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>
20 #include <linux/uidgid.h>
23 #include <net/netlink.h>
24 #include <net/net_namespace.h>
25 #include <net/netns/generic.h>
28 * Our network namespace constructor/destructor lists
31 static LIST_HEAD(pernet_list);
32 static struct list_head *first_device = &pernet_list;
34 LIST_HEAD(net_namespace_list);
35 EXPORT_SYMBOL_GPL(net_namespace_list);
37 /* Protects net_namespace_list. Nests iside rtnl_lock() */
38 DECLARE_RWSEM(net_rwsem);
39 EXPORT_SYMBOL_GPL(net_rwsem);
41 struct net init_net = {
42 .count = REFCOUNT_INIT(1),
43 .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head),
45 EXPORT_SYMBOL(init_net);
47 static bool init_net_initialized;
49 * pernet_ops_rwsem: protects: pernet_list, net_generic_ids,
50 * init_net_initialized and first_device pointer.
51 * This is internal net namespace object. Please, don't use it
54 DECLARE_RWSEM(pernet_ops_rwsem);
55 EXPORT_SYMBOL_GPL(pernet_ops_rwsem);
57 #define MIN_PERNET_OPS_ID \
58 ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *))
60 #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
62 static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS;
64 static struct net_generic *net_alloc_generic(void)
66 struct net_generic *ng;
67 unsigned int generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]);
69 ng = kzalloc(generic_size, GFP_KERNEL);
71 ng->s.len = max_gen_ptrs;
76 static int net_assign_generic(struct net *net, unsigned int id, void *data)
78 struct net_generic *ng, *old_ng;
80 BUG_ON(id < MIN_PERNET_OPS_ID);
82 old_ng = rcu_dereference_protected(net->gen,
83 lockdep_is_held(&pernet_ops_rwsem));
84 if (old_ng->s.len > id) {
85 old_ng->ptr[id] = data;
89 ng = net_alloc_generic();
94 * Some synchronisation notes:
96 * The net_generic explores the net->gen array inside rcu
97 * read section. Besides once set the net->gen->ptr[x]
98 * pointer never changes (see rules in netns/generic.h).
100 * That said, we simply duplicate this array and schedule
101 * the old copy for kfree after a grace period.
104 memcpy(&ng->ptr[MIN_PERNET_OPS_ID], &old_ng->ptr[MIN_PERNET_OPS_ID],
105 (old_ng->s.len - MIN_PERNET_OPS_ID) * sizeof(void *));
108 rcu_assign_pointer(net->gen, ng);
109 kfree_rcu(old_ng, s.rcu);
113 static int ops_init(const struct pernet_operations *ops, struct net *net)
118 if (ops->id && ops->size) {
119 data = kzalloc(ops->size, GFP_KERNEL);
123 err = net_assign_generic(net, *ops->id, data);
129 err = ops->init(net);
140 static void ops_free(const struct pernet_operations *ops, struct net *net)
142 if (ops->id && ops->size) {
143 kfree(net_generic(net, *ops->id));
147 static void ops_exit_list(const struct pernet_operations *ops,
148 struct list_head *net_exit_list)
152 list_for_each_entry(net, net_exit_list, exit_list) {
158 ops->exit_batch(net_exit_list);
161 static void ops_free_list(const struct pernet_operations *ops,
162 struct list_head *net_exit_list)
165 if (ops->size && ops->id) {
166 list_for_each_entry(net, net_exit_list, exit_list)
171 /* should be called with nsid_lock held */
172 static int alloc_netid(struct net *net, struct net *peer, int reqid)
174 int min = 0, max = 0;
181 return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC);
184 /* This function is used by idr_for_each(). If net is equal to peer, the
185 * function returns the id so that idr_for_each() stops. Because we cannot
186 * returns the id 0 (idr_for_each() will not stop), we return the magic value
187 * NET_ID_ZERO (-1) for it.
189 #define NET_ID_ZERO -1
190 static int net_eq_idr(int id, void *net, void *peer)
192 if (net_eq(net, peer))
193 return id ? : NET_ID_ZERO;
197 /* Must be called from RCU-critical section or with nsid_lock held. If
198 * a new id is assigned, the bool alloc is set to true, thus the
199 * caller knows that the new id must be notified via rtnl.
201 static int __peernet2id_alloc(struct net *net, struct net *peer, bool *alloc)
203 int id = idr_for_each(&net->netns_ids, net_eq_idr, peer);
204 bool alloc_it = *alloc;
208 /* Magic value for id 0. */
209 if (id == NET_ID_ZERO)
215 id = alloc_netid(net, peer, -1);
217 return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED;
220 return NETNSA_NSID_NOT_ASSIGNED;
223 /* Must be called from RCU-critical section or with nsid_lock held */
224 static int __peernet2id(struct net *net, struct net *peer)
228 return __peernet2id_alloc(net, peer, &no);
231 static void rtnl_net_notifyid(struct net *net, int cmd, int id, gfp_t gfp);
232 /* This function returns the id of a peer netns. If no id is assigned, one will
233 * be allocated and returned.
235 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp)
237 bool alloc = false, alive = false;
240 if (refcount_read(&net->count) == 0)
241 return NETNSA_NSID_NOT_ASSIGNED;
242 spin_lock_bh(&net->nsid_lock);
244 * When peer is obtained from RCU lists, we may race with
245 * its cleanup. Check whether it's alive, and this guarantees
246 * we never hash a peer back to net->netns_ids, after it has
247 * just been idr_remove()'d from there in cleanup_net().
249 if (maybe_get_net(peer))
250 alive = alloc = true;
251 id = __peernet2id_alloc(net, peer, &alloc);
252 spin_unlock_bh(&net->nsid_lock);
253 if (alloc && id >= 0)
254 rtnl_net_notifyid(net, RTM_NEWNSID, id, gfp);
259 EXPORT_SYMBOL_GPL(peernet2id_alloc);
261 /* This function returns, if assigned, the id of a peer netns. */
262 int peernet2id(struct net *net, struct net *peer)
267 id = __peernet2id(net, peer);
272 EXPORT_SYMBOL(peernet2id);
274 /* This function returns true is the peer netns has an id assigned into the
277 bool peernet_has_id(struct net *net, struct net *peer)
279 return peernet2id(net, peer) >= 0;
282 struct net *get_net_ns_by_id(struct net *net, int id)
290 peer = idr_find(&net->netns_ids, id);
292 peer = maybe_get_net(peer);
299 * setup_net runs the initializers for the network namespace object.
301 static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
303 /* Must be called with pernet_ops_rwsem held */
304 const struct pernet_operations *ops, *saved_ops;
306 LIST_HEAD(net_exit_list);
308 refcount_set(&net->count, 1);
309 refcount_set(&net->passive, 1);
310 get_random_bytes(&net->hash_mix, sizeof(u32));
311 net->dev_base_seq = 1;
312 net->user_ns = user_ns;
313 idr_init(&net->netns_ids);
314 spin_lock_init(&net->nsid_lock);
315 mutex_init(&net->ipv4.ra_mutex);
317 list_for_each_entry(ops, &pernet_list, list) {
318 error = ops_init(ops, net);
322 down_write(&net_rwsem);
323 list_add_tail_rcu(&net->list, &net_namespace_list);
324 up_write(&net_rwsem);
329 /* Walk through the list backwards calling the exit functions
330 * for the pernet modules whose init functions did not fail.
332 list_add(&net->exit_list, &net_exit_list);
334 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
335 ops_exit_list(ops, &net_exit_list);
338 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
339 ops_free_list(ops, &net_exit_list);
345 static int __net_init net_defaults_init_net(struct net *net)
347 net->core.sysctl_somaxconn = SOMAXCONN;
351 static struct pernet_operations net_defaults_ops = {
352 .init = net_defaults_init_net,
355 static __init int net_defaults_init(void)
357 if (register_pernet_subsys(&net_defaults_ops))
358 panic("Cannot initialize net default settings");
363 core_initcall(net_defaults_init);
366 static struct ucounts *inc_net_namespaces(struct user_namespace *ns)
368 return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES);
371 static void dec_net_namespaces(struct ucounts *ucounts)
373 dec_ucount(ucounts, UCOUNT_NET_NAMESPACES);
376 static struct kmem_cache *net_cachep __ro_after_init;
377 static struct workqueue_struct *netns_wq;
379 static struct net *net_alloc(void)
381 struct net *net = NULL;
382 struct net_generic *ng;
384 ng = net_alloc_generic();
388 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL);
392 rcu_assign_pointer(net->gen, ng);
401 static void net_free(struct net *net)
403 kfree(rcu_access_pointer(net->gen));
404 kmem_cache_free(net_cachep, net);
407 void net_drop_ns(void *p)
410 if (ns && refcount_dec_and_test(&ns->passive))
414 struct net *copy_net_ns(unsigned long flags,
415 struct user_namespace *user_ns, struct net *old_net)
417 struct ucounts *ucounts;
421 if (!(flags & CLONE_NEWNET))
422 return get_net(old_net);
424 ucounts = inc_net_namespaces(user_ns);
426 return ERR_PTR(-ENOSPC);
433 refcount_set(&net->passive, 1);
434 net->ucounts = ucounts;
435 get_user_ns(user_ns);
437 rv = down_read_killable(&pernet_ops_rwsem);
441 rv = setup_net(net, user_ns);
443 up_read(&pernet_ops_rwsem);
447 put_user_ns(user_ns);
450 dec_net_namespaces(ucounts);
457 * net_ns_get_ownership - get sysfs ownership data for @net
458 * @net: network namespace in question (can be NULL)
459 * @uid: kernel user ID for sysfs objects
460 * @gid: kernel group ID for sysfs objects
462 * Returns the uid/gid pair of root in the user namespace associated with the
463 * given network namespace.
465 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid)
468 kuid_t ns_root_uid = make_kuid(net->user_ns, 0);
469 kgid_t ns_root_gid = make_kgid(net->user_ns, 0);
471 if (uid_valid(ns_root_uid))
474 if (gid_valid(ns_root_gid))
477 *uid = GLOBAL_ROOT_UID;
478 *gid = GLOBAL_ROOT_GID;
481 EXPORT_SYMBOL_GPL(net_ns_get_ownership);
483 static void unhash_nsid(struct net *net, struct net *last)
486 /* This function is only called from cleanup_net() work,
487 * and this work is the only process, that may delete
488 * a net from net_namespace_list. So, when the below
489 * is executing, the list may only grow. Thus, we do not
490 * use for_each_net_rcu() or net_rwsem.
495 spin_lock_bh(&tmp->nsid_lock);
496 id = __peernet2id(tmp, net);
498 idr_remove(&tmp->netns_ids, id);
499 spin_unlock_bh(&tmp->nsid_lock);
501 rtnl_net_notifyid(tmp, RTM_DELNSID, id,
506 spin_lock_bh(&net->nsid_lock);
507 idr_destroy(&net->netns_ids);
508 spin_unlock_bh(&net->nsid_lock);
511 static LLIST_HEAD(cleanup_list);
513 static void cleanup_net(struct work_struct *work)
515 const struct pernet_operations *ops;
516 struct net *net, *tmp, *last;
517 struct llist_node *net_kill_list;
518 LIST_HEAD(net_exit_list);
520 /* Atomically snapshot the list of namespaces to cleanup */
521 net_kill_list = llist_del_all(&cleanup_list);
523 down_read(&pernet_ops_rwsem);
525 /* Don't let anyone else find us. */
526 down_write(&net_rwsem);
527 llist_for_each_entry(net, net_kill_list, cleanup_list)
528 list_del_rcu(&net->list);
529 /* Cache last net. After we unlock rtnl, no one new net
530 * added to net_namespace_list can assign nsid pointer
531 * to a net from net_kill_list (see peernet2id_alloc()).
532 * So, we skip them in unhash_nsid().
534 * Note, that unhash_nsid() does not delete nsid links
535 * between net_kill_list's nets, as they've already
536 * deleted from net_namespace_list. But, this would be
537 * useless anyway, as netns_ids are destroyed there.
539 last = list_last_entry(&net_namespace_list, struct net, list);
540 up_write(&net_rwsem);
542 llist_for_each_entry(net, net_kill_list, cleanup_list) {
543 unhash_nsid(net, last);
544 list_add_tail(&net->exit_list, &net_exit_list);
548 * Another CPU might be rcu-iterating the list, wait for it.
549 * This needs to be before calling the exit() notifiers, so
550 * the rcu_barrier() below isn't sufficient alone.
554 /* Run all of the network namespace exit methods */
555 list_for_each_entry_reverse(ops, &pernet_list, list)
556 ops_exit_list(ops, &net_exit_list);
558 /* Free the net generic variables */
559 list_for_each_entry_reverse(ops, &pernet_list, list)
560 ops_free_list(ops, &net_exit_list);
562 up_read(&pernet_ops_rwsem);
564 /* Ensure there are no outstanding rcu callbacks using this
569 /* Finally it is safe to free my network namespace structure */
570 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
571 list_del_init(&net->exit_list);
572 dec_net_namespaces(net->ucounts);
573 put_user_ns(net->user_ns);
579 * net_ns_barrier - wait until concurrent net_cleanup_work is done
581 * cleanup_net runs from work queue and will first remove namespaces
582 * from the global list, then run net exit functions.
584 * Call this in module exit path to make sure that all netns
585 * ->exit ops have been invoked before the function is removed.
587 void net_ns_barrier(void)
589 down_write(&pernet_ops_rwsem);
590 up_write(&pernet_ops_rwsem);
592 EXPORT_SYMBOL(net_ns_barrier);
594 static DECLARE_WORK(net_cleanup_work, cleanup_net);
596 void __put_net(struct net *net)
598 /* Cleanup the network namespace in process context */
599 if (llist_add(&net->cleanup_list, &cleanup_list))
600 queue_work(netns_wq, &net_cleanup_work);
602 EXPORT_SYMBOL_GPL(__put_net);
605 * get_net_ns - increment the refcount of the network namespace
606 * @ns: common namespace (net)
608 * Returns the net's common namespace.
610 struct ns_common *get_net_ns(struct ns_common *ns)
612 return &get_net(container_of(ns, struct net, ns))->ns;
614 EXPORT_SYMBOL_GPL(get_net_ns);
616 struct net *get_net_ns_by_fd(int fd)
619 struct ns_common *ns;
622 file = proc_ns_fget(fd);
624 return ERR_CAST(file);
626 ns = get_proc_ns(file_inode(file));
627 if (ns->ops == &netns_operations)
628 net = get_net(container_of(ns, struct net, ns));
630 net = ERR_PTR(-EINVAL);
637 struct net *get_net_ns_by_fd(int fd)
639 return ERR_PTR(-EINVAL);
642 EXPORT_SYMBOL_GPL(get_net_ns_by_fd);
644 struct net *get_net_ns_by_pid(pid_t pid)
646 struct task_struct *tsk;
649 /* Lookup the network namespace */
650 net = ERR_PTR(-ESRCH);
652 tsk = find_task_by_vpid(pid);
654 struct nsproxy *nsproxy;
656 nsproxy = tsk->nsproxy;
658 net = get_net(nsproxy->net_ns);
664 EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
666 static __net_init int net_ns_net_init(struct net *net)
669 net->ns.ops = &netns_operations;
671 return ns_alloc_inum(&net->ns);
674 static __net_exit void net_ns_net_exit(struct net *net)
676 ns_free_inum(&net->ns);
679 static struct pernet_operations __net_initdata net_ns_ops = {
680 .init = net_ns_net_init,
681 .exit = net_ns_net_exit,
684 static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = {
685 [NETNSA_NONE] = { .type = NLA_UNSPEC },
686 [NETNSA_NSID] = { .type = NLA_S32 },
687 [NETNSA_PID] = { .type = NLA_U32 },
688 [NETNSA_FD] = { .type = NLA_U32 },
691 static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh,
692 struct netlink_ext_ack *extack)
694 struct net *net = sock_net(skb->sk);
695 struct nlattr *tb[NETNSA_MAX + 1];
700 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
701 rtnl_net_policy, extack);
704 if (!tb[NETNSA_NSID]) {
705 NL_SET_ERR_MSG(extack, "nsid is missing");
708 nsid = nla_get_s32(tb[NETNSA_NSID]);
710 if (tb[NETNSA_PID]) {
711 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
712 nla = tb[NETNSA_PID];
713 } else if (tb[NETNSA_FD]) {
714 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
717 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
721 NL_SET_BAD_ATTR(extack, nla);
722 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
723 return PTR_ERR(peer);
726 spin_lock_bh(&net->nsid_lock);
727 if (__peernet2id(net, peer) >= 0) {
728 spin_unlock_bh(&net->nsid_lock);
730 NL_SET_BAD_ATTR(extack, nla);
731 NL_SET_ERR_MSG(extack,
732 "Peer netns already has a nsid assigned");
736 err = alloc_netid(net, peer, nsid);
737 spin_unlock_bh(&net->nsid_lock);
739 rtnl_net_notifyid(net, RTM_NEWNSID, err, GFP_KERNEL);
741 } else if (err == -ENOSPC && nsid >= 0) {
743 NL_SET_BAD_ATTR(extack, tb[NETNSA_NSID]);
744 NL_SET_ERR_MSG(extack, "The specified nsid is already used");
751 static int rtnl_net_get_size(void)
753 return NLMSG_ALIGN(sizeof(struct rtgenmsg))
754 + nla_total_size(sizeof(s32)) /* NETNSA_NSID */
758 static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags,
759 int cmd, struct net *net, int nsid)
761 struct nlmsghdr *nlh;
762 struct rtgenmsg *rth;
764 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags);
768 rth = nlmsg_data(nlh);
769 rth->rtgen_family = AF_UNSPEC;
771 if (nla_put_s32(skb, NETNSA_NSID, nsid))
772 goto nla_put_failure;
778 nlmsg_cancel(skb, nlh);
782 static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh,
783 struct netlink_ext_ack *extack)
785 struct net *net = sock_net(skb->sk);
786 struct nlattr *tb[NETNSA_MAX + 1];
792 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
793 rtnl_net_policy, extack);
796 if (tb[NETNSA_PID]) {
797 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
798 nla = tb[NETNSA_PID];
799 } else if (tb[NETNSA_FD]) {
800 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
803 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
808 NL_SET_BAD_ATTR(extack, nla);
809 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
810 return PTR_ERR(peer);
813 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
819 id = peernet2id(net, peer);
820 err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
821 RTM_NEWNSID, net, id);
825 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid);
835 struct rtnl_net_dump_cb {
838 struct netlink_callback *cb;
843 /* Runs in RCU-critical section. */
844 static int rtnl_net_dumpid_one(int id, void *peer, void *data)
846 struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data;
849 if (net_cb->idx < net_cb->s_idx)
852 ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid,
853 net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI,
854 RTM_NEWNSID, net_cb->net, id);
863 static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb)
865 struct net *net = sock_net(skb->sk);
866 struct rtnl_net_dump_cb net_cb = {
871 .s_idx = cb->args[0],
875 idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb);
878 cb->args[0] = net_cb.idx;
882 static void rtnl_net_notifyid(struct net *net, int cmd, int id, gfp_t gfp)
887 msg = nlmsg_new(rtnl_net_get_size(), gfp);
891 err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, id);
895 rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, gfp);
901 rtnl_set_sk_err(net, RTNLGRP_NSID, err);
904 static int __init net_ns_init(void)
906 struct net_generic *ng;
909 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net),
911 SLAB_PANIC|SLAB_ACCOUNT, NULL);
913 /* Create workqueue for cleanup */
914 netns_wq = create_singlethread_workqueue("netns");
916 panic("Could not create netns workq");
919 ng = net_alloc_generic();
921 panic("Could not allocate generic netns");
923 rcu_assign_pointer(init_net.gen, ng);
925 down_write(&pernet_ops_rwsem);
926 if (setup_net(&init_net, &init_user_ns))
927 panic("Could not setup the initial network namespace");
929 init_net_initialized = true;
930 up_write(&pernet_ops_rwsem);
932 if (register_pernet_subsys(&net_ns_ops))
933 panic("Could not register network namespace subsystems");
935 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL,
936 RTNL_FLAG_DOIT_UNLOCKED);
937 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid,
938 RTNL_FLAG_DOIT_UNLOCKED);
943 pure_initcall(net_ns_init);
946 static int __register_pernet_operations(struct list_head *list,
947 struct pernet_operations *ops)
951 LIST_HEAD(net_exit_list);
953 list_add_tail(&ops->list, list);
954 if (ops->init || (ops->id && ops->size)) {
955 /* We held write locked pernet_ops_rwsem, and parallel
956 * setup_net() and cleanup_net() are not possible.
959 error = ops_init(ops, net);
962 list_add_tail(&net->exit_list, &net_exit_list);
968 /* If I have an error cleanup all namespaces I initialized */
969 list_del(&ops->list);
970 ops_exit_list(ops, &net_exit_list);
971 ops_free_list(ops, &net_exit_list);
975 static void __unregister_pernet_operations(struct pernet_operations *ops)
978 LIST_HEAD(net_exit_list);
980 list_del(&ops->list);
981 /* See comment in __register_pernet_operations() */
983 list_add_tail(&net->exit_list, &net_exit_list);
984 ops_exit_list(ops, &net_exit_list);
985 ops_free_list(ops, &net_exit_list);
990 static int __register_pernet_operations(struct list_head *list,
991 struct pernet_operations *ops)
993 if (!init_net_initialized) {
994 list_add_tail(&ops->list, list);
998 return ops_init(ops, &init_net);
1001 static void __unregister_pernet_operations(struct pernet_operations *ops)
1003 if (!init_net_initialized) {
1004 list_del(&ops->list);
1006 LIST_HEAD(net_exit_list);
1007 list_add(&init_net.exit_list, &net_exit_list);
1008 ops_exit_list(ops, &net_exit_list);
1009 ops_free_list(ops, &net_exit_list);
1013 #endif /* CONFIG_NET_NS */
1015 static DEFINE_IDA(net_generic_ids);
1017 static int register_pernet_operations(struct list_head *list,
1018 struct pernet_operations *ops)
1023 error = ida_alloc_min(&net_generic_ids, MIN_PERNET_OPS_ID,
1028 max_gen_ptrs = max(max_gen_ptrs, *ops->id + 1);
1030 error = __register_pernet_operations(list, ops);
1034 ida_free(&net_generic_ids, *ops->id);
1040 static void unregister_pernet_operations(struct pernet_operations *ops)
1042 __unregister_pernet_operations(ops);
1045 ida_free(&net_generic_ids, *ops->id);
1049 * register_pernet_subsys - register a network namespace subsystem
1050 * @ops: pernet operations structure for the subsystem
1052 * Register a subsystem which has init and exit functions
1053 * that are called when network namespaces are created and
1054 * destroyed respectively.
1056 * When registered all network namespace init functions are
1057 * called for every existing network namespace. Allowing kernel
1058 * modules to have a race free view of the set of network namespaces.
1060 * When a new network namespace is created all of the init
1061 * methods are called in the order in which they were registered.
1063 * When a network namespace is destroyed all of the exit methods
1064 * are called in the reverse of the order with which they were
1067 int register_pernet_subsys(struct pernet_operations *ops)
1070 down_write(&pernet_ops_rwsem);
1071 error = register_pernet_operations(first_device, ops);
1072 up_write(&pernet_ops_rwsem);
1075 EXPORT_SYMBOL_GPL(register_pernet_subsys);
1078 * unregister_pernet_subsys - unregister a network namespace subsystem
1079 * @ops: pernet operations structure to manipulate
1081 * Remove the pernet operations structure from the list to be
1082 * used when network namespaces are created or destroyed. In
1083 * addition run the exit method for all existing network
1086 void unregister_pernet_subsys(struct pernet_operations *ops)
1088 down_write(&pernet_ops_rwsem);
1089 unregister_pernet_operations(ops);
1090 up_write(&pernet_ops_rwsem);
1092 EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
1095 * register_pernet_device - register a network namespace device
1096 * @ops: pernet operations structure for the subsystem
1098 * Register a device which has init and exit functions
1099 * that are called when network namespaces are created and
1100 * destroyed respectively.
1102 * When registered all network namespace init functions are
1103 * called for every existing network namespace. Allowing kernel
1104 * modules to have a race free view of the set of network namespaces.
1106 * When a new network namespace is created all of the init
1107 * methods are called in the order in which they were registered.
1109 * When a network namespace is destroyed all of the exit methods
1110 * are called in the reverse of the order with which they were
1113 int register_pernet_device(struct pernet_operations *ops)
1116 down_write(&pernet_ops_rwsem);
1117 error = register_pernet_operations(&pernet_list, ops);
1118 if (!error && (first_device == &pernet_list))
1119 first_device = &ops->list;
1120 up_write(&pernet_ops_rwsem);
1123 EXPORT_SYMBOL_GPL(register_pernet_device);
1126 * unregister_pernet_device - unregister a network namespace netdevice
1127 * @ops: pernet operations structure to manipulate
1129 * Remove the pernet operations structure from the list to be
1130 * used when network namespaces are created or destroyed. In
1131 * addition run the exit method for all existing network
1134 void unregister_pernet_device(struct pernet_operations *ops)
1136 down_write(&pernet_ops_rwsem);
1137 if (&ops->list == first_device)
1138 first_device = first_device->next;
1139 unregister_pernet_operations(ops);
1140 up_write(&pernet_ops_rwsem);
1142 EXPORT_SYMBOL_GPL(unregister_pernet_device);
1144 #ifdef CONFIG_NET_NS
1145 static struct ns_common *netns_get(struct task_struct *task)
1147 struct net *net = NULL;
1148 struct nsproxy *nsproxy;
1151 nsproxy = task->nsproxy;
1153 net = get_net(nsproxy->net_ns);
1156 return net ? &net->ns : NULL;
1159 static inline struct net *to_net_ns(struct ns_common *ns)
1161 return container_of(ns, struct net, ns);
1164 static void netns_put(struct ns_common *ns)
1166 put_net(to_net_ns(ns));
1169 static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns)
1171 struct net *net = to_net_ns(ns);
1173 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
1174 !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
1177 put_net(nsproxy->net_ns);
1178 nsproxy->net_ns = get_net(net);
1182 static struct user_namespace *netns_owner(struct ns_common *ns)
1184 return to_net_ns(ns)->user_ns;
1187 const struct proc_ns_operations netns_operations = {
1189 .type = CLONE_NEWNET,
1192 .install = netns_install,
1193 .owner = netns_owner,