1 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
3 #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>
9 #include <linux/sched.h>
10 #include <linux/idr.h>
11 #include <linux/rculist.h>
12 #include <linux/nsproxy.h>
14 #include <linux/proc_ns.h>
15 #include <linux/file.h>
16 #include <linux/export.h>
17 #include <linux/user_namespace.h>
18 #include <linux/net_namespace.h>
19 #include <linux/sched/task.h>
22 #include <net/netlink.h>
23 #include <net/net_namespace.h>
24 #include <net/netns/generic.h>
27 * Our network namespace constructor/destructor lists
30 static LIST_HEAD(pernet_list);
31 static struct list_head *first_device = &pernet_list;
33 LIST_HEAD(net_namespace_list);
34 EXPORT_SYMBOL_GPL(net_namespace_list);
36 struct net init_net = {
37 .count = REFCOUNT_INIT(1),
38 .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head),
40 EXPORT_SYMBOL(init_net);
42 static bool init_net_initialized;
44 * pernet_ops_rwsem: protects: pernet_list, net_generic_ids,
45 * init_net_initialized and first_device pointer.
46 * This is internal net namespace object. Please, don't use it
49 DECLARE_RWSEM(pernet_ops_rwsem);
51 #define MIN_PERNET_OPS_ID \
52 ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *))
54 #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
56 static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS;
58 static struct net_generic *net_alloc_generic(void)
60 struct net_generic *ng;
61 unsigned int generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]);
63 ng = kzalloc(generic_size, GFP_KERNEL);
65 ng->s.len = max_gen_ptrs;
70 static int net_assign_generic(struct net *net, unsigned int id, void *data)
72 struct net_generic *ng, *old_ng;
74 BUG_ON(id < MIN_PERNET_OPS_ID);
76 old_ng = rcu_dereference_protected(net->gen,
77 lockdep_is_held(&pernet_ops_rwsem));
78 if (old_ng->s.len > id) {
79 old_ng->ptr[id] = data;
83 ng = net_alloc_generic();
88 * Some synchronisation notes:
90 * The net_generic explores the net->gen array inside rcu
91 * read section. Besides once set the net->gen->ptr[x]
92 * pointer never changes (see rules in netns/generic.h).
94 * That said, we simply duplicate this array and schedule
95 * the old copy for kfree after a grace period.
98 memcpy(&ng->ptr[MIN_PERNET_OPS_ID], &old_ng->ptr[MIN_PERNET_OPS_ID],
99 (old_ng->s.len - MIN_PERNET_OPS_ID) * sizeof(void *));
102 rcu_assign_pointer(net->gen, ng);
103 kfree_rcu(old_ng, s.rcu);
107 static int ops_init(const struct pernet_operations *ops, struct net *net)
112 if (ops->id && ops->size) {
113 data = kzalloc(ops->size, GFP_KERNEL);
117 err = net_assign_generic(net, *ops->id, data);
123 err = ops->init(net);
134 static void ops_free(const struct pernet_operations *ops, struct net *net)
136 if (ops->id && ops->size) {
137 kfree(net_generic(net, *ops->id));
141 static void ops_exit_list(const struct pernet_operations *ops,
142 struct list_head *net_exit_list)
146 list_for_each_entry(net, net_exit_list, exit_list)
150 ops->exit_batch(net_exit_list);
153 static void ops_free_list(const struct pernet_operations *ops,
154 struct list_head *net_exit_list)
157 if (ops->size && ops->id) {
158 list_for_each_entry(net, net_exit_list, exit_list)
163 /* should be called with nsid_lock held */
164 static int alloc_netid(struct net *net, struct net *peer, int reqid)
166 int min = 0, max = 0;
173 return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC);
176 /* This function is used by idr_for_each(). If net is equal to peer, the
177 * function returns the id so that idr_for_each() stops. Because we cannot
178 * returns the id 0 (idr_for_each() will not stop), we return the magic value
179 * NET_ID_ZERO (-1) for it.
181 #define NET_ID_ZERO -1
182 static int net_eq_idr(int id, void *net, void *peer)
184 if (net_eq(net, peer))
185 return id ? : NET_ID_ZERO;
189 /* Should be called with nsid_lock held. If a new id is assigned, the bool alloc
190 * is set to true, thus the caller knows that the new id must be notified via
193 static int __peernet2id_alloc(struct net *net, struct net *peer, bool *alloc)
195 int id = idr_for_each(&net->netns_ids, net_eq_idr, peer);
196 bool alloc_it = *alloc;
200 /* Magic value for id 0. */
201 if (id == NET_ID_ZERO)
207 id = alloc_netid(net, peer, -1);
209 return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED;
212 return NETNSA_NSID_NOT_ASSIGNED;
215 /* should be called with nsid_lock held */
216 static int __peernet2id(struct net *net, struct net *peer)
220 return __peernet2id_alloc(net, peer, &no);
223 static void rtnl_net_notifyid(struct net *net, int cmd, int id);
224 /* This function returns the id of a peer netns. If no id is assigned, one will
225 * be allocated and returned.
227 int peernet2id_alloc(struct net *net, struct net *peer)
229 bool alloc = false, alive = false;
232 if (refcount_read(&net->count) == 0)
233 return NETNSA_NSID_NOT_ASSIGNED;
234 spin_lock_bh(&net->nsid_lock);
236 * When peer is obtained from RCU lists, we may race with
237 * its cleanup. Check whether it's alive, and this guarantees
238 * we never hash a peer back to net->netns_ids, after it has
239 * just been idr_remove()'d from there in cleanup_net().
241 if (maybe_get_net(peer))
242 alive = alloc = true;
243 id = __peernet2id_alloc(net, peer, &alloc);
244 spin_unlock_bh(&net->nsid_lock);
245 if (alloc && id >= 0)
246 rtnl_net_notifyid(net, RTM_NEWNSID, id);
251 EXPORT_SYMBOL_GPL(peernet2id_alloc);
253 /* This function returns, if assigned, the id of a peer netns. */
254 int peernet2id(struct net *net, struct net *peer)
258 spin_lock_bh(&net->nsid_lock);
259 id = __peernet2id(net, peer);
260 spin_unlock_bh(&net->nsid_lock);
263 EXPORT_SYMBOL(peernet2id);
265 /* This function returns true is the peer netns has an id assigned into the
268 bool peernet_has_id(struct net *net, struct net *peer)
270 return peernet2id(net, peer) >= 0;
273 struct net *get_net_ns_by_id(struct net *net, int id)
281 peer = idr_find(&net->netns_ids, id);
283 peer = maybe_get_net(peer);
290 * setup_net runs the initializers for the network namespace object.
292 static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
294 /* Must be called with pernet_ops_rwsem held */
295 const struct pernet_operations *ops, *saved_ops;
297 LIST_HEAD(net_exit_list);
299 refcount_set(&net->count, 1);
300 refcount_set(&net->passive, 1);
301 net->dev_base_seq = 1;
302 net->user_ns = user_ns;
303 idr_init(&net->netns_ids);
304 spin_lock_init(&net->nsid_lock);
305 mutex_init(&net->ipv4.ra_mutex);
307 list_for_each_entry(ops, &pernet_list, list) {
308 error = ops_init(ops, net);
313 list_add_tail_rcu(&net->list, &net_namespace_list);
319 /* Walk through the list backwards calling the exit functions
320 * for the pernet modules whose init functions did not fail.
322 list_add(&net->exit_list, &net_exit_list);
324 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
325 ops_exit_list(ops, &net_exit_list);
328 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
329 ops_free_list(ops, &net_exit_list);
335 static int __net_init net_defaults_init_net(struct net *net)
337 net->core.sysctl_somaxconn = SOMAXCONN;
341 static struct pernet_operations net_defaults_ops = {
342 .init = net_defaults_init_net,
345 static __init int net_defaults_init(void)
347 if (register_pernet_subsys(&net_defaults_ops))
348 panic("Cannot initialize net default settings");
353 core_initcall(net_defaults_init);
356 static struct ucounts *inc_net_namespaces(struct user_namespace *ns)
358 return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES);
361 static void dec_net_namespaces(struct ucounts *ucounts)
363 dec_ucount(ucounts, UCOUNT_NET_NAMESPACES);
366 static struct kmem_cache *net_cachep __ro_after_init;
367 static struct workqueue_struct *netns_wq;
369 static struct net *net_alloc(void)
371 struct net *net = NULL;
372 struct net_generic *ng;
374 ng = net_alloc_generic();
378 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL);
382 rcu_assign_pointer(net->gen, ng);
391 static void net_free(struct net *net)
393 kfree(rcu_access_pointer(net->gen));
394 kmem_cache_free(net_cachep, net);
397 void net_drop_ns(void *p)
400 if (ns && refcount_dec_and_test(&ns->passive))
404 struct net *copy_net_ns(unsigned long flags,
405 struct user_namespace *user_ns, struct net *old_net)
407 struct ucounts *ucounts;
411 if (!(flags & CLONE_NEWNET))
412 return get_net(old_net);
414 ucounts = inc_net_namespaces(user_ns);
416 return ERR_PTR(-ENOSPC);
423 refcount_set(&net->passive, 1);
424 net->ucounts = ucounts;
425 get_user_ns(user_ns);
427 rv = down_read_killable(&pernet_ops_rwsem);
431 rv = setup_net(net, user_ns);
433 up_read(&pernet_ops_rwsem);
437 put_user_ns(user_ns);
440 dec_net_namespaces(ucounts);
446 static void unhash_nsid(struct net *net, struct net *last)
449 /* This function is only called from cleanup_net() work,
450 * and this work is the only process, that may delete
451 * a net from net_namespace_list. So, when the below
452 * is executing, the list may only grow. Thus, we do not
453 * use for_each_net_rcu() or rtnl_lock().
458 spin_lock_bh(&tmp->nsid_lock);
459 id = __peernet2id(tmp, net);
461 idr_remove(&tmp->netns_ids, id);
462 spin_unlock_bh(&tmp->nsid_lock);
464 rtnl_net_notifyid(tmp, RTM_DELNSID, id);
468 spin_lock_bh(&net->nsid_lock);
469 idr_destroy(&net->netns_ids);
470 spin_unlock_bh(&net->nsid_lock);
473 static LLIST_HEAD(cleanup_list);
475 static void cleanup_net(struct work_struct *work)
477 const struct pernet_operations *ops;
478 struct net *net, *tmp, *last;
479 struct llist_node *net_kill_list;
480 LIST_HEAD(net_exit_list);
482 /* Atomically snapshot the list of namespaces to cleanup */
483 net_kill_list = llist_del_all(&cleanup_list);
485 down_read(&pernet_ops_rwsem);
487 /* Don't let anyone else find us. */
489 llist_for_each_entry(net, net_kill_list, cleanup_list)
490 list_del_rcu(&net->list);
491 /* Cache last net. After we unlock rtnl, no one new net
492 * added to net_namespace_list can assign nsid pointer
493 * to a net from net_kill_list (see peernet2id_alloc()).
494 * So, we skip them in unhash_nsid().
496 * Note, that unhash_nsid() does not delete nsid links
497 * between net_kill_list's nets, as they've already
498 * deleted from net_namespace_list. But, this would be
499 * useless anyway, as netns_ids are destroyed there.
501 last = list_last_entry(&net_namespace_list, struct net, list);
504 llist_for_each_entry(net, net_kill_list, cleanup_list) {
505 unhash_nsid(net, last);
506 list_add_tail(&net->exit_list, &net_exit_list);
510 * Another CPU might be rcu-iterating the list, wait for it.
511 * This needs to be before calling the exit() notifiers, so
512 * the rcu_barrier() below isn't sufficient alone.
516 /* Run all of the network namespace exit methods */
517 list_for_each_entry_reverse(ops, &pernet_list, list)
518 ops_exit_list(ops, &net_exit_list);
520 /* Free the net generic variables */
521 list_for_each_entry_reverse(ops, &pernet_list, list)
522 ops_free_list(ops, &net_exit_list);
524 up_read(&pernet_ops_rwsem);
526 /* Ensure there are no outstanding rcu callbacks using this
531 /* Finally it is safe to free my network namespace structure */
532 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
533 list_del_init(&net->exit_list);
534 dec_net_namespaces(net->ucounts);
535 put_user_ns(net->user_ns);
541 * net_ns_barrier - wait until concurrent net_cleanup_work is done
543 * cleanup_net runs from work queue and will first remove namespaces
544 * from the global list, then run net exit functions.
546 * Call this in module exit path to make sure that all netns
547 * ->exit ops have been invoked before the function is removed.
549 void net_ns_barrier(void)
551 down_write(&pernet_ops_rwsem);
552 up_write(&pernet_ops_rwsem);
554 EXPORT_SYMBOL(net_ns_barrier);
556 static DECLARE_WORK(net_cleanup_work, cleanup_net);
558 void __put_net(struct net *net)
560 /* Cleanup the network namespace in process context */
561 if (llist_add(&net->cleanup_list, &cleanup_list))
562 queue_work(netns_wq, &net_cleanup_work);
564 EXPORT_SYMBOL_GPL(__put_net);
566 struct net *get_net_ns_by_fd(int fd)
569 struct ns_common *ns;
572 file = proc_ns_fget(fd);
574 return ERR_CAST(file);
576 ns = get_proc_ns(file_inode(file));
577 if (ns->ops == &netns_operations)
578 net = get_net(container_of(ns, struct net, ns));
580 net = ERR_PTR(-EINVAL);
587 struct net *get_net_ns_by_fd(int fd)
589 return ERR_PTR(-EINVAL);
592 EXPORT_SYMBOL_GPL(get_net_ns_by_fd);
594 struct net *get_net_ns_by_pid(pid_t pid)
596 struct task_struct *tsk;
599 /* Lookup the network namespace */
600 net = ERR_PTR(-ESRCH);
602 tsk = find_task_by_vpid(pid);
604 struct nsproxy *nsproxy;
606 nsproxy = tsk->nsproxy;
608 net = get_net(nsproxy->net_ns);
614 EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
616 static __net_init int net_ns_net_init(struct net *net)
619 net->ns.ops = &netns_operations;
621 return ns_alloc_inum(&net->ns);
624 static __net_exit void net_ns_net_exit(struct net *net)
626 ns_free_inum(&net->ns);
629 static struct pernet_operations __net_initdata net_ns_ops = {
630 .init = net_ns_net_init,
631 .exit = net_ns_net_exit,
634 static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = {
635 [NETNSA_NONE] = { .type = NLA_UNSPEC },
636 [NETNSA_NSID] = { .type = NLA_S32 },
637 [NETNSA_PID] = { .type = NLA_U32 },
638 [NETNSA_FD] = { .type = NLA_U32 },
641 static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh,
642 struct netlink_ext_ack *extack)
644 struct net *net = sock_net(skb->sk);
645 struct nlattr *tb[NETNSA_MAX + 1];
650 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
651 rtnl_net_policy, extack);
654 if (!tb[NETNSA_NSID]) {
655 NL_SET_ERR_MSG(extack, "nsid is missing");
658 nsid = nla_get_s32(tb[NETNSA_NSID]);
660 if (tb[NETNSA_PID]) {
661 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
662 nla = tb[NETNSA_PID];
663 } else if (tb[NETNSA_FD]) {
664 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
667 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
671 NL_SET_BAD_ATTR(extack, nla);
672 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
673 return PTR_ERR(peer);
676 spin_lock_bh(&net->nsid_lock);
677 if (__peernet2id(net, peer) >= 0) {
678 spin_unlock_bh(&net->nsid_lock);
680 NL_SET_BAD_ATTR(extack, nla);
681 NL_SET_ERR_MSG(extack,
682 "Peer netns already has a nsid assigned");
686 err = alloc_netid(net, peer, nsid);
687 spin_unlock_bh(&net->nsid_lock);
689 rtnl_net_notifyid(net, RTM_NEWNSID, err);
691 } else if (err == -ENOSPC && nsid >= 0) {
693 NL_SET_BAD_ATTR(extack, tb[NETNSA_NSID]);
694 NL_SET_ERR_MSG(extack, "The specified nsid is already used");
701 static int rtnl_net_get_size(void)
703 return NLMSG_ALIGN(sizeof(struct rtgenmsg))
704 + nla_total_size(sizeof(s32)) /* NETNSA_NSID */
708 static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags,
709 int cmd, struct net *net, int nsid)
711 struct nlmsghdr *nlh;
712 struct rtgenmsg *rth;
714 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags);
718 rth = nlmsg_data(nlh);
719 rth->rtgen_family = AF_UNSPEC;
721 if (nla_put_s32(skb, NETNSA_NSID, nsid))
722 goto nla_put_failure;
728 nlmsg_cancel(skb, nlh);
732 static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh,
733 struct netlink_ext_ack *extack)
735 struct net *net = sock_net(skb->sk);
736 struct nlattr *tb[NETNSA_MAX + 1];
742 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
743 rtnl_net_policy, extack);
746 if (tb[NETNSA_PID]) {
747 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
748 nla = tb[NETNSA_PID];
749 } else if (tb[NETNSA_FD]) {
750 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
753 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
758 NL_SET_BAD_ATTR(extack, nla);
759 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
760 return PTR_ERR(peer);
763 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
769 id = peernet2id(net, peer);
770 err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
771 RTM_NEWNSID, net, id);
775 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid);
785 struct rtnl_net_dump_cb {
788 struct netlink_callback *cb;
793 static int rtnl_net_dumpid_one(int id, void *peer, void *data)
795 struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data;
798 if (net_cb->idx < net_cb->s_idx)
801 ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid,
802 net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI,
803 RTM_NEWNSID, net_cb->net, id);
812 static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb)
814 struct net *net = sock_net(skb->sk);
815 struct rtnl_net_dump_cb net_cb = {
820 .s_idx = cb->args[0],
823 spin_lock_bh(&net->nsid_lock);
824 idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb);
825 spin_unlock_bh(&net->nsid_lock);
827 cb->args[0] = net_cb.idx;
831 static void rtnl_net_notifyid(struct net *net, int cmd, int id)
836 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
840 err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, id);
844 rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0);
850 rtnl_set_sk_err(net, RTNLGRP_NSID, err);
853 static int __init net_ns_init(void)
855 struct net_generic *ng;
858 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net),
860 SLAB_PANIC|SLAB_ACCOUNT, NULL);
862 /* Create workqueue for cleanup */
863 netns_wq = create_singlethread_workqueue("netns");
865 panic("Could not create netns workq");
868 ng = net_alloc_generic();
870 panic("Could not allocate generic netns");
872 rcu_assign_pointer(init_net.gen, ng);
874 down_write(&pernet_ops_rwsem);
875 if (setup_net(&init_net, &init_user_ns))
876 panic("Could not setup the initial network namespace");
878 init_net_initialized = true;
879 up_write(&pernet_ops_rwsem);
881 register_pernet_subsys(&net_ns_ops);
883 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL,
884 RTNL_FLAG_DOIT_UNLOCKED);
885 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid,
886 RTNL_FLAG_DOIT_UNLOCKED);
891 pure_initcall(net_ns_init);
894 static int __register_pernet_operations(struct list_head *list,
895 struct pernet_operations *ops)
899 LIST_HEAD(net_exit_list);
901 list_add_tail(&ops->list, list);
902 if (ops->init || (ops->id && ops->size)) {
904 error = ops_init(ops, net);
907 list_add_tail(&net->exit_list, &net_exit_list);
913 /* If I have an error cleanup all namespaces I initialized */
914 list_del(&ops->list);
915 ops_exit_list(ops, &net_exit_list);
916 ops_free_list(ops, &net_exit_list);
920 static void __unregister_pernet_operations(struct pernet_operations *ops)
923 LIST_HEAD(net_exit_list);
925 list_del(&ops->list);
927 list_add_tail(&net->exit_list, &net_exit_list);
928 ops_exit_list(ops, &net_exit_list);
929 ops_free_list(ops, &net_exit_list);
934 static int __register_pernet_operations(struct list_head *list,
935 struct pernet_operations *ops)
937 if (!init_net_initialized) {
938 list_add_tail(&ops->list, list);
942 return ops_init(ops, &init_net);
945 static void __unregister_pernet_operations(struct pernet_operations *ops)
947 if (!init_net_initialized) {
948 list_del(&ops->list);
950 LIST_HEAD(net_exit_list);
951 list_add(&init_net.exit_list, &net_exit_list);
952 ops_exit_list(ops, &net_exit_list);
953 ops_free_list(ops, &net_exit_list);
957 #endif /* CONFIG_NET_NS */
959 static DEFINE_IDA(net_generic_ids);
961 static int register_pernet_operations(struct list_head *list,
962 struct pernet_operations *ops)
968 error = ida_get_new_above(&net_generic_ids, MIN_PERNET_OPS_ID, ops->id);
970 if (error == -EAGAIN) {
971 ida_pre_get(&net_generic_ids, GFP_KERNEL);
976 max_gen_ptrs = max(max_gen_ptrs, *ops->id + 1);
978 error = __register_pernet_operations(list, ops);
982 ida_remove(&net_generic_ids, *ops->id);
988 static void unregister_pernet_operations(struct pernet_operations *ops)
990 __unregister_pernet_operations(ops);
993 ida_remove(&net_generic_ids, *ops->id);
997 * register_pernet_subsys - register a network namespace subsystem
998 * @ops: pernet operations structure for the subsystem
1000 * Register a subsystem which has init and exit functions
1001 * that are called when network namespaces are created and
1002 * destroyed respectively.
1004 * When registered all network namespace init functions are
1005 * called for every existing network namespace. Allowing kernel
1006 * modules to have a race free view of the set of network namespaces.
1008 * When a new network namespace is created all of the init
1009 * methods are called in the order in which they were registered.
1011 * When a network namespace is destroyed all of the exit methods
1012 * are called in the reverse of the order with which they were
1015 int register_pernet_subsys(struct pernet_operations *ops)
1018 down_write(&pernet_ops_rwsem);
1019 error = register_pernet_operations(first_device, ops);
1020 up_write(&pernet_ops_rwsem);
1023 EXPORT_SYMBOL_GPL(register_pernet_subsys);
1026 * unregister_pernet_subsys - unregister a network namespace subsystem
1027 * @ops: pernet operations structure to manipulate
1029 * Remove the pernet operations structure from the list to be
1030 * used when network namespaces are created or destroyed. In
1031 * addition run the exit method for all existing network
1034 void unregister_pernet_subsys(struct pernet_operations *ops)
1036 down_write(&pernet_ops_rwsem);
1037 unregister_pernet_operations(ops);
1038 up_write(&pernet_ops_rwsem);
1040 EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
1043 * register_pernet_device - register a network namespace device
1044 * @ops: pernet operations structure for the subsystem
1046 * Register a device which has init and exit functions
1047 * that are called when network namespaces are created and
1048 * destroyed respectively.
1050 * When registered all network namespace init functions are
1051 * called for every existing network namespace. Allowing kernel
1052 * modules to have a race free view of the set of network namespaces.
1054 * When a new network namespace is created all of the init
1055 * methods are called in the order in which they were registered.
1057 * When a network namespace is destroyed all of the exit methods
1058 * are called in the reverse of the order with which they were
1061 int register_pernet_device(struct pernet_operations *ops)
1064 down_write(&pernet_ops_rwsem);
1065 error = register_pernet_operations(&pernet_list, ops);
1066 if (!error && (first_device == &pernet_list))
1067 first_device = &ops->list;
1068 up_write(&pernet_ops_rwsem);
1071 EXPORT_SYMBOL_GPL(register_pernet_device);
1074 * unregister_pernet_device - unregister a network namespace netdevice
1075 * @ops: pernet operations structure to manipulate
1077 * Remove the pernet operations structure from the list to be
1078 * used when network namespaces are created or destroyed. In
1079 * addition run the exit method for all existing network
1082 void unregister_pernet_device(struct pernet_operations *ops)
1084 down_write(&pernet_ops_rwsem);
1085 if (&ops->list == first_device)
1086 first_device = first_device->next;
1087 unregister_pernet_operations(ops);
1088 up_write(&pernet_ops_rwsem);
1090 EXPORT_SYMBOL_GPL(unregister_pernet_device);
1092 #ifdef CONFIG_NET_NS
1093 static struct ns_common *netns_get(struct task_struct *task)
1095 struct net *net = NULL;
1096 struct nsproxy *nsproxy;
1099 nsproxy = task->nsproxy;
1101 net = get_net(nsproxy->net_ns);
1104 return net ? &net->ns : NULL;
1107 static inline struct net *to_net_ns(struct ns_common *ns)
1109 return container_of(ns, struct net, ns);
1112 static void netns_put(struct ns_common *ns)
1114 put_net(to_net_ns(ns));
1117 static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns)
1119 struct net *net = to_net_ns(ns);
1121 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
1122 !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
1125 put_net(nsproxy->net_ns);
1126 nsproxy->net_ns = get_net(net);
1130 static struct user_namespace *netns_owner(struct ns_common *ns)
1132 return to_net_ns(ns)->user_ns;
1135 const struct proc_ns_operations netns_operations = {
1137 .type = CLONE_NEWNET,
1140 .install = netns_install,
1141 .owner = netns_owner,