet_ops_rwsem);
Eric W. Biederman5f256be2007-09-12 11:50:50 +02001074 error = register_pernet_operations(&pernet_list, ops);
1075 if (!error && (first_device == &pernet_list))
1076 first_device = &ops->list;
Kirill Tkhai4420bf22018-03-27 18:02:23 +03001077 up_write(&pernet_ops_rwsem);
Eric W. Biederman5f256be2007-09-12 11:50:50 +02001078 return error;
1079}
1080EXPORT_SYMBOL_GPL(register_pernet_device);
1081
1082/**
1083 * unregister_pernet_device - unregister a network namespace netdevice
1084 * @ops: pernet operations structure to manipulate
1085 *
1086 * Remove the pernet operations structure from the list to be
Oliver Pinter53379e52008-02-03 17:56:48 +02001087 * used when network namespaces are created or destroyed. In
Eric W. Biederman5f256be2007-09-12 11:50:50 +02001088 * addition run the exit method for all existing network
1089 * namespaces.
1090 */
1091void unregister_pernet_device(struct pernet_operations *ops)
1092{
Kirill Tkhai4420bf22018-03-27 18:02:23 +03001093 down_write(&pernet_ops_rwsem);
Eric W. Biederman5f256be2007-09-12 11:50:50 +02001094 if (&ops->list == first_device)
1095 first_device = first_device->next;
1096 unregister_pernet_operations(ops);
Kirill Tkhai4420bf22018-03-27 18:02:23 +03001097 up_write(&pernet_ops_rwsem);
Eric W. Biederman5f256be2007-09-12 11:50:50 +02001098}
1099EXPORT_SYMBOL_GPL(unregister_pernet_device);
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001100
1101#ifdef CONFIG_NET_NS
Al Viro64964522014-11-01 00:37:32 -04001102static struct ns_common *netns_get(struct task_struct *task)
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001103{
Eric W. Biedermanf0630522011-05-04 17:51:50 -07001104 struct net *net = NULL;
1105 struct nsproxy *nsproxy;
1106
Eric W. Biederman728dba32014-02-03 19:13:49 -08001107 task_lock(task);
1108 nsproxy = task->nsproxy;
Eric W. Biedermanf0630522011-05-04 17:51:50 -07001109 if (nsproxy)
1110 net = get_net(nsproxy->net_ns);
Eric W. Biederman728dba32014-02-03 19:13:49 -08001111 task_unlock(task);
Eric W. Biedermanf0630522011-05-04 17:51:50 -07001112
Al Viroff248702014-11-01 00:10:50 -04001113 return net ? &net->ns : NULL;
1114}
1115
1116static inline struct net *to_net_ns(struct ns_common *ns)
1117{
1118 return container_of(ns, struct net, ns);
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001119}
1120
Al Viro64964522014-11-01 00:37:32 -04001121static void netns_put(struct ns_common *ns)
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001122{
Al Viroff248702014-11-01 00:10:50 -04001123 put_net(to_net_ns(ns));
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001124}
1125
Al Viro64964522014-11-01 00:37:32 -04001126static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns)
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001127{
Al Viroff248702014-11-01 00:10:50 -04001128 struct net *net = to_net_ns(ns);
Eric W. Biederman142e1d12012-07-26 01:13:20 -07001129
Eric W. Biederman5e4a0842012-12-14 07:55:36 -08001130 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -07001131 !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
Eric W. Biederman142e1d12012-07-26 01:13:20 -07001132 return -EPERM;
1133
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001134 put_net(nsproxy->net_ns);
Eric W. Biederman142e1d12012-07-26 01:13:20 -07001135 nsproxy->net_ns = get_net(net);
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001136 return 0;
1137}
1138
Andrey Vaginbcac25a2016-09-06 00:47:13 -07001139static struct user_namespace *netns_owner(struct ns_common *ns)
1140{
1141 return to_net_ns(ns)->user_ns;
1142}
1143
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001144const struct proc_ns_operations netns_operations = {
1145 .name = "net",
1146 .type = CLONE_NEWNET,
1147 .get = netns_get,
1148 .put = netns_put,
1149 .install = netns_install,
Andrey Vaginbcac25a2016-09-06 00:47:13 -07001150 .owner = netns_owner,
Eric W. Biederman13b6f572010-03-07 18:14:23 -08001151};
1152#endif